...

' '
'±7
■

. tP
i.Vj ■' ■•/'■■ .? ■

.: K\
R; I
hx/.'dr.
\y a ¥ £
Wk

I P )f
rlrdo-^L
\

AQUILA
I
1A

1
Dudhoth
 L ia 14

MICHAEL DUMMETT

FREGE
Philosophy ofMathematics

(if/
cey
B'b'ioten;
\\
= F&c. d: Leii.^ C
Z- \ I
^<>\e Fiiosoftf/ <: i

DUCKWORTH
r

Second impression 1995

First published in 1991 by
Gerald Duckworth & Co. Ltd.
The Old Piano Factory
48 Hoxton Square, London N1 6PB

© 1991 by Michael Dummett

All rights reserved. No part of this publication

may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic,
mechanical, photocopying, recording or otherwise, Contents
without the prior permission of the publisher.

A catalogue record for this book is

available from the British Library.
Preface vii
ISBN 0 7156 0886 X (cased)
ISBN 0 7156 2660 4 (paper) 1
1. The Significance of Grundlagen
Printed in the United States of America 2. The Introduction to Grundlagen 10
3. Analyticity 23
4. The Value of Analytic Propositions 36
5. Frege and Dedekind 47
6. Numerical Equations and Arithmetical Laws 55
7. What is Number? 72
8. Units and Concepts 82
9. Two Strategies of Analysis 99
10. Frege’s Strategy' 111
11. Some Principles of Frege’s Strategy' 125
12. Frege and Husserl 141
13. Frege’s Definition of Cardinal Numbers 155
14. The Status of the Definition 167
15. Did Frege Refute Reductionism? 180
16. The Context Principle 200
17. The Context Principle in Grundgesetze 209
18. Abstract Objects 223
19. Part III of Grundgesetze 241
20. The Critique of Formalism 252
21. The Critique of Cantor 263
22. Frege’s Theory' of Real Numbers 277
23. Assessment 292
24. The Problem of Mathematical Objects 307
Phototypeset by Intype, London
Bibliographical Note 322
Index of Frege’s Writings 323
General Index 327
for Tessa, Nathan and Nicola Preface

A book of this title was advertised as forthcoming in Duckworth’s catalogue

for 1973, the year in which my Frege: Philosophy of Language was published. I
therefore feel some need to explain why it is coming out only now to all who
have been asking me, over the years, when it was going to appear. It was not
in fact until 1973 that I started to write, as a separate book, this sequel to the
earlier one. For the new book, I formed the plan of setting out systematically
the problems of the philosophy of mathematics, and considering in order
Frege’s responses to them, to the extent that he said anything relevant: the
architecture of the book was to be that of the subject, not of Frege’s writings,
that is to say of the subject as I saw it, not as Frege saw it.
I completed about two-thirds of the book in 1973. Though I was fortunate
to hold, for a few years, a Senior Research Fellowship at All Souls’ College,
other writing commitments, including the preparation, with much help from
Mark Helme and Charles Donahue, of the second edition of Frege: Philosophy
of Language and the composition of the introduction to it, which turned into
The Interpretation of Frege’s Philosophy, prevented me from attending to the
book, which remained untouched on my shelves, until 1982. In that year I was
awarded an Alexander von Humboldt-Stiftung prize for study in Germany,
and spent four months at the University of Munster in Westphalia, taking with
me the typescript of Frege: Philosophy of Mathematics. There are two well-
known reactions to reading what one has written long ago: to think, ‘How
brilliant I was then: I could never do that now’; and to wonder how one could
have written such poor stuff. Mine was the second, and I started to rewrite
the whole book, still on the same plan, from Chapter 3 onwards. To my
disappointment, I did not finish. During four long vacations, from 1983 to
1986, I gave my main attention to trying to finish the book. Each time, it was
difficult to recall just what my previous intentions had been, and each time I
failed. In 1985, I decided to extract all the material on Frege’s theory’ of real
numbers and publish it as a separate monograph, including both philosophical
and mathematical material, the latter including the solution, due to Dr Peter
vii
 viii Frege: Philosophy of Mathematics Preface &
Neumann, of the independence problem that troubled Frege;1 I then worked For this purpose, the people so competing with one another should not be
simultaneously on the monograph and on the main book. But even with this encouraged to believe in the good of anything but themselves as individuals;
excision, the latter grew beyond all reasonable size; and still I did not finish if they were to believe in society as a whole, they might form ideas about
either. protecting the weak or unfortunate that w ould clog the efficiency of the system.
During 1988-9, I enjoyed, for the first time in my life, a whole sabbatical A glance at the universities as they used to be revealed a social sector not
year. I was lucky enough to spend from September to June at the Center for functioning in this manner; it therefore obviously could not be functioning
Advanced Study in the Behavioral Sciences at Stanford. I went hoping to efficiently, or justifying the money spent on it, and hence must be transformed
complete two long unaccomplished tasks, one of them the Frege book, for in accordance with the model decreed by ideology.
which I took with me for an enormous pile of typescript and collection of The plan of the ideologues is to increase academic productivity' by creating
discs. It was a toss-up which of the two tasks I should start on first; but I conditions of intense competition. Those who compose what is known, in
happened to select the William James lectures on The Logical Basis ofMetaphy­ today’s unlovely jargon, as academic and academic-related staff are now to be
sics. I succeeded in sending off a completed text of that book to Harvard lured by the hope of gaining, and goaded by the shame of missing, extra
University Press just before I left in June. I had also to revise a much shorter payments and newly invented titular status. Their output is monitored by the
book, I Tarocchi Sicilian, in accordance with the suggestions of my then use of performance indicators, measuring the number of words published per
collaborator and now sorely missed friend, the late Marcello Cimino.2 The year. Wittgenstein, who died in 1951 having published only one short article
result of all this was that I did not devote one minute of my time at Stanford after the Tractatus of 1922, would plainly not have survived such a system.
\ to Frege: Philosophy ofMathematics, and crossed the Atlantic again with all my Those most savagely affected by the new regime are, as always, the ones on
! typescripts and discs unused. the bottom rung of the ladder: the graduate students working for their doctor­
For those who think in terms of completion rates, mine is disgraceful. ates. The degree of Ph.D. (in Oxford, D.Phil.) fitted rather awkwardly into
‘Completion rates’ - the very phrase is like a bell. British universities are in the system of doctorates as it had evolved in Britain out of the mediaeval one,
the course of being transformed by ideologues who misunderstand everything and was originally instituted here to satisfy’ the needs of foreign students, for
about academic work. The transformation is of course merely part of a trans­ whom it was a necessary professional qualification. Only in recent years has
formation of society as a whole. The official stance of the ideologues is that it become an indispensable minimum qualification for British academic posts
they do not believe that there is any such thing as society; in point of fact, in arts subjects: candidates for them stand little chance if they cannot also
however, they do not believe in anything else. They are concerned, for example, show, at the start of their careers, an impressive list of publications. Relentless
with the performance of ‘the economy’: not with whether individual people pressure is applied to students and their universities by the Government and
are prospering, but with the economy as a distinguishable system on its own. its agencies - the research councils and the British Academy - to force them
The successful performance of the economy will grossly enrich some, and to complete their doctoral theses within three years of graduating; but it is
deprive others of all hope or comfort: but the aim, if one is not to take a hardly needed. Nervously conscious from the start that they must jostle one
cynical view of it, cannot be either to reward those who scramble to the top another for the diminished number of posts, they are anxious to jump the first
of the economic mountain or to punish those who are cast on to the scrapheap hurdle of the Ph.D. degree as quickly as possible, and then rush to submit
at its foot, but simply to ensure efficient functioning of the economy as such. their unrevised theses for publishers to turn into books.
The vision which the ideologues have of the successful functioning of the The universities have no option but to co-operate in organising the squalid
economy or of any other social mechanism is that it works well only if operated scramble that graduate study has become, in introducing the new ‘incentives’
by human beings engaged in ruthlessly biting and clawing their way to the top, for their professors and lecturers and in supplying the data for the evaluation
where they will be able to obtain a disproportionate share of limited rewards. process. The question is to what extent they will absorb the values of their
overlords and jettison those they used to have. Once more, it is the graduate
Published in S.A. Adeleke, M.A.E. Dummett and Peter M. Neumann, ‘On a Question of
Fre 5*13 321°Ut Groups’, Bulletin of the London Mathematical Society, vol. 19, 1987, students who are the most at risk, for they are in effect being taught that the
rat-race operates as ferociously in the academic as in the commercial world,
,2 1’0wever> also had t0 devote much time to extensively revising La Storia dei Tarocchi and that what matters is not the quality' of what you write but the speed at
which had been commissioned by Bibliopolis of Naples in 1982, and had become seriously out
°f date m the mtenm. I was compelled to give priority to this task, because the director of the which you write it and get it into print. It is obviously as objectionable in a
Pobhshmg house. Signor Francesco del Franco, had promised to bring the book out bv Christmas capitalist as in a communist country that politicians should decide how the
1989, and wanted the revisions urgently. It has not vet appeared, but I still have hopes universities are to be run; but it is catastrophic when those politicians display
X Frege: Philosophy ofMathematics Preface xi
total ignorance of the need to judge academic productivity on principles quite more than three clear months of my sabbatical to run. I plunged straight away
different from those applicable to industry. Our masters show some small into work on the present book. Instead of revising, compressing and tailoring
awareness that, as in industry, quality is relevant as well as quantity: their the enormous amount of material I had already amassed, however, I ignored
performance indicators are sometimes modified by the use of more sophisti­ it altogether, and started writing afresh, on an entirely different plan, indeed
cated criteria, such as counting the number of references made by other writers virtually the opposite plan. Instead of arranging the book as one might arrange
to a given article. Frege would never have survived such a test: his writings a systematic non-historical treatise on the philosophy of mathematics, I com­
were very seldom referred to in his lifetime. It is not, how'ever, that quantity posed it as a close study of Frege’s texts: that is, of his Gnindlagen, followed
is not the only criterion, but that it is positively harmful. The reason is that by selected parts of the later Gnindgesetze. Into this w'ere to be inserted some
overproduction defeats the very purpose of academic publication. It long ago comparative matter concerning Frege and Dedekind, and Frege and Husserl:
became impossible to keep pace with the spate of books and of professional not, however, for its own sake, but as illuminating Frege’s texts. The Gnind­
journals, whose number increases every year; once this happens, their pro­ lagen is written with a deceptive clarity: it is in fact a very easy book to
duction becomes an irrelevance to the working academic, save for the misunderstand. My original plan for my book on Frege’s philosophy of math­
occasional book or article he happens to stumble on. This applies particularly ematics had left readers without a helpful guide to the subtleties, and artfully
to philosophy. Historians may be able to ignore much of their colleagues’ work concealed lacunae, in the argument of Gnindlagen'. I should do them much
as irrelevant to their periods; but philosophers are seldom so specialised that better service, I now thought, by providing one.
there is anything they can afford to disregard in virtue of its subject-matter. Furthermore, the new’ plan concentrated attention on w'hat w’as central to
Given their need for time to teach, to study the classics of philosophy and to Frege’s philosophy of arithmetic. 1 had not intended, at the outset, to wTite a
think, they cannot afford to plough through the plethora of not bad, not good treatise of length comparable to that of Frege: Philosophy of Language. What
books and articles in the hope of hitting on the one that will truly cast light had swollen the book to beyond that size wras a misguided ambition to achieve
upon the problems with which they are grappling; hence, if they are sensible, comprehensiveness: I had thought I must include everything relevant to the
they ignore them altogether. philosophy of mathematics. A great deal of this - such as a chapter on Frege’s
Academics who delivered their promised manuscripts twenty years late used philosophy of geometry - has now’ been excised. Among the casualties has
to cause us amusement; but it was a respectful amusement, because we knew been a discussion of Frege’s views on the consistency of mathematical theories,
the delay to be due, not to idleness, but to perfectionism. Perfectionism can in terms of his controversy with Hilbert, which had been written in 1973 and
be obsessive, like that which prevented Wittgenstein from publishing another survived successive revisions intact. It had in fact been published in Matthias
book in his lifetime, and probably would have done so however long he had Schirn’s collection Studien zu Frege/Essays on Frege of 1976: when he asked
lived; but, as the phrase goes, it is a fault on the right side. Every learned me for a contribution, I selected the most self-contained passage from the
book, every learned article, adds to the weight of things for others to read, uncompleted typescript of the book. Since it has been published, and repub­
and thereby reduces the chance of their reading other books or articles. Its lished in my Frege and Other Philosophers, its omission from here is no loss.
publication is therefore not automatically justified by its having some merit: The topic is indeed of some interest; and there were other discussions, unpub­
the merit must be great enough to outweigh the disservice done by its being lished and now suppressed, on topics of similar interest. I decided, however,
published at all. Naturally, no individual writer can be expected to be able that the attempt to discuss everything in Frege’s writings that bore on the
accurately to weigh the one against the other; but he should be conscious of philosophy of mathematics had resulted, and could only result, in a diffuse,
the existence of such a pair of scales. We used to be trained to believe that rambling book. I have tried to replace it by one that goes to the heart of Frege’s
no one should put anything into print until he no longer sees how to make it philosophy of arithmetic, setting aside everything not of central importance for
any better. That, I still believe, is the criterion we should apply; it is the only
that purpose.
means that exists of keeping the quality of published work as high as possible, Frege’s reputation as a philosopher of logic, of language and of thought has
and its quantity manageably low. The ideologues who in their arrogance force
grown steadily from about 1950 onwards; he is generally perceived as the
their misconceived ideals upon us attempt to make us apply virtually the
founder of analytical philosophy. Not so his reputation as a philosopher of
opposite criterion: publish the moment you can get editor or publisher to
mathematics. His wnrk in this field has tended to be equated with maintenance
accept it. We are compelled outwardly to comply with their demands; let us
of the logicist thesis, and consequently dismissed as a total failure; it is ironic
inwardly continue to maintain our own values.
that, in his last years, he would have concurred with this judgement. He would
Wlien I returned from Stanford in 1989, it was early June, and I still had
have done so because he had aimed at, and for a time had believed that he
xii Frege: Philosophy ofMathematics Preface X1U

had achieved, total success; but, since no one has achieved total success, it his contemporaries perceived, despite his unawareness of much that concerns
requires explanation why that judgement should be made now. Hilbert, too, us but wholly failed to strike him, or could not even be formulated until logic
propounded a programme that proved impossible of execution as he formulated had made further advances, he is, in my judgement, the best philosopher of
it; and his philosophy of mathematics, as a system, would have been tenable mathematics. This book is a historical study: but it has been w’ritten in the
only if that programme could have been carried out: yet no one regards belief that we can still profit greatly by reflecting on what Frege wrote about
Hilbert’s views on the subject as negligible. Probably the reason is that Frege’s the foundations of arithmetic, and therefore in the hope that it is not merely
work does not prompt any further line of investigation in mathematical logic, a historical study.
unlike the modifications of Hilbert’s programme studied by Georg Kreisel. It
M.D.
does not even appear to promise a hopeful basis for a sustainable general Oxford, July 1990
philosophy of mathematics: while it is appealing to be a neo-Dedekindian like
Paul Benacerraf, or a neo-Hilbertian like Hartry Field, neo-Fregeanism,
though espoused by Crispin Wright and by David Bostock,3 seems to most to
be considerably less attractive.
Various features of Frege’s work in the philosophy of mathematics have
contributed to the general neglect of it. An inborn obstinacy combined with
his increasing bitterness to make him ever less receptive to the ideas of others.
He had a great early interest in geometry, particularly projective geometry;
and in Grundlagen he alluded to non-Euclidean geometry in a perfectly reason­
able way, categorically affirming the consistency of elliptic geometry but observ­
ing that we cannot imagine such a space. Subsequently, he became a fierce
opponent of non-Euclidean geometry, descending, in a fragment of his Nachlass
of which it is to be hoped that he was not later proud, to comparing it, as a
pseudo-science, with alchemy. He allowed no merit to Hilbert’s Grundlagen
der Geometrie, nor, in his Grundgesetze, Volume II, to either Cantor’s or Dede­
kind’s theory of real numbers; and, although he lived until 1925, he paid
scarcely any attention to the work of his successors in mathematical logic. Some
explanations, psychological or intellectual, can be given for these attitudes. He
continued to regard geometry as the science of physical space, and so held
that there can be only one true geometrical theory. His early respect for
Cantor, manifested in Grundlagen, was repaid by the cruelty of Cantor’s mean-
spirited review of that book. Yet, whatever may be said in mitigation, these
evidences of the blindness and lack of generosity which were such marked
features of Frege’s work after 1891 combine with his great blunder in falling
into the contradiction to suggest that he cannot have much to teach us.
Nevertheless, his work in this field deserves great respect. It certainly cannot
be reduced to the bare statement of the logicist thesis. There is much that he
found worth saying, or said for the first time, that is either obvious to us or a
received part of very elementary logic or mathematics; but there is also much
that remains challenging. A good deal, indeed, is patently wrong; but of which
philosopher of mathematics is that not true? Despite his blindness to things
 CHAPTER 1

The Significance of Grundlagen

Die Grundlagen der Arithmetik is Frege’s masterpiece: it is his most powerful

and most pregnant piece of philosophical writing, composed when he was at
the very height of his powers. It was written as a prolegomenon to his magnum
opus, Grundgesetze der Arithmetik-. a first rough sketch of Part II of that work,
presented without unfamiliar symbolism and with a minimum of symbolism of
any kind, in the hope of reaching as wide an audience as possible. But it
occupies both a more central and a more problematic place in his work on the
philosophy of arithmetic than this intention would suggest. What he did not
foresee, when he was composing it, was that, in starting work on Grundgesetze,
he would be led to make fundamental changes both in his formal logical system
and in his underlying philosophy of logic. It is the system of logical and
philosophical doctrines that Frege elaborated as embodying these changes
which we think of as constituting his philosophy; and it was in the framework
i of this system that the two volumes of Grundgesetze were written. This suggests
that Grundlagen should be set aside as a brilliant but immature work, and that
we should study Frege’s philosophy of arithmetic primarily from his Grund­
gesetze. We cannot do that, however, because he chose not to carry out, on a
revised basis, a philosophical justification of his theory of natural numbers of
I the kind that had occupied most of Grundlagen-. Part II of Grundgesetze, which
corresponds to Grundlagen in subject-matter, is wholly formal in character,
being written almost exclusively in Frege’s logical notation, and thus entirely
omitting the philosophical argumentation. It was not that Frege had come to
consider such argumentation superfluous, for he supplies it at great length in
Part III of Grundgesetze, which treats of the foundations of the theory of real
numbers, a subject left untouched in Grundlagen. It must have been, rather,
that he considered that readers could easily transpose the argument of Grund­
lagen into the mode of his new system of philosophical logic. If so, he gravely
underestimated the difficulty of the task, which to this day creates problems
not easily solved. We have no choice, however, but to treat Grundlagen as
presenting the greater part of the philosophical underpinnings of the theory
of the foundations of arithmetic expounded in Grundgesetze, while bearing in
1
 2 Frege: Philosophy ofMathematics
1 The Significance of Grundlagen 3
mind that, if he had incorporated this material into Grundgesetze, he would
which Frege’s views changed, sometimes subtly and, in some instances, rad­
have subjected it to substantial modification.
ically. To recognise this, it is sufficient to compare what Frege wrote in
Grundlagen is deceptively lucid. That is not at all to say that it is deliberately
Grundlagen with the remark in the article ‘Booles rechnende Logik und die
misleading; only that it is so persuasively written, and so adroit in its selection
Begriffsschrift’, which in 1881 - only three years before the appearance of
of the rival news that are then so skilfully refuted, that it is easy to overlook
Grundlagen - he unsuccessfully submitted for publication, that ‘individual
the options that have not been presented to their best advantage, or at all, and
things cannot be assumed to be given in their totality, since some of them,
to misconstrue the architecture of the argument as it is developed from begin­
such as numbers for example, are first created by thinking’.1 We therefore
ning to end of the book. We have here to renew the course of that argument
cannot presume that what he wrote at one time during his early period he
so as to bring to light all that is not apparent on first reading.
would have continued to endorse at a later time, though it is natural to suppose
Grundlagen is written in the framework of a Kantian terminology', not used
that he regarded later thoughts as better. The greatest difficulty is to decide
by Frege in any of his writings after 1890, save those composed at the very
how much carried over from the early to the middle period. Naturally, when
end of his life. This terminology does not indicate his acceptance of any what he wrote in his middle period expressly corrected or modified something
specifically Kantian doctrines: indeed, despite the tone of deep respect he
he had said in the early period, we know exactly where we are: but what when
frequently, though by no means invariably, adopts when speaking of Kant, he
he was simply silent?
overtly discusses Kant’s news almost exclusively to disagree with them. Frege’s This question is particularly acute in relation to Grundlagen, because three
use of his terminology' may be due to a special effort to make himself under­
salient doctrines of that book were never afterwards explicitly reaffirmed by
stood by the professional philosophers; more probably, to his simply assuming
Frege, but never explicitly denied by him, either. The first is the ‘context
that a Kantian framework was the proper one within which to pose philosophi­
principle’, that it is only in the context of a sentence that a word has meaning.
cal questions. The brilliance of Grundlagen makes it easy to forget that it was,
This has been much discussed: I believe that a definitive answer can be given
after all, his first full-fledged incursion into philosophy.
to the greatly controverted question whether he repudiated or maintained it,
and shall give that answer in its proper place. The second is the adoption by
The status of Grundlagen Frege, in § 3 of Grundlagen, of the Kantian classification of true propositions
into analytic, synthetic a priori and a posteriori, and his recharacterisation of
The principal problem of Frege exegesis is to determine the relation between
these three classes. The very object of the book is stated, in § 87, as having
the writings of Frege’s early period, up to 1886, and those of his middle
been to make it probable that ‘the laws of arithmetic are analytic judgements
period, beginning in 1891. During the years 1887—1890, he published nothing,
and consequently a priori’; and yet, throughout his middle period, Frege never
but was engaged in thinking through afresh his system of philosophical logic
employed these or any equivalent terms. It is instructive to read the different
and redesigning, in accordance with it, the formal system he had presented in
Begriffsschrift. He announced his new' ideas in the lecture Function und Begriff way in which he stated the object of Grundlagen in the first sentence of his
of 1891. The principal changes in his philosophical logic were the introduction Introduction to Grundgesetze. ‘in my Grundlagen der Arithmetik I sought to make
of the far-reaching distinction between sense and reference, and the identifi­ it probable that arithmetic is a branch of logic and that no ground of proof
cation of truth-values as objects and as the references of sentences. The needs to be drawn either from experience or from intuition.’ Not only is this
principal changes in his formal system were the introduction of value-ranges,
more accurate, in that to call a proposition ‘analytic’, in the sense of Grundlagen,
and the obliteration of any formal distinction between sentences (henceforward is not to say that it is expressible in purely logical terms: more importantly, it
called by him ‘names of truth-values’) and singular terms (‘proper names’); relates, not to individual propositions, but to an entire theory, taken as a whole.
I the addition of a description operator was an important secondary development. It is possible that Frege came to be dissatisfied, either with the manner in which
During the middle period, lasting from 1891 to 1906, his thought evolved he had defined ‘analytic’ and ‘a priori’, or with those concepts themselves; if
little. Doubtless much of what he wrote was newly thought out: but there is I so, it is puzzling that he never said so, but, if not, equally puzzling that he
refrained from ever employing them again until 1924. The third doctrine never
no reason to suppose that he ceased, at any later time within this period, to
believe anything that he w'rote for publication at any time during it. The logical again heard of after Grundlagen is that which introduced the pregnant concept
basis of all the work of the middle period w'as presented complete and entire of a criterion of identity: ‘if we are to use the symbol a to designate an object’,
in Function und Begriff-, and it scarcely altered throughout the whole period. he pronounced in § 62, ‘we must have a criterion which decides in all cases
The early period, by contrast, wras one of considerable development, during
' Nachgelassene Schriften, p. 38, Posthumous Writings, p. 34.
4 Frege: Philosophy ofMathematics 1 The Significance of Grundlagen 5
whether b is the same as a, even if it does not always lie within our power to becomes necessary to check that the proofs can still be carried through. The
apply this criterion.’ This is an immensely important dictum: in this third impossibility' of what Frege here claimed, with misplaced confidence, to be
example, it is especially mysterious that the whole topic should apparently have able to do, he took nearly four years to discover. He should not be blamed
vanished from his thinking. for this. In 1902-3 he was occupied with his first series of articles against
More important than whether, or to what extent, Frege continued during Hilbert’s Grundlagen der Geometric, in 1903-4 with his article ‘Was ist eine
his middle period to maintain these three particular doctrines is the question Function?’, and in 1905-6 with his second series of articles against Hilbert:
whether or not we may take the philosophy of arithmetic expounded in Grund- far more serious, in 1904 his wife died. But the fact is that, as soon as Frege
lagen to be essentially that to which he subscribed during the middle period. enquired into the question w'hether the proofs of the theorems of Grundgesetze
That the actual logical construction of the theory of the natural numbers, and would still go through under tire weakened axiom (V7'), he would have found
of cardinal numbers generally, remained the same is beyond question, since that they did not: not even the proof of the theorem (111) that 0 does not
it is repeated in Grundgesetze in more detail but in essentially the same way equal 1.
that it is sketched in Grundlagen: what needs to be decided is whether the From his unpublished waitings, w-e can pinpoint the moment at which he
philosophical ideas remained the same, allowing for the more sophisticated discovered this catastrophic fact. In 1906 he began writing a reply to an article
philosophical logic Frege had elaborated in the meantime. This question can by Schoenflies on the paradoxes of set theory', w'hich had appeared in the
be answered by considering the architecture of Grundgesetze. January' issue of the Jahresbericht der deutschen Mathematiker-Vereinigung in that
year. The unfinished draft contains a reference to an article by Korselt that
The structure of Grundgesetze appeared in the March-April issue of the same journal; as the editors of the
Nachgelassene Schriften remark, and as his footnote references to ‘this journal’
Grundgesetze, as we have it, is divided into three Parts; but it is an uncompleted indicate, Frege obviously intended to submit his reply to the Jahresbericht. The
work. The division into volumes has scarcely any relation to the segmentation article w’as never completed and never submitted, however; but his plan for it
of the book: it looks as though Frege had an agreement with his publisher contains an item showing clearly that, w'hen he drew' it up, he still believed in
that a certain number of pages constituted a volume, and the publisher brought his solution to the contradiction:
out a volume as soon as he had copy amounting to that number of pages. At
any rate, the two volumes are of almost precisely the same length: Volume I Russell’s contradiction cannot be eliminated in Schoenflies’s way. Concepts which
has 254 pages of text, with 32 pages of Preface and Contents, making 286 in agree in their extension, although that extension falls under the one but not under
all, while Volume II has 253 pages of the main text, with 16 pages of Contents the other.
and 13 pages of the Appendix dealing with Russell’s contradiction, which we
know to have been added in proof, making 282 pages in all; perhaps Frege The draft breaks off before this point. A tiny fragment is headed ‘Was kann
withdrew a section in order to make room for the Appendix. Volume I contains ich als Ergebnis meiner Arbeit ansehen?’ (‘What can I regard as the outcome
all of Part I and about three-quarters of Part II; Volume II contains the rest of my work?’), and begins ‘Almost everything hangs together with the logical
of Part II and about two-thirds of Part III: possibly Frege planned a fourth notation {Begriffsschrifty-, it goes on to list various of his logical doctrines,
Part, or possibly Volume III, had it appeared, would have been shorter. remarking in passing that ‘the extension of the concept, or class, is not the
Volume III did not appear because Frege came to realise that his solution first thing for me’. There follows in the Nachgelassene Schriften a relatively
to Russell’s contradiction, set out in the Appendix, was inadequate. The last lengthy ‘Einleitung in die Logik’ (‘Introduction to Logic’), the stages in whose
paragraph but one of the Appendix, dated October 1902, reads as follows: composition have been dated by Frege himself, the first having been written
on 5 August 1906. Once during his early period, and again in 1897, he had
It would take us too far here to pursue further the consequences of replacing [the attempted to write a systematic exposition of his philosophical logic as a whole:
ongjnal axiom] by [the proposed modification] (V'). It must be acknowledged
that to many of the propositions auxiliary hypotheses will have to be added- but the Logische Untersuchungen of his late years are the first three chapters of a
there need be no anxiety that any essential obstacles to carrying out the proofs final attempt. The ‘Einleitung’ was his third attempt; and it follows very exactly
th’S' !l WU nevertheless be necessary to check thoroughly all the sketch contained in ‘Was kann ich als Ergebnis meiner Arbeit ansehen?’.
propositions discovered up to this point. “ It is plain enough what had happened. In the course of writing his anti-
Schoenflies article, presumably as the result of a belated enquiry into the
That of course is correct: when one of the axioms of a theory is weakened, it consequences for the proofs of Grundgesetze of the weakening of Axiom (V)
6 Frege: Philosophy ofMathematics 1 The Significance of Grundlagen 7
proposed in the Appendix, Frege had come to realise that his solution to Frege had arrived at a new philosophy of arithmetic, differing markedly
Russell’s contradiction did not work. As the final paragraph of that Appendix, from that expounded in Grundlagen. Arithmetic can no longer be taken as
and hence of Volume II of Grundgesetze, he had written: founded on logic alone; hence, as Frege maintains in two works written in the
last year of his life, since it remains a priori in character, it must rest, as Kant
We may regard as the fundamental problem of arithmetic the question, how do had taught, on pure spatial or temporal intuition. In these late writings he
we apprehend logical objects, and in particular the numbers.'' What justifies us in declares, what he must for some time have believed, that set theory is an
recognising the numbers as objects? If this problem has not yet been so completely
solved as I thought when I wrote this V olurne, I do not doubt that the way to its illusion generated by language, which misleads us into taking such a phrase
solution has been found. as ‘the extension of the concept fixed star1 as standing for an object. There was
not the time remaining to him to develop these new' thoughts into a complete
Now’ he was faced with the realisation that he had not even found the way to theory.
it. His life’s work had been to construct a definitive foundation for number
theory and analysis, so that their content and their justification need never The contents of Grundgesetze
again be thought problematic, and he had believed that he had succeeded:
All this explains why no further volume of Grundgesetze was ever published;
now' he had to acknowledge that he had failed. His task now wras to salvage
but one was obviously intended. If it had contained only the completion of
from the wreck whichever of his ideas remained undamaged, those, namely,
Part III, it would have been much shorter than Volumes I and II. Part I
not dependent on the notion of a class or extension of a concept. This task
expounds the formal system. It sets out the primitive vocabulary, formation
he, with great courage, immediately undertook, even though he eventually
rules, axioms, rules of inference and some definitions. It also contains an
lacked the heart to carry it through. We may thus set the date of his discovery' exposition of Frege’s system of philosophical logic, formulated with exactitude
that his solution of Russell’s contradiction w'ould not work between April and but without argument or justification, and, in terms of that, gives in detail the
early August, 1906. We need not suppose that he ever knew that the modified semantics of the system; references to ‘Uber Sinn und Bedeutung’,2 Function
system wras still inconsistent, though he may possibly have suspected it: if you und Begriff? and ‘Uber Begriff und Gegenstand’4 early in the book direct the
cannot prove that 0 and 1 are distinct, you are unlikely to be able to prove the reader to treatises in which he can find a justification of the apparatus
values true and false distinct, and may even be able to prove their identity. employed. Part II contains the logical construction of the natural numbers,
taken as finite cardinals, and proves various fundamental theorems concerning
The late period them and concerning the least transfinite cardinal, called by Frege (dte Anzahl
Endlos' (the number Endless), corresponding to Cantor’s Aleph-0. The incom­
It is from August 1906, then, that: we may date the beginning of Frege’s late plete Part III consists of Frege’s theory of real numbers, incompletely
period. Ven' little was published save the three essays forming the Logische
I expounded. Possibly, if Volume III had ever been published, it would have
Untersuchungen, ‘Der Gedanke’ and ‘Die Vemeinung’ in 1918 and ‘Gedanken- contained a fourth Part, dealing with complex numbers. From the usual stand­
gefuge in 1923. There is very little unpublished material, even, most of it point, it is trivial to construct the complex numbers, given the reals. Part III
concerned with the philosophy of logic rather than of mathematics (even when shows, however, that Frege wished to define the real numbers in such a way
it is applied to mathematics); Frege deliberately put aside the central problems as to make the possibility of applying them to physical reality manifest in their
of the philosophy of arithmetic. It was not until 1918 or 1919 that he rallied definition, and he may have wanted to do the same for the complex numbers;
enough to address himself once more to them. He explained the matter in a when they are defined as ordered pairs of reals, with the appropriate definitions
letter of that period which he wrote to Karl Zsigmondy: of their sums and products, their application within physics, and even the
mathematical theory of functions of a complex variable, are far from immedi­
ately evident.
However this may be, there is a signal difference, already alluded to, between
Parts II and III of Grundgesetze. Part III is divided into two halves. The
tid Ais workPwWch "°. 10ngiF °ffiC-ai!ly concernin& myself with the matter. 2 Vol. I, Preface, p. ix fn., and Introduction, p. 7 fn.
3 Preface, p. x, Introduction, p. 5 fn. and § 21, p. 36 fn.
■* Introduction, pp. 3 fn., 5 fn. and 8 fn.
 8 Frege: Philosophy ofMathematics 1 The Significance of Grundlagen 9
uncompleted second half is a formal development of Frege’s construction of make the necessary adjustments themselves. Hence, despite some serious
the real numbers; the first half is a prose justification of that construction. It uncertainties, we may consider Grundlagen as expressing, with fair accuracy,
is designed after the model of Grundlagen: alternative theories of the real Frege’s mature philosophy of arithmetic, not merely a superseded phase of his
numbers, including Cantor’s and Dedekind’s, are reviewed and criticised, so thinking.
that, in the course of the critique, Frege’s requirements for a correct theory
emerge; at the end of the discussion, those requirements are summarised and
an advance sketch is given of the construction to be developed within Frege’s
formal system in the second half of Part III. Part II, on the other hand,
corresponds only to that second half. It consists entirely of a series of proofs
and definitions within the formal system, together with the brief prose expla­
nations that accompany the formal proofs throughout Grundgesetze, and com­
pletely lacks any argumentative justification for the theory such as is provided
by the first half of Part III for the construction of the real numbers.
Why this asymmetry? The obvious answer is that previously suggested: that
Frege was satisfied that he had already provided such a justification in Grund­
lagen, and that there wras therefore no need to do it over again. No other
conclusion seems possible; and it is reinforced by the repeated references to
Grundlagen in Grundgesetze,5 together with the reference to ‘Uber formale
Theorien der Arithmetik’,6 a lecture given just after the publication of Grund­
lagen and fully expressing its point of view. In none of these is a note of
caution sounded, like Frege’s warning to the reader that Begriffsschrift no longer
corresponds to his present standpoint.7 Frege of course was well aware that
Grundlagen would need rewriting to adapt it to the later doctrines, and
especially the sense/reference distinction; in ‘Uber Begriff und Gegenstand’
he said, for example:8

When I wrote my Grundlagen der Arithmetik, I had not yet made the distinction
between sense and reference and hence, under the expression ‘judgeable content’,
grouped together what I now distinguish by the words ‘thought’ and ‘truth-value’.
I therefore no longer wholly approve of the explanation given on p. 77, although
I am essentially of the same opinion.

I Probably he did not realise how far-reaching such a rewriting would have to
be; there is much in the book that he would have considered wrrong, or at
least would not have cared to say, in 1893. But we may take it as certain that
he thought it sufficiently near his current views to make such rewriting redun­
dant, and assumed that readers aware of his later doctrines would be able to

5 In vol. I they occur on pp. viii-xi of the Preface, pp. 1 and 3 of the Introduction, p. 14 of § 9,
p. 56 of § 38, pp. 57-60 of §§ 40-6, and p. 72 of § 54.
6 Introduction, p. 3.
’ In the footnote on p. 5 of the Introduction.
8P. 198.
 2 The Introduction to Grundlagen 11
propositions has still to be solved: the value of Frege’s unsuccessful attempt
lies in its pinpointing the place where the difficulty lies.

Mathematical value
CHAPTER 2
The motivation for the work is set out again in §§ 1 and 2 of the main text,
where Frege represents his investigation as in line with the general drive
tow ards greater rigour in the mathematics of his time. In these sections, it is
The Introduction to Grundlagen the mathematicians whom Frege is principally trying to persuade; he had
chiefly aimed his remarks in the Introduction at the philosophers, although he
everywhere insists that the enquiry lies on the borderline between the two
subjects. There would be little point in dwelling on the reasons Frege gives
Frege begins his Introduction by persuasively setting out the need for an for undertaking the investigation, so obvious must its interest appear to almost
enquiry of the kind undertaken in the book. Mathematicians - the mathema­ all with any philosophical inclination, were it not that there has been a recent
ticians of his day - are, he says, unable to give any sensible answer to the movement, led by Philip Kitcher, to argue that it was indeed pointless. The
question what the number 1 is, or what a number in general is: and so they argument is that, unlike the clarification of the foundations of analysis, it
are incapable of explaining what arithmetic is about. This is a disgrace to the wras not needed for the resolution of antinomies hampering the progress of
science, which urgently calls for a remedy. Frege remarks that many will mathematics. This might be thought the expression of a philistine attitude
reckon such an enquiry not worth the trouble, supposing that the matter is towards philosophy on the part of certain mathematicians by anyone unaware
already well understood; but this only shows how deep the trouble lies - we that it actually proceeded from philosophers. Such philosophers reduce them­
do not even know that we do not know. selves to the status of the repairmen of the sciences, not needed until called
Until he received Russell’s letter of 16 June 1902, informing him of the in to clear up some confusion that is impeding the important w’ork of the
inconsistency of his formal system, Frege believed that he had found the scientists. The questions what the natural numbers are, and how we know-
definitive solution to the problems of the foundations of number theory and what we assume to be true about them, are of intrinsic interest, whether or
analysis: the definitive answers to the questions on what our knowledge of the not the answers contribute to progress within number theory’: since they go to
truths of those mathematical theories rest, and w'hat the two theories are about. determine what number theory’ is about, and w'hat its epistemic basis is, they
If he had really resolved these important philosophical problems, the value of lack interest only if either number theory’ itself is of no value, or philosophy
his work would be beyond all doubt; since the problems remain unresolved, as a whole is devoid of interest.
they can hardly be dismissed as trivial. Our task now is to answer three There is indeed a significant contrast between the contemporary' but inde­
questions: what Frege tried to achieve; w'here he failed and why; and how’ pendent w’ork of Frege and Dedekind on the foundations of number theory’;
much he actually established. Almost everyone recognises that mathematical the difference could certainly be characterised by saying that Dedekind’s
propositions differ in status from empirical ones: they are arrived at by a approach was more mathematical in nature, Frege’s more philosophical.
process of reflection and reasoning in which observation plays no part, and Plainly, contributions to the philosophy of mathematics are not to be judged
they are invested with a necessity that bars us from conceiving what the by how much they contribute to mathematics itself, any more than contributions
world would be like if they did not hold. Frege attempted to show that some to the philosophy of mind are to be judged by whether they advance the
I
mathematical propositions, those of number theory and analysis which he science of psychology. In any case, it is an illusion to suppose that Frege’s
jointly classified as ‘arithmetic’, had the same character as, and in fact were, foundational work was of no mathematical value, even if this judgement is
logical propositions; he never believed this to be true of the whole of mathemat­ restricted to number theory, with which Grundlagen is almost exclusively con­
ics. The set-theoretic contradictions rendered his attempt a failure. They did cerned, and not applied to his work on the foundations of analysis, presented
not, however, invalidate the whole attempt from start to finish: the argument only in the second volume of Grundgesetze. The illusion occurs for several
for the logical character of some simple propositions - numerical equations, reasons. One is the error of considering Grundlagen in isolation from the
for instance, or the commutative law for cardinal addition - remains unaffected. previous w’ork embodied in Begriffsschrift - a facile mistake arising from the
For the rest, the problem of explaining the special character of mathematical absence of logical symbols from Grundlagen-, but in fact the later wrork depends
10
 12 Frege: Philosophy ofAlathematics 2 The Introduction to Grundlagen 13
on the earlier, which had been carried out in preparation for it. It is not only
distinguishing what belonged to number theory proper from its logical foun­
that, in § 79 of Grundlagen, Frege borrows from Begrijfsschrift the celebrated
dations, precisely because he believed there to be no sharp line between
definition of the ancestral, to yield, in § 83, a definition of natural numbers as
arithmetic and logic. Nevertheless, in Grundlagen and in Grundgesetze, he
those objects for which finite mathematical induction holds good — a definition
presented proofs of a number of general propositions, labelled in Grundgesetze
which Frege saw as serving to eliminate appeals to intuition or to specifically ‘the basic laws of cardinal number (/tnzahl)'. Given that Frege was operating
arithmetical modes of reasoning. It is also that the possibility of completely with a successor relation rather than with a successor function, his ‘basic laws’
formalising mathematical proof underlies the entire programme, as is made had to be more explicit than the Peano axioms: he needed to prove that even'
clear in §§90 and 91: only by means of a formalisation that precludes a natural number had a successor, and that nothing had more than one. Given
surreptitious appeal to intuition can we attain certainty that the theorems of this difference, his basic laws in effect comprise the five Peano axioms, not,
number theory rest on a purely logical foundation. Plainly, inventing modern however, isolated as an axiom-set entailing number theory as a whole: the first
mathematical logic, and devising the very' first formal system, were major two, saying that 0 is a natural number and that a successor of a natural number
contributions to mathematics under any but the narrowest circumscription of is a natural number, and the fifth, embodying the principle of induction, are
what constitutes mathematics. incorporated into his definition of ‘natural number’ (‘finite number’ in his
A second reason for the illusion is that much of what Frege laboured to terminology), from which they are immediate, rather than being formulated as
make clear is now common currency with us: no one would now regard as theorems. The third Peano axiom, that 0 is not a successor, figures as theorem
anything but ludicrous the explanations of the concept of number that eminent 108 of Grundgesetze, and as part of theorem 6 of § 78 of Grundlagen, while the
mathematical contemporaries of Frege were satisfied to give, but he criticised fourth, that successor is one-many, appears as theorem 89 of Grundgesetze,
so trenchantly. It is possible also because the notorious failure of the most and as half of theorem 5 of § 78 of Grundlagen. Frege’s basic laws include
salient part of Frege’s programme - the reduction of arithmetic to logic, some propositions concerning the number 1, which of course is not a primitive
taken as a whole - obscures the success of another part forming an essential notion in the Dedekind-Peano axiomatisation (when 0 is taken as the starting-
preliminary to it. To describe him as reducing arithmetic to set theory, and point), and others not expressible in purely number-theoretic terms, but con­
then to disparage that reduction as unimportant, as is sometimes done, is to cerned with one-one correspondence. He ought, indeed, to have seen the
caricature both what he intended and what he accomplished. The description necessity for isolating certain of the laws as at least forming a plausibly suf­
has, as a background assumption, what no one now would doubt, that set ficient base for the derivation of all truths of number theory'; but although he
theory is an autonomous mathematical theory, in no way to be identified with did not do this, he came far closer than anyone had done previously to analysing
logic. Frege valued his reduction only so long as he believed it to be a reduction the basis of number theory'. Frege was not concerned to present number theory
to logic: as soon as he abandoned hope of a relatively simple means of avoiding as an axiomatised mathematical theory in the ordinary sense, and it is therefore
the contradiction, he deemed it to have been a mistake to treat the theory of not surprising that, regarded from that standpoint, Dedekind’s work was
classes as a part of logic; the reduction thereupon ceased to interest him. superior to his; it is nevertheless ludicrous to suggest that this aspect of Frege’s
Frege’s aim, as stated in Grundlagen, was to make it probable that the truths work was of no mathematical value.
of number theory are analytic, in the sense he gave to that Kantian term. To
do so, he did not propose to examine in turn all the theorems in some current

J textbook of number theory'. Rather, it sufficed to provide such a demonstration

for the fundamental principles of arithmetic: the rest would then follow of
itself. This therefore made it necessary for Frege to identify those fundamental
Psychologism
By a natural train of thought, Frege passes in his Introduction from explaining
the motive prompting his enquiry to the first of his many polemics against
/ principles. Surprisingly, no attempt had yet been made to isolate the laws, or psychologism. His opposition to it becomes ever harder for philosophers to
even the concepts, from which number theory could be developed, despite the comprehend, at a time when what passes in the American philosophical schools
universally admired example provided by Euclid of how this could be done for die ‘standard reading’ of Frege is itself whole-heartedly psychologists.
for geometry’. Frege’s pioneering work in this regard has been overshadowed At this stage of his career, Frege was interested solely in the content of our
by that of Dedekind. It is a valid criticism of him that he did not actually statements, and not at all in our grasp of that content. Later, he acquired a
n’Tr11? ™mber theoiy: as is now generally known, that was done by strong interest in the latter: his notion of sense, as set out in his writings from
Dedekind whose axiomatisation was adopted by, and named after, Peano. 1891 onwards, has to do precisely with understanding-, the sense of an
Frege did not do this, since he had no strong reason to be interested in expression is something that we grasp. What made it possible for him to go
14 Frege: Philosophy ofMathematics 2 The Introduction to Grundlagen 15
immensely further towards a satisfactory account of understanding than anyone containing the expression defined. Once we recognise that, we see that psy­
had done before in the history of philosophy, and certainly far further than chology - the description of inner mental operations or of their hypothesised
any of his contemporaries, was, however, that he had started with the notion physiological correlates - has no place in mathematics or logic.
of content, and that he therefore fashioned a theory of sense in accord with Frege, concentrating on the content of the expressions he was concerned to
that, rather than trying to explain content in terms of our grasp of it. analyse, determined that his definitions should be fruitful in the manner that
At the time Frege was writing, psychologism was not a mere tendency, but definitions in mathematics - at least, those given after the first two pages -
an explicit philosophical doctrine, consciously held and widely subscribed to: ordinarily were: and this meant that they must serve to determine when the
not Frege’s attacks on it, but those of Husserl in his Prolegomena zur reinen sentences containing them were true and when they were false. For that, on
Logik of 1900, the first part of his Logische Untersuchungen, first loosened its Frege’s new, was what distinguishes thoughts from eventhing else, namely that
grip on the German philosophical community. Even as no more than an they may meaningfully be called ‘true’ or ‘false’. Everything else is irrelevant
unconscious inclination, however, it is rather a natural one for a professional to the content of a thought - to what thought it is: when - and only when -
philosopher. When a child asks an adult what ‘sister-in-law’ means, the adult it is determined under what conditions a thought is true, it is thereby deter­
will not refer to any inner mental processes that accompany hearing the word mined what its content is.
‘sister-in-law’, but will tell the child in what cases one person is rightly said Grundlagen is, of course, a work of Frege’s early period, when he operated
to be another’s sister-in-law. That was what the child needed to be told: and with an undifferentiated, and not very precisely analysed, notion of content.
the adult tells him that because he is not concentrating on the notion of In the interval, from 1886 to 1890, between his early and middle periods, he
meaning, but takes it for granted. The philosopher, on the other hand, is, very developed his famous theory of sense and reference, which he expounded in
properly, perplexed by the notion of meaning. He quite rightly regards it as the middle period (1891-1906). Because he had first concentrated on the
an extraordinary thing, demanding explanation, that words - noises that issue notion of content, his theory of sense was elaborated from that model.
from our mouths or marks we make on paper - should have meanings. He Thoughts - the senses of sentences - are, on his account, intimately connected
naturally thinks that their possessing them depends on what goes on in our with the notion of truth - a notion belonging to the theory of reference. Our
minds. All that physically occurs when two people converse is that they alter­ grasp of the sense of an expression is our way of apprehending what its
nately make certain noises: the fact that they are exchanging thoughts, asking reference is - a particular way, out of various possible ways; and our grasp of
questions, giving information, raising objections, etc., must have to do, the the thought expressed by a sentence is constituted by our apprehension of the
philosopher concludes, with what takes place in their minds, where a connec­ condition for it to be true.
tion has been established between the noises that they emit and the ideas they The notion of understanding - of a grasp of sense - is of crucial importance
thereby express and convey. What makes the difference, he thinks, is that each to a philosophy of either thought or language. Thoughts can be conveyed, and
interprets the utterances of the other; and so he is driven to concentrate upon are conveyed by language; a philosophical account of communication is obvi­
the inner process of interpretation. The philosopher’s disposition to think in ously impossible without an account of what understanding is. That is why
this w ay is reinforced by his addressing himself to words whose meanings are Frege was quite right to interest himself in the notion in his middle and late
more fundamental, and frequently more abstract, than ‘sister-in-law’, and periods. Now understanding is a grasp of content: sense can therefore be
therefore harder to explain; either in conformity to a general methodology, or explained only as a way in which content is grasped. We can arrive at a
without explicitly noticing what he is doing, he substitutes for an explanation plausible account of sense only if we first have a workable conception of
ol content an account of mental operations accompanying the use or hearing content - of that which is grasped; and that is why Frege arrived, for the first
ol the word or leading up to a grasp of its meaning. time in the history of philosophical enquiry, at what was at least the beginnings
Frege made the simple observation that anyone would make if offered a of a plausible account of sense, and thus of understanding. Those w'ho started
psychologistic: account when he asked what ‘sister-in-law’ meant: you can with the conception of the inner grasp of meaning floundered in confused
PUrpOrted <explanation’- You cannot use it to decide that descriptions of irrelevant mental processes, achieving nothing towards explain­
ZTaXX 15 nOt’ thersister-in-law of somebody else; and so it has ing either the general notion of meaning or the meanings of specific
XhScat C°ntent WOrd> If a mathematical term is explained expressions.
involvine it- defin’ri °U CannOt appeal t0 t^e explanation to prove a theorem The notion of content, as used in the foregoing discussion, is ambiguous:
sTse th w can7’ £ PUrp°Se’ muSt be in at ^ast this it wavers betw een the realms of reference and of sense. That was unavoidable:
sense, that we can use them to determine the truth or falsity of a statement it was indeed ambiguous as Frege used it in his early period, and it wras his
 16 Frege: Philosophy ofMathematics 2 The Introduction to Grundlagen 17
perception of that ambiguity that drove him to make the sense/reference
and it must be plausible to attribute a grasp of it to the speakers of the
distinction. The notion of content cannot be definitely located in either realm:
language in virtue of their understanding the expression. Frege respected these
the content of a sentence is obviously not identifiable with its reference, which
constraints: it is in fact they which force the distinction between sense and
is merely its truth-value, nor with any structure that might be imagined as
reference. He confined himself, however, to giving an account of what we
made up out of the references of its parts, somehow' held apart from one grasp, leaving it for psychology to explain the manner of our grasping it.
another so that function and argument refrain from yielding the corresponding The fundamental principle of analytical philosophy is the priority, in the
value. Nor can the notion of content be straightforwardly identified with the order of explanation, of language over thought: the only route to a philosophical
later notion of sense, because it is not conceived as correlative to an act of account of thought is through an analysis of its expression in words or symbols,
grasping it. That is why Frege always later said that he made the distinction that is, a theory of linguistic meaning. So long as this principle remained in
between sense and reference within the notion of content. W’hen he used the place, it was possible to fill the two lacunas in different ways, or, more precisely,
latter notion in his early period, how'ever, his attention was almost always to make the means of filling the second depend on that of filling the first. On
directed outwards, as it wrere, on what w as needed for the truth of a statement, such an account, to be found in its clearest and most explicit form in Witt­
rather than on our apprehension of that condition in accordance with the genstein, the sense of an expression consists in its role within the complex
manner in which it w’as stated. social practice constituting the communal use of the language, a practice open
That is not to concede that the notion of understanding can rest on quite to view' and not in itself involving any hidden mental operations. An individual
so objectivist a base as Frege believed. Sense, on Frege’s account, is our way speaker’s grasp of that sense then becomes one ingredient in his ability,
of grasping what the reference is; and the reference is something in the acquired by training, to engage in that practice. On this approach, if any
objective world, quite independent of us or our awareness of it. To have a explanation were needed of a possession of this ability, it would not belong to
Fregean grasp of sense, w'e must have a conception of what it is for a statement the philosophical order, but w'ould properly pertain to psychology'; such an
to be true, independently of our means of recognising its truth. There appears, explanation would be altogether irrelevant to a philosophical account of linguis­
how ever, to be no non-circular way of explaining what it is to have such a tic understanding, and hence of thought. In this way, the structure of the
conception, or hence of giving an account of understanding that does not Fregean theory' is fully safeguarded. A reversal in the order of dependence
presuppose w'hat it purports to explain. If there is not, a possible remedy is to between the w'ays of filling the two lacunas does not necessitate abandoning
replace the notion of knowing w'hat it is for a statement to be true by that of the fundamental principle of analytical philosophy; but it is one step in a retreat
knowing w'hat w'ould rightly lead us to recognise it as true. Such a substitution back to psychologism. This reversal involves taking, as the basic notion, not
of what may broadly be called a verificationist theory' of meaning for Frege’s that of the language common to a community, but the idiolect of a single
truth-conditional one w'ould greatly narrow' the gap between sense and refer­ individual. A speaker’s mastery of his own idiolect is taken as consisting in,
ence. It would nevertheless preserve the essential structure of the Fregean or at least resembling, a knowledge of a theory of meaning for it, and his
theory, since it w'ould still explain meaning as a communally recognised feature grasp of the sense of a particular expression as a constituent of that complex
of expressions, and understanding as the grasp of that feature, rather than knowledge. The sense of the expression in the common language can then be
c aracterising meaning in terms of mental operations taken as constituting explained as its sense in a majority of a range of overlapping idiolects. The
understanding.’ irreversible retreat to psychologism takes place w’hen, as increasingly within
There are two lacunas in Frege’s account, at opposite ends: one at the end the analytical tradition, the fundamental principle is jettisoned, and thought

J \
° wnT’ Other at
°re^essinS>
°f 3 sPeaker’s of sense. What constitutes
sense that it does, that is, its sense in the
treated as prior, in the order of explanation, to language. This development
is due, in part, to the instability of the intermediate position: since individual
sense consk^ u °ngS‘? *n w^at ^oes an individual’s grasp of that speakers manifestly' have no explicit knowledge of a theory of meaning for
I
/ Sat sense o p enSi°n °f sense in ™ his attaching their idiolects, the questions whether they can in any sense be said to know'
of these two ouect'^ h himse^ciid not so much as mention the former such a theory', and, if so, what constitutes their knowledge, or, if not, in w'hat
relegate k to dsX^ b™1101^ latter only to brand it a mystery' and simulation of knowledge their linguistic competence consists, become pressing.
a grasp of sense bein'’ H?S, the°r>' ?resents sense as something to be grasped, Since all attention is focussed on the abilities of the individual subject, the
temptation becomes irresistible to attempt a direct explanation of that subject’s
knowledge or sowing closely
it must be somethingta 1X1/ °f a" e*Pressi°n ca" be: processes of thought, considered as unmediated by their linguistic expression,
and append to it a hypothesis concerning the connection he then makes
° content of knowledge or of apprehension,
18 Frege: Philosophy ofMathematics 2 The Introduction to Grundlagen 19
— words
between the of his idiolect and features of his thought. At this stage
------------------ ment of the attack made by Frege and by Husserl upon psychologism can
psychologism has in effect been fully reinstated, even if, as with many nine­
afford to overlook the incontestable truth of that dictum.
teenth-century thinkers, scientific respectability is thought to be maintained
by ritual obeisance to materialism and an assurance that, ultimately, all will
The conflict between Frege and Husserl over psychologism
reduce to neurophysiology'.
It is uncontroversial that much of philosophy' is concerned w i th the analy sis In Chapter 1 of his recent book on Husserl,1 David Bell puts up a valiant
of concepts; and certainly Frege’s Grundlagen is occupied to a large extent defence of Husserl’s Philosophic der Arithmetik of 1891, rating it as giving a
with the analysis of numerical and arithmetical concepts. If a sound analysis better account of its subject than Frege’s Grundlagen, published seven years
is to be given of a concept or set of concepts, it must proceed in accordance earlier and criticised by Husserl in his book. Bell’s motivation for this lies in
with a correct conception, even if only implicit, of what the analysis of a his repudiation of the conventional view that Husserl’s first book was imbued
concept requires. Any such conception stands to be vindicated by the general with the psychologism of which Frege, in his review’ of the book in 1894,
philosophy of thought; when the philosophy of thought is approached via the perceived it as a salient example. As we have seen, Husserl later rejected and
philosophy of language, its vindication will be provided by the theory of attacked psychologism in his Prolegomena zur reinen Logik of 1900; a prevalent
meaning. It is for this reason that the theory' of meaning acquired so fundamen­ opinion is that he was prompted to his change of view by the severe criticisms
tal a place in the architecture of philosophy as practised by the analytical of Frege in his renew'. Bell thinks, on the contrary', that there w'as no change
school: a correct theory of meaning will determine what is to count as an of view' on this matter. According to him (p. 81), the psychologistic component
adequate analysis of the meaning of an expression, and hence of a concept. of the Philosophic der Arithmetik is far more restrained than Frege misunder­
The new that the meaning of an expression in a language consists in its having stood it as being, and Husserl’s position in the Logische Untersuchungen is
identical or similar meanings in a large number of idiolects, and that its indistinguishable from that of the earlier book. It is a pity that Bell devotes
meaning in an idiolect is to be characterised in terms of the workings of the only a hurried paragraph to the Prolegomena-, he surely owed his readers an
individual subject’s mind, entails that the analysis of a concept must ultimately explanation of w’hat he thought Husserl was attacking, if not the doctrine that
be given in psychological terms: precisely the view combatted by Frege in the he had formerly held. Husserl’s footnote, in which he retracts his criticisms,
Introduction to Grundlagen. Nevertheless, adherents of the new' psychologism in his first book, of Frege’s anti-psychologism, tells in favour of the more usual
interpretation;2 but Bell seems to have overlooked it, wrongly saying (p. 137)
are bound to concede that, although Frege’s remarks are couched in highly
that Frege is mentioned only' once in the whole of the Logische Untersuchungen.
general terms, he w'as right at the level with w'hich he w’as directly concerned.
An author who began in a condition of deep philosophical confusion, but then,
His principal object in Grundlagen was to determine the justification of the
by heroic efforts, eradicated that confusion, is certainly more interesting than
propositions of number theory, and of others involving the natural numbers.
one who, throughout his life, remained in a state of confusion. That is why it
In the Introduction, he argued that, for this purpose, psychological accounts
is important for Bell to demonstrate that there was no confusion in the
are valueless, and must be replaced by definitions that specify the contribution
Philosophic der Arithmetik. He does not succeed.
made by the expression defined to the condition for the truth of a statement
Bell’s general defence of Husserl is that he was concerned first to give a
in which it occurs; considerations about the mechanism of an individual sub­
‘theory of our concepts and intuitions of numbers’ (p. 61), and distinguished this
ject s grasp of its meaning are beside the point. His arguments are so compel­
from an account of what the numbers are, which he intended to go on to
ling that modern psychologistic meaning-theorists cannot refuse to allow room
explain in a second volume never published nr composed. This is not wrell
for analysis at the level he w'as urging as the only relevant one. Any adequate
stated: for an accurate account of the concept oi number w'ould tell us all we
meaning-theory' must, after all, acknowledge the place of the concept of truth, had a right or need to ask about w’hat numbers, in general are. Bell means, I
and recognise that, for a great many statements belonging to the common think, that the first (and, as events proved, only) volume of the Philosophic der
language, and above all for those of mathematics, the criteria for their truth Arithmetik was devoted to the task of explaining how we form our concepts of
are held in common. Frege did not deny the possibility, or even the value, of individual cardinal numbers and of number in general; the second volume
psychological investigations. ‘It may indeed be of some use’, he says on p. vi would then have gone on to explain what the numbers are. Certainly, this
o his Introduction, ‘to examine the ideas and changes of ideas that occur description of the content of the first volume agrees very w'ell with what is to
ma ematical thinking; but he adds, ‘psychology should not imagine
1 David Bell, Husserl, London, 1990.
that it can contribute anything to the foundation of arithmetic’. No reassess- 2 Footnote to § 45 of the Prolegomena, which formed part I of the Logische Lntersuchungen.
 20 Frege: Philosophy ofMathematics 2 The Introduction to Grundlagen 21
be found there; but, when the project as a whole is so explained, its absurdity ogism in the Grundlagen is the warning in the Introduction (p. vi) not to ‘take
is manifest. For to explain what the numbers are is just to characterise the a description of the way in which an idea arises for a definition’. Definitions
general concept of number, so that the project would be first to say how we must be certified as genuine by being fruitful, which means that we may appeal
form that concept, and then to say what the concept is. This is evidently to them in the course of proving theorems (p. ix); but ‘a description of how’
impossible, however: there is no way of giving an account of how we form a we arrive at the object or concept in question’ can never serve this purpose
concept in advance of attaining clarity about what that concept is. (p. viii), and so cannot be substituted for genuine conceptual analysis. In
In fact, Bell’s understanding of Husserl’s project receives no support from particular, therefore, ‘a description of the inner processes that precede the
Husserl’s text. In his Preface, Husserl promises to devote Part 1 of his second formation of a judgement of number ... can never replace a genuine determi­
volume, not to some ontological counterpart to the psychological investigation nation of the concept’ (§ 26). Bell might object that Frege does concern himself
of Volume I, but to what he calls ‘quasi-numbers’, i.e. ‘negative, imaginary, with concept-formation in the Grundlagen, above all in the celebrated claim
rational and irrational numbers’, and Part 2 to the question whether it is the that by construing ‘a is parallel to b’ as an identity-statement, ‘we carve up the
natural numbers or one of these other number-domains that is governed by content in a way different from the original one, and thereby attain’ the ‘new
‘general arithmetic in its first and original sense’.3 There simply is no such concept’ of a direction (§ 64). But what differentiated such an account from
distinction in Husserl’s book between the number-concept and the objective one of the type used by Husserl was, as we shall see, that it did not serve in
number as Bell strives to make us believe; Frege was right in his renew to place of a true definition, but as a guide to arriving at one.
say that Husserl obliterates ‘the boundary between the subjective and the In the philosophy of Frege’s middle period (1891-1906), a more rigid
objective’,4 so that no clear differentiation between a number and a number­ doctrine marks the boundary between psychology' and logic. The notion of
concept remains possible. It is clear that Husserl took a pure number {reine sense is correlative to that of understanding, that is, of grasping a sense: what
Anzahl) to be an aggregate of featureless units, obtained from a more determi­ may be attributed to sense is constrained by the principle that sense can be
nate aggregate by mentally abstracting from the particular features of its mem­ grasped. But, in logic, we are concerned only with what the sense is; the
bers:5 the very conception whose incoherence was, as we shall see, demon­ mental act of grasping it, hard as that may be to explain, is a matter for
strated so conclusively by Frege in Grundlagen, §§ 34-44, and the terminus of psychology', and is of no concern to logic. Now understanding either is a
the Husserlian process of forming the concept of a particular number. When species of knowledge or is akin to knowledge: so, although Frege never spoke
Husserl says that ‘the arithmetician does not operate with the number-concepts of it as knowledge, we may express the point untendentiously by saying that
as such, but with the ... objects of these concepts’, it is not to introduce some the concern of logic, or, as we should say, of the theory' of meaning, is solely
objective entities distinct from his psychologically obtained number-concepts, with what a speaker knows about an expression in virtue of knowing the
but to suggest that the mathematician uses ‘5’ as a variable ranging over five­ language, with the content of his knowledge, and not with the manner in which
membered sets.6 Bell’s defence is based upon a distinction Husserl never he know's it, or in what his knowing it consists. This view' of Frege’s, as it
draws. stands, is surely not quite right: but we shall not properly appreciate why he
If Husserl had proposed an account of the process of forming the concept came to think it if wre do not view, as they truly were, the psychologistic
of number as something that could stand on its own, before a subsequent doctrines prevalent in his time, such as those advanced in Husserl’s Philosophic
account was given of the concept itself, the mistake would have been bad der Arithmetik.
enough: in fact, he substitutes his account of the process of concept-formation
/ tor a delineation of the concept. It is above all in making this substitution that
psychologism is objectionable; and it is precisely for this reason that Frege Methodological principles
opposes it so vehementiy. The characteristic expression of his anti-psychol- Frege concludes the Introduction to Grundlagen by enunciating three methodo­
logical principles: the psychological is always to be sharply separated from the
logical, the subjective from the objective; the meanings of words must be asked
after only in the context of sentences, not in isolation; and the distinction
between a concept and an object is alw'ays to be kept in view. The second of
these is the celebrated context principle, to be discussed at greater length at
reprin“d h roL x*1’ "*• the point at which Frege applies it. He wras vividly conscious of its connection
6 Philosophic der Arithmetik, pp. 201-2.
with his repudiation of psychologism. To ask after the meaning of a w’ord in
 22 Frege: Philosophy ofMathematics
the context of a sentence in which it may occur is to explain it in terms of its
contribution to what is required to determine such a sentence as true. To ask
after it in isolation is, as he remarks, at least to court a severe temptation to
explain it in terms of the mental images it evokes or the mental acts that
accompany our contemplation, or subserve our grasp, of it.
The third principle is unconnected, but embodies Frege’s rejection of the CHAPTER 3
procedure of postulation in mathematics. Definition of a general term, such
as ‘porcupine’ or ‘unicorn’, cannot of itself guarantee the existence of an
object to which it applies. That, if it is possible at all, requires independent
demonstration; and this applies as much when the definition rules out there Analyticity
being more than one object to which the term applies as in the general case.
This is perfectly obvious when the general term is an empirical one, and
equally within a mathematical theoiy. When the point is made concerning the
foundations of such a theory, it raises the whole question on what basis we In § 3 of Grundlagen, Frege gives his own characterisations of the two Kantian
recognise the existence of mathematical objects, a problem with which Frege dichotomies, the a priori versus the a posteriori, and the analytic versus the
wrestled, and by which he was in the end defeated. That cannot alter the need synthetic. He claims, in a footnote, that he is not wishing ‘to assign a new
for distinguishing between the specification of a general concept, whether a sense’ to the terms, but ‘only to hit off what earlier writers, and Kant in
mathematical one or not, and the assertion that there is an object falling particular, have intended’. This somewhat disingenuous disclaimer is corrected
under it. The third of Frege’s principles offers the least opportunity for in § 88, where Frege says that Kant was guilty of ‘too narrow a definition of
controverting it. the concept’ of analytic judgements, and that ‘on the basis of his definition,
the division into analytic and synthetic judgements is not exhaustive’, although
he concedes that ‘he seems to have had some inkling of the wider concept’
employed in Grundlagen. Frege wavers, in § 3, between treating the Kantian
terms as applying to ‘judgements’, ‘propositions’ (Sdtze'j and ‘truths’; he is
explicit that none of them applies to a false proposition. The basis of his
classification is the justification for the judgement: not how we in fact know
the proposition to be true, but the best justification of it that could be given.
He regards such a justification as a proof: and he envisages the proof as
deductive in character, with the crucial exception that, in the course of it,
appeal may be made to definitions of the terms involved. When such an
appeal is made, we must also take account of ‘the propositions on which the
admissibility of a definition depends’; the proofs of any such auxiliary propo­
I sitions must be included in the proof of the proposition into whose status we
are enquiring.
Austin’s example has here been followed of rendering Frege’s word ‘Sate’,
as used in this section of Grundlagen, by the ambiguous term ‘proposition’,
whose ambiguity it indeed shares in German. It is plain, however, that, in
allowing explicitly for definitions to be invoked in the course of the deduction,
Frege shows that he is characterising the status of sentences, not of their
contents. A definition states what an expression is to mean, or else what it is
already used to mean: a concept is not open to any stipulation. Concepts may
be analysed, but not defined; it is words and symbols that are the subjects of
definitions, and what is derived by means of them must be a verbal or symbolic
sentence.
23
24 Frege: Philosophy ofMathematics 3 Analyticity 25
Given such a proof, the status of the proposition will depend upon the initial trichotomy of judgements; perhaps his later abstention from the use of the
premisses of the proof. If the justification is complete, the initial premisses terms ‘analytic’, ‘a priori’ and their contraries is due to dissatisfaction with it.
will not themselves be capable of proof. In somewhat imprecise language, As he remarks in the footnote to § 3, ‘from mere individual facts nothing
Frege distinguishes among them between what he calls facts , and explains as follows’: if our knowledge is not to be confined to such individual facts as we
‘unprovable truths devoid of generality, the contents of which are predications observe to hold, it must include some general truths. The main text assumes
about particular objects’, and ‘general laws’ which, he says in a phrase almost that all justification proceeds by deductive reasoning. Any judgement that can
identical with one used by Lotze, ‘themselves are neither capable of proof nor be justified at all can be justified by a deductive derivation: he does not allow
need one’.1 If the initial premisses of any justification of the proposition include for the possibility of any other form of justification. The footnote indeed allows
particular facts, then the proposition is a posteriori; if it can be proved from that empirical induction may establish the truth of a physical law only with
general laws alone, it is a priori. Among a priori propositions, analytic ones probability; it is left unclear whether this would constitute a justification of the
are distinguished by being derivable from general logical law's, together, of law itself, or only of the proposition assigning it a certain probability. In the
course, with the definitions to which appeal may always be made in the unpublished fragment ‘Logik’, perhaps written about when Grundlagen was
course of a justification. If, however, the initial premisses, though consisting published, he indeed admitted the necessity for non-deductive justifications:
exclusively of general laws, necessarily include some ‘which are not of a general
logical nature, but relate to some special domain of knowledge’, the proposition, Now the grounds which justify the recognition of a truth often lie in other truths
already recognised. If truths are to be recognised by us at all, however, this cannot
though a priori, is synthetic. The implicit characterisation of a logical propo­ be the only kind of justification.
sition is thus that it involves only terms of universal application, whose use in
no wray delimits the domain in which the proposition holds good; they are, in He then qualified this concession by adding:
a later terminology, ‘topic-neutral’.
With uncharacteristic carelessness, Frege has framed his definition so as There must be judgements whose justification rests on something different, if
not to cover the initial premisses themselves. The criticism cannot be evaded indeed they need a justification at all.3
by declaring a one-line derivation whose premiss coincides with its conclusion
a limiting case of a proof, since Frege says explicitly that neither the particular Now Frege unwaveringly believed that any deductive proof must have a
facts nor the general law's are provable. An obvious extension of his definition starting-point in the form of initial premisses. A complete justification must
w’ould rate the particular facts as a posteriori, the general logical laws as therefore derive from premisses of which no further justification is possible:
analytic and the general laws belonging to a restricted domain as synthetic a propositions that we know without the need, and without the possibility', of
priori. A more serious failure on Frege’s part to make his own classification proof. If we can claim to know anything more than particular facts, therefore,
exhaustive is discernible if it is a classification of true propositions rather than if we know any general truths, we must know, without the need or possibility
of judgements, for he makes no allowance for there being true propositions of proof, some fundamental general laws. In the footnote, he cites the principle
that cannot be known at all. In the Preface to Grundgesetze he insisted that the underlying empirical induction as an instance of such a general law that is not
truth of a proposition is independent of its being recognised to be true: ‘being logical in nature; we know that he held the axioms of Euclidean geometry to
true is something different from being held to be true, whether by one, by have a similar status. Frege believed all this because he consistently rejected
many, or by all, and can in no way be reduced to it.’2 It follows that the the legitimacy of deriving a consequence from a mere supposition: all inference
meaning of the proposition must be given in terms of what will render it true, must be from true premisses. This excludes the use of reasoning under a
conceived of as independent of how we recognise it as true; it therefore hypothesis subsequently to be discharged by a rule of inference such as reductio
requires special argument if it is nevertheless to be maintained that every true ad absurdum. In ordinary practice, we apply this rule by first stating a hypothesis,
proposition is capable of being known by us to be true. Frege offers no such such as ‘Suppose 2 has a rational square root’. We then reason under this
argument: he therefore has no ground to rule out the possibility that there are hypothesis, drawing consequences dependent on it; when we finally derive a
truths that cannot be known either a priori or a posteriori. contradictory consequence, such as that some integer is both odd and even,
A whole epistemology is implicit in Frege’s refashioning of the Kantian we conclude to the falsity of the hypothesis, our conclusion of course no longer
being governed by it. According to Frege, however, this is not a correct account
J . h**’18”’ 5 ■- H"”"" L»» H-to of oeaer need nor
2 Vol. I, p. xv. 3 Nachgelassene Schriften, p. 3, Posthumous Writings, p. 3.
26 Frege: Philosophy ofMathematics
3 Analyticity 27
of any legitimate inferential procedure. On his view, any step ini our reasoning
nor about the way, perhaps erroneous, in which someone else has come to take
has to be asserted outright: what figured in the foregoing description as the it to be true, but about the ultimate ground on which the justification for taking
initial enunciation of a hypothesis should be considered as the formulation of it to be true depends.
the antecedent of each of a series of conditionals forming every step in the
argument except the final one. The penultimate step will then be of such a
Some doubt is cast by the qualification ‘perhaps erroneous’; but it is natural
form as ‘If 2 has a rational square root, some integer is both odd and even’,
to take Frege as meaning that an a priori proposition may be known a posteriori:
from which we then derive our conclusion ‘2 has no rational square root’.
otherwise the status of the proposition would be determined by any correct
Hilbert and Russell both followed Frege in formalising logic in accordance
justification that could be given for it. This suggests that a priori propositions
with this principle. Such a formalisation, exemplified both in Begriffischrift and
can be further subdivided into those that can be known a posteriori, and those
in Grundgesetze, does not directly address itself to the analysis of deductive
which, if known at all, can only be known a priori. To avoid triviality, we must
inferences, but constitutes a formal theoiy of logical truth: it begins with the
here exclude derivative knowledge - knowledge depending upon that of
axiomatic stipulation of certain logical truths, and derives others by means of
another or on the subject’s memory of having had that knowledge in the past.
a restricted number of rules of inference. For sentential or first-order logic, If I know' the truth of a theorem because I have been assured of it by a
the logical truths are represented by valid formulas, in higher-order logics, by trustworthy mathematician, my knowledge is not a priori: since any truth may
sentences formulated in purely logical terms: in either case, the specification be known by testimony, and hence a posteriori, we may disregard such know­
of what is to count as a valid argument from non-logical premisses to a non- ledge in the present context.
logical conclusion is only supplementary to (though not uniformly derivable Even on this understanding, the existence of propositions of the former kind
from) the central theory, which is a theory' of logical truth. cannot be questioned. If I know that John Trevor was born at Leighton
All this w’as changed by Gerhard Gentzen, wrho did not share Frege’s quite Buzzard, I shall agree with anyone who says that he was bom at Leighton
unjustifiable hostility to rules of inference that discharge hypotheses, and to Buzzard if anyone was; and I shall then also agree with anyone who says that
the reasoning under hypothesis that leads up to an application of such a rule. there is someone who was born at Leighton Buzzard if anyone was, but very'
The result was his formalisation of logic in natural deduction systems, whose likely without noticing that (by the standards of classical logic) this is analytic
direct concern wras with rules of inference and which dispensed with axioms in Frege’s sense: my reason for assenting to the proposition is that I know
altogether. In the light of such a formalisation, logically true sentences are a someone wrho actually was born there. A sentence w’hich instantiates a valid
mere by-product of the procedure necessary for drawing non-logical conse­ formula of first-order, or even of sentential, logic may be recognised as true
quences from non-logical premisses: they arise simply by successively discharg­ in the same way as a similar sentence that is not logically true, by evaluating
ing all hypotheses. It is thus not true that every deductive argument requires it in accordance with its structure after determining the truth-values of subsent­
initial premisses. Framed in terms of a natural deduction formalisation of logic, ences or the applications of constituent predicates; one need not notice, in the
analytic propositions could be defined as those logically derivable, with the process, that the outcome would have been the same whatever the subsentences
help of definitions, from the null set of premisses; such a formulation greatly or the predicates. Conversely, certain truths, such as ‘There are seven days in
re uces die analogy between them and synthetic a priori propositions, as Frege the wreek’ and ‘April comes after March’, are constitutive of the meanings of
the w'ords used to express them, and hence are not only true a priori, but
could only be known a priori.
A finer classification Into w'hich subclass should we put numerical equations? If I use my pocket
calculator to add 56179 and 43286, it appears that I now' have a posteriori
c!ass*^es trye propositions according as they can be known a priori or knowledge of an a priori truth. This case differs, however, from that of an
kn0m^",ll'MOny omitdns "« possibility that they cannot be instance of a valid formula involving empirical predicates. If a sentence of the
L P?“'S'S Iha' SU,US does ”°> d=P“d 0" *e grounds latter kind is recognised as true by determining the application of the predicates
on which they are in fact accepted: & and the truth-values of the subsentences, the recognition of its truth has been
effected in accordance with the w'ay in which its meaning w’as given. It is, as
it were, an accident, not intrinsic to our grasp of its meaning, that it was wired
in my s'n“-w' »•= ”»>
up in such a way' that it would have come out true w'hatever the extensions of
its predicates and the truth-values of its subsentences; that is why we can
 28 Frege: Philosophy ofMathematics 3 Analyticity 29
understand the sentence and recognise its truth without noticing that it is posteriori empirical propositions whose truth no experience has revealed to
analytic. If, on the other hand, we take the meaning of a numerical equation us.
such as ‘13! = 6227020800’ as given by the rule for computing the function, In § 148 of the Wissenschafislehre Bolzano had given a similarly non-epistemic
it is an accident that it should be possible to make an electronic machine definition of ‘analytic’. Bolzano’s classification was of propositions (what he calls
mimic the computation procedure. The interesting principle of classification lSdtze an sich"), not of sentences. This means that the w'ork to be done by
is not whether we can know the truth of the proposition a posteriori, but definitions, at the level of linguistic expression, has, as it were, already taken
whether we know' a priori that, if it can be knowoi at all, it can be known a place; just as Kant spoke of a subject as ‘containing’ a predicate, although the
priori: wre know this of the numerical equation, but not of the instance of the predicate might not be apparent in its verbal expression, so Bolzano thought
valid formula. We might say of propositions of w'hich we know this that they of a complex idea as containing its constituents. If we transpose from the mode
are ‘claimants to aprioricity’. A claimant to aprioricity need not be known to of sentences and their component w’ords to that of propositions and their
be true, or even be true; but if wre know' it a posteriori, w'e also know a component ideas, he in effect used the notion expressed by Quine as ‘essential
posteriori that it is true a priori. occurrence’. An analytic truth in the wider sense w'as for him a true proposition
/ containing at least one idea inessentially: no admissible replacement of that
idea by another would deprive the proposition of truth. An analytic truth in
Epistemic and ontic modalities the narrower sense was one in which all but the logical concepts occur inessen­
tially. Thus in § 197 he expressly observed that the two distinctions, analytic/-
Frege classified truths according to an epistemic principle, that is, by reference synthetic and conceptual/intuitive, cut across one another: there are instances
to how’ w’e can know them. ‘A priori’ and ‘a posteriori’ are naturally taken, as of all four combinations. The proposition, ‘This triangle is a figure’, exemplifies
Kant took them, as epithets which, in the first instance, qualify our knowledge; the class of intuitive analytic truths, since the use of the demonstrative in its
but Frege understood ‘analytic’ and ‘synthetic’ in an equally epistemic sense. linguistic expression indicates that the idea expressed is a mixture of intuition
In this, too, he was essentially in agreement with Kant, since although, in the and concept.
Kritik der reinen Vemunft, Kant defined an analytic judgement in terms of the In classifying propositions according to their intrinsic characteristics rather
relation between the concepts expressed by its subject and predicate, and a than how we can know them, we do so by reference to w’hat renders them
synthetic judgement as one that wras not analytic, his immediate comment was true. Frege’s later insistence that what renders them true is independent of
that only synthetic judgements extend our knowledge. our knowledge of them is matched by the manner in which, for example, he
Bolzano, in his Wissenschafislehre of 1837, had taken Kant to task for defining specifies the meaning of the universal quantifier, namely in terms of w'hat
any of these concepts by reference to knowledge. The details of his classifi­ makes a universally quantified statement true, and not at all in terms of how'
cation of propositions is of less significance than the principle on which it is we can recognise it as such. It is therefore surprising that he did not at least
based. In § 133, he distinguished ‘conceptual’ from what he variously called supplement his epistemic classification by an ontic one. The explanation is
‘perceptual’, ‘empirical’ or ‘intuitive’ propositions. He accepted Kant’s distinc­ surely that, with no ground for the assumption, Frege presumed that all true
tion, among ideas, between concepts and intuitions, modifying it only by the statements of arithmetic w’ere provable by us. The distinction between episte­
admission of mixed ideas, compounded of both. For him, a conceptual propo­ mic and ontic necessity is precisely that between proof-theoretic and model-
sition was one involving only pure concepts, an intuitive proposition one theoretic consequence. A logical formula may be called ‘provable’ if it is a
involving some intuition. He went on to remark that this distinction happens theorem of some axiomatic formalisation, or derivable from the null set of
‘nearly to coincide’ with that drawn by Kant between a priori and a posteriori hypotheses in a natural deduction system. A statement is analytic in Frege’s
judgements, ‘since the truth of most conceptual propositions can be decided sense if it is the definitional equivalent of an instance of a provable formula.
by pure thought, while propositions that contain an intuition can be judged If we transpose back from the mode of propositions to that of sentences, a
only by experience’. He nevertheless objected to Kant’s having replaced the statement is analytic in Bolzano’s sense if it is the definitional equivalent of
former distinction by the latter: ‘the former rests, not on the relation of an instance of a (model-theoretically) valid formula. Had Frege recognised
propositions to our cognitive faculty, but on their intrinsic characteristics.’ In that there might be arithmetical truths we are incapable of proving, he would
support of this claim, he observed that, by stating that all mathematical propo­ surely have accorded them such a status. Similarly, a statement will, for Frege,
sitions are judgements a priori, Kant had thereby included propositions that be synthetic a priori if it is the definitional equivalent of one deductively
we do not at present know, and that he would similarly have included as a derivable from the fundamental non-logical law’s; the corresponding ontic
 30 Frege: Philosophy ofMathematics 3 Analyticity 31
notion would be that of a definitional equivalent of a statement semantically synthetic a priori must, obviously, be correct ones; but, in Grundlagen, Frege
entailed by those laws. simply takes it for granted that we know a correct definition when we see one.
In virtue of the completeness of first-order logic, and the incompleteness It may have been his uncertainty' how to fill this lacuna that deterred him from
of that of second order, the epistemic and ontic notions will coincide for subsequently employing the terms ‘analytic’ and ‘a priori’, or repeating his
statements that do not involve higher-order quantification, but not for those definitions of them; but the difficulty' goes deeper than that, and could not be
that do. Frege of course never formulated the concept of completeness, partly escaped merely by abstaining from the use of the term ‘analytic’. In the
because he did not really think in terms of schematic letters; what look like Introduction to Grundgesetze, having stated it as the aim of Grundlagen to make
schematic letters in his logical notation are, officially, variables bound by tacit it probable that arithmetic is a branch of logic, he went on to claim that
initial quantifiers. At any event, he never attached any particular significance
to the first-order fragment of his logical theory'; for him, second-order quanti­ In this book this will now be vindicated by deriving the simplest laws of cardinal
fication was indispensable for the definitions of ‘natural number’ and of cardi­ numbers (Anzahlen) by logical means alone.
I
nal equivalence, and even for that of class-membership. The present notion
of ontic necessity has little to do with Kripke’s notion of metaphysical necessity', As Frege observed in § 4 of Grundlagen, the derivation necessitated a number
which relates to the behaviour of sentences when governed by modal operators of definitions:
interpreted non-epistemically. An example w'ould be the statement ‘It is now
Starting from these philosophical questions, we come upon the same demand as
4 o’clock G.M.T.’, made at a moment when it was true. Since ‘now’ and ‘4 that which has independently arisen within the domain of mathematics itself: to
o’clock GALT.’ are rigid designators, if it is now 4 o’clock G.M.T., there is prove the basic propositions of arithmetic with the utmost rigour, whenever this
no possible world in which it is not now 4 o’clock G.M.T.; hence it w'ould be can be done ... If we now try to meet this demand, we very' soon come upon
false to say, ‘It might not have been 4 o’clock G.M.T. now’, unless one propositions a proof of which remains impossible so long as we do not succeed
meant the remark in an epistemic sense, and so the original statement was in analysing the concepts that occur in them into simpler ones or in reducing
them to what has greater generality. Number itself is what, above all, has either
metaphysically necessary. A distinction related to these was made by Aquinas to be defined or to be recognised as indefinable. This is the problem to which
in discussing the ontological argument.4 The proposition ‘God exists’, he this book is addressed. On its solution the decision on the nature of arithmetical
maintained, is perse nota secundum se, but not per senota quoad nos, as it would be laws depends.
were the ontological argument valid; we can infer to its truth only from observ­
able, if highly general, features of the world. The epithet 'per se nota quoad nos’ For the proofs of the basic propositions of arithmetic to be convincing, the
plainly means knowable a priori’; whether a proposition is per se nota secundum definitions they appeal to need to be recognised as correct.
se, on the other hand, presumably depends on what makes it true. The notion Frege’s first explicit statement of the condition for a correct definition occurs
cannot be equated with analyticity in Bolzano’s sense. It could be assimilated in his review of Edmund Husserl’s Philosophic der Arithmetik of 1891. Husserl’s
to that of metaphysical necessity, since no one would want to assert that there book contained an extensive discussion of Grundlagen, and he sent a copy of
is a God, but that there might not have been—unless, again, he was speaking it to Frege in the year of its publication, together with offprints of two of his
in an epistemic sense, meaning that, for all he formerly knew, there may not articles of the same year. Frege wrote a friendly reply, expressing the hope
have been. It is not easy to hit on Aquinas’s exact meaning, since he does not that he ‘would soon find the time to reply to your objections’. He did not find
make it explicit; but he deserves credit for drawing a distinction of a kind not the time until 1894, just a decade after the publication of Grundlagen, when
subsequently made, so far as I know, by anyone before Bolzano. he published a devastating review of the book. Husserl had objected to Frege’s
way of defining number in Grundlagen that ‘what this method in fact allow's us
Definition to define are not the contents of the concepts of direction, shape and cardinal
number, but their extensions'.5 In the review, Frege replied that:
The most serious defect in Frege’s characterisations of the concepts of analytic-
defiXnPn°nClty hT>,inAS faiIUre t0 State 1116 conditions under which a Here a divergence is revealed between psychological logicians and mathematicians.
ie comse of thT ?efinitions t0 which he all°™ appeal to be made in For the former it is a matter of the sense of the words and of the ideas which
the course of that proof whose existence shows a proposition to be analytic or they fail to distinguish from the sense; for the latter, by contrast, it concerns the
subject-matter itself, the reference of the words?
4 Summa Theologica, part I, question 2, article 1. ’ Frege s AttemPt>-
 32 Frege: Philosophy ofMathematics 3 Analyticity 33
At this point Frege refers in a footnote to his essay ‘Uber Sinn und Bedeutung’; Admittedly, we often conceive of the matter the other way round, and many
he of course uses the term ‘idea’ to mean a mental image or the like. He teachers define: parallel straight lines are those which have the same direction.
The proposition ‘If two straight lines are both parallel to a third, they are parallel
continues: to each other’ can then very conveniently be proved by appeal to the analogous
proposition about identity. Only the trouble is that this is to stand the true state
The objection that it is not the concept, but its extension, that is defined, actually of affairs on its head. For everything geometrical must surely be originally given
affects all definitions in mathematics. For the mathematician, it is no more correct in intuition. Now I ask whether anyone has an intuition of the direction of a
and no more incorrect to define a conic section as the circumference of the straight line. Of a straight line, indeed; but do we distinguish in intuition the
intersection of a plane and the surface of a right circular cone than as a plane direction of the line from the straight line itself? Hardly. The concept of direction
curve whose equation with respect to rectangular co-ordinates is of degree 2. is first arrived at through a process of intellectual activity that takes its start from
Which of these two definitions he chooses, or whether he chooses another again, the intuition. On the other hand, we do have an idea of parallel straight lines.
is guided solely by grounds of convenience, although these expressions neither
have the same sense nor evoke the same ideas.
It has been argued by Gregory Currie that these observations rest entirely
Frege is here being very’ unfair to Husserl: Husserl had discovered the upon the peculiarities of geometry’, as Frege conceived of them; and certainly
paradox of analysis, w hich was so greatly to exercise G.E. Moore, and which it is written in such a way as to suggest this. Were this so, however, the passage
cannot be dismissed with such nonchalance as Frege manifests. Frege expressly would be entirely beside the point. In § 64, Frege is expressly invoking what
denies that a correct definition need capture the sense of the expression it he takes to be a case analogous to that with which he is directly concerned,
defines: it need only get the reference right. This criterion cannot always be the definition of the notion of (cardinal) number in terms of the relation of
readily applied: in the very case that Husserl was discussing, how is it to be cardinal equivalence; and he continues to discuss the matter in terms of the
determined whether Frege’s definition of‘cardinal number’ secured the correct analogy' until § 68, when he reverts to the true topic, on the assumption that
reference for it? The criterion is in any case far too weak to yield any reasonable the general points established for the analogy' apply also to it. If what he said
notion of analyticity’, defined as in Grundlagen, § 3: almost any proposition in § 64 depended on a feature of the analogy that differentiated it from the
could be shown to be analytic, given suitable choices of definitions for the principal case, namely its geometrical as opposed to arithmetical character, the
terms involved. This makes it likely that, by the 1890s, Frege had lost interest entire discussion would be vitiated. Frege obviously intends his readers to
in the status of individual propositions in favour of that of whole theories; we understand that to define ‘There are just as many Fs as Gs' to mean ‘The
saw’ that, in the opening sentence of the main text of Grundgesetze, he character­ number of Fs is the same as the number of Gs’ would be to stand the true
ised the aim of Grundlagen as that of showing arithmetic, in the singular, to state of affairs on its head in just the same way as to define ‘The line a is
be a branch of logic, rather than showing arithmetical truths, in the plural, to parallel tc the line F to mean ‘The direction of a is the same as the direction
be analytic. of F; if not, the discussion of direction would have no relevance to the problem
The alternative definitions of ‘conic section’, in Frege’s example, are not how number is to be defined. But the reversal of the true state of affairs, in
merely co-extensive, but provably co-extensive. This criterion W'ould allow us the case of number, could have nothing to do with intuition on Frege’s view,
to determine that two suggested definitions were equally correct; but, if no in the light of his claim to have shown that arithmetic in no way depends on
proposition involving the term defined can be proved without appeal to a intuition. Rather, Frege is here appealing to a general principle that nothing
definition of it, it would never allow us to determine any definition as correct should be defined in terms of that to which it is conceptually prior.
absolutely, since we could never prove the defining expression to have the Some twenty years later, a lecture course of Frege’s, ‘Logik in der Mathema-
same reference as that defined. tik’, contained a discussion of definition, and was preserved, in a version of
noHiXS\?C Pr°P0Sed in review of ^sserl was certainly 1914, among his surviving papers. The views here expressed differ both from
Id ? h Frege had, m mmd When Writin® G™dlagen-, for, although he those of Grundlagen and of the review of Husserl. Frege distinguishes between
X f criterion " ** b°Ok’ he » “ted'on a analytic definitions and what he calls ‘constructive’ ones; the latter are stipul- > i. ■ -

secure the correct f 6 S°eS beyon<^ tbe demand that the definition ative definitions, not responsible to anything, but laying down what a new’ w'ord
Z sbn must L dTCH- ThlS iS ** conceptual priority be respected: no or symbol is to mean, or the sense in which an author proposes to use an
expression must be defined in terms of one L„ ’
’ conceptually existing one. The attitude Frege expresses towards these constructive defi­
that is
Frege makes this explicit, in § 64, when he disc prior to it.
nitions coincides with Russell’s. From a logical standpoint, they are mere
direction in terms of ‘parallel’; —Jusses the means of defining abbreviations, since the defining expression and that defined will have the
 34 Frege: Philosophy ofMathematics 3 Analyticity 35
same sense: although their psychological importance may be great, •logically definitions of them, without stopping to ask after the criterion for such defi­
considered, they are really quite inessential’.’ As Eva Picard, has remarked, nitions to be correct. The reason for his use of a special jargon in Grundlagen
/ this is a far cry from the talk in Grundlagen of the fruitfulness ol definition. was quite different. His motive was to exhibit what he had argued to be the
Analytic definitions, on the other hand, are those that attempt to capture correct logical analysis of the familiar expressions. He had stressed that it was
the senses of existing expressions; we hear no more about such a definition’s to concepts that numbers attach, and statements of number relate: he therefore
needing to be faithful only to the reference. Frege maintains, however, that intended the verbal form ‘the number belonging to the concept F’ to bring
‘we shall be able to assert’ that the sense of the defining expression agrees this out more perspicuously than its everyday equivalent. Likewise, numerical
with that of the term it purports to define ‘only when it is immediately evident’.’ equality was a relation between concepts, rather than objects, and the form
He is here relying on his belief in the transparency of sense: anyone who ‘The concept F is equinumerous to the concept G’ presented itself as making
grasps the senses of two expressions must thereby know whether or not they this apparent in a way ‘There are just as many Fs as Gs’ did not. If this was
are the same. ‘How is it possible’, Frege asks, ‘that it should be doubtful Frege’s reason for employing his jargon, he was mistaken: but, then, he was
whether a simple sign has the same sense as a complex expression, when the still in a state of innocence, as yet unaware of the paradoxes with which he
sense of the simple sign is known, and that of the expression can be recognised grappled in ‘Uber Begriff und Gegenstand’.
from its composition?’, and answers, ‘If the sense of the simple sign is really It is therefore astonishing that, even thirty’ years later, Frege could have
clearly grasped, it cannot be doubtful whether it coincides with the sense of come so to depreciate the conceptual analyses that had formed so large a part
the expression.’10 When it is in this way immediately endent that the analysis of Grundlagen as to deny the very possibility’ of conceptual analysis save in rare
captures the sense already possessed by the expression analysed, it is better and unproblematic cases. How, at the time of w riting, he conceived of the
not to call it a ‘definition’, but to present it as an axiom. This will happen in definitions given in Grundlagen, and how' we ought to conceive of them, is best
very few’ cases, however, since very often wre do not apprehend the sense of left to be discussed when they have been renewed in more detail. For the
the existing term clearly, but only in a confused fashion ‘as through a fog’. In present, it is enough to be conscious that their status is a question unresolved
such a case, Frege recommends that we should simply use our proposed by Frege and critical to an evaluation of his work.
analysis as a stipulative (constructive) definition of a newly introduced w’ord
or sign, and always use the latter in place of the existing term.
At first glance, one might suspect that this was the strategy’ he had followed
in Grundlagen. He does not, after all, employ familiar terminology'. Instead of
speaking of ‘the number of Fs’, he says ‘the number belonging to the concept
F’; in place of ‘There are just as many Fs as Gs’, he says ‘The concept F is
equinumerous to the concept G’. But, plainly, in claiming to make it probable
that the truths of arithmetic are analytic, Frege did not intend merely to be
asserting the analytic character of a new theory, devised by himself to mimic
number theory as ordinarily understood: he obviously meant that what everyone
took to be the truths of arithmetic were analytic. In proving that every natural
number has a successor, for example, he had no doubt that he was proving
what any'one else w’ould have understood by the proposition: his definitions
enabled him to give such a proof, but did not confer upon the words a sense
in virtue of which they expressed some quite different proposition. In this, he
was simply following the standard practice of mathematicians, who, in order
to prove a theorem involving terms already in use, may begin by giving rigorous
^Nachgelassene Schriften, p. 226, Posthumous Writings, p. 209.
*’icard1’. ‘*7^ on Definition and Logical Proof, in C. Cellucci and G. Sambin
pp 227-30 ‘af p^228 ° ‘ ^a^osofia scienza contemporanee, vol. 1, Bologna, 1988,
’ .\achgelassene Schriften, p. 227, Posthumous Writings, p. 210.
Nachgelassene Schnften, p. 228, Posthumous Writings, p. 211.
 4 The Value ofAnalytic Propositions 37
should already know all consequences attainable by a sequence of such simple
steps, however long; when the theory was a first-order one, this would mean
all consequences whatever. As Frege remarked in § 88, the conclusions are
contained in the premisses, not as rafters within a house, but as the plant
CHAPTER 4 within the seed.
The solution necessarily lies in drawing an appropriate distinction between
form and content. All conceptual thought involves the apprehension of pattern:
The Value ofAnalytic Propositions a report of current observation singles out particular features from a multi­
farious field of perception, subsuming them under general concepts. Some
patterns force themselves upon us, but others need to be discerned. The
characteristic of a pattern is that it is there to be discerned, but that, to
apprehend that in which it is a pattern, we do not need to discern the pattern;
Analytic judgements extend our knowledge
it is essential to our discerning the pattern that we recognise that that in
Kant underestimated the value of analytic judgements, Frege says in § 88; and which we have discerned it remains unaltered. One can hear a poem without
in § 91 he concludes, in direct opposition to Kant, that ‘propositions that identifying the metre or the rhyme scheme; someone unfamiliar with the
extend our knowledge may have analytic judgements as their content’. The Fibonacci sequence may fail to detect the principle determining the terms.
value of analytic propositions and that of deductive inference are essentially When we become conscious of the metre or of the rule of generation, we
the same; as Frege remarked apropos of arithmetical truths in § 17, on what perceive that pattern in the poem or the sequence, which we recognise as still
was at that stage of the book only the hypothesis that they were derivable from the same poem or sequence as before.
logic: In the present case, we are concerned, not with that imposition of pattern
upon heterogeneous reality that constitutes conceptual thought, but with the
Each would then contain within itself a whole series of inferences condensed for discernment of pattern at a level one higher, namely in the thought itself. We
future use, and its utility would consist in our no longer needing to make the may grasp the content of certain propositions, and recognise their truth; but,
inferences singly, but being able to express the result of the whole series simul­
taneously. even when wre think of them at the same time, wre may well not perceive the
pattern revealed by a proof of which they are the premisses. We cannot,
The point of an analytic proposition, in other words, is to encapsulate an in general, say that a verification of the premisses constitutes a verification
inferential subroutine which, once established, may be repeatedly appealed to of the conclusion. An even number is perfect just in case it is of the form
without itself having to be repeated: it is not the truth of analytic propositions 2"_,(2" — 1), where the odd factor is prime. The processes of verifying that
which is in itself important, but heir service in easing our deductive transitions it has the latter form and that it is perfect are different. The proof consists of
from synthetic truths to other synthetic truths. Frege’s contradiction of Kant’s a method of arranging the two processes simultaneously so that a falsification
dictum thus represents his acknowledgement of the fruitfulness of deductive of either can be made to yield a falsification of the other; the possibility of
inference. such an arrangement depends on the fact that even the verification of such
Independently of whether mathematical truths are taken to be analytic or simple propositions consists in a sequence of steps of which the order is
synthetic, mathematics compels us to recognise the fruitfulness of deductive indifferent. To hit on the proof requires an apprehension of the pattern that
inference;; on whatever basis the axioms of a mathematical theory are accepted, makes such an arrangement possible.
the theorems are established by logical proofs. That deductive reasoning can Similarly with the problem of the bridges at Kbnigsberg. The major premiss
yield a Tvast' range of unexpected consequences is therefore incontrovertible: is the fact that a traveller crossed every bridge; the minor premisses are the
k’„_ is L-
the problem how, t0 eXpjajn ^5 without rendering the validity of such disposition of the bridges, and the fact that he traversed a continuous path;
reasoning problematic. It is tempting to explain the validity of simple inferential and the conclusion is that he crossed at least one bridge at least twice. A
steps byr appeal to the thesis that a knowledge of the premisses carries with it verification of the premisses would not, in general, involve verifying the con­
a know ledge of the conclusion. But, if we have already taken every step in the clusion. An observer might be stationed at every bridge, noting if the traveller
direction of Rome, wre must already be in Rome. If the thesis wrere true, we crossed it, and then going away; his continuous path might have been checked
by someone tailing him without noticing when he crossed a bridge. The
36
 38 Frege: Philosophy ofMathematics 4 The Value ofAnalytic Propositions 39
conclusion might be verified by again stationing an observer at every' bridge; ‘the Earth’ expresses a genuine component of the latter thought, and why the
as soon as one of them observes the traveller crossing his bridge a second connective ‘Either ... or and the two subsentences, all express genuine
time, he reports to base and the observations are abandoned. The proof components of the former; it is of such examples that Frege stated, in his
consists in a manner of arranging any sufficiently detailed observations of the middle period, that the sense of a part of a sentence is a part of the thought
traveller’s path in such a way as to verify simultaneously that he crossed every expressed by the whole. But the process of dissection referred to in the passage
bridge and that he crossed one of them twice. Here it is not a matter merely from ‘Booles rechnende Logik’ is not, in general, aimed at extracting such
of arranging any verifications of premisses and conclusion, but of arranging a components: it is a process of concept-formation, aimed at arriving at something
process that simultaneously verifies all three premisses, and one that simul­ new, which is why he had said, in the previous paragraph, ‘I admit the
taneously verifies the conclusion and the minor premisses; and, again, the formation of concepts as arising first from judgements’.
proof consists in the apprehension of a pattern permitting a comparative The process is described in Begriffsschrift, § 9, as follows:
arrangement of the two processes.
Frege believed, however, that every' proof could be broken down into If we suppose that the circumstance that hydrogen is lighter than carbon dioxide
extremely small steps, as taken in his formalised system; and we know that, as is expressed in our formalised language, we can replace the symbol for hydrogen
far as first-order inferences are concerned, he was demonstrably right. It by the symbol for oxygen or for nitrogen. By this means, the sense is altered in
such a way that ‘oxygen’ or ‘nitrogen’ enters into the relations in which ‘hydrogen’
was therefore necessary' to solve the problem of the fruitfulness of deductive formerly stood. By thinking of an expression as variable in this manner, it is
inference, not at the level of entire proofs, but at that of the simplest single dissected into a constant component, which represents the totality of the relations,
steps. Frege’s solution involved precisely the idea of discerning a pattern within and the symbol which is thought of as replaceable by another, and which signifies
a thought, or, rather, in the terminology’ of the early period, a judgeable the object that stands in those relations. I call the former constituent the function,
content, a pattern it shared with a certain range of other thoughts or contents. the latter the argument.
This was the process that led him to declare in ‘Booles rechnende Logik’
that:1 The same sentence or judgeable content can be dissected in different ways; a
simple example used by Frege is the proposition that Cato killed Cato:
Instead of putting a judgement together out of an individual as subject and a
previously formed concept as predicate, we conversely arrive at the concept by If we think of ‘Cato’ as replaceable at the first occurrence, the function is ‘to kill
dissecting the judgeable content. Cato’; if we think of ‘Cato’ as replaceable at the second occurrence, the function
is ‘to be killed by Cato’; finally, if we think of ‘Cato’ as replaceable at both
Why should he say this? If the judgeable content is complex, w’hy should we occurrences, the function is ‘to kill oneself’.
prefer the metaphor of dissecting it so as to extract the constituents to that of
putting it together out of those constituents? It is not, as some have thought, The process is succinctly explained, in essentially the same way, in ‘Booles
that Frege had some strange idea
• ■ of- our apprehending the judgeable content, rechnende Logik’; Frege there uses the variable ‘x’ to indicate the effect of
in the first instance, as a simple unit devoid of complexity: he scotches that treating a given term as replaceable by others. If, in the equation
interpretation in the very next sentence after that quoted above: 24 = 16

Admittedly, in order to be able to be so dissected, the expression of the judgeable we treat the ‘2’ as replaceable, we obtain the concept ‘4th root of 16’; if we
content must already be composite. treat the ‘4’ as replaceable we obtain the concept ‘logarithm of 16 to the base
2’. The talk of imagining a term as replaceable by others shows that the con­
The reason is, rather, that the metaphor of ‘putting together’ is appropriate stant part—what in Begriffsschrift, but hardly at all in later writings, Frege
to that complexity which we must apprehend in order to grasp the content at called ‘the function’ - constitutes a pattern common to all the sentences
^nIMoimP°Slb e-t0.SraSP 1116 thought exPressed by ‘Either Venus is larger obtained by making such a replacement.
rhn ht ■ • arS 1S -u^er ^an MercuiT’ save as a disjunction of two simpler In Grundlagen, the process of dissection is directly referred to only in § 70,
^oughts; it is impossible to grasp that expressed by ‘The Earth rotates’ save w'here Frege uses it to explain his general notion of what he here calls a
p dicatmg something of the object denoted by ‘the Earth’. That is why ‘relation-concept’. He does not now use the psychological language of imagin­
ing a term as replaceable by others, but speaks of ‘subtracting it: what remains,
' Nachgdassene Schriften, p. 18, Posthumous Writings, p. 17. when we subtract one term, is an expression for a concept, and, when we
 40 Frege: Philosophy ofMathematics 4 The V'ilue ofAnalytic Propositions 41
subtract two, one for a relation-concept; but, since either ‘demands a com­
of ‘and’. The point is stated clearly in Begriffischrift, § 9, where, immediately
pletion to make a judgeable content’, the two metaphors have exactly the same
after the sentence quoted above explaining his use of the terms ‘function’ and
‘argument’, Frege says:
This was not in fact the only process of concept-formation Frege was
/
prepared to admit. The process by which we attain to such concepts as shape, This distinction has nothing to do with the conceptual content, but is only a
direction and number itself, exhaustively described in §§ 63-9 of Grundlagen, matter of how we regard it.
is quite different; and in § 34 he seems prepared to allow that some concepts
can be attained by the process of abstraction. The concepts Frege believed to Dissection is therefore justly described as a process of concept-formation: it
be attainable only through judgements or complete propositions were those reveals something new, one pattern among many discernible in the proposition
expressed, in the first instance, by complex predicates, to any of which we and shared by it with others, but not, in general, intrinsic to a grasp of its
may, in interesting cases, equate some newly introduced simple predicate by content.
definition. The reason why such a concept has to be regarded as attained, not It is when they essentially involve the process of dissection that, in Grundla-
by being built up out of its constituents, but by the dissection of a proposition, gen, Frege regards definitions as fruitful. As he says in § 88:
is that, on Frege’s new, the sense of a complex predicate is not directly
derivable from its components. From the proposition ‘Either Jupiter is larger [Kant] seems to think of a concept as determined by co-ordinate characteristics;
than Neptune and Neptune is larger than Mars, or Mars is larger than Neptune but this is one of the least fruitful methods of concept-formation. Anyone who
surveys the definitions given above will scarcely find one of this kind. The same
and Neptune is larger than Jupiter’, we can extract the predicate ‘Either Jupiter holds of the truly fruitful definitions of mathematics, for example that of the
is larger than x and x is larger than Mars, or Mars is larger than x and x is continuity of a function. In these we do not have a sequence of co-ordinate
larger than Jupiter’, thus attaining the concept ‘intermediate in size between characteristics, but a more intimate - I should like to say, more organic - combi­
Jupiter and Mars’. But neither the connective ‘or’ nor the connective ‘and’, if nation of specifications. The distinction can be made intuitive by means of a
regarded as primitive, is to be explained as operating on two predicates to geometrical picture. If one represents the concepts (or their extensions) by regions
of a plane, what corresponds to a concept defined by means of co-ordinate
form a new complex predicate: each is explained only for the case in which it characteristics is the region common to all the regions representing those charac­
serves as the principal operator in a complete proposition. Hence the complex teristics; it is enclosed by segments of their peripheries. In giving such a definition,
predicate cannot be understood save as extractable from such a proposition as therefore, it is a matter - to speak pictorially - of using the already given lines in
that cited above: its sense may be seen as being given as a function carrying a new way to delimit a region. Nothing essentially new emerges from this.
the sense of the name ‘Neptune’ on to the thought expressed by ‘Either Jupiter
is larger than Neptune and Neptune is larger than Mars, or Mars is larger As Frege remarks in a footnote, the case is similar when the characteristics
than Neptune and Neptune is larger than Jupiter’, the sense of the name are connected by disjunction. He continues:
X enus on to the thought expressed by ‘Either Jupiter is larger than Venus
and \ enus is larger than Mars, or Mars is larger than Venus and Venus is The more fruitful determinations of concepts draw boundary lines which were
not previously given at all. What we shall be able to infer from them cannot be
larger than Jupiter’, and so on. We can regard it as such a function only predicted in advance; we are not in this case simply taking out of the chest what
because we already understand the complete propositions; it is in grasping their we had put into it.
contents that we direcdy advert to the meanings of the connectives ‘or’ and
‘and’. And from this he draws the conclusion that ‘the consequences derived advance
The process of dissection thus does not respect that structure in virtue of our knowledge’.
which we grasp the content of the proposition in accordance with its compo­ This, then, is Frege’s explanation of the fruitfulness, not merely of definition,
sition, what it yields is, in general, a feature which the proposition shares with but of deductive reasoning, and, with it, of analytic propositions. But why does
others but of which we did not have to be aware in order to grasp its content. he link the two? The reason is that dissection is necessary’ in order to recognise
Io understand the proposition ‘Jupiter is larger than Neptune and Neptune the validity of inferences. If we define ‘x is intermediate in size between y and
is arger than Mars, it is not necessary so much as to notice that the name £ to mean ‘Either y is larger than x and x is larger than z, or z is larger than
eptune occurs in both subsentences, let alone to conceive of the range of x and x is larger than j’, we need, if we are to draw the conclusion There is
propositions obtainable by replacing it in both occurrences by some other a body intermediate in size between Jupiter and Mars’, to be able to recognise
name, a at is necessary is to understand both subsentences and the meaning the complex three-place predicate as extractable from the proposition Either
 42 Frege: Philosophy ofMathematics 4 The V'due ofAnalytic Propositions 43
Jupiter is larger than Neptune and Neptune is larger than Mars, or Mars is Ranges of application
i
larger than Neptune and Neptune is larger than Jupiter: we have to discern
that pattern in it. This does not apply only when a definition is involved: since It would be a mistake, though a natural one, to suppose that Frege’s only
ground for maintaining the truths of arithmetic to be analytic was his detailed
it is, in general, to a complex predicate that a quantifier is attached in order
reduction of its fundamental laws to logical truths: for he has, besides, some
to form a quantified proposition, the operation of dissection must be conceived
general arguments, based on the universal applicability he ascribes to arithme­
as a necessary preliminary to the formation of a quantified proposition in the
/ tic. Grundlagen in fact advances two distinguishable theses about arithmetical
standard case. In order to frame the proposition ‘For some .v, Jupiter is larger
truths: that they are analytic, and that they are expressible in purely logical
than x and .v is larger than Mars’, the complex predicate ‘Jupiter is larger than
terms. On his own principles, neither implies the other. The presence of non-
,r and x is larger than Mars’ has first to be extracted from such a proposition
logical expressions in a formulation of the axioms of geometry does not, of
as ‘Jupiter is larger than Venus and Venus is larger than Mars’. This predicate
itself, prove those axioms to be synthetic; for there might be some system of
is not a component of the proposition from which it was extracted by dissection,
definitions connecting the geometrical terms the application of which would
in that we do not have to recognise its presence in order to grasp the content
render them derivable from logical first principles. Conversely, a synthetic
of the proposition; but it is a component of the quantified proposition. As
proposition might be expressible by means of logical notions alone. An example
Frege puts it, clumsily but clearly, in § 9 of Begriffsschrift:
would be Russell’s Axiom of Infinity, which says that there are infinitely many
When the argument is indeterminate, as in the judgement, ‘You can take an individuals: since, for Russell, neither numbers nor classes - what Frege
arbitrary positive integer as argument for “to be representable as the sum of four regarded as logical objects - are individuals, the analytic character of this
squares”, and the proposition will always remain correct’, the distinction between axiom can hardly be sustained. For a proposition to be analytic in Frege’s
function and argument becomes of significance as regards the content. sense, it must follow from the fundamental laws of logic, which neither need
nor admit of proof. These laws are ones we recognise, and must recognise if
Deductive reasoning is thus in no way a mechanical process, though it may we are to be able to reason: but Russell’s Axiom of Infinity is neither included
be set out so as to be checkable mechanically: it has a creative component, among these nor derivable from them. Indeed, it is very probably untrue.
involving the apprehension of patterns within the thoughts expressed, and Propositions differ, on Frege’s view, according to their range of applicability;
relating them to one another, that are not required for or given with a grasp the extent of that range is to be measured along two dimensions, corresponding
of those thoughts themselves. Since it has this creative component, a knowledge to the two features just considered: the modal status of a proposition, as a
of the premisses of an inferential step does not entail a knowledge of the posteriori, synthetic a priori or analytic, and the vocabulary needed for its
conclusion, even when we attend to them simultaneously; and so deductive expression. The second of these two dimensions relates to the region of reality7
reasoning can yield new knowledge. Since the relevant patterns need to be within which the proposition holds good: it may be true of material objects only,
discerned, such reasoning is fruitful; but, since they are there to be discerned, or, more generally, of spatio-temporal objects, or of all objects whatsoever. The
its validity is not called in question. other dimension relates, rather, to the degree of reality: the proposition may
Such was Frege s solution to the problem of the utility of deductive reason­ be true of everything there actually is, or of everything we can imagine, or of
ing. He is one of die veiy few to have faced the problem at all: J.S. Mill everything of which we can intelligibly think at all.
was another, but his solution failed completely. Whether or not the specific Arithmetical propositions, Frege argued, have maximal applicability along
explanation that Frege offered is adequate, it is surely along the right general both dimensions. They apply to all regions of reality: objects of every' type can
lines. All conceptual thought involves the imposition of form upon an amorph­ be counted. More exactly expressed, we may ask, of objects of every7 type, how
ous rea ty. on Frege s account, deductive reasoning requires the further impo­ many there are satisfying some given condition. The point is made in Grnnd-
sition o orm upon our thoughts. It is surely that conception that can alone lagen, § 24, where it is used to refute the empiricist view that number is a
explain how such reasoning can be at the same time fruitful and cogent in physical property; but its full implications are drawn in the lecture ‘Uber
virtue solely of the contents of the thoughts involved. formale Theorien der Arithmetik’ given by Frege in 1885, the year after the
book’s publication, which starts characteristically as follows:

Under the name ‘formal theory’ I wish here to consider two conceptions, of which
I agree with the first, but seek to controvert the second. The first says that all
arithmetical propositions can be derived purely logically from definitions alone,
 44 Frege: Philosophy ofMathematics
4 The Vline ofAnalytic Propositions 45
and consequently must be so derived.... Of all die grounds that tell in favour
of this view, I wish here to cite only one, that which rests upon the all-embracing if we find it impossible even to conceive of the contrary as an intelligible
applicability' of arithmetical theorems. Virtually everything that can be an object possibility, we have probably come to know it by the use of our logical faculty.
of thought may in fact be counted: the ideal as well as the real, concepts as well The argument was most cogently set out by Frege in one of Grundlagen's
as things, the temporal as well as the spatial, events as well as bodies, methods
as well as’theorems; even the numbers themselves can in turn be counted. Nothing purple passages, § 14, which is worth quoting in full; it must be borne in
is really demanded save a certain sharpness of circumscription, a certain logical mind, in reading it, that the phrase ‘the axioms of geometry’ meant, for Frege,
completeness. From that fact can be gathered this much, that the fundamental ‘the axioms of Euclidean geometry’.
principles on which arithmetic is constructed cannot relate to a narrower domain
whose peculiarities they express as the axioms of geometry express those of
what is spatial. Rather, those fundamental principles must extend to everything A comparison of truths in respect of the domains which they govern also tells
thinkable; and a proposition that is in this way of the greatest generality' is against the empirical and synthetic nature of arithmetical laws.'
justifiably assigned to logic. Empirical propositions hold of physical or psychological actuality, while geo­
metrical truths govern the domain of the spatially intuitable, whether actual or
the product of our imagination. The maddest fantasies of delirium, the most
The argument does not show that arithmetical terms and concepts can be daring inventions of legend or of the poets, which have animals speaking and the
reduced to logical ones. Rather, it shows that they are already logical in character. stars standing still, which make men from stones and trees from men, and teach
The only differentiation of logical notions from others ever considered by how one can pull oneself out of a swamp by one’s own forelock, are yet subject
to the axioms of geometry, as long as they remain intuitable. Only conceptual
Frege rested on their being unrestricted in the subject-matter to which they thought can in a certain fashion shake free of those axioms, when it assumes a
could be applied, rather than being confined to any particular domain of space of four dimensions, say, or of positive curvature. Such considerations are
knowledge. Once it is recognised that there is no segment of reality composed not in the least useless; but they completely abandon the base of intuition. If we
of objects that cannot be numbered, it is thereby recognised that the notion do call intuition to our aid in this connection, it is still the intuition of Euclidean
of number is a logical one. By itself, this as yet says nothing about the grounds space, of the only space of whose structure we have any intuition. It is then taken,
not for what it is in itself, but as symbolic for something else; for example, we
on which we accept the laws of arithmetic as true; it tells us only that they are call something a straight line or a plane which we perceive as curved. For
expressible in purely logical terms, or, rather, that they are already stated in conceptual thought we can always assume the opposite of this or that geometrical
purely7 logical terms. The definitions given in Grundlagen of arithmetical notions axiom, without involving ourselves in any self-contradictions when we draw
in terms of simpler ones are required more in order to make manifest the deductive consequences from assumptions conflicting with intuition such as these.
grounds of those laws than to establish that they are logical in nature. This possibility shows that the axioms of geometry are independent of one another
and of the fundamental laws of logic, and are therefore synthetic. Can one say
The laws of arithmetic have maximal generality' in the other dimension also: the same of the fundamental principles of the science of number? Does not
they apply to all that can be grasped by conceptual thought. The argument in everything collapse in confusion when we try denying one of them? Would thought
this case concerns the ground of our knowledge. Just as he was later to do itself then be possible? Does not the ground of arithmetic lie deeper than that of
in the essay ‘Erkenntnisquellen der Mathematik und der mathematischen all empirical knowledge, deeper even than that of geometry? The truths of arithme­
Natunvissenschaften’ which he wrote for publication in the last year of his tic govern the domain of what is countable. This is the most comprehensive of
all; for it is not only what is actual, not only what is intuitable, that belongs to it,
1 e’ re^e operated with a threefold classification of grounds of knowledge: but everything thinkable. Should not the laws of number then stand in the most
observation; spatial and temporal intuition; and our logical faculty. Observation intimate connection with those of thought?
can te us only how things actually are. Something w'e observe to be so may
e true of all that is imaginable, or even of all that is conceivable; but, if we
That the axioms of geometry can be denied without contradiction does not
now it by observing it to be so, we have no ground to suppose that it holds
good of more than what there actually is. A priori spatial or temporal intuition prove that they are synthetic: it is what is meant by saying that they are synthetic.
But, as Frege frequently pointed out in other connections (and as he was to
thpmUS °" m^S mU-St be we are either to apprehend or to imagine discover to his bitter cost), the fact that we have not come upon a contradiction
. .as m space or time; it cannot tell us of how they must be even if w'e are does not prove that none is lurking. Until a consistency proof was available,
unable to apprehend or iimagine them. Only of w’hat we know in virtue of our
no more could be said than that our not having encountered a contradiction
unaided logical faculty dilo we have any ground for supposing it to hold good
suggested that there was none, and hence that the axioms of Euclidean
of everything thinkable.
geometry were synthetic. Likewise, our inability' to describe coherently a state
co™erseb' that, if we find it impossible to imagine the contrary’ of affairs in which any of the laws of arithmetic failed does not demonstrate
of some general law,’ we have probably derived it from a priori intuition; and,
that they are analytic: a proof that they stand or fall with the laws of logic,
46 Frege: Philosophy ofMathematics
such as Frege attempted to give, is needed for that. The argument from th
applicability of arithmetic to everything that can be grasped by conceptual
thought was no more than suasive. It remains that Frege had to hand quite
powerful suasive argument in favour of the thesis he wished to establish: f0
it at least appears that we can make no intelligible sense, of the kind we can
make of a denial of the parallel postulate, of the supposition that the laws of CHAPTER 5
arithmetic might not hold.

Frege and Dedekind

Dedekind’s Was sind und was sollen die Zahlen? appeared four years after
Frege’s Grundlagen, but was certainly composed independently of it; in his
Preface, Dedekind states that a first draft of his book was completed by 1878
and privately circulated during the ensuing decade. In the Preface to the
second edition, of 1893, he paid a belated tribute to Frege’s book; it was ironic
that, in the Preface to Grundgesetze, published only a month earlier, Frege, call­
ing Dedekind’s book ‘the most profound work on the foundations of arithmetic
that has lately come to my notice’, had complained that, among others, its au­
thor appeared to be unacquainted with his own work.1 It hardly detracts from
the originality of Dedekind’s book to observe that it had been in part antici­
pated by Bolzano’s Paradoxien des Unendlichen, as he acknowledges; Frege, on
the other hand, appears to have known none of Bolzano’s writings.
Of the two, Frege’s book was by far the more philosophically pregnant and
perspicacious; but there is a clear sense in which Dedekind’s revealed much
more about the natural numbers. Dedekind was the first to state and justify
the general principles governing the definition of a function by recursion,
which he formulated for one whose values need not be natural numbers.2 He
used recursion to define addition, multiplication and exponentiation,3 and
proved the fundamental algebraic laws holding for them. A reader of Grund­
lagen, on the other hand, who has kept in mind the sustained discussion of
numerical equations in §§ 5 to 17, may be surprised to discover the book
coming to an end before addition has even been defined. Grundlagen purports
to make it probable that the truths of arithmetic are analytic; yet those whose
proofs are given or sketched do not include w'hat, in § 2, Frege had called
‘the simplest propositions holding of the positive integers, which form the
foundation of the whole of arithmetic’. The addition of cardinal numbers is,
admittedly, touched on at the end of Part II of Grundgesetze-,4 but even there

1 Grundgesetze, vol. I, p. viii and p. x, fn. 1.

2 Was sind und was sollen die Zahlen?, § 9.
3 Ibid., §§ 11-13.
4 Vol. II, §§ 33-44.
47
 48 Frege: Philosophy ofMathematics 5 Frege and Dedekind 49
it is not systematically investigated, and multiplication is never treated of at axiomatising number theory, although in fact what are known as the Peano
all. Frege acknowledged that Dedekind had earnedI his denvatton of the laws axioms were first enunciated by him in private correspondence.6 Instead, he
of arithmetic a great deal further than he himself had done but explained that gives a direct characterisation of structures that serve as models for the Peano
this was possible for him because he was not interested, like Frege, in giving axioms, calling them simply infinite systems’, which are what Russell later
formal proofs that exclude the possibility ol overs.ght and render us fully called ‘progressions’; the four conditions in the definition of a simply infinite
conscious of everything involved in them. system correspond closely to the Peano axioms.7 He feels obliged then to prove
Dedekind’s approach to the question posed in his title differs utterly from that the class of simply infinite sy stems is not empty. This he does by a piece
X Frege’s Dedekind tackled it more specifically in the spirit of a mathematician, of non-mathematical reasoning, his example is the sy'stem whose initial element
Frege more in that of a philosopher; Dedekind’s treatment was that of a pure is my self (mein eigenes Ich) and which is generated by the operation that carries
mathematician, whereas Frege was concerned with applications. Dedekind’s an object x into the thought that x can be an object of my thinking.8
/ central concern was to characterise the abstract structure of the system of He now comes, by a means not at all to Frege’s taste, to define the natural
natural numbers; what those numbers are used for was for him a secondary’ numbers. Dedekind’s philosophy of mathematics was that mathematical objects
matter. In this respect Frege, pioneer as he was, was old-fashioned. From are ‘free creations of the human mind’, as he says in the Preface. He neither
§ 18 to § 83 of Grundlagen, he occupies himself exclusively with the question, amplified nor defended this belief; but he adhered tenaciously to it. The idea,
‘What is number?’, and its ancillary’, ‘What are the individual numbers such widely shared by his contemporaries, was that abstract objects are actually-
as 0 and 1?’. Up to § 44, he reviews and criticises the answers of Mill, Kant created by operations of our minds. This would seem to lead to a solipsistic
/
and many other philosophers and mathematicians. All of these take for granted conception of mathematics; but it is implicit in this conception that each
I that, to say what number is, we must simultaneously’ explain what numerical subject is entitled to feel assured that what he creates by means of his own
equations and the like are about, and analyse the use of number-words in mental operations will coincide, at least in its properties, with what others have
empirical contexts to answer questions beginning ‘How many ... ?’. Frege created by means of analogous operations. For Frege, such an assurance would
does not challenge this assumption: he shares it. Kant took it for granted that be without foundation: for him, the contents of our minds are wholly subjective;
the symbol ‘5’ in the equation ‘5 + 7 = 12’ has an immediate connection since there is no means of comparing them, I cannot know- whether or not my
with the w’ord ‘five’ as it occurs in ‘I have five fingers on my left hand’; and idea is the same as yours. Even if this could be known, there could be no
Frege took it for granted, too. For both of them, arithmetical propositions are ground for declaring one person right and the other wrong, if their ideas
about numbers in the same sense of the w’ord ‘number’ as that in which w'e proved to be different: as Frege says in the Preface to Grundgesetze?
speak of the number of Jupiter’s moons or of inhabitants of Berlin.
Even the German language helped Frege to make this assumption appear It is impossible to ascribe to each person his own number one; for it would then
inescapable. In the footnote to § 4, he explained that he would be almost have first to be investigated how far the properties of these ones coincided. And
if one person said, ‘Once one is one’, and another, ‘Once one is two’, we could
exclusively concerned with ‘the positive integers, w’hich answer the question,
only register the difference and say: your one has that property', mine has this.
“How’ many?” ’, though, on his own principles, he ought to have said ‘the
non-negative integers’. In speaking of them, he usually employs the w’ord That is why thoughts, or judgeable contents, w’hich are communicable and
Anzahl’ rather than '’Zahl'. As his English translator, Austin, notes, ‘Anzahr can be judged by anybody true or false absolutely, rather than true for one
has the sense of ‘cardinal number’, but not its technical ring, being a quite person and false for another, are not to be viewed as contents of the mind: as
everyday w’ord; Austin is therefore driven to distinguishing lAnzahl’ from lZahl' he wrote in the ‘Logik’ of the 1880s:10
by writing ‘Number’ with a capital letter. Frege needed some verbal means of
distinguishing the natural numbers from rationals, real numbers, etc.; his A judgeable content ... is ... not the result of an inner process or the product
choice of the word Anzahl' for this purpose was powerful subliminal propa­ of some human being’s mental operation, but something objective, which means
ganda for the view that their essential characteristic is their use as finite
cardinals. . ^ao Wang, ‘The Axiomatisation of Arithmetic’, Journal of Symbolic Logic, vol. 22, 1957, pp.
Dedekind, by contrast, relegated that use of of them to aa w
them to ’holly subordinate
wholly subordinate H5-57.
status. In his book, he did not characterise the natural number system by »S,n<^ und was so^en die Zahlen?, § 6, definition 71.
S 3, theorem 66.
^P. xviii.
5 Grumlgeselze, vol. I, pp. vii-viii. Nachgelassene Schriften, p. 7, Posthumous Writings, p. 7.
50 Frege: Philosophy ofMathematics 5 Frege and Dedekind 51
something that is exactly the same for all rational beings, for all capable of grasping them in §§ 7—13: only in the final section, § 14, does he give an account of
it, just as the Sun, say, is something objective. the use of the natural numbers to give the cardinality of finite systems, by
using the same notion of one-one correlation employed by Frege in Grundlagen
One of the mental operations most frequently credited with creative powers and by Cantor in papers from 1874 onwards. In complete contrast to Frege’s
was that of abstracting from particular features of some object or system of method of defining the natural numbers, this application of them is not central
objects, that is, ceasing to take any account of them. It was virtually an to Dedekind’s way of characterising them; it is external, an appendage which
orthodoxy, subscribed to by many philosophers and mathematicians, including could have been omitted without damaging the theory as a whole. This diver­
Husserl and Cantor, that the mind could, by this means, create an object or gence is reflected in the way each defines the sum of two natural numbers.
system of objects lacking the features abstracted from, but not possessing any Dedekind does so by means of the recursion equations for addition; Frege, in
others in their place. It was to this operation that Dedekind appealed in order effect, as the number of members of the union of two disjoint classes. Dedekind
to explain what the natural numbers are. His procedure differed from die indeed proves such a union to have m 4- n members if the two classes had m
usual one. Husserl, in company with many others, supposed that each indi­ and m members respectively (§ 14, theorem 168); but, for him, it required
vidual cardinal number was created by a special act of abstraction: starting proof, rather than being immediate from the definition, and was a mere
with any arbitrary’ set having that number of members, we abstract from all addendum to his general treatment of addition in § 11.
the properties possessed by the individual members of the set, thus transform­ Frege and Dedekind were at odds over two interconnected questions:
ing them into featureless units; the set comprising these units was then the whether or not the use of natural numbers to give the cardinality’ of finite
relevant cardinal number. Cantor’s variation on this account was a trifle more totalities is one of their distinguishing characteristics, which ought therefore
complex: we start with an ordered set, and abstract from all die features of to figure in their definition; and whether it is possible, not merely to character­
the individual members, but not from their ordering, and thus obtain their ise the abstract structure of the system of natural numbers, but to identify the
order-type; next, we abstract from the ordering relation, and obtain the cardinal natural numbers solely in terms of that structure. Unlike Frege’s, Dedekind’s
number as an unordered set of featureless units, as before. Frege devoted a natural numbers have no properties other than their positions in the ordering
lengthy section of Grundlagen, § § 29-44, to a detailed and conclusive critique determined by their generating operation, and those derivable from them; the
of this misbegotten theory’; it was a bitter disappointment to him that it had question is whether such a conception is coherent.
not the slightest effect. Cantor, who might have been supposed to have read In his great book of 1903, Bertrand Russell criticised Dedekind’s theory as
Grundlagen, since he reviewed it, persisted undeterred with his abstractionist follows:13
account;11 Husserl, in his book of 1891, again subscribed to it, despite his
lengthy discussion of Grundlagen. It is impossible that the ordinals should be, as Dedekind suggests, nothing but
Dedekind, on the other hand, applies the operation of abstraction to an the terms of such relations as constitute a progression. If they are to be anything
arbitrary’ simply infinite system to obtain from it the system of natural num­ at all, they must be intrinsically something; they must differ from other entities
bers:12 as points from instants, or colours from sounds.... Dedekind does not show us
what it is that all progressions have in common, nor give any reason for supposing
it to be the ordinal numbers.... What Dedekind presents to us is not the
If, in considering a simply infinite system .V, ordered by a mapping 0, we entirely numbers, but any progression: what he says is true of all progressions alike, and his
disregard the particular nature of its elements, retaining only’ their discriminability demonstrations nowhere ... involve any property distinguishing numbers from
front each other, and having regard only to the relations to one another imposed other progressions. No evidence is brought forward to show that numbers are
• e mapping 0 which orders them, then these elements are called natural prior to other progressions. We are told, indeed, that they are what all progressions
numbers or ordinal numbers or simply numbers.
have in common; but no reason is given for thinking that progressions have
anything in common beyond the properties assigned in the definition, which do
The mapping <p is of course the operation that generates the system, corre­ not themselves constitute a new progression.
sponding to the successor function.
Having thus defined the natural numbers, Dedekind develops the theory’ of Russell is here obstinately refusing to recognise the role assigned by Dede­
kind to the process of abstraction. He thinks that, if we are to refer to the
natural numbers, or ‘the’ finite ordinals, we must thereby refer to quite specific
“h 73'"“ °f * PP' l33“*2’ b“
The Principles of'Mathematics, p. 249.
52 Frege: Philosophy ofMathematics 5 Frege and Dedekind 53
objects; but Dedekind would not deny this. He thinks, further, that if these a paradigm of any object which plays it, but by representing the relation that any
numbers are to be specific objects, they must possess properties other than third member of a progression bears to the rest of the progression.
the purely structural ones they have in virtue of their positions in the sequence;
but that is just what Dedekind would deny. He believed that the magical But whether 3 is the third or the fourth term in the sequence of natural
operation of abstraction can provide us with specific objects having only struc­ numbers depends whether you start with 1 or with 0. Frege started with 0,
tural properties: Russell did not understand that belief because, very rightly, because 0 is needed as a finite cardinal; Dedekind started with 1, for no
he had no faith in abstraction thus understood.14 especial reason; Husserl, eccentrically, started with 2, on the ground that
Mathematicians frequentiy speak as if they did believe in such an operation. neither 0 nor 1 is a number. The number 0 is not differentiated from the
One may speak, for example, of ‘the’ five-element non-modular lattice. There number 1 by its position in a progression, otherwise there would be no
are, of course, many non-modular lattices with five elements, all isomorphic difference between starting with 0 and starting with 1. That is enough to show’
to one another: if you ask him which of these he means, he will reply, ‘I was that w’e do not regard the natural numbers as identifiable solely by their
speaking of the abstract five-element non-modular lattice’. But, even if he positions within the structure comprising them.
retains a lingering belief in the operation of abstraction, his way of speaking It might be retorted that this objection depends upon a mere tactical mistake
is harmless: he is merely saving what holds good of any five-element non- on Benacerraf s part: if he had considered the structure <N, 0, 5, +, .> or
modular lattice. That is how neo-Dedekindians such as Paul Benacerraf, who <N, 1, 5, +, .> instead of <N, 0, 5> or <N, 1, 5>, the problem would
have understandably jettisoned the doctrine of abstraction, would have us not have arisen. It is not a mere question of tactics, however: if he had done
suppose it to be with the natural numbers. that, he would have been false to his own principles. If we are concerned, as
The system of natural numbers differs from the lattice in that, for many of Benacerraf is, with what it is that mathematicians are talking about, we have
their applications, for instance in giving a definition by induction upon them, to think, not merely of mathematical structures, but of how’ they are given to
it is essential to know that such a system exists. Dedekind recognised that us, that is, how they are characterised. Obviously, Benacerraf must have
necessity. If he had proved the existence of a simply infinite system by purely in mind a second-order characterisation, which alone yields a categorical
mathematical means, for instance from the theory of classes, could he not have specification of the structure of the natural numbers. Under a second-order
identified the system so proved to exist with that of the natural numbers? Or characterisation, however, there is no call to treat addition and multiplicaticn
was it essential that the system of natural numbers be what all simply infinite as primitive, since they are definable; the structure is completely determined
systems have in common?
by the Peano axioms, and needs no further determination in terms of other
Some mathematical objects’ really have only a pure structural identification, operations. Benacerrafs thesis is that structure is all that matters, since w’e
and thus, as Benacerraf argues concerning the natural numbers, are not genu­ can specify a mathematical object only in terms of its position in the structure
ine objects at all: there is no more such an object as the zero of ‘the’ eight­ to which it belongs. The thesis is false, and the example Benacerraf chose to
element Boolean algebra than there is such a point as the centre of‘the’ circle
illustrate it is the very' one that most clearly illustrates its falsity. The identity’
with unit radius. As far as the natural numbers are concerned, however, Frege
of a mathematical object may sometimes be fixed by its relation to w’hat lies
and Russell are right, and Dedekind and Benacerraf wrong: we take them as
outside the structure to which it belongs; what is constitutive of the number
too intimately connected with certain immediate applications of them to regard
3 is not its position in any progression whatever, or even in some particular
them as identifiable solely through the internal structure of the natural-number
progression, nor yet the result of adding 3 to another number, or of multiplying
system. Benacerraf writes:15
it by 3, but something more fundamental than any of these: the fact that, if
certain objects are counted ‘One, two, three’, or, equally, ‘Nought, one, two,
°f3i ?’ “biKt c“ >» “ element in then there are 3 of them. The point is so simple that it needs a sophisticated
some progression. What ,s peculiar to 3 is that it defines that role - not by being intellect to overlook it; and it shows Frege to have been right, as against
Dedekind, to have made the use of the natural numbers as finite cardinals
intrinsic to their characterisation. We shall see later that this represents, not
1 i
the theory might be idle we observe that thpro ’ . i ' in^nite s)stems, to dispel the fear that a trifling detail, but a fundamental principle, of his philosophy of arithmetic.
proof was essential for Dedekind: without a simply infin' SUCh SySte,m,On the c°nJrary>the '1
For all that, the thesis that Benacerraf is principally concerned to oppose,
tion of abstraction, we could not arrive at the that the natural numbers are quite specific objects, with which any correct
at Numbers Could Not Be’, Philosophical Review, vol. LXXIV, 1965, p. 70. analysis must identify them, by no means immediately follows from the falsity
54 Frege: Philosophy ofMathematics
of the pure structuralist thesis, as maintained by him It remains open a
they are specific objects, to be identified with ones characterisable in So7
different wav; or that they are specific objects, but characterisable On|v
numbers; or that, as Benacerraf believes, that they are not specific objects
all, even though capable of being characterised by reference to their app]icati 31
‘PPlicati,
rather than by pure structure. The question touches upon a critical issu^ CHAPTER 6
nevertheless: one, in fact, that Frege came to recognise as the most critic I
for his entire philosophy of arithmetic. If numbers are logical objects, and ye
capable of being defined as specific objects, the first of the above three possibili^
ies must be correct. In this case, die definition must represent them, not simnl • Numerical Equations and Arithmetical
as numbers, but as particular members of some more general range of logical
objects, classes or value-ranges. The process of definition must stop some
Lams
where, however. When it stops, how can the objects at which it stops be
identified other than as objects of whatever kind they are? And how can this
amount to anything but identifying them structurally, that is, by their particular
roles within that range of logical objects? Must we not eventually come upon The status of numerical equations
a fundamental realm of mathematical objects the only account of which will From § 5 to § 44 of Grundlagen, Frege occupies himself with a far-ranging
consist of a description, in the spirit of Dedekind, of its internal structure’ critique of the answers proposed by a wide cluster of philosophers and math­
Even if Dedekind’s account of number theory is to be rejected, may we not ematicians to the questions he has raised concerning arithmetic. This critique
be forced to offer a similar account of whatever lies at those foundations of is intended to make Frege’s own views, by the time he comes to present
andimetic with which Frege was concerned? This was the challenge with them, appear not merely plausible but inescapable, all alternatives having been
which Frege ought to have seen Dedekind’s work as presenting him; we shall demonstrated to be untenable. It is arranged with great skill, not merely to
see in due course how close he came to meeting it. accomplish this, but to establish in succession a number of positive points.

(a) Kant
§§ 5 to 17 are devoted to the status of arithmetical propositions. Frege begins
by drawing the obvious distinction between numerical equations and general
laws, and gives his attention to the former, considering only those that involve
addition. In a few devastating sentences he ridicules and utterly refutes Kant’s
account of these as synthetic but unprovable, though not classifiable as axioms
because of the infinite number of them and their lack of generality'. He begins
by observing that, when the numbers involved are sufficiendy large, as in
‘135664 + 37863 = 173527’, such equations are not self-evident. He thus
sounds for the first time a note of frequent occurrence in the critical sections
of Grundlagen'. this or that theory' fails for large numbers, or for the number
1, or for the number 0. Kant, Frege says, uses the lack of self-evidence on
the part of such equations as an argument for their being synthetic. But, he
comments, it rather goes to show that they cannot be unprovable: for how, it
they were, could we recognise them as true? Frege is here alluding to Kant s
remark:1

Kritik der reinen I ernunft, B 16.

55
56 Frege: Philosophy ofMathematics 6 Numerical Equations and Arithmetical Laws 57
The arithmetical proposition is therefore always synthetic, and this may be per­
As Frege comments, such a definition must be of the operation of addition:3
ceived the more clearly when one takes somewhat larger numbers ...
Grassmann in effect defines it as that operation for which the above general
equation holds. It needs to be remarked, although it is not by Frege, that such
Referring to Kant’s immediately preceding observation, which relates to his a definition presupposes the meaning of ‘<z 4- 1’ as already known: if a successor
favourite equation ‘7 + 5 = 12’:2 operation were explicitly invoked, the definition could take the form that ‘+’
was stipulated to be that binary operation that satisfies the two equations:
One has to go beyond these concepts [of seven and five] by calling in aid an
intuition corresponding to one of them, say of one s five fingers, or ... of five a + 1 = a'
points, ... a + b’ = (a + b)'
where la” denotes the successor of a. These then form the recursion equations
Frege comments, ‘Kant wishes to call in aid an intuition of fingers or points. for addition over the positive integers, just as they were subsequently given by
He thus runs the risk of making these propositions appear empirical, contrary Dedekind. Frege’s criticism is that ‘a + b’ would be an empty symbol, if there
to his own opinion; for an intuition of 37863 fingers is in any case not a pure wras no operation, or more than one, satisfying these conditions: ‘Grassmann
one.’ On Kant’s own theories, only a pure intuition could underlie a synthetic simply assumes without proof that this does not happen, so that the rigour is
a priori truth. The upshot is that, if numerical equations are a priori, they only apparent.’
must be provable. The definition would be justified by the theorem proved by Dedekind,
establishing the existence of a unique function specified by primitive recursion,
that is to say, by an arbitrary pair of recursion equations. It is precisely such
(b) Leibniz a justification which Frege is demanding. It would furnish an example of those
Frege then turns, in § 6, to Leibniz, who gives a purported proof of‘2 + 2 = 4’ ‘propositions on which the admissibility' of a definition rests’ to which he
from definitions of each number from 2 to 4 as the result of adding 1 to its referred in § 3. In his middle period, Frege developed a very’ rigorist view of
predecessor. He observes that Leibniz’s proof tacitly assumes the associative definitions, barring any that required some proposition to be proved before it
law for addition: ‘if this law is assumed, it may be easily seen that we can in could be admitted; quite evidently, when he wrote Grundlagen, he had as yet
this way prove every numerical equation in addition. The observation that no objection to such definitions, provided that the necessary' justification was
there is a gap in Leibniz’s proof is not a mere passing cavil at an oversight on supplied. Whether or not he here had in mind the specific justification that
his part, it is of critical importance for the ensuing discussion, as establishing Grassmann’s definition required, it is impossible to say. If he did, he was very
that even numerical equations cannot be proved without appeal to some general close to Dedekind’s demonstration of the validity of recursive definitions; but,
arithmetical law, and enables Frege to fasten attention on the character of even if so, he eventually preferred a definition connecting addition more
sue He pauses to commend Leibniz’s idea of defining each individual directly with cardinality.
positive integer from 1 and the operation of adding 1, remarking that he sees Thus, at the end of § 6, we have reached the conclusion that numerical
no o er way in w’hich they could be defined; here, then, is another positive equations in addition cannot be proved from definitions of the individual
result established at an early stage. positive integers alone. Granted that all those greater than 1 are to be defined
in terms of 1 and the successor operation, such equations can be proved with
the help of a general law, namely a special case of the associative law for
(c) Grassmann addition, or, otherwise expressed, the second recursion equation. This law
To give such proofs of numerical equations in addition, we need the associative may possibly be derivable from a suitable definition of addition; but, if so, that
law only in the special form definition will itself need to be justified by means of a general theorem, which
Frege does not state explicitly, but is in fact the theorem validating recursive
a + (b + 1) = (a + b) + L 3 Frege writes, slightly obscurely, that the objection that this definition of the sum is circular ^an
Perhaps be evaded if we sav ... that what he is intending to define is not sum but addition . the
Frege observes thal Grassmann tries to obtain the law in this form by definition. remark is accurate: the recursion equations do not of themselves constitute a e 0 ’
'•e. do not allow it to be eliminated from all contexts; but we may legitimately define addition to be
2B 15. the unique binary function satisfying those equations - though the definition requires jus i •
(Otherwise expressed, *+’ is second-order, but not first-order, definable from 1 and successor.)
58 Frege: Philosophy ofMathematics 6 Numerical Equations and Arithmetical Laws 59
definitions. The status of numerical equations, as analytic or synthetic, a number that we mention; for no suitable general principle covering all of them,
posteriori or a priori, thus for the time being remains undecided: as Frege and obtained by empirical induction, can be framed. Variations on these
observes at the beginning of § 7, it will depend on that of whichever general arguments are pressed in § 8: the upshot is that the definitions of individual
law is appealed to in proving them. numbers, in terms of 0 or 1 and the successor operation, ‘neither assert
observed facts nor presuppose them for their justifiability7’ (§ 9).

(d) Mill
Application
Or so one would think; but, as he goes on to remark, Mill denies this
dependency. Although one might have expected Frege at this point to enquire Since the numerical equations are derivable from those definitions with the
more closely into the justification of the associative law, the mention of Mill help of some general law, Frege proceeds in § 9 to ask after the nature of
diverts us along a different path; §§7-11 are now devoted to a critique of such laws. Can they be highly general laws of nature, arrived at by empirical
Mill’s empiricist philosophy of arithmetic. Mill appears to accept that each induction, as Mill maintains? Here we come upon a feature of Grundlagen
particular number after 1 must be defined as the result of adding 1 to its which inevitably causes any reader who has not also read Part III of Grundge-
predecessor; for he says that we may call ‘Three is two and one a definition setze to form an incorrect picture of Frege’s philosophy of mathematics. For
of the number three,4 and later that ‘Each number is considered as formed by Mill, the inductive truths governing the operation of addition are arrived at by
the addition of a unit to the number next below it in magnitude’.5 He holds, observing the results of physically adjoining two aggregates of physical objects.
however, that these ‘are definitions in the geometrical sense, not the logical; ‘What the name of number connotes is the manner in which single objects of
asserting not the meaning of a term only, but along with it an observed matter the given kind must be put together, in order to produce that particular
of fact’.6 The observed fact corresponding to the definition of ‘3’ is claimed aggregate.’ To form ‘the aggregate which we call fouf, for example, ‘two
to be that ‘collections of objects exist, which while they impress the senses aggregates of the kind called two may be united; or one pebble may be added
thus, 0 „ °, may be separated into two parts, thus, 00 <>’. ‘What a mercy7, then’, to an aggregate of the kind called three'-,9 in this way arithmetical laws are ‘in
comments Frege, ‘that not everything is nailed fast’ (§ 7). Having enquired reality physical truths obtained by observation’.10 This leads Frege to make
after the physical facts underlying the numbers 0 and 1, unfairly so because the wholly justified accusation:
Mill does not suppose them to be defined in this way, Frege invokes the
universal applicability of number: on Mill’s account, it w’ould be incorrect to Mill always confuses the applications that can be made of an arithmetical propo­
sition, which are often physical and do presuppose observed facts, with the
speak of three strokes of the clock, three tastes or three solutions of an pure mathematical proposition itself. The plus sign may indeed seem, in many
equation. (Strictly speaking, Frege is wrong to cite the clock, since, in asking applications, to correspond to a process of aggregation. But that is not its meaning:
how often it struck, we are calling for an ordinal, not a cardinal, number.) for in other applications there is no question of heaps or of aggregates, or of the
Mill gives a proof of ‘5 + 2 = 7’ after the same fashion as Leibniz’s proof relationship of a physical body to its parts, for example when the calculation
of 2 + 2 = 4’, making a similar surreptitious appeal to associativity7;7 but relates to events.
he asserts that such equations ‘do not follow7 from the definition itself, but
from the observed matter of fact.8 This piece of carelessness, characteristic As a criticism of Mill, one reiterated in the footnote to § 17, this is completely
of the great empiricist, enables Frege to ask where in the proof the observed apt. The point is repeated in § 16, where Frege says of applications of arithme­
fact should hav e been cited; if Mill had allowed that such equations did follow tic to physics:
from the definitions, but had claimed for them an empirical status on the
ground that the definitions themselves rested on empirical facts, his position It is ... a mistake to see in such applications the real sense of the propositions;
in any application a large part of their generality is always lost, and something
would have been stronger. Even so, Frege urges that he would have no escape particular enters in, which, in other applications, is replaced by something else.
from having to maintain that we observe facts relating to every individual
These remarks naturally induce the reader of Grundlagen to interpret Frege
4 System of Logic, book II, chap. VI, § 2. as an advocate of the inviolable purity of mathematics, for whom its external
' Ibid., book III, chap. XXIV, 6 5.
6 Book II, chap. VI, § 2.
7 Book III, chap. .XXIV, § 5. of Logic, book III, chap. XXIV, § 5.
* Book II, chap. VI, § 2. Ibid., § 7.
 60 Frege: Philosophy ofMathematics 6 Numerical Equations and Arithmetical Laws 61
applications are adventitious and irrelevant to their essence. It comes as an real numbers and is among the least read of his writings. It is there apparent,
enormous surprise to such a reader, therefore, to come in Grundgesetze}} upon not only from his explicit statements, but from his rejection of rival means of
the statement that constructing the real numbers, such as those of Dedekind and Cantor, and
from the method of constructing them which he himself adopts. It comes out
It is applicability alone that raises arithmetic from the rank of a game to that of clearly from his criticism of Cantor, in particular, that his fundamental ground
a science. Applicability therefore belongs to it of necessity.
for rejecting rival theories like Cantor’s was their failure to satisfy the demand
that the principle governing all possible applications of the real numbers should
This might be thought to represent a complete change of view; but a little be displayed by their definition. Any system of objects having the mathematical
later in Grundgesetze we find Frege criticising Helmholtz in terms very similar structure of the continuum is capable of the same applications as the real
to those in which he had criticised Mill in Grundlagen as one of those who numbers; but, for Frege, only those objects directly defined as being so appli­
‘confuse the applications of arithmetical propositions with the propositions cable could be recognised as being the real numbers. It is not only Frege’s
themselves’.12 He continues: theory of real numbers that is overlooked by one who neglects to study Volume
II of Grundgesetze, but a leading component of his general philosophy of
As if the questions as to the truth of a proposition and as to its applicability were
not quite distinct! I can very well recognise the truth of a proposition, without mathematics.
knowing whether any application can be made of it. This component is present in Grundlagen, too, but far from obviously. The
rival theories of natural numbers reviewed in that book do not, of course,
Helmholtz, like Mill, was a proponent of an empiricist philosophy of mathemat­ include that of Dedekind, which wanted another four years until publication.
ics. It is when he is criticising empiricism that Frege insists on the gulf between But Frege’s deepest objection to it would have been that it attempted to
the senses of mathematical propositions and their applications; it is when he characterise the totality of natural numbers purely in terms of its internal
criticises formalism that he stresses that applicability is essential to mathemat­ structure, and relegated their application as finite cardinals to an appendix to
the theory. For Frege, conversely, that was the salient type of application that 11
ics. Formalism, properly so called, is not considered in Grundlagen, only its
more timid cousin, postulationism: and therefore Frege has no occasion in could be made of the natural numbers, and hence must be made, as he made
that book to sound the latter of these two notes. it, central to their definition. It was sufficiently general for such a purpose,
At first sight, there is a flagrant contradiction between what F*-ege says in being quite unspecific as to the type of objects concerning which the question
the one connection and what he says in the other; but the appearance is ‘How many?’ could be answered by citing a natural number; for that reason,
illusory. Any specific type of application will involve empirical, or at least non- it involved no concept peculiar to any non-mathematical subject-matter. With­
logical, concepts alien to arithmetic; very often, it will depend upon empirical out the discussion of real numbers in Grundgesetze, Part III, it would be easy
presuppositions. To make such applications intrinsic to the sense of arithmeti­ to suppose that Frege’s definition of the natural numbers as finite cardinals
cal propositions is therefore to import into their content something foreign to in Grundlagen was due only to a certain traditionalism in his approach, or at
it, and to render their truth synthetic: that is the mistake of Mill and Helmholtz. most to his desire to characterise them, without appeal to the psychological
Vhat is intrinsic to their sense, however, is the general principle governing all process of abstraction, as quite specific objects. Doubtless both motives oper­
possible applications. That must accordingly be incorporated into the defi­ ated. Perhaps, too, the possibility of a purely structural characterisation had
nitions of the fundamental arithmetical notions. It is not enough that they be not so much as occurred to him. But the procedure which he adopted in his
defined in such a way that the possibility of these applications is subsequently 'Construction of the natural numbers was in complete consonance with what
provable; since their capacity to be applied in these ways is of their essence, was later to appear as one of the principal strands of his philosophy of math­
the definitions must be so framed as to display that capacity explicitly. ematics. Far from insisting on the purity’ of mathematics, and treating its
In Grundlagen, Frege did not expound this aspect of his philosophy of applications as philosophically irrelevant, he is, among all the philosophers of
arithmetic (in the sense in which ‘arithmetic’ embraces analysis as well as mathematics, that one who assigned to applicability’ its most central place.
number theory), and therefore it is not apparent to a reader of that book. He
set it forth only in Part III of Grundgesetze, which is devoted to his theory of
General arithmetical law’s
"Vol. II, §91. If the general laws of arithmetic were based on induction in the sriennhc
"Vol. II, § 137, fn. 2. sense, Frege argues in § 10 of Grundlagen, they would have to be armed
 62 Frege: Philosophy ofMathematics 6 Numerical Equations and Arithmetical Laws 63
• ns for addition; and it hints at the method Frege will adopt for defining
eqUa cent natural number, namely as comprising all and only those objects
Since the equations can be proved from the definitions of the individual thC inable from 0 by reiterating the successor operation.
numbers only with the help of some general law, we should thus lose the
whole advantage of those definitions, and should have to find some other way
of establishing the numerical equations. Besides, the natural numbers are not Intuition
all alike, as are points in space or moments in time. Each has its particular In § 12, assuming his reader’s agreement that he has refuted the new that the
properties: we cannot say that what happens at one place in the sequence of general laws of arithmetic are a posteriori truths, Frege asks whether they are
natural numbers must happen at any other, as we say that what happens at synthetic a priori or analytic. Kant, as he remarks, held them to be synthetic
any spatio-temporal position must happen at any other, if the conditions are a priori; and, in this case, Frege comments, we have no alternative but to
the same. invoke a pure intuition as the ultimate ground of our knowledge of them.
These considerations are really poindess: in § 10, Frege has, for a moment, Despite his previous sarcasm about the intuition of 37863 fingers, and despite
lost the thread of his argument. Once he has, by distinguishing the sense of his observation in this section that ‘we appeal too readily to inner intuition
an arithmetical proposition from its physical applications, refuted Mill’s idea when we cannot cite any other ground of knowledge’, Frege should not be
that arithmetical truths are attained by induction from the results of physical understood as disparaging the whole notion of pure intuitions. In Grundlagen,
operations, nothing more needs to be said about their supposedly inductive § 89, he says expressly that, ‘in calling geometrical truths synthetic and a priori,
character. It is wrong to argue, as Frege in effect does, that we cannot arrive [Kant] revealed their true nature’; we must conclude, from the comment in
at number-theoretic conjectures on this basis: Goldbach’s conjecture is an § 12, that he regarded our knowledge of them as resting on pure intuition.
obvious counter-example, and there are many more. It is equally obvious that Mentioning certain contemporaries who agreed with Kant about arithmetical
we do not trust such conjectures, which prove to be mistaken as frequendy as truths, Frege cites from Hankel’s book on analysis the phrase ‘the pure
they prove correct, and that we certainly do not assert them as true before we intuition of magnitude’.13 He comments:
hit on a proof, or at least have, in the shape of a computer proof, empirical
evidence that there is a proof; there is no a priori reason why the smallest If we consider everything that is called a magnitude: numbers (,-lnzahlen), lengths,
counter-example to a generalisation should not be very large. areas, volumes, angles, curvatures, masses, velocities, forces, intensities of illumi­
Before bringing § 10 to a close, Frege introduces an image, that of the nation, electric currents, etc., it is easy to understand how they can all be brought
under one concept of magnitude; but the expression ‘intuition of magnitude’, let
borehole, which has no probative force, but suggests an important principle. alone ‘pure intuition of magnitude’, cannot be recognised as correct.
He supposes that the drill has so far penetrated a sequence of very' different
rock strata, but that we have noticed that the temperature increases uniformly Ten years previously, at the beginning of his Habilitationsschrift (post-doctoral
with the depth; and he comments, somewhat unreasonably, that we cannot dissertation) of 1874,14 Frege had expressed the same view, arguing that ±e
presume that the temperature would continue to increase at the same rate with concept of magnitude had been gradually disentangled from intuition and that
further drilling. \\ e can indeed form the concept ‘what will be encountered by its connection with intuition had in any case been illusory: lines and plane
continued drilling, but can deduce only what is determined solely by the depth figures are intuitable, ‘but precisely what constitutes their magnitude, what
of a stratum, and that without needing to invoke empirical induction. Similarly, lengths and areas have in common, eludes intuition’. The concept of magnitude
we can form the concept ‘whatever is obtained by repeatedly increasing by 1’- has far too great generality' to be derivable from intuition.
The difference is that we simply encounter the strata reached by going on with In the ‘Rechnungsmethoden’ Frege discussed the concept of magnitude in
the drilling, whereas the numbers are constructed by the repeated addition of detail, and in Part III of Grundgesetze he attempted a precise mathematical
1, and hence their very natures are thereby determined; it follows that all their analysis of it. Here, however, it is really an irrelevancy, which Frege attempted
properties can be deduced from the specific way each was so generated. That, to disguise by listing Anzahlen (cardinal numbers) in the above quotation as
however, amounts to saying that the properties of each number follow from forming a species of the genus magnitude. Properly speaking, however, they
its definition It furthermore ‘opens up the possibility of proving the general I do not; but Frege seems to have been far less clear about this in Grundlagen
laws of numbers from the method of generation common to all of them’. This
is a way of saying that the general law's of arithmetic are to be proved by fTheorie der contpkxen Zahlsysleme, Leipzig, 1867, pp. 54-5.
uui v...v —..._____ „ des Grbssenbegriffes griinden’
Rechnungsmethoden, die sich auf eine Erweiterung
mathematical induction, as the associative law can be proved from the recursion Extension
(Methods of Calculation based on an L..... of the Concept of Magnitude’).
 64 Frege: Philosophy ofMathematics 6 Numerical
alone can furnish us with Equations andthe
intuitions. On Arithmetical
contra Laws 65
than he later was in Gnindgesetze. As he repeatedly remarks, a natural number parallel explanation. Having explained that a Dercent;gave a Precisely
senes to answer a question of the form ‘How many?’: more precisely, ‘How of that particular kind which involves awareness (Frmis an idea
many objects are there which satisfy such-and-such a condition?’. ‘How many he says that such a perception may be either a mit Bew^stsein),
miles to Babylon?’ is not of this form; Frege is wrong, in Gnindlagen, § 19, to cognition (Erkenntnis). The latter are subdw «**»*«;
assert that it is, saying that ‘a number that gives the answer to the question, the difference between which he explains as follows:” and “"“P's.
“How many?”, can also determine how many units are contained in a given
length’. The magnitude of a quantity’, on the other hand, may be specified by
citing a rational or real number, together with a unit of measurement; to ask An
Theobjective perception
former relates is atocognition.
directly an objectThis
andisis inA \ntUlt,on °r a concept
after it is to pose a question of the quite different form ‘How much?’ or ‘How indirectly by means of a characteristic (Merkmal) whichcan h' relates t0 *
great?’. If Frege did not draw the distinction sufficiently sharply in Grundlagen, things. ’ nicn can be common to several
he did so in Gnindgesetze, where he says, Anzahlen [cardinal numbers] answer
the question, “How many objects of a certain kind are there?”, while the real
numbers may be considered as numbers used for measurement, which state It was, rather, a thesis maintained by Kant that all our intuitions - our ideas
how great a quantity' is compared with a unit quantity’’.15 of individual objects - are sensible in character. This thesis Frege rejected,
The crucial question, therefore, is whether we can speak of intuitions of as he says expressly in § 89 of Grundlagen-.™ numbers, for him, are objects,
the natural numbers. Frege declares outright that we have no intuition of a and it is endent at least for very’ large numbers that they are not given to us
I
large number such as 100,000, and proceeds to consider Kant’s notion of intuitively, if this involves perception or sensory’ imagination. For this refutation
1
I
intuition. He quotes from Kant’s Logik (§ 1) his distinction, among ideas,
between intuitions and concepts:
of Kant’s thesis to be effective, however, it must have been established that
numbers are objects. In ‘Booles rechnende Logik’, Frege had remarked that
‘the number 3 is not to be regarded as a concept, since the question what can
An intuition is an individual idea [l/orstellung] (repraesentatio singularis), a fall under it is senseless’: no object can be three in number. He had no right,
concept a general idea (repraesentatio per notas communes) or reflective idea however, save as an argumentutn ad hominem, to invoke Kant’s dichotomy
(repraesentatio discursiva).
between objects and concepts, since he admits other logical types: relations
He comments that: (§ 70) and concepts of second order or level (§ 53). In a sense, he preserves
the dichotomy, since he regards both relations and second-level concepts as
Here there is absolutely no mention of any relation to sensibility, which, on the / concepts of a kind; but the fact that we cannot speak of an object’s falling
other hand, is associated with intuition in the Transcendental Aesthetic, and under the concept 3 has no tendency, on Frege’s own principles, to show that
without which intuition cannot serve as the principle of our knowledge of synthetic the number 3 is an object.
judgements a priori. This, then, is one refinement that Frege makes of Kant’s classification: the
class of concepts is to be subdivided, so as to admit, besides first-level concepts
Citing Kant’s statement that ‘it is therefore through the medium of sensibility
proper, relation-concepts and concepts of second level (called in Gnindlagen
that objects are given to us, and it alone furnishes us with intuitions’,16 Frege
‘of second order’). A second emendation is the distinction between objective
concludes that he used the term ‘intuition’ in a wider sense in the Logik than and subjective ideas, which in the footnote to § 27 Frege condemns Kant for
in the Kritik. Frege allows that, in the former sense, 100,000 might be called
failing to draw Unknown to Frege, precisely the same demand for a sharp
an intuition, since it is certainly not a general concept; but, in this sense, an distinction between the subjective and objective senses of the word lorstellung1
intuition cannot serve as the foundation of arithmetical laws.
had previously been made by Bolzano in his Wissenschaftslehre of 1837. Bolzano
Frege s complaint is misstated. Kant had no intention in the Kritik of
understood it as covering, in its subjective sense, sensations, mental images
making dependence on sensibility part of the definition of the word ‘Anschauung1
and the like, and, in its objective sense, constituents of what he called ‘propo­
(conventionally translated ‘intuition’ when used by Kant or Frege), or of using
sitions in themselves’, which correspond to what, in his middle perio , rege
it in the Kritik in any narrower sense than that given by the definition in the
called ‘thoughts’. He even laid down the same principle of differentianon as
Logik; if he had, it would have been the merest triviality to say that sensibility
’’Vol. II, § 157.
16Kritik der reinen t'emunft, B 33. "Kritik, B 376-7.
’ Quoting, this time, from the Kritik, B 75.
/
66 Frege: Philosophy ofMathematics 6 Numerical Equations and Arithmetical Laws 67
that on which Frege was to insist, namely that subjective ideas require a subject switch from speaking of numbers as intuitions in the sense of Kant’s Logik
or bearer, whereas objective ones do not.” to speaking of them as objects was not intended to mark the difference
From § 27 onwards, Frege’s terminology diverges markedly from Kant’s. between ideas and what they are ideas of, but that between ideas of the
Kant’s term 'Vorstellung (standardly translated ‘idea’ as used by Frege, but subjective variety and those of the objective variety. The content or meaning
‘representation’ as used by Kant, although Kant intended it as the equivalent iBedeutung) of an expression was for Frege at that time simultaneously its
of ‘idea’ and 'idee' as used in philosophical writing in English and French) is significance and what it signified: the distinction became apparent to him
henceforth reserved by Frege for subjective ideas, that is, elements of the only w’hen he drew’ his distinction between Sinn and Bedeutung, and he was
stream of consciousness such as mental images. Unlike Bolzano, who retained strictly accurate in saying that he had split the former notion of content into
the Kantian distinction between intuitions and concepts within the realm of those two components.20 This explains the oddity of his later terminology"
objective ideas, as well as within that of subjective ones, Frege chose to treat he chose to retain the term 'Bedeutung for that which the expression signifies.
the term ‘intuition’ as applying only to occupants of the subjective side of the It explains also why the term ‘concept’ plays so striking a double role in
classification; ‘concept’, on the other hand, was used by him exclusively for Grundlagen, being used sometimes for the sense of a predicative expression
certain kinds of objective idea. Thus, in Frege’s revised terminology', the. and sometimes for its reference. Naturally, no coherent exposition can be
opposition between intuitions and concepts is misconceived: what is correlative given of the doctrines of Grundlagen without acknowledging a distinction
to the notion of a concept is that of an object. between significance and what is signified; but, in reading the book, w-e must
Frege, in his early period, was alive to the difference between sign and thing bear in mind the fact that Frege was not himself making such a distinction.
signified; given the distinction between subjective and objective ideas, he as His failure to do so means that there wras at that time a radical incoherence
yet saw no need for any further differentiation between an objective idea and at the very heart of his thinking, though one that obtrudes very little in the
that of which it is the idea. This latter distinction is that between sense and argumentation of the book. It is for this reason that it is so misguided to try’,
reference, which, when he later came to draw' it, was a distinction within the as writers like Baker, Hacker and Shanker have done, to read into the work
realm of the objective. In § 12 he has not yet introduced the subjective/objec- of Frege’s early period a system of philosophical logic different from
tive distinction; this is what makes it possible for him to say, not that w'e may the theory' presented in his middle period, but equally worked out and
perhaps have an intuition of 100,000 in the sense of ‘intuition’ explained in articulated.
Kant’s Logik, but that 100,000 may perhaps be an intuition in that sense. It The conclusive proof that numbers are not intuitable had in fact been given
does not disturb us to miss, in Grundlagen, any differentiation between the before § 12. In § 5 Frege had said that, if wre had an intuition of 135664
idea and that of which it is the idea when he speaks of concepts; but the lack fingers, another of 37863 fingers and a third of 173527 fingers, the correctness
of any such differentiation for objects appears to us deeply shocking, and it is of the equation ‘135664 4- 37863 = 173527’ ‘would have to be immediately
with dismay that we read, in the footnote to § 27, that ‘objective ideas can be evident, at least for fingers’: for Frege, intuitions involve not only particularity,
divided into objects and concepts’. In Grundlagen, how’ever, Frege considered but immediacy. It could easily be maintained, how’ever, that arithmetic needs
the distinction between objective and subjective ideas to be sufficient; w'hen, to appeal to intuition, without believing that we have intuitions of the individual
later, he made the distinction between objective ideas and that of which they numbers; the only certain way to refute the claim is by framing definitions,
are ideas, he made it uniformly for objects and concepts. ‘Objective ideas’ and supplying proofs, that show' the recourse to intuition otiose. Bolzano had
were transformed into senses; objects and concepts were what such objective begun the process of eliminating intuition from analysis by proving something
ideas were ideas of. apparently obvious to geometrical intuition, namely' the mean value theorem,
The brilliance and clarity of Grundlagen, and the cogency of many of its stating that a continuous function on the reals must assume the value 0 at
arguments, make it difficult for us to take in the fact of Frege’s blindness, some point in an interval in w'hich it has both positive and negative values.
during the whole of his early period, to what seems to us an obvious need Bolzano expressly proclaimed the value of proving apparently obvious state­
for a distinction. He simply had no consciousness, until he formulated the ments, in order to establish on what they actually rest, and, in particular, that
principles of' LJ
his imiddle-period
■ ■ ■ ,.. theory, of the necessity for distinguishing ^ey d° not depend on spatial intuition; but, as he show'ed in his example of
between the significance of an expression and that which it signifies. The

” U'issenschaftslehre, §§ 19, 48 and 270-3.

Resuhat gen>dhren, wenigstens eine mile IVurzel der Glachung neg
 68 Frege: Philosophy ofMathematics 6 Numerical Equations and Arithmetical Lam M I
not forget that the propositions thus obtained, which mi^ht . 07
a continuous but nowhere differentiable function, what appears obvious may those given here, would yet assert far less than do the lart C°!nC,de verbally with
not even be true.22 hold good only in the domain of that intuition on which Cau?e WOu*d
One of Frege’s aims was to accomplish the same for number theory. Already wmen they were based.
in ‘Rechnungsmethoden’ he had pronounced on the difference in this respect
between arithmetic and geometry: A
temporal of sequence
conceptsequences, basedonontemporal
onebased intuition nnlv
spatial intuition could♦. he
° • >•. j Only t0
the concept of a sequence has a far greater general;^ SpatiaJ ones5 but
There is a remarkable difference between geometry and arithmetic in the way in between one term and the next may be of anv kind wh^’ SinceJthe relation I
which their fundamental principles are based. The elements of all geometrical
constructions are intuitions, and geometry' appeals to intuition as the source of its makes it definable by purely logical means. and
axioms. Since the subject-matter of arithmetic is not intuitable, its fundamental
principles cannot likewise spring from intuition. § 91 ofofGrundlagen,
anInexample a proposition thatcites
Frege might firstlast
theatvery theorem
sight be tab Besdf^sschrifi as
but which he has been able to prove ‘without ghn t0 be synthetic>
In the Preface to Begriffsschrift, having distinguished analytic and synthetic intuition’; applied to natural numbers, the theorem yiewTS frOm
propositions, he declared his ambition to demonstrate arithmetical laws to be
of the former kind: numbers, in §80,
Likewise, and remarks it isancestral
that the
he uses only by to define
means of thia °f natUral
possible to reduce mathematical induction an infc defimuon that it is
Having posed to mvself the question to which of these two kinds arithmetical mathematics, to the general laws“trf ori"’ an‘"fc'“« W®ntly peculiar to
judgements belonged, I had first to see how far one could get in arithmetic by the ancestral: ’ He Comments ™ <** definition of
means only of inferences based purely on the laws of thought, which rise above
everything particular. The path I followed was first to try to reduce the concept
of ordering in a sequence to logical succession, in order to advance from there to
the concept of number. So that nothing from intuition should intrude, everything Since the relation 0 has been left indefinite, the sequence is not necessarily to
had to depend upon the absence of any gaps in the chain of inference. be conceived in the form of a spatial and temporal ordering, although these cases
are not excluded. Some might perhaps regard another definition as more natural,
for instance: if, starting from x, we always transfer our attention from one object
Accustomed as we are to the geometrical representation of a function on to another to which it stands in the relation 0, and if in this way we can finally
the real numbers, we find it unsurprising that fundamental theorems of analysis reach y, we say that y follows x in the 0-sequence. This is a way of investigating
should once have been supposed to rest upon spatial intuition, so that it needed the matter, not a definition. Whether we reach y by transferring our attention may
the efforts of Bolzano and his successors such as Cauchy and Weierstrass to depend on a variety of subjective surrounding circumstances... Whether j follows
expel intuition from the theory'. It is less obvious that this should have needed x in the 0-sequence has in general nothing to do with our attention and the
to be done for number theory; but Frege rightly fastened on the concept of a conditions for transferring it.
sequence, which it was natural to discuss in terms of a temporal process of
moving from term to term. In Begrijfsschrift, Frege succeeded in giving a purely He concludes that ‘by means of my definition the matter is f'™ *
logical analysis of the concept by means of his celebrated definition of the
ancestral of a relation; and at the beginning of Part III of that work, he
emphasised its philosophical significance:
/ " a^ZI n iJd^d »
One sees from this example how pure thought alone, prescinding from any content between arithmetic and geometry': if in geometry g points, lines
given by the senses or even by an a priori intuition, is capable of eliciting, from derived from intuition, that is explicable from L represen-
a content that arises out of its own constitution, judgements that at first sight and planes intuited are not really particular ones, an s rithmetic, because,
seem to be possible only on the basis of some intuition ... The propositions tatives of their entire kind.’ Nothing similar is possiblem §
about sequences developed in the following far surpass in generality' all similar
ones that can be derived from any intuition of sequences. If anyone were to regard
it as more appropriate to use an intuitive idea of sequence as a basis, he should
as previously remarked, no one number is entire y
magnificent piece of rhetoric quoted in a previous p ’
-n wbjch Frege
e> it can be
I
declares that, although we cannot imagine non- uc clidean) geometry
“The example was given in his Funktionenlehre, written in about 1830, but not published until encompassed by conceptual thought, so that e ;nt0 confusion if
a century later, in Bernhard Bolzanos Schriften, vol. I, Prague, 1930, having been first reported by do not govern everything thinkable, whereas eveo wie(jge of these laws
M. Jasek in 1921.
we attempt to deny any of the laws of arithmetic. Our knowledge
 70 Frege: Philosophy ofMathematics 6 Numerical Equations and Arithmetical Laws 71
cannot, therefore, be based on intuition: the basis of arithmetic lies deeper on|y when their proofs were fully formalised. If we are content with the
even than that of geometry. § 15 cites Leibniz and \\ .S. Jewns as favouring ^j-maHsed proofs usual in mathematics, either of two opposite errors may
the analyticity of arithmetical truth. § 16 raises against this the difficulty posed Un ur Some seemingly self-evident step in the proof may in fact depend in
I
by the question, ‘How do the empty forms of logic come to disgorge so rich 0CttUon intuition, rather than representing a purely logical inference; or, con- I
a content?’ (in Austin’s fine, though free, translation). This is die problem of P^selv, some purely logical transition may be taken to rest upon intuition,
the value of analytic judgements, to which Frege is not yet ready to expound ^cause it fails to conform to any recognised form of logical inference, when
his solution: he contents himself, in § 17, with pointing out that it coincides •n fact it could be broken down into a sequence of shorter but purely logical
with that of the fruitfulness of deductive reasoning, since we can always 1 Hence Grundlagen itself could do no more than make it probable that
transform anv valid piece of reasoning into an analytic truth by framing the arithmetical truths depended upon the laws of logic alone: to establish that, it
conditional whose antecedent is the conjunction of the premisses and whose was necessary7 to present fully formalised proofs of them from fundamental
consequent is the conclusion. Such a transformation has die advantage that it logical principles; that, of course, was to be the task of Grundgesetze.
‘leads to a general proposition, which need not be applicable only to the facts Part 1 of Grundlagen ends with § 17. At the beginning of § 18, Frege records
immediately under consideration’. Frege, as a philosopher of mathematics once more the decision reached to define the individual numbers in terms of
placing the greatest emphasis on its applications, was highly conscious that the 1 and successor, and remarks that these definitions remain to be completed
ultimate point of establishing analytic truths is to enable us rapidly to infer by defining the number 1 and the successor operation. Further, even to derive
non-logical conclusions from non-logical premisses. numerical equations, we need in addition some general laws, which, in virtue
i of their generality, can follow only from the general concept of number
The complaint is sometimes made that Frege concentrated on the finished
/ mathematical product, not on the process of constructing it. Of this he was {Anzahl). It is to views concerning this concept that Part II is devoted.
perfecdy well aware: the questions he was concerned to ask concerned the
finished product; he assumed only that no considerations about the process of
arriving at it would invalidate his answers to those questions. § 17 contains a
quotation from Leibniz which exactly states Frege’s point of view: ‘it is here
a matter, not of the history of our discoveries, which is different in different
people, but of the connection and natural order of truths, which is always the
same.’
In the very first sentence of the Introduction to Grundgesetze Frege stated,
more accurately than he had done in Grundlagen itself, what he took the aim
of Grundlagen to be:

In my Grundlagen der Arithmetik I sought to make it probable that arithmetic is a

branch of logic and needs to borrow no ground of proof either from experience
or from intuition.

That w as indeed Frege s central concern; not to arrive at certainty concerning

the truths of arithmetic, but to establish the ground for our acceptance of them,
and, in particular, to refute the belief that intuition was among those grounds;
and, in this regard, he was following in the footsteps of Bolzano. Formalisation
of proofs would unquestionably increase certainty, when this was in any degree
m doubt; but that was not Frege’s objective in formalising them. His aim was,
ra er, to achieve certainty, not about the truth of arithmetical theorems, but
about what was needed to establish their truth. In § 90 of Grundlagen, Frege
conceded that he had not, in that book, conclusively demonstrated the deriv­
ability of the truths of arithmetic from the laws of logic alone: that could be
 7 What is Number? 73
takes up the greater part of Part III, from § 29 to § 44, which
completion
constitutes Frege ’s relentless critique of the abstractionist account of the

The beginning is unpromising. In § 19, Frege cites Newton as proposing

to define number as a ratio between quantities, which was to be his own
CHAPTER 7 characterisation of real numbers in Grundgesetze, Part III. Frege correctly
remarks that Newton’s explanation covers ‘numbers in the wider sense, includ­
ing fractions and irrationals’; but since, as we have already remarked, he does
What is Number? not appear in Grundlagen to distinguish these from cardinal numbers as sharply
as he does in Grundgesetze, he flounders somewhat, and fails to make the
simple point as cleanly as he ought.
After a brief dismissal, in § 20, of those who consider number indefinable,
In Grundlagen, § 18, Frege reiterates that the individual numbers are to be not because they know of any obstacle in principle to defining it, but only
defined in terms of 1 and successor; to complete the definitions, he says, we because attempts to define it have failed, Frege gets into his stride. The salient
need to define 1 and the successor operation. Furthermore, he repeats, we question of §§ 21 to 54 is: what is a number the number of? Alternatively
need general laws even to derive numerical equations. ‘Just because of their expressed, it is: of what is a number a feature? A subsidiary question is what
generality, such laws cannot follow from the definitions of the individual sort of feature a number is, and, in particular, whether it is an objective or a
numbers’, he declares, ‘but only from the general concept of number’; the subjective feature. We may interpret the question ‘What is a number the
word he uses for ‘number’ in the phrase ‘concept of number’ is 'Anzahl', i.e. number of?’ as follows. As Frege remarks in § 21, number-words occur in
‘cardinal number’. Frege was, indeed, right to think that, if we want to prove ordinary speech principally as adjectives. Tacitly, he assumes that any sentence
laws concerning general cardinal arithmetic, we have no choice but to define containing a number-adjective can be transformed into what he calls an ‘ascrip­
‘cardinal number’. We do not need laws of such generality in order to derive tion of number’ (Zahlangabe') (§ 47). An ascription of number is a direct answer
numerical equations concerning the positive integers, on the other hand; if we to a question ‘How many?’, and takes the form ‘There are (just) «...’. For
are concerned solely with the laws governing them, Dedekind’s book disproves instance, ‘The Kaiser’s coach is drawn by four horses’ (§ 46) can be expressed
Frege’s assertion. Provided that we know how to justify definition by recursion, as ‘There are four horses that draw’ the Kaiser’s coach’. The question then
and then define the arithmetical operations, addition, multiplication and any becomes ‘What, in general, is an ascription of number about?', or, alternatively,
others we need, by that means, we can derive the arithmetical laws holding ‘What is a number ascribed to?’.
for the positive integers from the fact that they consist in whatever can be To us, it appears that there is a simple method of answering this question,
attained from 1 by repeated application of the successor operation, that is, easily applied. Suppose we had to answer the analogous question what, in
from the principle of mathematical induction. Precisely this was implicit in general, a virtue is a feature of or is ascribed to. Expressions that we use to
Frege’s earlier remarks, in § 10, about the borehole; here that insight seems ascribe virtues are those like ‘... is honest’, ‘... is generous’, etc., in the sense
to have slipped his mind. °f ‘ ... has an honest (generous, etc.) character’. The w’ord or phrase that
That might seem to make little difference: for, if we defined the positive fills the gap in any such expression to complete the sentence denotes that to
integers in that way, and supplemented this definition by specifying what object ( which the virtue is being ascribed. Hence, to determine to what, in general, a
the number 1 is to be taken to be, and what operation is to be that of successor, virtue is ascribed, we have to ask after the widest range of terms that can fill
we should thereby have fully explained the concept of positive integer. If Frege the gap in a virtue-predicate so as to yield an intelligible sentence. Obviously,
had taken this path, he would certainly so have defined 1 and successor as to an expression belongs to this range if and only if it is a term denoting a person,
elucidate the use of number-words to answer questions of the form ‘How hence it is, in general, a person to whom a virtue is ascribed.
many ...?’; but it does not suit his purposes to take that path at this stage. fi course, in that case the technique merely gives an answer that could
Instead, he embarks on a review of answers to the question, ‘What is number? > ftave been given straight off; but this justifies the use of the technique in less
where, of course, this means primarily ‘What does it mean to speak of the 0 vious cases. In view of the generality of number - the fact that there is no
number of things of a given kind?’. The preliminary enquiry into answers to ^estriction on the type of objects of which we can say how many there are: -
this question occupies Part II of Grundlagen, comprising §§18 to 28; its e Can give no more specific answer to the question what, in general, can
72
 74 Fre^e: Philosophy ofMathematics
7 What is Number? I
the gap in an ascription of number than ‘a predicative expression’; a (
expression whose meaning consists in its applying, or not applying, to an use to explain this in terms of the number of visual representations of the
given object. Frege explains, in § 74, that, as he uses the term ‘concept’, the figures there may be: for we need to be satisfied that every figure is represented,
and none represented more than once.
general form of a judgeable content that is about an object is ‘a fa]ls Under
the concept F’: otherwise expressed, whatever can be true of or false of aa § 22 accomplishes something different. First, it establishes the obvious fact
that the plural subject of a sentence ending ... are green’ functions differendy
arbitrary’ object may be taken as standing for a concept. \\ e thus arrive at the i
from that of one ending ‘ ... are 1,000’. In the former, the predicate applies
desired conclusion: that to which, in general, a number is ascribed is a concept
distributively, that is, as holding good of each object to which the subject-term I;
This is the eventual answer to the question which Frege supplies in § 46
applies; in the latter, it applies, not to each of them, but to them collectively.
after rejecting in turn a sequence of different answers proposed by various This naturally leads to Mill’s proposal that the property denoted by the
philosophers and mathematicians. Indeed, to anyone who has been tempted number-word attaches to a composite entity denoted by the subject-term, an
by those answers, Frege’s solution comes as a revelation, resolving all the aggregate or, in Mill’s own terminology, an agglomeration. This is really the
difficulties they provoke and he has exposed. To us, however, the detour only serious rival to Frege’s own semantic analysis of such sentences. It is
through these fallacious answers may seem a piece of unnecessary business. everybody’s first thought, and arises from being too readily impressed by the
The correct answer could have been arrived at with no perplexity, simply by grammatical similarity between singular and plural subjects. On this naive
applying the technique just sketched. If it is not quite so obvious as the answer view, when the predicate applies distributively, the plural subject simul­
to ‘To what is a virtue ascribed?’, it seems nearly so: pages of discussion were taneously denotes each of the objects to which it applies: the sentence states
surely not needed in order to light upon it. that the predicate holds good of each such object. But, when the predicate
Such a judgement would be quite superficial. The technique invoked for applies collectively, as when we say, ‘Dodos are extinct’, or, ‘Gorillas are
answering the question is essentially Frege’s; and its validity depends upon becoming rare’, we have something resembling a singular sentence. The predi­
acceptance of Frege’s semantics. Until we have accepted that semantics, either cate is not intended to apply to each individual object: it is not of a kind which
in the rudimentary state it still assumed at the time when Grundlagen was it would make sense to apply to any individual object. Rather, it applies to
written, or in the more developed state in which it appeared in the writings (past or present) dodos or gorillas as a whole-, and so, trying to analyse the
from 1891 onwards, we do not have the answer ‘a concept’ available as a plural subject by analogy' with a singular one, we come up with the idea of
possible one; even if the technique for arriving at an answer be employed, we treating it as standing, in such a sentence, for a single composite object made
are not in a position to say that the expression filling the gap in a sentence of up out of the individual ones. Such was Mill’s theory’; and so did many’ think
the form There are n ... ’ stands for a concept. The arguments Frege uses, of the matter in Frege’s time.
in §§ 21-8,, in favour of his answer to the question ‘What is a number the Frege tackles this in § 23. It has two fundamental weaknesses. First, for an
number of? and against answers proposed by others, are arguments for adopt­ aggregate to exist, there must be some relation between its parts in virtue of
ing ’s analysis of ascriptions of number. Since that analysis is both syntactic which they cohere; but, to give the number of objects of a certain kind
an semantic in character, they are also suasions in favour of his semantic correctly, there need be no particular relation between them at all. ‘Do we
eopr. it is, among other reasons, because that theory is capable of giving a have to gather all the blind in Germany into an assembly for the expression
convincing account of ascriptions of number, and rival semantic theories are “the number of blind people in Germany” to have a sense?’ is how Frege
it* V n°r^ ta^e f°r ^rante<^ t^le correctness of a semantics at least satirically puts the point. This may fail to impress those who think in terms
, ?r re?ean lines, and do not so much as stop to consider one of of disconnected ‘sums’ of matter; they may find it harder to answer Frege s
thwe implicitly underlying the views Frege here so decisively refutes, question whether there are really any aggregates of proofs of a theorem or of
much mnrpUrh10n r°m t0 § ^8 is in fact very economical. §§ 22-5 do events. But the critical objection to the theory’ is that, to assign any definite
physical oronert^ a?SWer question posed in § 21, whether number is a number to an aggregate, we should have to know what it was to be considered
adequately dealt with by the appeal, in § 24, to the an aggregate of\ and there is no one way to regard an aggregate as composed
abstracted from v+ ? ,nurnber: ‘it would be remarkable if a property of parts. ‘A bundle of straw can be split up ... by separating it into individual
to events ideas and™ mgS C°uld be transferred without change of sense straws, or by making two bundles out of it’; and this is to ignore the tact that
shouTd X in Whit PtS • • •■ Tt iS absurd what is by nature sensib,e each straw consists of cells or of molecules. A plural subject does not. in any
could not intelligibly talk^fT511’16' ^nUmber Were a physical ProPerty’ context, denote a whole made up of parts; and so a number is not a property
e number of figures of the syllogism. It1S n of any such composite object.

■
76 Frege: Philosophy ofMathematics 7 What is Number? 77
The colour blue, Frege remarks in § 22, belongs to a surface independently i it. but all that shows. is. that. we. ought
to c not to have taken the aggregate as
of any choice of ours ... ; our way of regarding it cannot make the slightest
die subject of the ascription m the first place. The subjectivist account is
difference’ to it. This, he says, constitutes ‘an essential difference between refuted by the observation that an ascription of number is as
suifficiently
L ‘
colour and number’. In this section, he hammers home the point that ‘our way objectively truie or false as any proposition can be. As Frege says in § 26, ‘the 1 ■'
of regarding’ something does make a difference to the number to be assigned
to it. This is essentially the same point as that concerning aggregates: it is not
enough to know how to delineate the totality to be numbered; it must also be
botanist means to say something just as factual when he states the number of
pel
tals of a flower as when he states their colour. The one depends as little on i'
■■ *

our arbitrary choice as does the other. There is, therefore, a certain similarity
known what it is to be regarded as being a totality' of. In Frege’s famous between number and colour; but it consists, not in their both being perceptible
example: -—■> in external things, but in their both being objective.’
by the senses
That is enough to resolve the question whether or not number is in part
If I give someone a stone with the words, ‘Find the weight of this’, I have thereby subjective;
given him the whole object of his investigation. But if I place a pile of playing subj-. - yet _ . devotes a good deal of space to discussing the matter in
. Frege
cards in his hands with die words, ‘Find the number of these’, he does not know §§ 26 and 27. He does so in order to take the opportunity to state his views
whether I wish to discover the number of cards, or of complete packs, or of on a topic essential to his philosophy of mathematics, and, indeed, to his i
complete suits. I have not yet completely given him the object of his investigation philosophy in general: the opposition between the objective and the subjective.
by putting the pile into his hands; I must add a w'ord - ‘cards’, ‘packs’ or ‘suits’. One of the most evident features of mathematics is its objectivity': the validity
or invalidity of its proofs and definitions is determinable to the satisfaction of
(The example has here been slightly altered: instead of ‘suits’, Frege wrote all. Sometimes, indeed, this objectivity' appears to be breached: disputes arise
‘point-values for Skat’.) Likewise, by adopting different ways of regarding it, over the legitimacy or otherwise of this or that method of proof. But such
‘I can conceive of the Iliad as a single poem, as twenty-four Books, or as a disputes are conducted on the assumption that they can be resolved, and are I1
large number of lines’. The point is repeated in § 25, in response to Mill’s not to be left as matters of taste; and, despite the still unreconciled schism
observation that two horses ‘are a different visible and tangible phenomenon’ between classical and constructive mathematicians, they usually are eventually
from one horse: ‘one pair of boots may be the same visible and tangible resolved. Unless, therefore, a philosopher of mathematics is content, with
phenomenon as two boots; here we have a difference of number to which no Brouwer, to deny this objective character, he must come to terms with it.
physical difference corresponds.’ Those who hold mathematical structures and mathematical objects to be •1*
mental constructions, as Dedekind did in Frege’s day, obviously have the
Subjective and objective hardest task in doing so. A mental construction is effected, in the first instance,
by a single individual, not by observable operations, but in the privacy of his
All this naturally leads to the suspicion that the number assigned depends own consciousness. It therefore becomes necessary, for one who takes this
upon some subjective way of conceiving the matter; not, indeed, on the part view, to maintain that such constructions can be communicated to others and
of those, like ourselves, who already know the answer, but on that of anyone faithfully reproduced in their consciousness, so that all may judge alike of ’ 5
who had not thought about the matter before, or had thought of it only ineptly, propositions relating to them and reasoning concerning them. Mathematics,
an w o was innocent of semantic theories based on syntactic analyses after then, belongs, according to such philosophers, to the realm of the intersubjec-
pattern of mathematical logic, which is to say of a Fregean logic. Frege is tive: that which is common to all, but owes its existence solely to our mental
in fact able in §25 to quote Berkeley as arriving at the conclusion that ‘number activity.
... is no mg xed and settled, really existing in things themselves’, but ‘is Frege’s view left no place for a category of the intersubjective, intermediate ♦
entirely the creature of the mind’. The illusion that the number to be ascribed between the wholly objective and the radically subjective. For him, if something
depends upon the subjective choice of a way of regarding the matter is due is common to or accessible by all, it must be independent of all; conversely,
°UF S^ected the wrong subject for the ascription, namely an something whose existence depends upon the consciousness of any one must
XT O' die like. Just because the aggregate does not, by itself, determine e private to that one, and not communicable to others. This is the primary
constitoen °™bd T" “ ” aggr'Sate of’ that is- whaI “ “ “unt as a si* source of his opposition to psychologism, the attempt to explain the meanings
conX of Z T tW° cons,itu“ts.« are driven, so long as * o the logical operators, or of mathematical propositions, or meaning in general,
conceive of the number as attaching to the aggregate to sunnose that some ln terms of internal mental operations. If the meaning of a proposition essen-
sub^cuve conception is needed to determine KSe ascrib'd
aly Solved anything interior to consciousness, then, according to Frege, it
 78 Frege: Philosophy ofMathematics 7 What is Number? ■

could not be conveyed, at least wholly or with assurance: we should have n0 We customarily recognise the colour by the sensation to which it gives rise in
way of knowing whether what had been aroused in the mind of the one to s. and this sensation is subjective. But Frege has already stated, earlier in
whom the attempted communication had been made was or was not the same Z section, that colour is an objective quality, so that a botanist’s description
as the original in the mind of the one trying to communicate it. Even if such of a flower as having petals of a certain colour is a factual statement. In so far -■ ■

a proposition could be communicated, there could be no common basis for as we use a colour-word in such a way that it can be decided to the agreement
determining its truth or falsity: if it appeared true to one and false to another, of all whether it applies to a given surface or not, we are using it in an objective
they could merely acknowledge their difference, but not resolve it. For that sense, to designate an objective quality, not a subjective sensation. Even a
reason, psychology must be barred from logic and from mathematics: the only colour-blind person can grasp this sense, because he knows what determines
.1
result of its intrusion would be the dissolution into inextricable subjectivity of the application of the word: criteria employed in the laboratory’, or simply the
what should be objective and the same for all. common response of those with normal vision.
.An analysis of a concept and a description of the psychological operations From the standpoint of Wittgenstein’s discussion of a private language, J '
necessary for attaining it are two quite different things. Only the former is Frege is committing the error of supposing that we attach both a subjective Ii
relevant to the justification or the proof of mathematical propositions: it can and an objective sense to the colour-word: we ordinarily judge it to apply in
be appealed to in a proof, whereas the psychological description is impotent the objective sense to a physical surface by recognising that it applies in the
to yield any mathematical conclusions. subjective sense to the visual sensation. If, then, there is no such subjective II
A distinction needs to be drawn which could not be drawn in Grundlagen sense, but only the objective one on w’hich Frege insists, is it irrelevant to that
in the absence of the sense/reference distinction. In Frege’s middle period, objective sense that colours are observational properties? Frege’s idea is that 1i
he always contrasts the senses of expressions, the thoughts expressed by the colour-blind man can grasp the objective sense of ‘red’ or ‘green’ as well
sentences, with contents of consciousness such as sensations and ideas. The as anyone else; and, if that is so, the observational character of the colour so ■

former are in themselves communicable and therefore objective, consequently designated does not enter into the sense of the word. The colour-blind man
not depending for their existence on our grasping or expressing them or understands the word ‘red’ in the way wre all understand ‘magnetic’: he judges
judging them true or false; the latter are subjective and hence essentially its application to an object by the effects that object has. Among these effects
incommunicable. In the late essay ‘Der Gedanke’, a breach is indeed made in I are those upon the normally sighted; but, for the colour-blind man, the nor­
the thesis of the communicability of senses; Frege there maintains that the mally sighted have the same bearing upon the sense of the word that iron
pronoun I, when used in soliloquy — not w’hen speaking to another — expresses filings do upon the sense of the w’ord ‘magnetic’. Since the objective sense
a sense that only the thinker can grasp. Before this, however, he always must be the same for all, the normally sighted cannot construe it in any other
contrasted senses with ideas as not being contents of consciousness and as way than the colour-blind; so the {objective} sense they attach to it takes account
accessible to all: in grasping a thought, the mind lays hold of what exists of their capacity to recognise the colour of something just by looking at it, but
independently of the process of grasping it, and may therefore be the same is indifferent to this being their own capacity.
or all who do grasp it. A sense which is in itself objective may, however, relate The implausibility7 of this view may be relieved by a distinction Frege never
to something subjective, as when we speak of our own sensations: the sense drew: between the sense a w’ord has in the language to which it belongs,
is o jective, but the referent is not. For it to be possible even to contradict an common to many speakers, and an individual speaker’s grasp of that sense, i
assertion, the thought denied to be true must be the same as that asserted, which may be erroneous and may, even when correct, rest upon some associ­
an recognisably so, hence, together with their constituent senses, all thoughts ation or ability peculiar to himself. According to the explanations he gave in
must be objective, and it is only they that can meaningfully be characterised his writings after 1891, the sense a speaker attaches to an expression consists i
as e or a se. ut, for the truth-value of a proposition to be assessable by in the way in which its reference is given to him. That explanation, in itself,
criteria common to all, the references of its components must themselves be would allow sense to be wholly subjective, different speakers attaching different
invnlvp'^a 1 lt,C0nce™s an object that only one person can apprehend, or senses to the same expression, and only the reference being the same for all. !
onlv th Pre 1C-ate wbose ^Plication only one person can determine, then or Frege, however, the sense of an expression is objective, being part of the I I
In GIT0" 1 ‘V P°sition t0 judge k t0 be false. ought communicated by’ a sentence containing it. This is because it is not
to an eZeSoVl hWeVe[’ applies Ae term <obiective senSe’ ■ ei}°ugh for communication that two speakers associate the same reference
was objective and d ° middIe period> he would say that its referent wtth an expression: they must know', or at least be able to determine, that they
, nd discusses colour-adjectives in the light of that distinction. do- It is likewise not enough that a proposition cannot in fact be true according
 80 Frege: Philosophy ofMathematics 7 What is Number? gj
to one speaker's understanding oi it, and false accoiding to another’s: each P rth’s axis, the Equator and the centre of mass of the solar system. The
must judge of its truth and falsity by the same criteria, and acknowledge the Fauator is in no way subjective, nor was it created by our thought; but it is
same reasoning to that proposition or its negation as conclusive. t causally efficacious, you cannot trip over it, or claim to feel that you are
The matter is nevertheless more complex. The use of a language in com. crossing it- One might conclude from these examples that something that is f;
munication requires that the speakers know the generally accepted criteria for not wirklich may yet be acted on, since the Earth s axis is subject to precession >■

determining the reference of an expression, and hence the truth-value of and nutation; but the conclusion should be drawn only with caution. In § 46 h
sentences in which it occurs. It does not follow that those criteria exhaust the Frege rejects the idea that the number of inhabitants of the German Reich
manner in which its reference is given to any one speaker, or even that they can change from year to year: it is merely that one number is the number of
I ;
always could do. The normally sighted user of a colour-word knows that his inhabitants of the German Reich at the beginning of 1882, and another number
unaided judgements about its application count, for others, as a defeasible that of the inhabitants of the German Reich at the beginning of 1883; he says
piece of evidence, and, as a speaker of the common language, is content the same in the essay ‘Was ist eine Function?’ of 1904. So perhaps there is
himself so to treat them. But his mastery' of the word is based upon his ability' no such point as the centre of mass of the solar system, which changes position
to recognise something as red or yellow just by looking at it; the sense of the as the planets move: only one point that has that status at one moment, and
word is given to him as the name of a colour with which he is familiar, and another that has it at another moment.
his knowing what a colour is rests upon his ability to match colours. However that may be, it is to the class of objects that are objective but not
In Grundlagen, Frege did not yet have the conception of the sense of an actual {wirklich) that numbers belong. The failure to recognise that something f
expression as the way in which its reference is given; the example of colour­ can be wholly objective without being actual leads to grievous errors. In
i words is discussed in simpler terms. It is cited, not for its own sake, but to the Preface to Grundgesetze, Frege stigmatises this failure as the root of the
emphasise that that which can be judged by common criteria, and that about psychologistic conception of mathematics: if numbers are not actual, as on
which it is possible to judge by common criteria, must be objective, and hence, Mill’s empiricist conception, they can only be, for one who makes this mistake,
according to his conception of objectivity, cannot be the product of human contents or creations of consciousness.1 The recognition that there is no reason
f
J mental activity. Since both ascriptions of number and arithmetical propositions
are judged by common criteria, numbers must therefore be objective in this
strong sense.
‘why what has an existence independent of anyone making judgements about
it must be actual {wirklich), that is, capable of directly or indirectly acting on
the senses’ is fundamental to Frege’s entire philosophy of mathematics.
§ 26 ends with Frege’s characterisation of objectivity as ‘independence from
Actual and non-actual our experience, intuition and imagination and from the delineation of inner
images from the memory of earlier experiences’. § 27 spells out the subjectivist
They are objective: but they are not ‘actual’. § 26 does not serve only to draw consequences of the psychologistic interpretation of numbers: if numbers are
the contrast between subjective and objective as Frege sees it; it also introduces ideas, each individual has his own arithmetic, and cannot dispute with anyone
a distinction, of great importance in his eyes, within the realm of the objective, else whose arithmetic differs from his. Here Frege inserts his footnote, already
that, namely, between what is and what is not actual {wirklich). In most contexts, discussed, complaining of Kant’s failure to distinguish subjective from objective
the German adjective ‘wirklich' is properly translated ‘real’; but Frege emphati­ ideas, and records his own decision to use the word ‘idea’ in an exclusively
cally does not use it to mean ‘real’ as opposed to ‘fictitious’ or the like. It subjective sense. The conclusion is that ‘number is neither spatial nor physical,
serves, ra er, as his manner of distinguishing between whaf present-day • •. nor subjective like ideas, but non-sensible and objective’. And, with that,
philosophers usually call ‘concrete’ and ‘abstract’ objects, though his contem­ Frege is ready to bring his critical preliminaries to a close: he has extracted
poraries were more given to speaking, ambiguously, about ‘real’ as opposed to several positive principles from the discussion; he is on the verge of giving his
l!np, Frege.does not use term ‘ideal’, but his use of‘wirklich' is in solution to the problem w'hat number is and what it is ascribed to.
in S b obiect is wirklich for him if it is a causal agent; He does not bring it to a close, however. Before giving his own positive
in § he speaks of restricting the actual {wirklich) to ‘that which acts upon account, he interposes a lengthy but brilliant section demolishing the widely ■

as theiHmm^-^ en^a^es *n actions which may have sense-perceptions avoured abstractionist theory of numbers as sets of units.
‘acts’ and ‘a remote consequences’; the German words here translated
‘actual’ Tfip COgnate with as the English words are with
actual. The examples cited in § 26 of objects which are not wirklich are the 'VoL I, p. xviii.
 . calls a ‘concept-word wouldand
8’, itUnits Concepts
stand for a n™ . 83
,bject whatever (§ 29): whereas
Frege P pertJ’ Possessed by evety
ol
Jt is only that
assertion in virtue obtains a that
wisepossibility
Solonofis the sompttr
sense. l
The: contend "Ot ** W,’se the •i :
CHAPTER 8 as its extension mcreases; if the latter becomes all/mk 3 ConcePt diminishes
vanish altogether. mes aJ,-embracmg, its content must
I
Units and Concepts It is even worse to suggest, with Baumann, that whether something is one or
many depends on our way of regarding it (§ 30). In fact, we cannot admit such
a sentence as ‘Solon was one’ as well-formed, save when understood as
supplemented by some concept-word supplied by the context; and, if we did,
Numbers as sets of units we could not, from ‘Solon was one’ and ‘Thales was one’, infer ‘Solon and
I In the brief § 28 of Grundlagen, Frege rejects the conception of a number as
a set or plurality, using the now familiar objection that it fits the numbers 0
Thales were one’.
It is useless to attempt to explain ‘one’ as standing for such a property as
I j,
I. 1
and 1 particularly badly. That the number of objects of a given kind should that of being circumscribed, self-contained or undivided. ‘When we say that
be the set of those objects is sufficiently absurd to need no refuting: the section the Earth has one moon, we do not mean to specify that the Moon is circum­
serves principally to introduce Frege’s attack upon the widespread conception, scribed, self-contained or undivided’ (§ 32); for the moons of Jupiter are as
■ common, as we have seen, to Husserl and Cantor, among many others, of a circumscribed, self-contained and undivided as our own. If the word ‘one’
number as a set of featureless units. According to this theory, starting with a denoted a property of this kind, we should expect even animals to have some
set of objects to be counted, we abstract from all the specific properties of idea of it; but it is improbable that a dog ‘has even a dim awareness of the
those objects, thus mentally fashioning a purely abstract set whose members common element in the cases in which he is bitten by one larger dog and in
- units - have no properties whatever; that is the number of objects in the which he chases one cat’ (§ 31). Still less does it mean ‘thought of as undivided
original set. Plainly, if we had started with any other set containing just as or indivisible’; this makes the application of ‘one’ subjective once more,
many' objects as the first, we should by this means arrive at the same abstract whereas truth cannot be attained by thinking of things as they are not (§ 33).
set of units: such abstract sets thus have the essential characteristic we desire It follows that the notion is not, as Locke supposed, ‘suggested to the under­
of numbers, that the number belonging to any' set is the very same as that standing by every object without us, and every’ idea within’, but that it is,
belonging to any other set that has, as we say, the same number of members. rather, attained ‘by means of those higher mental powers that distinguish us
from the beasts’.
What does ‘one’ mean?
Abstractionism .
does b §§ 29 t0 44> t0 refute: and he ’
from § 29 to § 33 does no/d 1e®n*t*ve^>1 The first part of his discussion, The word ‘unit’ cannot, therefore, be explained as ^^t^j^eword ‘one’
of their possessing the property' denoted by e wor ^„der) Hobbes and
mention abstraction, but treat ea.Wld? dle full-blown theory, and does not
denotes no property. Frege quotes various strictly
as a ‘multitude of units’ r° i-j,° va&uer characterisations of a number (
Hume) as maintaining that units are to b^conS’ come about, since, t-
to. What, then, is a unit? Is it a UC I- S de^n’h°n, which he begins by alluding
are, identical with one another (§ 34). How' j:stjnct? Frege quotes his
denoted a property of objects dlat ‘S one,? ®ut die adjective ‘one’
Plainly, the actual objects to be counted must e . as means by
'»»* ».4Id, F ncti°"cd “ whaI’from 538 onmrds’ colleague at Jena, J. Thomae, as appealing o . one of a
Kz*:!1!?’1-21’ WE. °f •S-nwdoniM Am? .re 'Ober * . t
which the identity is achieved: he might, of cours , culiarities of the
des Herm Thomae’ (X Sdhuben d 899) and pp sVK ?°s,hunwus PP' 72jl1: i number of contemporary writers. We ‘abstrac r te ^ngs, disregard
substantial to his rtf. 76)’ second of these l °f auf die Ferienphud'fy' I individuals in a set of objects’ and, ‘in const enn , frOin one another’,
S refutat,on it in CnlndlTgen 7 hea'?' sarcasmi but none adds anjlh * i
die characteristics by which those things are is n i

82
■
 84 Frege: Philosophy ofMathematics 8 Units and Concepts 8$
Frege does not here launch any general attack upon the whole idea of abstrM. '
non8 he staph- remarks that such a process mil not vic d either a set of n„it
“r the number of things in the set, but only 'a general concept under whic ^emadcs. His essential, and crucial, contention in Z Aat
toe things fall’; 'the things themselves do not thereby lose any of ft* abstraction is (at best) a means of coming to grasp certain genefd *

peculiarities’. as a mental operation, it has no power to create abstract objects o

structures. Subsequently, in the review of Husserl, published in 1894
If for example, in considering a white cat and a black cat, I disregard the rejected abstraction altogether. That is, he rejected not only thesis
properties which distinguish them, I obtain, say, the concept cat. If I now bring thesis (2) as well; his words are consistent with his continuing to accent L
them both under this concept, and call them units, the white one still remains (1), but nothing he says hangs on that and there is no reason to assumf t^T
white all the same, and the black one black. The cats do not become colourless
as a result of my not thinking of their colours; they remain just as different as ai„ believed it. Thus he wrote, with hear, sarcasm,
they were before^ for all my resolring to draw no conclusions from their difference, example: 6
The concept ‘cat’, which has been attained by abstraction, indeed no longer
includes the peculiarities of either; but, just for that reason, it is a single concept.
f eX SoS ta £r»X»L"'r
The latter is particularly effective. We take less notice oUpX J’it
It is fatuous to base on this remark, and other similar ones in Grundlagen, ' vanishes. By causing one charactensttc after another t0 «nlsh. £ ™ '
F an onslaught on Frege as one of the chief nineteenth-century proponents of i
abstractionism, as do G. Baker and P. Hacker.2 Frege was one of the earliest ones abstract
more than concepts.
objects; Concepts,
they have too, are
only those thereforefrom
properties merely
ideaswhich we less
hXe ?
and most vigorous opponents of the doctrine of abstraction, at a time when it abstracted. Inattention is a highly effective logical toe h “ ,£
absent-mindedness of scholars. Let us suppose for example, there are stog sSe
was generally taken for granted among philosophers: if his attack was still not
by side in front of us a black and a white cat. We nay no attention tn 5,(16
as sweeping as it could or ought to have been, that is no ground for criticising dtey become colouriess, but are still sitting side ^sto win^X°“o
him as an arch-abstractionist. To the extent that he admitted the notion of
abstraction, that notion plays no role in any of the philosophical theses he was tfieir posture, they are no longer sitting, without, however, assuming a different
concerned to advance; it is no part of his argument in § 34 that we do arrive P.®,re’ but e,ach !5 st,H in 1116 same position. We cease to attend to their places-
haw Jl C°me dkeV0'd °f p0SItlon> but continue to be apart from one another We
at a concept by this means, but only that the most we could so attain would be ha thus, perhaps, attained from each of them a general concept of a cat
a concept. Such a criticism is not merely fatuous: it misses the point. We may I

consider three theses, in ascending order of strength:

Whereas in Grundlagen Frege had allowed that the concept ‘cat’ might be
attained in this way, here he mocks even that application of the operation' oi
(1) it is possible to attain a radically new concept by contemplating a
number of otherwise diverse objects falling under it; abstraction, by repeated application of which, he says, every o jec is
formed into an ever more bloodless ghost’:
(2) the attainment of the new concept in case (1) is effected by abstracting
from the properties differentiating the objects in question, i.e. by diverting Whereas, on my view, bringing an object under a concept is m®reb r66
of a relation which already obtained beforehand, here the o je come mor’e
the attention from them; altered thereby, so that the objects brought under the same concept become
similar to one another.
(3) the operation of abstraction referred to under (2) can also generate
obipc^rth11^13! c°nstrucbons> that is, abstract objects or structures of
placed th°Se Pr°perties abstracted from and have no others
leading proponent.

■Frege: Logical Excavations, Oxford, 1984, pp. 57-8.

3 Review
of Husserl, pp. 316-17.
 8 Units and Concepts gy
86 Frege: Philosophy ofMathematics
jered impossible (§ 40). In any case the manoeuvre is of no avail;
Are units identical? m°rC if points of space, for instance, differ only in their relations to one
,Of’ |C'Cnthev still differ, so that numerically equivalent sets of points may still
In unlike the review of Husserl, Frege expends little energy „„
contesting die whole idea of abstraction as a means of constructing abstract sn -te distinct (§ 41). Jevons proposes to regard the distinction between one
entities by mental operations, and devotes little attention to the process by
be ql" j another as consisting merely in ‘the empty form of difference’ (§ 44),
which a set of units is generated; instead, he concentrates on the product, that f to say that they differ, but not in any respect. Even if that were
is, on the nature of the set of units. He poses a fundamental dilemma for the ""lC table, it still would not help: if the units differ from one another at all,
conception of numbers as abstract sets of featureless units, namely whether the ^there may be distinct but numerically equivalent sets of units, and the
units are identical with or distinct from one another. The point of interpreting ^"ose of die entire theory is frustrated. The difficulty7 becomes particularly
numbers as sets of units, rather than taking the number of objects of a given Pcute when we consider the addition and subtraction of numbers. If numbers
kind simply to be the set consisting of those objects themselves, is obviously arc sets of units, it lies to hand to interpret addition as set union and subtraction
to guarantee that the number will be independent of the particular objects as set difference. Now7 numbers are specific entities (§ 38):
counted, being determined, as it ought to be, solely by how many of those
objects there are: if, say, there are just as many spoons as forks on the table, \\e speak of ‘the number one’, and indicate by means of the definite article a
/ the number of spoons on the table will be the very' same abstract entity' as the single, determinate object of scientific enquiry. There are not distinct number
/ ones, but only a single one. In 1 we have a proper name, and, as such, it is as
number of forks on the table. This requires that the set of units arrived at by incapable of a plural as ‘Frederick the Great’ or ‘the chemical element gold’ ...
abstraction from the set of spoons shall be the very same set of units as that Only concept-wwds can form a plural.
arrived at by abstraction from the set of forks. It seems to be possible to
guarantee this only if no trace of individuality7 is retained by the units: those Hence, if units differ from one another, and numbers are sets of units, each
derived by abstraction from the spoons must be identical with those derived number must be a specific set of particular units. Either any two numbers are
by abstraction from the forks. This can be so only if the operation of abstraction disjoint, or (as Jevons appears to suppose) larger numbers include smaller ones.
strips the original objects of all their properties: for the spoons, or, rather, In either case, it becomes impossible to explain how addition or subtraction can
what had originally been spoons, could hardly retain features differentiating yield the right set of units to constitute the number that we know7 results from
them from one another if they could no longer be differentiated from w7hat the arithmetical operation. Without a loophole for escape, Frege has shown
had been forks. That is one reason w'hy Frege is able to quote so many waiters how, even without calling in question the magical operation of abstraction, one
as maintaining the strict identity of all units with each other.
may reduce the entire theory7 to ruins.
The conclusion poses an obvious difficulty’ for the theory7; for, as Frege says,
we cannot succeed in making different things identical simply7 by operations
with concepts; but, if we did, wre should no longer have things, but only a The solution
single thing’ (§ 35). And he quotes W.S. Jevons as saying, ‘It has often been With § 44 Frege’s critique of untenable views comes to a triumphant end
said that units are units in respect of being perfectly similar to one another; (though there is to be a supplementary7 critique of postulationism in
t^ou^ they may be perfectly similar in some respects, they must be
§§ 92-103). It remains to provide positive answers to the outstanding ques
different in at least one point, otherwise they would be incapable of plurality’,
dons, namely:
every unit is identical with any (other) unit, there can only be one unit. As
Frege summarises the problem (§ 39), ‘if we try to make the number originate
Mat is a number a number of, that is, w'hat is number ascribed to?
• J”. 6 k°m inatl,on distinct objects, we obtain an agglomeration compris-
not theOnb|eCh ’USt tJ°Se properties differentiate them; and that is
Mat is (cardinal) number in general?

Mat, specifically, are the numbers 0 and 1?

Mat is the relation between a natural number and its successor.

Wtat,on of number to what is non-spatial and non-temporal is «« n § 46, the celebrated answer to the first of these questions is gi
88 Frege: Philosophy ofMathematics
1
8 Units and Concepts 89
L„ 7 eine Aussage von einern Begriffe? Annoyingly, this terse concept of itself performs the only function of gathering them together or of
fomtS difficult to translate. A ‘ZMmfM is what we have been calling
s singling them out that is needed (§ 48). Of course, to obtain a determinate
notation of number’, namely a sentence beg.nntng w.th an express.™ „f ! number, we must consider, not the general concept copse or tree, unless w'e are
™ form‘There are «... ’ (where ‘«’ represents a number-adjective). Angabe' wanting to give the number of all copses or trees in the universe: the relevant
Lore accurately rendered ‘statement’ than ‘ascription’; the latter has been
concept is that of a copse or a tree in such-and-such a place at such-and-
preferred here as more readily comprehensible. To translate the verb as such a time (§ 46).
‘contains’ which is what it literally means, is misleading, carrying as it does i Frege’s notion of a concept might seem at first like the traditional notion
the connotation ‘contains among other things’: Frege merely means to indicate of a universal, which can be predicated of particulars, but can also have
what, according to him, is the content of such a form of words. Accordingly, something predicated of it; but the two notions differ crucially. On the tra­
we may begin our rendering of the slogan by ‘The content of an ascription of ditional conception, if we are to predicate anything of the universal, the
number is ... ’. How, then, is ‘eine Aussage von einem Begriffe' to be translated? 'i predicate which expresses it when it is predicated of particulars must be
Austin’s version, ‘an assertion about a concept, imports a suggestion that is
not present, and omits one that is. By Aussage', Frege does not mean to single
i transformed into an abstract singular term: but Frege’s idea is that an ordinary
form of sentence, in which the concept-word occurs in predicative form, as a
A
out assertions, as opposed to questions, commands, etc.; Friedrich Waismann, general term, plural or singular, may serve to predicate something of the
misled, apparendy, by the word (which can indeed be used to mean ‘assertion’), concept for which it stands. Frege’s notion is very much broader than the
actually criticised Frege, quite erroneously, for restricting his account of num­ 11 traditional notion of a universal. The general form of a judgeable content
ber-words to assertions, and not providing for their use in interrogative and which treats of an object a is la falls under the concept F’, he tells us in § 70.
imperative sentences.4 Frege uses Aussage’ here to mean ‘predication’: he Taken together with the Begriffsschri.fi doctrine of the extraction of concepts
intends to convey that, in an ascription of number, something is predicated of from judgeable contents, which allows any proper name or other singular term,
a concept, in analogy with the sense in which the sentence ‘Julius Caesar I occurring anywhere in a sentence, and, when it occurs more than once, at any
was ruthless’ is used to predicate something of the individual, Julius Caesar. i* selected number of places, to be regarded as variable, this yields a very wide
Unfortunately, the English phrase ‘a predication of a concept’ is most naturally conception of what is to count as standing for a concept, embracing far more
understood as applying to the act of predicating a concept of some object. The than can be represented by any abstract noun-phrase.
best we can do, therefore, is to render the slogan as ‘The content of an Frege’s doctrine concerning concepts does not fully emerge in Grundlagen-,
ascription of number consists in predicating something of a concept’. we have to go to the writings of his middle period to grasp it properly.
Certainly, some elaborations were new: he would not have said, at the time of
So, then, there is Frege’s solution: what a number is ascribed to is a concept.
When we regarded it as ascribed to a complex, an aggregate, it seemed that s Grundlagen, that concepts were functions from objects to truth-values, and is
e number to be ascribed depended on our subjective way of regarding it: as t unlikely to have thought of them, as Baker and Hacker suppose, as functions
I
one copse, or as five trees; as four companies, or as five hundred men. But | from objects to judgeable contents; nor, of course, did he distinguish between
ere is nothing subjective about it: it is the concept, copse or tree, company or concepts and the senses of concept-words. For all that, much of what he said
C t?Ve ,invo^e *n tbe ascription of number, that determines objectively about them after 1890 must already have been present in his mind when
as nhiprri111 U F°r a concept is not a subjective idea, but is Grundlagen was written. An adherent of the traditional doctrine of universals
we niediZ^ an °bJeCt: i"dePendendy of anything we may think or imagine, and particulars would certainly agree that we cannot understand an abstract
something of ^/L3 concept as or as falsely as we predicate ! term for a universal unless we take it as standing for something that can be
it is to be snlit ° a^Kregate does not of itself determine how predicated of particulars; the role of the universal as predicate is prior to its
j role as subject. It was not part of that traditional doctrine, however, to hold
unambiguous^
of^Sd Jafi unSoTL “ S° is the concept that determines
Thk components: AU that
that any intelligible predication of some characteristic to a universal must be
able to be expressed by a sentence in which the universal is represented by a
can be numbered pn ?e concept Or another: hence objects of all kinds
be no physical relationth*111^ t0 be ascribed t0 a ^ven concept, there need
if predicative expression or general term. It would be generally agreed that the
need to^Xt^nL oTri d° analogue does hold for mathematical functions. We can intelligibly say, e.g.,
(
‘The sine function is everywhere differentiable’, only because the same thought
‘ Einfih^ i operations upon our ideas of those objects: the
can be expressed by a sentence invoking only the functor ‘sin ()’, containing

I an argument-place. It is for this reason that it is impossible to say of a function

90 Freget Philosophy ofMathematics 8 Units and Concepts 91
what can meaningfully be said of a number, or conversely; and Frege was transitive’ (applied to a function or a relation). Certainly no one, not even
certainly of the same view in respect of concepts and objects. He later went Frege, has attempted to observe it when using natural language. But Frege’s
much further, indeed, denying that such an apparent singular term as ‘the sine purpose - admittedly obscured by the use of the jargon - was to analyse
function’ was well-formed at all. Only an incomplete or unsaturated’ sentences - expressions of judgeable contents - as they stand. When we predi­
expression - one with an argument-place - could stand for a function or a cate something of a concept, we do not need, and in fact ought not, to
concept; hence a functor or concept-word ought never to appear without its transform the expression for the concept into the grammatical subject. Such
argument-place. a transformation is indeed effected by Frege’s jargon, which serves to empha­
This of course led to the paradox which Frege attempted to dismiss, and sise that a predication is involved by presenting it in the form most familiar to
failed to resolve, in ‘Uber Begrift und Gegenstand, to the effect that the us for predications, namely that in which the subject of the predication is
concept horse cannot be a concept, since the phrase the concept horse, as a grammatically a singular term: but the point of the jargon is not to assimilate
singular noun-phrase with the definite article, can only’ stand for an object. sentences predicating something of a concept to ones predicating something
From the remark already quoted from Grundlagen, § 70, that la falls under the of an object, but to highlight the analogy between them, despite the obvious
concept F’ is the general form of a proposition about an object, it is apparent difference of verbal form and the difference of logical level, on which Frege
that Frege was as yet oblivious of this difficulty. He has in mind sentences insists as firmly as on the analog}'. It is possible to make such second-level
representable symbolically as ‘F(rt)’, which, according to his doctrine, include predications only because our language, as we ordinarily employ it, allows us
all sentences containing a singular term, at least when viewed in a particular to form sentences embodying those predications, in which the expression for
way. He emphatically does not mean that such sentences invoke a relation, the concept remains predicative in form, and hence clearly recognisable as an
that of falling under, obtaining between objects and concepts. The concept is expression for a concept. It was Frege’s clear simultaneous recognition of the
predicative by nature, requiring an argument for completion, and hence of analogy between ‘The Danube is long’ and ‘There are five sheep in the
itself couples, as it were, with the object to make a complete judgeable content. meadow’ and of the difference of level between them that enabled him to
It is sometimes asserted that the metaphor of saturation wras not used by Frege introduce a type unknown to Kant, that of properties of concepts, or, better,
before 1891: in fact, though it does not occur in Grundlagen itself, it can be concepts of second level or order (Grundlagen, §§46 and 53). We ascribe
found in his letter to Anton Marty', written in 1882 w'hen the book was nearly properties to (first-level) concepts, as well as ascribing properties to objects;
finished, where he says: but the kind of property' a concept can have is utterly different from a property'
possessed by an object, and expressed by a form of words differing radically
A concept is unsaturated, in that it requires something that falls under it; hence
it cannot subsist by itself alone. That an individual falls under it is a judgeable in logical character. If we observe the injunction never to use, as an expression
content, within which the concept appears as the predicate, and is always predicat­ for a concept, one that is complete and lacks an argument-place, then it
ive. In this case, in which the subject is an individual, the relation of subject and becomes impossible even to try to say of a concept what can be said of an
predicate is not a third thing added to the two of them, but belongs to the content object, or conversely.
of the predicate, which is what makes the latter unsaturated.
It is the usual practice nowadays to describe the semantic value of a one-
The denial that the relation of subject to predicate, or of object to concept, is place predicate as being a set, viz. a subset of the domain, rather than as an
a third ingredient seems to be contradicted by Frege’s representation of, say, unsaturated entity like a concept in Frege’s sense. What is here meant by a
‘set’ is what Frege normally called a ‘class’; except occasionally in late writings
concenter38 T IthaC3’ “ °f IthaCa 11131 U falls Under 1116 dating from the time when the phrase ‘set theory’ had become standard, Frege
concept place where Odysseus was set ashore’; but the point of that represen-
always used ‘set’ to mean an aggregate made up out of individuals — what
to make ScitVl *“7° One Possible anal^s of that sentence, not
same holdS fn nT- alluded t0‘ As alreadY o^rved, the partisans of the calculus of individuals call a ‘sum’. A set, in this sense, has
parts, rather than members, and the part-whole relation is transitive; there
concepts and the like. C Grundlasen about numbers belonging to therefore cannot be such a thing as an empty' set, and there can be no
like‘^Sinefunction’and’ultimately><the^pt distinction between an object and a set consisting wholly of that object.5 A
class, on the other hand, is for Frege the extension of a concept:6 it is the
excessively severe- ft w/hT °f functl0ns and of concepts - may be thought
concept which determines what is and what is not a member of the extension,
predicates masquerading as firsZl^el o' a,.ban on second-level
ones, such as ‘is continuous’ or ‘is 5 See the review of Schroder (1895), pp. 433-7.
6 Ibid., p. 455.
92 Frege: Philosophy of Mathematics 8 Units and Concepts 93
and thus mav rule out from membership the member of a member. As with certain objects fall under it and others do not. Class-membership is a relation
numbers, only a concept can articulate the class into its members in a determi­ between objects which ought to be expressly mentioned whenever it is invoked.
nate way. It is because a concept can be empty that there is such a thing as The relation between an object and a concept which obtains when the former
the empty class; and it is because the extension of a concept need not itself falls under the latter, on the other hand, is what we might call a formal relation
fall under that concept that we can distinguish a unit class from its sole (in analogy with Wittgenstein’s notion of a formal concept): it does not need
member. to be, and is not, invoked in an ordinary predication, which is why there is no
An extension of a concept is not, indeed, the same thing as the concept. relation that fails to be symbolised in the form ‘F(«)’, and why an explanation
This is not because concepts are intensional rather than extensional. Frege of the truth-grounds of such a proposition in terms of concepts is preferable
identified himself with extensionalist logicians, as against the intensionalist to one in terms of classes.
ones,7 and declared that the relation between concepts analogous to that of Fregean semantics undermines the superficial similarity between singular
identity’ between objects was that of being co-extensive.8 The analog}' can only and plural. A grammatically singular noun-phrase may of course be functioning
lie in the principle of extensionality: whatever holds good of a given concept as a concept-word, and will always be so functioning when preceded by the
must hold good of any co-extensive one. The difference is, rather, one of indefinite article (Grundlagen, § 51): but a plural noun-phrase, even when
level: classes are objects, and, as such, are denoted by singular terms. They preceded by the definite article, cannot be functioning analogously to a singular
can therefore themselves fall under first-level concepts, and hence belong to term. There are, of course, complex objects; but their continued existence
the extensions of such concepts. But the notion of a concept is prior to that depends on the maintenance of some relation between their components.9 If
of the extension of a concept: we can only form the notion of a class via that the troops cease to obey their officers, the army dissolves, with not a man lost;
of a concept which determines what is and what is not a member of it; a class if the bicycle falls to bits, there is no longer a bicycle, though all the parts are
can be given only as the extension of a concept. That is why, in the formal there; even a pile of dust ceases to be when the wind scatters the particles.
system of Grundgesetze, Frege takes as primitive, not, as in our conventional But a plural subject of predication or ascription cannot stand for any such
systems of set theory, the symbol for the membership relation, but that for composite object, both because it presupposes no relation between the objects
class abstraction. (Strictly speaking, it is a more general abstraction operator, alluded to, and because it determines which those objects are in a way In
forming expressions for value-ranges - functions in extension; but the general­ which no composite object is uniquely articulable into components. There is
isation may be passed over for the present.) Membership is, in Grundgesetze, no such thing as a ‘plurality’, which is the misbegotten invention of a faulty
a defined relation: a is a member of b if there is some concept F such that a logic: it is only as referring to a concept that a plural phrase can be understood,
falls under F and b is the extension of F. Here is one of the many places where
because only a concept-word admits a plural. But to say that it refers to a
Frege found second-order quantification indispensable. Thus, from Frege’s,
concept is to say that, under a correct analysis, the phrase is seen to figure
surely correct, viewpoint, we do not dispense with the notion of a concept by
predicatively. Thus ‘All whales are mammals’, correcdy analysed, has the form
making the semantic value (reference) of a predicate a class: we still need the
‘If anything is a whale, it is a mammal’, and ‘The Kaiser’s carriage is drawn
concept to explain what a class is and what it is to belong to it. If we explain
by four horses’ the form ‘There are four objects each of which is a horse
a sentence of the form *F(a)’ as being true just in case the object denoted by
a belongs to the class associated with ‘F’, we are merely saying that the object I that draws the Kaiser's carriage’ (§§ 47 and 46). On this analysis, no one has
tails under some concept with whose extension ‘F’ is associated: we could subsequently found an improvement, the only plausible variation being that
which would substitute, say, ‘any organism’ for ‘anything’ in the first and
y ‘TT 1116 concept directly 1116 Predicate, and say that the
bec^L th Uher th?'OnCept Indeed’ we shaU do better to say the latter, ‘organisms’ or ‘animals’ for ‘objects’ in the second, importing an explicit
tit J XeZatOb!i CtpWkCh constitutes extension does not really enter into circumscription of the domain into the quantifications.
whose two narfc fi/t nngin^Jt *n> we obscure the unity of the proposition,
Z/Sve bv „a„ ,08t'I?eCaUSe were made •» d° »• A concept is
predicative by nature, which is to say that it is constituted by the fact that

Bedeuiung', apropos of Husserl's rentt’ofstbrod un^bllshed ‘Ausfiihrungen uber Sinn und 9 In his letter to Russell of 28 July 1902, Frege discusses this notion of a composite object,
Wntmgs, p. 122. f Schroder. Naehgelassene Schriften, p. 133, Posthumous under the name of a ‘whole or system’, using just this example of an army; see G. Frege,
“ See the review of Husserl, p. 320. Wissenschaftlicher Briefoechsel, ed. G. Gabriel, H. Hermes, F. Kambartel, C. Thiel and A. Veraart,
Hamburg, 1976, p. 222, or the English translation by Hans Kaal, Philosophical and Mathematical
Correspondence, ed. B. McGuinness, Oxford, 1980, p. 140.
 Frege: Philosophy ofMathematics 8 Units and Concepts 95
94
relevant concept; there is no difficulty in counting the rectangles in some given
The number of red things diagram. But the application ol the predicate ‘red’ does not presuppose or
There is however, a grave defect in Frege’s answer to the question ‘To what depend on any prior articulation of the candidates for its application into
is a number ascribed?’. At the beginning of the 1885 lecture ‘Uber formale discrete (even if overlapping) objects. This fact is in turn due to the existence
Theorien der Arithmetik’, he said that ‘nothing is really required’ for a number of a level of language below and prior to that which can be regimented by the
to be assigned to things of a certain kind ‘save a certain sharpness of circum­ syntax of predicate logic, a level at which there is no reference to objects, but
scription, a certain logical completeness’. In Grundlagen, § 54, he recognises only ostension: when, pointing to some surface, I say, ‘This is red’, I am not
that not all concepts have this feature. The counter-example he there cites is referring to an object, because I have, and need, no criterion for recognising
the concept ‘red’. Such a concept does not determine what is to count as a what I point to as the same again (§ 62), or for determining whether what
single object falling under it; the totality of red things in the room is, in this someone else is pointing to at the same moment is or is not the same thing
respect, no better than an aggregate. Asked to count the red tilings in the as that at which I was pointing. That does not mean, indeed, that my statement
room, we do not know how to begin: if, say, the wallpaper pattern has a red cannot be contradicted. But, if you say, ‘You are wrong: that is not red’, the
background, are we to count the background as one red thing, or the connected first question to be settled between us is not whether you are pointing at the
red regions on the wallpaper as a large number of red things, or what? If a same thing (red or otherwise) as I was, but whether what you are pointing at
red-headed man is present, is the hair of his head one red thing, and each is the same colour as what I was pointing at; if it is, you were genuinely
eyebrow another, and so on, or should we count the individual hairs? We are contradicting me, and, if it is not, our statements can be reconciled. A concept
in a greater difficulty than when handed a pile of playing cards with the like ‘somnolent’ or ‘spherical’, by contrast, is applied only to determinate
instruction, ‘Count these’. Frege’s response is inadequate: he says that ‘to objects, and thus does not belong to the most primitive level of language, but
such a concept no finite number belongs’; but the fact is that there is no to its second level, which is representable by a quantificational syntax; and it
infinite number, either. Quite unlike the question, ‘How many natural numbers can therefore be intelligibly asked how many somnolent or spherical things
are there?’, the question, ‘How many red things are there in this room?’ has there are in a room at a given time. It is only to such concepts that numbers
no answer, at least if there is anything red in the room; even to cite a transfinite can be ascribed. A predicate like ‘red’, belonging to the most primitive part
number in answer to a question ‘How many?’, the concept requires ‘a certain of language, can still be used at the next level, and applied to determinate
sharpness of circumscription’. Frege is at fault, not merely in giving no objects; but, because it does not presuppose any articulation of reality into
adequate response to the difficulty' in Grundlagen, but in never reverting to a separate objects, we cannot intelligibly ask what number attaches to it.
matter that called in question his solution to the problem to what a number
is, in general, to be ascribed.
Aggregates and concepts
The fact is that he had not the resources to resolve the difficulty within the
framework of his logic. It is not the vagueness of the predicate ‘red’ that causes Bell’s defence of Husserl’s Philosophie der Arithmetik against Frege’s criticisms
the trouble. Assume it to be being used only of opaque surfaces, and its sense interested us in Chapter 2 because it helped to illuminate the issue of psycho­
to have been so sharpened that, of every' such surface, it is determinate whether logism. It also bears on Frege’s doctrine that it is to concepts that numbers
or not it is re all over, still there is no saying how many red things there are attach. In his book,10 Bell goes so far as to defend, against Frege’s criticism
in the room. And yet, given Frege’s definition of ‘the number of Fs\ it is in his review,11 Husserl’s contention that 0 is not a number, on the ground
nnnX- t0 dlat> ^or any concept F, there is a determinate that ‘nought’ is a negative answer to the question ‘How many?’;12 he balks,
eiZr rM °b’eCtS mg Under iL By h^0thesis> each object is, determinately, though, at concurring with Husserl’s denial of that status to 1, on an analogous
numL dfi°7 r : Sc° What can Stand in Ae way of *ere being a specific i
ground. Bell misses Frege’s point, however; both he and Husserl have the
ThT dhficulw a°r tranfsfinite\ whjch is ^t of ail the objects which are red? matter the wrong wray round. If we deny that there is such a number as 0, we
semantic^fca 1Z fr°m ?eJact *at a ^gean semantics - a classical can and must hold ‘none’ to be a negative answer to ‘How many?’. The same
X " *e T\CtUre °f Standard (higher-order) predi- goes for 1 and for Aleph-0, vis-a-vis the answers ‘one’ and ‘denumerably
objec/some objects Zh’11 already detemunately articulated into individual many’; the scholastics held it to be a priori impossible that the human race
be part of aZge^^ °^eCtS’ as one rectangle may
10 D. Bell, Husserl, London, 1990, p. 70.
11 Review of Husserl, pp. 327-8.
12 E. Husserl, Philosophie der Arithmetik, Halle, 1891, p. 144.
96 Frege: Philosophy ofMathematics 8 Units and Concepts 91
had always existed, because there would then be no number that was the of tliem.,',> Despite this, his explanations of the notion of an aggregate are
number of all men there had ever been. But Frege’s point was that, by treating tantalisingly inadequate; this is due especially to his concentrating in the first
0 and 1 (and Aleph-0) as numbers, we run into none of the antinomies that instance on aggregates of objects of perception. It is hard to discover which
result from treating ‘never’ as the name of a time or ‘nobody’ as the name of features aggregates are supposed to retain in a more general context, and
a person. We therefore do not have to hold that ‘none’ is a negative answer, hence whether Bell’s observations about the notion are faithful to Husserl’s
and so cannot infer, from the premiss that it is, that 0 is not a number. There conception or not. Even supposing them to be so, it is natural to wonder if
are cases in which we should run into antinomies. One known to Frege is the any difference is left between an aggregate and a concept. Bell insists that an
number of red things in the room; another, not known to him, is the number aggregate is of a more concrete character than a concept or its extension, as
of cardinal numbers. Frege thinks of it: a suitable aggregate can be photographed or travel by
Bell believes - falsely, as we saw in Chapter 2 - that Husserl, having aeroplane.20 Since he seems to nurture a degree of nominalist prejudice, this
explained the genesis of the concept of number, had a further explanation to predisposes him in favour of the aggregate.
give, namely of what the numbers are, and he attempts, on Husserl’s behalf, The obvious reply, from Frege’s standpoint, is that it is the members of the
to reconstruct it; but the attempt yields no clear explanation. Bell boldly claims aggregate, the objects falling under the concept, that are photographed; or, if
superiority for Husserl’s notion of an aggregate (Inbegrijf) or plurality (Vielheit) the point is that they are photographed together, we are talking, not about an
over Frege’s notion of a concept (Begrijf); according to him, Husserl’s use of aggregate, but a system, whose cohesion requires that certain relations obtain
the former notion enabled him to close the lacuna in Frege’s semantic theory, between its components.21 Bell’s answer is that this objection springs from the
its inability to treat plural terms in analog}' with singular ones.13 For the notion Fregean blindness to plural terms. Statements about aggregates are simply
of an aggregate to perform the task assigned to it, however, the features which statements with plural subjects, and fall into two types: those that ascribe a
distinguish it from that of a concept, as understood by Frege, and on which distributive property, and those that ascribe a collective one.22 A distributive
its superiority to the latter might be claimed to rest, are progressively jettisoned property holds good of the individual members of the aggregate, while a
by Bell in favour of the features of concepts. Thus although an aggregate is collective property does not. This is Frege’s distinction between saying that
a whole made up of parts,14 and the part-whole relation is transitive, its the tree has green leaves, and saying that it has 1,000 leaves;23 but Bell has a
articulation into members is determinate15 — just the very feature that makes particular reason for holding that the ascription of a collective property relates
Frege think it essential to take a concept as that to which a number attaches; to an aggregate, not to a concept. For a statement about an aggregate never
a member of an aggregate is a part of it, but not all its parts are members. makes ‘ineliminable reference to a single thing called an aggregate’, but is
Again, the existence of an aggregate does not, as Frege supposes, depend on always a statement about several individual objects at once. The difference
any relation’s objectively obtaining between its members.16 Frege, discussing between predicating a distributive property and predicating a collective one is
Grund!agen’ § 23’ had assumed, surely rightly, that the constituents of that the distributive property is itself ascribed to the members of the aggregate,
aii ian aggregate or agglomeration must be related by physical proximity; whereas the collective property is not one that individual objects can possess;
but BeU correctly explains that, for Husserl, the unity of an aggregate arises but a collective predication can still be analysed as a statement about the
purely from a psychological act of ‘collective combination’.17 members of the aggregate, but as one ascribing some other property to them.
Although Husserl allows that membership of an aggregate is frequently
What is here meant by ‘a statement about the members’? From the examples
m \en7Td by P°ssession of a common Property, and that it is always possible
given by Bell, it is apparent that, in such a statement, the predicate expressing
XXS 18\C°nCept U"der Which faU aU and only members of a given
membership of the aggregate appears only as applying to individual members,
tha^agscrintiohne<!1SfeniPhktlC WaS WrOng and he is riSht in thinking
therefor^ ? of number relate to aggregates, not to concepts: 'The number ” Philosophic derArithmetik, p. 185; emphasis Husserl’s. On pp. 188-9 he argues that, in Frege’s
herefore attaches, not to the concept of the objects being counted, but to the aggregate example, the number four attaches, not to the concept ‘horse that draws the Kaiser s carriage , but
to its extension. It is possible only with the hindsight afforded by the writings of Frege’s middle
period to perceive clearly what is wrong with this objection, namely that Frege was not, like
13 Op. cit., pp, 63ff. Husserl, an intensional logician, and that his notion of a concept is already an extensional one;
14 Op. cit., p. 38. but Husserl is also mistaken, at least on Frege’s view, in equating, as he does, the notion of the
15 Ibid., p. 78. extension of a concept with his own notion of an aggregate.
“ Ibid., pp, 50-1. I 20 Bell, op. cit., p. 65.
21 See note 9.
22 This citation from Bell’s book, and those immediately following, are from pp. 68-9.
23 Grundlagen, § 22.

!
98 Frege: Philosophy ofMathematics
never with a bound variable in its argument-place. It is, however, evident that
a transformation of the original statement about die aggregate into one that
is by this criterion, about its members will be possible only when we know
which individual members the aggregate has; but, since this knowledge is not
part of the content of the original statement, such a transformation, even when CHAPTER 9
possible, is not an analysis.
This is very clear from Bell’s attempt to make out his claim that Husserl’s
notion of an aggregate is a better tool for analysing die concept of number
than Frese's notion of a concept. He proposes to explain such an ascription of
number as 'There are at least three coins in the fountain’ as meaning ‘F(n)
Two Strategies ofAnalysis
& F(/>) & F(r) & a * b & b * c & a * c', where, he says, ‘ “F” designates
the property of being a coin in the fountain, and “a”, “A and d are arbitrary
names’. He proclaims this analysis to be superior to Frege’s, on the ground
A false start
that it involves ‘only objects and their properties’, and does not invoke abstract
entities like concepts. It is obscure why the property of being a coin in the Rejecting formalism, Frege acknowledges that mathematical propositions have
fountain should be regarded as less abstract than the concept coin in the a content, rendering them true or false; his aim in Grundlagen is to make
fountain. Bell appears temporarily to have forgotten that the use of ‘concept’ explicit the content both of ascriptions of number and of the propositions
in Frege’s terminology is only the consequence of his step-by-step adaptation encountered in number theory. Number-words occur in two forms: as adjec­
of that of Kant, that the two-argument analogue of a concept is a relation, tives, as in ascriptions of number, and as nouns, as in most number-theoretic
and that Frege equated the concepts under which an object falls with its propositions. When they function as nouns, they are singular terms, not admit­
properties.24 In any case the analysis collapses when we ask what ‘a’, <’b' and ting a plural; Frege tacitly assumes that any sentence in which they occur as
‘ri name if there happen not to be any coins in the fountain. Plainly, the adjectives may be transformed either into an ascription of number - a sentence
analysis results from ignoring the distinction between saying that there are at beginning ■with ‘There are’, followed by a number-adjective - or into a more
least three coins in the fountain and seeing that there are. It is precisely complex sentence containing an ascription of number as a constituent part.
because Husserl was preoccupied with cases in which we have all the individual Plainly, any analysis must display the connection between these two uses: it
members of an aggregate in view, or at least in mind, that, in struggling to would not do to give separate explanations of number-adjectives and of
understand his notion of an aggregate, one is in danger of losing one’s grip numerals functioning as terms, without protiding for any explicit relation
on that otherwise endent distinction. Bell, in endeavouring to vindicate the between them. Otherwise, we should be unable to appeal to the equation
notion, comes close to assimilating it to the Fregean notion of a concept; but, ‘5 + 2 + 0 = 7’ to justify inferring that there were seven animals in the field
in trying to preserve some difference between them, he finally unfits it for the from the fact that there were five sheep, two cows and no other animals
work of being that to which a cardinality is attributed in any ascription of there. Evidently, there are two alternative strategies. We may first explain the
number. As employed by Husserl in the Philosophic der Arithmetic the notion
adjectival use of number-words, and then explain the corresponding numerical
hvn Kegate 1S s*mply ^coherent, and remains so even after being explained terms by reference to it: this we may call the adjectival strategy. Or, conversely,
of mimhHUlSer n°t-SiVe US “ alternative t0 Frege’s account of ascriptions
of number, but only an irremediable confusion. we may explain the use of numerals as singular terms, and then explain the
corresponding number-adjectives by reference to it; this we may call the
substantival strategy. A radical version of the adjectival strategy would be to
refuse to take numerical terms at face-value. According to this strategy, equa­
tions and other arithmetical statements in which numerals apparently figure
! as singular terms are to be explained, not merely in terms of adjectival uses of
number-words, but by transforming them into sentences in which number­
words occur only adjectivally. On this view, numerals are only spurious singular
24 See ‘Ober Begriff und GegensUnd’, p. 201. terms, when they apparently function as such: equations and other number-
i
)
theoretic sentences have a grammatical structure that belies their true logical
99
 i
100 Frege: Philosophy ofMathematics 9 Two Strategies ofAnalysis 101
structure, which is revealed only when they have been transformed into ver­ In § 56, Frege rejects these definitions, saying that they ‘suggest themselves
sions containing only number-adjectives. The analogously radical version of so spontaneously in the light of our previous results that an explanation is
the substantival strategy is, on the face of it, less attractive. According to it, it needed why they cannot satisfy us’. He begins on the third definition, saying
is ascriptions of number and other sentences in which number-words occur that it will be the first to arouse qualms. He agrees that, by appeal to this
adjectivally that have a misleading surface form: they contain hidden references definition and the second one, it will be possible to explain the expression:
to the objects denoted by numerical singular terms.
In § 55, Frege essays the adjectival strategy, which, as he says, suggests the number 1 + 1 belongs to the concept F,
itself very’ naturally. That is to say, he experiments with three definitions of
expressions involving the adjectival use of number-w’ords, stated without appeal and hence also the expression:
to any prior definitions of numerical notions, and, in particular, without appeal­
ing to the use of numerical singular terms. In stating them, he makes heavy the number 1 + 1 + 1 belongs to the concept F,
use of his jargon. Instead of saying, ‘There is just 1 F’, or ‘There is just 1
object which is F’, or even ‘There is just 1 object falling under the concept and so on; that is, expressed without the jargon, to explain first ‘there are just
F’, he says, ‘The number 1 belongs to the concept F’. This, of course, obscures 1 + 1 Fs’ and then ‘there are just 1 + 1 + 1 Fs’, and so on. That is to say,
the fact that these are adjectival uses of number-words that he is defining; for, the third definition serves as a pattern for constructing further definitions, in
in his jargon, the number-words precisely appear as singular terms - ‘the the sequence ‘there are just 2’, ‘there are just 3’, and so forth: but Frege does
number 1’, ‘the number O’, ‘the number (s + 1)’ and so forth. When we strip not consider it to be in itself a proper definition. His objection to it is that
the terminology’ of the jargon, writing ‘There are just n Fs' for ‘The number ‘strictly speaking, the sense of the expression “the number n belongs to the
n belongs to the concept F’, and ‘a is an F’ for ‘<i falls under the concept F’, concept G” ’ - that is, of the expression ‘there are just n Gs’ - ‘is just as
the definitions assume the following forms: unknown to us as that of the expression “the number (n + 1) belongs to the
concept F” ’ - that is, of the expression ‘there are just (m + 1) Fs’. He states
‘There are 0 Fs’ is to mean that:
this objection very badly, in the notoriously inept words ‘we can never ...
for all x, x is not an F;
decide by means of our definitions whether the number Julius Caesar belongs
to a concept, whether that famous conqueror of Gaul is a number or not’. No
‘There is just 1 F’ is to mean that:
one reading the book for the first time can have seen this remark as making
it is not the case that, for all x, x is not an F, and, for all x and y, if x
much sense, let alone as relevant. Nevertheless, Frege’s objection to the third
is an F and y is an F, then x = y;
of the proposed definitions is perfectly sound.
There are just (n + 1) Fs’ is to mean that: He comes more to the point when he says:
for some x, x is an F, and there are just n objects distinct from x which
are rs. It is only an illusion that we have defined 0 and 1; in reality we have only fixed
the sense of the phrases
If we use the notation ‘3„/ t the number 0 belongs to
to mean ‘there are just n x’s such that’, and
similarly for other indices, we may write these as follows: the number 1 belongs to;

‘3n x Fx" means that V.v -> Fx; but we are not licensed to pick out the 0 and the 1 in them as self-subsistent objects
I' recognisable as the same again.
I
‘31 x Fx' means that V.V -n Fx & VxVj (Fx 8c Fy x = j);
In §§ 56-61, Frege is attempting to establish that arithmetical statements
‘3.L1X Fx’means that 3 x [Fx & 3 , j/ (Fj & x # j)]. containing numerals must be taken at face-value, and hence that the radical
adjectival strategy is not feasible. Ontologically expressed, he is trying to
Symbols like ‘3? and ‘3 ’ e establish that numbers must be regarded as objects. If the radical adjectival
m logical literature as ‘numerically USUaHy kn°Wn strategy could be made successful, they could not be so regarded: apparent
singular terms for numbers would be spurious, and the only admissible uses
i
IQ2 Frege: Philosophy of Alathematiis 9 Two Strategies ofAnalysis 103
of number-words would be as adjectives, overt or in disguise. Instead of certainly not entitled to replace a part of each by a variable, any more than we
regarding (cardinal) numbers as obiccts, we should have to be concerned only can replace all but the common parts of the words ‘cowl’, ‘cowrie’, ‘coward’
with the’ corresponding second-level concepts denoted by esprcss.ons l,ke and ‘cowed’ by a variable. The point is very clearly put by Frege, in a similar
•there is just one’, 'there are just three’, ‘there are denumerably many, and connection, in the Appendix concerning Russell’s paradox which he added to
so on It is for this reason that the passage irom § 55 to § 61 is ot such great Volume II of Grundgesetze. In the body of the book, he has of course defined
importance: its purpose is to prove the crucial thesis that numbers are obiccts. cardinal numbers as classes; here he is discussing the possibility of ‘regarding
It is not immediately apparent what the precise content ol this thesis is. but class-names as pseudo-proper names, which would therefore in fact have no
it certainly involves that numerical terms have to be taken at face-value, and reference’. His comment on this proposal is:
cannot be explained away. Hence Frege’s anxiety to convict the definitions he
proposed in § 55 of being logically amiss. They would then have to be regarded as parts of symbols that would have a
In the comment about picking out 0 and 1 as self-subsistent objects, how­ reference only as a whole. One may indeed consider it advantageous, for some
ever, he is assuming what he is trying to prove: a valid objection to the third purpose or other, to construct different symbols so that certain parts of them
of the proposed definitions is entangled with a petitio principii. Anyone with coincide, without thereby making them into complex ones. The simplicity of a
symbol demands only that the parts that one may distinguish within them do not
his wits about him, who sees nothing wrong with the definitions of § 55, will have a reference on their own. Even what we are accustomed to construe as a
reply that he was not in the least concerned to pick out 0 and 1 as self­ numeral would in such a case not really be a symbol at all, but only an inseparable
subsistent objects, that he knows that the phrases ‘the number 0 belongs to’ part of a symbol. A definition of the symbol ‘2’ would thus be impossible; one
and ‘the number 1 belongs to’ - better, ‘there are 0’ and ‘there is just 1’ - would have instead to define several signs that contained ‘2’ as an inseparable
have been defined only as a whole, and that there is no licence to extract the constituent, but were not to be thought of as composed of ‘2’ and some other
part. It would then be illicit to let such an inseparable part be replaced by a letter
numerals ‘0’ and T’ from them and treat them as having a sense independently [by this, Frege means a variable]; for as regards the content, there would be no
of the wholes of which they are part: and he will add that he was not intending complexity'. The generality' of arithmetical propositions would thereby be lost.
to do so. Frege himself view's number-adjectives as essentially occurring only
in the context ‘there are just... since he believes that to explain ascriptions In just the same way, the expressions ‘there are just two’, ‘there are just three’,
of number is thereby to explain all adjectival uses of number-words. Hence, and so on, contain a common part, as do the symbols ‘32’, ‘3? and the rest:
if ‘0’ and ‘T are to be given independent senses for use in other contexts, but, when each has been defined only as a w'hole, their possession of a common
those can only be contexts in which they function grammatically as singular
part is logically without significance, and w’e cannot retain that common part,
terms, at least in Frege’s eyes. Now no one pursuing a radical adjectival while replacing the remaining constituent by a variable.
strategy' will wish to recognise any such context as genuine; and so he will be
This is a cogent objection to the third definition as formulated in Grundlagen,
unmoved by Frege’s denial to him of the right to envisage ‘0’ and ‘1’ as
having independent senses. Someone pursuing an adjectival strategy' with a less § 55: but it presents no obstacle to the general strategy’. The remedy lies to
ambitious goal will not hope to eliminate numerical singular terms, but only hand: the variable must stand not in place of a part of one of the defined
to explain them: but he thinks the w'ay to do this is by first explaining the symbols, but in place of the w'hole symbol. Each such symbol stands for a
corresponding numerically definite quantifiers, and then to explain the numeri­ concept of second level, and so the variable will range over second-level
cal terms by appeal to them. He, too, will be unmoved by Frege’s protest: he concepts: we have to define an operator which transforms one quantifier
does not want to remove ‘0’ and ‘1’ from the contexts in which alone they (expression for a second-level concept) into another, and w'hich may therefore
ave so far been explained: he merely wants to work up to explaining them be applied, not merely to the numerically definite quantifiers, but to all quanti­
for other contexts. 6 fiers. This is difficult to express in words, but easy to write symbolically. If we
use the symbol<+’ to represent our operator, the emended definition will take
i” ” 56 and 57.’ he had demonstrated the definitions
defectonlvin tV .a- a faulty; in fact, however, he had pointed out a the form:
d'finW0nS' The and Sc“"d definitions
conStXE T "um'rals ‘°’ “d the ‘M+a Fv’ is to mean that 3v [Fx & M, (Fy & x * j')].
variable When the nb ' k°n
whe««been defined’ “Erese r,ghtl>'
Prec'se'y d13'. by replacing them with a
as
1 The outcome of this will be that *3i+xF.v’ has the force of‘There are just 2
es, we can attach a sense to each only as a whole: we are therefore Fs\ ‘3,*.v Fx' that of ‘There are just 3 Fs’, and so on, just as desired, while
‘3 x Fx' will mean ‘There are at least 2 Fs, ‘3'*x FxJ will mean ‘There are at

»
 I

104 Frege: Philosophy ofMathematics 9 Two Strategies ofAnalysis 105

least 3 Fs’, and so on. Again, if we take ‘Hr Er’ as an abbreviation for Sf4 (3i y Fy, Gx)
‘Vr (x is a man -> Er)’, then ‘Hr 0x’ will stand for that concept of second
level under which fall all first-level concepts under which all men fall. The would say that there were just two Gs (i.e. two objects falling under G), and
formula ‘H+ xFx’ will then say that all men fall under the concept F, together we could define:
with at least one other object that is not a man. As for 'V' x Fx’, it says that
there fall under the concept F all the objects there are, and one more besides, 32x Gx Sfa (31 y Fy, Gx).
and thus cannot ever be true; but this is no objection to our definition of the
operator Likewise, we could write:
In point of fact, however, even that definition is unsatisfactory from the
standpoint of good notation, and, for the same reason, conflicts with the Sf,v [Vj ()' is a man Fy), Gx]
syntactic and semantic principles of Grundgesetze. As for notation, ‘+’ suffers
from being applicable only to a single quantifier-symbol like ‘3’ or ‘V’, whereas to mean that all men, and at least one other object, fell under the concept G:
there also exist what may be called complex quantifiers, in the broad sense of the notation would allow us to avoid having first to define the special quantifier
the term ‘quantifier’ relevant to present purposes; ‘Vx (x is a man —> <Px)’ was ‘Hr 0x’.
an example of such a complex quantifier. That is why, before applying it In fact, there is an easier way to go about it. Instead of defining a successor
was necessary to define a single symbol ‘H’ as equivalent to the complex operation, represented by <+’ or ‘S’, we may define the binary quantifier ‘The
quantifier. Such a necessity points to a defect in notation. Fs consist of the Gs and one other object’ (i.e. ‘Everything falling under G
The operator *+’ transforms a quantifier into another quantifier: it therefore falls under F, and just one object falls under F but not under G’). This, like
stands for a function from second-level concepts to second-level concepts. ‘Most Fs are Gs’ and ‘There are just as many Fs as Gs’, stands for a relation
Yet, on Frege’s principles, as elaborated in Grundgesetze and other works of between first-level concepts. If wre write it as ‘D, (0x, IFx)’, we may readily
the middle period, the only admissible functions are those which have objects define it thus:
or truth-values as values. The only admissible operators, under this principle,
are those having sufficiently many argument-places that, when they are all Dx (Fx, Gx) 3x [-> Gx Sc Vj (Fy Gyvy = x)]. i
filled, either a singular term or a complete sentence results. The reason is that
an incomplete expression can be arrived at only by extracting it from a complete We may then define ‘S2 x 0.x1 by:
one. To comply with this doctrine, we need an operator with two argument­
places, one admitting an expression for a second-level concept (a simple or 32Er 3G [3| x Gx & Z)A (Fx, Gx)].
complex quantifier), and the other admitting one for a first-level concept.
reat notational cumbrousness results, however, from employing an operator The definition is of second order, requiring quantification over first-level
o s type, while observing Frege’s requirement that every expression shall concepts; but Frege had, of course, no objection to that.
V'* ,tS Ue ar^ment-places, filled, if necessary, by bound vari­
ables. Our binary operator would then take the form:
Frege’s sleight of hand
Sfa. (=■_), Fy, 0x), The upshot of all this is that Frege was right to object to the proposed
definition of ‘There are just (« + !)..,. ’, but had no case against those of
Here 7” and ‘x’ represent the variables bound by ‘5’ while and ‘There are 0 ... ’ and ‘There is just 1 ... ’; and, moreover, that the third
definition, though unsound, can be corrected without affecting its essential
spirit. Frege aimed, how’ever, at much more than an emendation of one
defective definition: he aimed to show all three definitions erroneous, and

As a result,
S* (M, Fy, Gx) 3x [Gx & M, (y * x & Gy)].
i thereby to prove a radical adjectival strategy unfeasible, because numbers have
to be recognised as being objects. In this, he utterly failed, in fact, § 56 may
be stigmatised as the weakest in the whole of Grundlagen. The arguments lack
all cogency: they more resemble sleight of hand. This is not to suggest that
106 Frege: Philosophy ofMathematics 9 Two Strategies ofAnalysis 107
they were deliberately designed to take in the reader: rather, Frege, impelled where the prime symbol denotes the successor operation. When we wished,
by his desire to establish that numbers are objects, seems to have been taken however, to state the analogue of the proposition
in by his own jargon. When ‘There are 0 Fs’ and ‘There is just 1 F’ are
expressed as ‘The number 0 belongs to the concept F’ and ‘The number 1 n is the number of Fs & m is the number of Fs -» n = m,
belongs to the concept F’, it looks more plausible to complain that the defi­
nitions do not entitle us to pick out 0 and 1 as self-subsistent objects; without we should need to take a little care. Where ‘N’ and ‘iM’ are free variables for
the jargon, it would have been apparent that they were not meant to and did quantifiers, ranging over second-level concepts, we cannot simply write:
not need to.
The same holds good for the complaint that (1) N, Fx & M.v Fx VG (Nx Gx Mr Gx),

W e cannot prove, by means of the putative definitions, that, if the number a since it obviously is not true: there are at least nine planets, and there are
belongs to the concept F and the number b also belongs to it, then necessarily fewer than a hundred planets, but the second-level concepts denoted by ‘there
a = b. There would then be no way to justify the expression 'the number which are at least nine’ and by ‘there are fewer than a hundred’ are far from being
belongs to the concept F\ and it would be altogether impossible to prove a (
numerical equation, since we should not be able to get hold of any determinate co-extensive. We need a means to restrict the generalisation to those second- I
number. level concepts which correspond to cardinal numbers, namely those under
which fall precisely those first-level concepts having some specific cardinality'.
Identity’, as Frege understood it, is a relation between objects: the complaint But Frege offers not the slightest reason for thinking that this cannot be done;
therefore assumes what has yet to be proved, that numbers are objects. If they in fact, no obstacle whatever stands in the way of doing it. Obviously, it will
are to be so regarded, then indeed we shall have to be able to express the be possible only if we have available the notion of cardinal equivalence, which
relation holding between a number n and a concept F when there are just n Frege later defines in terms of one-one mappings, expressed in natural lan­
objects falling under F, and to prove that, for any concept F, there is only one guage by sentences of the form ‘There are just as many Fs as Gs’; we may’
such number; and certainly the proposed definitions do not suffice for this. write it symbolically as the binary’ quantifier ‘=.v (0x, ’Fx)’. A quantifier
But Frege has supplied no ghost of an argument for supposing that they cannot ‘Mv 0.v’ then serves to answer the question ‘How many?’ if the following holds
be suitably supplemented. good generally:
It is possible to interpret Frege’s complaint in another way, not directly
involving reference to numbers as objects. According to the principle stated (2) Mx Fx & « (Fx, Gx) Ma G.r.
subsequently by Frege, the analogue, for first-level concepts F and G, of the
relation of identity’ between objects is co-extensiveness, as expressed by: Formula (2) will hold good of quantifiers expressing such notions as ‘there
are less than a hundred ... ‘There are infinitely many ... ‘There is an
Vr (Fx Gx). odd number of ... ’ and so forth; the more important condition is that the
quantifier should serve to give a definite answer to the question ‘How many?’,
Presumably the analogue mil hold for second-level concepts. For example to which it •will do if the stronger condition:
express dial what corresponds to identity holds between the s"ZX'v‘d
“S' may ™ ° d“«'d b> ‘There (3) Mv Fx Kv (Fx, Gx) M.v G.r]

holds generally. Formula 1 may then be proved to hold for all quantifiers I
VF (3,FxFx 31 xFx); ‘M, 0.v’ and ‘N, <J»x’ that satisfy’ (3) for all F and G.
Admittedly, Frege’s complaint relates, not to the definitions in their amended
form (that is, with the amended version of the third one), but to the original
^\XXZmb“i “ ~ ~ " definitions as formulated in § 55; the idea of employing general variables for
second-level concepts has therefore not been introduced. But, even so, were
the third definition not objectionable on other grounds, there would be reason
0' = 1,
to introduce such variables, and no obstacle to doing so; and, by their means,
 I

108 Frege: Philosophy of^Mathematics 9 Two Strategies ofAnalysis 109

i part of the second-level predicate; but no argument has yet been given why
it would be possible to devise a way of restricting the quantifiers considered
in the formula (1) to those in the sequence we should want to separate it, let alone construe it as a numerical term.
Frege continues by urging that, in number-theoretic statements and equa­
‘3o.v <Px\ ‘3, x <Px\ ‘32.v 0.v’, tions, number-words occur, for the most part, in substantival form, as singular
terms; we have therefore ‘to regard the concept of number in such a wray that ■

that is to say, to those assigning a natural number as the cardinality of a it can be used in science’. He then has to answer the question how, if we are
concept. In fact, given Frege’s subsequent definition of finite number, it to treat number-words as ‘proper names’, we can construe their use as adjec­
would be easy: all that would be needed would be the analogue of that tives. He replies that this is easily done. ‘The sentence “Jupiter has four
definition in the adjectival mode. That, indeed, would involve quantification moons”, for example, can be converted into “The number of Jupiter’s moons
over third-level concepts. This would be inexpressible in natural language, is four” ’, where the word ‘four’ functions as a proper name of the number
and the notation would thereby become exceedingly cumbrous - given always 4; the transformed sentence has the form of an identity-statement. Frege’s
the requirement that all argument-places be explicidy visible; but there would solution of the problem of relating the adjectival and substantival uses of
be no conceptual difficulty. The reader to whom the topic is new cannot be number-words to one another is thus, apparently, to treat the adjectival uses
expected to perceive this. He may well be persuaded by Frege that there is a as disguised forms of the substantival ones. A sentence like ‘Jupiter has four
difficulty in principle, simply because he himself cannot see the way out of a moons’ or ‘There are four moons of Jupiter’ does not appear, from its surface
merely apparent difficulty. This, together with the confusing effect of his form, to contain any reference to the number four, regarded as an abstract
jargon, is why Frege’s argument in §§ 56-7 has the character of sleight of (non-actual) object. Nevertheless, according to the analysis Frege here sug­
hand. gests, its surface form is misleading: w'hen its deep structure is uncovered,
Apart from the valid point about the third definition, Frege’s arguments in it can be recognised as really being a statement of identity between the
§ 56, however charitably interpreted, at most point out that the proposed object denoted by ‘the number of Jupiter’s moons’ and that denoted by ‘the I
!
definitions do not accomplish all that we need, and that massive supplement­ number 4’.
ation will be necessary; but this should in any case be obvious. Whether the The contention has a high degree of implausibility. Worse, its acceptance
supplementation is conceived as involving the substantival use of numerical undermines Frege’s appeal to the surface forms of number-theoretic sentences.
terms, or as shunning it in favour of an exclusively adjectival use, Frege says If it is legitimate for analysis so to violate surface appearance as to find in
nothing to show that we could not build upon the base provided by the first sentences containing a number-adjective a disguised reference to a number
two proposed definitions and an emended version of the third. considered as an object, it wrould necessarily be equally legitimate, if it were
possible, to construe number-theoretic sentences as only appearing to contain
singular terms for numbers, but as representable, under a correct analysis of
How Frege handles number-adjectives
their hidden underlying structure, by sentences in which number-words
In § 57 Frege explains that ‘when, in the proposition “the number 0 belongs occurred only adjectivally. The possibility is that aimed at by the radical
to the concept F”, we consider the concept F as the real subject, 0 is only an adjectival strategy, which, for all his rhetoric, Frege has done nothing, in
element in the predicate’. For this reason, he says, §§ 56-7, to prove unfeasible. If the appeal to surface form, in sentences of
natural language, is not decisive, then it cannot be decisive, either, when
I have avoided calling a number like 0, 1 or 2 a property of a concept The applied to sentences of number theory. Frege has merely expressed a prefer­
’ “"'•“bsis,em <**“■> precisely >• • mere ence for the substantival strategy, and indicated a means of carrying it out: he
has in no way shown the adjectival strategy impossible, as he is purporting to
have done.
i
fct"! pr'varica,ira »f § 56- Considered as an
r
identified with
identified with aa second
second-level i °b’eCt c.a",be I
numbers are to be re„.d,d as objX Some defensive moves i
are to be regarded as
i In the remainder of this passage, §§ 58—61, Frege does no more than defend
number-word, but, when phrased in a natural wav.
the thesis that numbers are objects against objections, reiterating that an object t
may be objective but yet not actual, nor, in particular, spatial, and observing
 I
I
110 Frege: Philosophy ofMathematics
that even actual objects may not be capable of being imagined, and so, in his
sense of the word ‘idea’, may have no idea associated with them, or, at best, j
an irrelevant or manifestly inadequate one. The impossibility of forming an
idea of its content is no ground for denying meaning to a word, lor meaning
is not constituted by ideas or mental images. At this place (§ 60), Frege invokes CHAPTER 10
the context principle, stated in the Introduction.

We must always keep a complete sentence before our eyes. Only in it do the
words really have a meaning. The inner images which may hover before us do
not necessarily correspond to the logical constituents of the judgement. It is Frege's Strategy
enough if the sentence as a whole has a sense; it is through this that the parts
obtain their content also.

Whatever may be thought of the context principle thus strongly enunciated, The linguistic turn
Frege’s general contentions concerning objects, in §§ 58-61, are evidently
quite sound. They successfully defend the thesis that numbers are objects The highly unsatisfactory' passage from § 55 to § 61 of Grundlagen is followed
against fallacious objections; but they do nothing to establish that thesis, so by the most brilliant and philosophically fruitful in the book, and the most
crucial for Frege’s philosophy of arithmetic, and it remains, at the end of § 61, important for Frege’s philosophy of mathematics, and, indeed, his philosophy
wholly devoid of cogent justification. generally. It extends from § 62 to § 69, and is highly significant, not merely
for the understanding of Frege’s own philosophy of mathematics, but for the
philosophy of mathematics in general.
Having made, in § 55, what, in §§ 56-7, he then rejected as a false start,
Frege now adopts a fresh strategy. Of these inspired sections, § 62 is arguably
the most pregnant philosophical paragraph ever written. It does not merely
introduce the important notion of a criterion of identity', considered as associ­
ated with any proper name or other singular term: it is the very' first example
of w'hat has become known as the ‘linguistic turn’ in philosophy. Frege’s
Grundlagen may justly be called the first w'ork of analytical philosophy.
After § 61, Frege assumes that he has shown that numbers are objects, and
must be treated as such. Since they are objects, he begins his new enquiry by
posing the Kantian question, ‘How are numbers given to us?’. Kant’s doctrine
was, of course, that objects can be given only through sensible intuition. Frege
has, however, already rejected the notion that number is any kind of perceptible
feature of things, or that numbers are objects of w hich we can have intuitions.
i
The problem is therefore an acute one, particularly for anyone influenced by'
Kant, as few philosophers were not at the time when Frege wras writing.
J
Jr • His solution was to invoke the context principle: only in the context of a
r
sentence does a word have meaning. On the strength of this, Frege converts
} i the problem into an enquiry' how' the senses of sentences containing terms for
numbers are to be fixed. There is the linguistic turn. The context principle is
stated as an explicitly linguistic one, a principle concerning the meanings of
! words and their occurrence in sentences; and so an epistemological problem,
with ontological overtones, is by its means converted into one about the
meanings of sentences. The context principle could have been given a non- i

111
112 Frege: Philosophy ofMathematics I 10 Frege's Strategy H3

linguistic formulation. It would then have said that we do not conceive of as he explains in § 65, where he speaks of the law as a ‘definition’ of identity.
objects save as ingredients in states of affairs, or that we cannot apprehend an In his middle period, he rated identity indefinable, on the ground that every
object save in the course of recognising something as holding good of it. But definition must take the form of a identity-statement, generalised or otherwise.
Frege gave it, from the outset, a linguistic formulation; and so, when he comes Nevertheless, the reply that, since wre know what the sign of identity means, J
to invoke it, he makes the linguistic turn. He offers no justification for making we have to define the operator ‘the number of 0s’, would at this stage be
it, considers no objection to it and essays no defence of it: he simply executes quite unnatural on the part of anyone who has gone along with Frege so far.
the manoeuvre as if there were no novelty to it, and does it so skilfully that The doctrine of criteria of identity involves that we shall explain terms of the
the reader scarcely perceives the novelty. \ et it was in fact unprecedented in form ‘the number of Fs by explaining when two such terms denote the same
the history of philosophy. Plenty of philosophers — Aristotle, for example — number, or, in other words, when an identity-statement connecting them is
had asked linguistic questions, and returned linguistic answers: Frege was the true. If so, it wrould hardly be reasonable to propose explaining that by first
first to ask a wow-linguistic question and return a linguistic answer. If it wrere defining the operator, ‘the number of 0s’, which we may call the ‘cardinality
on the strength of Grundlagen, § 62 and its sequel alone, he would still deserve operator’: for that would appear to render the doctrine wholly nugatory’. Rather,
to be rated the grandfather of analytical philosophy. we have to lay down the truth-conditions of statements of the form
!
(1) the number of Fs = the number of Gs
Criteria of identity
The principle of criteria of identity’ enunciated in § 62 states that: in some non-question-begging way.
The notion of a criterion of identity, which Frege introduced into philo­
If the symbol a is to designate an object for us, we must have a criterion that will
in even' case decide whether b is the same as a, even if it is not always within
sophy, has been widely employed by analytical philosophers in recent decades;
our power to apply this criterion. but it is far from simple to explain, and Frege himself, who never mentioned
it again after Grundlagen, provided little help. A criterion of identity for Cs
Accordingly, numerical terms must be provided with a criterion of identity'; ought not in general to be equated with a necessary and sufficient condition
and this means that there must be a determinate, non-circular condition for for the truth of identity-statements connecting terms for Cs, although it fre­
the truth of an identity-statement connecting them. Frege does not at this quently is. Such an interpretation would lead us to say that the criterion of
point discuss the meanings of numerical equations, however. Instead, he tacitly identity for countries is that they should have the same boundaries: for the
assumes that the fundamental type of terms standing for numbers consists of truth of Tran and Persia are the same country’ stands or falls with that of Tran
those of the form ‘the number of Fs’, or, in his jargon, ‘the number belonging and Persia have the same boundaries’. But this is a criterion we can apply only
when we already know what ‘Iran’ and ‘Persia’ denote: it will not help us to i
to the concept F’, without attempting to justify this choice. The choice is,
after all, natural, given the demand for a criterion of identity. If we w'ere asked decide whether Edinburgh and Birmingham, or Kiev and Moscow', are or are !
for a criterion of identity for distances, we should not explain how to judge not in the same country. The same mistake is, I think, involved in Davidson’s
the truth of statements of the form T metre = 39.37 inches’: we should claim that having the same causes and the same effects is the criterion of
explain what determined whether the distance between P and £was the same f identity for events,1 and, far more disastrously, in Ayers’s claim that spatio­ i
as that between R and S. So the question becomes how to specify the condition temporal continuity is the criterion of identity for material objects:2 we cannot
for the truth of a statement of the form ‘The number of Fs is the same as the know the causes and effects of an event until we know what that event
number of Gs’. comprises, and we cannot trace the path of an object through space unless we
The sequence of thought leading up to this question makes it utterly unnatu- are already able to identify that object from one moment to another. On the
rep y at is point that, since we already know' the meaning of ‘is the contrary, the criterion of identity for objects of a given sort is something we
’ f at,lsnceded 1S t0 spec*fy 016 meaning of a term of the form ‘the
number of Fs. As a matter of fact, that is precisely what, in § 68 Frege 1 Donald Davidson, in his ‘The Individuation of Events’, originally published in N. Rescher I
(ed.), Essays in Honor of Carl G. Hempel, Dordrecht, 1969, reprinted in D. Davidson, Essays on
he hTelf a!Jy Aaion and Events, Oxford, 1980, pp. 163-80, says, ‘We have not yet found a clearly acceptable
amng of is the same as, namely as given by Leibniz’s law: criterion for the identity of events. Does one exist? I believe it does, and it is this: events are
identical if and only if they have exactly the same causes and effects’ (p. 179).
X=y VF (Fx 2 Michael Ayers, ‘Individuals without Sortals’, Canadian Journal of Philosophy, vol. IX, 1 74,
Fy), PP-113-48.

i
 I i
I

114 Frege: Philosophy of Alathetnatiis 10 Frege's Strategy H5

have to learn before we are in a position to know what a term tor an object able; and yet it is only his skill in steering his readers in just the direction in
of that sort denotes. It must therefore be thought ol as determining the which he wants them to go that prevents us from being amazed at the step he
condition for the truth of identity-statements connecting terms whose use has is taking. The whole drift of the argument in §§ 55-61 had appeared to be to
not vet been completely fixed; we know that they are meant to function as reject, not merely the radical adjectival strategy, but an adjectival strategy of
singular terms, and we know something of when they can be used and row any kind. Yet the form of sentence ‘There are just as many Fs as Gs’ must
the"truth of certain statements involving them is determined, but the criterion clearly be placed on the adjectival rather than the substantival side of the
of identity must be supplied if we are to be able to use them as full-fledged divide. It contains no number-adjective, indeed; but, just as a sentence of the
form, ‘There are four Fs, assigns a property to a first-level concept, so one
terms.
of the form, ‘There are just as many Fs as Gs’, states a relation between two
first-level concepts, that of equinumerosity: unlike a sentence of the form,
An alteration of course ‘The number of Fs is the same as the number of Gs’, it involves no reference
The impetus of the discussion pushes us to take a further step. e are aiming to or quantification over numbers treated as objects. It is far from apparent
at arriving at an explanation of the cardinality operator via a specification of why, if a sentence like ‘Jupiter has four moons’ is to be explained as meaning
the truth-conditions of a statement of the form (1). The cardinality operator ‘The number ofJupiter’s moons is 4’, one like ‘Jupiter has just as many moons
has been tacitly accepted as the fundamental operator forming terms for as there are figures in the syllogism’ should not be explained as meaning ‘The
numbers. It follows, therefore, that, to avoid circularity, our specification of number of Jupiter’s moons is the same as the number of figures of the
truth-conditions should appeal only to expressions not involving numerical syllogism’. Yet, having proposed the first explanation in § 57, Frege here
terms, viz. singular terms denoting numbers presented as objects. That is vehemently rejects the latter. He indeed insists on the fundamental status of
precisely the kind of specification that Frege gives. A sentence of the form (1) the equivalence:
is to be specified to be equivalent to the corresponding sentence ‘There are
just as many Fs as Gs’, or, in Frege’s jargon, ‘The concept F is equinumerous (2) , (Fx, Gx) card, [Fa] = card, [Gv],
to the concept G’. At the outset, indeed, in § 63, Frege makes this appear so
innocuous a step that he conflates it with the definition of ‘There are just as where ‘card, [Fa]’ symbolises ‘the number of Fs’. But he argues emphatically
many Fs as Gs’ in terms of one-one mapping to mean ‘There is a one-one that the direction of explanation must be from left to right, from the adjectival
map of the Fs on to the Gs’. We should, however, keep the two steps distinct. to the substantival form. His is therefore a mixed strategy, neither purely I

The first question is whether ‘There are just as many Fs as Gs’ should be adjectival nor purely substantival.
I
explained as meaning ‘The number of Fs is the same as the number of Gs’,
or, conversely, ‘The number of Fs is the same as the number of Gs’ explained !
Numbers and directions
as meaning ‘There are just as many Fs as Gs’. Frege decides in favour of the
latter direction of explanation in § 65. More precisely, he decides in favour As already observed, Frege does not argue directly that the direction of
of explaining The direction of the line a is the same as the direction of the explanation in (2) must be from left to right, but argues instead for the priority
line b as meaning The line a is parallel to the line F rather than conversely, of ‘The line a is parallel to the line F over ‘The direction of a is the same as
adding that the discussion can in essentials be transferred to the case of the the direction of F. In § 64 he adopts the expository’ device of switching the [
identity of numbers. The proposed explanation would obviously be fruitless if discussion from the question with which he is actually concerned, namely by'
it were not then possible to give a definition of‘is parallel to’ without appealing
I
what means the second-level operator ‘the number of 0s’ is to be introduced I
to the notion of a direction: but Frege does not trouble himself to discuss how and explained, to the discussion of what he trusts will be perceived as an i
such a definition should be framed. The two steps are distinct, even though analogous case, namely how to introduce and explain the first-level operator
the first would be useless if the second were impossible. Likewise, it would ‘the direction of £’. He is explicit about his intentions in a footnote to § 65:
be useless to explain ‘The number of Fs is the same as the number of Gs’ as
I here speak of parallelism in order to be able to express myself more conveniently
ZelnT aS CS’ if * Were nOt then P°ssible t0 and to be more easily understood. What is essential to these discussions will easily
aU that the n „ mvolvin&terms for numbers as objects; but, for i
an mat, the two steps are distinct. be able to be transferred to the case of identity of numbers. ■

Frege succeeds in making all this seem entirely natural, and all but inescap-
He continues his whole discussion in terms of this analogue until he arrives i
>
 t
U5 Frege: Philosophy ofMathematics I 10 Frege's Strategy H7
at the final conclusion almost at the very’ end, half-way through § 68; only at
I
I' evident, as well as the stated, purpose of the passage from § 64 to the middle
that point does he revert to the real topic, without more ado applying the i of § 68 is to conduct a discussion directly relevant to the central problem, how
conclusion he has reached concerning the correct means of introducing the I the cardinality operator is to be introduced, by treating of an almost perfectly
direction-operator to that of introducing the cardinality' operator. This pro­ analogous case. The analogy was not perfect, and Frege’s discussion gave
cedure rests upon the assumption, which he expresses in the footnote, but inadequate recognition to its imperfection: but any other interpretation of this
makes no attempt to argue, that the two operators are analogous in the i elevant passage reduces it to nonsense.
respects.
In fact, they are not. The analogy' fails on two counts. By far the more
important will be scrutinised in detail later. This turns on the fact that the I Problems
I
argument-place of the cardinality operator is to be filled by' a concept-word, One problem that arises from the passage extending from § 62 to § 69 is thus
that of the direction-operator, on the other hand, by a singular term standing to scrutinise the supposed analogy' between the concept of a direction and that
for a line; this discrepancy in level makes a significant difference to the upshot of a number. How did Frege intend us to transpose the argument for the
of the discussion. The second failure of the analogy' lies in the fact that one conceptual priority of the notion of parallelism over that of a direction into
operator belongs to arithmetic and the other to geometry. Frege was never one for the conceptual priority of the notion expressed by ‘just as man/ over
tired of emphasising the difference in character and status between the two that of a number? Does the fact that the cardinality’ operator is of second order,
branches of mathematics: he might therefore be expected to have taken care while the direction-operator is of first order, make a significant difference to
to include nothing in his reasoning about the direction-operator that made Frege’s argument?
appeal to its specifically geometrical content. He failed to take such care. He I
The passage as a whole is concerned to explore the possibility of introducing
argues in § 64 for the conceptual priority' of the relational expression ‘£ is the cardinality operator by outright stipulation of the equivalence (2) between
parallel to over the term-forming operator ‘the direction of £’. The argument ‘There are just as many Fs as Gs’ and ‘The number of Fs is the same as the
he gives is that we have intuitions of straight lines and an ‘idea’ of parallel number of Gs’, assuming the former to have been antecedendy explained in
lines, but no intuition of a direction, whereas everything geometrical must be
terms of one-one mappings. Much of the discussion is conducted in terms of
intuitive in origin: hence the operator must be defined in terms of the relation
the analogue, namely a means of introducing the direction-operator by stipulat­
of being parallel, rather than lines being defined to be parallel if their directions
coincide. ing the equivalence between ‘The line a is parallel to the line F and ‘The
direction of a is the same as the direction of F, assuming the former to
According to Frege, geometry' rests on intuition, while arithmetic does not.
be already understood. In §§ 63-5 such an explanation is defended against
It follows that the argument for the conceptual priority of the notion of
objections; but, at the last moment, Frege decides that there is one objection
parallelism over that of a direction cannot be adapted, without being greatly
modified, to a proof of the conceptual priority of the notion expressed by ‘just against which no defence can be provided. He therefore abandons the proposal
as many’ over that of a number. We could not argue that we have intuitions in favour of an explicit definition of the direction-operator, and, by parity of I

of concepts, and an ‘idea’ of their equinumerosity, but no intuitions of numbers. reasoning, of the cardinality operator. This decision comes as a shock to the
reader, because, in § 62, the passage has opened with the terse enunciation i
The argument needs to be thoroughly recast for this case; and Frege has
overlooked the necessity for indicating how it should be recast. of two principles which seem to make the proposal, rejected in §§ 66-9,
The attempt has been made to defend him by maintaining that, contrary to mandatory. The first was the context principle, offered as supplying the answer t
his express declaration in the footnote, he did not intend to transfer the to the initial question how numbers are given to us; § 62 had opened as
argument from the one case to the other? If he had not, then §§ 64-7 would follows: *
> t
a" ,rreievant excursus into the foundations of geometry, and
IF aritiirneti Ae " °f § 68’ he abrUptly lurches back int0 a discussion of ___ t is a number
How, then, ________ us, iftowe can have no idea or intuition of
to be,jgiven
arithmetic, his pronouncement that the cardinality operator must be defined it? Only in the
_ context of a sentence do the words mean anything. It therefore
for^heZdT5 t0 direCti0n-°Perat0r ™dd have been an assertion becomes a matter of explaining the: sense of a sentence in which a number-word
, he had given no argument whatever. Obviously, this is wrong. The occurs.

- the British Journal fir Only two sections previously, Frege had glossed the context principle by
observing that ‘it is enough that the sentence as a whole should have a sense;
 118 Frege: Philosophy ofMathematics 10 Frege's Strategy H9
I

it is through this that its parts obtain their content’. The principle thus appears three methodological precepts that have governed the composition of the whole
to demand that we should not attempt to assign a meaning to terms lor book, and cited again, with a flourish of trumpets, in § 60. Moreover, the
numbers on their own, independently of the particular forms of sentence in interpretation would have it that the two principles were ones in which Frege
which they occur, but should fix the meanings of those terms by laving down indeed believed, but which turned out to be irrelevant to his project. If that
the senses of such forms of sentence: it could hardly be foreseen that the were so, they would hardly figure again in his final summary of his argument
upshot of the whole enquiry would be an outright definition of the cardinality (§§ 106-8); but the context principle is reiterated once more in § 106, as a
operator, considered apart from any particular context. fundamental principle without which it is impossible to avoid a physicalist f
The second principle enunciated in § 62 was that of criteria of identity, and interpretation of number without falling into a psychologistic one. Frege thus >
this was presented as determining which sentences involving terms formed by considers it to have been an essential aid to arriving at his conclusions. I
means of the cardinality operator we need in the first instance to explain: Moreover, in §§ 106-7, Frege lays the same stress as before on fixing the I

sense of an identity-statement (a ‘judgement of recognition’). We thus have

We have already settled that number-words are to be understood as standing for an acute exegetical problem to resolve. It is: what, then, is the purport of the
self-subsistent objects. There is thereby given to us a category of sentences which
must have a sense, namely sentences which express recognition. If the symbol a context principle, and what that of the doctrine of criteria of identity, when
is to designate an object for us, we must have a criterion which decides in all something that appeared to be the consequence, and the intended conse­
cases whether b is the same as a, even if it is not always within our power to apply quence, of both is in the end repudiated?
this criterion. In our case we must explain the sense of the sentence
‘The number belonging to the content F is the same as that which belongs to The strategy in detail
the concept G’;
These are by no means the deepest, let alone the only, problems raised by
that is, we must render the content of this sentence in another way, without using
i
I §§ 62-9. To discuss them adequately, however, we must have in view Frege’s
I
i. ■
the expression entire strategy for defining the basic arithmetical notions in logical terms and
‘the number belonging to the concept F\ deriving from the laws of logic the fundamental principles of arithmetic as so
interpreted. We may list the arithmetical notions dealt with in Grundlagen as |
\\ e shall thereby provide a general criterion for the identity of numbers. When follow’s, setting those which belong with the adjectival use of number-words
we have thus obtained a means of laying hold on a determinate number and on the left, and those which belong with their substantival use on the right. I ;
recognising it as the same again, we can give it a number-word as its proper I
name. Those on the right involve either terms for numbers or individual variables
ranging over them; those on the left contain no numerical terms, and we are
not required to take the range of their individual variables to include numbers.
This had seemed quite explicit: terms of the form ‘the number of Fs’ are
to be explained by laying down the sense of an identity-statement connecting
There are just as many Fs as Gs the number of Fs
two such terms, and hence not directly; for if they were explained directly,
n is a number
there would appear to be no question of laying down the sense of the identity­
There are 0 Fs the number 0
statement which would, instead, have to be derived from that explanation.
And yet, that is what in the end Frege does: he gives an explicit definition of the number 1
the operator used to form such terms for numbers. What, then, was the point There is one more F than there are Gs n is a successor of w
?°JPrinCipleS’ ±e COntext PrinciPle and doctrine of criteria n>m
n m
dvevsened mere 1°^ SUCh A firSt th0U^ht miSht be **
9 Planation ofthe P.e™ve’ not as Prescriptive: they established that an n is a finite number ■
>
Seme ? Card,nal,t*' °perator bY stipulating the sense of an identity-
nrnved 7° "JV was not be ruled out a priori, even though it eventually Frege begins his chain of definitions at the top of the left-hand column, bj
defining ‘There are just as many Fs as Gs’ to mean ‘There is a one-one map
of the Fs on to the Gs’ (§§ 63 and 72); the notion of a one-one map is itself f

on the conte„ priMiple> defined in §§ 70-2. He then crosses to the right-hand column; as stated above,
! he first suggests explaining ‘the number of Fs’ by stipulating that ‘The number
t i
120 Frege: Philosophy ofMathematics 10 Frege's Strategy 121
) axiom, is immediately derivable, while the first and second Peano axioms,
of Fs = the number of Gs’ is to be equivalent to ‘There are just as many Fs
stating that 0 is a natural number and that the successor of a natural number
as Gs’, but then rejects this idea, and defines ‘the number of Fs explicitly’to
is again a natural number, are equally readily derived.
mean ‘the class of concepts G such that there are just as many Fs as Gs
68 and 72): the extension of a second-level concept is a class of concepts, This, then, is Frege’s mixed strategy of definition. The chain of definitions
just as the extension of a first-level concept is a class of objects. Frege then starts at the top of the left-hand side, moves across to the top of the right­
defines the predicate bz is a (cardinal) number’ to mean ‘for some F, n is the hand side, and then proceeds down that side, with the left-hand expressions
number of Fs’ (§ 72). The number 0 is defined to be the number of objects other than the top one defined in terms of the corresponding right-hand
not identical with themselves (§ 74), and the number 1 as the number of expressions. The strategy thus rests on three fundamental principles. The first,
objects identical with the number 0 (§ 77). The corresponding expressions on not stated explicitly, is that all notions on the right-hand, substantival, side are
the left-hand side, ‘There are 0 Fs’ and ‘There is just 1 F’, have in fact been ultimately derivable from that of the cardinality operator: the most basic
satisfactorily defined in § 55; but these definitions have been rejected in favour numerical terms are those of the form ‘the number of Fs’, and all other
of explaining them instead in terms of their right-hand counterparts, that is, arithmetical notions are to be defined by means of them. The second principle
as meaning, respectively, ‘The number ofFs = 0’ and ‘The number of Fs — 1 . is not only stated by Frege, but vigorously argued for by him in §§ 62-9: it is
The relation expressed by ‘» is a successor of nd is defined in § 76 to mean that the adjectival (left-hand) notion of cardinal equivalence expressed by
‘For some F and some ,v, .v is F and n is the number of Fs and m is the number ‘There are just as many Fs as Gs’ is conceptually prior to the cardinality
of objects distinct from ,v that are ‘F’. In this sense, a successor is not, in gen­ operator, which accordingly must be defined in terms of it: that is w’hy the
eral, the next greatest cardinal number: each transfinite cardinal will be its own chain of definitions must start on the left-hand side. The third principle is
successor; until this relation has been proved to be many-one, the definite ar­ stated by Frege in § 57, but only cursorily argued for. It is that all other
ticle before ‘successor’ remains unjustified. The binary quantifier expressed by adjectival notions must be defined in terms of their substantival (right-hand)
‘There is one more F than there are Gs’ is not in fact defined by Frege. ‘The analogues. The argument, such as it is, is that only so is it possible to establish
Fs comprise the Gs and one other object’ was suggested above as a suitable the required logical connection between substantival and adjectival notions.
replacement for the defective definition of‘There are just (n + 1) Fs’ suggested Essentially the same strategy is followed in Grundgesetze, Part II, save that
in § 55; from that, ‘There is one more F than there are Gs’ can easily be there Frege does not bother to introduce particular symbols for ‘is a number’
defined with the help of ‘There are just as many Fs as Gs’. Frege would, of and ‘is a finite number’, since he can express those notions quite tersely
course, define ‘There is one more F than there are Gs’ to mean ‘The number without them; but he calls attention in w’ords to the means he has for expressing
of Fs is a successor of the number of Gs’. The expression ln > nd is not used them.4 In the logical system of that book, the notion of a concept is generalised
in the above table in the sense ‘n is a larger cardinal number than nd, but in to that of a function yvith arbitrary objects as values; since the truth-values
place of Frege’s locution ‘n follows m in the natural series of numbers’, which true and false are treated as objects, a concept can then be regarded as a
holds if n can be reached from m by a finite number of steps (at least one) function all of whose values are truth-values. Every’ first-level function has a
i
going from a number to a successor of it: this notion is obtained by applying ■ value-range, belonging to the domain of objects; the abstraction operator,
die Begrijfsschrift definition of the proper ancestral to the successor relation, as forming terms for value-ranges, is primitive, and a class is the value-range of
is done in Grundlagen, §§79 and 81, and hence as meaning ‘m falls under a concept. In Grundlagen, appeal to the notion of a class is cut down to the
even- concept hereditary’ with respect to the successor relation under which very minimum: it is used for the definition of the cardinality operator and for
any successor of m falls’: a concept is hereditary’ with respect to a relation if nothing else. In Grundgesetze, however, it is used very’ freely, and this gives a
every object falls under it to which another object falling under it stands in somewhat different form to several of the definitions. XV ithout introducing any
t at relation. The expression ‘m 3= nd likewise represents Frege’s locution additional primitive, it is possible for Frege to introduce the notion of a double
n belongs to the natural series of numbers beginning with tn’, defined in § 81 value-range’, or extension of a binary function/(£, £): this is the value-range
to mean n > m or n - m'. Finally, ‘n is a finite number’, which is Frege’s of the unary function which maps any object a on to the value-range of the
St0’ m " lsAnaT?nuraber’’can be defined to mean o’ unary function/^, a). In the special case that the binary’ function is a relation
numbeHfittak definitlOn ,S of course ** « wifl be a finite (natural) (Beziehung), the double value-range thus obtained serves as the extension of
under which 0
“t From Ss Ae ’
’^7 heredita,y resPect to successor
1 lf °frd,"ar>'mathematical induction holds good for
, e principle of induction, which constitutes the fifth Peano
i that relation; Frege calls the extension of a relation a Relation. Since a function

Vol. I, § 42, p. 58, and § 46, p. 60.

122 Frege: Philosophy ofMathematics 10 Frege’s Strategy 123
I
is required to be defined for all possible arguments, however, Frege has to
substitute, for the notion of a function with a restricted domain, that of a
many-one relation; he had done just the same in Begnjfsschrijt and Grundlagen.
I the statement ‘The number of Fs is the same as the number of Gs’ (§ 73).12
It will be recalled that §§ 63-7 had been expended on an intricate discussion
of whether it was possible to introduce the cardinality operator by stipulating
He first defines this notion;5 more precisely, he defines the notion of being that equivalence outright, terminating in a decision that it was not possible
the extension of a many-one relation. He then defines the notion of a mapping:6 and in the consequent explicit definition of the operator. Now’, having given that
not, however, as in Grundlagen, as a one-one relation mapping the objects definition, Frege immediately exploits it to derive that very same equivalence.
falling under one concept on to those falling under another, but as the exten­ Furthermore, he never directly invokes the definition of the cardinality operator
sion of a one-one relation mapping the members of one class on to the for any other purpose: all that follows depends only on the equivalence proved
members of another. as the very first theorem. Crispin Wright devotes a whole section of his book,
Having defined the (extension of the) converse of a relation,7 he then Frege’s Conception of Numbers as Objects,™ to demonstrating that, if we were to
defines the cardinality operator.8 This is done in a similar way to that used in take the equivalence in question as an implicit or contextual definition of the
Grundlagen-, the difference is that, in Grundgesetze, the operator is a first-level cardinality’ operator, we could still derive all the same theorems as Frege does.
one, and a number is given as a class of classes rather than of concepts. The I He could have achieved the same result with less trouble by observing that
operator can be applied to an arbitrary object a\ its value will be the class of I Frege himself gives just such a derivation of those theorems. He derives them
classes that can be mapped one-one on to a, in other words, of classes having I all from that equivalence, with no further appeal to his explicit definition.
the same number of members as «; of course, if a is not a value-range, it will The proofs of the remaining theorems stated in Grundlagen present no
have no members. The Grundlagen notion of a class whose members are difficulty, save for the last. The first (§ 75) states that if there are no Fs and
concepts is undoubtedly a bizarre one, adopted by Frege in that book in order no Gs, then the number of Fs is the same as the number of Gs, namely 0,
not to use the notion of a class (of an extension of a concept) save where it and, conversely, that, if 0 is the number of Fs, then there are no Fs.14 There
became strictly necessary. In Grundgesetze, however, he did not want, and saw’ follow six theorems enumerated in § 78.15 Theorem (1) states that, if n is a
no need, to incorporate into the system a higher-level abstraction operator successor of 0, then n = I.16 Theorem (2) says that if 1 is the number of Fs,
forming terms for the value-ranges of second-level functions; hence his defi­ then something is F, and theorem (3) that, on the same hypothesis, anything
nition of cardinal numbers in that book more closely resembles that which that is F is identical with anything that is F (in other words, that not more
Russell was to give. Allowing for this difference in the definition of the than one thing is F).17 Theorem (4) is a joint converse of these, saying that if
cardinality operator, the Grundgesetze definition of the (extension of the) suc­
at least one thing is F and at most one thing is F, then the number of Fs is 1 .'8
cessor relation’ is just as in Grundlagen', the same applies to the definitions of
We arrive at something more interesting with theorem (5), which states that
the (cardinal) numbers 0 and I.10 Finally,11 he defines the ancestral of a
relation, essentially as in Begriffsschrift-, it will come as no surprise that he
actually defines the operation which converts the extension of a relation into
I
i
the successor relation is one-one: this constitutes essentially the fourth Peano
axiom.1' Theorem (6) says that any cardinal number other than 0 is a successor
of some number.20 The proposition that 0 is not a successor of anything, which
the extension of the ancestral of that relation. He then uses the ancestral, in
application to the successor relation, to obtain the notion of a finite cardinal is the third Peano axiom, is not expressly stated in Grundlagen, but is proved
number. in Grundgesetze, Volume I.21 The last theorem stated in Grundlagen (§ 83),
The theorems whose proofs are given or isketched ’ though not the last in the order of proof, is that no finite number follow s itself
in Grundlagen are as in the natural series of numbers. That is to say that the sequence of natural
follows. First, having defined the cardinality c
' operator, Frege immediately
proves that the statement ‘There are just as many Fs
> as Gs’ is equivalent to 12 corresPon<3>ng theorems of Grundgesetze, vol. I, are (32), § 65, p. 86, and (49), § 69, p.

13 Aberdeen, 1983, section xix, pp. 154-69. fl ,7Q

5 These appear in Grundgesetze, vol. I, as theorems 94, § 97, p. 128, and 9/, § , P-
Vol. I, §37, p. 55.
" § 38, p. 56. ’’ These are listed together in symbolic form in Grundgesetze, vol. I, § 44, pp. 30- •
’ § 39, p. 57. ’’This is theorem 110 of Grundgesetze, vol. I, § 103. p. 132.
6 § 40, p. 57. ^The corresponding theorems of Grundgesetze, vol. 1, are 113, § 103, p. > ’ ■ ’
’ § 43, p. 58.
'"§§ 41 and 42, p. 58. J* This 15 E’ven ’n Grundgesetze, vol. 1, as theorem 122, § 107, p. 136.
" In §§ 45 and 46, pp. 59-60. ’ This forms theorem 90 of Grundgesetze, vol. I, § 95, p. 127.
It figures in Grundgesetze, vol. 1, as theorem 107, § 101, p. 131.
' As theorem 108, § 103, p. 131.
124 Frege: Philosophy ofMathematics
numbers does not form a cycle, returning to itself.22 This theorem, as being
the first listed here in which the generalisation is restricted to the natural
numbers, is therefore also the first that requires an appeal to induction. It
does not yet establish that the sequence of natural numbers is infinite, since
it has to be shown that it does not terminate in a number that has no successor. i CHAPTER 11
To this purpose, Frege sketches in §§79 and 82-3 the proof of the crucial
theorem that every natural (finite) number has a successor.23 Every theorem
so far is likely to be quite easily provable on the basis of anything worthy of
the name of a construction of arithmetic; the infinity of the sequence of Some Principles of Frege's Strategy
natural numbers, which, in Frege’s construction, depends on the existence of a
successor to every’ natural number, has a far more uncertain status. Since
every' number is, for Frege, the number of objects falling under some concept,
he has to cite, for any given natural number n, a concept such that the number Contextual definition
of objects falling under it is a successor of n. For this purpose, he takes the
concept ‘natural number C n’ (where the relation is just the converse of The proposal discussed by Frege in Grundlagen, §§ 62-7, is to introduce the
the relation > defined above). There is one number < 0, namely 0 itself, and operator ‘the number of 0s’, not by defining it explicidy, but by means of a
the number 1 is a successor of 0. Likewise, there are two numbers, namely 0 contextual definition, namely by stipulating a sentence of the form ‘The
and 1, each of which =£ 1; and 2 is a successor of 1. In §§ 79 and 82—3, Frege number of Fs is the same as the number of Gs’ to be equivalent to ‘There are
shows how, by induction, to establish the general theorem that the number of just as many Fs as Gs’, where the latter is in turn explained by an explicit
natural numbers C to any given.natural number n is a successor of «;24 and definition in terms of one-one mapping; from § 64 onwards, the discussion is
from this the desired theorem that every natural number has a successor conducted in terms of the analog}’ with directions. In his middle period, Frege
follows at once by existential quantification. 25 became hostile to contextual definitions, and to ever}’ form of definition other
than straightforwardly explicit ones. It is quite evident, however, that, at the
time of writing Grundlagen, he felt no such hostility, and, moreover, that he
conceived of his context principle as licensing contextual definitions; his remark
in § 60, that, in accordance ■with the context principle, it is sufficient that a
sentence should have a sense as a whole, from which its parts derive their
content, and that ‘this observation ... throws fight on many difficult concepts,
such as that of the infinitesimal’, leaves no room for doubt on this score; this
is reinforced by Frege’s reference to Grundlagen towards the end of his renew
of Hermann Cohen’s Das Prinzip der Infinitesimal-Methode und seine Geschichte.
It is therefore unsurprising that Frege should treat the proposal for a contextual
definition of the cardinality operator with complete seriousness.
His later objection to contextual definition was expressed in Grundgesetze by
the use of a mathematical analog}’.1

It is evident that the reference of an expression and of one of its parts do not
always determine the reference of the remaining part. One therefore ought not
to define a symbol or word by defining an expression in which it occurs, the
remaining parts of which are already known. For an enquiry would first be
necessary’ whether any solution for the unknown - I here avail myself of a readily
understandable algebraic metaphor - is possible, and whether the unknown is
uniquely determined. As has already been said, however, it is impracticable to

' Vol. II, § 66.

 I

126 Frege: Philosophy ofMathematics 11 Some Principles ofFrege’s Strategy 127

make the justifiability of a definition depend upon the outcome of such an enquiry, contextually has a surface form belying that which the contextual definition
which, moreover, may perhaps be quite unable to be carried out. The definition i assigns to it: the classic example is Russell’s definition of the description
must, rather, have the character of an equation solved for the unknown, on the
other side of which nothing unknown any longer occurs. operator. Russell permits this operator to retain the outward form of a term­
forming operator of second level, that is, to be attached to a predicate. The
whole point of the Theory7 of Descriptions, however, is to deny that definite
When he wrote Grundlagen, Frege plainly had not yet developed any objec­ descriptions, that is, the apparent terms formed by means of the description
tion to definitions whose justifiability' depends on the proof of some proposition: operator, are genuine terms at all. Given Russell’s notation, it w'ould therefore
he had spoken equably of such a possibility7 in § 3. Would he then have be impossible to give an explicit definition of the description operator, since
required, in order to justify a contextual definition, a proof that (to continue any such definition would be forced to accept its apparent form as genuine.
the metaphor) it had a solution, or, more exigently yet, that it had a unique On Russell’s theory7, the description operator actually functions as a binary
solution? This, though very differently expressed, proved in the end to be his quantifier. Were Russell to have adopted a notation in which it explicitly
objection to the proposed contextual definition of the cardinality operator. Put figured as such, e.g. by writing
more exactly, the contextual definition had a solution, but not a unique one;
it had therefore to be replaced by an explicit definition, providing a determinate h [£v, Gr]
solution. There is no hint, in the text of Grundlagen, that from this any general
objection to contextual definitions can be derived, and Frege’s remarks in § 60
for ‘The F is G’, there would be no obstacle to giving an explicit definition:
make it very doubtful that he thought so. It was just that this particular
the need for a contextual definition arises from the mismatch between the
contextual definition, and others of analogous form, did not fulfil the require­
apparent and the real form of the expression defined. That is not at all the
ments that we are entitled to demand of a definition.
case with Frege’s proposed contextual definition, however. The cardinality
The stipulation that the direction of a line a is to be the same as that of a
operator has the same overt form as Russell’s description operator, namely
line b just in case a is parallel to b does not determine whether the direction
of a line is itself a line or something quite different: this contextual definition that of a term-forming operator of second level: and Frege takes it without
indeed has a solution, but it is far from unique. Even if the requirement were reservation to be of just that form. Numerical terms, including those formed
to be made that every’ direction should itself be a line, the stipulation would by use of the cardinality' operator, stand in all cases for objects.
in no way determine which line any given direction was to be; it could, in fact, Having framed the proposal for a contextual definition of the cardinality
be any line whatever. A convenient choice would be to take some point as the operator, Frege proceeds to consider three objections to it. The first two he
origin 0, and identify the direction of any line a with that line through 0 that rejects; the third, to the reader’s surprise, he sustains, and so adopts instead
was parallel to a; even so, any point could serve for this purpose as the origin. his explicit definition. The first objection is stated in § 63 as follows:
The contextual definition might well be defended on the ground that we do
not need to know anything about directions save w'hat it tells us: as long as The relationship of identity' does not occur only among numbers. From this it
we know that the direction of a is the same as that of b just in case a is parallel seems to follow' that we ought not to define it especially for this case. One might
to b, we are quite indifferent to what, specifically, the direction of a may be, well think that the concept of identity' would already have been fixed previously,
and that from it and from the concept of cardinal number \AnzaM\ it must follow
or any other facts about it. Frege makes plain in § 66 that this defence w7ould when cardinal numbers are identical with one another, without there being any
not satisfy him at all. It is an inexcusable defect in a proposed definition of need for this purpose of a special definition besides.
the direction-operator that it fails to tell us what, specifically, the direction of
a given line is to be: and hence it must be replaced by an explicit definition This goes very much to the heart of the powerful, but obscure, principle of
which does tell us that. criteria of identity7: does the principle require that identity7 be defined separately
. reason’ unaccePtable to Frege, for giving a contextual definition, may for each of a multitude of cases? Frege’s answer to the objection contains a
thus be that it does not have a unique solution, and we do not wish to specify
one. Another might be that, although the definition has a unique solution, we resounding ‘No’ to this question:
“ £ve,*h' r'S™'“sit- There is, however, a quite different
ranting a definition as a contextual one: namely, that the talk of a
o u ion is quite misplaced. This will occur when the expression defined

I
intention is to construct the content of a judgem therefore not wishing
identity on either side of which a number stan s.
128 Frege: Philosophy ofMathematics 11 Some Principles ofFrege's Strategy 129
to define identity especially for this case, but, by means of the already known ■s *
concept of identity, to attain that which is to be regarded as being identical. This I
contextual definition of the cardinality operator would needTo be'uSfcTbv
admittedly appears to be a very unusual kind of definition, which has not yet been showing cardinal equivalence - the relation exore«ed «• . 1 ned by
adequately noticed by logicians; but a few examples may show that it is not
;oncepSeql,iVa'enCe rela'iOn °f SK°nd <an tXZbe^Tn’
unheard of.

In the following section, § 64, Frege then gives various examples that he c aims Frege here commits a Wunder, easily overlooked. Having „bse„ed that “we
as analogous, including the concepts of direction and of shape; length and
t
colour are mentioned in § 65. By presumption, what holds good for any one I
of these cases will hold good for all. Frege proceeds, in the same section, to
state his grounds for holding that the concept of parallelism is prior to that of All other statements about directions would first have to be defined, and for these
a direction, so that the latter must be defined in terms of the former, and not definitions we can impose the rule that the intersubstitutability of the direction of
a line by that of another one parallel to it must be ensured. 4;
conversely; by analogy', the concept of a number must be defined in terms of
cardinal equivalence, rather than defining ‘There are just as many Fs as Gs
to hold when the number of Fs and the number of Gs coincide. It is natural to regard this remark as obvious. If it is worth introducing terms
Frege’s reply to the first objection can be seen only as an endorsement of the for directions at all, w'e shall surely want to say things about directions other
general principle of contextual definition, that procedure which, very' similarly' I than that they are or are not identical. To do this, w'e must introduce suitable
I
described, he repudiated in A olume II of Grundgesetze. The proposal is not to I predicates. This we shall surely do by means of further contextual definitions,
define the cardinality' operator on its own, and then, by putting this together equating statements assigning properties to, or relations between, directions
with the already known meaning of ‘is the same as’, to arrive at the sense of with ones assigning corresponding properties to, or relations between, fines.
‘The number of Fs is the same as the number of Gs’, by addition, as it were. For such a definition to be legitimate, the relation of parallelism must be a
Nor is it to give a sense to ‘is the same as’, but only for this special context. congruence relation with respect to the property of, or relation between, lines
Rather, it is just because we already know what ‘is the same as’ means in all that it invokes: and therefore we shall be involved in giving a chain of contextual
contexts that, by stipulating what ‘The number of Fs is the same as the number definitions, one for each context in which we want terms for directions to
of Gs’ is to mean, we can arrive, by subtraction, at the meaning of ‘the number occur, and, with it, a chain of proofs that parallelism is a congruence relation
of Fs’: the very procedure subsequently condemned by Frege in the passage with respect to various properties and relations.
cited above from Grundgesetze. In the earlier passage, there is no hint that he For example, w'e might propose to define the direction of a to be orthogonal
saw anything wrong with it. to the direction of b just in case the line a is perpendicular to the line b. This
But, if the sign of identity does possess a meaning which it retains in all definition w'ould be in order if we were concerned only with plane geometry,
contexts, and this meaning is already given, namely by Leibniz’s law', as Frege so that all the lines considered w'ere on a single plane: but, to show it to be
explains in § 65, a second objection arises: how can we be sure that our in order, we should have to show' that being parallel was a congruence relation
definition does not conflict with the general laws of identity? The objection with respect to perpendicularity. That is, we must show that, if a is parallel to
cannot receive a general answer: we need to examine the particular proposed o', and b to b', and if, further, a is perpendicular to b, then also a' is
contextual definition; in § 65, the matter is being discussed apropos of the perpendicular to b'. If, however, we were concerned with three-dimensional
contextual definition of the direction-operator in terms of the relation of being space, we should not be able to prove that, since two lines can be perpendicular
parallel. For this case, we need specifically to show, for any lines a and b, that,
if a is parallel to b, the term ‘the direction of a’ can be replaced, in all contexts, only if they intersect; and so the definition would be inadmissible.
In the same wray, wre shall wish to be able to say, not merely that cardinal
by the direction of b’ without change of truth-value. Frege here remarks that
numbers n and m are equal to one another (that is, are identical), but also that
we have not as yet provided for the occurrence of such terms in any other
°ne is less than or equal to another in magnitude; as is well known, we, shall
context than an identity-statement connecting two of them. For such contexts,
then need to define ‘less than’ to mean ‘less than or equal but not equal. e
relation^that i<e °ne showing parallelism to be an equivalence shall also need to define the operations of cardinal arithmetic - addition,
multiplication and exponentiation. To take the first step, we may well define
the number of Fs to be less than or equal to the number of Gs just in case
 >•

11 Some Principles of Frege's Strategy 131

130 Frege: Philosophy ofMathematics
unable to prove the number of Fs to be the same as the number of Gs when
there is a one-one map of the Fs into the Gs, that is, on to some (possibly all)
there were just as many of the one as of the other: in this case, we should
of the Gs. To justify this definition, we must show that cardinal equivalence
have no choice but to make a second contextual stipulation. It is true that we
is a congruence relation with respect to mapping into: if there are just as many
cannot avail ourselves of the same device of existential quantification when we
Fs as Hs and just as many Gs as As, and the Fs can be mapped one-one into
come to define the operations of cardinal arithmetic. In that case, we must
the Gs, then the Hs can be mapped one-one into the As. I
It is, however, a mistake to suppose that, once the original contextual justify our definitions by proving cardinal equivalence to be a congruence
relation with respect to analogous operations on sets (in Frege’s terms, on 1 1
definition has been given, any further such definitions are needed. Ue need
only define a direction p to be orthogonal to a direction q if there exist lines concepts). But it is still cardinal equivalence that is the pivot on which the
definitions turn.
J
a and b such that a is perpendicular to b, and p is the direction of a and q the
direction of b, and then our definition is unquestionably legitimate, without
the need for any proof. This will not, of course, save us any real work: we Why numbers had to be objects
shall still need to prove, on the plane, that, if the direction of a is orthogonal
to the direction of b, then a is perpendicular to b, something that will not be It is only when we have surveyed the chain of definitions Frege actually gives,
so, on this definition, in 3-space. But it serves to bring out the force of the and of theorems he actually proves, that we can see why he had, in Grundlagen,
principle which Frege’s insight lighted on, that, in determining concepts like I § 56, to reject a radical adjectival strategy; why, that is, he had to take sentences
number and direction, the criterion of identity is the first feature to be fixed. I containing apparent singular terms for numbers at face value. Contrary to the
We might well question this for cardinality. To lay down when two sets are impression he contrives to convey in §§ 56-7, the radical strategy’ can be
to be said to have the same number of members is w’ell known not to determine pursued for a considerable distance: the definitions and proofs Frege actually
I
unambiguously when one should be said to have fewer members than, or at
most as many members as, another. If we defined in the usual w’ay w’hen there
were at most as many Fs as Gs, on the other hand, we could stipulate that
gives in §§ 71-83 can readily be mimicked in the adjectival mode. To illustrate
this, some laxity of notation is desirable: if wre w-ere to retain the bound
variables needed to satisfy Frege’s principle that no concept-expression ought
I
I

there were just as many Fs as Gs if there w ere at most as many, and also at ever to appear deprived of its argument-place, the formulas w’ould become too
most as many Gs as Fs; the Schroder-Bernstein theorem would guarantee that cluttered with bound variables to be readable. For purposes of exposition,
the notion so defined would coincide with the usual one, as defined in terms therefore, we may, when convenient, omit the argument-places, waiting
of mappings on to. This might w'ell lead us to think that cardinal equivalence ‘ 3 ( (F)’ for ‘There is just one F\ lF ~ G’ for ‘There are just as many Fs
is not after all the fundamental notion. as Gs’, and so on.
Such a thought would, at least in one clear sense, be mistaken. It is true As already indicated, we may start with the analogue of the predicate is a
that we need an independent definition of the binary quantifier ‘There are at cardinal number’. Where ‘AF ranges over second-level concepts, we may use
most as many Fs as Gs’. It is also true that we could, if we liked, define ‘There lCard(/W)’ to mean that ‘M(F)’ holds good just in case F is of some one
are just as many Fs as Gs’ from it. But the latter notion - that of cardinal particular cardinality. ‘Card(/W)’ may thus be defined to mean:
equivalence - is still what we need in order to arrive at cardinal numbers, that
■

drfni,1™ “X cardinality operator, whether by means of a contextual VF [M(F) -> VG (/Vf(G) > F ~ G)]
oZ;. '™?' ' c We have k’ ” need ”° contextual definition
card naTnumbe,.T f” r™8 relati°ns by between The numerically definite quantifiers ‘2 o’ and ‘3i’ maY now defined just as
°f direCIiOnS’ we > number " in § 55. We may use ‘Succ(N, Al)’ as the analogue of ‘n is a successor of m ,
I and define it to mean:

VF (M^) -> 3 •' V* & My &

Now, where ‘A ranges over MmMevel concepts, and ‘F, ‘£’ over second-
ievel ones, we may define lN > AF to mean:

as many. as VA [(VP (Succ(P, Al) A(P)) &

VPV£(A(0 & Succ(P, jQ) K(P)) -> A(A)]
 11 Some Principles of Frege’s Strategy 133
1
132 Frege: Philosophy ofMathematics
belonging to the domain of the individual variables, that he was enabled to
(N > M' may then be set as:
spin the infinite sequence of natural numbers out of nothing, as it were. There
must be at least 0 objects, and hence the number 0 exists. Since the number !
N>A1 vVF[N(F) WJ 0 exists, there is at least one object, and so the number 1 exists: and so on
indefinitely. It is in order to prove the infinity of the natural-number sequence
Finally, the analogue of ‘w is a finite number’ may be written as INat(A')’, and that Frege is compelled to construe numbers as objects, and not for any of
defined to mean: the spurious reasons he cites in §§ 56-7.
It may naturally be said that there can be no such thing as a purely logical
TV 2 o. proof that there are denumerably many objects - logical objects - unless a I.
strong existential assumption was built into the theory at the outset. Existential
With these definitions, we may readily prove the analogues of all the the­ assumptions, when not explicitly stated by means of the existential quantifier,
orems proved by Frege up to Grundlagen, § 78, together with that of the are incorporated into a theory by the use of term-forming operators and the
theorem stated in § 83, that, for every natural number «, -> n > n. The principles assumed to govern them. In the theory of Grundlagen, this is done
analogues of the remaining two theorems cannot be proved, however. These by use of the cardinality operator, or, when this is defined in terms of classes, j
are the theorem establishing the infinity of the sequence of natural numbers, by the abstraction operator by means of which class-terms are formed. If there •. I
to the effect that every natural number has a successor, and its lemma, saying are n objects altogether, there will be 0(n) cardinal numbers, for a suitable
that, if n is a natural number, the number of numbers m such that n 3* m - function 0: when n Aleph-0, we may put 6(n) = n + 1. Since the cardinal I
that is, of numbers from 0 to n inclusive - is a successor of n. Since, according numbers are themselves objects, we must have 0(ri) n; and this can be so !
to the radical adjectival strategy, we are not taking numbers to be objects, we only when n 2= Aleph-0. In a similar way, Frege’s use in Grundgesetze of the
cannot so much as frame the lemma. In place of numbers, we have (numerical) abstraction operator for forming value-range terms imposed a restriction on
second-level concepts. To say how many second-level concepts there are the cardinality of the domain. If there are n objects in the domain, there will
satisfying a given condition - to say, for example, for given TV, how many be n" value-ranges; since value-ranges are objects, we must have n” n. There
second-level concepts Al there are such that TV & Al - we should need an is only one value of n for which this inequality holds, namely n = 1; but since
expression for a fourth-level concept, T, say. But this T could not be a Frege had assumed the existence of at least two objects, the two truth-values,
successor of TV, for they are concepts of different level, and it is only of a a contradiction resulted, and the system was inconsistent.
concept of second level that we can meaningfully say that it is a successor of From this we see more clearly the content of Frege’s thesis that numbers
some other second-level concept. Now, since numbers are not being taken to are objects. The argument used in § 57, from the surface appearance of
be objects, our theory will not contain any presumption that there are infinitely numerical terms in arithmetical statements, is in any case weak, since it is not
many objects: it will be perfectly consistent with the hypothesis that there are, mandatory to respect surface form. Even if it were allowed full weight, however,
say, only 100. If there were only 100 objects, the second-level concept 3100 it would prove too little. We could respect the surface forms of arithmetical I
would have no successor; for, if it had a successor, the condition
statements in a two-sorted theory, distinguishing, say, a domain of actual
(wirkliche) objects from a domain of non-actual ones, or of non-logical objects
3 -V 3 wo J (Fy & y # x) from a domain of logical ones. If the cardinality operator were allowed to be
attached only to predicates applying to objects of the first domain, we could
would have to hold for some suitable F, and there would then after all be at not form the term ‘the number of numbers tn such that n tn’, and so could
least 101 objects. It was for precisely the same reason that Russell, whose not prove that there were infinitely many natural numbers. If we were permitted
theory of types required that numbers, as classes of classes, be segregated to form terms by attaching the cardinality operator to predicates applying to
from individuals, was compelled to adopt an Axiom of Infinity, stating that objects of the second domain, in which the numbers were located, then o
there were infinitely many individuals, in order to guarantee that there were course we could prove the infinity' of the natural number-sequence just as
Frege does; but there would then be little point in distinguishing the two f.
infinitely many cardinal numbers; in fact, if value-ranges are excised from the
logical system of Grundgesetze, the result is a form of the simple theory of domains. Frege’s thesis that numbers are objects does not imply merely that
types. It is only because Frege reckoned numbers among objects, that is, as expressions for numbers have the logical status of singular terms: it implies,
further, that such terms stand for objects belonging to the sole domain over
 11 Some Principles of Frege's Strategy 135
134 Frege: Philosophy ofMathematics I

Ba 3A (p = dir(a) & q = dir(fc) & a is perpendicular to b).

which the individual variables range. Such objects therefore fall within the
scope of the first-level quantifiers. Frege’s explanation of the universal quanti­
We may then convert
fier in Grundgesetze is of a resolutely ‘objectual’ character, in Quine’s termin­
ology’:2 dir(r) x dir(d)
Xfii <P(a)
into
is to denote the value true if the value of the function is the value
Ba 3A (« // c &. b // d & a is perpendicular to b).
true for every’ argument, and otherwise the value false.

In order that quantified statements should have a determinate truth-value, all In this manner, the expanded language, involving reference to and quantifi­
the objects in the domain must be, as it were, already in place, independently cation over directions, can be translated into the original language, involving
of which of them is denoted by some term that can be framed in the formal only reference to and quantification over lines.
I
language. Such a two-sorted theory’ provides the only context in which the stipulation
This bears on the question whether the modes of introducing the direction­ of the criterion of identity’ for directions genuinely constitutes a contextual
operator and the cardinality operator reviewed, and ultimately rejected, in definition, in the sense of one supplying a means of eliminating any occurrence
§§ 63-7 of Grundlagen are genuinely contextual definitions, in the standard of the expression defined. It certainly does not represent Frege’s intention,
sense of permitting the elimination of the defined expression, by transforming however: he surely wished to add the direction-operator to a one-sorted
any sentence containing it into an equivalent one not containing it. If the language whose individual variables ranged over lines, and many other objects
proposed contextual definition of the direction-operator is construed as intro­ besides, and whose single domain would then be taken also to include direc­
ducing a two-sorted language, with one sort of individual variable {a, b, ... ) tions, and tacitly to have included them all along. The vagueness of the
ranging over a domain of lines and a new sort (p, q, ... ) ranging over a background against which the proposed contextual stipulation is supposed to I
domain of directions, in which the direction-operator can be attached only to be given makes it difficult to discuss; but we may simplify our discussion by
a term or variable for a line, the elimination is easy. If we write ‘dir(£)’ for supposing that, in the original language, the variables were capable of being
‘the direction of £’, we have first to transform any quantification over directions: interpreted as ranging only over lines. Now, in the language expanded by the
addition of the direction-operator, we have two choices, if we are not to
abandon at the outset all chance of eliminating that operator: we may either
identify directions with lines, or differentiate each from the other. We may
into a quantification over lines: begin by’ considering the first of the two options. It would be entirely’ contrary’
to Frege’s principles to place any restrictions on the occurrence of terms for
Vrt ( ... dir(fl) .. . ). directions in the argument-places of the already existing predicates. Hence, if
we are hoping to eliminate such terms from all contexts, we must add assump­
Then any subformula of the form
tions strong enough to identify specifically the line with which any given
direction is to be equated, say by the device, already mentioned, of taking the
dir(a) = dir(Z>) direction of a to be the line through some particular point 0 parallel to a. We
shall, in other words, have to meet Frege’s third objection, and lay down
can be replaced by
enough to determine, not indeed whether England is the direction of the
Earth’s axis, since, in our artificially restricted example, we have the means of
a // b. referring only to lines and their directions, but whether the Earth s axis is or

sx: zz defined * is ,o
quantifier, in place of Frege’^concLi3'6 Standard modern
> <• -
for the universal
ls not the same as its direction. If we do this, however, the contextual
stipulation

dir(«) = dir(A) a//b

■ I
 11 Some Principles of Frege's Strategy 137
136 Frege: Philosophy ofMathematics
3 v [Line(x) & (-> Line(x) v -i Line(dir(x)))].
will become otiose, since we shall be able to derive it as a theorem, moi eover,
we shall be in a position to define the direction-operator explicitly. i
The alternative is the stipulation that, for every' x, die direction of ,v is not Since *Line(x) & -> Line(x)’ is a contradiction, and Line(dir(x))’ is a
a line. To formulate this within the theory, we should have to introduce a of
consequence c. the axioms that had been added, this would reduce to:
predicate, ‘Line®’, meaning ‘£ is a line’, and lay down a number of axioms
3x Line(x),
serving in effect to give the one-sorted theory the force of a two-sorted one.
One such axiom would be:
which could be rendered in the language of the original theory' as
V.v -i Line(dir(.v)).
3x x // x.
Another, to sterilise reiterations of the direction-operator, might be: I
Plainly, such a transformation would not count as a mere elimination of the
V.v (-1 Line(x) —> dir(.v) = x). direction-operator in the usual sense, since it would involve tampering with
the (one-sorted) variables of quantification in the passage from the new theory
Furthermore, for each predicate of lines, we should need to adopt an axiom to the old one. I
of the form: Thus even the stipulation of the criterion of identity for directions cannot
count as a contextual definition proper save against the background which
I.
V.vVj(.v//j- Line(.v) & Line()')). Frege undoubtedly did not intend. This, it may be said, explains why Frege ■

rejected it, his programme being to demonstrate arithmetical statements to be

analytic in the sense of being reducible to logical truths with the aid of
By these means, we should obtain a theory whose theorems might be translated
definitions. In any case, it does not differentiate the supposed contextual defi­
into theorems of the original theoiy which contained no direction-operator.
nition of the direction-operator from that of the cardinality operator; for that,
We should nevertheless be unable to eliminate the direction-operator by prov­
ing, within the new theory', the equivalence of statements containing it with too, w’ould become a genuine contextual definition only if it were part of a
two-sorted theory'. Suppose wre have a second-order, one-sorted language,
:i
statements not containing it. The reason is that, in the process of mapping
theorems of the new theory' into theorems of the old, we should not be able with individual variables x, j, ...»in which is expressed a theory of any kind
to leave the quantifiers intact, but should have to translate them. Thus a you please. To this we add individual variables of a second sort, namely
statement of the form ‘Vv A(x)’ would first have to be transformed into: number-variables tn, n,..., and the cardinality operator, construed as forming
terms denoting numbers (elements of the second domain) when attached to
V.v [Line(.r) (A(x) & A(dir(x)))J. predicates applying to elements of the original domain. The cardinality operator
‘card, [0.v]’ will be governed by:
Wlien ‘A(x)’ had been transformed into a formula ‘B(x)’ not containing the !
direction-operator, and ‘A(dir(x))’ into another such formula lC(x)’, we could “v [Fx, Gx] card,. [F.v] = card, [G.v].
finally render the statement ‘V.v A(x)’ of the new theory as the statement
We can now translate every statement of this expanded theory' into a statement
Vx(x//x-> B(x) & C(x)) of the original theory', first rendering a quantification of the new sort.
I
V» A(m)
"Awl St S: e"S,en“ F°r inSUnCe’"" as
S v -i Line(x).
VF A(cardv [Fx]).
This we should first have to transform into:
In such a theory, we could not prove that there are infinitely many
 i

/1 Some Principles of Frege's Strategy 139

138 Frege: Philosophy ofMathematics l fact that the cardinality operator is of second order, while the direction-
i
numbers, since we could not even frame a term for the number of numbers The
js of first order, thus proves to be no irrelevancy, as Frege would
operator
less than or equal to a given number»: the cardinality operator can be attached think, but of critical importance: it is just for that reason that the
onlv to predicates applying to the original objects, not to those applying to have us
introduction of the cardinality operator embodies a far stronger ontological
numbers. But that only shows what we already knew, that such a two-sorted
assumption, namely that the domain of objects over w'hich our individual
theory was not what Frege had in mind. .... •ables range is infinite. That is what is involved in regarding numbers as
Eliminability of the newly introduced operator is not the principal point, vain to treat it as embodying that assumption was a heavy burden for Frege
however. Even though, in a one-sorted theory, the criterion of identity for objects:
allotted to logic. Scattered amongst his writings are muted invocations
directions is not a contextual definition, properly so called, we have a ready to have
of an argument for taking numbers to be objects, different from those he states
/ means of constructing a model of the new theory', given a model of the original °n Qrundlagen. This is that numbers can themselves be counted, and that in
one; the easiest way to do so is that already canvassed, by identifying directions number theory' we often need to speak of the number of natural numbers
with lines. That is made possible by the ontological parsimony of the theory
satisfying a given condition. Even on Frege’s principles, it does not direcdy
of directions: it does not demand the existence of any' more of the new objects
follow from the fact that numbers can be counted that they must be objects,
- directions - than there were of the old ones - lines. The theory' of cardinal
on the ground that a cardinal number is the number of objects falling under
numbers is very' far from being ontologically parsimonious, on the other hand:
some first-level concept; for Frege observes more than once that concepts,
it requires the existence of n + 1 new objects — numbers — given n objects of
f> the original kind, and hence, in a one-sorted theory such as Frege intended,
of the original kind and the new kind taken together. If the model for the
too, can be counted. There is more force in the observation that we frequently
wish to relate a natural number to the number of numbers standing to it in a
original theory’ wras finite, a denumerable model would indeed suffice for the certain relation, as when we define Euler’s number-theoretic function (pin) to
expanded theory' to which numerical terms had been added; but that would be the number of numbers «£ n and prime to it. But the fact is that even this
obviously not be a model for whose construction we had employed only the argument is not compelling. In Grundgesetze, Frege admits binary functions,
resources required for the construction of the original one. Even if we could and therefore relations, whose arguments are of different levels;3 there is
not in the usual sense eliminate the direction-operator, we could, by reinter­ therefore no reason why we should not consider a one-one mapping of things
preting the quantifiers, translate statements involving directions into statements of one level on to those of another. Functions such as 0(n) may be dealt with
not involving them. We cannot do this for statements involving numbers. Since even more conveniently. We need only define the characteristic function
the cardinality operator is of second level, occurrences of it can be embedded %(w, h) for the relation ‘are co-prime to one another’, putting
within the scope of other occurrences in a much more complicated way than
could happen with the first-level direction-operator. Consider the crucial term 1 if (w,h) = 1
card„ [h > m], essential for the proof of the infinity of the sequence of natural Z(w, ») =
numbers. The relation 3= between n and m is the ancestral of the successor 0 otherwise
relation, whose definition involves two occurrences of the cardinality operator.
No programme of eliminating the cardinality operator by appeal to the sup­ and then defining 0(h) by:
posed ‘contextual definition’ - the criterion of identity for numbers - could
possibly succeed in eliminating these inner occurrences of the operator. The
reason is that they will be found to stand in contexts of the form
n
(pin) = 2 «)•
m=l
card,. [Fx] = k,
This could also be written, without appeal to «), as:
nmS^n0”1’’ ‘F’ br‘ j’ “ ’ b0U”d ™able; and the ‘«»“«tlial definition’
on one side T °f "i "1’ “ e<l“a,i°n ” which a "“Mrical term stands
eodd L , n7T. °" tb' other- No strategy of definition
0(h) = S 1-
namral nLZ P™f °f °( °f 0<h</h
Meed Z c '‘.r “ J™* This may be seen as a spectal case - (w,»)=l
indeed, the crucial case - of Frege’s third objection.
’Vo1- M22, p. 38.
140 Frege: Philosophy ofMathematics
This leaves Frege without a proof that numbers are objects: only a strong
motive for taking them as such.

CHAPTER 12

Frege and Husserl

How sound was Frege’s strategy of definition? We have seen him left with a
motive, but no justification, for taking numbers to be objects; but perhaps the
motive was justification enough in itself. If number theory requires the exist­
ence of infinitely many objects, it is more appropriate to take those objects to
be numbers, as Frege did, than to presuppose, with Russell, the existence of
infinitely many non-logical objects. This apart, of all possible strategies, the
decision to treat numbers as objects rules out only one, the radical adjectival
strategy according to which the surface appearance of arithmetical statements,
with the apparent numerical terms they contain, is illusory; every other means
of connecting the adjectival and the substantival uses of number-words, every
other choice of which notions are prior and which to be defined in terms of
those, is left a possibility by that decision. Having read Grundlagen, one may
well believe that Frege has shown his definitional strategy to be mandatory';
but that only reflects his skill in presenting his task, at each stage, so that the
step he actually takes appears inevitable; in fact, he argues far less in favour
of the route he adopts than the reader has the impression that he does. Even
for the conceptual priority of cardinal equivalence over the notion of a cardinal
number, he argues, as we have seen, only in terms of a faulty analogy. The
justification of the strategy he follows is therefore best studied by considering
the objections of a critic. Seven years after Grundlagen appeared, Edmund
Husserl published his Philosophic der Arithmetik, in which Frege s book is
subjected to strong criticism, not on the ground that the definitional equival­
ences were false, but on the ground that they flouted the true relations of
Priority between the concepts involved. Three years later yet, Frege reviewe
usserl s book, vigorously retorting to his criticisms. The exchange provi es
a perfect basis for our enquiry.

Husserl’s objections ’dered as a definition

In Chapter VI of his book, Husserl first § 65’ LeibrUZ’S
°f identity, which is how Frege presents i t
 12 Frege and Husserl 143
142 Frege: Philosophy ofMathematics
! philosophical defence. Nevertheless, Husserl, in 1891, thought proper to
‘stands the true state of affairs on its head’, as he expresses it, borrowing
Frege’s own phrase.' The question is not one of the extensional correctness attack it.
of the definition, but of conceptual priority: the only ground for assuming the I.

replaceability' of one content by another in all true judgements, Husserl argues, The paradox of analysis
is their identity-; we are therefore not entitled to define their identity7 as consisting before, Husserl does not contest the extensional correctness of the defi­
in that replaceability. Husserl is in fact mistaken in contending tfiat Frege I
I. nition, or even deny that this extensional equivalence is a truth of logic: he
made the Leibnizian definition of identity’ basic to his construction of the
admits3 that ‘it lays down a necessary and sufficient condition in the logical sense,
concept of number; although Frege allowed it as a genuine definition, all that
valid in all cases, for the obtaining of equality’. But this is not enough, in his
concerned him was that the lanrs of identity' consequent upon it should hold,
view, for the definition to be admissible. The possibility of a one-one mapping
and, further, that the equals sign in mathematics be construed as denoting
of one plurality7 on to another does not, he says, ‘constitute their equinumerosity',
strict identity. That is why, in his review,2 he was able to concede that Leibniz’s
but only guarantees it’: ‘to know that their numbers are equal does not in the
law does not constitute a definition.
least require knowing that it is possible’ to map one on to the other, and so
Husserl’s remarks about Leibniz’s law sene to illustrate his concern, in a
‘the one piece of knowledge is in no way identical with the other’. Here
great part of his discussion, precisely with the question of conceptual priority,
Husserl is at the threshold of the paradox of analysis that so exercised G.E.
that is, with the question which of some pair of notions should be taken as
serving to explain the other. As we have seen, Frege, in his review7 of Husserl’s Moore: it is precisely the problem, which we left in abeyance, of the status of
book, ignores such questions altogether, blandly maintaining that the only those analytic definitions with which Grundlagen abounds, but of which, as we
criterion for a correct definition in mathematics is that it maintain the reference saw earlier, Frege failed ever to give a satisfactory account. The notion
of the defined expression. In Grundlagen, how7ever, he had given it a central expressed by sentences of the form ‘There are just as many Fs as Gs’ is a
place in his argument, contending, in § 64, that we must on these grounds commonplace one: in what sense is it to be analysed in terms of that of a one-
explain ‘The direction of a is the same as the direction of b’ as meaning the one mapping, as explained in detail by Frege in §§ 71-2? Surely someone can
same as ‘a is parallel to b', and that the converse order of explanation, defining understand the phrase ‘just as many’ without having the remotest idea of what
parallelism as identity' of direction, would ‘stand the true state of affairs on its a one-one map may be; how, then, can an explanation of his understanding
head. Plainly, extensional correctness is not here the consideration, either: by appeal to one-one mappings possibly be correct? Should we say that the
the correctness of the equivalence is taken for granted, the problem being the explanation brings to light what he tacitly, but not overtly, knows? Or should
proper direction of explanation. ‘\Miat is essential in this discussion can easily we, rather, say with Husserl that it merely sets out a logically necessary and
be transferred to the case of numerical identity’, he tells us in the first footnote sufficient condition, without penetrating to that in which his understanding
to § 65; we must therefore assume that Frege was tacitlv also contending that actually consists?
conceptual priority' required that we explain ‘The number of Fs is the same
as the number of Gs’ as meaning the same as ‘There are just as many Fs as
I
Counting
s, an not conversely. His answer to Husserl in the review was disingenuous,
at least as regards his intentions when he wrote Grundlagen. While Frege blurred the distinction between defining the cardinality' operator
in f^°eS °n *n ChaPter I to discuss the definition of equinumerosity7 in terms of equinumerosity7, and defining equinumerosity’ in terms oi one-one
of maPpmg. This definition, given by Frege in full in §§ 70-2 mapping, Husserl ignored it altogether; and this makes it difficult to discern
the precise intention of his criticisms. One natural interpretation is that he
wen 7S’’ ™ T TT‘With htai in § 63 he atlrib“,es i< <» as objected to the adjectival-to-substantival direction adopted by Frege tor mtrod-
1883 h Came >!' J” 18’2' SchrM<;7 “ 1873 a”d Ca”‘OT
Stole (1885) 7‘ . ?!r p » “ 'ar'y as 1874; Husserl Schroder and ucing the cardinality operator, believing that ‘There are just as mam s as
Gs' should be explained as meaning that the number of Fs and e nun\ eJ"
deU™ hl I F7^ G—« *e of Gs are the same rather than conversely. Husserl lists three means y w ic
piece of mathematical orthodoxy, though
Frege undoubtedly gave it its most equinumerosity ’ may be established. The first of these can be understood o y
—. exact formulation and its most acute
in the light of Husserl’s espousal of the abstractionist theory of numbers as

18,1, p. 104.
3 °P- cit., p. H4.
144 Frege: Philosophy ofMathematics
12 Frege and Husserl 145
sets of units demolished by Frege in §§ 33-44 of Grundlagen- in the fifth
section of Chapter Mil of his book, Husserl discussed Frege’s criticisms of Abstraction
the theory but it is hard to see how' he can have supposed that he had met If we construe Husserl’s argument as previously suggested, namely as favouring
them The first of Husserl’s three methods of establishing tw o pluralities to a definition of ‘There are just as many Fs as Gs’ in terms of the cardinality
be equinumerous begins with the psychological act of abstraction; having operator, viz. to mean ‘The number of Fs is the same as the number of Gs’,
reduced each plurality to a set of units, we may map the units in the first set his tactics were at fault. The whole point of the abstractionist theory' was to
one-to-one upon those of the other. The second method is much simpler: we explain how one arrives at the same number if one starts from any set of a
apply the operation of one-one mapping directly' to the original concrete sets, given cardinality: the number being a set of featureless units, attained by'
without any prior act of abstraction. This second method is, therefore, precisely abstracting from the characteristics of the elements of the original set, no
that which Husserl is denying to contain the very’ meaning of‘equinumerous’. difference can appear between any one such abstract set and another save how
It is, however, the third method which is preferable to either of the other two, many such units it contains. Husserl indeed concedes that the act of abstrac­
both as yielding more information and as corresponding to what ‘just as many’ tion, as performed in his first method of establishing numerical equivalence,
means: we count the members of each plurality, thereby determining not only does not yield the actual number of objects in each plurality'. Presumably this
whether there are just as many elements of the one as of the other, but, must be interpreted as meaning that, although we have attained the number,
specifically, the number of elements in each; ‘the simplest criterion for equality we do not yet know which number it is: we are in the position of Alice when
of number is just obtaining the same number when one counts the sets to be the White Queen tested her knowledge of addition by asking, ‘What is one
compared’. Husserl’s thesis closely resembles the answer that a child would and one and one and one and one and one and one and one and one and
give when first asked the question, ‘What does it mean to say that there are one?’. In this respect, however, Husserl’s theory’ is at no disadvantage as against
just as many nuts as apples in the bowl?’; almost any child will reply, ‘It means Frege’s, according to whose definitions ‘There are four horses in the field’ is
tantamount to ‘There is one more than one more than one more than one
that, when you count each of them, you will get the same number’.
horse in the field’. The act of abstraction by which we pass from a set to the
Frege has no difficulty, in his renew, in disposing of these objections. The
number of its members involves no reference to any one-one mapping, or
second method just is that of establishing a one-one mapping, w'hile the first
! indeed to a comparison of any kind of the original set with any other; and so
uses such a mapping indirectly. So, how’ever, does the third: ‘the author I it explains in what the number of members of the set consists 'without appeal
forgets that counting itself rests on a one-one correlation, namely between the
to such a comparison. Having arrived in this manner at the number of Fs and
number-words from 1 to n and the objects of the set’ (p. 319). This retort is
I at the number of Gs, there will be no need to correlate the units belonging to one
evidently wholly justified. Our sequence of number-words, as wre employ them
in counting, forms a kind of universal tally with respect to W'hich we can I with those belonging to the other, as in Husserl’s first method of establishing
numerical equality: for the two sets of units will be strictly identical.
compare the cardinalities of different concepts, and thus provides a means of
If the procedure of abstraction had really worked as it was supposed to do
pving, in the finite case, a specific answer to the question ‘How many are
on this theory, Husserl would have had a perfect rebuttal of the definition in
ere‘ ’ $PeHfic answers to questions of forms such as ‘How long?’, ‘How far?’,
terms of one-one mapping. One powerful argument for a thesis that one notion
ow eavyr, etc., always demand the adoption of a conventional standard of is conceptually prior to another is the possibility of defining the first without
comparison, the question ‘How many?’ appears to be no exception. Husserl reference to the second. Frege has to hand a definition of equinumerosity
obwusty has in mind only finite sets. To accommodate the case of infinite
independent of the cardinality operator, and he tacitly' denies the possibility
’ ere °re’ S ^es*s might be generalised to the following: to say that of defining the cardinality operator independently' of equinumerosity; it the
■llnw’m ’USr-aS ma?yFs as Gs is to say that a definite answer to the question, abstractionist theory had been sound, ‘the number of Fs' could have been
‘How manv r 310 ke the same as a definite answer to the question, explained without reference to equinumerosity, and the scores would have
not resolve theVffi ’ Gn^ortuitately for defenders of Husserl, this does been equal on this count. A second argument for conceptual priority is greater
that w’e can exnl °f *S aPPeal to the notion of equinumerosity simplicity: since it is obviously simpler to explain ‘the number of Fs without
many?’. W at constitutes a definite answer to a question ‘How considering any comparison of the Fs with the Gs, for some other concept G,
Husserl’s team would thereby have defeated Frege’s. Had it been possible to
sustain the abstractionist theory, such a victory could have been secured.
Why, then, did Husserl not adopt these tactics? He did not see the full
146 Frege: Philosophy ofMathematics 12 Frege and Husserl 147
force of Frege’s refutation of abstractionism; but he saw enough to grasp that doing that? Indeed, if, in view of the conceptual priority of the latter, it is
the units, although featureless, must in some way retain their distinctness, in necessary to define ‘the number of Fs’ in terms of ‘just as many as’ rather
the form of some shadow}’ remnant of the particular objects from which they than conversely, w hy is it legitimate to explain ‘there are two’ in terms of‘the
had been engendered; and so he spoke of correlating like-numbered sets of number 2’ and not conversely, when the former is evidently conceptually prior
units instead of merely registering that they are identical, without observing to the latter? It looks as though Frege invoked conceptual priority when it
that, with this concession, the entire abstractionist theory falls to the ground. suited his definitional strategy to do so, and ignored it when it did not.
The proper way to respond to Husserl’s criticism, thus interpreted, is to
enquire whether Frege’s strategy could have been reversed. Could he have
Another interpretation of Husserl started with definitions of the numerically definite quantifiers, proceeded to
There is, however, an alternative way of interpreting Husserl’s discussion, as define from them the corresponding cardinal numbers, and only then have
arguing, not for the definition of an adjectival notion, that of equinumerosity, defined the cardinality operator, perhaps explaining equinumerosity in terms
in terms of a substantival one, that of cardinal numbers, but for a reversal of of it? Frege strives so hard to make his actual strategy appear the only possible
the downward direction of Frege’s sequence of definitions, as set out in the one that we gain some insight into the explanatory force of his definitions if
table in Chapter 10. In particular, Frege’s official definitions explain each in this way we explore an alternative one.
numerically definite quantifier in terms of the cardinality operator and the As already remarked, there is no difficulty in defining as many of the
corresponding cardinal number: ‘There are two apples on the table’ is numerically definite quantifiers as we wish without invoking cardinal equival­
explained as meaning ‘The number of apples on the table is 2’. Since the ence or the corresponding cardinal numbers. There is, for example, no need
cardinality operator and the individual numbers are defined in terms of equinu­ to explain ‘There are denumerably many Fs’ as meaning that the Fs can be
merosity, this means that ‘there are just two’ and ±e rest are defined, ulti­ mapped one-to-one on to the natural numbers; we may simply define it to
mately, in terms of ‘just as many’; and Husserl may be understood as urging mean that the Fs can be enumerated.4 Accepting, for present purposes, Frege’s
that it should be the other way round. On this interpretation, Husserl’s princi­ identification of a cardinal number with the class of concepts having that
pal objection is not so much to defining cardinal equivalence in terms of one- cardinality, we could then define the number 2 as the class of concepts F such
one mappings, but in taking cardinal equivalence as the fundamental notion. that there are just two Fs, and so on for all other cases. Vlien we need to
In his Chapter VII, which includes an examination of Frege’s definition of the generalise over the natural numbers, we shall need the relation of successor:
cardinality' operator, he argues that the sense of an ascription of number to a this can be defined essentially in Frege’s way, but with no appeal to cardinal
set does not consist in classifying it with a group of numerically equivalent equivalence. We should define ‘h is a successor of ni’ to mean is the class
sets; we are concerned only with the set itself, and not its relation to other of Fs such that, for some .v, .v is an F and the concept “an F other than .v ’
sets. In this, Husserl is quite right. Frege, in the celebrated example of the belongs to »z’; and wre should then be able to define ‘is a natural number
waiter laying plates and knives on the table in Grundlagen, § 70, pointed out from the successor relation and the number 0 exactly as Frege does. With this
the possibility' of establishing that there are just as many things of one kind as in hand, w;e could with the help of a description operator define a restricted
of another without determining how many of each there are. Husserl is here cardinality operator ‘the finite number of Fs’ to mean the natural number n
pointing to the fact that it is possible to specify how many things there are of such that Fis a member of but this would of course be undefined w henever
a given kind without adverting to whether or not there are just as many as there wrere infinitely many Fs.
things of any other kind; more exactly, that it is possible to explain what is As long as our concern is solely with the arithmetic of the natural numbers,
meant } such a specification without adverting to the notion of cardinal Husserl’s thesis, when interpreted in this second way, is thus vindicated. Frege
equivalence. Both are right. It is not, indeed, possible to explain the procedure differed from Dedekind in believing that the natural numbers had to be
o counting save by appeal to one-one mapping: but each particular numerically presented as finite cardinals; but it is possible to present them as the ven
definite quannBer, including ‘there are dennmerably many may be classes of concepts with w’hich he identified them in Grundlagen without im ok-
iS fi"vd Wlth°“;?">'such aPPeal’ as Frege himself had shown in § 55 for all ing the relation of equinumerosity, or, therefore, the cardinality operator, t
If d‘e P°KlMi,y °f defining numerical equivalence without would be only at the point at which we wished to define the general operator
Z 1‘"T “ Iht qUMti<’n 'H°» ma">'?’is a good argument for
4 ‘The Fs can be enumerated’ must here be taken to mean that, for some x and some one-one
am lust „■, ?0SSMity °f definin® egressions of the form ‘there relation R, no j- stands in the relation R to x, and the Fs consist of those obiects ~ to which .
lust n without invoking numerical equivalence not a good argument for
stands in the ancestral R* of R.
 I
I
148 Frege: Philosophy ofMathematics
I
12 Frege and Husserl 149
‘the (cardinal) number of Fs’ that we should be forced to define equinumeros­ definition is in terms of there being a suitable mapping. Waismann’s objection
ity: Frege’s thesis that the second-level relation of equinumerosity is necessarily can easily be reformulated as being that Frege owed us a criterion for the 1
prior in the order of explanation to that of the (unrestricted) cardinality operator existence of relations, and that no such criterion can be framed without I ■

cannot be gainsaid. If, following him, we wish to identify all cardinal numbers circularity. The problem of the range of second-order quantification is indeed
with maximal classes of equinumerous concepts, we have no way of doing so a serious and difficult one. For Frege a sufficient condition for the existence •r.
save by first saying when two concepts are to count as being equinumerous. of a relation satisfying a given condition is the possibility of constructing a
Frege was therefore right that, if it is the arithmetic of cardinal numbers in I linguistic expression for such a relation, w'here any expression containing two II
general that concerns us, cardinal equivalence must be treated as the funda­ I argument-places for singular terms counts as standing for a relation. It needs
mental notion. The sense in which this is true comes through even more little reflection to see that, whenever there are finitely many Fs, and just the
clearly when we do not blindly follow Frege’s strategy of definition, but explore same number of Gs, sufficient knowledge w’ould in principle enable us to
a plausible alternative, one that respects relations of conceptual priority better construct such an expression, though, w'hen the number of Fs was sufficiently ♦
than does Frege’s own. But, equally, the sense in which it is not strictly true large, this would be practically impossible. Doubts may be harboured concern­
fails to appear when his strategy is followed: by setting the question ‘What is ing the notion of possibility in principle, which is one of intense philosophical
a (cardinal) number?’ at the outset of his enquiry, Frege causes us to overlook interest and difficulty; but it is a mistake to consider it a philosopher’s inven­
the fact that this question need not be answered in full generality if we wish tion. Asked what it means to say that the population of China is about
only to present the natural numbers as what he took the finite cardinals to be, 1,008,000,000, almost anyone would say that it means that, if you were able i;
and hence as serving on appropriate occasions to answer questions of the form i to count the inhabitants of China, you would reach approximately that figure; i
‘How many?’. I the possibility here referred to is possibility in principle. In the present case,
i there can be no harm in invoking the notion to explain the stated condition
I for the existence of a relation; it w ould be simple to give an inductive argument
The status of the definition
I to show’ that, whenever there wrere n Fs and n Gs, there existed a relation
We have, thus, two definitions, that of equinumerosity' in terms of one-one I mapping the Fs one-to-one on to the Gs.
mapping, and that of the cardinality’ operator in terms of equinumerosity. More In the infinite case, however, the matter is less simple. Here the linguistic
I
exactly, we have three, the intermediate one being the rejected definition, in I
criterion is certainly inadequate: in Russell’s famous example, given denumer- ■

terms of equinumerosity, of the identity of cardinal numbers. All of these ably many pairs of socks and of shoes, we cannot specify any mapping of the
definitions raise problems concerning their status: in this chapter, we may
consider only the first of the three. I
I
socks on to the shoes, although, intuitively, there are just as many of each.
The platonist characterisation of the totality’ of all relations over a given domain j
If equinumerosity’ (cardinal equivalence) is to be a fundamental notion, then is, however, notoriously questionable. It does not follow that there is any
I it must itself be defined if the project of deriving arithmetic from logic is to
be realised; and no definition has ever been proposed save that which was I
vicious circle. When quantifiers are understood classically, the Axiom of Choice
is said to hold good on a fully platonistic conception; we may, conversely,
already standard by the time Frege wrote Grundlagen. Apart from Husserl’s, construe it as giving a criterion for the existence of functions or of sets
very few objections to it have ever been raised. Waismann criticised it as (including relations construed as sets). One who declines to believe that there
circular, on the score that it would be too restrictive to say that there are just are in fact as many functions or sets as the Axiom of Choice requires may fall
as many Fs as Gs only if there actually is a one-one map of one on to the back on the notion of possible existence; from this standpoint, the Axiom of
other: we can claim only that, whenever there are just as many of each, i Choice tells us what functions or sets, and, in particular, what one-one map­
there could be such a mapping.5 There is, Waismann argues, no non-circular pings, there could be, so far as the cardinalities of other sets allow. It tells us,
explanation of the kind of possibility’ involved. What is meant is that there in other words, when there are enough members of each of two sets—enough
could be a mapping as far as the number of Fs and of Gs is concerned-, and thus shoes and enough socks, for example—for there to be a one-one map of either
the definition goes in a circle. i one to the other; and it does so without circularity. No case is known in which
The objection is readily answered. Frege invokes no modal notions: his I two sets have, intuitively, the same cardinality, and yet the Axiom of Choice is
5 F. Waismann, Einfiihrung in das mathematische Denken, Vienna, second edition, 1936, pp. 77-8, insufficient to prove their equivalence; so we may put Waismanns objection
I English translation by Theodore J. Benac, Introduction to Mathematical Thinking, New York, 1951, aside as not posing a genuine difficulty.
pp. 108-9.
We do much better to admit, with Husserl, the logical correctness and utility’
I
I
I
150 Frege: Philosophy ofMathematics 12 Frege and Husserl 151 M
of the criterion, and enquire into its credentials as an analytic definition. No
ki
number-words. Because we are all taught to count before we acquire any other
better example of such a definition could be sought: the notion defined is one concepts belonging to the theory of cardinality, this idea is likely to be new to
understood by all, and yet the definition is tar from obvious. Certainly it is the ordinary speaker. When he has considered it, he will probably make
not one that occurs to everyone immediately upon being asked what just as Waismann’s objection, that the numerical equivalence of one set with another
many nuts as apples’ means: it has already been observed that a child, who I guarantees that their elements could be correlated, but not that any such I.
certainly knows the meaning of the phrase, is likely to answer by talking about I correlation already exists. At this stage it will be necessary to convey to him
counting the nuts and the apples. He will, however, ordinarily recognise the the very broad sense in which the word ‘correlation’ is being used, and in
possibility of establishing that there are just as many of each by pairing them which a correlation is being said to exist. After a good deal of explanation, he
off, when it is pointed out to him. Can it be claimed that this definition fits may then be brought to agree that, whenever there are the same number of
the criterion of correctness given by Frege in ‘Logik in der Mathematik’, Fs as of Gs, a one-one correlation exists between them; but he will still deny,
namely that the equivalence is obvious to everyone who understands both the like Husserl, that that is what he has all along meant by the statement that
defined expression and its definition? the number is the same.
The best way to decide whether anyone could fail to recognise the alleged In ‘Logik in der Mathematik’, Frege requires that the defining expression
synonvmy is to enquire whether anyone has failed to recognise it. The answer should have the same sense as that defined; so we must ask after his criterion
is plainly affirmative. Husserl was one who failed to do so, and Waismann for two expressions’ having the same sense. About this he wavered somewhat,
another: it is striking how faithfully their objections echo those likely to be but his favoured criterion for the synonymy of two sentences was that anyone
made by an ordinary speaker. A third witness is Bolzano, w'ho observed that \
who understood both should recognise their extensional equivalence, i.e.
the real numbers in the open interval (0, 5) can be mapped one-to-one on to coincidence of truth-value.7 To interpret this consistently with Frege’s other
those in the open interval (0, 12), but denied that there were just as many in views on sense, we must require the recognition of equivalence to be immediate,
the former interval as in the latter, on the ground that it was a proper subset rather than consequent upon reflection; for reflection might include the recog­
of it.6
nition of a logical proof of equivalence, and to allow it would result in equating
The defence may impugn the testimony of these witnesses. It may move synonymy with analytic equivalence. For Frege, however, synonymy must be I
that Bolzano’s, for example, be set aside, on the ground that, regarded as an a much narrower relation; otherwise analytic judgements would not extend
expression of everyday speech, ‘just as many’ has a definite sense only as I
denoting a relation between concepts with finite extensions, and hence that our knowledge.
It was Frege’s general belief that anyone who grasps the sense of each of
Bolzano’s uncertainty was not about how' it does apply to the infinite case, but two expressions must know- whether or not their senses coincide. In the
i
about how- it ought to be extended to that case. And it is true that, while the
formulation referred to above, however, he does not require that anyone who
ordinary speaker would undoubtedly consider ‘Infinitely many’ to be an answer
understands two synonymous sentences should recognise them as synonymous,
to the question ‘How many.''’, he w’ould have no inkling how to decide whether
only as extensionally equivalent; presumably the subject might assign distinct
it is a definite answrer like ‘A hundred’ or an indefinite one like ‘At least a
senses to them on the basis of some faulty’ conception of what constitutes the
hundred. What, then, would be the ordinary speaker’s reaction to our ques­ !.
sense of a sentence. On this ground, Husserl might be dismissed as a witness
tion? Asked outright what it means to say that there are just as many Fs as
Gs, or, as he is likely to express it, that there are the same number of Fs as for the prosecution, since, although he denied that ‘There are just as many
Fs as Gs’ means ‘There is a one-one map of the Fs on to the Gs’, he agreed
of Gs, he will probably reply, like Husserl, that it means that, if you counted
that, on logical grounds, they are equivalent. Can the ordinary speaker’s
the Fs and also the Gs, you would get the same number. When it is pointed
out to him that you can sometimes tell that the number is the same without i testimony be dismissed for the same reason?
i The difficulty is that he does not start with a full understanding of the
ding out what it is, he will doubtless agree that a one-one correlation i defining expression; and the explanation needed to convey it to him involves
supplies a sufficient criterion for numerical equivalence, but will urge that
a great deal that would otherwise have been the product of reflection. He is
such an equivalence may obtain even when no correlation is to be had. To
surely right to resist the contention that the explanation merely brought to
is rege s retort to Husserl may be adduced, that counting consists in setting light w'hat from the outset he has understood ‘just as many’ as meaning: the
up a corre ation between the elements of a set and an initial segment of our I explanation has expanded his conceptual repertoire, and therefore cannot
•»>. H 20-1; MM. 7 See, for example, ‘Kurze Ubersicht meiner logischen Lehren’, 1906, Nachgelassene Schriften,
p. 213, Posthumous Writings, p. 197.

I
 I
152 Frege: Philosophy of Mathematics 12 Frege and Husserl 153
while someone’s failure to make them would count for us as showing that he
embody only what he always consciously took the phrase to mean. The real
fault lies, however, with Frege’s criterion of synonymy, which commits tlie error did not fully understand the expression; and if, thirdly, no rival definition is
of asking after the sense of the sentence in isolation from related sentences. A possible which has a better claim to capture the sense of that expression, as
speaker’s grasp of the sense of a sentence comprises an ability to perceive its commonly understood. To this extent, analysis is necessarily reconstruction;
connections with other sentences which are close neighbours in the web of but not to an extent justifying Frege’s counsel of despair in ‘Logik in der
language; asked, out of context, to explain the sense of a sentence, he may Mathematik’, that we should abandon the everyday expression in favour of
well overlook some of those connections, perhaps by asking himself how he some newly introduced technical term.
could tell that the sentence was true in a given case, and omitting to ask what, Thus, when the criterion of the correctness of definitions is their faithfulness
knowing it to be true, he would inter from it. An analogous oversight is likely to our everyday understanding of the defined terms, the question whether the
to occur if you ask someone how ‘inefficient’ and ‘incompetent’ differ in
f proposed definition of a single term is correct lacks a sharp sense. We obtain
meaning: you will probably get a subtle distinction, but not the remark that a sharper sense, how’ever, when we ask after the correctness of a system of
the construction ‘incompetent to do such-and-such’ is admissible, while the definitions of a range of interconnected terms. This is because our ordinary
parallel construction with ‘inefficient’ is not. Suppose a child has been taught judgements about an individual’s understanding of an expression obey no clear
to count, and, asked to say how many cakes there are on a plate, correctly principle. We say that someone understands an expression, or that he fully
replies, ‘Twelve’. If his mother now says, ‘Good: there are just enough to go understands it, if he knows concerning it what most speakers of the language
round’, referring to the forthcoming children’s party, the child will hardly know, and is able to use sentences containing it, and closely related sentences,
count as understanding what the statement ‘There are twelve cakes on the as competently as do most speakers. This leaves principles governing the
plate’ meant if he proves not to grasp the idea of their going round among meanings in the language of some words as a matter for experts, of which a
twelve children. In such a case, he has merely been trained as a counting­ speaker may be ignorant without forfeiting the claim to understand those
machine, but does not understand the ascription of number which results from words: hence Putnam’s division of linguistic labour. It also parcels out the
the operation of counting. ‘There are just enough to go round’ is one wray of knowledge required for a perfect master}' of the language in an uneconomical
saving, ‘There is a one-one correlation’; so Frege was not far astray, after all, way. In a systematic account of the knowledge needed for total mastery, we
in incorporating the notion of a one-one correlation into his analysis of ascrip­ should divide it without overlap into what determined the meanings of the
tions of number. As Husserl saw, we can explain what ‘There are twelve ..." various individual w’ords, together with the principles governing phrase- and I.
means without alluding to equinumerosity; in practice, we should not regard 1
sentence-construction: we should not assign the same piece of knowledge to
anyone as understanding the phrase if he failed to perceive that from ‘There
are twelve Fs’ and ‘There are twelve Gs’ it can be inferred that we can associate
the Fs with the Gs in such a wray that just one F is associated with each G.
I two different words, as knowledge a speaker needed to possess if he was to
know the meaning of either. We should make our account as economical as
possible, by not repeating the same item, which a speaker needed to know if
he was to have perfect mastery of the language, as governing the meaning of
A speaker’s understanding of an expression is thus only incompletely
revealed by what he says when asked to explain it; moreover, there are no more than one word or device of sentence-formation: it would be sufficient
precise principles determining the distribution of the items comprised in his that everything one had to know for perfect knowledge of the language as a
knowledge of the language among his component understandings of those of whole should, on the account wre w'ere constructing, be comprised in the
its distinct component expressions that are closely related to each other. For knowledge determining the meaning of some word or form of expression. If
these reasons, there is often no determinate answer to be given to the question we followed our intuitive judgements about what is required for someone to
whether a proposed analytic definition does or does not capture the ordinary be said to understand a given expression, on the other hand, we could not
sense of the defined expression. Analysis makes explicit connections which we conform to such a maxim of economy: we should have frequently to reckon
make but are not, out of context, conscious of making; it supplies, for principles the same piece of linguistic knowledge as determinative of the meanings of
to which we have been trained to conform, formulations we do not have the different words.
present vocabulary’ to give. An analytic definition can therefore be required to What holds good for a hypothetical account of the knowledge required for
o no more than to come as close as possible to capturing the existing sense. a complete mastery of a language holds good also for a system of definitions
1 It does so if, first, any ordinary speaker can be brought to agree that it provides of interrelated terms such as Frege’s definitions in Grundlagen of basic arith­
a necessary and sufficient condition for the application of the expression metical expressions. Even when, as in Frege’s case, the expressions in question
defined; if, secondly, it appeals only to connections which we make in practice, are already in use, and one aim of the definitions is to be faithful to their
154 Frege: Philosophy ofMathematics
ordinary senses, it suffices, for that system of definitions to be correct, that it
comprises everything that must be implicitly known by anyone w ho understands
all those expressions: the system is not to be judged by how it parcels out that
knowledge among the defined expressions, and hence it is no criticism that
the definition of any one expression does not furnish everything that a speaker
would ordinarily be thought to need to know to be acknowledged as under­ CHAPTER 13
standing that expression. Particularly is this so when, as in Frege’s case, the
system of definitions is intended to sen e as a basis for deductive proofs, and I
t
hence, in particular, when these are mathematical definitions, no matter how
widely used in everyday contexts be the expressions they serve to define. Such
Frege's Definition of Cardinal Numbers
a system of definitions serves, not merely to make explicit what is implicit in
our everyday understanding, but to systematise it. This gives a further reason
why analytic definitions involve reconstruction - which yet is in no way arbitrary'.
The derivation, by means of a definition belonging to such a system, of some What Frege thought was wrong with the contextual definition
individual statement from the laws of logic alone is therefore unreliable as an We may call the proposition that ‘The number of Fs - the number of Gs’ is
indication of the logical necessity of that statement, as ordinarily understood; equivalent to ‘There are just as many Fs as Gs’ the ‘original equivalence’. We i
but this is no defect when the purpose is like Frege’s, to demonstrate the have seen that the supposed contextual definition of the cardinality operator,
analytic character of an entire theory. consisting of a stipulation that the original equivalence is to hold, and discussed
If, in devising such a system of definitions, we conform to the Bolzano/Frege at length by Frege in §§ 63-7 of Grundlagen, is in reality no definition at all.
maxim to prove whatever is capable of proof, we may find ourselves with some It is not, how'ever, exactly for this reason that he rejects it in §§ 66-7, but
freedom to choose what to allocate to the definitions and what to the theorems I
rather because it does not determine the truth-value of every’ admissible
immediately derivable from them: the contents of such theorems may legit­ sentence containing the operator. Specifically, it fails to determine the truth­
imately be among the things which intuitively form part of the meanings of value of any sentence of the form ‘The number of Fs = q\ when the term lq’
the terms defined. Thus, if the abstractionist account had been viable, it would is not itself of the form ‘the number of Gs’. Such sentences, Frege insists,
have been possible to prove from the definitions based on it that, if two sets A
cannot be set aside as inadmissible: for any two singular terms, the statement
and B have the same number n of members, there will be a one-one map of
of identity connecting them must have a sense, and so any legitimate definition
A on to B. For let N be the unique abstract set of n featureless units, and let
must provide a determinate truth-value for it. In particular, we have noted that
<p and ip be the restrictions to A and B respectively of the operation of the purported contextual definition fails to provide any means of eliminating the
abstraction. Then <p maps A one-to-one on to N, and ip maps B one-to-one
cardinality operator from a sentence of the form ‘The number of Fs = x\
on to N: so the composition of <p with the inverse of tp maps A one-to-one
where ‘x’ is either a free or bound variable. If it had been a genuine definition,
on to B. But this, while rendering more plausible Husserl’s contention that
it would have done so; in that case, Frege’s objection to it w'ould not have
one-one correlation is no part of the meanings of number-words, would not
arisen. His objection is, however, stronger than just that the stipulation fails
conclusively vindicate it: the content of such a theorem may well be something
to provide a method of eliminating the operator from all contexts: it faults the
ffiat must be known, at least implicitly, by anyone w'e should admit to have a
full understanding of such words. stipulation for failing to pass a less stringent test. We could express the test II
as that of determining the reference of terms of the form ‘the number of Fs’.
A pair of recursion equations for addition or multiplication fails, within a il
first-order language, to provide any means of eliminating the addition or !I
i multiplication sign from all contexts; but it uniquely determines the function I
it denotes. Frege’s objection is to the effect that the stipulation he has been
I considering fails uniquely to determine the mapping of concepts on to objects
I
effected by the cardinality operator.
The context r _r„7 r
principle, presented in § 62 as furnishing the essential key to
the problem how numbers are given to us, rules out as spurious all problems
155
 )

156 Frege: Philosophy ofMathematics 13 Frege's Definition of Cardinal Numbers 157

about what an expression stands for that cannot be expressed within the the truth or falsity of a sentence like ‘The number of planets is Julius Caesar’, !•
language, or, in other words, that cannot be stated as questions about the and thereby fails to determine the references of numerical terms. We may
truth-value of some sentence of the language containing that expression — perhaps take it for granted that sentences of this kind are to count as false.
about what its truth-value is, or at least how' it is determined. W e cannot, as We may urge that in practice no one will confuse the number of planets with
it were, stand in thought outside our language, and mentally apprehend the Julius Caesar; but, as Frege remarks in § 66, that is no thanks to the attempted 'i

reference of the expression; and so it is no defect of a given manner of contextual definition. This means that we have not attained a unique specifi­
introducing the expression into the language that it does not enable us to cation of the reference of numerical terms formed with the cardinality operator:
establish such an extra-linguistic mental association of expression and referent. since we have failed to make any stipulation determining whether or not Julius
Grasping the reference of an expression just is grasping certain principles Caesar is the number of planets, we have not said what the number of planets
governing the determination of the truth-values of sentences of the language is, that is, what the term ‘the number of planets’ stands for.
containing it. All legitimate questions about the reference of a newly introduced Why should this matter? A first inclination might be to say that it is necessary
term 7’ will therefore be ones that can be framed within the language. We to say what is not a cardinal number, as well as what is one, if we are to
can legitimately ask whether the object for which T stands is of a given kind, generalise about cardinal numbers. For example, if we merely wish to show
say an organism, because this is just to ask whether the sentence 7 is an that certain specific tasks can be effectively performed, we may rest content
organism’ is true; and it is to be answered by appeal solely to the principles with laying down some merely sufficient conditions for an operation to be
that have been laid down for determining the truth-value of such a sentence. effective; but, if we wish to prove some proposition concerning all effective
We can legitimately ask whether ‘f has the same reference as some other term operations, or if, in particular, we wish to show that some task cannot be
7’, because this is just to ask whether the sentence 7 is the same as s’ is true; effectively performed, we need to have a necessary as well as sufficient con­ ■

and the question is to be answered in the same way as the previous one. But dition for the effectiveness of an operation. So, likewise, it might be thought,
that is all. If all those questions can be answered successfully, then the term if we wish to prove something to hold good of all cardinal numbers, or that ! i
has a reference. There is no further test it can be required to pass, such as there is no cardinal number satisfying a given condition. But this was not at
proriding us with a means of imagining its referent, or a possibility of encoun­ all Frege’s motivation. Having given his explicit definition of the cardinality
tering or contacting it: all nominalist objections on scores such as these spring operator, which supposedly overcame the problem by uniquely determining
from the rice of considering the meaning of the term in isolation. the truth-value of every sentence of the form ‘The number of Fs = q', Frege
By the same token, however, we have not fixed the reference of the term uses that definition, as we have seen, solely to derive the original equivalence,
until w'e have supplied the means of answering all those questions of this kind and, from that, the equivalence between ‘The number of Fs = the number of
that are formulable within the language. This involves, in particular, that, to Gs’ and ‘There is a one-one map of the Fs on to the Gs*. No further appeal ■

have fixed the reference of the term, we need to have laid down determinate is ever made to the definition of the cardinality operator, because it is not
conditions for the reference of the newly introduced term T to coincide with
that of any other given term in the language: in other words, for the truth of
needed for the proof of any proposition in which Frege is interested. To prove
that there is a cardinal number satisfying a given condition, it suffices to show
iI
any identity-statement formed by putting T on one side of the sign of identity
and any arbitrary term of the language on the other. It may or may not be
that there exists a concept F such that the number of Fs satisfies that condition;
to prove that every cardinal number satisfies some other condition, it is enough
i
obvious that our intention, in introducing T, included its not being taken to to show that, for every concept F, the number of Fs satisfies that condition. iI II
stand for the Moon; but, unless we have provided for the falsity (or, if we That could have been done on the basis of the original equivalence, perhaps
wish, for the truth) of the sentence 7 is the Moon’, we cannot claim to have laid down as an axiom, together with the definition w'hich Frege gives of‘n is
fixed the reference of f, since we have not stipulated whether or not it stands a cardinal number’, namely as meaning ‘For some F, n is the number of Fs’.
for the Moon. That it is sufficient to rely on this latter definition, along with the original
That is the fault that Frege finds with the purported contextual definition equivalence, of course depends on which conditions we are concerned to show
of die direction-operator. It affords us no means of determining the truth or
falsity of a sentence like ‘The direction of the Earth’s axis is England’; and,
some cardinal number, or all cardinal numbers, to satisfy. If Frege had wished
to establish whether any cardinal number had crossed the Rubicon, he w'ould
I
in failing to do so, it fads to determine the references of terms for directions. have had to enquire whether Julius Caesar wras a cardinal number; but, of
, . ,at> implication, is the fault he finds with the purported contextual course, he wished to establish only those propositions fundamental to arith­
definition of the cardinality operator: it affords us no means of determining metic. For this purpose, it was sufficient to know only two things about cardinal

I
15 8 Frege: Philosophy ofMathematics
numbers: (1) that something is a cardinal number just in case, for some F, it
13 Frege's Definition of Cardinal Numbers
were expressed in a language all of whose terms possessed a reference, and
159 11
is the number of Fs; and (2) that the number of Fs is the same as the number hence all of its sentences a truth-value; its rules of inference could be acknowl­ !
of Gs if and only if there is a one-one map of the Fs on to the Gs. edged as valid only if they governed such a language. Frege did not aim at
Frege’s ground of objection to the contextual definition he discussed at such completeness in his formal theory; but the stipulations that lay down the I
length was thus not that he needed, in order to earn’ out his programme of references of the primitive expressions had, for him, to be complete in the I
deriving the theory of cardinal numbers from purely logical principles, to sense of providing a determinate reference for every term formed by means
determine the truth of any sentences of the form The number of Fs = q, of them, if the formal proofs were to be relied on as incontestably valid.
where was not itself of the form 'the number of Gs : it was simply that he It is evident that Grundlagen leaves unfinished business to be dealt with: not
considered it essential to secure a determinate reference for every’ expression just details, but a fundamental matter. The explicit definition of the cardinality­
in his theory, and, in particular, regarded it as a requirement upon a legitimate operator enables us to determine that Julius Caesar is not a cardinal number
definition that it specify a unique reference for the term defined. This attitude only if we are able to determine that he is not the extension of a second-
is plainly apparent in Part I of Grundgesetze. A good example is his treatment level concept, or, in the revised Grundgesetze version, of a first-level one. In
of his description operator ‘\’. This he treats formally as a first-level operator, Grundgesetze, the extension of a concept is a special kind of value-range.
to be attached to singular terms. The axiom governing it, Axiom VI, merely Unlike the notion of a cardinal number, that of a value-range is not, however,
lays down that, when u is a unit class whose sole member is a, then \ u is a. introduced by definition; if it had been, this would only have pushed the
In expounding the semantics of the formal system, on the other hand, Frege problem back one step further. The task must therefore be assigned to the
is careful to provide for the case in which u is not a unit class;1 in this case, semantic stipulations which provide the interpretation of the abstraction oper­
\u is to be u itself. This stipulation is not needed for proving anything in the ator: they must be so framed as to determine whether Julius Caesar is a value­
formal theory that Frege needed to prove; if it had been, it would have been range, or, more generally, the truth or falsity of any identity-statement having
incorporated into the axiom, as it could easily have been. It is not needed a value-range term on one side and a term of any other kind on the other.
because Frege never attaches his description operator to any term that cannot This is precisely the problem Frege faced, and attempted to solve, in § 10 of
be proved to denote a unit class; he nevertheless believed that his semantic Grundgesetze, Volume I. On its successful solution depended for him the
interpretation of the formal language had to provide a reference for all well- legitimacy’ of his use of terms for value-ranges, and hence of his entire concep­
formed definite descriptions. tion of logical objects and how they are given to us, and thus of the ontological
Why should he have thought it necessary, or even worth w’hile, to supply status of numbers of all kinds. Russell’s contradiction forced him to acknowl­
stipulations that were not to be embodied in the formal axioms, and hence edge that the stipulations in §31 of Volume I did not suffice ‘to secure a
never appealed to in any’ formal proof? The reason lies in the very purpose of reference in all cases’ to the terms of his symbolic language, as he admitted
carrying out proofs within a formal system. The primary advantage of the .
in his first reply to Russell of 22 June 1902. The question raised at the
formal system is not the possibility of effectively verifying that the proofs are beginning of § 62 of Grundlagen, how numbers are given to us, was, by Frege’s
I
formally’ correct; it is that xve can be certain that they are valid. We can be own lights, only partially answered in that book. The essential problem still >.
certain of this because, starting from sound fundamental law’s of logic, they remained, and could not be airily dismissed by saying, ‘I assume that it is
proceed in accordance only with valid rules of inference. We recognise that known what the extension of a concept is’, as Frege tried to do in the footnote

h
these logical law’s are sound and these rules of inference valid only by seeing to § 68 of Grundlagen-, on the contrary, wrestle with it as he did, he failed in
that the former are true and that the latter transmit truth from premisses to the end to solve it, and his failure led to the disaster of the contradiction.
conclusion. If not every’ w’ell-formed expression has a reference, then not every
formal sentence will have a truth-value. The rules of inference will then
sometimes lead from true sentences to those devoid of truth-value; since there Criteria of identity
is no guarantee that they will never allow us to derive false sentences from This, then, was Frege’s motive for abandoning the attempt to give a contextual 1 ■

those that are devoid of truth-value, we shall not be able to rely on formal
definition of the cardinality’ operator and adopting his explicit definition in : ■

deductions beginning with true premisses not to arrive at false conclusions.

terms of classes. The introduction of the notion of classes was a disastrous
The axioms of the formal theory could be acknowledged as true only if they
step, destined to bring his w’hole enterprise down in ruins. Frege appears to
have had some inkling of its perilous nature, since in Grundlagen he refrains
' Vol. I, § 11.
from invoking the notion save in this one place; by the time he came to write
 160 Frege: Philosophy ofMathematics 13 Frege's Definition of Cardinal Numbers 161
in a one-sorted formal theory, we took the direction-operator as primitive,
Grundgesetze, he had unfortunately lost those qualms. A proposal of Wright’s
there would be nothing to prevent us, when giving the intended interpretation
would render the step unnecessary, so far as the theory of cardinal numbers
of the formal language, from dividing the domain into directions and other
was concerned; on a strong reading of this proposal, the so-called contextual
definition already resolved the Julius Caesar problem, and already determined elements, the former being denoted by direction-terms and the latter by terms
that no cardinal number could be identified with anything not explicitly given
I of other forms. There would, indeed, be no obstacle to incorporating such a
as a cardinal number.2 Frege should therefore, on this view, have contented stipulation into the formal theory itself: we need only add a primitive predicate
himself with adopting the contextual definition, and thus have spared himself ‘D(£)\ with an axiom of the form ‘D(dir(a))’ and other axioms guaranteeing
the disaster that overtook him. Wright does not suggest how he could have that lD(t)’ shall always be false when t is a term of any other form.
avoided appeal to classes in his theory of the real numbers. Of course, if Even so, Frege’s remarks might be defended on the ground that such a
Wright’s argument is correct, Frege’s explicit definition was not merely segregation of directions from all other objects could not be accomplished by
unnecessary, but actually mistaken, because it does identify cardinal numbers anything properly called a definition. However this may be, it is not the stronger
with objects not explicitly given as such, namely with ones given as extensions but the weaker thesis that troubles Wright. The weaker thesis is that, given
of certain second-level concepts. That definition yields the original equivalence that the direction-operator has been introduced by means of the so-called
as a consequence; but, according to Wright, if that equivalence had been contextual definition, stipulating the equivalence between 'dir(fl) = dir(£)’ and
adopted as a definition, or at least a stipulation, it would have ruled out Frege’s is parallel to b\ it will still be legitimate subsequently to stipulate further
definition of numbers as extensions of concepts in advance. that a direction-term denotes the same object as a term of another kind.
Frege’s answer to the suggestion that the contextual definition already con­ Crispin Wright’s belief is that the contextual means of introducing the direc­
tains a solution to the Julius Caesar problem - or, rather, that the correspond­ tion-operator rules out any such identification; by contrast, Frege’s argument,
ing contextual definition of the direction-operator contains a solution to the on the weaker interpretation, is that such an identification is left as an open
problem of England and the Earth’s axis - is given in § 67 of Grundlagen. possibility, and that therefore something explicit must be done if we are to
exclude it. On the face of it, Frege must be right. We have seen that it is
If we were to try saying: q is a direction if it is introduced by means of the perfectly consistent with the stipulated equivalence to treat direction-terms as
definition set out above, we should be treating the way in which the object q is denoting particular fines: how can it be maintained that the stipulation in fact
introduced as one of its properties, which it is not. The definition of an object
does not really predicate anything of it, but only lays down the meaning of a prohibits any such identification?
symbol. When this has been done, it can be converted into a judgement that does Did Frege’s criterion of identity for directions leave him free to identify
treat of the object; but it now no longer introduces it, and stands on the same them with classes of lines? Wright’s principle, stated for directions, is:
level as other statements about it. If one were to choose this way out, one would
be assuming that an object could be given in only one way; for otherwise it would (LF) S is a sortal concept under which directions fall only if there are singular terms ‘s’
not follow from the fact that q was not introduced by means of our definition that and ‘t’ denoting S’s such that the truth-conditions of‘s = t’ could adequately' be
it could not be so introduced. All statements of identity such as equations would explained as those of some statement to the effect that certain lines are parallel.
therefore come down to saying that what is given to us in the same way should
be recognised as the same. This is so self-evident and so unfruitful, however, that
it is not worth while stating. In fact, we could not derive from it any conclusion Frege’s whole point was that the criterion of identity for the directions does :■ I
that differed from all of our premisses. Rather, the significant and manifold utility not settle whether a direction can be a country; Wright, appealing to (ZX),
of equations rests on the fact that we can recognise something as the same again
although it is given in a different way.
thinks that it settles that it cannot. Now take 5 as class, and ‘s’ and ‘t’ as both
of the form ‘the class of lines parallel to the line r’: does 5 satisfy the condition
fI
stated in (EX')? Wright wavers, but remarks that, if we construe ‘adequately'
Frege is here refusing, on general grounds, to take the easy way out by simply explain’ strictly, a negative verdict must be given: we have also to know that
stipulating that no term formed by means of the direction-operator shall be
classes are identical iff co-extensive.
taken as standing for an object denoted by a term of any other kind. His words Wright’s discussion rests on his adopting, and ascribing to Frege, a ‘picture
are ambiguous between a strong thesis and a weaker one. The strong thesis of the natural numbers as unique, indiependent, determinate objects (p. 115).
is that any such stipulation would be illegitimate; and it is that thesis which Frege indeed believed that we ought to identify them with such objects: he did
f e appears to need for his conclusion. As such, it appears unjustified. If, even not think that our existing manner of speaking of them already determined with
’ C. Wright, Frege’S Conception ofNumbers as Objects, Aberdeen, 1983, pp. 113-29, 178-9. which objects we must identify them.
The oddity of Wright’s position is that the legitimacy of a stipulation de­
pends upon whether it follows or precedes another. The eccentricity' of one
 162 Frege: Philosophy ofMathematics 13 Frege's Definition of Cardinal Numbers 163
ellipse coincides with that of another just in case they are similar. I his follows standing for proto-properlies. Among these feature-placing predicates are some
from a definition of eccentricity as the ratio between the distance between the with two argument-places, as when, using two pointing gestures, we say, ‘This
foci and the length of the major axis; but, on Wright’s view, that definition is darker than that’; these may be said to stand for proto-relations.
would be illegitimate if we had first stipulated the condition for the eccentricity It is from this basis that we advance to the next higher level of language, at j
I
of two ellipses to be the same: the geometrical criterion would preclude the which we refer to and quantify over objects; and it is at this higher level that
identification of the eccentricity with a real number, with which is associated we have use for the first time for the concept of identity. When a learner of
a quite different criterion of identity. the language is first introduced to the notion of an object of a given primary
Frege’s introduction into philosophy of the notion of a criterion of identity kind - a kind not characterised as a species of some already familiar genus -
embodied a profound insight. This is that, whenever we speak of objects of the relevant criterion of identity must be conveyed to him. This will normally
anv kind, we must have in the background a principle for determining what be done by example. The necessity for his grasping that criterion is most easily
is to count as the same object of that kind. We are speaking of an object seen in the case of words such as ‘letter’ (in the sense of‘letter of the alphabet’,
whenever we use a singular term, and also whenever it would be in place to not of ‘epistle’), which are ambiguous between different criteria of identity:
call for a singular term, by means of a question of the form ‘Which one?’. the sense in which there are twenty-six letters in the English alphabet differs
First-level quantification presupposes a domain of distinguishable objects, as in just this respect from that in which ‘letter’ is a six-letter word. At this first
do the corresponding expressions of generality in natural language: we cannot stage in the introduction of sortal concepts, and thus of a means of referring
intelligibly ask whether there is anything in the room that was made in Hong to and talking about objects, the associated criteria of identity may also be
Kong unless it is determinate whether or not something formerly made in said, in a rough sense, to coincide with those for the obtaining of an equivalence
Hong Kong can be identified with anything now in the room. It is part of relation: but, of necessity, it is only a proto-relation, not a genuine relation
Frege's idea that the criterion of identity for objects of a given kind is not a
between objects, since we are not yet in a position to handle expressions for
consequence of the way that kind of object is characterised, but has to be
genuine relations. Admittedly, the criterion whereby we judge that the second
expressly stipulated as part of that characterisation. This is of course not true
letter in the word ‘letter’ is (in one sense) the same as the fifth might be said
when the characterisation itself used the concept of a more general kind of
object - a sortal concept, in the standard terminology - for which it is already to coincide with that for the obtaining of an equivalence relation between
determinate what is to count as the same object falling under that concept. letters in another sense; and the criterion for the identity of letters in that
When we introduce the concept of a prime number, we do not have to lay sense to coincide with that for an equivalence relation between letters in the
down when p is to count as the same prime number as q\ it is already provided sense in which we ask after the number of letters on a printed page; and the
that it xvill be so counted just in case it is the same number as q. But as we criterion of identity for letters in this last sense’ with that for an equivalence
ascend the hierarchy of ever more general kinds of object, we must reach one relation between w'hat Nelson Goodman would call ‘letter-inscriptions’. But
that was not characterised as a species of some genus; and, when the concept this does not correspond at all to how we in fact learn to apply those criteria
of an object of this kind was introduced, the relevant criterion of identity must of identity’; and, in any case, we virtually never talk or think about letter­
i inscriptions (though we might say, ‘The letter “B” has fallen off the sign
have been expressly stipulated.
In §§ 63-8 of Grun(Hagen, Frege is concerned with a large class of cases in “.MEMBERS ONLY” ’).
i We may call objects whose associated criterion of identity was not explained
which the criterion of identity for objects of a certain kind, such as directions,
shapes or numbers, consists in the obtaining of a certain equivalence relation in terms of an equivalence relation betw’een objects of another kind (or between
between objects of another kind, or, in the case of numbers, between concepts. concepts) ‘primary objects’, and those w'hose criterion of identity is so explained
There is, however, a more basic case: if there were not, we should have no ‘secondary’ ones. Wright is certainly correct in supposing that wre should not
conception of objects at all, since we should not be able to appeal to an admit any transsortal identification of primary objects, any identification of one
equivalence relation between objects of another kind or between concepts primary object with another having a quite different criterion of identity. We
under which objects fall. In this more basic case, we employ what Strawson have no use for any such identification, and should for an excellent reason
called feature-placing predicates, attached, not to a proper name or other resist it. Objects of different kinds admit different predicates: in the sense
f term denoting an object, but to demonstrative expressions indicating, vtith only in which there are twenty-six letters in the alphabet, we may ask after the i
rough precision, a presently observable region of space; since such predicates, origin of the letter ‘J’, but it would be senseless to ask after the origin of
when so used, are not being applied to objects, we may describe them as the third letter of the word ‘jejune’. If we admitted transsortal identification
I
 I
164 Frege: Philosophy ofMathematics 13 Frege's Definition of Cardinal Numbers 165 i

objects of a given kind into the language. Nevertheless, Frege’s position is !

of primary objects, confusion would ensue; and tor this reason we do not even
envisage it. sound: in the case of secondaiy objects, nothing stands in the way of our
For secondary’ objects, the matter stands differently. They are standardly identifying them with objects given in some other way, provided that we
referred to by the use of an operator, like Frege s direction-operator, attached respect the criterion of identity. Within natural language, we seldom make such
to a term for an object of the kind over which the equivalence relation is I
transsortal identifications: in more systematic types of discourse, they are
defined: the terms formed by means ot this operator may be called fundamen­ ! extremely common. Very many secondary objects may be identified with real
I
tal terms’, and the domain of the equivalence relation the fundamental numbers, whether within some interval such as [0, 1] or otherwise: the eccen­
domain’. Since secondary’ objects are identified by reference to a fundamental tricity of an ellipse, the probability of an event, the length of a vector in
domain, they cannot be picked out by straightforward ostension. If someone Euclidean //-space, and so on. Frege, indeed, would not regard such identifi­
points, saying, ‘That book’, the context may not be sufficient to dispel uncer­ cations as arbitrary: since all represent a ratio of one kind or another, they
tainty about the criterion of identity he has in mind, but, this apart, there can conform to a common principle. A quite different kind of identification is with
be no unclarity about what he is referring to. If, how ever, he points and says, a representative element of the fundamental domain; this was already illustrated
‘that shape’, the ambiguity does not turn on the criterion of identity' for shapes: by die identification of the direction of a line with a line through the origin. r
the question may need to be asked, ‘The shape of what?’. His ostension will Any such transsortal identification will make available a new range of predicates
have worked only if there happened to be a salient object in the direction of applicable to the objects so identified, in addition to the fundamental predi­
the pointing gesture. For this reason, there is not the same obstacle as with cates. But, unlike w’hat would happen in the case of primary objects, the
primary objects to transsortal identification. possibility of applying these new predicates, which may or may not remain a
To grasp a sortal concept covering secondary' objects of a given kind S, two possibility only in principle, will cause no confusion.
conditions must be known. First, we must know what constitutes the fundamen­ It appears at first sight as though, in rejecting the contextual ‘definitions’ of
tal domain. Knowing this, we shall know to which expressions the operator the direction-operator and the cardinality operator in favour of explicit ones,
may be applied in order to be sure of obtaining a fundamental term denoting Frege is betraying the principle he laid down in § 62 concerning criteria of
an object of kind 5; and so the circumscription of ±e fundamental domain identity. That principle apparendy required that the condition for the truth of
may be said to constitute the criterion for the existence of objects of the kind such an identity-statement as ‘The direction of a is the same as the direction
5. Concerning directions, we must ask: Do all lines have directions? Is it only of b' must be stipulated outright, not derived from our prior understanding of
lines that have directions? Concerning numbers, we have to ask: Does every' ‘is the same as’, together with a definition of the direction-operator; and yet,
concept have a number belonging to it (for instance, the concept ‘red’, or, it is in the latter way that Frege eventually proceeds. He defines ‘the direction
again, the concept ‘cardinal number’)? If not, what restrictions must be imposed t of a' as denoting the class of lines parallel with a: from this, we are led to
on the predicate lF(£)’ if the term ‘the number of Fs’ is to stand for anything? derive the condition for the identity of directions from that, which we already
The second essential piece of knowledge wre need is the criterion of identity.
know, for the identity of classes. Specifically, it follows direcdy from our
This will be given by some equivalence like Frege’s original equivalence for
definition that the direction of a is the same as that of b just in case any line
the identity of numbers, or the analogous equivalence for the identity of I c is parallel to a if and only if it is parallel to b, and indirectly, in virtue of the
directions; it will serve to equate the condition for the identity of the secondary' I

fact that parallelism is an equivalence relation, that this will be so just in case
objects with that for the obtaining of a certain equivalence relation between
elements of the fundamental domain. a is parallel to b.
It is evident from the foregoing discussion, however, that the inconsistency
These twTo conditions — the criterion of existence and the criterion of identity is no more than apparent. What Frege objects to is the idea that w’e could
- form the sole preparation we need for speaking of objects of the kind 5. Of
define ‘« is parallel to b' as meaning ‘The direction of a is the same as the
course, in order to speak of them, we need to know some predicates that can direction of F; and this, of course, he does not do, but, rather, uses the relation
be applied to them; the introduction of such predicates is straightforward.
of parallelism to define the direction-operator. His thesis is thus not that the
They will be defined or explained in such a way that the two criteria comprise
meaning of an identity-statement connecting two terms formed by means oi
all the information concerning objects of the kind 5 necessary to determine
the operator to be introduced must be stipulated outright, but, rather, that the
their application to an object denoted by a fundamental term: a predicate so
equivalence relation over the fundamental domain must be already understood
defined or explained may be called a ‘fundamental predicate’. For essential
purposes, this is all that is needed in order to introduce reference to secondary
i before the operator can be defined. That equivalence relation is to be identified
by asking after the condition for the truth of such an identity-statement.
 166 Frege: Philosophy ofMathematics
The equivalence between the identity-statement and the obtaining ol the
equivalence relation between the two given elements ol the fundamental
domain - the relevant original equivalence — will then provide a criterion for
the correctness of any proposed definition of the operator: such a definition
will be correct only if it yields that equivalence. The definition must provide CHAPTER 14
a way of determining the truth-value of an identity-statement connecting one
of the new terms with any other term whatsoever. In so doing, it may equate
the reference of a term of the new kind with that of one of a different kind:
all that is demanded of it is that it respect the criterion of identity embodied
in the original equivalence, and that it admit the operator as defined for every
The Status of the Definition
argument for which wre want it to be. In arriving at a definition of such an
operator, the identification of the relevant criterion of identity is therefore an
essential first step; very often the eventual definition will appeal to the equival­
Logical abstraction
ence relation in terms of which that criterion is formulated. But this is all that
the doctrine of criteria of identity requires: it does not demand that the original The passage from § 62 to § 69 of Grundlagen is the most important in the
equivalence incorporating the criterion of identity be itself the subject of a book. It contains the kernel of Frege’s whole logicist philosophy of arithmetic,
direct stipulation. In § 68 Frege proposes a uniform method for defining any of and it is there that its weakness is to be seen, as well as its strength; and
a whole class of operators, including the direction-operator and the cardinality therefore we are not finished with it yet. As was observed in the last chapter,
operator, namely by taking it as forming terms for equivalence classes. If wre the method Frege adopts for giving an explicit definition of the direction­
assume it to be already known what a class is, as Frege expressly says that he operator, and likewise an explicit definition of the cardinality operator, namely
is assuming it to be known, this is a case of equating the referent of a term to define directions as equivalence classes of lines and cardinal numbers as
of the new kind with that of one already understood, that is, of transsortal equivalence classes of concepts, is presented by him as highly general. So
identification; it is a mistake to suppose this to involve any violation of the indeed it is, and has become a standard mathematical device used by everyone;
doctrine of criteria of identity as Frege intended it to be understood. but Frege wras one of the very first to isolate it as a general device, and to
Frege undoubtedly took it for granted that there is nothing problematic perceive its wide applicability. If wTe wish to introduce a new' type of object,
about our referring to primary objects such as mountains, trees, people, cities but not as a subspecies of some already familiar type, and can formulate the
and stars, or about the determinateness of reference of the proper names and criterion of identity for objects of this new type as the obtaining of some
other singular terms by means of w'hich wre refer to them. It is obviously true equivalence relation between objects of some already known kind, this method
that an adequate theory of the mechanism of reference, and even an adequate enables us to identify the new objects as equivalence classes of the old ones
exposition of the concept of a criterion of identity, must treat of primary under that equivalence relation. Very often, the most natural way of forming
objects; but Frege, at least in his capacity as philosopher of mathematics, ought I
terms for the new objects will be by means of an operator/to be attached to
not to be reproached for not hating said enough to determine howr such an a term a for one of the objects in the fundamental domain; we shall then
account should go. The notion of a criterion of identity, as he introduced it, define this operator by setting each such new term fid) formed by means of it
is applicable to singular terms for objects .of every kind, as he said, which is
as standing for the equivalence class to which the referent of a belongs.
why it is wrong to represent it, as some have done, as relating only to logical
This device has since been labelled, not very happily, ‘definition by abstrac­
objects or to those that are not actual. It remains that, in Grundlagen, he was
tion’. Frege would have disliked this terminology; but he was fully aware
concerned directly only with its application to secondary objects, and we can that the device accomplished, in a legitimate wray, w'hat others attempted to
hardly cavil at his restricting his discussion to them in a book about the
accomplish by means of the operation of psychological abstraction. Both types
foundations of arithmetic. of abstraction aim at isolating what is in common between the members of
any set of objects each of which stands to each of the others in the relevant
equivalence relation: Frege’s logical method by identifying the common feature
with the maximal set of objects so related to one another and containing the
given objects; the spurious psychological operation by deleting in thought

167
 168 Frege: Philosophy ofMathematics 14 The Status of the Definition 169
everything except that common feature. In an important respect, it is a matter (Bl) There are just as many Fs as Gs
of regret that Frege hit on this device, since it prompted him for the first time (B2) The number of Fs is the same as the number of Gs.
to introduce classes into his logical system, and so led eventually to die
catastrophe of Russell’s contradiction; yet the invention of what we may call We may call this the ‘synonymy thesis’. Someone who held that (Al) is just a
‘logical abstraction’ was a highly significant contribution on his part to die disguised way of saying what (A2) says, and (Bl) a disguised way of saying
logically rigorous practice of matiiematics without the intrusion of appeals to what (B2) says, and hence that (Al) should be defined as meaning (A2), and
psychological operations, let alone to spurious ones. (Bl) as meaning (B2), could of course cheerfully endorse the synonymy thesis:
Frege was not in a position to recognise one difference, important to us, there could, on that view, be no objection to it. (A2) can undoubtedly be
between the second-level cardinality operator and the various first-level oper­ regarded as stating that a certain relation obtains between the lines a and b,
ators to which he compares it. A definition by logical abstraction effects a so that it is perfectly in order to introduce an abbreviation for the expression
partition of the domain of the equivalence relation: if the lines on the Euclidean of that relation. Likewise, (B2) can unquestionably be regarded as stating that
plane form a set, then each of the equivalence classes of lines on the plane, a certain second-level relation obtains between the concepts F and G, so that
under the relation of parallelism, is a set. But, when cardinal numbers are it is again quite in order to introduce an abbreviation for the expression of
defined, after the manner of Grundgesetze, as equivalence classes of sets under that second-level relation. It was, however, precisely against this direction of
the relation of equinumerosity, then, by the standards of von Neumann- explanation that Frege set his face: he insisted that the second member of
Bemays set theory, every' cardinal number other than 0 will be a proper class, each pair should be explained in terms of the first, and not the first in terms
since its union will be the universe. In standard set theory, therefore, Frege’s of the second. Is the synonymy thesis consistent with this view of the order in
cardinal numbers could not themselves be members of classes, and his proof w'hich it is necessary to explain them?
of the infinity of the natural-number sequence would be blocked: that is why On the face of it, it is not. (A2) contains terms for two lines, and it is
cardinal numbers, in standard set theory, are not defined in Frege’s way, but therefore uncontroversial that it may be viewed as stating a relation between
as sets each representative of its cardinality, comprising all ordinal numbers lines; (B2) contains two predicates, and it is therefore uncontroversial that it
of lower cardinality. may be viewed as stating a second-level relation between concepts. But (Al)
contains no terms for directions, and (Bl) no terms for numbers, and hence
The status of the original equivalence neither appears to admit an analysis as a statement of identity between direc­
tions or between numbers. The only way in which we can so construe (Al)
In § 64 of Gnindlagen Frege wrote: and (Bl) is by regarding them as disguised ways of expressing (A2) and (B2);
The judgement ‘The straight line a is parallel to the straight line b', in symbols
i and this is precisely what Frege denies.
a // b, can be regarded as an identity-statement. If we do this, we attain the
I The only alternative way of defending the synonymy thesis appears to be to
concept of a direction and say, ‘The direction of the straight line a is identical maintain the opposite, namely that (A2) is not to be taken at face value, but
with the direction of the straight line b'. We thus replace the symbol // by the construed as an idiomatic way of expressing (Al), and likewise for (B2) and
more general symbol = , by distributing the content of the former symbol to a (Bl). This, however, also runs counter to Frege’s evident intentions. If it were
and to b. We split up the content in a way different from the original way, and correct, there could be no objection whatever to the proposed contextual
thereby obtain a new concept.
definitions: they would explain the sense of (A2) and (B2) in the most direct
This wray of characterising the transition appears to commit Frege to holding manner possible. But then (A2) and (B2) wrould not really be identity-state­

I that the judgeable content of the two sentences

(Al) a is parallel to b
(A2) The direction of a is the same as the direction of b
ments at all, but merely idiomatic sentences disguised as identity-statements;
and the terms for directions and for numbers that occur in them would not
be genuine singular terms, but only what Frege calls in the Appendix to
Grundgesetze sham proper names. The whole point, however, is that Frege
intends them to be taken as genuine terms, standing for objects, and subject
to all the logical operations, invoking quantified sentences and the identity­
coincides: the two sentences have the very same content, or, in the terminology
sign, that real singular terms obey: it is just for this reason that he finds the
of his middle period, express the same thought. Clearly, this is also meant to
apply to the pair with which Frege is really concerned, namely
contextual definitions w'anting. This option is therefore likewise closed, there
seems no consistent way in which the synonymy thesis can be maintained.
 14 The Status of the Definition 171
170 Frege: Philosophy ofMathematics
in their case. For pair (D), by contrast, Frege apparently held that the direction
Nevertheless, other examples occur in Frege’s writings. In Gnindlagen, § 54,
- of explanation runs from (2) to (1): in this case, it is from the implausibility
he says that an ‘affirmation of existence is nothing other than denial of the
number nought’: this appears to imply that I of the claim that the problem stems.
Why, then, did Frege assert the identity of content of the far from evidently
synonymous members of all these pairs? One reason lies in the fact that it is
(Cl) Dwarf elephants exist
(C2) The number of dwarf elephants is not 0 II a consequence of his preferred criterion for the identity' of the thoughts
expressed by two different sentences. This criterion is that anyone who grasps
form a similar pair. Perhaps more significantly, he says in § 57 that ‘one can the thought expressed by the one sentence and that expressed by the other
convert the sentence “Jupiter has four moons” into “The number of Jupiter’s must immediately recognise either as true if he recognises the other as true;
moons is four” ’; although he does not state specifically what remains invariant a criterion better expressed by saying that he must immediately recognise that
under this transformation, the suggestion apparently is that the former sentence both must have the same truth-value. As we have seen, to accord with Frege’s
may really be construed as being a disguised form of the latter, and hence that intentions, the immediacy must be stressed: if time for devising a proof of
they form yet another pair: either statement from the other were to be allowed, w'e should have merely a
criterion for the analytic equivalence of the two sentences, which, for Frege,
(DI) Jupiter has four moons is a weaker relation. Frege never explicidy applies this criterion save to prove
(D2) The number of Jupiter’s moons is four. that two expressions do not have the same sense, as he does in Function und
Begriff itself; but, since he claims it as a sufficient as wrell as necessary condition
More important than these is a remark in the lecture Function und Begriffof for identity of sense, it may be presumed to have influenced him in supposing
1891. In that lecture, Frege explained the changes in his formal and philosophi­ that the pairs (A) to (E) consist of synonymous forms of sentence. There is
cal logic that he had made during the silent years that separated his early from no doubt that their synonymy follows from the criterion. Anyone who under­
his middle period. In particular, he explained the introduction into his formal stands both (Bl) and (B2), for example, must straightaway recognise them as
logic of the new notion of a value-range, where a value-range is to a function equivalent. About the pair (E), our reluctance w'ould reside in not admitting
as a class is to a concept. The fundamental principle governing value-ranges the presupposition that every function has a value-range, but this was never
is that which was to be embodied in the celebrated (or notorious) Axiom V of Frege’s uncertainty: even the modification to Axiom V proposed in the Appen­
Grundgesetze, stating the equivalence between the generalised identity ‘For dix to Grundgesetze maintains this presupposition, merely allowing that two
every a, fla) = gfctf and the identity-statement ‘The value-range of f is the functions will have the same value-range if they have the same value for every'
same as that of g'. It has frequendy been observed, with perfect justice, that argument other than that common value-range. If Frege’s criterion for identity
this equivalence is formally analogous to the original equivalence between (Bl) of content is correct, then, modulo Frege’s beliefs about value-ranges, all the
and (B2) discussed, and rejected, in Grundlagen, as a possible contextual pairs (A) to (E) consist of two sentences with a single content.
definition of the cardinality operator. Frege was, of course, perfecdy clear that In ‘The Limits of Intelligibility: a Post-verificationist Proposal’,1 Christopher
the equivalence involving value-ranges could not rank as a definition of the Peacocke declares that w’hat he calls ‘Frege’s Principle’ is indisputable. He
abstraction operator used to form terms for value-ranges. Nevertheless, he formulates the principle thus:
says, in Function und Begriff, that a particular sentence stating the co-extensive-
ness of arithmetical functions ‘expresses the same sense, but in a different Content p is identical with content q just in case: necessarily any rational thinker
way* as one stating the identity of their value-ranges (pp. 10-11). We may judges that p iff he judges that q.
therefore take as our final pair
From left to right, the principle, as so formulated, is indeed indisputable: it
(El) For every a,ffa) = g(a) follows immediately from the laws of identity'. From right to left, it is somewhat
(E2) The value-range off = the value-range of g. ambiguous. Frege formulates the principle as one determining the identity of
foe contents of two sentences, and includes the condition that the thinker should
These five pairs are not, according to Frege, entirely analogous. For the grasp the content of both sentences. It is by no means indisputable that it
I
pairs (A), (B) and (E), the direction of explanation runs from (1) to (2); this follows from the necessity that one who does so grasp the content of both
is evidently true also of pair (C): and it is this fact that creates the problem Christopher Peacocke, Philosophical Review, vol. XCVII, 1988, pp. 463-96; see p. 471.
 172 Frege: Philosophy ofMathematics 14 The Status of the Definition 173
should recognise them as having the same truth-value that the content of each plainly required for an understanding of a sentence of that form: it w’ould
is the same. For this, it is also necessary that anyone who grasps the content therefore follow that more is required for the understanding of a sentence
of either should thereby grasp the content of the other: not, indeed, that he than a knowledge of its content. No hint of any such distinction is present,
should know that it is the content of the other sentence, but that he should however, in the writings of Frege’s middle period, nor, indeed, in the less
be acquainted with the thought it expresses. This requirement makes the systematic discussions of his early period. Only two features, not constitutive
principle of little use in determining that two sentences are indeed synonymous; I of the thought an assertoric sentence expresses, are ever allowed by him as
Frege surely revealed a sure instinct by employing his principle only to negative required for an understanding of the sentence: the assertoric force attached
effect, to demonstrate non-synonymy. to it, distinguishing it from an interrogative or other utterance; and the tone
It follows from Frege’s criterion for identity of content that all five of our attaching to the words or phrasing, which evokes associated ideas or affects
pairs consist of two sentences expressing the same thought. And yet this the hearer’s expectation of what is coming next, but is irrelevant to a judgement
conclusion runs counter to intuition. The reason is that someone could well of truth or falsity. To grasp the thought expressed by a sentence of the form
understand the first sentence of any pair without being in a position to under­ (A2), we must indeed recognise that it is true or false according as the
stand the second. Of course, it in no way detracts from the claim that two corresponding sentence of the form (Al) is true: but, if this is all that is known,
sentences are synonymous that someone might understand one without under­ the sentence (A2) may be no more than an idiomatic w'ay of expressing (Al), its
standing the other: he might simply not know the w'ords. In cases (A) to (E), constituent singular terms only sham proper names in positions not admitting
however, he might understand the first member of a pair without being as yet variables bound by quantifiers. Are we, then, to say that, when only this much
capable of grasping the thought expressed by the second, because he was not is known, the content of the sentence has been grasped, although more is
as yet in possession of a concept constitutive of the content of that second needed for a grasp of the thought it expresses? There is no warrant for any
sentence: for instance, because he lacked the concept of a direction or of a such distinction; but, in any case, it does no justice to the data of the problem.
cardinal number. That, indeed, is what Frege implies when he says that, by By treating the content of a sentence as not exhaustive of the thought it
splitting up the content in the new way, we thereby attain a new concept. Yet
expresses, it allows that, when w’e have grasped the content of (A2), wre have
this possibility is patently ruled out if the content of the one sentence is the still not attained the concept of a direction; and it thus leaves unexplained
same as that of the other: if it is the same, everyone who grasps the content
how' w'e are supposed to come by that concept by splitting up that very' content
of the one by that very act grasps that of the other, even if, through his
ignorance of the w ords occurring in it, he does not know that it is the content in a new w'ay.
of the other. Frege at no time proposed a distinction between the sense of a sentence
and its content that would allow' us to hold that, w'hile a grasp of the sense of
How' is it possible to escape from this dilemma while still maintaining
a sentence of, say, the form (B2) involved having the concept of a number,
Frege’s criterion for identity of content? One route would be by maintaining
which a grasp of the corresponding sentence (Bl) did not, still the content of
that, contrary' to appearances, a grasp of the concept in question is required
for a grasp of the content of the first member of the pair: for instance, that a the tw'o sentences coincided. In Begriffischrifi, however, he did maintain a
grasp of the concept of a direction is required for a grasp of the content of a distinction between the content of a sentence and the way in w hich that content
is regarded — the pattern we discern in it. One and the same design may be
sentence of the form (Al) as well as for a grasp of that of one of the form
(A2). But that w'ould be implicitly to maintain that (Al) is no more than a seen either as an array of white circles on a black background or as an array
disguised form of (A2), and hence must be explained precisely as being of black Maltese crosses on a white background: the design remains the same,
equivalent to (A2); and that reverses the order of explanation on which Frege and w'e perceive it as remaining the same, w'hile wre organise it now in one
so strongly insists. Besides, to hold that anyone who grasps the content of (Al) way, now in the other. It was precisely by adopting a new way of regarding an
must already have the concept of a direction would contradict his thesis that i already given content, namely by hitting on a particular way of dissecting it
F it is by making the transition from (Al) to (A2) that we first acquire that into one or more variable parts and a constant part, that Frege held that we
concept. This way out of the dilemma therefore appears to be blocked. can arrive at a new’ concept. This, therefore, is surely the model for Frege s
The only alternative is to hold that a grasp of the concept is not required contention that it is by splitting up the content of (Al) or of (Bl) in a new’
for a grasp of the content of the second member of the pair: that a grasp of way that we attain the concept of a direction or of a number.
e concept of a direction is not required for a grasp of the content of a To grasp the content of ‘Cato killed Cato’, we do not have to ha\e the
sentence of the form (A2). But how can this be? A grasp of the concept is concept of suicide: wre have only to know the content of the name Cato and
"'hat it is for a person x to kill a person y. We arrive at the concept of suicide
 14 The Status of the Definition 175
174 Frege: Philosophy ofMathematics
of content. When this distinction is made, the content of a sentence may be
by considering both occurrences in the sentence of the pioper name Cato as
taken as determined by what results from any transformations of it licensed
simultaneously replaceable by another name, say Brutus , and so apprehending
by the linguistic devices it contains. Since such transformations convert both
the pattern common to that sentence and the sentence Brutus killed Brutus’.
•Cato committed suicide’ and ‘Cato killed himself’ into ‘Cato killed Cato’, the
Having done so, we can give a definition, laying down that a committed
suicide’ is to be equivalent to killed a'. Io understand the sentence Cato I content of all three sentences, so understood, will be the same: but the senses
of all three will differ, because they involve distinct linguistic mechanisms. In
committed suicide’, we have to have the concept of suicide, and implicitly' to
particular, a grasp of the sense of a given sentence may require possession of
know its definition; that is, we must know' that Cato committed suicide is
a concept not required for a grasp of the sense of another sentence with the
true if and only if‘Cato killed Cato’ is true. The same indeed holds good for
an understanding of a sentence containing a reflexive pronoun: you understand same content.
the sentence ‘Cato killed himself’ only if you are able, with the additional This serves to explain how Frege could have come to think that the two
premiss ‘Whoever killed Cato was a scoundrel’, to deduce ‘Cato was a scoun­ members of each of the pairs (A) to (E) had the same content; but it does not
drel’. vindicate that contention. The analogy' between our attainment of a concept
We may refer to such expressions as reflexive pronouns as ‘linguistic like that of suicide, or that of the continuity of a function (Frege’s favourite
devices’: a linguistic device, in this sense, is an expression or phrasing an example), is a false one. The process by which we come to view ‘Cato killed
I
understanding of which consists in the grasp of a principle whereby any Cato’ as saying, of Cato, that he killed himself, is explained in the Begriffsschrift
I
sufficiently simple sentence involving it is equivalent to some other sentence in terms of a linguistic operation, and could be explained in no other way. If
not involving it. The qualification ‘sufficiently simple’ is required because ‘A we have any conception of distinct occurrences of the content of the name
Roman senator killed himself’ is not equivalent to any ordinary' English sen­ ‘Cato’ within the judgeable content, or, in the later terminology', of the sense
tence not containing either the reflexive pronoun or the word ‘suicide’: but of the name within the thought the sentence expresses, or any conception of
we understand it because we know that it follows from a sentence such as what it would be to replace one constituent of the content or of the thought
‘Cato was a Roman senator and Cato killed himself’, and also know ‘Cato by some other, it can only be by analogy' with the linguistic expression and its
killed himself’ to be equivalent to ‘Cato killed Cato’. In a language employing components and with operations upon it: in drawing this analogy, we rely upon
the quantifier/variable means of expressing generality, the qualification would Frege’s principle that the composition of the sentence reflects the composition
be unnecessary': every' sentence containing a reflexive pronoun could be trans­ of the thought. In order to grasp the content of the sentence, it is unnecessary
/ formed into an equivalent one not containing it. Linguistic devices, in this that the possibility of any particular way of dissecting it should have occurred
sense, include definable expressions like ‘suicide’, an understanding of which to us: but that possibility is intrinsic to the structure of the sentence, and
is tantamount to knowing their definitions. Frege’s explanation of this mode of concept-formation depends upon this fact.
!
An understanding of a sentence containing a linguistic device obviously i By contrast, no similar operation on (Al) or (Bl) can exhibit it in the form
requires a know ledge of how that device works. The functioning of the device of an identity-statement connecting two singular terms: no singular term is to
may therefore reasonably be regarded as a constituent of the sense of the be discerned in either sentence. We can base a definition of the phrase
sentence. So regarded, the sense of the sentence ‘Cato committed suicide’, or ‘committed suicide’ upon our dissection of the sentence about Cato: but no
of‘Cato killed himself’, is more complex than that of‘Cato killed Cato’, since definition can be framed that will effect a transformation of (Al) into (A2) or
it requires a knowledge of the working of one or other linguistic device. This, °f (Bl) into (B2). Precisely that wras the negative conclusion ot Frege’s exami­
however, conflicts with the thesis, which appears intuitively obvious and which nation of the proposed contextual definition that would stipulate outright that
Frege often asserted, that the sense of a defined expression coincides with I (A2) w'as to be equivalent to (Al). We may justifiably speak of the transition
that of the expression used to define it. This thesis implies that the sense of from sentences of the form (Al) or (Bl) to those of the form (A2) or (B2) as
the sentence Cato committed suicide’, equated by Frege with the thought the a process of concept-formation: but it is a different process from that w'hich
! sentence expresses, coincides with that of‘Cato killed Cato’. Plainly, the thesis leads to the concept of suicide, to that of continuity', or to other concepts
holds good only under a coarser application of the notion of sense than that explicitly definable by complex predicates of first or higher level. When, as in
according to which a grasp of the sense of a sentence containing the word cases (B) and (E), the transition involves the introduction of an operator of
suicide or the w-ord ‘himself’ requires an understanding of that word. Oddly second level, it depends upon a possibly dangerous ontological assumption.
enough, Frege gave no sign of ever having noticed this; but it provides a Sentences of the form (Bl) or (El) make no demands upon the domain of
ground for distinguishing between a finer notion of sense and a coarser notion objects over which our individual variables range. By contrast, the passage
 176 Frege: Philosophy ofMathematics 14 The Status of the Definition \T]
from (Bl) to (B2) demands recognition of the domain as containing at least 'ust as many Fs as casc, any claim to have captured the meaning
denumerably many objects, while that from (El) to (E2) makes an unrealisable 1 ttachcd to phrases of the form ‘the number of Fs’ by ordinary speakers of
demand upon the size of the domain, as Frege learned to his cost. Had (E2) ! the language would be palpably unjustified. To such speakers, the very idea
really had no greater content than (El), Frege’s Axiom V, which states their [ of a class of concepts, or even of a class of classes, would be remote; even
equivalence, would have been unassailable; and then it could have given rise when it had been explained to them, none would agree that the class of
to no inconsistency. concepts equinumerous to the concept F was just what he had been intending
In Grundgesetze, Frege makes no such claim: he nowhere suggests that the to refer to when he spoke of the number of Fs.
thoughts expressed by the two sides of Axiom V are identical. We must How, then, did Frege see the definition? With what right did he offer it in
presume that by 1893 he had come to acknowledge to himself that the thesis the course of a purported demonstration of the analyticity of arithmetical
which he had so vividly expressed in Grundlagen for the pair (A), and by truths? It is here that we come upon a conventionalist strain in Frege’s thinking.
implication for the pair (B), and had in 1891 extended to the fundamental pair He believed that, to enable rigorous logical proofs to be given, rigorous defi­
(E), had been an aberration incompatible with his other doctrines. It clashed, nitions were necessary, and, further, that the common understanding of estab­
in particular, with the doctrine, constantly repeated by Frege during his middle lished expressions did not always fully determine how they were to be rigor­
period, that the sense of part of a sentence is part of the thought expressed ously defined. A rigorous definition had, in particular, to specify uniquely7 the
by the whole. This doctrine means nothing if it does not mean that a grasp reference of the expression defined. If it was a predicate or functional
of the thought depends on a grasp of that constituent sense. To grasp the I expression, this required it to be defined for every object as argument; if it
thought expressed by (B2), one must have the concept of a cardinal number, was a singular term, it must be assigned a unique object as referent. But
or, in other words, must grasp the sense of the cardinality operator. To grasp common usage did not always serve to circumscribe the application of a
the thought expressed by (Bl), one need never have attained the concept of predicate or functional expression to every7 possible argument; nor, for a
a number: the sense of the cardinality operator is no part of that thought. It singular term, did it always provide a criterion for whether or not the referent
follows that Frege’s criterion for identity of content is defective: it is without of that term should be identified with a given object or not In such a case,
doubt a necessary7 condition, but certainly not a sufficient one. That Frege the sense attached to the expression by ordinary7 speakers was defective: when
never publicly acknowledged that he had been wrong to maintain, for these it became necessary to supply a rigorous definition of it, the defect had to be
five pairs, the same content for both their members, is no proof that he did remedied. The definition must respect the ordinary sense of the expression
not change his mind: he was never very good at confessing past errors. defined, to the extent that it was determinate; but, in remedying the deficiency7,
/ there was nothing to which we must hold ourselves responsible, and hence
f Frege’s definition of numbers the remedy might be chosen with a view only to convenience.
Frege makes this attitude quite plain in § 69 of Grundlagen. He comments:
Grundlagen is full of analytical definitions: definitions of expressions in common
use, the apparent purpose of which is to capture the senses which they bear That this definition is correct will perhaps be hardly evident at first. Do we not
j are commonIy used. We have already scrutinised one of these: the think of the extension of a concept as something quite different.'
definition of numerical equivalence, that is, of the binary quantifier ‘There are
k\aSfim-any " * ‘ as • • • • that case, doubts could be raised about whether He goes on to remark that a standard form of statement concerning the
the definition could be claimed as rendering explicit something at least implicit extension of a concept is that it is wider than that of another, that is, that one
in any ordinary7 speaker’s understanding of the phrase ‘just as many’. There class includes another. After observing that no number, defined as he has
f was no doubt, however, that every such speaker could be brought to recognise,
defined numbers, can include any other number, he concedes that there may
at least for the finite case, the equivalence of defining and defined expressions. be a case in which the extension of the concept ‘equinumerous to the concept
he definition could be said to impose greater systematisation on what an F was more or less inclusive than the extension of some other concept, which
ordinary speaker grasps only hazily; but it could not be accused of importing
could not then be a number on his definition. He admits that ‘it is not usual
With Frege’s definition of the cardinality operator, to call a number more inclusive or less inclusive titan the extension of a
S “ C°ns,rM 111 Krms f°7 ordinal numbers, the
concept’. But this is not for him a decisive objection: he retorts that ‘there is
ioXiata ds,d'ffereH'He defl"“,the numbe7 °f F1’ bf his mtthod °f
logical abstractton, and hence as ’the class of concepts G such that there are nothing to prevent us from adopting such a way of speaking, if the case shou
happen to arise’.
178 Frege: Philosophy of .Mathematics 14 The Status of the Definition 179
Precisely the same attitude is displayed by Frege when, later in the book, expression, how could an appeal to it be justified in demonstrating the
he is discussing briefly how we ought to go about introducing numbers of alyticity of some statement, as ordinarily understood? The answer is obvious.
other kinds - rational numbers, real numbers, complex numbers and so on. If the proof of the statement does not depend essentially on the features of
This section is principally devoted to an attack on what he here calls ‘formal­ he definition that have been arbitrarily chosen, it will turn only on those
ism’, but would be better called ‘postulationism’: the idea that we need do no features that were responsive to the sense of the expression as it is ordinarily
more than to lay down the laws that we want tire numbers in question to obey used; and then the statement, as ordinarily understood, will have been
and the conditions that we want them to satisfy, and may then assume their adequately demonstrated. Precisely this is how it is with Frege’s definition of
existence without more ado. This, he says in § 102, is to ‘proceed as if mere (cardinal) number. As we already saw', the definition serves only two purposes:
postulation were its own fulfilment’; it does not even establish the consistency to fix the reference of each numerical term uniquely; and to yield the original
of the postulated assumptions. Even if it did, that would not be enough to equivalence. In deriving the arithmetical laws by means of the definition, only
justify the appeal to an auxiliary mathematical theory' to prove a proposition the original equivalence is appealed to; no other feature of the definition plays
belonging to another. In § 101 he invokes his favourite example, the derivation the least role. It is the original equivalence that embodies the received sense
of the formulae for the cosines and sines of specific multiples of d from de of the expression ‘the number of ... ’; hence, if we grant the legitimacy of
Moivre’s theorem that defining all terms for numbers by the use of that operator, Frege’s demon­
stration of the analyticity of those laws is in no way impugned by the admitted
(cos 0 + i sin 0)" = cos nd + i sin nd. partial artificiality of the definition he gives of it. Benacerraf s problem simply
does not arise for Frege. He can happily assert all four of the following
Here a theorem involving complex numbers is used to prove formulae involving propositions: (i) that the laws of arithmetic can, by means of definitions, be
only real numbers. To guarantee the truth of those formulae, the mere consist­ derived by purely logical means from the fundamental laws of logic; (ii) that,
ency of the theory' of complex numbers does not suffice: we want to know' that in giving those definitions, we must be faithful to the received senses of
the formulae are true, not merely that they are consistent. The example could arithmetical expressions; (iii) that our definitions must completely fix the ident­
be better chosen: the point is cogent. ity’ of the natural numbers as specific objects; and (iv) that the received senses
How, then, does Frege believe that the complex numbers ought to be of numerical terms do not impose any one specific identification of the natural
introduced? His discussion of this, in § 100, is almost light-hearted: but a numbers.
serious purpose lies behind it. He directs attention, first, to the fact that we
are not starting with already given meanings of the signs of addition and
multiplication, as applied to the complex numbers; rather, ‘the meanings of
the words sum ’ and “product” are extended simultaneously with the intro­
duction of the new numbers’. He then describes w'hat he regards as the proper
procedure as follows:
i
We take some object, say the Moon, and define: let the Moon multiplied by itself
. e . . ’ e . n ba'e> >n the Moon, a square root of-1. This definition appears
legitimate, since from the previous meaning of multiplication nothing follows
concerning the sense of such a product, and hence, in extending this meaning,
we may make arbitrary' stipulations. We also need, however, the product of a real
number and the square root of-1. Let us, then, choose instead the temporal
rb™H°f °tne ,se.cond [° be a S(luare root of-1, and designate it by i. We can
we interval of 3 seconds’ etc- Which object shall
this case? ^'13 What meaning is to be given to the plus symbol in

If, in such a case, f^4 de

' ^' n'don had certain features due only to the choice
of ±e one giving ft. definition, and wilg~
I sense of
 15 Did Frege Refute Reductionism? 181
ation, it >s cited in § 62 simply because Frege wants to explore the possibility
of defining die cardinality operator contextually, in order to show where the
suggestion is at fault. Once the idea has been abandoned, and the correct
solution adopted, the context principle has no positive part to play.
If we take §§ 63-9 by themselves as an enquiry into the proper definition
CHAPTER 15 of the cardinality operator, they will bear that interpretation. If we add § 62,
however, the enquiry becomes one into the way numbers are given to us. The
i immediate answer provided is the context principle; it is on the strength of
Did Frege Refute Reductionism? that principle that the enquiry is converted into one concerning the senses of
sentences containing numerical terms - the first instance of the linguistic turn,
as already remarked. The interpretation no longer fits very well. According to
it, the context principle would not be an answer to the initial question at all,
The context principle has been much discussed. What does it mean, and what but merely a false clue, misleading us towards a mistaken answer; and there
I
role does it play in Frege’s argument? Even more copious discussion has been i would no longer be any ground for converting the enquiry into a search for a
i
devoted to the question whether Frege continued to maintain it in his middle linguistic definition. The superficiality of the proposed interpretation is con­
or mature period from 1891 to 1906. The questions are intertwined. Either firmed if we look more widely in Grundlagen. To have selected the context
the principle did not play a crucial role in the argument of Grundlagen-, or principle in the Introduction as one of the three methodological maxims
Frege still adhered to it in Grundgesetze-, or the structure of his thinking about guiding the entire investigation would have been quite unwarranted if it had
the philosophy of arithmetic underwent a radical change from one book to the played only the subsidiary role which the interpretation allots to it. This is
other. The emphasis given to the context principle in the Introduction to confirmed in Frege’s summan' of the whole course of the book’s argument in
Grundlagen makes the first of these three options highly implausible. Before §§ 106-9; in § 106, he cites it as a fundamental principle, providing the key
asking which of the other two we should choose, wre must therefore examine both to the problem how numbers are given to us and to that of finding the
the significance of the principle as it figures in Grundlagen. right definition of the cardinality operator. If Frege saw it thus, he must have
There can be no doubt, from Frege’s formulation of it in § 60: construed it as a guide towards the correct definition of that operator, rather
than as a principle, sound in itself, tempting us to give an incorrect one.
It is enough if the sentence as a whole has a sense; it is through this that its parts Indeed, in § 106 he ascribes to the context principle a deeper significance
/ obtain their content.
yet: it is what enables us to steer between the rock of empiricism and the
whirlpool of psychologism. As we saw, in the Preface to Grundgesetze this same
that he took it, at that time, as licensing contextual definitions. If any such
role is assigned to the different principle, also stressed in Grundlagen, that
doubt existed, it would be dispelled by the accompanying footnote:
objectivity does not entail actuality (Wirklieiikeif). The two principles must
The problem [concerning infinitesimals] is not... to produce a segment bounded therefore be closely connected: at least in Grundlagen, Frege must have seen
by two distinct points whose length is dr, but to define the sense of an equation I a grasp of the context principle as an essential condition for recognising that,
like df(x) = g(x) dr. to possess the status of being objective, an object does not have to be actual.
If we do not recognise this, we shall commit either the empiricist mistake of
The attempted contextual definition of the cardinality operator suggested in I
taking numbers to be actual, and so either physical aggregates or physical
I § 63 is not rejected because it is a contextual definition; that feature of it is
justified in § 65. It is rejected, rather, because it fails to solve the Julius
properties, or the psychologistic one of regarding them as subjective, the
products of human mental processes. We cannot have a correct view of w hat
Caesar problem; and though this defect will be shared by all similar proposed numbers are unless we understand that they are objective, but that they are
contextual definitions of those operators which need in fact to be defined by actual: but, if we fail to grasp the context principle, we shall be unable to
means of logical abstraction (by equivalence classes), there is no suggestion
that all contextual definitions will suffer from analogous defects. A hasty see how these two propositions can be true together.
reading of § § 62—9 w'ould therefore prompt us to interpret the context principle I Now objects which are objective but not actual are precisely what are now
called abstract objects. The salient characteristic of abstract objects is taken
as simply amounting to a justification of contextual definition. On this interpret- to be that they have no causal powers; and this is essentially Frege s criterion
180 I

I
 182 Frege: Philosophy ofMathematics IS Did Frige Refute Reduaionism? .g
for an object’s not being actual (tvirklich). In virtue of this characteristic, ours;nominalist
in The was
challenge to
what corresponded standard
not the
it was in Frege’s d
classification * has become
abstract objects stand in high disfavour in certain philosophical circles. Lacking of the human mind’ by those who, recognising that th/ nUmberS as ‘Crcations
causal powers, it is said, they cannot sene to explain anything. Moreover, we
of the physical universe, thought that there was nothing^ COnstituents
can have no evidence for their existence. For, since they lack causal powers,
Both lines of thought rest, equally, on a refusal to L t0 be>
everything would appear exactly the same to us if they did not exist. It follows
that the hypothesis of their existence is completely groundless, since we can objective but non-actual objects. We shall fail m j °gnise numbers as
no! appreciate that, for him,'such a reZ i“ “ if W d°
explain nothing by it and can have no evidence in favour of it: it must therefore
as taking the existence of the Equator to be a a™. a?d 8X055 an error
be dispensed with. does not succeed in explaining anything. Indeed^ we shall liTto^
This line of argument - which we may call ‘the nominalist challenge’ - can
have no force when the existence of the abstract objects is taken to be an large component of truth in the context principle if we fail t i^?
that the nominalist challenge is as crude a mistake as tha acknowledSe
analytic truth, as Frege took the existence of numbers to be. An analytic truth Equator; which is not at all to say that the matter is unproX^tic^
is not required to explain anything in order to be believed: it simply defies
disbelief. If it is analytically true that the natural numbers exist, we cannot
intelligibly ask how things would appear to us if they did not exist: their non­
existence is literally unthinkable; the question how things would be if some Meaning, sense and reference
self-contradictory proposition held good requires no answer.
This retort does not go to the heart of the matter, however. It is not in
I There can be no doubt, from § 62, that Frege was appealing to the context
i principle to justify our belief in the existence of the numbers. To ask, ‘How
every case an a priori truth that some non-actual object exists; and yet, even
are numbers given to us?’, is to ask by what means we apprehend them. It is
when it is not, the nominalist challenge is a paradigmatic example of what
therefore an epistemological question; but since it includes the question what
Wittgenstein meant by comparing philosophical perplexity to the bewilderment
of a primitive confronted with a sophisticated machine. In Grundlagen, Frege’s entitles us to suppose that there are any such things, it is also an ontological
one. Now the context principle is, in the first instance, a principle concerning
examples of objective but non-actual objects are the Equator and the centre
of mass of the solar system. The existence of the Equator is certainly an a meaning. It is by fixing the sense of a differential equation that the expression
posteriori truth. It depends on the fact that the Earth has poles, which in turn ‘dr* obtains its content or meaning; it is by fixing the senses of sentences
depends on the unquestionably contingent fact that it spins about an axis. Yet, containing numerical terms that such terms obtain theirs. The notion of
if someone argued that to assume the existence of the Equator explains nothing, f
meaning or of content employed by Frege in the early period diat includes
that, moreover, since it has no causal powers, everything wrould be exactly the the writing of Grundlagen was an undifferentiated one. In particular, Frege
!
same if it did not exist, and that therefore we have no reason to accept the made no distinction at that time between the meaningfulness of a singu ar
hypothesis of its existence, we should gape at the crudity of his misunder­ term and its denoting something: it either had a meaning or content, consisting
standing. in its standing for some object, or it was meaningless. The sentence eo
What should we say to correct the objector’s misunderstanding? He is trying Sachse is a man” is the expression of a thought only if “Leo Sachse designates
to conceive of the Equator as an actual object that has been stripped of its something’, as he wrote in his very early comments on Lotze s Logi . e
causal pow’ers; naturally, then, he cannot see what grounds we can have for question whether, by fixing the senses of sentences containing numerica terms>
believing in such an object. We have to teach him that it is an altogether we thereby guarantee those terms a sense, but not necessarily a reference, cou
f different kind of object. We can do that only by patiently explaining to him not arise for Frege at this stage of his thinking. That is why, althoug e neier
the use, or the truth-conditions, of sentences containing the term ‘the Equator’; reiterated the context principle in so many words, we cannot discuss it, as i
such an object as the Equator is given to us only by means of our grasp of is used in Grundlagen, without asking w’hat became of it - or w at s ou a
what can meaningfully be said about it and when it is true to say it. When we become of it — once the distinction between sense and re erence a
have given these explanations, he will grasp that there is nothing problematic introduced. Should we regard it as a principle concerning sense, or concerning
about the existence of the Equator; that its existence is not a hypothesis, but reference, or both? Did Frege continue to maintain it as a principle con g
stands or falls with the proposition that the Earth rotates about an axis. Or, if sense, and is it plausible as so interpreted? Did he continue to mam a
he does not, we may abandon him to self-congratulation on his resistance to
platomstic superstition. 189> and Posthumous Writings,
I 1 ‘Siebzehn Kemsatze zur Logik’, no. 10,
P-174, where it is incorrectly printed as part o
9.
i
 15 Did Frege Refute Reductionism? 185
184 Frege: Philosophy ofMathematics
This is surely correct as exegesis of Frege. For him, a subsentential expression
a principle concerning reference? If so, is it plausible when interpreted in this
jscssing a sufficient unity to be recognised as having a reference at all, must
way? pOS!
‘saturated’ or ‘unsaturated’. An unsaturated expression is one which
As a principle concerning sense, we may take the context principle as stating either be
be understood, must be conceived as containing one or more argument-
that the sense of an expression relates exclusively to its role in sentences, and tO U1‘~”—
places; it cannot therefore be taken as referring to an object, since an object
consists in its contribution to the thought expressed by any sentence in which
does not have arguments: if it refers to anything, it refers to a concept, relation
it occurs. So understood, it is indisputable that Frege continued to maintain
or function. A saturated expression is a ‘proper name’, in Frege’s terminology,
it in Grundgesetze. In Volume I, § 32, he wrote, ‘The simple names, and those
or what we less misleadingly speak of as a singular term. Only such an
that are themselves complex, of which the name of a truth-value consists
expression can be a candidate for having an object as its referent; if it has a
contribute to the expression of the thought, and this contribution, on the part
of any one of them, is its sense'-, by a ‘name of a truth-value’ Frege here means referent at all, that referent must be an object. Since the term that refers to
what is ordinarily called a sentence (considered apart from assertoric force), it has no argument-places, the referent cannot have arguments. What has
while the simple or complex names composing it are the constituent subsenten­ arguments is a function; an object is anything that is not a function. There
tial expressions. It follows that general stipulations adequate to determine die
i can be no question of acknowledging an expression as having a referent, and
senses of a range of sentences in which a given expression occurs must suffice then examining the referent to discern whether it is an object, a concept or a
to fix the sense of that expression, as it occurs in those sentences; since the I function of some other kind. If we did not know that, we did not understand
I
thought expressed by a sentence is partially constituted by the occurrence in the expression; and then we did not know what it was for it to have a referent.
it of a constituent with that sense, such stipulations must enable us to isolate It is true that we might query' Wright’s use of the term ‘syntactic’. If he
the contribution which that expression makes to the sense of the whole. In a were referring to expressions of Frege’s logical symbolism, the term would be
I completely accurate: in it, the distinction between ‘proper names’ and, for
language adapted to the carrying out of deductive inference, as Frege was
convinced that natural language is not, an expression ought, he believed, to example, names of first-level functions of one argument is a syntactic one in
bear the same sense in all contexts: it ought, that is, to make a uniform the strictest sense. So understood, however, the syntactic priority thesis would
contribution to the senses of all sentences containing it. The context principle, be utterly banal: the context of Wright’s discussion demands that we interpret
as one relating to sense, amounts to the conceptual priority of thoughts over him as referring to expressions of natural language. The question therefore
their constituents: the constituents can be grasped only as potential constituents arises to which principles of syntactic classification he is appealing. From the
i
of complete thoughts. That principle governed Frege’s thinking from start to standpoint of traditional syntax, ‘every actor’ is a singular noun-phrase, of the
finish: however expressed, it w;as one to which he was constant. same grammatical category as ‘the Pole Star’, but it is certainly not a candidate
What, then, does the context principle say if we interpret it as a principle for referring to an object. Wright is invoking a Fregean syntax for natural
concerning reference? Crispin Wright’s Frege's Conception ofNumbers as Objects2 language, which may not exist in its entirety, but is in any case a syntax devised
is largely devoted to this question. Both the meaning of the principle and with an eye to semantics. We may perhaps say that an expression play's the
Frege s attitude to it during his middle period are far more problematic w'hen kind of ‘syntactic role’ Wright has in mind if, to use Quine’s term, when
it is so understood than wrhen it is regarded as relating solely to sense. For regimenting’ sentences of natural language in the notation of predicate logic,
the present, we may address ourselves to the first of these two questions only, we should find it advisable to treat that expression as a singular term; or, at
assuming, merely as a w'orking hypothesis, that Frege continued to maintain least, to do so when the regimentation is designed for relatively superficial
the principle, in content if not in formulation, in Grundgesetze-, vve can reserve purposes. Certainly this would require that the expression should conform to
for later an enquiry' into the degree to w-hich the hypothesis holds good. standard rules of inference involving quantified sentences; and certainly we
V right interprets the principle as what he calls ‘the thesis of the priority of could not devise tests, formulated wholly by reference to natural language, for
syntactic over ontological categories’.3 He explains this thesis as laying down its fulfilling that ‘syntactic role’ unless we were permitted to appeal to the
intuitive validity of simple inferences carried out in natural language. Whether
We can devise comprehensive tests involving such an appeal is a debatable
the question whether a particular expression is i " " ______ matter, and one debated by Wright. His term ‘syntactic role’ must therefore
a candidate
is entirely a matter of the sort of syntactic role which it playstoinrefer
wholeto an object
: sentences. e Understood somewhat loosely; we may perhaps leave problems of syntactic
2 Aberdeen, 1983. classification to be dealt with as they arise, contenting ourselves with the
3 Op. cit., p. 51. reflection that vve can in practice judge reasonably well whether or not an
 186 Frege: Philosophy ofMathematics 15 Did Frege Refute Reduction? 187
expression would count for Frege as being a ‘proper name’, even if we cannot of variousor,sentences
equator, alternatively, becausethe
containing have provided
we term w'hich arecondirin
in t ? rf°r the truth
precisely formulate the principles underlying our judgements. condition for the Earth to have an equator is that it shn i? satisfied- The
So far, so good; but, given a candidate for being an expression that refers axis, which it does. Such a sentence as ‘We have crossed” b Up°n an
to an object, we naturally want to know what decides whether it is to be elected be judged true if the speaker has made a journey at the “ t0
or not. The question became an increasingly pressing one for Frege. In ‘Uber closer to one Pole and at the end of which he was do t he Was
Begriff und Gegenstand’ syntactical criteria induced him to deem ‘the concept this sometimes happens. It is accordingly not within hT t0 Other’ and
horse' to refer to an object, whereas, as he came to see, the right solution philosopher to deny that the Earth has » ?1
would have been to declare such phrases misbegotten; in his last years, he is such a thing as the Equator; and we need nnlv ’ a e f re’ ^re
expressed the view that phrases of the form ‘the extension of the concept a' category to which the expression ‘the Equator’ belo^ 10gicaJ
are pseudo-proper names referring to nothing.4 Wright has an answer for Equator is not a concept or a relation or t, Junction but Sn n f u”
this question, too, which he takes to be a further component of the context as an object. This is the reasoning which Wright is ad™ be ,classified
principle: rather plausibly, as reproducing Frege’s thinking; and XndnLn
account as cogent. B ancmg on his own
If it [a given expression] plays that sort of role [sc. the ‘syntactic role’ previously
mentioned], then the truth of appropriate sentences in which it so features will
be sufficient to confer on it an objectual reference.
■

A serious omission
Wright is of course here assuming that, by some unspecified but admissible i
I Now Wright is not particularly concerned with the Equator, which, indeed,
means, we have determined the truth-conditions of sentences containing the
expression. What is an ‘appropriate sentence’? If T is the expression in ques­ he does not mention. As the title of his book indicates, its aim is, by appeal
tion, then it might be a sentence like lT exists’ or ‘There is such a thing as to the context principle thus understood, to vindicate Frege’s method of
T. Alternatively, if we follow Frege’s principle that a sentence containing an introducing the cardinal numbers by fixing the senses of sentences containing
empty' singular term is devoid of truth-value, it might be held sufficient that numerical terms. More exacdy, he wishes, on the strength of Frege’s context
any sentence containing T is to be evaluated as either true or false. principle, to vindicate a means of fixing the senses of such sentences — and
The idea behind Wright’s suggestion is clear. ‘The Equator’ is a candidate so guaranteeing a reference for numerical terms - that Frege rejects, namely
for being a proper name referring to an object because it behaves like a the ‘contextual’ method discussed, largely by means of the analogy with terms
I
for directions, in §§ 63-7 of Grundlagen-, for Wright believes the Julius Caesar
I singular term; and it is to be admitted as in fact having an object for its referent
because it satisfies the condition we have laid down for the Earth’s having an problem to have been only a spurious obstacle to that procedure. The conse­
quence is that, for Wright, Grundlagen left no unfinished business. Frege’s
4 See ‘Erkenntnisquellen der Mathematik und der mathematischen N’aturwissenschaften’, Nach- actual procedure did leave unfinished business: the cardinality' operator was
gelassene Schrifien, pp. 288-9, Posthumous Writings, pp. 269-70. The passage is worth quoting at defined in terms of extensions of concepts, themselves archetypal logical, and
length. ‘A property of language that endangers the reliability of our thinking is its tendency to
form proper names to which no object corresponds ... A particularly noteworthy example of this ence non-actual, objects. The question then remained to be resolved how
is the formation of a proper name in accordance with the pattern “the extension of the concept vvtenSi°nS concePts are given to us, how terms for them are to be introduced;
a", e.g. “the extension of the concept fixed staP’. In virtue of the definite article, this expression right s view of the matter enables him to evade this question entirely.
seems to designate an object; but there is no object that could be so designated linguistically. i et, even if his short wray with the Julius Caesar problem were sound, he
From this have arisen the paradoxes of set theory, which have brought set theory to naught. I
myself, in attempting a logical foundation for numbers, succumbed to this delusion, by trying to ad no right to ignore the question about extensions of concepts altogether,
construe the numbers as sets ... It is indeed difficult, perhaps impossible, to test every expression t was in Volume I of Grundgesetze that Frege dealt with the business left
which language presents to us for its logical harmlessness. A great part of the philosopher’s work
thus consists - or at least ought to consist - in a struggle with language ... The same expression un nished in Grundlagen. Extensions of concepts are introduced in the later
* “the extension of the concept fixed staP’ ’ °°k as special cases of value-ranges. The question therefore becomes: how
OnTsuch^s11 >et anOther Wa' the dan8erous tendency of language to form pseudo-proper names. are va^Ue~ranges given to us, and how are terms for value-ranges to be guaran-
ee a reference? We are here still proceeding on the assumption, as yet
. ... * *,dle concePt fixed staP’ ’ unscrutinised, that, in Grundgesetze, Frege still maintained the context principle,
y itself, from the definite article the appearance arises that an object is designated, or, what is
the same thing, that ‘the concept fixed staP' is a proper name; whereas “concept fixed staP’ as a principle concerning reference. If so, then it must have been by appeal
designates a concept and hence stands in the sharpest contrast with a proper name.’
that he justified his introduction of value-ranges. This is prima facie
■

I [
 188 Frege: Philosophy ofMathematics 75 Did Fre& Ref™ Reductionism? 189
plausible, since there is an exact formal analogy between the cardinality- oper­
for
to the Julius (the
numbers original
Caesar equivalence),
problem and governing
(a stipulation idJtit k S°me S°lution
supplementing
ator and the abstraction operator. Both are term-forming operators of second a numerical term on one side and a term of another sort 1^'Stat^ments wth
level, to be attached to an expression for a concept or for a function; both are appearances, this would exactly resemble Frege’s To aU
governed by a criterion of identity stated in terms of a second-level relation abstraction operator in Grundgesetze. In this case we thefntr°kducin?
between the relevant concepts or functions. The abstraction operator differs options: to reject the context principle altogether- to — ?ave three
from the cardinality- operator, on Frege’s account, in being primitive, and that it does not vindicate the procedure Wright has in m a™ but decIare
hence incapable of being explicitly defined. Since, by the time of writing a restriction upon it that distinguishes the cardiX ipeZr f“ a
Grundgesetze, Frege had repudiated the whole conception of contextual defi­ tion operator. Wright does none of these thinev hp ■ Om ^de­
nition, he does not so much as entertain the possibility- of defining the abstrac­ principle in full generality, understood as ht intemreteilTd TT
tion operator contextually: it is to be a primitive symbol governed by an axiom
and by stipulation of its reference in the metalanguage. But this leaves it in appeal to itintothe
introduced justify ascribing
foregoing withouttostopping
a reference
manner, numerical terns u1 dered as
to exnkin
no different case from the cardinality operator, considered as introduced in ently similar manner of introducing value-range terms w h*'anfppar'
the way Wright favours, namely by the ‘contextual’ procedure canvassed in contradiction. He owes us such an explanationXc^mtr th ' *
Grundlagen, §§ 63-5, and rejected in §§ 66-7. It makes no difference, for I introducing the cardinality operator he envisages would oWate w “a „°f
present purposes, whether this procedure is described as a ‘definition’ or not.
Hence, if the context principle, as expounded by Wright, is enough to validate notion of a class suppl,es no excuse for his failure to provide that explanadon
the ‘contextual’ method of introducing the cardinality operator, it must be
enough to validate a similar means of introducing the abstraction operator.
This is why the mere fact that, on his new, it is unnecessary to define the
cardinality operator in terms of classes or of value-ranges does not entitle Contextual definitions
Wright to ignore the problem of the abstraction operator. For Frege’s method For Wright, two conditions are together necessary- for recognising an
of introducing the abstraction operator - that is, of introducing value-ranges expression T as referring to an object: that T fulfils the ‘syntactic role’ of a
- was, notoriously, not in order. It rendered his system inconsistent; and that singular term; and that yve have fixed the truth-conditions of sentences contain­
inconsistency- forced him eventually to acknowledge that his entire enterprise ing Tin such a way that some of them come out as true. It does not, apparently-,
had failed.5 If the context principle, as stated by Wright, were sound, there matter by what means we fix those truth-conditions. In particular, it is plain
could have been no inconsistency. More exactly, w-e should distinguish between that Wright would admit the case in which we do so by means of a genuine
the general principle and an application of it to justify ascribing a reference contextual definition, or chain of genuine contextual definitions, provided that I
to value-range terms as Frege introduces them in Grundgesetze. The context these leave intact the syntactic role of the putative singular term T, as, for
principle, as formulated by Wright, requires the truth-conditions of sentences example, Russell’s theory of descriptions fails to do for definite descriptions,
containing the terms in question to have been fixed: if they have not been t does not matter yvhether, in the sentence into which the definition transforms
fixed, then the context principle does not genuinely apply. We have then only a sentence containing T, there is any longer a singular term corresponding to
two options: either the context principle is not unreservedly sound; or if, in • what matters is that the sentences to be transformed are so explained that,
introducing the abstraction operator in the way he did, Frege took himself to so far as the logical behaviour of those sentences is concerned, T is not
r be guided by the context principle, then his application of the principle was unmasked as only a spurious singular term. This means, primarily, that the
erroneous, since he failed to satisfy the conditions it lays down. It is not open ay\s relating to quantifiers must remain valid. On Russell’s theory-, we cannot,
to us to defend both the context principle in general and this application of (2r examplc, infer ‘The King of France brushes his teeth at night’ from
it. our task must be to diagnose Frege’s error and, if necessary, to delimit the 'cryone brushes his teeth at night’; that is why, on that theory-, ‘the King of
scope of the context principle accordingly. But this throws the gravest doubts rance does not have the syntactic role of a singular term. We cannot accuse
upon Wright s claims. We may take him as concerned to vindicate, by appeal right of being unfaithful to the intentions of the author of Grundlagen;
to e ^text principle, a method of introducing the cardinality operator which F^e.s remarks about infinitesimals in § 60 clearly show that he is not.
Frege did not in fact adopt: namely, by laying down the criterion of identity d fi ma^es clear the admissibility-, from his standpoint, of contextual
e nitions by- choosing, as his central example, precisely the introduction of
me - letter to Honigswald of 4 May 1925. der Arithmetik the whole structure collapsed
H about
ms for directions, as discussed by Frege in §§ 63—5. He considers the
I

1
i
 190 Frege: Philosophy ofMathematics 15 Did Frege Refute Reductionism? 191
truth-conditions of sentences about directions as determined by three sets of into sentences involving numerically definite quantifiers - but still insist
them
stipulations: (fiat those terms arc genuine singular terms standing for non-actual objects,
the numbers.
(1) the identity condition that the direction of a is the same as the direction of b Wright distinguishes three ways of regarding such a system of contextual
if and only if a is parallel to b; definitions: an austere way, a robust way and an intermediate way. The austere
(2) a series of stipulations to the effect that a predicate A, is to hold good of the
interpretation is that of an intolerant reductionist. Such a reductionist claims
direction of a just in case some known predicate F, holds good of a, where that to explain sentences about directions by translating them into sentences
parallelism is a congruence relation with respect to F,; about lines, or sentences about numbers by translating them into ones involving
expressions for and quantification over second-level concepts, commits us to
(3) for each such A(, the stipulation that ‘For some d, t\(.d)' is to be true if and only if denying that there are any such objects as directions or numbers. Such an
‘For some a, F («)’ is true.
explanation, according to him, precludes discernment of any genuine semantic
There is a little uncertainty' over whether these stipulations are intended to structure in the sentences so explained; they merely possess misleading surface
govern a two-sorted or a one-sorted language. In Wright s first formulation of forms.
them, he speaks of ‘quantification over directions’, and uses ‘d' as a bound The robust interpretation, which Wright favours, is that the contextual
variable ranging over directions, and ‘a’ as one ranging over lines.6 When the definitions succeed in conferring upon sentences containing the terms contex­
stipulations are later repeated, ‘a’ is still used as both a free and a bound tually defined senses which warrant our viewing them as having just that
variable over lines, but ‘.r’ as a bound variable over directions and also a free semantic structure which their surface forms suggest. The defined terms are
variable over lines.7 The point is not, however, of crucial importance; Wright genuine singular terms, with a genuine reference, albeit to abstract objects.
makes plain that he is wishing to consider a set of stipulations together The intermediate interpretation is attributed by Wright to me, on the basis of
constituting a complete contextual definition, enabling sentences involving what I wrote in my Frege: Philosophy of Language. Ultimately, Wright fails to
i
reference to or quantification over directions to be transformed into ones find this intermediate view coherent: he doubts if there is any tenable position
relating only to lines. between the austere and robust interpretations. As concerns contextual defi­
We could also consider an explanation of numerical terms, and of arithmeti­ nitions, properly so called, I shall here maintain an intermediate new, perhaps
cal sentences containing them, by means of the interpretation of numbers as one more austere than that which W’right had in mind. I shall, however, spend
concepts of second level, in line with Frege’s abortive definitions in § 55 of no time in discussing either how faithfully Wright represents the views I
Grundlagen. Wright’s objection to this is the standard one, that it would make expressed in Frege: Philosophy of Language, or how far those I advance here
the infinity' of the sequence of natural numbers depend on there being infinitely diverge from them.
many objects other than numbers (or classes); but this does not affect the The intermediate interpretation, as I here understand it, is that of a tolerant
present point. Ifwe so interpreted arithmetical statements, we should have to reductionist. He holds that the contextual definition serves to explain what it
agree that how many cardinal numbers there were depended upon how many means to say, ‘There is a direction orthogonal to those of lines a and b, or
non-logical objects there were: the question presently at issue is whether our ‘There is a prime that divides both 943 and 1357’, rather than to show that
so interpreting them would be compatible with our maintaining that those we ought not to sayr things of that kind. He therefore agrees that it would be
cardinal numbers w'hich did exist were objects. The interpretation is naturally wrong to say that neither directions nor numbers exist, even if we adopt the
described as embodying a refusal to take numerical terms at face value, as contextual definitions: yrou cannot consistently combine the assertion that there
being genuine singular terms, and that is how it was described when it was is a number satisfying a certain condition with the declaration that there are
discussed in an earlier chapter. If the context principle, as Wright understands no numbers whatever. He recognises further that ‘ “31” refers to an object
it, is correct, however, the description is tendentious: we could explain sen­ can be construed untendentiously as simply the equivalent, in the formal mode,
tences containing numerical terms in exactly the same way — by transforming of‘There is such a number as 31’, and hence as uncontroversially true. What
e denies, however, is that the notion of reference, as so used, is to be
»
understood realistically.
Mat does this denial amount to? The difference between Frege’s early
6C.WrightFrege’s Conception ofNumbers Penod, during which Grundlagen was written, and the middle period which
Ibid., p. 67. as Objects, pp. 29-30.
e«ends over the composition of both volumes of Grundgesetze is not merelv
extends
 15 Did Frege Refute Reductionism? 193
192 Frege: Philosophy ofMathematics
fflav be made to rest upon the theory of Bedeutung as a base, the sense being
that, in the latter, he distinguished between sense and reference within the
the way >n which the Bedeutung is given to us. In Grundlagen, no such theory
former inchoate conception of content or meaning. Certainly his failure,
is envisaged. Frege was certainly interested there in the semantic analysis of
during the early period, to draw any overt distinction between the significance
particular forms of sentence, as in the discussion of ascriptions of number
of an expression and what it signifies left an incoherence in his thinking
(Zahlangabcn) in § 46 or of ‘All whales are mammals’ in § 47; but he had no
which he rectified only when he drew the sense/reference distinction: that
is why it is so grievous an error to attempt to extract from the writing of his apparatus that would supply the resources for any general theory. In Grund-
earlv period a system of logical doctrines as articulated and consistent as that lagen, the semantic discussion is treated as being on the same level as the
expounded in the middle period. There is, however, a deeper difference. In sentences on which it bears. That is why Frege so unconcernedly employs his
a clear sense, Frege did not even aim, in his early period, at constructing a jargon, speaking of the number 4’s belonging to the concept moon ofJupiter
philosophical theory of logic. The context principle, as enunciated in Grund- instead of speaking of there being four moons of Jupiter: expressions which
lagen, can be interpreted as saying that questions about the meaning we should think of as having their home in the metalanguage are not kept
(Bedeutung) of a term or class of terms are, when legitimate, internal to the
language. We know the meaning of a term, say ‘the Equator’, when we know
segregated from expressions of the object-language, but substituted for them
or jumbled together with them, because Frege has no distinction between I
the conditions for the truth of any sentence containing it; that is all we need metalanguage and object-language in mind.
to know, and all we can know. Hence, to determine the meaning of a term, In Grundgesetze, all is different. There there is the sharpest distinction i
what we have to do, and all that we have to do, is to fix the senses of between the object-language, which in this case is Frege’s formal language, and
sentences in which it occurs. Reference therefore does not consist in a mental the metalinguistic stipulations, stated in German, of the intended references of
association between the term and the object, considered as apprehended by expressions of the formal language. Now the conception of sense and reference
the mind independently of language; nor can it consist, we may add, in the with which Frege operated throughout his middle period was as follows. The
existence of a causal chain leading from the object to an utterance of the sense to be attributed to an expression depends on what is involved in grasping
term. It follows that any legitimate question about the meaning of a term, thoughts expressed by sentences containing it. To grasp the thought expressed
that is, about what we should call its reference, must be reducible to a by a sentence is to know what determines it as true or as false. The references
question about the truth or otherwise of some sentence of the language. To of the component expressions constitute their respective contributions to the
ask whether a term ‘a’ denotes something with spatial location is to ask determination of its truth-value; and the sense of any one of them constitutes
whether the sentence ‘a is somewhere’ is true; to ask whether ‘a’ and ‘b’ the particular way in which its reference is given to one who grasps the thought.
have the same reference is to ask whether ‘a = is true; to ask whether ‘«’ Our conception of the way its truth-value is determined is therefore itself
has a reference is to ask whether ‘There is such a thing as a' is true. articulated, in a manner corresponding to the articulation of the thought and
Questions about the meaning or reference of a term that cannot be thus of the sentence expressing it: we have a particular way of conceiting of a
formulated in the material mode are illegitimate and derive from attempting j certain object as being picked out by each singular term, a particular way of
r I
to ask after its meaning in isolation. In particular, there can be no further conceiving of a relation as obtaining or failing to obtain between any two given
specifically philosophical enquiry needed, beyrond the relevant enquiry within objects, and so on, which jointly yield for us a particular manner in which one
the subject-matter to which the term relates, in order to establish whether
or not it stands for anything. or other truth-value is arrived at.
When the sense of a term is given to us by means of a contextual definition,
All this accords very' well with Wright’s account, which we may therefore however, this model ceases to apply. Our grasp of the thought expressed by a
recognise as in large degree a faithful exegesis of Frege’s use of the context sentence containing the term is mediated by our knowledge (possibly only
principle in Grundlagen. This the proponent of the intermediate interpretation I implicit) of how to arrive at an equivalent sentence not containing that term.
acknowledges by allowing that, so understood, the claim of a term to have a I The notion of the reference of the term, as determined by its sense, plays no
reference is not impugned by its hating been introduced by a contextual role in our conception of what determines the thought as true or false, nor,
definition. Viewed against the background of Frege’s middle period, however, erefore, in our grasp of the thought; the attribution of reference to the term
the whole framework of the discussion is altered. Now the notion of Bedeutung
be ^eFensible, when tolerantly viewed, but is semantically idle.
incorporates a whole theory’ of how the truth-value of a sentence is determined tight strives valiantly to resist this conclusion, and, more stou y yet, to
in accordance with its composition, while the notion of sense serves to sketch tesist the austere interpretation. Again, we cannot say, on behalf of the Frege
how’ a theory- of what we know' when we understand an expression or a sentence 01 Grundlagen, that he is wrong. At the end of his review, published in 18W,

i I
 194 Frege: Philosophy ofMathematics 15 Did Frege Refute Reductionism? I95
of Hermann Cohen’s book on infinitesimals,8 Frege wrote: and in which it may be grasped. When these different ways of grasping a single
Lse are allowed to vary to the extent that one involves apprehending that an
As regards the foundation of the differential calculus, we must, in my opinion, object of a certain sort is being referred to, W'hile the other is compatible with
go back to the concept of a limit as understood in analysis, which, owing to his j ’0rance of the very concept of objects of that sort, the link that exists in
misunderstanding of it, the author disparages as a ‘negative’ one. I recently Frege’s theory' between sense and reference has been snapped; and now it is
indicated, in my Grundlagen der Arithmetik (p. 72, fn. 1), how by means of such
a foundation it is possible to secure for the differential a certain self-subsistence
quite unclear what kind of thing the sense of an expression is at all. The claim
[Selbstdndigkeit]. of‘ontological priority’ for sentences containing the contextually defined terms
/
divorces the notion of reference from that of sense, just as the proponent of
If we disregard non-standard analysis, virtually everyone would construe an the intermediate interpretation maintained. If someone can understand the
explanation of differentiation by appeal to limits in a reductionist spirit, as i statement about lines without being aware that it involves any reference to a
i
showing that we need not understand differential equations as involving any direction, he has a conception of what would determine it as true into which
I reference to infinitesimals; Frege, on the contrary, appears to think that, in reference to the direction does not enter. That means that the attribution of
the light of the context principle, such an explanation vindicates the conception such reference is semantically idle, precisely in accordance with the intermedi­
of ‘dr’ as denoting an infinitesimal quantity. Yet we are exploring a region of ate interpretation.
Frege’s thinking which we know in advance cannot be rendered fully coherent, All this, however, is seeing the matter from the perspective of Grundgesetze,
since otherwise he would not have fallen into inconsistency; it is this feature when a semantic theory is in place, informed by the distinction between sense
! of the topic which Wright appears persistently to forget. and reference. The perspective of Grundlagen differs altogether. The doctrine
He opposes the austere interpretation by standing the usual argument for of Grundlagen is, in effect, that there is no metalinguistic standpoint: such
it on its head. He proposes that it is better to regard a sentence ostensibly only would-be metalinguistic statements as ‘The term “a” refers to an object’
I about lines, but equivalent by contextual definition to one about a direction, as reduce to, and can only be understood as, statements such as ‘There is such
having a misleading grammatical form, than so to regard the one about direc­ a thing as «’ which belong to the language itself, a language which we cannot
tions, as the reductionist does: the former sentence, he claims, ‘achieves a allude to as the ‘object-language’, because no other language is under consider­
reference to a direction without containing any particular part W’hich so refers’.9 ation. Thus, in Grundlagen, the context principle amounts to a repudiation of
This appears to fly in the face of Frege’s insistence that it is through our prior the possibility of a semantic theory explaining, as from outside the language,
understanding of propositions stating that lines are parallel that we attain the the mechanism by which its sentences are determined as true or as false.11
concept of a direction. If that is so, then we can understand such a proposition !|
The context principle, as understood in Grundlagen, therefore admits only a
before we have that concept; and how could wre understand it if it involved a thin notion of reference, that notion according to which ‘ “The direction of
reference to something of w hich we as yet had no conception? Wright’s answer a ’ refers to something’ is indisputably true, because it reduces to ‘The line a
is that, while the statement about lines has epistemological priority, that about has a direction’, and ‘ “The direction of a” refers to the direction of a' trivially
directions has ‘ontological priority’:10 but what can this mean? If, indeed, two true, because it reduces to ‘The direction of a is the direction of a'. The
sentences have the very same sense, and one involves reference to a direction, context principle of Grundlagen is thus strictly7 analogous to the redundancy7 ■

the other must do so as well: genuine reference to an object must be an theory of truth, that theory which admits only the thin notion of truth according
intrinsic feature of the sense, rather than characterising merely the manner of to which ‘ “Cleanliness is next to godliness” is true’ reduces to ‘Cleanliness
its expression. We saw earlier, however, that Frege was w'rong, on his own
principles, to hold that a definitional transformation leaves sense wholly " I mean, of course, ‘repudiation of the possibility of a semantic theory altogether, such a theory’
unaltered. This cannot be true if sense incorporates everything concerning the einR °ne t^at would explain ...’, and not ‘repudiation of the possibility that a semantic eon
linguistic expression that goes to determine its reference. The alternative is to 'Quid be able to explain ..if I had meant the latter, I should have written ‘the possibility of a
*e°?'’s fining...’. A decade or so ago, the risk that my sentence would be
deny that the sense simply is the w7ay in w'hich the referent is given, by ^understood would not have occurred to me; but now that few writers or publishersjtmce a
admitting different ways in which one and the same sense can be expressed Romp’0 t^'e d*st*nct*on between a gerund and a participle, so that phrases like due to ’
Sent COns!antly aPpear in print, it has become substantial. I preferred, however,
freoii^tiaS ^rst wrote ’*> while adding this footnote, to resorting to inelegant perip r i .
8 H. Cohen, Das Prinzip der Infinitesimal-Methode und seine Gesehichte, Berlin, 1883. their m J tymark that they see no point in observing grammatical rules, so ong as ,
9 C. Wright, op. cit., p. 32. strin? _ea,ninf’’ This is like saying that there is nothing wrong with using 8 raz05 ewress
10 Ibid., p. 31. ‘heir m ° °ng as t*le s,r'nB >s cut; bv violating the rules, they make it difficult or o
^"■meamng without ambiguity . ’
3
 196 Frege: Philosophy ofMathematics 15 Did Frege Refute Reductionism? 197
is next to godliness’. It was on the basis of the context principle, so understood, •opher or geometer busied himself with it. Whether, and, if so, how, it is
that Frege, who at that time allowed contextual definitions as legitimate,’ Possible to explain abstract terms otherwise than by contextual definition, and
rejected an austere view of them, as his remarks about differentiation demon­ what substance the notion of reference has, when applied to them, are que­
strate. From the perspective of Grundlagen, there is no more substantial notion stions deferred to the subsequent chapters.
of reference, and hence there is no room for an intermediate view, which Frege’s realism about mathematics was already in place in Grundlagen: the
cannot even be stated in die terms allowable by the Grundlagen doctrine: this famous ringing declaration in § 96 that ‘the mathematician ... can only dis­
is what prompts Wright to judge it to be incoherent. Are we dien to credit cover what is there, and give it a name’ leaves us in no doubt about that. Yet
i
I Wright with giving a faithful interpretation of the Frege of the Grundlagen 1
we may well feel dubious whether he was entitled to his realism. What the
I
when he attributes to him a robust way of construing all legitimate means of I rhetoric means may be glimpsed by contrasting empirical with mathematical
introducing new terms into the language, including contextual definition? concepts. To make an empirical concept, say comet, sharp, we need a criterion
The answer depends on what ‘robust’ means. If it means simply ‘not austere’, for whether an object given directly or indirectly by observation falls under it,
I
so that a robust view is merely the negation of the view that a reference may and a criterion for whether an object falling under it and given by a certain
not be ascribable to die new terms at all, then Wright’s interpretation is indeed observation is the same as an object falling under it and given by some other
faithful to Grundlagen. But, when we recall that his formulation of the context observation. A realistic conception of the external world assures us that, once
principle involved the ascription of an 'objectual reference’, we may doubt this we are satisfied that the concept is sharp in these respects, we need do no
mild reading of the word ‘robust’. From the standpoint of Grundgesetze, there more to guarantee determinate truth-values for quantified statements involving
is a more substantial notion of reference than the thin one allowed by Grund­ it, statements to the effect that there is a comet satisfying some condition, or
lagen, that notion, namely, employed in the semantic theory: to ask whether that all comets satisfy some other condition. In general, the determination of
an expression has a reference in this sense is to ask whether the semantic theory the truth-values of our sentences is effected jointly by our attaching particular
assigns one to it, or needs to do so, where reference is a theoretical notion of senses to them and by the way tilings are. We do not need to specify what
that theory. Wright’s notion of objectual reference appears to be just such a comets there are, once we have rendered our concept of a comet sharp: reality
substantial notion, at home in Grundgesetze but not in Grundlagen. This is not, does that for us, and reality7 therefore determines the truth or falsity of our
of course, to criticise Wright for adopting a semantic viewpoint, which he is
quantified statements. So, at least, realism assures us.
entirely justified in doing; but one cannot combine this with embracing the Hardly anyone is realist enough about mathematics to think in the same
whole doctrine of Grundlagen.
way about quantified mathematical statements. A fundamental mathematical
W right argues further that the austere view7 ‘is not even an option’ when
concept, say real number, which determines the domain ot quantification of a
the contextual definition does not serve to introduce new7 expressions into the
mathematical theory7, must indeed have a criterion of application and a criterion
language, but to explain existing ones.12 This depends, however, on whether
of identity7. Given a mathematical object, specified in some legitimate way, we
the contextual definition is put forward as giving the senses we already attach
must know7 what has to hold good of it for it to be a real number; and, given
to sentences containing the expressions defined, or merely to show how we two such specifications, we must know the condition for them to pick out the
could explain them without disturbing our existing use of them. A thorough­
same real number. Few suppose, however, that, once these two criteria have
going reductionist will indeed claim that the mere possibility of contextually
been fixed, statements involving quantification over real numbers have thereby
efining the direction-operator shows that there are no such things as direc-
all been rendered determinately true or false; to achieve that, it would be
ons; but we may cheerfully reject even the intermediate view of the matter,
generally agreed that further specifications on our part were required, in some
ere ormu ate , if the contextual definition fails to give the senses we fashion circumscribing the totality of real numbers and laying down what real
-° SeTenreS ab°Ut directions> for then we have no reason to numbers there are to be taken to be. Frege was, perhaps, an exception to this
it must he' m t0 be semantically idle. In that case, however, generalisation. For certainly7, as we shall see, his stipulations concerning the
tha7 accords °tSSt t0 what ^ose actual senses are, in some manner value-ranges comprised by the domain of the formal theory7 of Grundgesetze go
ScaZ JwT ? a reference *at is semantically idle: I
no further than supplying criteria for something’s being a value-range and for
that the v ’ dn nght ZS’ SimpIy brush question aside on the ground I foe identity7 of value ranges; nothing resembling a circumscription ot the
that the word direction’ was already in use in the language before any philo-
domain occurs. Of course, since value-ranges are logical objects, the truth­
Ibid., p. 68. values of quantified statements of the theory7 would not then be determined
by what value-ranges there happened to be, as those of quantified statements

I
 15 Did Frege Refute Reductionism? 199
198 Frege: Philosophy ofMathematics
•fore postponed. His courage in tackling the most difficult case for his
about comets are determined by what comets there happen to be, but by what to be admired; but he appears to regard it as the typical case, which
value-ranges there must be. view
re]y is not. Certainly the cases that concern us in considering Frege’s
This is no more than speculation about how Frege thought, for the sake of 'h’losophy of arithmetic - those of cardinal numbers and of value-ranges -
attaching some non-metaphorical content to his rhetoric. If it is correct, Frege ph'°not instances of contextual definition. Wright makes a good case for
was profoundly mistaken; so strongly realistic an interpretation of mathematical T king that, at the time of writing Grundlagen, Frege would have been willing
statements cannot be sustained. But, whatever be the right understanding of 1111 ascribe a reference to contextually defined terms on the strength of the
the realism of Grumilagen, on what does it rest? Only on ascribing to mathemat­ tOntext principle. He errs in supposing that this would have been the substan­
ical terms the thin notion of reference that is all that Grumilagen admits. The
tial notion of reference used in Grundgesetze, rather than the thin one that is
proponent of the intermediate new of terms introduced by contextual definition
all the context principle of Grundlagen allows. In the absence of a semantic
- the view for which I have here argued - maintains that that thin notion of
theory or of any desire for one, we cannot even say what it is to put a realistic
reference will not bear the weight of a realistic interpretation of those terms;
interpretation upon any given range of sentences. Once Frege had such a
and, since Grundlagen does not allow a more substantial notion of reference
theory, he disallowed contextual definition altogether. Even if this be waived,
to be ascribed to any terms, however introduced, there can be no basis for
we are bound, in the presence of such a theory, to acknowledge that the robust
realism about mathematics, or, indeed, about anything else. Within the frame­
view is untenable for terms for which contextual definition is the only way in
work of Grundlagen, it cannot be asked whether the notion is semantically idle
which to explain them, or the one most faithful to our ordinary’ understanding
when applied to contextually defined terms, because it is not semantically
of them: the intermediate view is the closest we can come to accommodating
operative, whatever it be applied to.
Realism is a metaphysical doctrine; but it stands or falls with the viability' Wright’s claims.
of a corresponding semantic theory’.13 There is no general semantic theory' in,
or underlying, Grumilagen-, the context principle repudiates semantics. That
principle, as understood in Grumilagen, ought therefore not to be invoked as
underpinning realism, but seen as dismissing the issue as spurious. There is
a semantic theory in Grumigesetze-, and it is in the light of that theory that we
must assess the conclusions drawn by Wright from the context principle.
Admittedly, we have so far advanced very’ little distance, and are not yet in a
position to pronounce on abstract terms in general from the standpoint of a
theory’ of reference like that of Grundgesetze. We can, however, surely conclude
that, where the notion of reference is semantically’ idle, it cannot be appealed
to in justification of realism. Full-fledged realism depends on — indeed, may
be identified with — an undiluted application to sentences of the relevant kind
of a straightforward two-valued classical semantics:’4 a Fregean semantics, in
fact. This excludes an explanation of certain terms by a rule for transforming
sentences containing them into equivalent sentences containing neither them
nor any corresponding terms, which is what a contextual definition will yield.
It is on contextually defined terms that Wright chose to take his stand; concern­
ing them,, at least, we must judge that his case fails.
Wright’s further arguments relate, not specifically to contextually defined
terms, but to terms for abstract objects generally; consideration of them is

must eo^roueh^mMn^nffUfkP’aCeS ^at .On'J TOUt.e ,0 a v’nc*ication or refutation of realism

is riven in the It enQyir)’ into the right form of semantic theorv; the argument f
I
* T"‘ Cambridge.
14 See M. Dummett, ‘Realism’, Synthcse, Vol. 52 (1982), pp. 55-112.
 16 The Context Principle 20J
,1 apparently proceeded to derive that condition frnn, .
he thenof the direction-operator. But this appearance was illuso^wJ ’
nition c
was claiming , arrive at a grasp of the concept of lines" being pX. The
directions are, . - r
criterion of identity, stated in terms of that relation, was not itself a definition
CHAPTER 16 Or part of a definition, of the direction-operator, but a condition for the
correctness of a definition of it: such a definition could be correct only if the
criterion were derivable from it. Something very similar holds good’in the
The Context Principle present case. The flaw in the foregoing argument that the context principle is
not needed to justify ascribing a reference to an explicitly defined term lies in
its neglect of what is required to recognise a proposed explicit definition as
correct - in our case, a definition of the cardinality operator. If we supposed
The role of the context principle in Grundlagen that an assignment of a reference to a term consisted in a mental association
of the term with the referent, apprehended independently of language, we
On the face of it, the explanation of an expression - in our case, the cardinality should never allow Frege’s definition of cardinal numbers as extensions of
operator - by means of an explicit definition renders otiose an appeal to the concepts. We do not directly apprehend extensions of concepts, and certainly
context principle to justify7 ascribing a reference to it. This claim does not not extensions of second-level concepts, or of concepts under which extensions
depend on interpreting the context principle as doing no more than declare of other concepts fall. Once we have grasped the context principle, however,
legitimate explanations by means of contextual definition. It does not matter we recognise that this is quite the wrong way to think about the matter. What
whether the alternative was a contextual definition, or a series of them, or is needed is a definition that will fix the truth-conditions of sentences in which
some other form of definition, or yet an explanation not amounting to a numerical terms occur. It may be recognised as correct provided (a) that it
definition. Whatever the alternative, it appears that, when we can define a confers determinate truth-conditions on every7 admissible such sentence, and
term explicitly, we do not need to appeal to the context principle, or any other, (b) that it confers the right truth-conditions on those of such sentences for
to warrant the ascription to it of a reference, proriding, of course, that we which there are ‘right’ truth-conditions. Particular truth-conditions may be
acknowledge the definiens as having a reference. It therefore appears perplex­ considered right for a sentence containing numerical terms if they are those
ing that, haring eventually arrived at an explicit definition of the cardinality required by the sense we ordinarily attach to that sentence, supposing that we
operator, Frege should, in § 106, emphasise the context principle as an essen­ do ordinarily attach a sense to it. It so happens that, provided that we give
tial step on the route to that definition. suitable definitions for other arithmetical expressions, condition (b) can be
A bad explanation would be that the cardinality operator is defined in terms fulfilled as long as the cardinality operator is so defined as to satisfy the original
of extensions of concepts, that extensions of concepts are in turn to be
explained, in Grundgesetze, as forming a special kind of value-range, and that equivalence
the ascription of reference to terms for value-ranges can be justified only by
The number of Fs = the number of Gs
appeal to the context principle. In Grundlagen, Frege is plainly not in the least
occupied with the question how to justify ascribing a reference to terms for if and only if.
extensions of concepts: he is simply taking the notion of the extension of a there are just as many Fs as
concept for granted as unproblematic. His citation of the context principle in t, epntences containing numerical
§ 106 does not read like a glance ahead to the completion of the unfinished Moreover, the senses we ordinarily attach rtOrPnces of those terms uniquely-
business: it is meant to remind the reader of an indispensable step in the terms do not suffice to determine the refere*c d determine them
preceding argument. Hence, while condition (a) demands that our e joes not violate
The resolution of the perplexity7 is not far to seek: it resembles the resolution uniquely, we are at liberty' to do so in any m origjnai equivalence -
of that concerning Frege’s appeal to the criterion of identity for directions. condition (b). Thus, as before, the denvabihtyfor the correctness
die criterion of identity for numbers - becomes text principle teaches
Frege appeared to claim that we could not derive, from a knowledge of what
°f a definition of the cardinality operator. , a orjginal equivalence can
the direction of a line is, the condition for two lines to have the same direction:
is to be satisfied with a definition from which the
200
i
 16 The Context Principle 203
2Q2 Frege: Philosophy ofMathematics
containing it, but only of certain ones: those of a particular simple form,
be derived, or, more exactly, with any definition fulfilling our two conditions. haracteristic for the expression in question. The contribution of the expression
If we do not acknowledge the context principle, we are certain to reject a
to the thoughts expressed by other, more complex, sentences is then grasped
definition of the cardinality operator such as Frege gives. Having understood
nd can be explained, by reference to the senses of those simpler characteristic
the principle, we shall realise that a definition of such a kind accomplishes
sentences. An obvious case is that of a predicate. Someone may be credited
everything that a definition of numerical terms can be required to do or can
with a grasP of die sense of 1116 Predicate if he knows the condition for it to
apply to any one given object, which is to say that he has a general understand­
ing of atomic sentences in which it figures. He need not understand all of
A compositional interpretation of the context principle them, of course, since there may be many proper names of objects that he
To grasp the sense of an expression is to apprehend the contribution that it does not know; but he must understand some singular terms picking out
makes to the thought expressed by any sentence in which it occurs. But w’hat objects of which the predicate may intelligibly be said to hold good, and thus
is it to know this? Must we understand every sentence in which the expression have the general conception of referring to such an object in the course of
occurs? Obviously not: for the understanding of such sentences wall depend applying the predicate to it. The speaker’s grasp of the sense of the predicate
on our grasping the senses of other expressions occurring in them. But suppose does not, however, require him to understand quantified sentences containing
we do understand all such sentences: does our understanding of them constitute it. His coming to understand them is wholly a matter of his coming to grasp
our understanding of the given expression? Again, obviously not. Our under­ the senses of the quantifiers. That will depend upon his already having the
standing of them - indeed, if the expression has a uniform sense, of any one conception of the predicate’s being true or false of any one given object; and
of them - is a sufficient condition for our grasping the sense of that expression, that in turn is derived from his prior general understanding of atomic sentences
since one cannot grasp the thought expressed by a sentence unless one grasps containing it. ■

the senses of all its constituent expressions. But we grasp the sense of a The classic case is that of the logical constants (including the quantifiers).
sentence by knowing the senses of its constituent expressions: it is because we A logical constant can stand within the scope of another logical constant. Frege
already know’ those senses that we are able to understand new’ sentences we perceived, however, that, in order to give the sense of a logical constant in all
have never encountered before, expressing thoughts we have never previously contexts, it is sufficient to describe its contribution to determining the truth­
entertained. Our understanding of such a sentence cannot therefore constitute conditions only of sentences of w’hich it is the principal operator. For the
our understanding of its constituents: wre must already have been able to isolate truth-conditions of complex sentences depend systematically on those of their
the contributions made by them to the thoughts expressed by other sentences immediate constituents, the subsentences to w'hich their principal operators
containing them, so as to put them together to compose that expressed by the are applied; here wre must treat an immediate constituent of a quantified
new sentence. But what is it to come to grasp in advance this sense attributable sentence as an application of the (in general complex) predicate to any one
to a specific expression capable of occurring in a wide variety of sentences? Is specific object. Hence, in analysing the sense of a complex sentence in which
it to learn the sense of that expression taken on its own? That would violate some logical constant figures as a non-principal operator, we have successively
the context principle (considered as applying to sense). It is meaningless to to consider ever simpler constituents; by the time we attend to the contribution
speak of grasping the sense of an expression conceived as standing on its own, made by the given logical constant, we shall be considering a constituent of
independently of any sentence in which it occurs. Its sense just is its contri­ which it is the principal operator.
bution to thoughts expressed by sentences of w’hich it is part; to regard the The context principle applies to all expressions, and in particular to all
expression as standing on its owm, independently of any sentence, is to destroy singular terms, those for actual as well as those for non-actual objects. It
the whole conception of its possessing a sense. erefore rules out that conception of a grasp of the meaning of a proper name
The escape from this dilemma requires us to regard sentences, and the as consisting in a direct mental apprehension, unmediated by language, of the
thoughts they express, as ordered by a relation of dependence: to grasp the 0 ject named and an association of the name with it. On the contrary, an
thoughts expressed by certain sentences, it is necessary first to be able to grasp nn erstanding of the name, as of all other expressions, comprises a grasp o
those expressed by other, simpler, ones. To grasp the sense of a given at determines the truth-value of a member of some characteristic range o
expression requires us to be able to grasp the thoughts expressed by certain fences containing it. There is no such thing as an immediate apprehension
sentences containing it: if it did not, we should be able to grasp that sense in tJ" ° ’eCt’ d *s onbr by coming to grasp the use of proper names, or o er
isolation, contrary to the context principle. Not, however, of all sentences s’ referring to them that we form any conception of objects as persisting
 204 Frege: Philosophy ofMathematics 16 The Context Principle 205
constituents of a heterogeneous, changing reality and as identifiable as the 1 definition’; it was not a contextual definition, nor a definition of
same again. Mere presentation of the object fails to determine how- it is to be ,contcxt‘.13 inCC permit elimination of the cardinality operator from
any k’nd> s^ce yyright was therefore quite wrong to treat contextually defined
recognised as the same again, unless some criterion of identity with which we
every senton case for evaluating the claim of the context principle to
are already familiar is presumed; and we can become familiar with such a
criterion only by coming to master the use of terms for objects of that sort. terms as ^ription of referCnce to terms for abstract objects.
When the term stands for some actual object, we may take the most basic justify an ■ tQ apprOach the question afresh for the crucial cases of
characteristic sentences containing it, a mastery of which is required for an " c t us vajue ranges Wright undoubtedly has strong grounds for attribu-
understanding of the term, to be what were called in Frege: Philosophy of nufflbersjm robust view for these cases. Frege quite explicitly claimed, in
Language ‘recognition statements’: that is, statements of the form ‘This is T', t*nl’ t(/ to have secured determinate references for value-range terms;
where T is the term in question, or ‘This 5 is T, where 5 is a sortal such as Grundgese jntentjon jn Grundlagen to do so for numerical terms. If
‘person’, ‘street’, ‘river’, etc., that is, a general term carrying with it a criterion °bVi° were any suspicion that he meant an attribution of reference to them to
of identity. The presence of a demonstrative pronoun or adjective is essential ?erC derstood as a mere fafon de purler, it would be dispelled by his explanation
to a recognition statement; such recognition statements can be regarded as i
be "crundgesetze of the (first-order) quantifier, which proceeds along what
basic, when the object is an actual one, because actual objects are characterised I
have become standard objectual lines. The domain over which the individual
as ones that can affect our senses, and hence as ones that we can perceive. I
variables of the formal system of Grundgesetze are to be taken as ranging
Doubtless, someone whose use of a proper name was confined to recognition consists it appears, solely of the two truth-values together with the value-
statements might be denied as yet to have a complete grasp of the sense of ranges i e the referents of sentences and of terms formed by means of the
the name; to attain that, he must surely learn the use of atomic sentences abstraction operator. Yet his stipulation regarding the universal quantifier, in
applying some predicate to the object named. But the understanding of such § 8 of Grundgesetze, Volume I, reads thus:
predications may plausibly be thought in turn to rest upon a grasp of predi­
cations in which the object is identified demonstratively; if so, recognition ‘_ i— 0(a)’ is to refer to the value true if the value of the function 0(c)
statements are more basic than any other sentences containing the name. It is is the value true for even' argument, and to refer to the value false
also true that we use a great many proper names of objects that no longer
exist, or are too large, too small or too remote for us to be able to perceive.
otherwise.
But it is at least arguable that our conception of what it is to pick out an object
by the use of such a name is founded upon our graspJ f the use of names of Frege does not here give a substituUonal explanati^ for even
objects of the same or related sorts which we can encounter or perceive. does not say that ‘ is to have the value W has
value-range term ‘f that can be constructed in die^onnal 1
the value true. He conceives of every functiona -T which is defined
Abstract terms constructed in the language as having as its ire er vajue just
Non-actual objects cannot be perceived, and they cannot be indicated, save on every object in the domain; 1---- — , r jn every other
by deferred ostension, by means of demonstratives. Recognition statements, case this function has the constant value true, an . dut, bv his stipu-
properly so called, cannot therefore exist in their case. The definability of one case. There is therefore no doubt that Frege is ass determined a
expression in terms of others is only an extreme instance of the relation of lations concerning the abstraction operator, e as determined just
dependence of sense; more usually, it is merely that the understanding of genuine reference for every’ value-range term, u
certain sentences presupposes an understanding of others, without there being which objects compose the domain. ^e Abated by its
any possibility of replacing the former by the latter. In neither of the cases in The discussion in the foregoing paragrap ®PPe chanter, between the
which we are primarily interested - that of terms for numbers and that of neglect of the radical difference, stressed in the 1 d.fference is t00 wide
terms for value-ranges — does the explanation Frege offers take the form of a framework of Grundlagen and that of Grundgesetze. „cripdon of reference,
contextual definition. As we saw, this would still have been true even if he to allow the context principle, taken as warranting Qne framework and
had not gix en an explicit definition of the cardinality operator, but had rested I to have the same content when understood wi in ^nised, that Frege
content with the method of introducing it, by means of the original equivalence within the other, even on the assumption, yet to certainiy true that, in
alone, canvassed in §§ 63—5 of Grundlagen and misleadingly called by him a continued to maintain the principle in Grundgese * • ;n the sense ot
Grundgesetze, Frege took a robust view of value-r g
 16 The Context Principle 207
206 Frege: Philosophy ofMathematics
. n0 question can be raised whether it really stands for anything: the object
claiming to have secured for them a reference in the substantial sense employed 7 which it stands is given to us through our understanding of the term, which
within the semantic theory of that work. In the framework of Grundlagen, on • Fn turn constituted by our grasp of the senses of sentences containing it
the other hand, no such wholly robust view of numerical terms was available, 15 We may say in reply that, even on the most resolutely internalist view, there
because that substantial notion of reference was missing. Wright may neverthe­ . a further question to be settled, especially when a term-forming operator,
less claim that, even at that stage, Frege’s view of them had a certain degree 'nd therewith a whole range of new terms, are being introduced: the question
of robustness, in that the ascription of reference to them was no mere fafon of suitably determining the domain of quantification. This, however, was
de purler; on the contrary', Wright’s use of the expression ‘objectual reference’ something that Frege persistently neglected, a neglect which, as we shah see,
may be defended on the ground that numerical terms are understood as proved in the end to be fatal.
denoting elements of the domain of the individual variables, and quantification F Given Frege’s insistence on there being non-actual objects, as objective as
over that domain is understood objectually. Obviously, the latter claim can actual ones, referred to by means of abstract terms and belonging to the same
hardly be sustained by quoting the explanation of the quantifier given in domain of quantification as actual ones, Wright’s case for a robust interpre­
Gnindgesetze; the claim is nevertheless undoubtedly correct. If it were other- tation of the context principle as Frege intended it to be understood in
vise, it would be unintelligible that Frege allows the cardinality operator, in Grundlagen may well seem difficult to gainsay. It cannot be gainsaid if‘robust’
Grundlagen, to be attached to predicates applying to cardinal numbers, an
is taken to mean merely that there are genuine objects corresponding to the
operation upon which his proof of the infinity of the natural numbers depends;
terms under consideration, and that these objects fall within the domain of
Grundlagen makes no sense unless we take the referents of numerical terms
the individual variables. If, on the other hand, it is taken to mean that the
to be full-fledged objects falling within the domain of quantification. The
much more loosely expressed stipulation in Begriffsschrift, § 11, concerning the identification of an object as the referent of such a term is an ingredient in
quantifier reads: the process whereby the truth-value of a sentence containing it is determined,
Wright’s interpretation must be rejected as importing ideas from Frege’s
J—-i—0(d) signifies (bedeutef) the judgement that the function is a fact whatever middle period foreign to Grundlagen. On the Grundlagen view, we can ask
we take as its argument. whether the truth-conditions of sentences containing a term of the kind in
question have been fixed, and for a statement of those truth-conditions; we
Fairly clearly, this, too, is intended to express an objectual interpretation of cannot ask after the mechanism by which the truth-values of those sentences
the first-order quantifier, an interpretation that Frege appears to have put on are determined, nor, therefore, after the role of the given term in that mech­
it throughout his career. anism.
This semi-robust feature of the context principle, as understood even in When no definition is involved, there can be no question of an austere view
Grundlagen, nevertheless lays an extra burden on the explanations by means in Wright’s sense. When we shift to the perspective of Grundgesetze, there may
of which a new range of terms is introduced, when these do not take the form be some view to be taken that falls short of full robustness: one that allows
of an explicit definition. It is not necessary only to determine the truth­ that a reference has in some sense been secured for the terms being introduced
conditions of sentences containing those terms; it is necessary also to determine into the language, but denies that such reference can be construed realistically.
those of all sentences involving quantification, that is, to determine the domain Such a less robust view could not be called an ‘intermediate’ view, because
of the variables so as to include referents of all such terms. there is no austere view to stand on the other side of it to the robust view':
We must bear firmly in mind that, in Grundlagen, Frege drew no distinction ere *s only the nominalist view' according to which the putative abstract terms
between an expression’s being meaningful and there being something it stands ought to be expunged from the language altogether, or at least not only denied
for. The answer to the question what it is required to have for it to denote a r5^erence but declared incapable of occurring in true sentences.
something is therefore the same as that to the question what must be known o reject nominalism is to declare that abstract terms, as such, are unobjec-
if we are to understand it: the questions are not differentiated. The answer, 'onable. To recognise that there is no objection in principle to them requires
in both cases, is: determinate truth-conditions for sentences containing it* c nowledgement that some form of the context principle is correct, since
That is why the context principle - as employed in Grundlagen - makes a cinltra*Ct °^ects can neither be encountered nor presented. The context pnn
term s possession of a meaning internal to the language: we need only satisfy naiJ *n ^Ct a^so Sterns terms for actual objects, since a grasp o a proper
ourselves that truth-conditions have been fixed for all the sentences in which com6]111' 0^65 an understanding of its use in sentences, and thus of a re ative y
the term may occur, and no further question remains to be settled. In particu- Plex segment of language. The nominalist is unaware of this, however.
 208 Frege: Philosophy ofMathematics
he is old-fashioned, he harbours a mythical conception of a mental connection
between name and bearer. 11 he is more up to date, he entertains the equally
superstitious belief that, for me to refer to an object, that object must have
acted to initiate a causal process that eventuated in my utterance of the name.
In either case, his conception of reference prompts him to regard names of CHAPTER 17
actual (concrete) objects as thoroughly legitimate, and names of abstract ones as
spurious. To recognise abstract terms as perfectly proper items of a vocabulary
therefore depends upon allowing that all that is necessary' for the lawful
introduction of a range of expressions into die language is a coherent account The Context Principle in Grundgesetze
of how they are to function in sentences, even when those expressions have
the form of singular terms; and this is a version of the context principle. It is
as yet unclear, however, whether acknowledging the context principle to this
extent - the extent necessary for a repudiation of nominalism - carries with o question is: how did the serpent of inconsistency enter Frege’s paradise?
it a commitment to a robust or realist conception of reference as ascribed to Terms for logical objects must obey the same principles as all other
abstract terms, or is compatible with a less robust, non-realist conception of expressions. In accordance with the context principle for sense, the senses of
it; nor whether, in the latter case, the satisfaction of some further condition such terms consist in the contribution they make to the senses of sentences
will warrant a realist conception. in which they occur. The context principle further requires that those senses
This question cannot be answered without a clearer idea of how abstract cannot be thought of as given antecedently to the senses of all such sentences:
terms can legitimately be introduced into the language, when not by contextual they are given by the manner in which the truth-values of certain basic
definition. We know’ in advance, however, that Frege’s method of introducing characteristic sentences containing them are determined. The truth-conditions
them was not legitimate: it could, and in the case of value-range terms did, of more complex sentences containing those terms are then to be regarded as
lead to inconsistency. In relation to it, therefore, it is poindess to debate understood by appeal to (or explicable in terms of) those of the more basic
whether a more or less robust interpretation should be put on the notion of sentences. The contribution of one of the terms in question to the sense of
reference as applied to terms so introduced. This is in effect what Wright any such more complex sentence can then be conceived as its contribution to
does, on the plea that, when the method is used solely to introduce terms for
the senses of one or more (possibly infinitely many) basic sentences on which
natural numbers, no contradiction will ensue. But, if the method can lead to
the sense of the complex one depends. This holds good, as we have seen, for
contradiction, it is patendy unsound. Our task is therefore to locate the error
in Frege s procedure, and enquire whether it could be repaired: only then primitive predicates, for logical constants and for names of actual o jects. ,e
shall w’e be in a position to discuss abstract terms, including mathematical questions to be answered in any specific case are: (1) which are e asic
ones, in general. sentences? (2) what are the truth-conditions of these basic sentences, an ow
do we grasp them? (3) how do the senses of the complex sentences epen
on those of the basic ones? In the case of names of actual objects, e m
basic sentences are, or can plausibly be taken to be, recognition state™®n®'
For terms standing for logical objects, there are no such sentences. ’
then, on Frege’s account of the matter, are the basic sentences c arac
for the senses of such logical singular terms? . identitv-
In Grundlagen, the answer is plain enough: they consist o
statements in which a term of the kind being introduced figures ’bo
whtch a term of that kind appears on either side of the i en 1 ce
ose ’n which such a term appears only on one side. The ^ut
sen-es to give the truth-conditions of identity-statements of the •F
>s rejected as a claimant for providing on its own a comp e e - truth­
pl ?ew terms for cardinal numbers because it fails to ete ^Julius
nditions of identity-statements of the second type (that is,

209
 17 The Context Principle in Grundgesetze 211
2Jo Frege: Philosophy ofMathematics
L nd-md or Julius Caesar, is a number at all, and, if so, to what concept it
Caesar problem). We are hindered from examining the matter any further by aS. We can determine that the number of planets is odd, and is a perfect
Frege’s choosing to remedy the defect by resorting to an explicit definition. bC °nre since those propositions can be expressed as propositions about the
Explicit definitions are always in order, provided that the expressions used in
sqUt ,’t Dlanef, but we cannot determine whether or not it has a monarchy or
the definiens, and taken as already understood, are themselves in order. This
Xs assassinated in the Capitol.
forces us to attend to the completion, in Gnindgesetze, of the unfinished busi­
Frege proceeds to back this up with an argument to the effect that, given
ness of Gnindlagen. nv assignment of referents to value-range terms, a permutation of them would
\ye have accordingly to abandon our methodological assumption that, in
I not disturb the criterion of identity (i.e. that Axiom V would remain true).
/ Gnindgesetze, Frege maintained the original context principle, but now under­
stood as relating to reference, and examine how he actually tackled the unfin­ This argument appears to flout the context principle: for a similar argument
ished business. He did so in Volume I, §§ 9, 10 and 31, with relevant remarks would defeat any claim to have fixed the reference of the primitive vocabulary
in §§ 29 and 30. The context principle, as formulated in Gnindlagen, allots a of any formal language (provided, in the general case, that the extensions of
primaev, with respect to meaning, to sentences: it is only in the context of a the primitive predicates were also subjected to the permutation).1 When Frege
sentence that a word has meaning. We saw diat, in § 32 of Gnindgesetze, Volume gives his solution to the problem, however, it fails to meet the objection from
I, Frege continued to allot a primacy to sentences, with respect to sense: the the permutation argument; but it plainly rests upon a generalised context
sense of an expression is its contribution to the thought expressed by a sentence principle. ‘How is this indeterminacy’ of reference ‘to be overcome?’, he asks,
in which it occurs. He refers to sentences, in this paragraph, as ‘names of and answers:
truth-values’. This is because, in the theory1 of reference of Gnindgesetze, no
categorial distinction, theoretical or formal, is drawn between truth-values and By determining, for each function as it is introduced, what values it obtains for
value-ranges as arguments, just as for all other arguments.
objects, or between sentences and singular terms: truth-values are just two of
the objects in the domain (even if particularly distinguished ones), and a
He proceeds to do this for each of the functions that have so far been
sentence is simply a singular term that happens to have a truth-value as its
introduced, namely the relation of identity, the horizontal function and the
referent. There is thus no primacy allotted to sentences, with respect to
negation function. The references of these have been specified as follows. In
reference. In so far as the primacy of sentences is an integral part of the
§ 5, the horizontal function----- £ was laid down as being that whose value is
content of the context principle, as it figures in Grundlagen, Frege did not
the value true for the value true as argument, and the value false for any other
maintain that principle, as relating to reference, in Gnindgesetze.
object as argument. In § 6, the negation function —r~£ was stipulated to have
He did maintain a generalised context principle, however, which is funda­
the value false for the value true as argument, and the value true for any other
mental to his attempted justification for introducing the ? ostraction operator.
In Volume I, §§ 3 and 9, he states the principle (embodied, in § 20, in his object as argument. Finally, identity of course figures in Frege’s system as a
binary’ function £ = £ whose value is always a truth-value. In § 7 the value of
celebrated Axiom V) that the value-range c/(e) of a function/(£) coincides with
the value-range dg(a) of a function g(£) just in case, for every a,/(a) = #(<!)- this function, for which wre may’ here use the name ‘the equality function, was
stated as being the value true when the same object is taken both as the first
This is the criterion by which we can recognise a value-range as the same
again, when it is designated by a value-range term, that is, one of the form argument and as the second, and the value false in every’ other case.
WTiat is the assumption implicitly underlying Frege’s alleged solution of the
c0(e). In § 10, Frege raises the question whether the principle is sufficient
to determine the reference of each value-range term. His answer is that it is
problem
problem of derprminino- the
of determining rpfprpncps of
th a references value ra g terms? It is evidently
of value-range realism.
not, on the ground that, by appeal to it,

we can as yet neither decide whether an object is a value-range, if it is not given

to us as such, or, if a value-range, of what function, nor in general decide whether
a given value-range has a given property, if we do not know that this property is
i reference has not been adequately determined, P . wv form; but.the tion argument
connected with a property of the function to which it belongs. no longer adhered to the context principle for reference in perm** was „
to the problem of fixing the reference of valuf'ra''1 ting that his appeal to notjon of
on its own ground counters that suggestion by i . his understan J ration of itwth
Thisjs the Julius Caesar problem again. From the criterion of identity between ^rration. That aberration may be taken as >^n/0Xing a robust interpret*!
reference was indeed robust, but also the difficult}
numbers, we cannot determine whether
------- ■ an object not given as a number, such adherence to the context principle.
 2|2 Frege: Philosophy ofAliithcnuitus 17 The Context Principle in Grundgesetze 213
that a singular term of the formal language has reference if the result Of then proceeds to argue as follows. First, the negation function can
ErCgL t of account, since its argument can always be taken as being a truth-
inserting it into the argument-place of any functiona expression of the lan.
be left That is to say, the value of the function ~i~£ will always be the same
guage has a reference. Let us label this assumption GCP (lor generalised value. 111 (----- £); we therefore need not consider the result of inserting
context principle’). It is a generalisation of the context principle of Grundlagen as that °*.range term directly into the argument-place of the expression
in that neither sentences nor predicates play any distinguished role. It says,
a 'a t>e The case of the horizontal function may be reduced to that of the
roughly, that the term in question has a reference provided that every more
complex term of which it is a constituent has a reference. In the formal tn- function: for the value of the function-- £ is always the same as
language of Grundgesetze, sentences are treated syntactically as singular terms; eqUa f the function £ = (£ = £)• We have therefore to consider only the
and GCP accords them no role in the theory' of reference distinct from that that ]°tv function. Thus, in the end, despite the greater generality of the
of other complex singular terms. In that theory of reference, names of truth­ eqaa Ling principle, the problem comes down once more to determining the
values have no semantic role distinguishing them from names of other objects, i ^/-conditions of identity-statements, that is, to solving the Julius Caesar
despite the fact that they do have a special place in the Grundgesetze theory of j problem for value-ranges.
sense. E We have to stipulate the value of the equality’ function for any case in which
Not only has the context principle of Grundlagen been generalised: it has one of its arguments is given as a value-range and the other is not. Frege here
also acquired a stronger sense. For the notion of reference -with which Frege obsenes that ‘we have so far introduced as objects only the truth-vah.es and
is here operating is no longer the thin notion of reference (meaning or content) die value-ranges’; that is to say, all terms other than value-range terms, formed
of Grundlagen, under which metalinguistic statements about reference were to bv means of the abstraction operator, have been stipulated to have truth-values
be understood by reducing them to statements of the object-language: it is the as their referents. Hence, Frege argues, the matter reduces to the question
substantial notion of reference which serves as the central notion of his seman­ ‘whether either one of the truth-values is a value-range’.
tic theory'. Moreover, the principle has become more specific. A mere generalis­
ation of the principle, as stated in Grundlagen, would say that a term will have
a reference if we have supplied a reference for every more complex term The Julius Caesar problem solved
containing it; but Frege now' claims that a reference will have been secured Frege now gives an argument to show that we are at liberty to make a transsortal
to it provided only that wre have supplied a reference for every term formed i entification of the value true with any arbitrary value-range, and of the value
by inserting it into the argument-place of every’ primitive unary functional fase with any other arbitrary- value-range distinct from it. The argument is
expression, and of every’ unary first-level functional expression formed by filling essentially as follows. Suppose that we have a domain of value-ranges, with
one argument-place of a primitive binary first-level functional expression. none o which the two truth-values coincide. (More rigorously stated, suppose
Remarkably, for two such formally' distinct versions of the principle, the appli­ at we have a model of the system, in which no value-range term has the
cations Frege makes of them reduce to much the same: the solution of the same enotation as any sentence.) Choose any two extensionally non-equivalent
Julius Caesar problem. the^*?nS ar*d exPressibIe in the system. Define a function A' which maps
It so happens that all three primitive function-symbols listed by Frege as e t a ue true on to the value-range of h, and conversely, the value false on to
having already been introduced serve to form sentences; the functions to which mar'a,Ue ran£e di and conversely, and every’ other object on to itself, We
they refer have only' truth-values as values. In Frege’s argument, however, this
sente re^nterPret terms of the system as follows: we continue to treat
fact is not specifically alluded to. It is undoubtedly important to that argument
valu^nCeS aS den°ting one or other truth-value, just as before; but we take a
that the values of those functions have been explained, unproblematically» ejem ran6e term to denote the result of applying the function A to that
as objects with which we are presumed to be already familiar, and hence internet ° ■ ^orna'n which it denoted in the original model. The resulting
independently of the notion of a value-range; but the fact that these objects
two tr,r*i,at10? satisfy Axiom V, and will yield a model in which the
are truth-values is not treated as being of any especial importance.
Besides GCP, a further assumption underlies the argument of § 10. This s "ruth-values are also value-ranges.
value-/ "e are at ^^erfy t0 identify the values true and false with any two
is that, if the result of inserting a term into the argument-place of every class We ch°ose> Frege elects to identify the value true with its unit
primitive functional expression has a reference, then the result of inserting it unit cla/ ;t1Ue’range horizontal function) and the value false with its
into the argument-place of any functional expression will have a reference.
the valu S 6 Va^ue-ranSe of the function which maps the value false on to
’’’ e may' call this the ‘compositional assumption’. e true, and every' other object on to the value false.) Therewith, the
214 Frege: Philosophy ofMathematics 17 The Context Principle in Grundgesetze 215
Julius Caesar problem for value-ranges is solved: the truth‘i or falsity of a (iv) an expression for a second-level function which takes a first-level
statement identifying a value-range with a truth-value will now be determined function of one argument as its sole argument has a reference if the result
by the criterion of identity' for value-ranges, since each of the truth-values now of inserting in its argument-place a referential expression for a first-level
is a value-range. The task of determining the references of value-range terms function of one argument always has a reference.
is thereby completed, for the time being; for, as Frege remarks, ‘As soon as
there is a further question of introducing a function that is not completely' A high degree of circularity7 is evident in these stipulations: to determine
reducible to the functions already known, we can stipulate what values it is to whether an expression for a unary7 first-level function has a reference, we have
have for value-ranges as arguments; and this can be regarded as being as to know7 whether, when we insert in its argument-place a singular term that
much a determination of the value-ranges as of that function’. has a reference, it yields a more complex term that again has a reference; and
to determine whether a singular term has a reference, we have to know
Conditions for referentiality whether, when w7e insert it in the argument-place of a functional expression
I that has a reference, it yields an expression that has a reference. Frege is in
In §§ 29-31, Frege returns to the matter in more detail, essaying, in § 31, a effect satisfied, how'ever, that, provided that there are some expressions of
proof that even- singular term of his symbolism has a determinate reference; w'hich we can assert outright that they have a reference, we can use his
this would include as a corollary' that every7 sentence had a determinate truth­ stipulations as inductive clauses from w'hich to derive that other, in general
value. From this fuller treatment, we can discern his intentions more precisely. more complex, expressions have a reference.2 The stipulation governing the
In § 29, he lays down general conditions for an expression of each logical type
reference of singular terms clearly embodies the GCP.
to be said to have a reference. Frege of course regarded a binary function as
a function of two arguments, not as a function of one argument whose value
was again a function of one argument; as already noted, he admits only Frege’s consistency proof
functions whose values are objects (including truth-values). But, for the pur­
poses of these sections, he allows for the formation of an expression for a Armed with the stipulations of § 29, Frege proceeds in § 31 to set out his
function of one argument by inserting a singular term in one of the argument­ I purported proof that every7 term of his symbolism has a reference. He deals with
places of an expression for a function of two arguments. Then, if for ease of the interdependence of function-symbols and singular terms by establishing the
formulation we say that an expression is ‘referential’ if it has a reference, his referentiality of the primitive function-symbols with respect to a fragment of
main stipulations are as follows: the language containing terms for truth-values only. Relatively to this fragment,
expressions for the horizontal, negation, equality and conditional functions are
all referential, the functions for w'hich they stand hating determinate values
(i) an expression for a first-level function of one argument has a reference for truth-values as arguments. The treatment of the universal quantifier is
provided that the result of inserting a referential term in its argument­ W'orth quoting. We have to ask, Frege says,
place is always again a referential term;
whether it follows generally from the fact that the function-name ‘0(f)’ refers to
(ii) a singular term (‘proper name’) has a reference if something that ‘_ ___ 0(0)’ has a reference. Now ‘0(f)’ has a reference if, for
of inserting it m
(a) the result of in the argument-place of a referential every referential proper name ‘d’, ‘0(d)’ refers to something. If so, this reference
expression for a first-level function of one argument is always a referential is either always the value true (whatever ‘d’ refers to), or not always. In the first
term; and case ‘_— 0(fl)’ refers to the value true, in the second to the value /atse. It
thus follows generally from the fact that the inserted function-name ‘0(f)’ refers
(b) the result of inserting the given term in either of the argument­ to something, that ‘— 0(0)’ refers to something. Consequently the function­
places of a referential expression for a first-level function of two arguments name — (p(d) is to be included in the circle of referential names. This
is a referential expression for a first-level function of one argument; 0 ows in a similar w'ay for ‘—JL.—iiP' (t?(/3))’-

(iii) an expression for a first-level function of two arguments has a refer­ the ,fir!jt sentence § 30 runs: ‘These propositions are not to be construed as definitions of
ence if the result of filling both of its argument-places with referential assum0rdV t0 haVe a reference” or “to refer to something", because their application always
singular terms always has a reference; serve to SOme names have already been recognised as having a reference; they can however
step by step, the circle of names so recognised.’ The word 'names here covers
214 Frege: Philosophy ofMathematics 17 The Context Principle in Grundgesetze 215
Julius Caesar problem for value-ranges is solved, the truth or falsity of a (iv) an expression for a second-level function which takes a first-level
statement identifying a value-range with a truth-value will now be determined function of one argument as its sole argument has a reference if the result
by the criterion of identity for value-ranges, since each of the truth-values now of inserting in its argument-place a referential expression for a first-level
is a value-range. The task of determining the references of value-range terms function of one argument always has a reference.
is thereby completed, for the time being; for, as Frege remarks, ‘As soon as
there is a further question of introducing a function that is not completely A high degree of circularity is evident in these stipulations: to determine
reducible to the functions already known, we can stipulate what values it is to whether an expression for a unary first-level function has a reference, we have
have for value-ranges as arguments; and this can be regarded as being as to know’ whether, when we insert in its argument-place a singular term that
much a determination of the value-ranges as of that function . has a reference, it yields a more complex term that again has a reference; and
to determine whether a singular term has a reference, we have to know’
Conditions for referentiality whether, when w:e insert it in the argument-place of a functional expression
that has a reference, it yields an expression that has a reference. Frege is in
In §§ 29-31, Frege returns to the matter in more detail, essaying, in § 31, a effect satisfied, however, that, provided that there are some expressions of
proof that even singular term of his symbolism has a determinate reference; which we can assert outright that they' have a reference, we can use his
this would include as a corollary' that every sentence had a determinate truth­
stipulations as inductive clauses from which to derive that other, in general
value. From this fuller treatment, we can discern his intentions more precisely. more complex, expressions have a reference.2 The stipulation governing the
In § 29, he lays down general conditions for an expression of each logical type
reference of singular terms clearly embodies the GCP.
to be said to have a reference. Frege of course regarded a binary function as
a function of two arguments, not as a function of one argument whose value
was again a function of one argument; as already noted, he admits only Frege’s consistency proof
functions whose values are objects (including truth-values). But, for the pur­
poses of these sections, he allows for the formation of an expression for a Armed with the stipulations of § 29, Frege proceeds in § 31 to set out his
function of one argument by inserting a singular term in one of the argument­ purported proof that every term of his symbolism has a reference. He deals with
places of an expression for a function of two arguments. Then, if for ease of the interdependence of function-symbols and singular terms by establishing the
formulation we say that an expression is ‘referential’ if it has a reference, his referentiality of the primitive function-symbols with respect to a fragment of
main stipulations are as follows: the language containing terms for truth-values only. Relatively to this fragment,
expressions for the horizontal, negation, equality' and conditional functions are
all referential, the functions for w'hich they stand having determinate values
(i) an expression for a first-level function of one argument has a reference for truth-values as arguments. The treatment of the universal quantifier is
provided that the result of inserting a referential term in its argument­ w'orth quoting. We have to ask, Frege says,
place is always again a referential term;
whether it follows generally from the fact that the function-name *0(£)’ refers to
(ii) a singular term (‘proper name’) has a reference if something that ‘_ ___ 0(C)’ has a reference. Now ‘0(i)’ has a reference if, for
(a) the result of inserting it in the argument-place of a referential I
every referential proper name ‘d’, ‘0(zl)’ refers to something. If so, this reference
I is either always the value true (whatever ‘zl’ refers to), or not always. In the first
expression for a first-level function of one argument is always a referential I case ‘—— 0(C)’ refers to the value true, in the second to the value false. It
term; and I thus follows generally from the fact that the inserted function-name l0(£)’ refers
(b) the result of inserting the given term in either of the argument- 1 to something, that ‘— 0(C)’ refers to something. Consequently the function-
p aces of a referential expression for a first-level function of tw'o arguments ^ame ‘ — 0(C) is to be included in the circle of referential names. This
is a referential expression for a first-level function of one argument; follows in a similar way for ‘—JL—(<?($))’•

(nt) an expression for a first-level function of two arguments has a refer­ fasj sentence of § 30 runs: ‘These propositions are not to be construed as definitions of
ence i e resu t of filling both of its argument-places with referential I assuW°rdSL Ut0 haVe a reference” or “t0 refer t0 something”, because their application always
( Se mes . 1 some names have already been recognised as having a reference; they can however
singular terms always has a reference; I
mean t0 rV!den’ steP bY steP- the circle of names so recognised.’ The word ‘names’ here covers
1
mg ul expressions of all logical types.
218 Frege: Philosophy ofMathematics J 7 The Context Principle in Grundgesetze 219
by '<p'. Then _Q,. (5(0) is the value true only when for every suitable value-range ol 5. then 5 (the value-range of //)’. This quantified sentence will
argument the value of our second-level function is the value true. have the value true just in case we obtain a true sentence by filling the
argument-place of our second-level function-name by any arbitrary first-level
Obviously, this is a comment, not a stipulation, since it is not laid down what function-name. In particular, we might fill it by the function-name 7/(£)’, in
—i— Q, (5(/>)) is to be when tine condition is not fulfilled; and no expla­ which case we should obtain the sentence ‘if the value-range of h is the same
nation is given of what constitutes a ‘suitable argument’. The comment must as the value-range of //, then //(the value-range of//)’. Since the antecedent is
be taken as appealing to the pronouncement in § 20: readily demonstrable, this is equivalent to ‘//(the value-range of //)’, which,
Now we understand by ‘-J?/t(F)’ the truth-value of one’s always ob- when partially spelled out, is ‘for every 5. if the value-range of It is the value­
— 'tf(rt)
taining a name of the value true whichever function-name one inserts in place of ‘g’
range of 5, then/} (the value-range of //)’. This, however, is precisely the
in *—T------------ O'(T)’
sentence whose truth-value we are trying to determine. Thus the stipulations
l—i.----- d’(a). intended to secure for it a determinate truth-value go round in a circle. If,
Despite the lack of generality, this is the closest Frege comes in Grundgesetze instead of 7/(£)’, we had here taken ‘^(£)’, abbreviating ‘for even 5, if £ is the
to stipulating the reference of a term formed by means of the second-order value-range of 5, then not 5(£)’, we should, with a little help from .Axiom A,
quantifier. His amazing insouciance concerning the second-order quantifier have obtained the Russell contradiction.
was the primary reason for his falling into inconsistency. The argument of Without second-order quantification, Frege’s formal system would be para­
§ 31 requires a proof of the legitimacy of the general operation for forming lysed, but the set-theoretic paradoxes would not be derivable. A model for the
second-level function-names. Suppose given a sentence or other singular first-order fragment of the theory could be arrived at in the following way.
term. As a preliminary for attaching an initial occurrence of the second-order Let Dy consist of the two truth-values together with the natural numbers. For
quantifier, we must form from it a second-level function-name by omitting any «, let D„+| be the union of D„ with the set of all its finite and cofinite
one or more occurrences of some first-level function-name. To say that the subsets. The domain D is then to consist of every member of any of the sets
resulting second-level function-name had a reference would be to say that D„, for any finite n. In the resulting model, a natural number k, considered
every result of filling its argument-place with a first-level function-name had as an element of D (or a member of the transitive closure of an element of
a reference; it would then follow’ that the sentence resulting from attaching D), is to be identified with the set of all subsets of D having exactly k members.
the second-order quantifier had a reference. But how could the premiss be It will be found that D contains all value-ranges definable by means of the
established? It might occur to us to reason by induction that, if every term limited vocabulary’. (One might have expected, instead of ‘finite and cofinite
containing n occurrences of the second-order quantifier has a reference, and subsets’, to have to say ‘functions, taken in extension, whose values differ from
if it is determinate, for every’ such term, whether or not its referent is the value one another for only finitely’ many arguments’; but, since a set is to be taken
true, then a term with n + 1 occurrences of the second-order quantifier must as the value-range of a function having only truth-values as values, and all
have a truth-value as its referent. But such reasoning, intended to fill a gaping Frege’s primitive function-symbols other than the abstraction and description
void in Frege’s proof, would be fallacious. operators denote such functions, it is sufficient to construct the entire model
The fallacy lies in the fact that, in considering the results of filling the out of sets.) In view’ of the consistency' of the fragment of the language without
argument-place of the second-level function-name, w'e have to consider all the second-order quantifier, it is therefore pertinent to ask w’hether the proof
first-level function-names as candidates for filling that argument-place, and °f § 31 w'ould have been valid for that fragment.
these will include ones with an unbounded number of occurrences of the The most natural diagnosis of the error in the proof is that Frege fails to
second-order quantifier: the induction hypothesis therefore does not suffice pay due attention to the fact that the introduction of the abstraction operator
for our purposes. Suppose, for instance, that wre abbreviate the first-level brings with it, not only new’ singular terms, but an extension of the domain.
function-name ‘for every ft, if £ is the value-range of 3. then $(£)’ as W-6 As w’e saw previously, it may be seen as making an inconsistent demand on
Vi e may then consider the second-level function-name ‘if the value-range of the size of the domain D, namely that, where D comprises n objects, we should
h is the value-range of <j>, then 0(the value-range of //)’; by attaching the have n" n, which holds only for n = 1, whereas we must have n 2, since
second-order quantifier, we obtain ‘for every 5, if the value-range of h is the the tw’o truth-values are distinct: for there must be n" extensionally non­
equivalent functions of one argument, and hence n" distinct value-ranges. But
M here use English in place of Frege’s symbolism for the sake of claritv: I intend the expression
to be understood as wntten m the primitive notation of Grundgesetze, however.
ts assumes that the function-variables range over the entire classical totality
I 0 functions from D into D, and there is meagre evidence for attributing such

i
 220 Frege: Philosophy ofMathematics 17 The Context Principle in Grundgesetze 221
a conception to Frege. His formulations make it more likely that he thought down, since otherwise a predicate might be interpreted inconsistently, as being
of his function-variables as ranging only over those functions that could be true of an element as picked out by one index and false of the same element
referred to by functional expressions of his symbolism (and thus over a denu­ as picked out by another.
merable totality of functions), and of the domain D of objects as comprising In our case, the domain is to consist of the two truth-values and of value­
value-ranges only of such functions. We therefore have to locate the error in ranges. The simplest choice of an index set for the value-ranges is that of
his attempted proof more precisely. first-level functional expressions of one argument of the formal language. Two
Although the assimilation of sentences to terms in the Grundgesetze blocks function-names will be taken to be indices of the same value-range just in
an overt statement of the context principle as allowing a distinguished role for case, for every' singular term, the results of inserting that term in the argument­
'// sentences, it is surprising how close in practice Frege comes to conforming to places of those function-names refer to the same element. This, of course,
such an ungeneralised version of the principle. The natural way to lay down merely repeats Frege’s own stipulations, save for taking the function-names
the semantics of a formal language is to start by delineating the domain of the into the metalanguage as indices. Doing so makes glaring a circularity' in the
/ individual variables, and then give the intended interpretations of the primitive procedure: wre cannot determine w'hether two functional expressions are indices
symbols in relation to that; and this has of course become the standard way. of the same value-range until we have fixed the interpretations of the primitive
Frege does nothing of the kind. He indicates that the two truth-values are to symbols, and we cannot safely do this until we know which indices relate to
belong to the domain; and they are the only two objects that he assigns directly the same value-ranges. If we decline to follow' Frege’s lead, and specify
to terms as their referents. Having first introduced various symbols capable of instead that no value-range is identical with either truth-value, we can take the
forming only terms referring to truth-values, he then introduces value-range extensions of the horizontal, negation and conditional functions as determined.
terms. He does not stipulate directly what the referents of these are to be, and But the equality function is a different matter: its value for value-ranges as
employs no vocabulary for doing so. Instead, he argues that he has provided arguments, considered as given by their indices, depends on determining
a reference for each value-range term, by means, in effect, of having deter­ whether those indices denote extensionally equivalent functions, which is a
mined the truth-value of any sentence in which that term occurs. How would matter of the truth or falsity' of a universally quantified identity-statement. In
it have been if he had gone about the task in what has become a more orthodox general, in order to obtain a determinate interpretation of a formal language,
manner? we must first specify, without circularity', what the elements of the domain are
The intended domain of a formal language may be specified either by to be, before we go on to specify the intended interpretations of the primitive
external or by internal means. An external specification characterises it as predicates; and this applies even if the only primitive predicate that gives any'
comprising certain objects with which we are presumed already familiar - the difficulty' is the sign of identity. To specify the domain, we must at least have
natural numbers, for example. We shall then in general need some means an index set; if wre do not assume that distinct indices always determine distinct
of singling out in the metalanguage particular elements of the domain so elements, wre must say when they do and when they do not. It is only after we
characterised. An internal specification requires only a comprehensive means have so specified of what the domain is to consist that we are at liberty to
of singling out any element of the domain, without identifying those elements specify the relation denoted by the identity-sign in the simple manner adopted
with objects given in any other way. An external specification is not, of course, by Frege, namely as holding between an element a and an element b of the
external to language as such - it is internal to the metalanguage - but only to domain just in case they are the same. Frege, on the other hand, omits to
the object-language. An internal specification nevertheless appears in better specify the domain, and, having explained the sign of identity in this manner,
accord with Frege’s employment of the GCP in Grundgesetze'. he certainly
proceeds to lay down the condition for the truth of a statement of identity
infers from that principle that an external specification is not required. In order
etween value-ranges under the guise of fixing the reference of the abstraction
to effect an internal specification of the domain, we need indices, drawn from I
I operator. That will depend upon the truth of a universally' quantified statement,
an already known index set, by means of which to pick out individual elements:
0 COmplexity depending on the function-names out of which the value-range
we can, when convenient, first specify the index set, without presuming that
terms were formed. The truth-value of that statement will in turn depend
different indices determine distinct elements, and subsequendy lay down the
upon the application of some complex predicate to every element of the
condition for two indices to determine the same element. After the domain
°main, and hence, in effect, upon the truth-value of every result of inserting
has been specified, the primitive predicates and individual constants have then
a vafoe-range term in its argument-place. Since these statements are likely to
to be interpreted with respect to that domain; but this can be done only when
the criterion of identity between elements with different indices has been laid '°ve further identity-statements between value-range terms of unbounded
°mplexity, Frege’s stipulations are not well founded: the truth-value of an
222 Frege: Philosophy ofMathematics
identity-statement cannot be construed as depending only on die references
of less complex terms or on the truth-values of less complex sentences.
Although there is in fact no danger of inconsistency in the fragment of Frege’s
system with only first-order quantification, he has provided no valid proof of
its consistency, because he has not succeeded in specifying the references of CHAPTER 18
all its terms. For that reason, he has failed to justify die introduction of value­
ranges.

Abstract Objects

What the contradiction signified for Frege

The inconsistency of Frege’s Grundgesetze system was not a mere accident
(though a disastrous one) due to carelessness of formulation. He discovered,
by August 1906, that it could not be put right within the framework of the
theory, that is, with the abstraction operator as primitive and an axiom govern­
ing the condition for the identity of value-ranges: but the underhing error lay
I
much deeper than a misconception concerning the foundations of set theory.
It was an error affecting his entire philosophy.
1
The context principle is crucially important to the philosophy of arithmetic
presented in Grundlagen\ and its generalised version is of equal importance to
Grundgesetze. Frege was engaged in completing the work that Bolzano had
begun, of expelling intuition from number theory and analysis (while lea-sing
it its due place in geometry). Bolzano had thought it important to prove
fundamental results in real analysis - the mean value theorem, for instance -
I by methods proper to the subject, and so without appeal to geometrical
I intuition, even though, when conceived in terms of their geometrical represen­
i tation, they appeared self-endent. It seems obvious to intuition that the graph
i of a function which assumes a negative value for x = 0 and a positive value
for x = 1 must cross the .v-axis somewhere in the interval; but it can be proved
without appeal to intuition, and therefore must be. This is partly because we
owe it to the subject to prove any truth we are concerned to assert if it is
capable of proof; but also because what appears self-evident may not be true.
It seems obvious that a curve contained within a finite interval must have a
tangent at all but finitely many points; by being the first to construct a continu­
ous function nowhere differentiable within an interval, Bolzano showed this
to be false. In Grundgesetze Frege characterised the aim of Grundlagen as having
een to make it probable that arithmetic is a branch of logic and does not
need to borrow any ground of proof from experience or from intuition’.1 The
U11 question with which Grundlagen, § 62, opens is, ‘How, then, are numbers

' Gmndgesetze, vol. I, p. 1.

223
 224 Frege: Philosophy ofMathematics 18 Abstract Objects 225

to be given to us, if we can have no ideas or intuitions of them?’. If intuition Frege been concerned only with number theory, and not also with analysis,
was really to be expelled from arithmetic, an answer to this question was and had he been able to solve the Julius Caesar problem for numbers, as he
imperatively demanded. If it could not be provided, then nothing would remain thought he solved it for value-ranges, then it would not have impaired his
but to fall back, after all, upon either a physicalistic or a psychologists concep­ logicist programme to take the numerical operator as primitive. But, in defi­
tion of number, or else to do what was contrary' to all Frege s instincts,2 but nition by logical abstraction (by means of equivalence classes), Frege believed
which in the last year of his life he felt himself driven to do, reduce arithmetic that he had found a uniform method of achieving similar results in a range of
to geometry. It was the context principle that enabled him to explain how analogous cases; and classes and value-ranges were of other uses as well. To
numbers could be given to us, yet neither by intuition nor by inner perception. make use of them therefore afforded a great economy of apparatus.
Frege came to call classes and value-ranges, including of course numbers What mattered philosophically, however, was not the definition in terms of
of all kinds, ‘logical objects’. Why ‘logical’? The term here does not have classes, but the elimination of appeals to intuition, a condition for which was
/ reference to what is required for inference: Frege was not claiming, therefore, the justification of a general means of introducing abstract terms, as genuinely
that overt reference to classes is essential for reasoning. Cardinality generally, referring to non-actual objects, by determining the truth-conditions of sen­
and numbers in particular, indeed enter into deductive inference, and so tences containing them. The contradiction was a catastrophe for Frege, not
numbers qualify on this score as logical objects; but classes - extensions of particularly because it exploded the notions of class and value-range, but
concepts - are to be recognised as logical objects independendy of the identifi­ because it showed that justification to be unsound. It refuted the context
cation of numbers as special cases of them. It would have seemed obvious to principle, as Frege had used it.
anyone at the time that the notion of the extension of a concept was a logical It is for this reason that the ontology' of the late essay ‘Der Gedanke’ is so
one: it was precisely in a treatise on logic that one would expect to encounter different from that of Grundlagen. In Grundlagen objects are divided into
it. Admittedly, what was said about extensions of concepts in such a treatise subjective ones - ideas - and objective ones, and the latter into actual ones -
could probably be expressed in terms of the extensional properties of concepts, material objects like the Earth - and non-actual ones like the Equator and the
rather than of their extensions considered as objects and hence as falling under numbers. In Grundgesetze the emphasis is upon logical objects, although Frege
k
further concepts; but that does not of itself impugn their logical status when warns that the realm of the objective may not be exhausted by physical and
so considered. logical objects.
The term ‘logical’, in the phrase ‘logical objects’, refers to what Frege always
picked out as the distinguishing mark of the logical, its generality: it does not We can distinguish between physical and logical objects, admittedly without an
relate to any special domain of knowledge, for, just as objects of any kind can i
i exhaustive partition being thereby given.3
be numbered, so objects of any kind can belong to a class. The style of
objection to logicism now exceedingly frequent is therefore quite beside the In ‘Der Gedanke’, however, the ‘third realm’ of objects that are not, or not
point: the objection, for instance, that set theory is not part of logic, or that it fully, actual appears to consist exclusively’ of thoughts and their constituent
is of no interest to ‘reduce’ a mathematical theory to another, more complex, senses. These could not be dispensed with; and there was no need to dispense
one. Much of this derives from hindsight, i.e. from the view we have of set with them, since their existence could be recognised without appeal to anything
theory7 long after the discovery of the paradoxes; but in any case it mistakes resembling the context principle. For the non-actual objects of Grundlagen
Frege s aim. Byr Frege’s criterion of universal applicability, the notion of and the logical objects of Grundgesetze, however, Frege no longer had any
cardinal number is already a logical one, and does not need the definition in philosophical justification; and so they quietly vanished from his ontology.
terms of classes to make it so. He did not himself speak of a ‘reduction’. What The problem how to introduce abstract objects would have been avoided if
w e call a reduction has two parts: the proofs, from absolute fundamentals, of Frege could have dispensed yvith mathematical objects altogether by construing
what we should ordinarily take as basic arithmetical laws; and the definition numbers of various kinds as concepts of second or higher order, beginning by
in terms of classes. The first is to ensure that we do not need any appeal to wilding on the rejected definitions of Grundlagen, § 55 (with the third of them
intuition at some early point - something not guaranteed by the mere fact that suitably amended). This was in effect Russell and Whitehead’s solution, or
number is a logical notion. The definition in terms of classes is not needed to ^°uld have been if Principia had been developed within the simple theory7 of
s ow arithmetic to be a branch of logic. To this extent, Wright is correct. Had > Wes, rather than the ramified theory required by the vicious circle principle.
ri metical theorems would then have been interpreted as yet more unprob-
2 See Grundlagen, § 19.
’ cnndgesetze, vol. II, § 74.
I
■

J
 226 Frege: Philosophy ofMathematics 18 Abstract Objects 227
lematically logical in character, and as admitting of yet more direct application. successor of m. It is also this feature which blocked the original equivalence
It is unclear whether Frege ever seriously considered this possibility; but, if from being part of a genuine contextual definition of the cardinality operator,
he did, the dependence of arithmetical statements for their truth on the gut just for this reason, it did not succeed in determining the truth or falsity
existence of infinitely many non-logical objects must surely have been for him of every statement of identity between numbers. An obvious example would
I be the status of Aleph-0 (called by Frege in Grundlagen The largest
a fatal obstacle.
number would evidendy be the number of all objects, i.e. the number belonging
to the concept identical with itself, the question whether this number was the
Discovery or invention? same as or different from the number of natural numbers would be left quite
i undetermined by Frege’s stipulations. His belief that he had in his possession
AATiere, then, do we stand? We cannot retreat to the nominalist fatuity of
regarding a belief in the existence of the Equator as a baseless and unexplana- a means of determining the truth-value of every statement of a formal theory'
j tory superstition. At the other extreme, Crispin Wright’s claim that we should - that is, of finding a uniquely appropriate model for it - was grossly overblown.
I accord to contextually defined abstract terms a genuine, full-blown reference Even had Frege felt able to renounce his claim to be able to prove the
to objects must be dismissed as exorbitant. Yet Frege’s attempted justification infinity of the sequence of natural numbers, and so evade the problem of
of the introduction of a range of abstract terms not explicable by contextual justifying the introduction of mathematical objects by construing cardinal num­
(or other) definition proved to be fallacious. It seems that there is no ground bers as second-level concepts, and real numbers as, say, second-level relations,
left to stand on. his second-order logic would still have been impredicative: from a statement
/ If Frege had been able to devise a solution to the Julius Caesar problem beginning with the second-order universal quantifier he allows the inference
J
for numbers that he considered as adequate as his solution of the problem for of the statement resulting from the insertion of any first-level function-name,
value-ranges, then, as already remarked, his programme for number theory including one again invoking second-order quantification, in the argument­
alone would not have been essentially affected by taking the cardinality operator place of the second-level function-name to which the initial second-order
t as primitive and as governed by the original equivalence (the criterion of quantifier was attached. The choice between predicative and impredicative
identity for numbers) as an axiom. There would then have been no inconsist­ theories involving quantification over functions, properties and relations or sets
ency; but ±e claim to have determined the truth-value of every sentence of is sometimes said to depend upon whether mathematical entities are regarded
the theory would have been as fallacious as the parallel claim for the language as created by our thinking or as existing independently of us. We are then at
of Grundgesetze. The criterion of identity' governs identity-statements connect­ a loss to know how to resolve a metaphysical issue couched in these metaphor­
! ing terms formed by attaching the cardinality operator to predicates defined ical terms. Was the monster group discovered as Leverrier discovered Neptune?
over a domain that includes the referents of those terms (the numbers); the Or was it invented, as Conan Doyle invented Sherlock Holmes? How can we
criterion is expressed by means of a statement involving those predicates. decide? And can the legitimacy' or illegitimacy' of a certain procedure of
Suppose it had been presented as an axiom governing the cardinality operator, reasoning within mathematics possibly depend on our answer?
taken as primitive, as Axiom V governs the abstraction operator in Grundgesetze", A mathematician, impressed by the ineluctability of mathematical proof, and
and suppose that, on the strength of the context principle, Frege had invoked by the unexpectedness of many mathematical results, may be impelled to insist
it as serving partiaEy to determine the references of numerical terms, as, in that he discovers them. A philosopher, struck by the contrast between the
Grundgesetze, he takes Axiom V to do for value-range terms. This would mean, intellectual labour of the mathematician and the manipulation by' the astron­
in effect, that he was taking it as a partial specification of what cardinal omer of physical instruments like telescopes and spectrometers, may feel
numbers there were; the specification would be completed by a solution to the I
equally strongly' constrained to regard the former as engaged on invention. Y et
Julius Caesar problem. So considered, it would be as objectionably impredicat- this appears to have little to do with whether the mathematician employs or
ive as the analogous specification for value-ranges: for the truth of any state­ abjures non-constructive methods. Though they' differ about what constitutes
ment of identity between numbers would depend on the extensions of two
Mathematical proof, it remains as ineluctable for the intuitionist as for the
predicates defined on a domain which included the cardinal numbers, and classical mathematician; the results obtained by the former may surprise him
whose composition the axiom was supposed to be playing an essential part in
as intensely as those obtained by the latter. It is pointless to debate whether
determining. It is just this feature which enabled Frege to prove, from the
original equivalence, the infinity of the sequence of natural numbers, by e mathematician resembles the astronomer more closely than the novelist.
showing, for any natural number n, that the number of numbers =£ w is a (the) e resembles neither in any illuminating way: no enlightenment is to be
attained by choosing between two such inappropriate similes.
 18 Abstract Objects 229
228 Frege: Philosophy ofMathematics
was non-denumerable, or not even well-ordered by a known relation; this
The contrast between mathematical and empirical enquiry concerns not so
much the discover)’ of individual objects as the delineation of the area of question may here be passed by. All that matters here is that such a picture
search. The astronomer need have no precise conception of the totality of _ or some more sophisticated one - is needed if impredicative second-order
celestial objects: he is concerned with detecting whatever is describable in quantification is to be understood: the impredicative comprehension axiom
physical terms and lies, or originates, outside the Earth s atmosphere, and he cannot, by itself, provide such understanding.4
need give no further specification of this ‘whatever’. In mathematics, by con­ The doctrine that, in mathematics, existence means consistency7 was at one
trast, an existential conjecture, to have any definite content, requires a prior time espoused by Hilbert:5 to ask whether mathematical entities of a given
circumscription of the domain of quantification. The difference between predi­ range exist is simply to ask whether any contradiction would follow from
cative and impredicative second-order quantification is not between a cautious supposing them to exist. Frege utterly repudiated such a view7. For him, the
primary error in the widespread practice of simply ‘postulating’ the existence
I
/
and a bold assumption about what mathematical entities exist: it is between
an axiomatisation that is self-explanatory and one that is not. If we are given
a first-order theory of which we suppose ourselves to have a determinate
of some range of mathematical objects lay in the unwarranted presumption
that it w7as consistent to do so: he insisted that ±e fact that no contradiction
interpretation, its extension by the addition of predicative quantification over lay on the surface was no proof that none was lurking unobserved. (In Grund-
properties and relations defined over the elements of the domain needs no gesetze, he attempted to heed his own warnings by providing a proof of consist­
further explanation. By assumption, we already have a clear conception of what ency; his tragedy lay in its being fallacious.) He argued, further, that the only­
it is for, say, a formula with one free variable, expressed in the language of way in which consistency could be proved was by demonstrating the existence
the original theory (supplemented, if necessary7, by terms denoting all the of the required range of mathematical objects, that is, in our terminology,
elements of the domain) to be true or false of any one arbitrary7 element. The by providing a model: we cannot infer existence from consistency, but only
< domain of the new property-variables can then be regarded as indexed by consistency7 from existence. In thinking this to be the only way to prove
those formulas: if they are governed by an axiom of extensionality, two such consistency, w7e know7 him to have been wrong. (Not, indeed, wildly wrong,
properties will be identifiable if the formulas indexing them are co-extensive. since most consistency proofs that do not provide a model for the whole theory
An impredicative second-order extension of the original theory7, by contrast, proceed by determining a model for any finite subset of ±e axioms, and Frege
would not be self-explanatory7, since to attempt an analogous explanation would I had no compelling ground to consider infinite axiom-systems.) In any case,
involve vicious circularity: we have already to know the range of the second- Frege considered that, even if consistency could be proved without providing
order quantifiers if we are to know what it is for a formula with one free
variable, but involving second-order quantification, to be true of an element ■* Such a more sophisticated picture is provided by progressing through the (cumulative) ramified
of the domain. So construed, the vicious circle principle makes no assertion hierarchy of properties of individuals into the transfinite, until a fixed point is reached. Here
second-order properties are those expressible by means of quantification only cner individuals and
about what does or does not exist: it merely distinguishes between what does first-order properties; properties of order a those expressible by means of quantification only over
and what does not require further explanation. Impredicative second-order individuals and properties of order less than a: a fixed point is an ordinal p such that all properties
quantification is most usually taken to gain whatever intelligibility it has from of order + I are already of order /3. The impredicative comprehension axiom will then be
satisfied if the property-variables are taken as ranging over the properties of order /J. This
a picture we find it natural to employ. This picture invokes, first, the conception picture is assuredly not that which originallv prompted the assumption that the impredicative
of a completed arbitrary7 assignment of values true and false (or numbers 1 and comprehension axiom can be satisfied, but, if it is acceptable as coherent and determinate, justifies
0) to the elements of the given domain. ‘Arbitrary’ here means that the that assumption. Whether it is so or not depends upon the determinateness and coherence ot the
assignment does not depend upon any general rule: the values are assigned conception of the totality- of ordinals necessarv to prove the existence of the fixed point p.
Hilbert says this, for example, in his letter to Frege of 29 December 1899: ‘For as long as I
randomly (perhaps pictured as effected by successive random choices). ‘Com­ ate been thinking, writing and lecturing about these things, I have always said ... : If the
pleted means that we are conceiving of such an assignment as allotting values arbitrarily posited axioms, with all their consequences, do not contradict one another, then they
to all the elements of the domain. Having formed this conception, we proceed ai\ true> and then the things defined bv the axioms exist. That is for me the criterion of truth
to form the further conception of the totality of all such arbitrary assignments:
this totality forms the domain of quantification of the property-variables, under­
I an of existence. The proposition “Every' equation has a root" is true, or the existence ot roots
■s proved, as soon as the axiom “Every equation has a root” can be added to the remaining axioms
2 ar,thmetic without a contradiction’s being able to arise by means of any consequences drawn
stood as satisfying an impredicative principle of substitution or comprehension yerr>rOm ^’^ert then refers to his lecture ‘Cber den Zahlbegriff, published in the Juhresberidil
axiom. W hether such a picture really does yield a coherent and determinate eutschen Mathematiker-l'ereinigung, vol. 8, 1900, pp. ISO—4-, and reprinted as Appendix M to
■,9Mdlagen der Geometrie, 7th edn?In that lecture he advanced the very same thesis, as he says
conception of a domain of second-order quantification is a notoriously debat­ nu k etter’ ’n ‘I carried out, or at least indicated, the proof that the system of all ordinary real
able question, especially w’hen the intended domain of the first-order theory7 n and dlat on the other hand the system of all Cantorian powers or of all Alephs
exist — as Cantor also asserts in a similar sense, but in slightly different words.
 230 Frege: Philosophy ofMathematics 18 Abstract Objects 231
weenhouse’ gas altering the climate - by the year 2005. Britain would only fulfil
a model, what mattered was the existence oi a model rather than mere formal that commitment if other, unspecified nations promised similar restraint.
consistency? In this, he was surely correct. Whatever mathematicians profess,
mathematical theories conceived in a wholly formalist spiiit are rare. One such
Save for ‘Margaret Thatcher’, ‘air’ and ‘sea’, there is not a noun or noun
is Quine’s New Foundations system of set theory, devised with no model in
phrase in this paragraph incontrovertibly standing for or applying to a concrete
mind, but on the basis of a hunch that a purely formal restriction on the
object (is a nation a concrete object, or a gas?). Ordinary literate people readily
comprehension axiom would block all contradictions. The result is not a
understand such paragraphs; few would be easily able to render them in words
mathematical theorv, but a formal system capable of sen ing as an object of
involving reference only to concrete objects, if indeed they can be so rendered,
mathematical investigation: without some conception of what we are talking
or even to understand such a rendering if presented with it. An ordinary
about, we do not have a theory, because we do not have a subject-matter. In
reader’s comprehension of the abstract terms does not consist in the grasp of
this case, the theory is finitely axiomatisable, so that a consistency proof without
constructing a model is unlikely; it is in any case a first-order theory, so that, any such procedure of translation, but in a knowledge of how those terms
I function in sentences: no reason whatever exists for supposing him to attach
by the completeness theorem, its consistency would entail the existence of a i
I a reference to ‘Margaret Thatcher’, but not to ‘the climate’ or ‘air pollution’.
model. But, if an angel from heaven assured us of its consistency, we should
still not have a mathematical theory until we attained a grasp of the structure The notion of reference to an object is employed to mark the difference in
of a model for it. linguistic function between a singular term and a predicate or relational
expression; and that difference is as salient in the sentence ‘Carbon dioxide
is a compound’ as in ‘Margaret Thatcher is a woman’. One can know a great
The status of the context principle deal about Margaret Thatcher without ever having had to identify her; but, to
The GCP is surely incoherent. Indeed, it can scarcely be called a principle, understand a personal name, one has to know that there is such a thing as
since it embodies no criterion for distinguishing those terms whose reference identifying its bearer. There being such a thing is what constitutes it as
r- is to be fixed by fixing that of more complex terms containing them from those
whose reference is to be given outright. But since, in Grundgesetze, the only
referring to its bearer, and explains our understanding of its use in predicating
something of its bearer. Identification of a county, say as that in which one is,
referents that Frege does specify outright are truth-values, the GCP there of a gas, say as being emitted from an exhaust pipe, of a political group, say
reduces in practice to the restricted context principle for reference. This as holding a meeting, all differ greatly from identifying a person, because
/ cannot be construed as entitling us to ascribe any but the most nominal kind counties, gases and political groups are things of very different kinds from
of referentiality to contextually defined terms; but since, in his mature period, people: but such identifications occur, and play the same fundamental role in
Frege repudiated contextual definition altogether, this is not too great a con­ our discourse about such things as the identification of people plays in our
cession. discourse about them. To deny to those things the status of objects, and to
V» e have, then, to restrict our considerations to abstract terms not under­ the corresponding expressions the function of referring to them, is to fall
stood by means of a method of transforming sentences containing them into into nominalist superstition, based ultimately on the myth of the unmediated
r
1 sentences from which they are absent. For such terms, the restricted context presentation of genuine concrete objects to the mind.
principle, considered only under its general formulation, rather than in the
light of the applications Frege makes of it, is scarcely open to question.
Informal discourse is permeated by abstract terms. Here is a paragraph taken The context principle in mathematics
at random from the front page of a national daily newspaper: The language of the mathematical sciences differs markedly from that of
Margaret Thatcher yesterday gave her starkest warning yet about the dangers of everyday discourse: it could be said that the semantics of abstract terms
g obal warming caused by air pollution. But she did not announce any new policy •furcates, according as we are concerned with one or the other. In the first
to combat climate change and sea level rises, apart from a qualified commitment P ace, terms capable of definition are likely to be introduced by definition,
that Britain would stabilise its emissions of carbon dioxide - the most important erms not introduced by definition can therefore be expected to be indefinable
Within the framework adopted for the theory in question; and this makes it
more problematic to understand how the senses of sentences containing them
are. fixed. In the second place, and more importantly, the concept of identifi-
atlOn is harder to apply to mathematical objects than to abstract objects of
 18 Abstract Objects 233
232 Frege: Philosophy ofMathematics
sentences containing terms for the objects of the theory, that those objects are
the kind referred to in informal discourse; and this renders it more difficult riven to us as the referents of those terms. In this manner, Frege thought, he
to justify the ascription of reference, in any full-blooded sense, to mathematical had hit on a universally applicable method of justifying the assumption of the
terms. existence of logical objects, as the objects of a mathematical theory, of any
Crispin Wright and others are right to see the example of directions, used range whatever. He had, however, fatally overlooked the circularity of the
by Frege in Grundlagen, as having guided his thinking in tliis matter. Even entire procedure: that of specifying the criterion of identity in terms of the
though he had no use for the conception of reference as effected by a mental truth of sentences of the theory, and, more generally, that of attempting
association between a term and its referent, considered as apprehended inde- I simultaneously to specify the domain and the application of the primitive
pendendy of language, he did allow a distinction between objects of which we
predicates to its elements.
could, and those of which we could not, have an intuition. Of a line, an
Frege’s discussion in Grundlagen, §§ 62-9, taught him, as he supposed, a
intuition was possible; of a direction, it was not. The distinction corresponds,
second lesson: namely, that all logical objects, or at least all those needed in
f I
at least roughly, with that between an object that can be an object of ostension,
and one that cannot. Hence the transition from speaking of lines to speaking
mathematics, could be defined by logical abstraction, except, of course, the
■
classes needed for such definitions. In Grundlagen, the notion of the extension
of their directions could be effected only by coming to grasp the senses of
sentences in w’hich directions are referred to; and in such sentences, a direction of a concept is introduced very tentatively: the idea of using it w'as evidently
could not be referred to directiy by a demonstrative phrase, but only by a one that had only recendy occurred to Frege. By the time Grundgesetze was
phrase picking it out as the direction of some line (possibly itself identified written, it had taken firm root: the general problem how to justify7 the introduc­
/ tion of logical objects of any one kind had been reduced to the particular
demonstratively). Having rightiy perceived that the fundamental class of such
sentences was that of statements of identity between directions, Frege leapt to problem how to justify the introduction of classes, or, rather, of value-ranges.
the conclusion that the basis for introducing any new' range of abstract terms Thus Frege was not merely in possession of a general strategy for justifying
must consist in the determination of the truth-conditions of identity-statements the introduction of mathematical objects: he believed that, by applying that
► * involving them. In a certain sense, this was not far from the truth. It led, strategy to justify' the introduction of value-ranges, he had justified the intro­
however, to the root confusion that allowed him to believe that he could duction of all of them, once for all.
simultaneously fix the truth-conditions of such statements and the domain Despite the lack of importance generally attached to modelling whole math­
over which the individual variables w'ere to range. This belief was a total ematical theories within set theory, we have not, in this latter regard, moved
illusion. To arrive at an interpretation of a formal language of the standard very far beyond Frege’s viewpoint: definition in terms of equivalence classes
kind, employing an essentially Fregean syntax, we have first to attain a grasp has been adopted as a standard device, to be applied whenever available. In
of the intended domain of the individual variables: it is only after that that we the main respect, however, that of the specification of the domain of a formal
can so much as ask after the meanings of the primitive non-logical symbols. theory, our perspective is utterly different from Frege’s. Nobody followed
The confusion into which Frege fell did not invalidate the context principle Frege in this matter; virtually none attempted even to understand him.7 We
as such; but it fatally vitiated his application of it. It is this confusion that recognise no universal and unquestionable method of specifying the domain
constitutes the central flaw in his entire philosophy of arithmetic. It wras in of a theory, but, on the contrary, acknowledge it as problematic how' this is to
consequence of it that he believed that he had discovered an incontrovertible be done and when it is possible to claim to grasp a domain. In the present
means whereby to fix the senses of all sentences of any precisely formulated context, a specification of a domain as consisting of objects presumed already
mathematical theory (or at least of any not demanding appeal to intuition). All known - the real numbers, for instance - is irrelevant: what matters is how it
that wras essentially needed, according to this conception, was, first, to fix the is to be specified when it is not taken as already known. Whether we have a
criterion of identity for the characteristic terms of the theory, and then to lay formalised mathematical theory or merely an as yet uninterpreted formal system
down a further criterion to determine whether any such term was to be taken epends on whether we have some intuitive grasp of the structure of a model
as having the same reference as a term for an object of any other kind (such Or theory'. The conception of the cumulative hierarchy, for instance,
as a truth-value) to which the theory was required to allude. Given this, the
p-J1™5 Hi,.bert> in 1904, merely criticised Frege for adopting a notion of set that led to the
domain of the theory was determined; the interpretation of other symbols of
D krikS’ "’lbout bothering even to refute Frege’s attempted emendation of his Axiom \: see
the theory would then be unproblematic, and the truth-values of all sentences Inter. “ ert’, ‘Ober die Grundlagen der Logik und der Arithmetik’, I erhandlungen des Dritten
of the theory would thereby be determined. By appeal to the context principle, 17+.ReW"a . AIatllefna'iker-Kongresses in Heidelberg tom 8. bis 13. August 1904, Leipzig, 1905, pp.
one could then infer, from the determinateness of the truth-conditions of all 129—38* rePr*nted *n J- van Heijenoort (ed.), From Frege to Gddel, Cambridge. Mass., 196/, pp.
 18 Abstract Objects 235
234 Frege: Philosophy ofMathematics
actual use is made of the notion of reference in Frege’s purported procedure
renders Zermelo-Fraenkel set theory a genuine mathematical theory-, in con­ I
fixing die truth-conditions: he tells us that, when he has carried it out, the
trast to Quine’s New Foundations. Quite obviously, however, that conception for
does not give us a precise, but only a highly generic, idea of a model for the terms being introduced will have a reference, but the procedure makes no
appeal to any relation between them and elements of the domain. As a result,
theory: it relies on the problematic notions ot the power set and of the totality we have no conception of any analogue of the notion of identifying an object
of ordinal numbers. If we think of the elements of a domain as individuated as the referent of a term which plays a role in all other cases, abstract or
by being associated with the elements of some index set, then Zermelo-
concrete; the referents of the newly introduced terms cannot be thought of in
Fraenkel sets may be indexed by well-founded trees, in which no two or more
any other way than simply as the referents of those terms, and hence the
nodes determining isomorphic subtrees lie immediately below' the same node.
analog}' with other cases, which ought to sustain all uses of the notion of
Here, characteristically, while the notion has some effect in conveying an
reference, is here lacking.
intuitive grasp of the structure described by the theory, the indices are not in
The problem what constitutes a legitimate method of specifying the intended
general finitely representable, and the conception of the index set involves the
same difficulties as that of the domain it indexes. Our grasp of what is meant domain of a fundamental mathematical theory - one we do not treat as relating
by speaking of all such trees remains indeterminate in the same two respects: unproblematically to an already known totality of mathematical objects -
the height of the trees (where the height of a node is the smallest ordinal remains intractable; Frege’s philosophy of mathematics contributed precisely
greater than the heights of the nodes below it), and the totality of trees of any nothing to its solution, and is in that respect gravely defective. Certain ways
given height. Set theory is, of course, the most problematic case, principally of specifying the truth-conditions of the statements of a theory dispel its
because an intuitively natural model tor it must be of such enormously high apparent ontological implications; in such cases, the context principle does not
cardinality (from the viewpoint of an ordinary Earth-dweller). The problematic of course apply. This occurs when the statements of the theory are so inter­
cases are precisely those from which we derive our conception of that cardinal­ preted that the terms they contain are not taken at face value as denoting
ity which the domain must intuitively have: any index set must have the same elements of the domain; an example is the Kreisel-Troelstra interpretation of
cardinality, and hence a challenge to our claim to grasp the domain can never the intuitionistic theory of choice sequences, under which no reference to or
be conclusively rebutted. This applies equally to the natural numbers, from quantification over choice sequences remains. To adopt such an interpretation
w’hich we derive our conception of a denumerable totality. No refutation can is not, indeed, merely to admit that it yields a truth-preserving translation, but
be devised to defeat, on his owm ground, a finitist who professes not to to treat it as giving the meanings of the statements it interprets. But, when the
understand the conception of any infinite totality: Frege wras mistaken in intended meanings of the statements of the theory are explained by first laving
supposing that there can be a proof that such a totality exists wrhich must down what the domain comprises, and then interpreting the terms of the
convince anyone capable of reasoning. theory as denoting particular elements of that domain, Frege’s context principle
In any given case, the difficulty is to know what w’e ought to allow as sufficing is entirely correct in pronouncing that there is no further problem of warranting
to convey a determinate conception of a domain of mathematical objects: this the ontological implications of the theory: whether or not the purported expla­
is one of the principal sources of divergent practices within mathematics, as nation is legitimate may be problematic, but, if it is, there is no further problem.
well as of disagreements in the philosophy of the subject. Some comments on This is because any such explanation will necessarily embody some means of
this are reserved for the final chapter; but Frege evaded all such discussion individuating particular elements of the domain, whether within the semantic
because he falsely believed that he had a short cut taking him straight to the account, by appeal to an index set or otherwise, or as the denotations of
final goal. It remains of importance not just to dismiss his view as totally terms of the theory belonging to some canonical range. In contrast to Frege’s
misconceived, but to pinpoint his error. The context principle allows us to procedure, such a method will supply an analog}' to the intrinsically analogical
ascribe a reference to mathematical terms provided that we have fixed the notion of identifying an object as the referent of a term; since this notion plays
truth-conditions of sentences in which they occur; but Frege was completely a unctional part in the interpretation, we are justified in ascribing a genuine
mistaken about how we can go about fixing such truth-conditions. His mistake re erence> robustly understood, to the terms of the theory, which Frege’s own
leaves us in perplexity about the content of the context principle: for, although rocedure gave no acceptable grounds for doing.
his method of fixing the truth-conditions was not a contextual definition, it
ci 1 6 our prolonged enquiry into the validity of the context prin-
gives rise to exactly the same doubt about how it can justify the ascription of asPfoJlCOnsidered aS ^e^rimating the use ofterms f°r abstract objects, is then
a genuine reference, robustly conceived, to the terms in question as we rightly
feel concerning terms introduced by contextual definition. That is because no
Principle, as used in Grundlagen, really tells us no more than that the
 18 Abstract Objects 237
236 Frege: Philosophy ofMathematics
This reply misconstrues F rege s conception of the relation between the
use of such terms is legitimate if we have succeeded in assigning truth­
notions of sense and of reference, which are indissolubly linked on his under­
conditions to sentences in which they occur. A nominalist would contest even
standing of diem. The theory' of sense rests upon the theory of reference as
this1 but from no sober point of view can it be disputed. What must be
disputed, however, is Frege’s - and Wright’s - idea of what is sufficient for a base; and Frege was right to conceive of them as so related. A semantic
determining the truth-conditions of sentences containing terms of a newly theory is not justified solely by its according the right truth-values to the
introduced kind. Impredicative specification of the conditions for die truth of sentences of the language, in the light of the way things are: it has also to be
identity-statements involving one or two such terms is not sufficient, contrary adapted to serve as a base for a correct theory of sense. In the celebrated § 32
to Frege’s belief and to that ot his disciple Crispin W right. It fails to fix truth­ of Grundgesetze, Vol. I, Frege referred to his preceding stipulations of what the
conditions for all sentences containing the new terms, when these terms are references of the expressions of his formal language were to be, and then said:
formed by attaching an operator to a predicate or functional expression; and
it fails to do so because of the lack of an independent specification of the Every such name of a truth-value expresses a sense, a thought. That is, it is
domain, which it attempts, but fails, to circumscribe simultaneously with its detemnined by our stipulations under which conditions it refers to the True. The
determination of the truth-conditions of sentences containing the new' terms. sense of this name, the thought, is the thought that these conditions are fulfilled.
When the context principle is construed, not merely as legitimating the use
of abstract terms, but as justifying the ascription of reference to them, the It therefore matters how the stipulations concerning reference were formu­
question has substance only when the notion of reference is understood as lated: they needed to be framed in such a way that a grasp of them would
belonging to a semantic theory for the language as a w'hole. In this case, the yield ±e intended sense. The passage can be read as saying that sense cannot
foregoing remarks about the method adopted by Frege and favoured by Wright be stated, but only shown by the manner in which the reference is stated. We
continue to apply: the procedure now attempts, but fails, simultaneously to do not need to draw so strong a conclusion from it: it is consistent with holding
circumscribe the domain and to determine the references of the new terms. that it is possible to state informatively what is required for a grasp of the
How, then, would it be if the new terms were introduced, not by a circular sense of a given expression. But it certainly requires that the sense must be
procedure of that kind, but by a genuine contextual definition? In such a case, shown by the way the reference is laid down in the semantic theory’. The point
no view stronger than an intermediate one could be taken of a claim that a of a semantic theory, and what is required for it to be a correct theory, is that
reference had thereby been conferred upon them; the reference so conferred it should be capable of serving as a base for a correct theory of sense.
would be reference only in the thin sense of Grundlagen, since the notion The conception of a semantic theory' - in Frege’s case, his theory' of Bede-
would play no role in the semantic account of how the truth-values of sentences utung - as describing the mechanism whereby the truth-values of sentences
containing the terms are determined. are determined is most easily understood when it can be taken as relating to
The retort might be made that there are not two notions of reference, a the canonical means by' which we decide their truth or falsity'. A realistic
thin one and a substantial one: there is only one notion, the thin one according theory like Frege’s, however, cannot be understood in this fashion; it must
to which ‘the direction of a' refers to the direction of a, whether or not the
nevertheless be taken as reflecting our grasp of the truth-conditions of sen­
direction-operator has been explained by means of a contextual definition.
tences. The notions of sense and of reference are thus in symbiosis: the
The illusion that any more robust notion exists arises, according to this reply,
semantic theory' is a base for the theory' of sense, but must for that very' reason
from linking the theory of reference too tightly to the theory of sense. The
reference of a singular term, like that of any other expression, is its contribution
e constructed with an eye to its role as such a base. A realistic semantic
to the determination of the truth-value of a sentence in which it occurs, in eory must thus be understood as embodying our conception of how the
ru -values of our sentences are determined by' the way things are. This
virtue of how things are. It need have nothing to do with our means of coming
to recognise the truth-value of such a sentence, which indeed depends upon peonies obvious when it is put by' saying that our grasp of the thought
our grasp of its sense. If the sense of the term has been given to us by ^'pressed by a sentence involves a grasp of its semantic structure: to know the
means of a contextual definition, then, admittedly, our route to recognising the nse of each component expression — that is, of each semantically significant
sentence as true may go through an initial transformation of the sentence, in 5^L°?ent ~ We muSt know its logical type, and hence the type of thing to
accordance with the contextual definition, into another sentence in which the Un(jC 11 refers5 and to grasp the thought expressed by' the whole, we must
term does not occur; but this has nothing to do with how the sentence is in ser?tand these components are related to one another in the sentence
objectively determined as true, as this is explained by the theory of reference. gras^ ^at t^Ie‘r re^erents together determine it as true or as false. A
0 the sense of the sentence thus comprises, but is not exhausted by, an
 18 Abstract Objects 239
238 Frege: Philosophy ofMathematics
The notion of reference, as applied to singular terms, is operative within a
understanding of how its truth-value is determined in accordance with its ,crnantic theory, rather than semantically idle, just in case the identification
structure, as this is explained in the theory of reference. of its referent is conceived as an ingredient in the process of determining the
We should distinguish between what the reference of an expression is, which truth-value of a sentence in which it occurs. Hence the context principle, if it
is independent of how it is given to us, and how the theory' of reference needs is to warrant an ascription of reference to a term, robustly understood, must
to state what its reference is, if it is to accord with Frege’s implicit requirement, include a further condition if it is to be valid. It is not enough that truth­
in § 32 of Grundgesetze, Vol. I, that the stipulation of its reference should show conditions should have been assigned, in some manner or other, to all sen­
what its sense is. If an expression is introduced by an explicit definition, or is tences containing the term: it is necessary' also that they should have been
taken to be understood by a tacit appeal to that definition, then its reference specified in such a way as to admit a suitable notion of identifying the referent
is whatever is the reference of the defining expression, because that is what
of the term as playing a role in the determination of the truth-value of a
the definition lays down. The referent is then given to us as the referent of
sentence containing it. With that further condition, the context principle ceases
the defining expression; and so, in order to show what the sense of the defined
to be incoherent, and gains the cogency Frege took it to have: it will then no
expression is, the semantic theory (theory' of reference) must stipulate its
longer give even the appearance of validating the means of introducing value­
reference precisely by means of that definition: the definition must be incorpor­
range terms w'hich led Frege into contradiction, the analogue of which for
ated into the theory’ of reference, and not just into the theory' of sense.
numerical terms Wright seeks to defend.
How, then, does it stand with terms introduced by contextual definition? In
the theory’ of Grundgesetze, the question does not arise, because contextual The additional condition will probably always be met by appropriate expla­
definition is rejected as an illicit procedure. If, however, it were admitted, nations - not amounting to contextual definitions - of terms for dependent
then, by the same token, the contextual definition would have to figure within abstract objects: those objects, like the Equator, whose existence is contingent
the theory’ of reference; for we should take the process determining the truth­ upon the existence and behaviour of concrete objects. It is probably also the
value of a sentence containing a term so defined as going via a transformation case for all legitimate means of specifying the domains of mathematical theor­
of the sentence which deleted (rather than replaced) the term. The theory of ies, the existence of whose elements is required by the mathematical character
reference would therefore assign no reference to the contextually defined term, of the theories to be independent of any matters of contingent fact. We must
but only a truth-value to sentences containing it; and this would correspond here leave in abeyance the question w’hether there in fact exist legitimate, non­
to the fact that determining the referent of the term would play no part in the circular methods of specifying the domains of the fundamental ±eories of
i
determination of the truth-value of such a sentence, as we conceived of this. classical mathematics. Frege never advanced the context principle as having
But would this not conflict with the generalised context principle of Grundge- the advantages conceded by Russell to the method of postulation: it merely
setze't Could wre not construe the contextual definition, not as denying a refer­ indicated nohat honest toil w'as called for. It was his error and his misfortune
ence to the term, but as fixing its reference by determining the values of then to have misconstrued the task, an error for which he paid with the
various functions for its referent as argument? We might try replying that the frustration of his life’s ambition.
latter interpretation would be in place only if the semantic theory required us The discussion has here been conducted throughout as if the distinction
to ascribe a reference to the term, and other terms of the same form, and that between concrete and abstract objects, or Frege’s corresponding distinction
this would happen only when the domain of quantification could be grasped etween actual and non-actual objects, w'ere a sharp dichotomy7, as it is usually
only as comprising the referents of those terms; we could add that, in such a assumed to be. In fact, it is nothing of the kind, but rather resembles a scale
case, we could not have a contextual definition, but only a contextual stipulation upon w’hich objects of varying sorts occupy a range of positions. The criterion
within the semantic theory, probably one of the misbegotten variety employed o causal efficacy cannot be unequivocally applied in all cases: Frege himself
by Frege. It is better simply to acknowledge that, if Frege had admitted c into difficulties over it, in ‘Der Gedanke’, concerning thoughts, which he
contextual definitions, there would have been a conflict, but to blame this on ^anted to classify as non-actual, but could not deny some influence upon
the incoherent character of the GCP. Indeed, within the theory of reference hjgntS’. f°r’ s°nieone judges some thought to be true, that may well affect
of Grundgesetze, the context principle for reference would be incoherent even
howeCtI°nS’ ThiS hardly bears upon the assessment of the context principle,
when the relevant contexts were restricted to sentences. When the notion of
acc e'er’ esPecially7 in application to mathematical objects, which, on any
reference is the instrument of a serious semantic theory, serving as the base
ont, occupy the extreme abstract end of the scale.
for a theory' of sense, the context principle simply cannot be sustained in full
generality', against that background, it is useless to mount a defence of it. °ut aU C°ntext Pfmciple, understood as including the further condition, rules
grounds for cavil at construing mathematical theories as having abstract
 240 Frege: Philosophy ofMathematics
objects for their subject-matter. Proposals to treat such objects as symbols, as
mental constructions, as fictions, as creations ol the human mind, are shown
by it to rest upon coarse misunderstandings of the functioning of our language:
as Frege held, they are given in drought, but not created by drought. When
wre have accepted this, we see how short a distance we have advanced. The CHAPTER 19
real problems of the philosophy of mathematics are far more specific: the
existence of abstract objects was never more than a pseudo-problem, and,
when we have recognised it as such, the real problems remain. Great credit
is due to Frege for enunciating the principle which exposes it as a pseudo­ Part III of Grundgesetze
problem, to which many other philosophers of madrematics have laboured to
produce pseudo-solutions. This credit ought not to be denied on the ground
of his mistake in applying his own principles, and so in effect treating it as
/'
providing too easy a solution to the real problems, as well as a definitive answer As already remarked, the division of Grundgesetze into volumes bears no relation
to the spurious ones; he suffered sorely enough for that mistake, after all. to the architecture of the book. Volume I contains Part I and most of Part II,
Part I being concerned with the syntax and semantics of the formal system,
/ and Part II with the entirely formal presentation of Frege’s foundations for
the theory' of natural numbers, together with the smallest transfinite cardinal.
Volume II, published ten years later, contains the remainder of Part II, and
about three-quarters of Part III, entitled ‘The real numbers’, together with
the Appendix on the Russell paradox.1 Frege obviously intended a third
P volume; the last words of the main text (§ 245) read ‘The next problem will
now be to show’ that there exists a positive class, as indicated in § 164. The
possibility will thereby be opened of defining a real number as a ratio of
/ quantities of a domain belonging to a positive class. And we shall then also
be able to prove that the real numbers are themselves quantities belonging to
the domain of a positive class.’ If Volume III had contained only the conclusion
’ It seems likely that most of vol. II was already written in 1893, or shortly afterwards. Most of
Frege’s references are to works published before that year. Of his own writings, the only exceptions
are his letter to Peano of 1896, published in the Rivista di matentatica for that year, and his t rrr
die Zahlen des Herrn H. Schubert of 1899; both are cited in brief footnotes (to §§ 65 and 153) that
could easily have been added subsequently. Frege cites only three works from after 1893 b* other
authors: Peano’s reply to his letter, published in the same issue of the Rivista-, J. Thomaes
Elementare Theorie der analytischen Functionen einer complexen Verdnderlichen (Halle) in its secon
edition of 1898; and Alfred Pringsheim’s article ‘Irrationalzahlen und Konvergenz unendhcher
Prozesse’ in the Encyklopddie der mathematischen Wissenschaften, vol. I, pp. 47-146, onginaUy
pubhshed (together with Schubert’s article) in Heft 1, issued in 1898. A long footnote to § ?
discusses Peano’s replv, and mav well also have been added later. Thomae’s book is discussed
and quoted from at great length;’but the first edition, containing all the cited passages, had been
J1 lshed in 1880. Pringsheim’s article is discussed briefly in § 72; its omission vou not impair
e continuity of the section of w'hich it forms part. ,
con?0 frJagments of Frege’s Nachlass obviouslv meant for inclusion in vol. II of Grundgesttze, on
h\Xmed With Princ>Ples of definition, the other with the notion of a
Schriften, pp. 164-81, Posthumous Writings, pp. 15.2-66^ .
2 (1897? dle de^nition °f the implication sign given in Peano’s ^ma 1 found
0897), and the latter refers to a book of E Czuber published in 1898. Ne.ther fragment foumd
fragment**0 *jC k°ok- discussion of definition in §§ 55-67 corresponds very 1
Pr<£h? ; und Frege in f^t treated the second topic in his ‘Was ist erne Function, of 1 •
> these were rejected additions to a text already substantially complete.

9J.1
 19 Part III of Grundgesetze 243
242 Frege: Philosophy ofMathematics
,tributes positively to answering the question. In Part III.l of Grundgesetze,
of Part 111, it would have been extremely short. Possibly Frege had in nund a col
Part IV dealing with complex numbers. It may be thought (hat that would
sections follow no logical sequence. Each after the first, which concerns
the
have been pointless, since it is easy to define the complex numbers in terms general principles of definition, is devoted to a particular rival mathematician
of the reals; but Frege was much concerned with applications, and the appli­ or group of mathematicians: (/>) to Cantor, (c) to Heine and Thomae, (d) to
cations of complex analysis are by no means immediately evident from the Dedekind, Hankel and Stolz, and (e) to Weierstrass. From their content, the
representation of complex numbers as ordered pairs of reals, or even from the reader cannot but think tliat Frege is anxious to direct at his competitors any
geometrical interpretation, which in any case he would have disliked as involv­ criticism to which they lay themselves open, regardless of whether it advances
ing die intrusion into arithmetic of something dependent upon intuition.2 his argument or not. He acknowledges no merit in the work of those he
However this may be, what we have is an uncompleted Part III, largely criticises; nor, with the exceptions only of Newton and Gauss, is anyone quoted
neglected even bv Frege’s admirers.3 vvith approbation. The Frege who wrote Volume II of Grundgesetze was a very
Unlike Part II, Part III is divided into two halves: III.l, entitled ‘Critique different man from the Frege who had written Grundlagen-. an embittered man
of theories of irrational numbers’, is in prose, while the uncompleted III.2, whose concern to give a convincing exposition of his theory of the foundations
entitled ‘Theory of quantity’, is occupied, like the whole of Part II, with formal of analysis was repeatedly overpowered by his desire for revenge on those who
proofs and definitions. In III.l, Frege attempted to do for real numbers what had ignored or failed to understand his work. The consequence is that the
he had done for natural numbers in Grundlagen, §§ 5-54. The same general reader is not directed, as in Grundlagen, along a path appearing to lead irresist­
plan is followed. Existing theories are surveyed and subjected to criticism; a ibly to Frege’s own theory' as the only surviving possibility. Instead, he finishes
svnopsis (§§ 156-9) is devoted to drawing the moral from the failure of the the critical sections so wearied by the relentless carping at every' detail of the
various theories so reviewed; and a brief concluding section (§§ 160—4) sket­ theories examined - almost always warranted, but never generous and fre­
ches the theory Frege intends to put in their place. quently irrelevant - that he has very’ little idea what fundamental objection
Unhappily, the attempt woefully miscarried. The critical sections of Grund- Frege has to them.
lagen follow one another in a logical sequence; each is devoted to a question
concerning arithmetic and the natural numbers, and other writers are cited
only when either some view they express or the refutation of their errors Weierstrass
The most lamentable example is the last critical section (e), on Weierstrass.
2 See Grundlagen, § 103; also § 19. Leaving his great contributions to the foundations of analysis unmentioned, it
3 In 1913, Philip Jourdain wrote an impertinent and monumentally tactless letter to Frege,
saying, ‘In your last letter to me you spoke about working at the theory of irrational numbers. Do descends rapidly into the grossest abuse.4 Frege’s criticism is justified, but the
you mean that you are writing a third volume of the Grundgesetze der Arithmetik't Wittgenstein and tone is unforgivable, directed as it is at a great mathematician; and since it is
I were rather disturbed to think that you might be doing so, because the theory of irrational aimed at what Weierstrass says about the natural numbers, it is completely out
numbers - unless you have got a quite new theory of them - would seem to require that the
contradiction has been previously avoided; and the part dealing with irrational numbers on the
of place.
new basis has been splendidly worked out by Russell and \Miitehead in their Principia Mathematical The section serves no purpose but to advertise the ill manners Frege had
Jourdain had obviously never looked at vol. II of Grundgesetze, or he would have realised that acquired. No reader could have gone on from it convinced that the theory of
Frege had already gone far towards expounding his own theory' of real numbers.
It is possible that the work to which Frege had been referring in the letter to which Jourdain real numbers to be expounded was the only one remaining in the field; none
was replying was that contained in the manuscript entitled ‘Das Irrationale, gegriindet auf Anzahl- could view the five first sections of Part III.l as presenting any ordered train
klassen’, irretrievably lost through American bombing and Heinrich Scholz’s failure to make a o reasoning, but only as attacking all Frege’s rivals in no particular sequence,
copy of it; see Albert Veraart, ‘Geschichte des wissenschaftlichen Nachlasses Gottlob Freges und
seiner Edition , in Matthias Schim (ed.), Studien zu Frege/Studies on Frege, vol. I, Stuttgart, 1976, b n ess replaced by a serious examination of Weierstrass’s theory of real num-
p. 98, no. 76. , ers ~ essentially superseded by that of Cantor - the section on him ought to
The only modern studies of Frege’s theory of real numbers are: Franz Kutschera, ‘Freges ave been deleted in its entirety.
Begriindung der Analysis , Archie fur mathematische Logik und Grundlagenforschung, vol. 9, 1966,
pp. 102-11, reprinted in M. Schirn (ed.), op. cit., pp. 301-12; G. Currie, Frege: an Introduction
to hit Philosophy, Brighton, 1982, pp. 57-9; idem, ‘Continuity and Change in Frege’s Philosophy
of Mathematics’, in L. Haaparanta and J. Hintikka (eds.), Frege Synthesized, Dordrecht, 1986, pp- 4 tjp
345—73; S.A. Adeleke, M.A.E, Dummett and Peter M. Neumann, *On a Question of Freges questionh°- had never thought about the subject in his lite were woken from sleep vith the
Right-Ordered Groups, Bulletin of the London Mathematical Society, vol. 19, 1987, PP- ^ose of W • "at *s number?”, he would in his first confusion give voice to expressions similar to
vol 8 1987* Pete2^4Sim0ns’ *FreSe’s Theory of Real Numbers’, History and Philosophy ofLogic, it Vears lat6'6-^355’’ FreSe "rote in § 149, and was so pleased with the remark that he repeated
Posthumn,. ir'-n. his lect(Jre series ‘Logik in der Mathematik’ (Xachgelassene Schriften, pp. 238-9,
,ls filings, p. 221).
 19 Part III of Grundgesetze 245
Frege: Philosophy ofMathematics
244 ndcd that they satisfy certain laws which we may lay down in advance as
pr°c WC want them to satisfy, there may be some latitude in the precise form
Principles of definition
die definition. He believed, however, that each operation should be defined
Even had the section on Weierstrass been deleted, die impact of Part III.l ° e and once only, and that the definition should cover all cases, so that the
would still have been muffled by its disordered arrangement. It might well be Oration is defined for all objects whatever, and not just those to which we
thought that the first section («), entitled ‘Fundamental principles of definition’, °P interested in applying it; an incomplete definition will allow us to form
was quite out of place. It is written as if it were an afterdiought that should terms for which no reference has been provided. Moreover, ‘if we have no final
have been included in Part I of the book; but this is not wholly so. The first definitions, we likewise have no final theorems’ (§ 61). If a partial definition of
fundamental principle, to which Frege devotes §§ 56-65, is the principle of multiplication may come subsequently to be supplemented by a further defi­
completeness; the second, that of simplicity of the expression defined, occupies
nition relating to a new, or wider, domain, we cannot affirm with assurance
only § 66. The principle of simplicity rules out contextual definition. That of
that 9 has only two square roots: for we cannot know that multiplication may
completeness rules out piecemeal definition; the practice that occurs, as Frege
not be defined over some new domain in such a way that for some element}
expresses it (§ 58), ‘when a symbol is first defined for a restricted domain, and
of it, distinct from +3 and -3, we shall have t.t = 9.
then used in order to define the same symbol once more for a wider domain’;
Frege’s opposition to piecemeal definition thus serves as a ground for
for example (§ 57), when ‘the definition is given ... for the positive integers
rejecting the conception, common in mathematical textbooks of his time, of
... and after many theorems a second definition follows ... for the negative
integers and O’. In a footnote to § 58, Frege cites Peano as endorsing the the introduction of new sorts of number as the extension of an existing number­
procedure and declaring it indispensable.5 system by the adjunction of new elements; so regarded, it is highly pertinent
It is this procedure wrhich is relevant to the main topic, how the real numbers to the main topic, and may well stand as the first section of his critique. It
should be introduced. The most natural w'ay to think of the introduction does not, however, supply very' strong grounds for rejecting the conception in
of negative integers, fractions, irrationals and finally complex numbers is as question; and Frege was aware that it did not. Certainly he was right that
successive additions to already given number-systems. In this case, as the operations applicable within distinct domains are distinct operations, and
number-system is extended, the arithmetical operations of addition, multipli­ should in principle be denoted by distinct symbols. But, even if we grant that
cation and exponentiation must be extended, too: defined originally for non­ it should be impossible to form terms lacking a reference, and even if we also
negative integers, they must be extended by new definitions, first to the negative grant that all functions should be defined for all objects as arguments, Frege
integers, then to fractions, then to irrational numbers and eventually to complex could perfectly well have handled the process of extending a given number­
ones. It was to this conception that Russell repeatedly objected in his Introduc­ domain had he believed that it provided the right framework for the transition
tion to Mathematical Philosophy. According to Russell, each new number-system, from natural numbers to real numbers. He was forced to represent functions
considered as subject to the basic arithmetical operations, contains a subsystem defined over restricted domains as relations, or, more usually, as the extensions
isomorphic to the previous one but not identical to it: the complex number of such relations, and employed the device repeatedly in Grundgesetze. Given
1 + Oi is a distinct object from the real number 1; the rational 1/1 is not to the domain of natural numbers, it would be easy to define a ternary relation
be identified with the integer +1, nor that again with the natural number 1. A(», w, k) as holding only between natural numbers and as obtaining just when
Frege shared Russell s opposition to the conception of the successive exten­ n + tn = k; the symbol + could then be defined by means of the description
sion of the number-system: that is why he is careful to use different symbols operator. If, say, the signed integers were thought of as resulting from adjoining
for the natural numbers 0 and 1, which were for him cardinal numbers, and e negative integers to the domain of natural numbers, a new ternary relation,
for the real numbers 0 and 1, writing the numerals for the natural numbers Inc U(hng the former one, could then be defined over the signed integers, and
with slashes through them. He was well aware that, when a new number­ a new symbol for addition defined in terms of it: the procedure parallels those
system is introduced - that of the real numbers or of the complex numbers - rege does employ, and could not be objected to. He shows himself conscious
the arithmetical operations have to be specifically defined for them, and that, at piecemeal definition can readily be avoided without any fundamental
variaHektterT \ t0 Frege’s letter’ ‘If what is t0 be defined containS an3]^ *n structure of the domains to which the operation so defined is
definitions of that n-n c * UnCJ? ^ose tetters, it appears to me necessary to give conditional
kindS o \deufin,tlons with ^otheses - and to give as many definitions as for ie W^en he remarks (§ 60) that ‘it is easy to avoid a plurality of definitions
bl definedkwhen ?and T °n wh,ch "V*™ °Ut ^ration. Thus the symbol a + b will first dom°^e and same symbol. Instead of first defining it over a restricted
L": ‘ ’h'" •» *n (j0 a!n’ and then using it for the purpose of defining itself over a wider
ain ~ instead, therefore, of defining the same thing twice - we need only
246 Frege: Philosophy ofMathematics 19 Part III of Grundgesetze 247
choose different symbols, and restrict the reference of the first, definitively, to positulation and creation
the narrower domain.’ '), aa serious discussion of Cantor’s theory of real numbers, is com-
It may have been because Frege realised that the rejection of piecemeal Section (^)>
Ttely misplaced. It contains matter far more informative about the grounds
definition had, in itself, few substantial consequences that section («) is not P?1 C Frege’s own theory than any other of the critical sections, and its proper
presented as having much bearing on the main topic, it could nevertheless Ice is at the end of the review of rival theories, not towards the beginning.
have been made to appear a relevant opening to the renew of prevailing
The next section (c) is very long, and constitutes a brilliant examination of
theories about the introduction of real numbers if the short exposition of the
formalism as a philosophy of mathematics, the only one Frege ever undertook.
principle of simplicity had been placed first, and the longer passage about the
principle of completeness more explicitly addressed to the problem of introduc­ It earns its place in a survey of theories of real numbers in part from the fact
ing new kinds of numbers.
that one of his criticisms of formalism is that it is incapable of handling the
Frege had in fact a much stronger ground for opposing the conception of concept of an infinite sequence. It does much more than that, however: it is
introduction as extension than his objection to piecemeal definition. His theory the one passage in Frege’s writings in which he makes explicit his views on
of real numbers differs from all others - from those of Weierstrass, Cantor, the application of mathematics. It thus has an importance independent of its
Dedekind and Russell - in omitting the intermediate steps of introducing the context; but it is^also crucial to its context, since the motivation for Frege’s
rationals and even of the signed integers: he goes straight from the natural theory' of real numbers cannot be understood save in the light of his view of
numbers to the real numbers, positive and negative, without first defining any application. Since his criticism of Cantor turns on precisely this, the placing
domain less extensive than the reals but more extensive than the natural of section (c) after section (b) instead of before it was a serious error of
numbers. His reason for doing so is that he identifies kinds of number by arrangement.
their applications: and, in the two published volumes of Grundgesetze (where In §§ 140-7 of section (d), Frege engages in a critique of the method of
complex numbers are not mentioned), he acknowledges only two kinds. Having ‘postulating’ the existence of whatever new' numbers - negative, rational, etc.
remarked in § 157 of Volume II that ‘we have interpreted real numbers as - are needed at a given stage. This critique closely resembles that contained
ratios of quantities’, he goes on to say: in the brief section of Grundlagen (§§ 92-104) on ‘Other numbers’. In both,
Hankel is used as an example of w'hat Frege is criticising; in Grundgesetze,
Since cardinal numbers (Anzahleri) are not ratios, we have to distinguish them Otto Stolz serves as a second example. It is in place in a general critique of
from the positive whole numbers. It is therefore not possible to extend the domain
methods of introducing the real numbers, since such postulation was a favourite
of cardinal numbers to that of the real numbers; they are completely disjoint
domains. The cardinal numbers answer the question, ‘How many objects of a device among mathematicians of the day for proceeding from the positive
given kind are there.’’, whereas the real numbers can be regarded as measurement­ integers to more extensive number-systems. Like most of the other sections,
numbers, which state how large a quantity is as compared with a unit quantity. however, it is out of sequence. Since it tells us less about Frege’s positive
views than the critique of formalism, it ought to come second, after section
While Frege makes clear, in this as in other passages, the ground of differen­ (fl) and before section (r).
tiation between natural numbers and real numbers, he rather oddly never The objections Frege raised to postulationism in Grundlagen were in line
makes quite explicit the reason why his practice diverges from others in not with Russell’s famous remark that ‘the method of “postulating” what we want
recognising a distinct domain of rationals or even of integers. It is apparent, has many advantages; they are the same as the advantages of theft over honest
oweyer, from his section (b) on Cantor that he saw the rationals, at least, as toil .6 cannot ‘proceed as if mere postulation were its own fulfilment :7 we
pnncipa y sen ing to answer the question, ‘How great?’, that is, as giving the aAe to show’ that there is a system of numbers, with operations of addition
magnitude of a quantity relatively to a unit quantity, and therefore as not and multiplication suitably defined upon them, satisfying the conditions we
warranting recognition as forming a domain distinct from that of the real es're- we were content to do no more than postulate such a system, w e
num ers. preliminary' distinction between the questions, ‘How great?’ and Ou d not be sure that its existence did not contain a hidden contradiction;
How many. , could have been appended to section (a): this would have
Postulation therefore demands at least a proof of the consistency of w hat is
oriented the reader in the direction in which Frege wished him to go.
• °^tu ated- However, the only way to establish consistency, according to Frege,
0 prove that a system of that kind does exist; postulation thereupon loses

7 f'rnndla< ^cllon 10 Moihemalieal Philosophy, London, 1919, p. 71.

ge’ ^'^1^>
 19 Part III 0/Grundgesetze 249
248 Frege: Philosophy of Mathematics r equal, within its realm, to God’s. It was by the word of his mouth that
all its point. In any case, even if we could prove the consistency of the ^heavens were made; God had only to say, ‘Let there be light’, and there
hypothesis in some other way, it would not give us what we need. Frege the " ht. So, likewise, for the postulationist, a mathematician has only to say,
invokes his favourite example of de Moivre’s theorem as a case in which there be a square root of 2, or of-1’, and there is one. So, throughout
theorems about real numbers can be deduced from one about complex num­ L tion (A FreZe speaks’ nOt °f P°stulation’ but of creation {Schaffen or
bers.8 For us to have the right to regard the theorems about real numbers as Tli(ipfu»g)> although he does use the former of the German terms in Grundlagen
true, it is not enough that the hypothesis drat complex numbers exist should Io Secondly, he has acquired a more sophisticated view of the whole matter.
not be self-contradictory, and that, if they existed, de Moivre’s theorem would ^Grundlagen, he had no doubts about the possibility of proving the existence
hold good of them: we need to know that they do exist. To prove that the of any mathematical system we need; for, as yet, he took for granted the
rational, real and complex numbers exist, Frege says, we shall have to proceed availability of extensions of concepts as logical objects familiar to all. In
as we did for cardinal numbers. We must first fix tine criterion of identity for Grundgesetze, all mathematical objects were indeed to be defined as value­
the numbers to be introduced; we shall then be able to define them, as before, ranges: but he now recognised that the existence of value-ranges themselves
I as extensions of concepts.9
The discussion in Grundgesetze follows very- similar lines; essentially all the
required justification, and could not be taken for granted. Their existence
could not be proved by defining them as something else; and so, in §§ 146-7,
same points are made. Although in Grundlagen Frege had called those who he faces the question whether his own method of introducing them was not
engage in the method of postulation ‘formalists’, he here recognises that the an instance of postulation or of mathematical creation. He asserts that it is
formalism he attacked in section (f) is a much more radical doctrine: the
not, or, at any rate, that ‘it is quite different from the lawless, arbitrary creation
postulationists do not deny that mathematical symbols serve to denote objects
of numbers by many mathematicians’. Without the means provided by value­
/ distinct from themselves (§ 145). Instead of the example from de Moivre’s
ranges, ‘a scientific foundation for arithmetic would be impossible’, he says; it
theorem, Frege considers a case in which a theorem is proved by appeal to
serves to attain ‘the ends that other mathematicians mean to achieve by the
an auxiliary object not mentioned in the theorem but belonging to the same
domain: specifically, a natural number invoked in a proof in number theory'. creation of new numbers’. These two paragraphs suggest an uneasiness that
was to be proved well founded.
I To prove that, ifp is prime, the congruences x" = 1 and xd = 1 (mod/>) have
the same roots, where d is the greatest common divisor of n and p - 1, we
need to appeal to a primitive root of p.w It is not sufficient, for the sake of Dedekind
f the proof, to postulate that a primitive root exists: we need to prove that it exists
f (§ 140). As Frege remarks, the case is similar to that of the ‘construction’ of The first two paragraphs of section (d) are devoted to Dedekind’s theory of
an auxiliary line in a geometrical proof. real numbers. The first, § 138, acknowledges that he is no formalist, in the
The example is an instance of Fregean overkill. No postulationist, however sense of Heine and Thomae; so the critical discussion is confined to § 139.
brash, ever supposed that he might, in number theory, postulate the existence Frege briefly describes the celebrated idea of a cut in the rational line. He
of a number, or, in geometry', of a line, that he needed for a proof: only whole then quotes Dedekind as saying;11
systems of mathematical objects - particularly number-systems obtained by
i adjoining elements to an existing one - are postulated in this manner. Frege’s Now whenever a cut C4b.4z) occurs which is not generated by a rational number,
point is, of course, that there is no essential difference between the procedures; conaruct (erschafferi) a new irrational number which we regard as completely
but he makes it by affecting to suppose that the postulationists themselves see enned by the cut (4b _42); we shall say that the number a corresponds to this
no difference. cut’ Or ^at it generates this cut.
The discussion in Grundgesetze diverges from that in Grundlagen in two
. 7Fe attacks this as an instance of mathematical creation; leaving Dedekind
notable respects. In Grundlagen, Frege speaks principally of‘postulation’. But,
on v’ he brst gives his example of the primitive roots of/>, and then launches
i postu ation is its ov n fulfilment, the mathematician possesses a creative
« critique of Hankel and of Stolz.
’ Grundlagen, § 97. Poi 7^ ?VaS uncluestionably right to criticise Dedekind for resorting at this
* Grundlagen, § 104. to construction’. Russell independently made the very same criticism.
10 Bx Fermat’s theorem, if p is prime ;and’ does
' not divide a, a"-' = 1 (mod />). When /> — 1 is X’unJeS’etekinid’ S,eli«keil “nd "rationale Zohlen, Brunswick, 1872. § 4, ‘Creation of the Irrational
“ “h ** *■ ‘» «■“ '• pn™
12 B. Ru-?.’, „ The italics are Dedekind’s.
> Principles ofMathematics, London, 1903, § 267.
 79 ^"^/Grundgesetze
strass the whole concluded by section W on c 2S1
250 Frege: Philosophy ofMathematics
What right have we to assume the existence of such numbers? What reason have as it now stands. one of the effects of Part UI°
we for supposing that there must be a position between two classes of which one
is wholly to the right of the other, and of which one has no minimum and the
other no maximum?

A mathematician has no right to announce that he will construct a range of

new mathematical objects satisfying certain conditions unless he provides, or
can assume known, a method of defining them accordingly.
As Russell saw, however, Dedekind’s appeal to construction is unnecessary:
the real numbers, rational and irrational, can be identified with the cuts
themselves, or rather, with, say, those whose lower class contains no greatest
element (or, more simply, with the lower classes of such cuts). The case,
as Frege ought to have recognised, is quite different from one in which
a mathematician postulates a system of numbers satisfying certain general
conditions. Dedekind had provided a totality’, composed of classes of rationals,
with which the real numbers could be correlated one to one; he had done all
the honest toil required. Frege could have perceived the possibility of emending
the theory’ so as to dispense with the need for any construction as well as
Russell did. In his review of Cantor, after objecting to Cantor’s introduction
of order-types by psychological abstraction, he commented that ‘what Herr
Cantor aims to get hold of can be unobjectionably defined’; had he chosen,
he could have said the same of Dedekind. Dedekind’s resort to construction
was not a means of avoiding labour. It was due solely to his philosophical
orientation, according to which mathematical entities are to be displayed as
creations of the human mind.
We rightly think of Dedekind’s as one of the leading theories of real
numbers. Frege’s curt dismissal of it undermines the claim of Part III.l of
Grundgesetze to survey the range of available theories of the real numbers on
offer, and was a discourtesy to his readers as well as to Dedekind. To the
former he owed it to provide a serious ground for rejecting Dedekind’s theory,
suitably emended, instead of a captious one fastening on an inessential detail.
The fact pointed out above, that both Cantor’s and Dedekind’s theories were
unsatisfactory' from Frege’s standpoint, in that they depended upon a prior
definition of the rationals which Frege, for arguable reasons, declined to see
as composing an independent number-system, would have provided one such
ground, that Frege could have expounded had he chosen. The two paragraphs
on Dedekind, together with the whole section on Weierstrass, illustrate how
gravely’ Frege’s embitterment impaired his ability to emulate in Part III
Grundgesetze the great achievement of Grundlagen.
An emendation of Part III.l would then have started with a slightly revised
section (a), followed by section (d) without the two paragraphs about Dedekind.
After this would have come section (c) on formalism. Ideally, this would have
been followed by serious studies of the theories of Dedekind and of Weier"
 20 The Critique of Formalism 253
luce irrationals, we need infinitely many numbers’, whereas ‘formalist
intIhmetic has only a finite set of numerical figures’. The claim of radical
f a)ism is to dispense with the need to vindicate the possession either of
fornl ing or of truth by mathematical statements, or the possession of reference
Mathematical terms. Thus in § 87 Frege quotes Heine as saying, ‘I call
CHAPTER 20 tain tangible signs numbers: the existence of these numbers is thus not in
Ceestion’. It is then impossible, Frege argues, for the formalist to explain what
he understands by an infinite sequence of numbers without abandoning his
claim to this alleged advantage. Frege derives much amusement from the
The Critique of Formalism formalists’ contortions as they attempt to extricate themselves from this
dilemma.
More sophisticated formalists than those with whom Frege had to contend
would admit that they were concerned with abstract symbol-types rather than
Section (r) of Grundgesetze, Part III.l, running from § 86 to § 137, is occupied
physical tokens, and would avoid talking about infinite sequences of terms for
with a critique of radical formalism, as represented by Frege’s colleague at
rational numbers. Instead, they would consider what theory’ would be
Jena, Johannes Thomae, and by E. Heine. Radical formalism is the doctrine
that the formulae of a mathematical theory do not express genuine statements developed by one who believed in rational numbers distinct from but denoted
that can be true or false, but are merely uninterpreted strings of figures - by such terms, and in infinite sequences of them. They would next think how7
letters in an alphabet, in the generalised sense - which the mathematician to axiomatise that theory, and then how to formalise it. At that point, they
manipulates according to prescribed rules. Radical formalism may be pro­ would throw away the meaning that had guided them to the formal theory,
pounded as a local or as a global thesis: that is, as applying to some one or and declare mathematicians to be concerned solely with the production of
more mathematical theories, independently of what holds good about others, formal proofs within that formal theory. Frege cannot be blamed for the naivety
or as one holding good, of necessity7, of all mathematical theories. This section of the formalism of his day; but his third objection could not be raised against
of Grundgesetze is the only passage in Frege’s writings in which he offers a a more sophisticated version of it. We learn from it little about his own
critique of radical formalism. conception of the system of real numbers, save that it is not merely an infinite
He offers three main objections to it. They are: totality, but contains elements that are themselves infinite in character, in the
sense that they are not in general capable of being specified by a finite
(1) that it cannot account for the application of mathematics; description.
(2) that it confuses a formal theory with its metatheory;
(3) that it can give no coherent explanation of the concept of an infinite
sequence.
Theory and metatheory

Infinite sequences Frege s second objection, interspersed with other matter relating to the specific
formulations of Heine and Thomae, is expounded at great length, and occupies
Of Frege’s three objections to formalism, the third, developed in §§ 121-36, §§ 93-119, Nothing in the argument bears particularly upon the nature ot real
is directly relevant to the theory of real numbers; it is made apropos of the numbers. It is an objection to global formalism, not to a version directed only
introduction of irrational numbers into formalist arithmetic, which, as Frege at.a Particular mathematical theory: and it is Frege’s principal ground tor
says in § 124, is effected by both Heine and Thomae in a manner superficially rejecting it.
r^se^ irJS that of Cantor, by means of infinite sequences of rationals satisfying
§ 88 he quotes Thomae as comparing arithmetic, as the formalist con
the Cauchy condition for convergence. But the fact that the stipulations have
>'es it, with the game of chess; and in § 93 he draws the contrast, w ic ,
to be understood in the light of the governing formalist conception makes their
theory crucially different from Cantor’s, Frege argues; and in § 131 he speaks chp 6 rernar^s> Thomae fails to do, between the game and its theory. i ei er
oi the incorrigible disparity between what the introduction of irrationals thevtj1110768 nor Pos*tions on the chessboard express anything, in particu ar,
demands and what formalist arithmetic can offer’, due to the fact that ‘to KaiL rOt exPress thoughts that can be evaluated as true or false. Given the
0 chess, however, nothing can stand in the way of our developing the
252
 20 The Critique of Formalism 255
254 Frege: Philosophy ofMathematics
. „ nf mathematical theories. He has not succeeded in abolishing mean-
theory of chess: this consists of meaningful propositions about chess moves
mathematics: he has merely relabelled it.
and chess positions, capable of demonstration in the same way as mathematical
theorems. In the same wav, given an uninterpreted formal theory', nothing can T be more precise, he has merely restricted the subject-matter of what he
stand in the way of our constructing its metatheory, whose subject-matter will •oared to recognise as pure mathematics to the study of formal systems'
be the formulae and formal proofs of the uninterpreted theory. The propo­ ‘ilnure mathematics, for him, is metamathematics. The interest of a derivation
sitions of the metatheory will, again, be meaningful statements, established by 3 P formal system lies in its being a demonstration that the formula derived
deductive proof. The formalist has no way of preventing this metatheory from 10 3 be derived in that system. The formalist has not, therefore, expelled
arising: if we have any reason to be interested in the formal theory, we shall Waning from mathematics: he has merely shifted the mathematical proposition
be interested in metatheoretical results concerning it. The formalist can allow from the content of the formula to the metatheoretical statement that it is
no place for the metatheory in his philosophy of mathematics, however. It is derivable. For the interest of such a statement is continuous with that of other
not a mathematical theory', on his understanding of what a mathematical theory metatheoretical statements not demonstrable by means of a derivation in the
is, for it has content: but it is not an empirical theory, either, and still less is formal system, such as that a certain formula is not derivable, or that neither
f it an application of the formal theory'. it nor its (formal) negation is derivable, or that every' formula of a certain form
The formalist cannot block the development of a mathematical theory: he is decidable. This is because the statement that a given formula is derivable
can only formalise it. Frege did not have to respond to a formalist answer to is still a meaningful a priori statement in a language. As such, nothing can
his objection about the theory’ of chess, that it, too, could be formalised, as inhibit us from applying to it the usual logical operations of negation, generalis­
could the metatheory' of a formalised mathematical theory. If he had, it would ation and the rest, or, indeed, from generalising at a yet higher level to
have been open to him to reply that formalisation of the theory' of chess would such propositions as that every' consistent formal system of a certain class is
in no way prevent anyone from asserting the meaningful propositions of the incomplete. There is no gulf within mathematics comparable to that between
intuitive theory: he wants, not to derive a formula in a formal theory, but to a position on the chessboard and the proposition that mate cannot be forced
assert, for example, that it is impossible to force mate with a king and two w ith two knights: the principle of duality' is a theorem of projective geometry,
knights. The formalist must consider such a proposition as an application of and does not differ from, say, Desargues’s theorem as something of an utterly
the formalised theory of chess - an application of a formal theory made before
different character.
that formal theory’ existed. (This links Frege’s first objection to formalism with
Global formalism, in its radical version, promised to clear up all the problems
his second, that it can give no good account of the application of mathematics.)
In the same w’ay, the proposition that (formalised) Peano arithmetic is consist­ of the philosophy of mathematics by ruling them out of order: once meaning
ent (Gentzen), or that, if it is consistent, there is an undecidable sentence had been expelled from mathematics, those problems could simply no longer
(Gbdel), is to be regarded as an application of the formalised metatheory. (We arise. Once formalism had shed its crudest formulations, it could not continue
should not take even the second of these propositions as an application of the to maintain, as Heine attempted to maintain, that mathematics is solely con­
formal system itself, since we do not have a derivation in that system of its cerned with actual marks made with ink or chalk; it had to allow its subject-
arithmetised version, but only a proof in the metatheory that it is derivable.) matter to consist of strings of symbols considered as types, and thus of objects
The formalist is not merely engaged in drawing the boundary between as abstract as the natural numbers. Reflection on Frege’s crucial distinction
pure and applied mathematics in an unusual place: he is creating a hitherto etween chess and the theory’ of chess, and between a formal theory' and its
unrecognised region. For the intuitive theory' of chess differs from applications metatheory', rapidly dispels the remaining claim of radical formalism, that
of mathematics in the physical sciences in being itself a priori. Applications of mathematicians are not concerned to make meaningful assertions, and a fortiori
mathematics in physics depend on facts established by empirical observation, not t0 make meaningful assertions a priori: with that, global formalism ecapor-
or on theoretical hypotheses: but the theory of chess depends only on such ates as a tenable interpretation of mathematics.
premisses as that the king must move out of check, which is not a theoretical
hypothesis but a rule constitutive of the game of chess, just as the axioms and
rules of Peano arithmetic are constitutive of that formal system. The formalist Application
is thus compelled to recognise a realm of meaningful a priori theories, to Fre ge’s first
Paragraph 0^.of h°V° Fadical formaIism is expounded in §§ 89-92. These
s are
which he denies the title of pure mathematics, but which contain theorems ea* numbers wh- kt imPortance for the understanding of Frege’s theory of
and proofs of theorems, and would be regarded by everyone else as clear re also of high im e ev’dently had vividly in mind while writing them. They
Portance for a topic central to his philosophy of mathematics
 20 The Critique of Formalism 257
256 Frege: Philosophy ofMathematics p]ay a lesser role, or none at all, since we shall now be concerned with
generally, since they form the only sustained discussion of the application of u objects of the theory to which the application is being made: application
mathematics in the whole of his writings. Scattered remarks on the subject, thC therefore be regarded as consisting primarily of the instantiation of highly
such as those to be found in Griitidlugen, § 9, can easily give the utterly false "eneral truths of logic. Evidently, a formalist can allow no place for application
impression that he regarded mathematics as concerned with an ideal realm of L so conceived.
pure abstract objects having far less bearing upon empirical reality dian the The formalist could object that he is not bound by Frege’s conception of
galaxies have on terrestrial affairs. -plication: he can propose an alternative conception, according to which
The formalist expressly news a mathematical theory as a type of game. pplication consists in supplying a specific interpretation for an intrinsically
Frege insists that it is not a game, but a science in the general sense of a uninterpreted formal system. In practice, such interpretations will display a
sector in the quest for truth. What makes it a science, he claims, is precisely common pattern. Although, officially, the symbols of the formal system are all
that it is capable of being applied: ‘it is applicability alone’, he declares at the of them unmeaning marks, they will unofficially be subject to a syntactic
end of § 91, ‘that raises arithmetic from a game to the rank of a science.
classification: logical constants will be recognised as such, and, usually, as
Applicability therefore belongs necessarily to it.’
bearing the specific meanings of the operators of classical logic; individual
The formalist, according to Frege, cannot explain, or even recognise, the
applicability’ of a mathematical theory’. He cannot do so because, for him, the constants, predicate letters, function symbols and so on will all be viewed
formulas of the theory’ express no thoughts: they are not meaningful statements, as belonging to their respective syntactic categories. In short, the notion of
to be judged true or false, but mere formal objects. ‘Why can one not make interpretation employed will be that used in standard model theory, rather
an application of a chess position?’, Frege asks, and answers, ‘Obviously than that appropriate for someone trying to break an intricate code or decipher
because it does not express a thought’. ‘Why can one make applications of an unknown script; if there proved to be a successful interpretation of this
arithmetical equations?’, he continues, and replies, ‘Only because they express latter sort that did not respect the apparent syntax of the formal language, it
thoughts’. The formalist chooses to regard the equations as stripped of their would not be intuitively regarded as an application of the mathematical theory,
content, and hence as of the same nature as positions on the chessboard. He but merely as an astounding coincidence. Furthermore, the pattern common
thereby ‘excludes from arithmetic that w’hich alone makes it a science’. to the various interpretations would be likely to be discernible in much more
Why does Frege think it necessary’, for a mathematical formula to be applied, than the syntactic structure of the formal language: when made explicit, it
that it express a thought? Plainly’ because he takes the application of a math­ would closely resemble the meaningful version of the theory as Frege conceived
ematical theorem to be an instance of deductive inference. It is possible to of it. It demands proof that the formalist has at his command a conception of
make an inference only from a thought (only from a true thought, that is, from application genuinely distinct from that of Frege; and the proof seems unlikely
a fact, according to Frege): it would be senseless to speak of inferring to the to be forthcoming.
truth of some conclusion from something that neither was a thought nor In any case, the formalist cannot, consistently with his position, regard the
expressed one. We do not, of course, call every inference an ‘application’ of applications of a mathematical theory as any part of the business of mathemat­
its premisses: it is in place to speak of application only when the premisses ics. For Frege, it is here that the gravest error of formalism lies; but his
are of much greater generality than the conclusion. position is a subtle one. He would agree that specific applications of arithmetic
Frege tacitly took the application of a theorem of arithmetic to consist in are not the business of arithmetic, but only of the particular science within
the instantiation, by specific concepts and relations, of a highly general truth which those applications are made. That is the reason for his rebuke to Mill,
of logic, involving quantification of second or yet higher order: if the specific ln Grundlagen, § 9, for ‘confusing the applications that one can make of an
concepts and relations were mathematical ones, we should have an application antiunetical proposition ... with the pure mathematical proposition itself. It
within mathematics; if they were empirical ones, we should have an external
s is remark, above all, that may mislead the reader into supposing Frege to
application Mathematical theories could not themselves consist solely of logical
a',e beIieved arithmetical propositions to have senses entirely unconnected
tru s imo ving only higher-order quantification, since they required reference
to mathematical objects (which Frege believed he could analyse as logical ith their applications and to have regarded those applications as whoUy
o jec s , a ove a in order to maintain the extensional distinctness of the cat^rna^t0 matbematics and of no concern to it. His discussion of the app i
concePts and relations quantified over, which might collapse on one another Mathematics in Grundgesetze, Volume II, §§ 89-92, by contrast, reveals
if the domain of the individual variables were allowed to be too sparse. When Writ; tltU^e diametrically opposite to this. Had he altered his view e tween
we are conceme with applications, however, the objects of the mathematical nieang and writing the second volume of Grundgesetze. y no
s- h a footnote attached to § 137, at the very end of the section on e
 258 Frege: Philosophy of Mathematics 20 The Critique of Formalism 259
formalists, and devoted to Helmholtz,1 another exponent of an empiricist asks at the end of § 91, ‘to exclude from arithmetic what alone makes
philosophy of mathematics, Frege accuses Helmholtz of making the very same ^regCience?’ But the applicability of arithmetic sets us a problem that we need
mistake as that of which, in Grundlagen, he had accused Mill. The mistake is 3 Stye- what makes its applications possible, and how are they to be justified?
that of‘confusing the applications of arithmetical propositions with the propo- U’S°might seek to solve this problem piecemeal, in connection with each
sitions themselves, as if the question concerning die truth of a thought and tjcUiar application in turn. Such an attempt will miss its mark, because
that concerning its applicability were not quite distinct’; Frege objects that ‘I P?at explains the applicability of arithmetic is a common pattern underlying
can very' well recognise the truth of a proposition without knowing whether I Wn ats applications. Because of its generality, the solution of the problem is
shall be able to make any application of it’. a. efore the proper task of arithmetic itself: it is this task which the formalist,
It may seem that, when Frege is criticising formalism, he treats the appli­ who regards each application as achieved by devising a new interpretation of
cations of a mathematical theory' as intrinsic to it, but, when he is criticising the uninterpreted formal system and as extrinsic to the manipulation of that
empiricism, he treats them as extrinsic; but, surely, he could not be so careless system, repudiates as no part of the duty of arithmetic. ‘But what is then really-
as to juxtapose expressions of these incompatible views. In fact, his views are achieved by this?’, Frege asks at the beginning of § 92, and answers:
not incompatible. What are extrinsic to arithmetic are all particular applications
of it: these relate to restricted domains of knowledge, and, as Frege says in Admittedly arithmetic is relieved of some work; but is the task thereby removed
Grundlagen, ‘often ... presuppose observed facts’. The mistake of Mill and of from the world? The formalist arithmetician seeks to shift it on to the shoulders
Helmholtz consists in taking such particular applications as integral to the of his colleagues the geometer, the physicist and the astronomer; but these decline
with thanks to occupy themselves with it: and so it falls between these sciences
senses of arithmetical propositions. What is intrinsic to arithmetic, by contrast, into the void. A clean demarcation between the domains of knowledge may be
is the general principle that explains its applicability and hence determines the good; but it ought not to be carried out in such a way that one domain remains
common pattern of all particular applications. The mistake of the formalists over, for which no one will undertake the responsibility.
consists in ignoring this, or, at best, reckoning it not to be the business of
arithmetic.
On the contrary, Frege concludes,
In §§87 and 88, Frege alludes to his conception of the real numbers, as
being ratios between quantities, which he has already introduced in the preced­
ing section (b) on Cantor. In § 92 he makes telling use of it in explaining his it is reasonable to demand this work of the arithmetician, in so far as he can
accomplish it without encroaching on those special domains of knowledge. For
ideas about the relation of arithmetic to its applications. ‘We know,’ he says, this he needs, above all, to attach a sense to his formulas; and this will then be
that the same ratio between quantities (the same number) can occur in of so general a kind that, with the help of geometrical axioms and of physical and
connection with lengths, with temporal durations, with masses, with moments astronomical observations and hypotheses, it can find manifold applications within
o inertia, etc. This makes it probable that the problem how we are able to these sciences.
ma e use o arithmetic is to be solved, at least in part, independently of those
sciences wi in which the application is made.’ Frege is here asserting that So far from having accorded scant respect to the applications of mathematics,
of real numbers> though various, are not simply Frege was, of all philosophers of mathematics, the one who gave the greatest
Should t°US- On the"ontra^’ they display a common pattern. Arithmetic
attention to the topic.
denend?1nnCOnCern particular Rations, even when they do not
or masVXChTingent WS’ invoIve concePts alien to it, like length
knowledge Tt n t0 ^eornetry’ physics or some other special domain of Waismann’s critique of Frege
hand' ,0 *■“ With Wittgenstein, Friedrich Waismann, writing from a Wittgensteinian stand­
underlies all particular uses tf thTreTn ” "S'” °f sene’ahty’ point, was one of the few to comment on Frege’s critique of formalism. In
It is its caoaciw .YC numbers as measures. § 91> Frege wrote:
The formalistrelrd^thPP * raises arithme^ to the rank of a science,
lhe formalist regards this as irrelevant to arithmetic itself: ‘is it well done’,
Why can no application be made of a chess position? Obviously, because it
Frege says, ‘I have scarreh^erer^nmnn^”?11 er'tenntn«theoretisch betrachtet’, about which presses no thought. If it did so, and even' chess move conforming to the rules
essay’. ’ e'er encou^red armhing so unphilosophical as this philosophical WnrrHponded to a transition from one thought to another, applications of chess
Would also be conceivable.
 250 Frege: Philosophy ofMathematics 20 The Critique of Formalism 261
In the course of a long discussion of Frege’s argument, Waismann commented alists’ view, Waismann s allows no place for this: just that is what is wrong
as follows:2 jm Frege’s standpoint. Waismann believed that he had seen much
with it froi
than Frege; but he had not seen as far.
further L-
'What then has to be added, in order for a mathematical equation to express a
thought-content? Application, and nothing more. It is mathematics when the
equation is used for the transition from one proposition to another; otherwise it Retrospect and prospect
is a game. To say that a move in chess expresses no thought is hasty; for it wholly
depends on us. After the critical sections, Frege concludes Part III.l with two brief further
sections. Section (/), entitled ‘Retrospect and prospect’, and running from
Waismann goes on to imagine positions of chessmen being used to represent & 156 to § 159, aims to draw the moral of the long examination of other
the disposition of troops in a battle, and continues: dieories of real numbers that has preceded it; section (g), entitled ‘Quantities’,
and running from § 160 to § 164, gives a preliminary' sketch of the theory' of
‘Because a chess move expresses no thought, one cannot apply it.’ Would it not real numbers to be developed formally in Part III.2. Section (/) contains few-
be more correct to say that because we have not provided an application for it,
the chess move does not express a thought? surprises, and may be reviewed at this point.
Frege begins in § 156 by recalling the gross methodological errors: formal­
Waismann was not a formalist, and did not deny that a mathematical propo­ ism; the mistake of thinking that merely defining a concept guarantees the
sition has a sense. Yet the position he here adopts is barely distinguishable existence of an object falling under it; and the belief that its existence will be
from that of the radical formalist who construes application as imposing an secured provided that the concept can be shown to be consistent. Its inconsist­
empirical interpretation upon a hitherto uninterpreted formal calculus. Wais­ ency cannot be relied on to be obvious, however, and so its consistency needs
mann denies that we first confer a sense on the proposition, and then, in the to be proved. The only known way of proving it is by finding an object that
light of that sense, make various applications of it: rather, we make the appli­ falls under the concept; ‘until a quite new' principle for proving freedom from
cations, and thereby give it a sense - a truly Wittgensteinian idea. What is so contradiction is discovered, we can advance no further along this route’.
applied must therefore be an uninterpreted formula, just as the formalist In § 157, Frege reiterates that he can, by means of value-ranges, achieve
supposes: the only difference between him and Waismann is that he expressly what other mathematicians hope to gain by creative definitions.3 Furthermore,
denies that the application has anything to do with mathematics. he has already fixed on construing real numbers as ratios of quantities, and
It is difficult to find a psychologically convincing example of what Waismann hence on quantities as being the objects between which such ratios obtain; he
has in mind; but we might imagine a child who has in school been taught remarks in a footnote on his agreement with Newton in this respect.4 There was
computations with fractions in a purely formal manner. Waismann and Frege quoted in the last chapter the passage from § 157 in w'hich Frege distinguishes
would agree that he does not yet attach any sense to the equations: they merely cardinal numbers (Anzahlen), as answering the question, ‘How many objects
figure in a calculating game. He is then for the first time shown how to of a certain kind are there?’, from real numbers, as used to say how' great a
measure lengths with a ruler, and how to construct rulers divided into tenths, given quantity is compared to a unit quantity, and concludes that the two kinds
twelfths and ninths of an inch. Certainly he now attaches a certain sense to of number form disjoint domains.
the equations: but is his experience a refutation of Frege’s view? Plainly not. In § 158 Frege warns against making essential appeal to geometry'. ‘If arith­
e temporal order of his attaching that sense to them and his mastering the metical propositions can be proved independently of geometrical axioms, then
application is not to the point: what matters is that, although he has now begun ey must be. To do otherwise would be needlessly to belie the autonomy of
to understand equations involving fractions, he does not yet fully understand anthmetic and its logical nature.’ Writers on geometry' sometimes begin by
He idT T d° S° beCauSe 1116 sense he has learned t0 attach t0 such
nn Ah™ >Ck, tHe generallt>’ squired of the full arithmetical sense: it is not
speaking of the line segment a and later use ‘fl’ to denote the number giving
e ength of the segment; this results in the confused idea that a numerical
aDolicarinn-'th Am6ubUt & geornetr'cal one. It relates to just one kind of Jin ol does or can refer to a line segment. It refers, rather, to a ratio oi
number, m k * aCquire a ^asP of Ae ^eneral use of rational antities, here of the length of the segment to that of a unit segment: a ratio
numbers to give the magnitude of quantities of different kinds. Like the
can also obtain between masses, between temporal durations, etc. The
Grund/ar^ s'i°n^S’ abstraction is to be used in defining the real numbers, as foretold in
English translation bySecond edn > Vienna’ 1936, P' 74O
> J Benac, Introduction to Mathematical Thinking, New York, 1951, P- 24°’
e Grundlagen, § 19.
 2^2 Frege: Philosophy ofMathematics
real number is thereby detached from all particular types of quantity, and, „
“'in’sTo?”'®'re”Xw" approach as lor this reason being intermediate
ln§ l^yrege e f founding the theory of irrational numbers on
between the oW df arithmetical method introduced ‘in more
mcMdme?’ Characterising this intermediate approach he says:
CHAPTER 21

From the former « f geTmerted qu^fo'

quantities orthereby come doser to the mom The Critique of Cantor
and from all particular deftct apparent in
Sateither" measurement does not appear at all; or else it is tacked on in a purely
m ilwnn without anv inner connection founded upon the essence of
'“be It follows from the fatter approach that it has to be separately stated for
Section (b) of Grundgesetze, Part III.l, occupying §§ 68-85, is devoted to a
™ of quantin- what it is to measure it and how one may thereby obtain a critique of Cantor’s method of introducing irrational numbers, which was a
number A eeneml criterion is then completely lacking for when the numbers can modification of that of Weierstrass.1 In it, Frege goes to the heart of his
be used as measurement-numbers and for the form that this application of them
will then take. dissatisfaction with existing definitions of the real numbers; we leam much
from this section of his ground for adopting his own divergent approach. The
This passage contains the only explicit formulation in Frege’s writings of a discussion is nevertheless not clean: it does not confine itself to matters of
methodological maxim implicit in his practice. It is not enough that an arith­ fundamental principle, but includes many objections to mistakes on Cantor’s
metical theory should undertake to state and justify the general principles part due solely to carelessness and easily remedied.
governing its application. It is necessary, further, that they should not be
‘tacked on’ as an appendage to the theory, as Dedekind did with his foundations Fundamental series
for number theory and Cantor did with his method of introducing the real
numbers; rather, they belong to the essence of number, and hence should be Ca?or r? expounded his theory of real numbers in 1872.2 He did so afresh
made central to the way the numbers are defined or introduced. m § of his Grundlagen of 1883.3 Frege principally examines the exposition
We may thus hope, Frege concludes, neither to relegate the treatment of in ^ntors Grundlagen, but, in §§ 75 and 85, quotes also from the article of
e application of real numbers to the various special domains of knowledge lo/Z.
nor to su y arithmetic ■with objects, concepts and relations borrowed from Qma‘n course Frege’s argument begins in the first two paragraphs,
those sciences, thus endangering its autonomy and its essential character. The 5 76- th ’ 311 • t^en breaks off for a partial digression running from § 70 to
particular applications are indeed not the concern of arithmetic; but a treatment the se tT m.a^n ar^ument then resumes at § 77 and continues until the end of
demanded oTk undertying all applications may legitimately be cleanly dis “t * $5^though the main line and the digression cannot be quite
eramml ^ose®. ®ecdon (/) with a problem about the execution of his pro­
1 'pL’ ,
can wp fi ^W> * r.e.erence t0 geometrical and physical quantities is forbidden’, importance and co”s‘dered >n this chapter in greater detail than others, in part because of its
mus?dn^qUantltieS Stand’ One t0 th6 other * the ratio V2, which we translation.’ ° because’ together with sections (e), (/) and (g), it is not available in English
suXeX^ eHitenCe°firrati0nalnumber is tobe proved? He postpones
Annalen Matzes aus der Theorie der trigonometrischen Reihen’, AfarAewnw-
suggesting any solution to this problem until § 164. rePrinted in G r ’ 1 PP- 123-32, the relevant sections being §§ 1 and 2, pp. 123-8;
F Zennelo RPri- Gesammelte Abhandlungen mathematischen und philosophischen Inhalts, ed.
3G. Cantor rZ j>32, rePrinted D80, pp. 92-102 (§§ 1-2 are on pp. 92-7).
^Print of the’Fifth" • 7* e-'ner a^Seme<nen Mannichfaltigkeitslehre, Leipzig, 1883. This was a
jfheniatische Ann >artIC , *n his series ‘Uber unendliche, lineare Punktmannichfaltigkeiten’,
Frege cites HankeFs Them-' a (f’^ndlungen do uT’ino XXI, 1883, pp. 545-91, and reprinted in turn in his Gesammelte
thoJeh hRrf he means Tories ofC,^ Zal'/e^ysteme of 1867 as an example of the former; e those of the \P y*H he cited here throughout as Grundlagen; the page numbers will
e °es not mention them by name^ ^ede^‘nd> introduced simultaneously in 1872, fwwwe//f^^Mn^PaJate edition, with those of Mathematische Annalen, followed by those of
 2 J 7'he Critique of Cantor 265
264 Frege: Philosophy ofMathematics
. -n the way Uligens understands him, namely as confusing the sign with
Cantor defines a fundamental series to be a sequence of rationals satisfying • signifies, but that there is nevertheless something correct in Illigens’s
the Cauchy condition for convergence/ Frege begins by citing this definition, what n In § 76 we learn that this consists essentially in the absence from
together with Cantor’s additional remark, I call [this] a fundamental series ob'^tl°s theory of the principal thing, the real numbers themselves, at least
and associate to it the number b to be defined .■ Frege makes play with the
he interpretation of Cantor’s words that Frege is currently assuming. Since,
question whether, in this sentence, Cantor intended by ‘number’ a numerical
°n that interpretation, the symbols such as are dispensable, we have only
symbol or the referent of such a symbol. He considers first the former interpret­
on fundamental series, and no real numbers. A fundamental series might
ation, which he understands as meaning that the symbol is to denote the series
itself; as he remarks, under this interpretation nothing essential is accomplished to determine a particular real number, if we knew what real numbers
by selecting a particular symbol to denote the series. He goes on to quote Syere’ but just this is what we have not been told.
" This conclusion leads Frege, in § 77, to doubt whether the interpretation
Cantor as distinguishing three cases:
f Cantor’s phrase ‘I associate to [the fundamental series] the number b’ which
(1) for any given positive rational q, there is a term of the series such he has hitherto been assuming can accord with Cantor’s true intention. He
that the absolute value of even’ subsequent term is less than q; therefore now conjectures instead that Cantor intended, not to select a mere
label for each fundamental series, but to associate with it a number; such a
(2) for some positive rational q, there is a term of the series such that number would not, in general, be a rational. ‘These numbers’, Frege com­
every subsequent term is greater than p; ments, ‘are therefore in part new ones, that have not as yet been considered,
and they are to be determined precisely by the fundamental series ■with which
(3) for some negative rational -q, there is a term of the series such that they are connected.’ This emended interpretation is so obviously correct that
every subsequent term is less than -Q. the reader may feel some impatience at the time wasted by scrutinising Cantor’s
theory in the light of the old one; but, from Frege’s standpoint, the new
‘In the first case’, Frege quotes Cantor as saying, ‘I say that b is equal to interpretation hardly improves matters. The burden of Frege’s complaint
nought, in the second case that b is greater than nought or positive, and in against Cantor, so interpreted, is that at no point does he proride any account
the third case that b is smaller than nought or negative.’ Frege justly complains of how the new numbers are to be defined, nor of how, when they have been
of these definitions on the ground, first, that, in each, two expressions (‘nought’
defined, their association with the fundamental series is to be specified.
and ‘equal to’, ‘greater than’ or ‘less than’) are being defined simultaneously,
In §§ 79-81 Frege quotes Cantor’s immediately following remarks:'
and, secondly, that in any event these expressions must be taken as already
known and hence not open to further definition. He fails to point out the easy Now come the elementary operations. If (ar) and (<?',,) are two fundamental series
remedy, namely to delete the words ‘greater than nought or’ and ‘smaller than by which the numbers b and b' are determined, it is demonstrable that («,. ± a\.)
nought or in the second and third, and, in the first, to substitute for ‘is equal and (av.a'v) are also fundamental series, which therefore determine three new
to nought’ a simple predicate such as ‘vanishes’. numbers that serve me as definitions of the sum and difference b ± b' and of the
In § 70 Frege cites an article by Eberhard Uligens criticising Cantor’s Product b .b'.
theory, it is this that leads to the digression. As Frege remarks, Uligens adopts
the same interpretation of the phrase ‘I associate to it the number b' as he is sightly odd that Frege should here omit Cantor’s corresponding definition
currently assuming, taking to be a symbol denoting the fundamental series, irision; for, while excluding division by the real number 0, or by a funda-
rege reports Uligens as objecting that a series of numbers cannot itself be a ?ta series that vanishes (converges as a sequence to 0) - thus otfending
quantity, and hence that the terms ‘greater’ and ‘smaller’, as applied to them, ainst Frege’s principle that a function must be everywhere defined - Cantor
cannot have a sense analogous to that in which they express relations between whichCtSj,t0 Prov*de for the case when a term of the fundamental series by
rationa numbers. Frege s comment is that Cantor does not have to be under- Jnistak 6 divisor is given is 0. Here, then, Frege had an unquestionable
°verlnni-t0,ComPiain °f> although, again, one easily rectified; he surprising)
say‘Xence'nient t0 f01'°W CaW°r’S terminol°^' and speak of a ‘series’ where we should now ~ s, e °PPortunity.
Cantor’s
5 G. Cantor, Grundlagen, p. 23 (567, 186). — s next two sentences are once more quoted by Frege in full:
6 E. Uligens, ‘Zur Weierstrass’-Cantor’sch,ien Theorie der Irrationalzahlen’, Matheinatisdie The
Annalen, vol. XXXIII, 1889, pp, 155-60. mentary operations upon a number b given by a fundamental series (ar)
’G.C,
ant°r’ Gnintllagen, pp. 23-4 (568, 186-7).
 21 The Critique of Cantor 267
2^^ Frege: Philosophy of Mathematics
and a directly given rational number a are included in the above stipulations by S 83 to play die Same he a’s0 Pracdsed on Hilbert, using nonsense w’ords
in \ e of ‘equal’, ‘nought’ and so on in order to show that Cantor has not
k Onh-^ow come ^definitions of being equal to, greater than and smaller than, in P er]y succeeded in defining anything.
as benveen two numbers b and b’ (of which b may = 4): one says, namely, that pr<The foregoing criticisms of Cantor by Frege are perfectly sound, butintoler-
b = b1 or b > b' or b < V according as b - b is equal to nought or greater or
h|y laboured. As with Dedekind, Russell made very similar criticisms, going
less than nought. ao far as to say outright that ‘there is absolutely nothing in the above definition
Frege interprets these stipulations as accomplishing no more than to supply of the real numbers to show that a is the real number defined by the fundamen-
hints concerning which numbers are to be associated with particular fundamen­ ul series whose terms are all equal to a. The only reason why this seems self-
tal series: in no case do they specify the association uniquely, nor, more evident is that the definition by limits is unconsciously present, making us
importantly, do they supply us with any new numbers to associate with those think that, since a is plainly the limit of a series whose terms are all equal to
fundamental series that do not converge to rational limits. a, therefore a must be the real number defined by such a series.’8 Russell’s
Frege subsequently half-admits that, in making the first of these two objec­ point is that, contrary to Cantor’s intentions as interpreted by Frege, the
tions, he has gone too far. He is interpreting Cantor, apparently correctly, as rationals and the real numbers form disjoint domains: ‘a fundamental series
adding irrational numbers to the already given system of rational numbers: of rationals defines a real number, which is never identical with any rational.’
some of the numbers to be associated with fundamental series will be already We cannot but deprecate the disagreeable tone of Frege’s criticisms of
known ones, namely rationals, and others will be new ones, the newly intro­ Cantor, although, in view of Cantor’s mean-spirited review of his Grundlagen
duced irrationals. In § 78 Frege allows that the stipulation that the number der Arithmetic we may forgive it. In view of the fact that, as we shall see, he
associated with a fundamental series whose limit is 0 shall be ‘equal to nought’ did, in part of his section (fi) on Cantor, go to the heart of his disagreement
must mean that the (rational) number 0 shall be associated with every such with him, w'e may also forgive the heavy-handed style of the peripheral criti­
fundamental series, at least if we understand Cantor as meaning ‘identical’ by cisms (which no doubt did not seem peripheral to Frege) just reviewed. It is
‘equal’. In § 81 Frege recognises it as Cantor’s intention that the same number obvious to us how Cantor might have emended his theory so as to escape
shall be associated with two fundamental series (a„) and (a'v) if the fundamental Frege’s objections. Having distinguished his three classes of fundamental
series (av - a'v) has the limit 0. Since it is also plainly Cantor’s intention that series, as vanishing, positive and negative, he should, without yet speaking of
the rational number a shall be associated with the fundamental series (a'v) numbers correlated to fundamental series, have defined the difference oper­
every term a'v of which is a, the association of rational numbers with fundamen­ ation upon them, proving that (bv - b' v) is a fundamental series if (by) and
tal series has been fully provided for. (b'v) are. This would have enabled him to define (bv) to be equivalent to (b’v)
Frege insists, however, on his second objection. Cantor’s stipulations just in case (bv — b'v) vanishes. The next step wwld then have been to show’
achieve, at best, only an association of rational numbers to certain fundamental this relation of equivalence to be transitive and hence a genuine equivalence
series: it does not yield any definition of irrational numbers. Furthermore, the relation. The way would then have been open to define the real numbers by
concession depends upon assuming that, as Cantor uses it, ‘equal’ means logical abstraction, that is, as equivalence classes of fundamental series; the
.j.dentical • and Frege professes to be doubtful whether it can mean that, for, elementary operations’ on real numbers could then have been defined in
i it did, it w ould not be open to be defined. But, if it does not mean ‘identical, terms of the corresponding operations on fundamental series, after showing
then the stipulations do not even determine that the number 1 is to be equivalence to be a congruence relation with respect to the latter operations.
associated with the fundamental series (av) for which a = v/{y + 1) for every aced with a theory presented as Cantor presented his, we automatically
v\ or so Frege claims in § 84. v ^ans orm it in such a manner as that just suggested. This is because rigour
In general Frege says in § 83, ‘the words “equal”, “greater”, etc., are left nanresentadon ts second nature to modern mathematicians. It wras not second
in a perpetual oscillation between being known and being unknown’; thus, he 0 re. Cantor, as his simultaneous introduction of the three elementaiy
“Xg»S °ne moment 1116 words “equal”, “greater”, “smaller , abum?01?’ °n Pundamental series and on the numbers correlated with them,
and tb an Pr° U>Ct aPPear as known, immediately thereafter as unknown antly demonstrates; nor was it second nature to most of the mathema-
as,lno™’ W11*” Cantor defines these words as applyW »
XXT,'" h' “,purporti"6 •» introduce, it is because we cannot help
mb' that ™ ^onBly such a symbol « , reprinted in G. Cantor, Gaammdte Abhand-
Cantor s b to have some specific content, Frege argues in § 82. He proceed and appreciZr r his own Grundlagen, §§ 85-6, Frege had written in terms of high respect
PPreciat>On of Cantor’s Grundlagv of 1883.
 2 J The Critique of Cantor 269
268 Frege: Philosophy ofMathematics
«rf.Sent definite quantities, but need not do so,10 and has little
ticians of the time. Frege had won through to what we think of as a modern
sig05 remark.
standard of rigour (if we trouble to think of it at all) by hard reflection on
troulible ridicu & s it rcaiiy stand with the assertion that numerical symbols
legitimate methods of defining mathematical notions, we can therefore hardly ‘How, then’ titics?’, Frege asks in § 73. His answer brings him closer to the
blame him for his increasingly ponderous insistence on it, even though we designate AlUanwn views concerning real numbers than anything else in the
have little to learn from it. What we can complain about is his failure to heart oc.f his o uj * Qfiindgesetze, and is worth quoting in full.
indicate how Cantor’s theory’ could have been emended, which to him, the
crit- sections
critical
great exponent of definition by logical abstraction, must have been as obvious
as it is to us. i et us look at the applications of arithmetical laws in geometry, astronomy and
hvsics. Here numbers in fact occur in connection with quantities (Grossen)" such
Fortunately, criticisms of this nature do not exhaust Frege’s observations
as lengths, masses, intensities of illumination and electrical charges; and, upon
on Cantor’s theory of real numbers; in part of section (/>) he explained the deep superficial consideration, one might think that the same numerical symbol referred
grounds for his rejection of it. Together with the remarks about application in now to a length, now to a mass, now to an intensity of illumination. This would
the subsequent section on the formalists, this constitutes the essential prelimi­ then appear to support Pringsheim’s assertion that between the numerical symbols
nary to Frege’s own theory' of real numbers. and the quantities there exists a certain connection, but only a loose one. Let us
j examine this more carefully. What is it that we really apply, when we make use
of an arithmetical statement? The sound of the words? Groups of special figures,
■' Frege’s digression consisting of printer’s ink? Or do we apply a thought-content that we connect
with those words or with those symbols? What do we prove, when we prove an
In the digression running from § 70 to § 76, the trend of Frege’s remarks is arithmetical statement? That sound? Those figures? Or that thought-content’
not explicitly distinguished from his general contention, that Cantor’s construc­ Why, of course, this last. Ven’ well, then: we must have a particular thought as
tion of the real numbers is logically faulty'. In the course of the digression, he the content of the statement, and this we should not have, if the numerical symbols
nevertheless offers reasons why he would not have found Cantor’s theory and number-words occurring in it referred, now to this, now to that.
acceptable even if it had been emended in the way suggested; it is for this If we look more carefully, we notice that a numerical symbol cannot by itself
reason much the most interesting part of his critique of Cantor. on its own denote a length, a force or the like, but only in combination with an
expression designating a measure, a unit, such as a metre, a gram, etc. What,
Frege’s starting-point in § 70 is the remark he cites from the article on
then, does the numerical symbol on its own refer to? Obviously a ratio of quantit­
Cantor by Illigens that the Cantorian symbols such as ‘b\ which he takes to ies.12 This fact lies so close to hand that it is not surprising that it has long been
denote fundamental series, cannot denote quantities (Quantitaten), as the recognised. If, now', we understand by ‘number’ the referent of a numerical
rational numbers do; he quotes Illigens as saying, ‘The symbols for series of symbol, a real number is the same as a ratio of quantities. Now what have we
i
numbers lack the capacity to become concepts of quantity', in spite of the labels gained by defining real number as ‘ratio of quantities’? At first it seems merely that
attached to them by the various definitions’, and, in § 71, as concluding that one expression has been replaced by another. And yet a step forward has been
Cantor’s theory does not enable us to say what a line V2 metres long may taken. For, first, no one will confuse a ratio of quantities with a written or printed
be. As Frege observes, if this, as stated, were a valid objection, it would convict symbol; and so one source of countless misunderstandings and errors is blocked.
ccondly, the expression ‘ratio of quantities’ or ‘ratio of one quantity to another
Cantor of a confusion between sign and thing signified, or, as we should say, quantity serves to indicate the manner in which real numbers are connected with
between use and mention; but he defends Cantor against the charge, remark­ quantities. Admittedly, the principal w’ork remains to be done. We have as yet no
ing, however, that ‘there is nevertheless something true in this objection’. ore than words w'hich indicate to us only approximately the direction in which
Frege s difficulty, as he clearly perceives, is that it is Illigens himself who Dre S° y^On*s t0 be sought. The reference of these words has yet to be more
constantly confuses sign and thing signified: he has, therefore, to tread carefully J^isety fixed. But we shall now no longer say that a number or numerical symbol
in offering him partial support. sav U0W a now a mass, now an intensity of illumination. \\ e sh
Illigens s observations prompt Frege to comment on the use of numbers as to anntk^’ ^at a can have to another length the same ratio as a mass has
' mer m*ss, or as an intensity of illumination has to an intensity of illunu-
measures of quantity. He first rebukes Illigens for speaking of rational numbers
as symbols for quantities. ‘According to linguistic usage one calls lengths, 3> P- 55^nnBshe,m> article in the Encyklopddie der mathematischen Wissenschaften, vol. I, part A, no.
surface areas, angles, temporal intervals, masses and forces “quantities”. Is if II p
en correct to say that either the number 2/3 or the numerical symbol “2/3 own account the word ‘Grosse', and employs
denotes a certain length, or a certain angle, or, indeed, both?’, he asks in § 71. two words ‘r the,Writings of others; but he draws no distinction between the meanings
The German j 35 emP'oyed bv him, will here always be translated ‘quantity .
In § 72 he quotes A. Pringsheim as explaining that the rational numbers are inOg?ate w>th the J°Ld/Or ‘ratio’ is VerMHnis’, meaning a proportional relanon. Since it
t'le Phrase for ‘irraf ^a]3 ra^ona3 number, there is not even the appearance o
 21 The Critique of Cantor 271
270 Frege: Philosophy ofMathematics to quote and cr’t*c*se Pr’nSsheim s essay, seeing that this exclusively
nation; and this same ratio is the same number and can be denoted by the same troub CSS the rational*. The reason is that Frege sees the problem, not as that
numerical symbol. ^Hning how irrational numbers can be used to give the magnitude of a
ofe •<! on the assumption that we already know how the ratiorials can be
Frege is still interpreting Cantor as using ‘/»’ to denote a fundamental series,
r this purpose, but as that of explaining how real numbers generally,
and so he concludes the foregoing passage as follows: used. °r rational or irrational, can be so used. In order to assess Illigens’s
If Illigens understands ratios of quantities (Grone//), or, what we can now regard ■ • • m that Cantor’s real numbers cannot be used, as the rationals are used,
as synonymous, real numbers, by the word ‘quantities’ (Quantitiiten) as he uses it, Crltl\e the measure of a quantity, we must first ask in what this use of the
and if he means that the symbols for series of numbers do not, on Cantor’s theory, t0 onals consists. That will tell us what, essentially, the rational numbers are.
denote ratios of quantities, he is right. In Cantor’s definition there occur only the Sen we know this, we shall thereby know what, essentially, the real numbers
fundamental series and the number b, and the latter is the symbol for the series
of numbers. Nothing is here said about a ratio of quantities. The symbol for the are since the primary application of the real numbers is the same as that of
series of numbers simply denotes the fundamental series and accordingly ought the rationals.
not also to denote a ratio of quantities, for it would then be ambiguous. It was, for Frege, the same for the rationals as for the natural numbers: the
per way to define them was one that incorporated the principle underlying
Here we have Frege’s preliminary exposition of the foundation of his theory
the salient application that we make of them. The salient type of application
of how the real numbers should be defined. A word is in place concerning his
use of the word ‘quantity’. It is a little difficult to state the precise meaning of the rationals occurs, on his view, when we say that something is 5/8 inch
of this word in natural language; but Frege so uses it that a phrase like ‘2.6 long or weighs 3/4 kilogram. Russell defined the rationals as ratios of integers,
metres’ designates a specific quantity of one kind, ‘5.3 seconds’ a quantity of and was therefore forced to treat the real number 5/8 as an object distinct
another kind, and so on. He thus takes quantities to be objects, distinct from from the rational 5/8. Frege wrould, however, have regarded Russell’s definition
numbers of any kind. There cannot be two equal quantities, on this use: if as based on too restrictive a new of the application of the rationals:14 he saw
two bodies are equal in mass, they have the same mass. Quantities fall into them as being used, in general, to answer questions that could be meaningfully
many distinct types: masses form one type, lengths another, temperatures a answered by citing an irrational number. Hence he thought it illegitimate first
third. to define the rationals before introducing irrational numbers: we must ask at
Frege does not use any w’ord corresponding to ‘magnitude’, in the sense in the outset how, in general, a real number can serve to give the measure of a
which one mayr ask after the magnitude of a quantity. It is best compared with quantity.
nouns like cardinality’ and ‘whereabouts’: when one gives the whereabouts of To answer this question for any specific type of quantity - say temporal
an object, one names a place, not a whereabouts; there are no such objects as durations - wre must suppose given an order relation and an operation of
whereabouts. The cardinality of a set is given by naming a cardinal number; addition upon those quantities: it must be known what it is lor one duration
since one can say what it is for a set to have three, or denumerably many, to be greater than another, and for one to be the sum of two others (namely
members without comparing it with any other set, we may say that its cardinality ^hen a temporal interval whose length is the first is divisible into two subinter-
is an intrinsic property. The magnitude of a quantity can be given only as a 'als whose lengths are the other two respectively). A crude answer might then
ratio between that quantity and some other taken as unit: it is therefore an e that there is an order-preserving map of the quantities of that type on to
extrinsic property. e (non-negative) real numbers carrying addition of quantities into addition
It is remarkable that Frege nowhere calls explicit attention in Grundgesetze
nun?mbers‘ ^inly, this does not yet provide a basis for a definition of real
to the fact that, unlike Cantor and Dedekind, he is proposing to define the
mult’ S'nCe SUcb maPPings are ^ar from unique: given any such mapping,
real numbers without taking the rationals as already known.13 In §§ 70-6 he
not?1Cation bY any positive factor will yield another. It is only when we map,
is assessing igens s objection that Cantor’s real numbers are not quantities,
the e,quantities themselves, but their ratios, that we obtain a unique mapping:
as e rationa s are. specifically, that there is no way to explain their use to
give the magnitude of a physical quantity. We might therefore wonder why he a unit o Umbers rePresent, not the quantities, but their ratios, unless we select
As pUantity and express every other quantity as its ratio to the unit.
” He mentions it explicitly in his letter to Russell of 21 Mav 1903 savine ‘As it seems to me, §75 tnfi^ °bserves> the point is well known; and yet he has occasion, in
S a:f 'iX’Toi(1) T "r nw^i i, ,Th n ^auh with Cantor for overlooking it. In his article of 1872, an o
vo], [j c arithmetical ratios are naturally not meant’, Frege says in footnote 1 to G
> a ter asking in the main text for a definition of ‘ratio in gene
2^2 Frege.' Philosophy ofMathematics 21 The Critique of Cantor 2T3

calls both rational numbers and the real numbers he constructs

. ‘nunKrical cnerality of arithmetic by allusion to any specific type of empirical appli-
magnitudes’ (Zaldengriissen), an expression he does not use >n his Grun(ilageil
cation.
In § 2 of the article he explains how his real numbers can be used as measures
of distance, and gives the definition. Can Cantor explain the applications of real numbers?
We express this by sating: The distance from the point o of the point to be determined These general principles do not in themselves embody any objection to Can­
is equal to b, where b is the [fundamental] scries ... of corresponding numerical ’s theory, or corroborate that of Illigens: we must ask whether the use of
magnitudes. % numbers to assign the magnitude of a quantity can be explained on the
basis of the theory. In §§ 74-6 Frege considers Cantor’s brief retort to Illig-
Frege comments: ns’s article,16 which he finds unclear. He quotes Cantor as there saying, ‘It
was never asserted by me or by anyone else that the signs b, b', b", . were
In the first place, the mistake is here to be noted, that the unit is nowhere
mentioned in the defined expression, although it is necessary for the specification. concrete magnitudes (Grossed) in the proper sense of the word. As abstract objects
From this there may arise the delusive appearance that b, b', b",... are distances, of thought (abstracte Gedankendinge) they are magnitudes only in the improper
whereas it can only be a matter of ratios; and such ratios can occur just as well or transferred sense of the word.’
with strengths of electric current, with amounts of mechanical work, etc. Frege surmises that by ‘abstract objects of thought’ Cantor means what he
himself means by ‘logical objects’, and remarks that, if so, ‘there seems to be
No doubt, if Cantor ever read these comments, he was outraged by the a good agreement between us on the subject’. Unfortunately, he adds, Cantor
tendentiousness of the criticism, since, although the unit of measurement is fails to define any such logical objects. More important in the present context
left unmentioned in the definition, it is expressly adverted to in the preceding is Cantor’s distinction between concrete magnitudes and magnitudes in the
passage in which Cantor set the stage for it. However this may be, Frege was abstract or transferred sense. Presumably, by ‘concrete magnitudes’ Cantor
right that the point is crucial for attaining a correct characterisation of the real means particular quantities - areas, masses and the like; but the notion of an
numbers on the lines he proposed. A correct definition of the natural numbers abstract magnitude is left woolly. Real numbers are not, for Frege, obtained
must, on his view, show how such a number can be used to say how many by considering what Cantor calls ‘concrete magnitudes’ and abstracting from
matches there are in a box or books on a shelf. Yet number theory has nothing their specific type: they are obtained as ratios of such concrete magnitudes, in
to do with matches or with books: its business in this regard is only to display which concrete magnitudes of whatever type may stand to others of the same
what, in general, is involved in stating the cardinality of the objects, of whatever type.
sort, that fall under some concept, and how natural numbers can be used for
In his reply to Illigens, Cantor claims that wre are in a position to arrive at
the purpose. In the same way, analysis has nothing to do with electric charge
an exact quantitative determination of concrete magnitudes properly so called,
or mechanical work, with length or temporal duration; but it must display
such as geometrical distances, with the help of the abstract magnitudes b, b',
e general principle underlying the use of real numbers to characterise the
> •..; this, he says, must be regarded as decisive. Thus, Frege comments,
magnitude of quantities of these and other kinds. A real number does not
the application to geometry’, far from being a mere agreeable extra, is decisive,
directly represent the magnitude of a quantity, but only the ratio of one quantity
ut, if it is decisive, this tells against Cantor’s theory’, because this decisive
to ano er of the same type; and this is in common to all the various types. It
eature does not occur at all in his definition of numerical magnitude.’ The.
is ecause one mass can bear to another the very same ratio that one length
bears to another that the principle governing the use of real numbers to state construction of the real numbers comes first in his theory’, without reference
° k USe assi^n magnitudes to quantities. ‘It is only after the b, b , b ,
e magmtu e o a quantity, relatively to a unit, can be displayed without the
t0 16 j° Part’cu^ar tyPe of quantity. It is what is in common to all aye been introduced that the determination of distances by means of
umerical magnitudes is given’, Frege say’s in § 75; ‘that manner of introducing
nf Uses’ an 0 J ^lat> which must be incorporated into the characterisation
saidnUrner^Ca^ magnitudes is purely arithmetical, but does not contain what is
aS mathematical objects: that is. how statements about
n to. be decisive; the instructions for determining distances by means of
e a otte a sense which explains their applications, without violating
magnitudes contain what is decisive, but are not purely arithmetical.
nence the goal that Cantor has set himself is missed. In the definition
v, la/z, p. i2?'
• vi. vte 127, sc‘”s de; Abhandlungen,
Gesammelte d" p. 96. P. 476J mit BeZUg auf den Aufsatz ■ • • \MathematischeAnnalen, vol. XXXIII.
274 Frege: Philosophy ofMathematics 2J The Critique of Cantor 275
we have the fundamental series on the one hand, and die signs b, b .
on the other, and nothing besides . ,
This criticism-is not captious: it contains the core ol Frege s objection to
Cantor’s theory. It resembles that which he may be presumed to have felt to
Dedekind’s foundations for number theory in Has sind mid was sollen die
Zahlen?: die principle underlying the application, ot natural numbers or of real > " ' express this by saying: Th, d!sun„
numbers, should not be derived as a corollary , but should be incorporated into « ,V.Ib, where b ,s the sene, (I, corresponding numerical maptta™
the manner in which they are introduced. But, in Cantor s case, there is a I

further feature. The general principle governing applications should be formu­ Frege s first objection has already been mentioned, namely that the definition
lated and vindicated; to explain and justify its operation only in a specific type contains no reference to the unit distance; but he acknowledges that this defect
of case, such as distance, is to import somediing alien into arithmetic. could be easily rectified, and then asks:
What would the proper procedure have been, according to Frege? ‘The
matter would stand differently’, he continues in § 75, ‘if we had a purely But which expression is really being defined? It must be assumed as known what
arithmetical or logical definition of ratio, from which it could be concluded the distance of a point from another point is; the so-called numerical magnitudes
that there are ratios, and, among them, irrational ones. Then what is decisive (b) have already been introduced; and the word ‘equal’ must also be already
would be comprised in this definition, and the determination of a distance by
iI known. Thus everything in the defined expression is known, and, if all were in
order, the sense of the sentence The distance from the point o of the point to
means of a unit and a ratio (a real number) would have only the status of an be determined is equal to b’ would likewise have to be known, so that a definition
illustrative example, which could be dispensed with.’ would be at least superfluous, and thereby erroneous.
The position, then, is this. Cantor thinks that it tells decisively in favour of
his theory that the real numbers he constructs can be used to specify distances; On the face of it, the argument appears unsound. From Cantor’s standpoint,
Frege thinks that the claim that this is decisive undermines the theory, since he has his real numbers, introduced as determined by fundamental series. He
no provision for that use has been expressly made in the construction. Frege’s is assuming that we know what it is for the distance of a point from the origin
argumentation so far does not, however, refute Cantor’s claim that his real to have a rational ratio to the given unit distance. That does not entitle Frege
numbers can be used to give the distance between two points, or sustain to say that we know' in general what the distance of any one point is from any
Illigens’s claim that they cannot. To decide this, the details of Cantor’s justifi­ I
other: what has to be determined is what it is to assign an irrational number
cation of his claim must be examined. to the ratio of the distance of a point from the origin to the unit distance. This
Illigens had based his criticism primarily upon Cantor’s Grundlagen, although is therefore something still apt for definition, which is carried out by reference
he also mentions the article of 1872. In his reply, Cantor appeals to the original to a fundamental series which determines that irrational number.
article, since, in § 2 of that, he had provided an argument to show that his Despite appearances, however, Frege is correct: the definition, as stated, is
real numbers could be used to give the distance between two points. Frege’s
I
I circular. Given Cantor’s assumptions, there is indeed, for each term of a
account of this is quite accurate. ‘It is assumed as known how a distance is fundamental series, a point whose distance from the origin has that ratio to
determined by a rational number’, he says. Cantor considers the distance of
the points on a given straight line from some point o on the line chosen as
i the unit distance; we might call this sequence of points the ‘corresponding
series’. The condition we are required to consider is that the points of the
origin, wrhere the points on one side from o are being regarded as having a J
corresponding series approach ‘infinitely near’ the point in question. On the
positive distance from o and those on the other a negative distance. ‘If this i ordinary understanding of ‘approach infinitely near’, the phrase refers to the
distance has a rational ratio to the unit of measurement’, Cantor says, ‘it is i distances of the successive terms of the corresponding series from the given
expressed by a rational number’, thus making the assumption stated by Frege. point. By assumption, the distance of the given point from the origin does not
Hence, as Frege remarks, each term of a fundamental series corresponds to have a rational ratio to the unit distance; hence neither does the distance of
a definite distance and hence to a definite point on the line. ‘As the fundamen­ I the given point from any term of the corresponding series, since the terms of
I
tal series proceeds, these points approach without limit a certain point, which the fundamental series itself are all rational numbers.
is thereby uniquely determined’, Frege says. Cantor’s own formulation is that, Like his failure to make e:explicit mention, in the definition, of the unit
when a point w'hose distance from the origin does not have a rational ratio to distance, this piece of carelessness on Cantor’s part could easily be remedied.
the unit distance, and ‘when the point is known by a construction, it is always The phrase ‘approach infinitely near to the given point’ could be replaced by
276 Frege: Philosophy ofMathematics
a formulation in terms of intervals with rational end-points: for any positive
rational number r, there is such an interval die ratio ot whose lengtli to the
unit distance is r, containing the given point and all terms of die corresponding
series from some term onwards. Frege, for all die time he spends on Cantor’s
theory, does not have the patience to locate the inaccuracies precisely, or to CHAPTER 22
consider how they could be corrected.
The fact is that Cantor’s whole procedure here is offensive to Frege, in that
ratios of distances are defined piecemeal; more exactly, only the second half
of the definition is given, the first being assumed as already known. Frege’s
blanket condemnation of piecemeal definition is sometimes pedantically
Frege's Theory of Real Numbers
applied; but, in the present instance, it does not rest on pedantry. Cantor’s
problem, how to explain the use of irrational numbers as measures, given the
use of rational numbers for this purpose, is not for Frege the fundamental The concept of quantity'
one. By taking it as known what is meant by a rational ratio of one distance
to another, Cantor has assumed the basic notion requiring analysis: once we By the end of section (/) of Part III.l of Gnindgesetze it has been fully
have analysed the notion of a ratio of distances, we should have no difficulty established that Frege is proposing to define the real numbers, positive and
in explaining how a real number, whether rational or irrational, can be assigned negative, as ratios of quantities. The last section (g), comprising §§ 160-4,
to such a ratio. As Frege says in § 76, ‘These [fundamental] series can serve sketches in outline how he intends to explain the notion of a ratio of quantities.
to determine ratios, but only after we have learned what a ratio of quantities The first question is naturally what a quantity is. This, he claims, has never
is: and that is precisely what we are lacking’. yet been satisfactorily stated. ‘When we scrutinise the attempted definitions,
Cantor has not only assumed the principal notion to be explained, but has we frequently come upon the phrase “of the same type” or the like. In these
assumed it without sufficient generality. What is required is an explanation, definitions, it is required of quantities that those of the same type should be
not of ±e specific notion of a ratio of distances, but of the general notion of able to be compared, added and subtracted, and even that a quantity be
a ratio of quantities of some one type: real numbers can then be presented as decomposable into parts of the same type.’1 To this Frege objects that the
precisely such a ratio, without importing anything into the definition from phrase ‘of the same type’ says nothing at all: ‘for things can be of the same
outside arithmetic. Because Cantor’s construction of the real numbers does type in one respect, which are of different types in another. Hence the question
not present them as ratios of quantities, he can do no more than illustrate yvhether an object is of the same type as another cannot be answered ‘Tes’
their use to give the magnitude of a quantity case by case; and this has the or “No”: the first demand of logic, that of a sharp boundary, is unsatisfied.
consequence that he has to appeal to non-arithmetical notions (in his example, ‘Others’, Frege continues, ‘define “quantity” by means of the words
to geometrical ones). At the end of § 76, Frege concludes his digression thus: “greater” and “smaller”, or “increase” and “diminish”; but nothing is thereby
achieved, for it remains unexplained in what the relation of being greater, or
We have first to know the ratios of quantities, the real numbers; we can then the activity of increasing, consists.’ The same goes for words like addition,
discover how we can determine the ratios by means of fundamental series. It is ‘sum’, ‘reduplicate’ and ‘synthesis’;2 ‘when one has explained words in a
odd to ascribe to the correlation of the symbols b, b', b",... any creative power.
Bringing geometry’in is therefore decisive, since by doing so one gains hold of particular context, one ought not to fancy that one has associated a sense with
that content which takes all the strain. But then what is decisive belongs to them in other contexts. One here simply goes round in a circle, as it seems’
geometry, and Cantor s theory is by no means purely arithmetical by always defining one w'ord by means ot another which is equally in nee o
definition, without thereby coming any closer to the heart of the matter.
The mistake underlying all these attempts consists, Frege says in § , in
posing the question w-ronglv. The essential concept is not that of a quantity
but of a type of quantity, or, as he prefers to say, a quantitative domain

1 Frege here refers to Otto Stolz as an example.

2 The last of these Frege quotes from Hankel.
277
278 Frege: Philosophy ofMathematics 22 Frege i Theory of Real Numbers

(Grbssengebiet):3 distances form one such domain, volumes form another, and ctaracterisalion of a quantitative domain. AU the persuasive skill he showed
so on. ‘Instead of asking, “What properties must an object have in order to in CnaMw* and elsewhere ,n convincing readers that he had ghTthe
I
be a quantity?” ’, Frege says, ‘we must ask, “What must be the characteristics correct analysts of mluttiee concepts here deserts him. He was of comse
of a concept for its extension to be a quantitative domain? ; something is a entirely nght m ins,st,ng that the concept to be explained is that of a quantitat-
quantity, not in itself, but in virtue of belonging, with other objects, to a class ive domain, not that of an individual quantity; but those at whom he jeered in
constituting a quantitative domain. § 160 were quite right to seize on the addition and comparability of quantities
of a given type as central features, whether those quantities are absolute or
distinguished as positive and negative. It is essential to a quantitative domain
Quantitative domains
i of any kind that there should be an operation of adding its elements; that this
§ 162 opens with the abrupt declaration that, to simplify the construction, ‘we i is more fundamental than that they should be linearly ordered by magnitude
shall leave absolute quantities out of account, and concentrate exclusively on is apparent from the existence of cyclic domains like that of angles. The point
those quantitative domains in which tliere is an inverse’, that is, which contain was put very' forcefully in Frege’s Habilitationsschrift of 1874. He first remarks
positive and negative quantities. Temporal distances provide a natural example that ‘one will not give a beginner a correct idea of an angle by placing a
of the latter, in that they have a direction; temperatures provide a good drawing of one before him ... One shows [him] how angles are added, and
instance of the former, since, while they have a natural zero, there can be no then he knows what they are.’7 He subsequently generalises the point, saying
temperature lower than absolute zero. Given a domain of absolute quantities, that ‘there is so intimate a connection between the concepts of addition and
we can indeed always associate with it a domain of signed ones, ‘by considering of quantity that one cannot begin to grasp the latter without the former’.8
e.g. one gramme as 4- one gramme, i.e. as the relation of a mass m to a mass We know, then, that there must be defined on any quantitative domain, in
m' when m exceeds in' by one gramme’, as Russell and Whitehead put it; and, the general sense that includes absolute and cyclic ones, an operation playing
as they continue, given a zero, wre can get back to the absolute domain, since the role of addition, and, on most such domains, a linear ordering playing the
‘what is commonly called simply one gramme will ... be the mass which has role of an ordering by magnitude; but we do not yet know which operation
the relation + one gramme to the zero of mass’.4 Frege, however, does not and which relation these will be, nor which objects can be elements of a
even trouble to offer this much of an explanation. The restriction impairs his quantitative domain. Frege, however, proceeds immediately to offer answers
claim to give a comprehensive analysis of the concept of quantity, as also does to the first and third of these questions; the second, concerning the ordering
his neglect of cyclic domains such as the domain of angles;5 the magnitudes relation, receives a corresponding answer in Part III.2. Because he has decided
of all these, relative to a unit, are after all also given by real numbers. to confine himself to quantitative domains containing negative quantities, he
Frege immediately quotes an extensive passage from Gauss.6 This discusses follows Gauss in requiring such a domain to consist of permutations of some
the conditions under which positive and negative integers mayr be assigned to underlying set and in taking the addition operation to be composition, under
I
elements of some totality. Gauss says that the integers must be assigned, not which the domain is closed; since it will also be closed under inverse, it will
to objects, but to relations on an underlying set of objects with a discrete linear !
be a group of permutations, and, when the ordering is suitably defined, an
ordering, unbounded in both directions. The relations are those any one of ordered group. (Frege nowhere uses the term ‘group’ in Grundgesetze, although
which any object in the set has to another separated from it in a specific he must have been familiar with it.)9 . ,
direction in the ordering by a specific number of intervening objects; thus This faUs very far below Frege’s usual standards of conceptual analysis. It
these relations are closed under composition and inverse, and include the could be argued that ‘quantitative domain’ should be understoo as a pure y
identity relation as a zero, and form, in fact, a group of permutations. structural term, on the ground that any group that has the right group structure,
Frege seizes upon these suggestions as supplying the main features of his as subsequendy analysed by Frege, will admit application of the notion of rauo
as a relation between its elements, and an assignment of real numbe« 0
The term quantitative domain’ appears very early in Frege’s writings, with essentially the ratios, whatever those elements may be, and whatever the group operate> *
same meaning, namely in his Habilitationsschrift of 1874, Rechnungsmethoden, die rich auf eine
bnveiterung des Grissenbegriffes griinden. But this is not Frege’s position: he requires the elements to be permutation
i
4 A.N. Whitehead and B. Russell, Prindpia Maihematica, vol. Ill, 1913, part VI, ‘Quantity’,
i 7 Rechnungsmethoden, p. 1. (See note 3.)
5 Dealt with by Russell and Whitehead in section D of their part VI.
8 Ibid., p. 2. Of Heinrich Weber’s Lehrbuch der Algebra, which
6 C.F. Gauss, review of his own ‘Theoria residuorum biquadraticorum: Commentatio secunda’, ’ For example, from the second volume
Werke, vol. II, Gottingen, 1863, pp. 175-6. I
appeared in 1896.

i
 22 Frege's Theory of Real Numbers 281
280 Frege: Philosophy ofMathematics
interval with precise end-points, rather than with fuzzy ones? It would be
and the group operation to be composition, although he leaves the underlying absurd to say that we impose the system of natural numbers upon reaS bu
set uncharacterised; but he oflers no good argument tor the requirement.
it is not at all absurd to say the same about the mathematical continuum’ We
Group-theoretically, there is no loss of generality, since every group is isomor­ i
are not given physical reality as a set of instantaneous states arrayed in a dense
phic to a group of permutations; but since these are, in general, permutations
complete ordering: we apprehend it only over temporal intervals The idea of
on the elements of the original group, this is not explanatory. The question is
precisely on what underlying set the permutations Frege identifies as elements discontinuous change is not, of itself, conceptually abhorrent; we commonly
of a quantitative domain operate. In new of the generality required, this cannot think of ourselves as experiencing it, as when darkness succeeds illumination
be specified in the formal definition; but we need to have an idea what that i when the light is switched off. More exact examination shows that such
set will be, in representative cases, before we can accept or even understand I changes, at the macroscopic level, are in fact continuous; but that does not
Frege’s analysis of the notion of quantity. When the domain consists of spatial make the idea of such simple discontinuities absurd. We could, for instance,
understand the idea that the colour of a surface might abruptly’change from,’
or temporal distances, there is no problem: the underlying set is naturally
taken to comprise points or instants. What, however, when the domain consists say, red to green. What is conceptually absurd is to apply to such a change
of masses? The suggestion of Whitehead and Russell, as it stands, represents the distinction that can be made with Dedekind cuts, asking what colour the
signed masses as permutations on absolute masses; if we follow it, we need to surface was at the instant of change: there are not two distinct possibilities,
know what a domain of absolute quantities is before we can know what a according as it was then red or then green. Yet more absurd would be the
I idea of the surface’s being red through an interval, save at one particular
domain with positive and negative quantities is. It might be proposed that the I
underlying set should be taken to consist of the physical objects to which moment, when it was green. These are not physical absurdities, violating well
absolute masses are assigned. We could not then assume, however, that the known laws of physics: they are much deeper absurdities, conceptual absurdities.
group of permutations with which Frege identifies the quantitative domain And they suggest that the mathematical continuum fits physical reality some­
contained all the elements it was required to have to be a quantitative domain what imperfectly, yielding apparent logical possibilities that are no possibilities
I at all. We are familiar with the thought that quantities obtained by differen­
on his definition: it is not true a priori that, for every' conceivable mass, there
is an object that has that mass. I tiation, like velocity and acceleration, do not possess their values at any particu­
Frege has thus not achieved a convincing analysis of the concept of a lar moment in logical independence of what their values are at all other
quantitative domain. His illustration, in § 163, does not greatly help: it is the moments; but the foregoing examples suggest that the same is true of all
usual one, used by Veronese, Holder and Cantor, of distances along a straight I quantities, even the fundamental ones, so that these are not ‘loose and separ­
ate’, as Hume absurdly said. But, if so, the mathematical continuum is not the
line; the underlying set comprises its points, and the permutations forming the
quantities of the domain are displacements along it. The example shows, correct model for physical reality, but only one we use because we do not ha\ e
indeed, that some quantitative domains conform to Frege’s model; it is power­ a better. In regarding real numbers as ‘measurement-numbers, Frege was
less to show that all can be so characterised. An adequate general characteris­ treating of a wholly idealised conception of their application, instead of giving
ation of the notion of quantity would pay much more attention to how it is an analysis of our actual procedures of measurement and their underlying
applied in practice; it would also embrace absolute domains, cyclic domains, assumptions. By doing so, he skimped the task he had set himself.
and domains of vectors of more than one dimension. Frege is so anxious to
press on to his definition of real numbers that he ignores all quantitative
Holder
domains save those that have the structure of the real line; as a result, he
offers a highly defective analysis of the concept on which he fastens so much Frege was not as out of step with other mathematicians as he
attention. Possibly this deficiency would have been corrected in a Part IV two years before the second volume of Grundgesetze aPPe^ ’ , ,0
which never saw the light of day. published an article treating of much the same topic as art
What would not have been corrected was the philosophical naivety of taking 10 O. Holder, ‘Die Axiome der Quantitat und die Lehre a^atiscL undphysikalische
it for granted that eveiy quantity has a precise value representable by the gen der Koniglich Sdchsischen Gesellschaft der ^issenschaften zu Letp^g. the iOrdan-Hdlder theorem
assignment to it of a real number relatively to a unit but discoverable by us i
i
Klasse, vol. 53, 1901, pp. 1-64. It was this Otto Holder expresses himself as of the
only to within an approximation. We are led to adopt this picture by devising i >s (in part) named. In his article, Holder does not mention K < L„wr quite unaware of
same opinion as he in regarding arithmetic as purely lope . footnote (p. 2, fa. 1) that
ever more accurate methods of measurement; but with what right do we i the advances in logic that Frege had pioneered, and remar
assume that its limit is a point, and not an interval, or at least that it is an i arithmetical proofs cannot be rendered in any exisnng lopeal ca
i
 282 Frege: Philosophy ofMathematics 22 Frege’s Theory of Reul Numbers 283
A comparison between them is extremely instructive. Holder is aiming at a
general theory of measurable quantity. He is as explicit as Frege about the
need for generality, and criticises earlier work by Veronese for failing to
separate the general axioms of quantity from the geometrical axioms governing
segments of a straight line.11 Holder characterises absolute quantitative ing to that cut associated with it. p
domains, without a zero quantity; he does so axiomatically in terms of an In the second part of his paper, Holder applies his theory to everybody’s
operation of addition, assumed associative, and a linear ordering relation, favourite example, of directed segments of a straight line. The interest of the
assumed dense, complete and left- and right-invariant, both taken as primitive. example, for him, lay in its indicating how to handle dual domains of opposite
Such a domain is then an ordered upper semigroup, although, like Frege, quantities, together forming a domain of positive and negative quantities of
Holder does not use explicit group-theoretic terminology-. He appears to have the kind Frege concerned himself with; but we need not follow the details of
been the first to give a correct proof of the archimedean law from the complete­ Holder’s treatment.
ness of the ordering, and also to prove the commutativity- of addition from the It is a matter for the deepest regret that neither Frege nor Holder ever
archimedean law. As we shall see, Frege obtained similar theorems in his Part ■
became aware of the other’s work. Had he had to comment on Holder’s theory
III.2; but Frege’s theorems are more powerful than those of Holder, because in his Part III.l, he could not have dismissed it so lightly as he in fact dismissed
his assumptions are considerably weaker.12 Dedekind’s theory: it shows very clearly how that theory can be applied to
For n a positive integer, and a a quantity, Holder easily defines the multiple
na in terms of addition. He proceeds to characterise the notion of a ratio
J ratios of quantities. In doing so, it also brings out more sharply than before
the exact nature of Frege’s objection to such a theory as Dedekind’s. Holder,
between two quantities, and associates a real number with every such ratio. I like everyone else except Frege, first defines the rationals, essentially as ratios
Unlike Frege, however, he does not construct the real numbers by this means. i
between positive integers, and then defines the real numbers in terms of them.
Rather, he first defines the positive rational numbers, in effect as equivalence I For that reason, although the principles underlying the use both of rationals
classes of pairs of positive integers.13 He then takes the real numbers to be and of irrationals to give the magnitude of a ratio between quantities are very-
defined by Dedekind’s method, which he sets out without Dedekind’s own direct, they are still external to the definitions of the numbers themselves.
appeal to mathematical creation, identifying the real numbers w-ith the corre­ I Frege, by insisting that rationals and irrationals should be defined together,
sponding cut in the rational line in which the low-er class has no greatest
element.14
I made it necessary' that that application of them be internal to their definition.
Put in that way, the difference between Frege and Dedekind, once we set
The correct definition of ratio, given addition and therefore multiples, was
aside the matter of free creation by the human mind, becomes much narrower
well known, having been framed by Euclid,15 and Holder appeals expressly to than one might suppose from Part III.l of Grundgesetze. There is a significant
it; it allow-s the comparison of ratios between pairs of elements of different methodological uuicrcnvc. for
lutuiuuuiugiuai difference: iui Frege,
i legs., the U1VVXJ of quantity
U1V theory- — --- an integral *part
is-------
domains, provided each has an operation of addition, but Holder confines of the foundations of analysis, not a mere addendum of interest primarily to
himself to comparisons within a single domain. Intuitively-, we shall w-ant to *V UllMULlVlllJ KZi WllUATUiU) — —--- -----

applied’•mathematicians.
’ • • • But the mathematical
’ j:w=—>«/-« kprnmps
-«----- 1 difference becomes moremore slen-
slen­
associate the rational number n/m with the ratio of a to b w-hen ma — nb. I der. In particular, if he had reached the point in Part III.2 of defining ratios,
Euclid defines a as having the same ratio to b as c has to d when, for all
Frege would have had to use the Euclidean definition, or some ng very e
positive integers n and w, ma is smaller than (equal to, greater than) nb if and it, and would thus have come Quite close tto------
quite close Dedekind’s conception oi the real
Op. cit., p. 37, fn. 1; see G. Veronese, ‘Il continue rettilineo e I’assioma V d’Archimede’, numbers.
Atti della Reale Accademia dei Lincei, series 4, memorie della classe delle scienze fisiche, matematiche
e natural!, vol. 6, 1889, pp. 603-24.
If the conjecture that vol. II of Grundgesetze was already written w-hen vol. I was published is The existence of a quantitative domain
correct, Frege could have had the priority if he had published sooner; but the mathematical
community would not have accorded it to him, because nobody troubled to read vol. II. § 159 conc^udes Part III.l, Frege resolves the doubt expressed in
He actually says, rather vaguely, that all equivalent pairs ‘represent, in accordance with our Quantit ” Or<^er t0 ensure the existence of the real numbers, at least one
(arbi^ary) interpretation, an object which we designate a rational numbeC (op. cit., p. 20). atlve domain must be proved to exist, containing quantities bearing
. . e Phrase used is again slightly vague: a cut ‘can be regarded as representing’ a rational or
irrational number, and, in the first case, ‘identified with it straight out’, and, in the second, ‘called 17 In thpraC^ete^ exPress*ons occur in Euclid’s definition, but are here superfluous.
an irrational number straight out’; op. cit., p. 22. the affin.-H, .n°te to p. 29, Holder ven- properly points out that Dedekind himself acknowledged
15 Euclid, Elements, book V, definition 5.
. m the Preface to H as sind und teas sollen die Zahlen?
 22 Frege’, Theory ofReal Numbers 285
284 Frege: Philosophy ofMathematics
and belongs to one of A and B but not the other, or is nm
irrational ratios to one another; for, if it did not, the real numbers, defined as belongs either to both A and B or neither; f e and
ratios of elements of such a quantitative domain, would all be equal to one
another and to the null relation. Furthermore, the proof must use only logical
(ii) t = r + s if 0 is free, and t = r + s + 1 if 0 is not free
resources. As in all cases, the domain will consist of permutations on an
underlying set. Frege observes that the set underlying such a domain must
This definition is intended to determine the relation as holding between
have a cardinality higher than the class of natural numbers; he mentions the
<r,A> and </, C> just in case a + b = q where <r, A> <s B> and
fact (not proved in Part II) that the number of classes of natural numbers is
I <t, C> intuitively represent the real numbers a, b and c respectively. As
greater than the number of natural numbers, but fails to make any acknowl­
Frege observes, we now have such relations corresponding to every' pair
edgement to Cantor.18 He therefore proposes to use classes of natural numbers
<5, B>\ taken together with their inverses, these correspond one to one to
in specifying the underlying set.
If we temporarily assume the irrational numbers known, Frege continues, the positive and negative real numbers; and to the addition of the numbers b
we can regard every positive real number a as representable in the form and b' corresponds the composition of the corresponding relations. ‘The
class of these relations (Relationen)’, he says, ‘is now a domain that suffices
for our plan’, but adds that ‘it is not thereby said that we shall hold precisely
F 2-"* to this route’.
k=l He could not hold precisely to it, because, in the coming series of formal
definitions, he requires a quantitative domain to consist of permutations on
where r is a non-negative integer, and »b n2, ... form an infinite monotone an underlying set; that is to say, he requires the relations it comprises to be
increasing sequence of positive integers. This amounts to giving the binary' one-one, all to be defined on the same domain and to have a converse domain
expansion of a - r (in descending powers of 2, as a decimal expansion is in identical with their domain. The relations mentioned in § 164, and formally-
descending powers of 10); the expansion is chosen to be non-terminating, so defined above, are not, however, permutations: the operation of adding the
that 1/2 is represented by the infinite series 1/4 + 1/8 + 1/16 ... Thus to positive real number b carries the positive real numbers into the real numbers
every positive number a, rational or irrational, is associated an ordered pair, greater than b. In Volume II, he does not reach the formal proof of the
whose first term is a non-negative integer r and whose second term is an existence of a quantitative domain. If, when he did, he had wanted to use
infinite class of positive integers (which suffices to determine the sequence); additive transformations, he w'ould have had to take the underlying set to be
►
these may be replaced respectively by a natural number and an infinite class isomorphic to all the real numbers, positive, negative and 0, which would haxe
of natural numbers not containing 0. been somewhat more complicated; if he had wanted the underlying set to be
This, then, is the underlying set; the permutations on it are to be defined isomorphic just to the positive reals, he would have had to use multiplicand e
in some such way as the following. For each positive real number b there is a transformations, which would have been very much more complicated to dehne
relation holding between other positive real numbers a and c just in case a + b with the resources available. There is, of course, no actual doubt at ei
= c. This relation can be defined, Frege says, without invoking the real could be done.
numbers a, b and c, and thus without presupposing the real numbers. He does
not here give the definition; the following should serve the purpose. Suppose
given an ordered pair <s, B>, where 5 is a natural number and B an infinite The formal treatment ,
class of natural numbers not containing 0: we want to define a relation between nt in Part IU-2, much has
When the reader comes to the formal deve °P™e_ ntjtatjve domain; and
similar such pairs <r, A> and <t, C>. Let us first say that a natural number been settled. The first problem is to character^ M satisfying a number
n is free if, for every m > n such that m belongs both to A and to B, there is he knows that it must be an ordered group ° P considerable; it is a
a number k such that n < k < m belonging neither to A nor to B. We may of conditions. The mathematical interest of tn .j^ 3S already noted,
then say that our relation holds if the following two conditions are fulfilled: thoroughgoing exploration of groups wi or ei p . Holder jjer jinn the paper
theorems more powerful than those Prcn^ ultimate purpose: he cou
(i) for each «, n belongs to C if and only if n is positive and either is free discussed above. The interest is not due to. Freg* ^nve domain to
18 This omission is truly scandalous; Frege would never have displayed such ill manners at the simply have laid down all the conditions 5.. The interest is due, ra er,
time of writing Grundlagen.
satisfy and incorporated them in a sing e e
 22 Frege's Theory of Real Numbers 287
286 Frege: Philosophy ofMathematics tions - in a quantitative d°main must satisfy the commutative and associative
to Frege’s concern for what we should call axiomatics, that is, for intellectual ,s j_|e then proves that composition of relations is always associative.20 It is
h?no means always commutative, as he remarks. A special case in which it is
economy: as he explains in § 175, he wants to achieve his aim by making the I
fewest assumptions adequate for the purpose, ensuring that those he does bYfirst singled out by Frege, namely the class consisting of a relation p together
make are independent ol one another. Hence, although a quantitative domain lS’th all its iterations p\p, p\(p\p)> • • ■ (Here the symbol j is used for composition,
will prove in the end to be a linearly ordered group in the standard sense, in U l lace of Frege’s own; no attempt will be made to reproduce his symbolism.)
which the ordering is both left- and right-invariant, many theorems are proved p \ uses his definition of the ancestral to express membership of this class
^ r±out reference to natural numbers (and hence to multiples of the form np
concerning groups with orderings not assumed to be linear or to be more than
"* used by Holder). Even when p is a permutation, the class of its positive
right-invariant. 35 holes will not always be a group. In this connection, Frege defines an
Before we proceed further, a word is in place concerning Frege’s formal
apparatus. A reader unfamiliar with it may have felt uncertain whether his m tant notion, that of the domain of a class P of relations. This consists of
quantitative domains contain objects, relations or functions. The answer is that 1p1P° ther with the identity’ and the inverses of all members of P. If P is the
they contain objects, but objects which are extensions of relations. The formal 1 • f all multiples of a permutation p, its domain will of course be the cyclic
system of Gnindgesetze contains expressions for functions both of one and of C aSS ° generated by p; but it should be noted that the domain of a class of
two arguments; these include both one-place and two-place predicates, that ^ermutations will not always be the group generated by it, since it is not
is, expressions both for concepts and for binary relations (Beziehungen). There reouired to be closed under composition.
is, however, no special operator for forming terms for value-ranges of functions The next problem is how to introduce the notion of order. Frege chooses
of two arguments: this is accomplished by reiterated use of the abstraction t an it hv defining the conditions for a class to consist of the positive elements
operator (symbolised by the smooth breathing on a Greek vowel)19 for forming o°f a group of permutations on which there is an ordering, and defining e
terms for value-ranges of functions of a single argument. Thus ‘e(e + 3)’ rlprhitr in terms of that class. His first approach is to introduce the notion
denotes the value-range of the function that maps a number x on to x + 3; of X he calls a pM class. A posMval chss is a class of pe—ns on
and so 'as(e 4- a)’ denotes the value-range of the function that maps a number some underlying set satisfying the following four conditions.
j on to e(e + y). This ‘double value-range’ is then taken by Frege as the
extension of the binary' function of addition. In the same way, (e(e < 3)’ (1) if p and q are in P, so is
denotes the class of numbers less than 3, while ‘de (e < a)’ denotes the value­
range of the function that maps a number y on to the class of numbers less (2) the identity e is not in P;
than y. This, being the double value-range of a relation (Beziehung), in this
case the ‘less-than’ relation, is identified by Frege with its extension, standing (3) if p and q are in P, then p\q~ is in the domain of P;
to it as a class to a concept; the extension of a relation, being an object, is
called a Relation, to distinguish it from a relation proper. This is just an (4) if p and q are in P, then p~\q is in the domain of P.
example of how, throughout Grundgesetze, Frege is able to work with value­
ranges in place of concepts, relations and functions. A quantitative domain Here ‘/r’ denotes the inverse of p. If P « a Pos^ grated by P- Frege
contains Relationen - extensions of relations - rather than relations in the true foregoing definition, the domain of P will be th gr ? less q lf
sense: specifically, extensions of one-one relations on an underlying set. We goes on to introduce an order relation on ' rektion < thus
may, for brevity, call these ‘permutations’; throughout Part III.2, Frege works and only if q\p~ is in P. It follows immediately TV p[r < for any element
exclusively with value-ranges of various kinds, concepts, relations and functions defined is right-invariant,21 that is, that if P «’
hardly ever making an appearance. For this reason, the word ‘relation’ itself
will henceforth be understood in the sense of 'Relation', namely as applying to 20 Composition of relations was
the extension of a relation (Beziehung) in the proper sense.
Frege begins by announcing that addition - that is, composition of permu- «„:x: css p’X‘
in the ^-relation iff x is the mother of.), the” would be wn ' Frege
” Some commentators on Frege write the smooth breathing over Greek consonants, which grandfather of y. In standard group-theoretic that notadon, o t0 a noBtion
the operation to be applied first being writteni first- bs.^ confasing to^
looks extremely odd. Of course, there is no logical mistake; but Frege never used Greek consonants defined his order relation to be /e/t-invanan , the variables
as bound individual variables, and it would surely have offended his sense of propriety to write a
that accords with Frege’s in respect of the or
breathing over them if he had.
I
 22 Frege's Theory of Real Numbers 289
288 Frege: Philosophy ofMathematics
I ment r of P which is an upper rim of J in P and is not less than any other
r of die group, and, further, that P is die set ot elements of die group greater ner rim of A in P which belongs to P. Since < linearly orders P, there can
than the identity e (the set of positive elements). Furthermore, it follows easily at most one upper bound, in this sense, of a class A‘. it is the greatest lower
from (1) and (2) that < is a strict partial ordering of the group (i.e. is transitive h nd in our sense, of the complement of A. If A is such as to contain every'
and asymmetrical). I fnent of P smaller than any element it contains, Frege’s upper bound of A
Frege is, however, extremely worried that he is unable to establish whether C 31 be its least upper bound in the usual sense. Frege’s formulation of the
or not condition (4) is independent of the other three. In fact, it is;22 Frege, W1 dition for < to be complete in P is that, if some member of a class A is
uncertain of the point, proceeds to prove as much as he can, from § 175 to C°n noer rim of A in P, but there is an element of P not in B, then some
§ 216, without invoking clause (4), and calls attention, in § 217, to the fact
member of P is an upper bound of A in P.
that at that stage he finds himself compelled to do so.
If clause (4) does not hold, the domain of P will not constitute the whole Frege continues his policy of avoiding appeal to clause (4) even after introdu-
group generated by it, which will in fact be the domain together with the the notion of a positive class. Oddly, he does not raise the question
C'hether clause (4), if independent of clauses (1), (2) and (3), remains indepen-
elements p-\q for p and q in P. We may nevertheless still consider die order
relation as defined over the whole group. Clause (3) in effect says that < is a d t after the addition of the assumptions of completeness and density; as we
strict linear ordering of P, and is equivalent to the proposition that it is a strict hll see it does not. Frege is concerned with the archimedean law, that, for
upper semilinear ordering of the group. This means that it is a strict partial anv positive elements p and q, there is a multiple of p which is not less than
ordering such ±at the elements greater than any given one are comparable, The formulates it with the help of the class of multiples of an element
and that, for any two incomparable elements, there is a third greater than both mentioned above. The most important theorems that he proves are as follows;
of them: pictorially, it may branch downwards, but cannot branch upwards.
Clause (4) says that < is a strict linear ordering of the negative elements (those Theorem 635 (§ 213). If < is a complete upper semilinear ordering, then
less than e), and is equivalent to the proposition that < is a strict lower
the archimedean law holds.
semilinear ordering of the group (where this has the obvious meaning). (3)
and (4) together are therefore tantamount to the proposition that < is a strict
linear ordering of the group. If the ordering is left-invariant, clause (4) must Holder had derived the archimedean law from the completeness of the order­
r hold, since, ifp < q, by left-invariance e < p~\q, i-e. p~\q is in P. (The converse, ing in his paper of two years earlier, but he was using considerably stronger
however, does not hold: a group may have a right-invariant linear ordering assumptions than Frege’s, namely that the ordering is dense, left-invanant and
that is not left-invariant.) Frege’s independence problem thus amounts to linear. The completeness of the ordering is needed to obtain the rea^ numbers,
asking whether there is a group with a right- but not left-invariant upper but it is the archimedean law that is important in the subsequent eorei” ’
/ Frege employed, though did not name, an interesting and fruitful con p,
semilinear ordering that is not linear. Since in fact there is, the theorems that
he takes care to prove without invoking clause (4) hold for a genuine class of namely that of a restricted kind of left-invariance s\hic we may -
groups.
The notion of a positival class wras only' a preliminary approach to that w'hich
Frege wants, namely that of a positive class. This is a positival class P such then p\q < p\r. The next theorem uses this notion.
that the ordering < is dense and complete. To characterise the notion of
completeness, Frege has of course to define the notion of the least upper :himedean ordering,
Theorem 637 (§ 216). If < is an upper semilinear, arcl
bound of a subclass A of P. His definition does not agree with what appears
to us the obvious wray of defining the notion. He uses as an auxiliary notion then < is limp.
what we might call that of an ‘upper rim’ of the class A\ r is an upper rim of . of Frege’s avoidance,
A in P if and only if A contains every member of P less than r (Frege gives These two theorems have been so expressed in 'i do appeal to
no verbal rendering of this notion, but only a symbol). What he calls an ‘upper in their proofs, of appeal to clause (4). The ne
bound (obere Grenze) or simply ‘bound’ of A in P is now defined to be an it.
, Z2Se®.S’A'Aje,eke> M.A.E. Dummett and Peter M. Neumann, ‘On a Question of Frege's
about Kight-Ordered Groups’, Bulletin of the London Mathematical Society, vol. 19, 1987, pp. Theorem 641 (§ 218). If < is a linear, archimedean ordering,
513-21, theorem 2.1.
left-invariant.
 22 Frege's Theory of Real Numbers 291
290 Frege: Philosophy ofMathematics essentially Euclid’s, all the same. This definition would give the criterion of
Theorem 689 (§ 244). If < is a dense, linear, archimedean ordering, then identity for ratios; we might therefore naturally expect Frege then to define
the group is abelian: that is, the commutative law holds for composition. the real numbers by logical abstraction, i.e. as equivalence classes of ordered
pairs of quantities. This, however, would not yield the result he demands in
Holder also derived commutativity from die archimedean law, but he had to his last sentence, that the real numbers should themselves form a quantitative
assume left-invariance, whereas, for Frege, left-invariance was automatic by domain, because they would then have to be extensions of relations, which are
Theorem 641. The assumption of density is unnecessary; but Frege’s appeal not, for Frege, classes of ordered pairs, but double value-ranges. He would
to it in his proof is not a fault, since different proofs are needed for the two therefore have had to use a variation of the method. A real number would
cases. have to be the relation between a quantity r and a quantity s of the same
With the help of Frege’s theorem 637, a further improvement can be
domain which obtained when r stood to s in the same ratio as some fixed
obtained, namely the quantity p stood to another fixed quantity q of the same domain, i.e. when the
pair <r, s> stood in the relevant equivalence relation to the pair <p, q>.
Theorem. If < is an archimedean, upper semilinear ordering, < is linear
If we imagine the axioms governing value-ranges to be quite different,
and the group is abelian.23
yielding a consistent theory’ analogous to ZF set theory', and Frege’s notion of
Thus clause (4) is no longer independent in the presence of the assumption an ordered pair replaced by the modem one, there would be no trouble about
of completeness, or even just of the archimedean law, which then suffices to any of the work in Volume II, Part III, and none about the proof of the
prove commutativity. existence of a positive class. The definition of the real numbers as ratios would,
With theorem 689, Frege reached the end of the quest for a proof of the however, be blocked, because their domain, as relations, would be the union
commutative law announced at the very' beginning of Part III.2, and therewith of all domains of positive classes, and the class of such domains would certainly
the end of Volume II (save for the Appendix on Russell’s paradox). A quantitat­ be a proper class. This, of course, was precisely the fate of Frege s definition
ive domain, in the narrow sense, could now with assurance be identified with of cardinal numbers, including the natural numbers. The paradoxes o f se
the domain of a positive class. theory imposed limits quite unexpected by him upon definition by logical
In his brief concluding § 245, Frege announces as the next task to prove abstraction.
the existence of a positive class, along the lines indicated in § 164. That, he
says, will open up the possibility of defining real numbers as ratios of quantities
belonging to the domain of the same positive class. ‘And we shall then also
be able to prove that the real numbers themselves belong as quantities to the
domain of a positive class.’
The missing conclusion of Part III.2 would have been laborious, but would
ha've presented no essential difficulties. The device of § 164 would have had
to be amended a little; but this would have required nothing but work. Frege
would have had essentially to use Euclid’s definition of when the ratio of a
quantity’ p to another quantity q of some domain D coincided with that of a
quantity r to a quantity s, both belonging to a domain E, whether the same as
D or, distinct from it. He would not have defined a phrase containing ‘the
same or coincides with, but would have defined an equivalence relation
between ordered pairs of quantities. (He had defined an ordered pair in
Volume I, § 144, as the class of relations in which the first term stood to the
second.) Nor, when he had hitherto refrained from appealing to the natural
numbers in characterising multiples of quantities, would he have been likely
to start doing so at this point; but his definition would of necessity have been

23 See S.A. Adeleke, M.A.E. Dummett and Peter M. Neumann, op. cit., theorem 3.1.
j
 23 Assessment
293
The application of mathematics

Of such applications' - Riemannian geometry and general relad’C fo “'a

CHAPTER 23 favourite example - and some attempt explanations in terms of the evoluttaL
advantages of an accord between human patterns of thought and the strucm2
of reality. Frege’s objective was to destroy the illusion that anv miracle occurs
Assessment The possibility of the applications was built into the theory- from the outset
its foundations must be so constructed as to display the most general form of
those applications, and then particular applications will not appear a miracle
Frege did not in practice carry out his own principles in this regard with
How should we evaluate Frege’s philosophy of mathematics? Strictly speaking, complete success. He failed to proride a sufficiently general analysis of domains
he did not have a philosophy of mathematics: he never enunciated general of measurable quantities, or a justification of the analysis he gave; and he
principles applicable to all branches of mathematics, or to all branches save failed to explore the physical and metaphysical presuppositions underlying the
geometrv; he never claimed to have more than a philosophy of arithmetic. In assumption that such a domain has a complete ordering, i.e. that every physical
this he does not compare very unfavourably with others, Hilbert for example. quantity has a precise determinate magnitude given by a real number (relatively
What he lacked in scope, he made up for in breadth of coverage and in to a unit quantity). Even his definition of the natural numbers did not achieve
precision. We are usually too impressed with the really creative ideas of Hilbert the generality for which he aimed. He assumed, as virtually everyone else at
or of Brouwer to pay much attention to the patchy or uncomincing soil in the time u-ould have done, that the most general application of the natural
which they are rooted. We pass over Hilbert’s sloppy account of the constitution numbers is to give the cardinality7 of finite sets. The procedure of counting
of the natural numbers and the content of finitistic mathematics, and readily does not merely establish the cardinality of the set counted: it imposes a
forgive him his failure to make precise the notion of a finitistic proof. We particular ordering upon it. It is natural to think this ordering irrelevant, since
overlook the inadequacy7 of Bromver’s repeated explanations of the genesis of any two orderings of a finite set will have the same order type; but, if Frege
the natural-number sequence, and ignore his solipsism and his failure to had paid more attention to Cantor’s work, he w’ould have understood what it
achieve a coherent account of the relation between mental constructions and revealed, that the notion of an ordinal number is more fundamental than that
their symbolic formulations. In Frege’s writings, by contrast, everything is lucid of a cardinal number. This is true even in the finite case; after all, when we
and explicit: when there are mistakes, they are set out clearly for all to count the strokes of a clock, we are assigning an ordinal number rather man
recognise. a cardinal. If Frege had understood this, he would therefore have charactensed
Frege had answers - by no means always the right answers, but invariably the natural numbers as finite ordinals rather than as finite cardin . e w
definite answers - to all the philosophical problems concerning the branches well aware that Cantor was concerned with ordinal rather than cardinal num­
of mathematics with which he dealt. He had an account to offer of the bers in the first instance;2 but, since he never carried his own stu
applications of arithmetic; of the status of its objects; of the kind of necessity
attaching to arithmetical truths; and of how' to reconcile their a priori character transfinite arithmetic further than to prove some theorems a ou .
Anzahl Endlos’), he dismissed the difference as a mere divergence ot inte ,
with our attainment of new knowledge about arithmetic. His view of the status
of the numbers, ontological and epistemological, proved to be catastrophically and never perceived its significance. . • nav scant attention
wrong; for the last nineteen years of his life, he himself acknowledged it to An exception to the rule that philosophers of mathema P . conneCtioni
have been wrong, and regarded that as bringing with it the collapse of his to its applications is Wittgenstein. He criticised rege 161;
entire philosophy of arithmetic. In spite of efforts like those of Crispin Wright , '^ee Mark Kac and Stanislaw Ulam, Mathematics and ^rmanv mathematical
to defend it, we can clearly see that his view of this question was in error: but There is little doubt that the “external world” has been the so . ,e(J' qUjle independen j o
we have not supplied any very7 good alternative. In answering the remaining
questions, wre have not, save in one crucial respect, advanced very far beyond
Frege at all.
292
 294 Frege: Philosophy ofMathematics
23 Assessment 295
without apparently, having understood him, and certainly without appreciating
how far more sophisticated Frege’s view was than his own. He described
Frege’s new, maintained against die formalists, as being that ‘what must be occurred at a late stage in the development of a theory- it is the IT—”5
added to the dead signs in order to make a live proposition is something Of the process of rendering it fully rigorous. What, Tn to Ito
immaterial, with properties different from all mere signs, and retorted, But acknowledgement is that an analysis of the general form of the appHcatiZ of
if we had to name anything which is the life of the sign, we should have to
say that it was its use') As a critique of a passage in which Frege said that it a theory is the proper business of mathematics; no other science is competent
i to undertake it, so that, while it remains undone, the mathematical theory has
was applicability alone that raised arithmetic above the rank of a game, this
not yet been supplied with adequate foundations. If, then, we hit upon an
remark is astonishing; but equally astonishing is the crudity of Wittgenstein’s
application of a type not provided for in the existing foundations of the relevant
conception of the application of mathematics, which would do very well for
theory, we need to analyse what made that application possible, and in the
explaining why ‘B-Q6’ is not a mere mark on paper. An adequate account of
light of that revise the foundational part of our theory, or prove a more general
the application of mathematics must, after all, not merely explain how it can
representation theorem, accordingly.
be that mathematics is applied, but must do so in a way that does not make
There is an unfortunate ambiguity- in the standard use of the word ‘struc­
it puzzling that there can be such a thing as a pure mathematician. Wittgenstein
hankered after a view of mathematical formulas as not expressing propositions,
ture’, which is often applied to an algebraic or relational system - a set with
certain operations or relations defined on it, perhaps with some designated
true or false, but as encoding instructions for computation, although he did
not attempt to show how such an interpretation could be carried through;4 if elements; that is to say, to a model considered independently of any theory
which it satisfies. This terminology- hinders a more abstract use of the word
it could, the existence of pure mathematicians would indeed be hard to explain.
Indeed, Wittgenstein’s view closely resembles a bad, outmoded method of ‘structure’: if, instead, we use ‘system’ for the foregoing purpose, we may
speak of two systems as having an identical structure, in this more abstract
teaching mathematics in school, which drilled the pupils in techniques of
sense, just in case they are isomorphic. The dictum that mathematics is the
computation without explaining to them why they worked, far less proving that
they did or even indicating that such proofs were possible. Frege, by contrast, study of structures is ambiguous between these two senses of ‘structure’. If it
ascribed to mathematical sentences a sense, which we grasp by apprehending is meant in the less abstract sense, the dictum is hardly- disputable, since any
what will determine them as true or as false. The sense, if correctly explained,
model of a mathematical theory- will be a structure in this sense. It is probably
usually intended in accordance with the more abstract sense of ‘structure’; in
I is intimately connected with the possible applications of the theory to which a
this case, it expresses a philosophical doctrine that may be labelled ‘structur­
given sentence belongs, and so such applications lose their mystery; but it also
presents a problem independent of all applications, namely whether the sen­ alism’.
Even so, the term ‘structuralism’ still admits a stronger and a weaker
tence is true or false, and it is therefore likewise unmysterious that this problem interpretation, comparable to the two interpretations of the phrase iormal
may be studied for its intrinsic interest, regardless of any extra-mathematical
theory’ proposed by Frege in his lecture ‘Uber formale Theorien der Arithme-
use that may be made of it.
tik’ of 1885. On the stronger interpretation, structuralism is the doctrine that
Frege s precept obviously should not be taken as ruling out the theory of a
mathematics in general is solely concerned with structures in the abstract
class of algebraic systems defined by their structure, that is, closed under
sense, that is, with systems left no further specified than as exemplifying e
isomorphism, such as groups, rings, Boolean algebras and the like: what
structure in question. This doctrine has, again, two versions. According to the
encapsulates the general principle of possible applications of any such branch more mystical of these, mathematics relates to abstract structures, distmguisnea
of algebra is a representation theorem. Nor can Frege be read as preaching
by the fact that their elements have no non-structural properties, e a ®
’L. Wittgenstein, The Blue and Brown Books, Oxford, 1958, p. 4. four-element Boolean algebra is, on this view, a specific system, wi 1‘
. ,-.k " Re.mark* nn‘he Foundations ofMathematics, Wittgenstein asks, ‘Might we not do arithmetic
elements; but, for example, the zero of the algebra has no 0 er■ p _
. tk U • a'?n^ l '^Ca utter‘nS arithmetical propositions, and without ever having been struck
of r 7 ■ multiPlication and a proposition?’, and comments that ‘it is a matter than those which follow from its being the zero of that 00 ea
me«Cla r®Ia,10ns^P (original edn., ed. G.H. von Wright, R. Rhees and G.E.M. is not a set, or a number, or anything else whose nature ‘
‘PennT™? Pa? ’/’’‘V’ § 4’ revis!d edn-’ 1978> Part !> aPP- JII> § 4>- Elsewhere he says, algebra. This may be regarded as Dedekind’s version ot X^have
can an app,'ed mathematics without any pure mathematics. They him, the natural numbers are specific objects; but ey, arr )
time The e«-Pa descrlbed by certain moving bodies and predict their place at a given
edn., pan III, ji 15, wis^ mathematics may be quite foreign to them’ (original no properties save those that derive from their position in the
i
infinite system (sequence of order-type a>).
 296 Frege: Philosophy of Mathematics 23 Assessment
297
That there can be abstract objects possessing none but structural properties
is precisely what is denied by Paul Benacerraf: the denial is his ground for
Mathematics must preserve its logical virginity intact- and h J
holding that numbers cannot be objects. His is therefore the more hardheaded
version of structuralism, one misattributed by Russell to Dedekind himself. belonging to physics or any other of the sped,,
purity of a mathcmaocal theory, whether in its foundations or its 1’““’^'
According to it, a mathematical theory', even if it be number theory or analysis
which we ordinarily take as intended to characterise one particular mathematical ture. For that reason, mathematics should have nothing overtly to do uXT
system, can never properly be so understood: it always concerns all systems details of any specific application of any of its theories or concepts Henre dT
with a given structure. The difference between, say, number theory and group principle governing the applications of a mathematical theory, which is to be
theory, on this view, is merely that the structure with which the former is incorporated into its foundations, must be formulated in completely general
concerned is specific; that is, its subject-matter consists of a class of systems terms: it relates to the structure of those applications, and in no way to their
isomorphic to one another. It is part of such a view that the elements of the specific contents.
systems with which a mathematical theory' is concerned are not themselves It is this generality that must be respected both by those who think, like
mathematical objects, but, in a broad sense, empirical ones; it is not the Frege, that the claim of mathematics to be a science derives from its appli­
concern of mathematics whether such systems do or do not exist. cations, and that the true meanings of its propositions relate to them, and bv
One of the weaknesses of the hardheaded version of structuralism is that, those who think it is justified by those applications. Because the radical formal­
while it may not be for mathematics to say whether or not there exist any
ists make application external to mathematics, each application has to be
systems exemplifying the structures that it studies, the subject would appear
treated separately, consisting as it does, for them, in devising a particular
futile unless there was a strong chance that they would exist. The more mystical
version might seem to escape this difficulty, holding as it does that the purely empirical interpretation of a formal calculus; there can therefore be no general
abstract systems are free creations of the human mind. For Dedekind, however, principle. The same holds good for a neo-Hilbertian like Hartry Field. From
the process of creation involved the operation of psychological abstraction, the standpoint of a highly selective nominalism, which abhors real numbers,
which needed a non-abstract system from which to begin; so it was for him a but countenances space-time points, and space-time regions, as being suffi­
necessity, for the foundation of the mathematical theory, that there be such ciently physical entities, he rejects all claims that such a theory as real analysis
systems. That was why he included in his foundation for arithmetic a proof could actually^ be true. On his view, it must, rather, be justified indirectly, as
of the existence of a simply infinite system, which had, of necessity, to be a possessing a stronger property than that of being consistent with a scientific
non-mathematical one. theory within which it is applied, namely that of yielding a conservative exten­
Dedekind thus shared with the hardheaded structuralist the need to maintain sion of that theory when adjoined to it. We have thus to show that anything
that we can find infinite systems of objects - systems isomorphic to the natural statable in terms of the scientific theory and provable from the composite the­
numbers and others isomorphic to the real numbers — in nature; and the thesis ory could have been proved from the scientific theory alone.
is questionable. It may be held, indeed, that time, for instance, has the structure The notion of a conservative extension makes sense only if the theory to be
of the continuum; but this seems more a matter of our imposing a mathematical extended is formulated in a language more restricted than that of the exten e
structure on nature than of discovering it in nature. In his late essay ‘Er- theory. Hence, to give sense to Field’s claim, he has to make the poor claim
kenntnisquellen’, Frege made as robust a declaration as did Hilbert, at just to be able to reformulate scientific theories so as to avoid any apparent re
the same time, in ‘Uber das Unendliche’, that the infinite could never be ence to the spurious objects of the mathematical theory' such as real numi ,
found in empirically given reality; but he did not manifest the same conviction indeed, if the nominalistic motivation is to be satisfied, reference to■ ah other
in Part III of Grundgesetze, which rests on the assumption that there are abstract objects unacceptable to a nominalist of his persuasion , can
domains of physical quantities isomorphic to the real numbers. eliminated. This reformulation is the harder of Field s two tas ■
However this may be, the two types of structuralism — the mystical and the so frame physical theories as to eschew all abstract ob>ect% sountjness of
hardheaded — are variants of the strong interpretation of the view that math­
ematics is about structure. Frege rejected this strong interpretation of structur­ These difficulties would vanish for anyone convince ;ustification of
Frege’s invocation of the context principle to yield a gm presents
alism, primarily because it conflicted with his concern for applications: the
general type of application to be made of a system such as the natural, the abstract (non-actual) objects. The existence of ma ema jjewrOte Grund-
real or, presumably, the complex numbers was, for him, constitutive of those special problems, however, as Frege was already aware^w pain-
lagen, became more vividly aware in writing Gru g‘se ~'
 298 Frege: Philosophy of Mathematics 23 Assessment
299
fully more aware vet by Russell’s contradiction. Hence, even for someone free theories.5 But this argument cuts both ways. If it
of qualms about abstract objects in general, Field’s objective retains an interest
as a means of indirectly justifying appeal to specifically mathematical objects
and structures.
Field envisages the indirect justifications at which he is aiming as being
sssssrsSSS
better explanation could there be than that the theorems of Zmath
theory are analytically truer No doubt, in view of the difficulties maAemafca
obtained piecemeal; but this violates Frege’s principle of generality, that a objects pose for logicism, this explanation is not available- but it is bette t
uniform explanation be provided for all the applications that may be made of search for some related property possessed by mathematical theorems than o
any given mathematical theory. Consistency is an absolute property of an acquiesce in a case-by-case justification of applications of them. Of this we mav
arithmetical theory; so is analyticity: such a theory- has the one property or the be quite certain, for any given mathematical theory-: either Field’s programme
other irrespective of whatever other theory-, physical or mathematical, it may cannot always be carried out; or it can, and there is some general explan­
be applied to. Frege argued consistency to be too weak a property to w-arrant ation of that fact which will, of itself, constitute a justification of that theory
our concluding to the truth of propositions derived by applying the theory, If Field’s general objection to abstract objects were replaced by a restricted
and claimed analyticity as necessary for this. Conservativeness, on the other objection to mathematical ones, would anything then remain of Field’s strategy
hand, is not merely an intermediate property-: it is relative to the theory within for avoiding reference to such objects? That strategy would be considerably-
which the application is made. Hence success in one case would not guarantee simplified, but the programme would not become altogether otiose. Field’s
success in another: it therefore appears that the programme would have to be strategy- is to reformulate a given scientific theory- in nominalistic terms, and
carried out separately for each scientific theory in which the mathematical then to prove a representation theorem for the reformulated theory- in terms
theory- found application. Even if we accomplished the task for all existing of real numbers (or of whatever objectionable mathematical objects are
scientific theories, it would have to be done afresh for any new theory that appealed to in the usual formulation). The nominalistic reformulation replaces
was devised that made use of our mathematical theory. But this is contrary to references to such quantities as temperatures by predicates applying to space­
reason. The theory- of functions of a real variable, for example, is surely not time points; for instance, one stating intuitively that the temperature at j is
one that requires separate justification for each application that is made of it: intermediate between that at x and that at z, one stating intuitively that the
to whatever extent it needs justification, it must be justifiable once for all, in
difference between the temperatures at x and j is equal to that between those
at z and m, and one stating intuitively that the temperature at x is less than
such a way as to be available both for the formulation of a scientific theory
and for use in conjunction with it. that at y. (To deal with mass in this w-ay, we need to consider density- at a
point.)6 With the general objection to abstract entities waived, there is no
Suppose that some institute undertook to try to carry out Field’s programme,
reason w-hy the reference to quantities that occurs in the usual formulation
vis-a-vis the theory of real numbers, for all know-n scientific theories: and should be eliminated. Those quantities would not be postulated to be rep­
suppose that it achieved definite results in every case. If it established that, resented by real numbers, how-ever, or by numbers ot any other kind, the
for one or more scientific theories, the programme could not be carried out, properties of the quantities treated of in the theory that result from their
we should have to conclude that the theory of real numbers required a justifi­
numerical representation w-ould have to be stated directly, so as to ow o
cation of a kind different from that emisaged by Field. If, on the other hand, the subsequent proof of a representation theorem. In this way, the scienti c
it was shown that the theory- of real numbers yielded a conservative extension
when added to any one among all known scientific theories, we should surely theory w-ould still require reformulation. , ,
The non-nominalistic modification of Field s programme us s e
suspect that some general principle w-as involved, and that we were wasting provides a glimpse of how- the generality- principle mig t e reins ■
our time tackling each scientific theory individually. Indeed, the repeated Frege’s characterisation of the real numbers as ratios of quantities is
success of the programme would demand a general explanation. Presupposing then, given a far better analysis than he provided of when proftert s assi^
the feasibility of his plan of reformulating physical theories, Field argues that
by a physical theory- to bodies (or to space-time regions or p
those who consider mathematical theories to be true necessarily or a priori
must allow that they have the w-eaker property that, added to any other theory ’ H. Field, Science Without Numbers, Oxford, 1980, PP- 12—13- of respect for conceptual
w-hatever, they will yield a conservative extension of it; the fact that so many To treat mass density as primitive certainly violates the req • m3SS as product
have held them to have the stronger property is, he thinks, suasive evidence priority, since we normallv think of density as mass/volume, evorbitant when imposed on
that they have the weaker one, at least relatively to reformulated physical of density and volume. It may be retorted that the requtrem
formulations of physical theories.
 23 Assessment
applications of it which Frege demanded should hP • 301
300 Frege: Philosophy ofMathematics
classified as quantities, a general representation theorem could be proved for foundations. Only by following this methodological n mCOrp°T!lt^ mto its
theories satisfying those conditions; such an analysis is supplied by measure­ of the theory be prevented from assuming the guise f? Can Rations
ment theory. We should then have an account of the application of real so can philosophers of mathematics, and indeed ^^ulous; only
numbers that diverged far less widely from that at which Frege aimed, but fell apprehend the real content of the theory; Admittedly th * °f sub’ect>
short of supplying, than that proposed by Field himself. principles governing the of theis dffi 1 * mathtmati«l
It would not even be necessary; within such a modified version of Field’s
general
theory should be incorporated intoapplication
their definition
programme, to eliminate all reference to real numbers widiin tire physical ible, always to follow in practice. It remains clear tlJ k Perhaps imposs'
theory itself, since they could be treated as ratios between the quantities of it is the most direct way of embodying in the foundah'/6”?^ PraCticable>
which the theory treated, rather than as given antecedently by tire mathematical analysis of what renders its applications possible- but the the°13’ an
theory7. Such real numbers would not be mathematical objects, for the distin­ advanced by Frege in this connection is that’such an™S5mpOrtantthesis
guishing characteristic of mathematical objects is that their existence presup­ embodied hr some means, as being the proper business of ma*e”afa Jd
poses nothing about empirical reality; these real numbers would, rather, be drat which renders ,t a sconce - a genuine sector in the ,uest for™*
dependent abstract objects in the same sense as that in which the Equator failed, by quite a large margin, to achieve the analysis that his nhdnt
depends for its existence upon contingent features of the world: which real principies demanded, even in the two comparatively elementary casisShe
numbers existed would depend upon which quantities existed, according to
the theory. If we call the Equator an abstract geographical object, real numbers tackled, but he surely pointed the direction we need to take.
so explained could be called abstract physical objects. For that reason, their
existence would be unproblematic in the light of the context principle. What
would now be violated would be the purity7 of the mathematical theory; for the
theory of these real numbers would no longer be a branch of pure mathematics, Platonism and logicism
but a fragment of the physical theory. atean obii« e d°C[rineJthat theories relate to st,terns of
All this leaves the problem of mathematical objects unresolved; but, if we theories are d t eMSbng indePendendY of us, and that the statements of those
set that problem aside, we can surely say that Frege’s ideas concerning the doctrine has an ™ately tru,e or faIse independently of our knowledge. This
application of mathematics were surely sound in outline. It cannot be by a series Dhilosnnh ■ i 0 J1ous aPPeai to the pure mathematician, but raises immediate
of miracles that mathematics has such manifold applications; an impression of teri j oh- P^ob ems- ^ow can we know anything about this realm of imma-
a miraculous occurrence must betray a misunderstanding of the content of the univer S. nd b°W can Pacts about it have any relevance to the physical
theory7 that finds application. Frege was right to hold that it belongs to the unde bow’ in other words, could a mathematical theory; so
task of mathematics to analyse the principles in accordance with w'hich each On ,JS 00 ’ be applied? Logicism is not a natural ally of platonism, because,
mathematical theory7 is capable of being applied, not separately for each appli­ de f 6 m°St natUra^ ^ew oP logic, there are no logical objects: it was a tour
cation, but in a general fashion that will cover them all. The genesis of most an u°FCe °n Frege’s part to combine a vehement advocacy of platonism with
mathematical theories was due in the first instance to the need to arrive at a late nrjSen ed logicism about number theory7 and analysis. The most celebrated
logical analysis of one or another empirical problem. Certainly the development inwh^ L°^ate opPlat°msm, Kurt Gbdel, presented it in a non-logicist form,
of the theory7 requires us to ‘leave the ground of intuition behind’: we do not math 1C ’ ?ndeed’ prompts the two objections concerning the applications oi
have a properly mathematical theory7 until we have ceased to rely upon our no v ematics and our knowledge of it. To the former he had, so far as I know,
apprehension of the perceptible or experiential, and have attained that gener­ op 5^ good solution; the latter he solved by postulating a faculty of intuition
ality7 w'hich Frege would think entitled the theory to be recognised as a branch rais Stfact obiects, in analogy7 with the perception of material ones. But this
of logic - the generality7 demanded by structuralism in its weaker sense. At f0 S e farther problem, why proof is so salient in mathematics. The search
this stage, the theory is likely to admit a wider class of applications than those to tb1^3^01115 f°r set theory7, recommended bv Gbdel, might be compared
which originally prompted its development; but, when we have reached the reaIm °bsen'ations made by astronomers; but, if the analogy with the physical
stage of setting the theory7 upon firm foundations, wre must not be tempted by7 on <su u Sound, what would explain why mathematicians spend so little ome
the strong \ ersion of structuralism to lose sight both of the original applications the rn °bservati°ns, and so much on eliciting by means of complex deductions
and of possible future ones. The historical genesis of the theory7 will furnish an sPecn|n!-eqUenCes oi Pacts already observed? Why, indeed, do they not e a ora
indispensable clue to formulating that general principle governing all possible a lve theories which need testing by further obsenations, as na
 23 Assessment
302 Frege: Philosophy oj Mathematics 303
even easier than Russell’s did to see where’® cons^ U
scientists do, and why do they demand incontestable proof rather than high
probability, as a warrant for asserting a mathematical proposition? Uncertainties ity in accordance with our narrower conception of logical truth
about the formation of stars, or die behaviour ot Cepheid variables, do not to appeal to the status of natural numbers as objects solely in'orS tr
reflect any haziness in our grasp of die concept oi a stai, but only a defect in the infinity of the natural-number system; we may conclude that he succS
our knowledge of the behaviour of stars. Likewise, if die analogy between in showing to be uncontroversially analytic all arithmetical propositions1 t
physical and ideal objects were sound, our uncertainty about the continuum do not require the existence of infinitely many natural numbers - essential
hypothesis need show no haziness in our concept of a set, but only in our
knowledge of w hat sets God has chosen to create; for presumably ideal objects finitistic statements in Hilbert’s sense. A proposition may be said to be uncon­
A troversially analytic if it is analytic on the narrower conception, according to
are as much God’s creation as physical ones. Physical objects have many
properties neither revealed by’ immediate observation, nor deducible from which there are no logical objects: the thesis that all arithmetical truths are
those so revealed; we can hope to discover them only by making further uncontroversially analytic would be that of non-platonist logicism.
observations, and, from a realist standpoint, cannot be certain of discovering Non-platonist logicism was not a possible route for Frege because it allows
all of them even then. Were the analogy’ sound, mathematicians w’ould treat no access to the infinite totalities he took to be essential for mathematics. It
the ideal objects which they study in a similar way: the inappositeness of such is not merely that he w’ould have been unable to prove that there are infinitely
a description of their activities serves to point the lameness of that analogy. many natural numbers - it could after all be objected that his alleged proof is
Frege, in virtue of his logicism, had none of these objections to face. If the circular: it is that we should have no reason to suppose it true that there are
natural new is taken of logic, according to which there are no logical objects, infinitely many natural numbers. That is why his combination of logicism with
the logicist programme, if it could be carried out, W’ould provide an interpret­ platonism, had it w orked, would have afforded so brilliant a solution of the
ation of all mathematical statements in the language of higher-order logic. problems of the philosophy of mathematics. The logicism explained how
This interpretation would dispense with all mathematical objects, w’hich w’ould mathematics could be applied, how’ we could know mathematical propositions
disappear in favour of higher-order properties and relations; wre should thus to be true, and w’hence their necessity derived; the platonism justified the
have a non-platonist logicism. The application of mathematical statements existence of mathematical objects and clarified their status. Frege’s idea was
► would then be quite unproblematic: application w’ould simply consist in instant­ that such objects should always be defined as extensions of concepts directly
iation of universally quantified formulas. This was, in effect, w’hat Whitehead related to the application of the mathematical theory concerned: concepts to
and Russell attempted in Principia Mathematica, since their classes are only do with cardinality7 in the case of the natural numbers (and other cardinal
surrogate objects, affording a disguised means of speaking of (higher-order) numbers), concepts concerning the ratio of one quantity’ to another in the case
properties and relations; the ramified hierarchy of types seeks to evade the of the real numbers. In this wray, application could be understood as being no
objections to the impredicative character of higher-order quantification. Their more problematic than it w’ould be according to non-platonist logicism. it
attempt ran against the difficulty that W’ould have supplied the only valid would not consist in pure instantiation of formulas of higher-order logic, but
ground for Frege’s insistence that numbers are genuine objects, the impotence would involve deductive operations so close to that as to dispel all myster)
of logic (at least as they understood it) to guarantee that there are sufficiently about how’ application was possible. A mathematical theory, on this view, oes
many surrogate objects for the purposes of mathematics, forcing them to make indeed relate to a system of abstract objects existing independent!) o us. ej
assumptions far from being logically true, and probably not true at all: to are not, however, pure abstract objects in the sense in which we spea o pu
secure the infinity of the natural-number sequence, they had to assume their sets (sets all the members of whose transitive closures are also sets): me
axiom of infinity, and to secure the completeness of the system of real numbers, objects characterised in such a wav as to have a direct: connection
they had to assume the axiom of reducibility. logical concepts relating to any one of the particular domains o “
Frege failed to establish the logicist thesis, as he himself understood it; but, Physical universe among them. They could not otherwise have the applications
when we declare that he failed to establish it, we are inclined to forget that
his interpretation of it was more generous than ours, just because he believed that they do.7
in logical objects, and we, taking a narrower view of what logic is, do not. On a ’Thus Frege’s cardinal numbers must be thought of as «n“in,n? that *e
actual objects. This indeed conflicts with his implication m GrunW• ^pi^on. though
his definition, a statement is analytically true if it can be derived, by the help n ’ects of the theory are restricted to truth-values and value-rang: • cardjnai numbers
of definitions, from a logical truth; a broader or a narrower conception of will ’?a^vertent *n the context, makes nonsense of his plain ,n. .. remarl; applies to his rea
''11 be those involved in empirical ascriptions of number; a s.mdar rem
analyticity must then result from adopting a broader or a narrower conception
numbers.
3Q4 Frege: Philosophy ofMathematics 23 Assessment
One reason why it is convenient to express mathematical theories in terms to that
over of sets
such specifying
(which truth-conditions
we shah be unable do even in '•nV°!,v’.n^ Quantification
fortostatement
of objects such as numbers of various kinds is that non-logical abstract objects if we have failed to provide for the possibility of ther^ matbctn^s
frequently figure in the physical theories to which the mathematical ones are natural numbers). If we do not follow Frege in th infiniteI? many
applied. The cause of nominalism cannot be advanced by dispensing with becomes that of specifying truth-conditions outright fi* Ae problem
mathematical objects such as real and complex numbers, but allowing point­ numbers. It does not help to rephrase the problem as th Sftatfements ab°ut real
instants in physical space-time to continue to infest the physical theory; conditions for statements about what real numbers th fpecif^'inS truth-
/; if
abstract objects do not deserve entry' visas, they cannot improve their case by
producing passports issued by physics rather than mathematics. The converse,
of course, does not hold: if the context principle licenses reference to abstract of logical, and hence of mathematical objects Dc hT°J existence
objects in general, that does not imply a liberty to assume the existence of to be solved piecemeal, in a different way fo^d^S ** Pr°blem *
mathematical objects of all kinds. Why, then, does there appear to be a simultaneously for all possible cases, as Frege hoped- his ?!
compelling need for mathematical objects? The need arises from the concern it should not be allowed to obscure all orbJ P d’ , fai ure t0 solve
of mathematics with infinity. It has to be concerned with infinity' because of arithmetic. aI1 Other aspects his philosophy of
the generality of its applications: even if we were fully convinced that everything
to which mathematics would ever be applied would be thoroughly' finite, we
cannot set an upper bound in advance on the number of its elements, or a
lower bound on the ratio of its magnitudes. There cannot be infinitely many The fruitfulness of deductive reasoning
properties or functions unless there are infinitely many objects to start with; Deductive inference patently plays a salient part in mathematics. The ■correct
infinity must be injected at the lowest level. Granted, for a particular appli­ observation that the discovery of a theorem does not usually proceed m acco,
cation, the mathematical theory' might borrow its objects from physics, or ance with the strict rules of deduction has no force: a proof has to be set ou
whatever other empirical science it was being applied to, if that science claimed in sufficient detail to convince readers, and, indeed, its au or, o t
an infinity of them to dispose of. This, however, would both violate the purity deductive cogency. The philosophy of mathematics is concerned wrf> te
of the mathematical theory', and offend Frege’s principle of generality': as he
product of mathematical thought; the study of the Proc^ss 0 P’° “ ty)eon-
insisted, that theory is not of itself concerned with particular applications, but
with the general principle underlying them. It must therefore be justified once concern of psychology’, not of philosophy. Althoug a eoiy axiomatise
for example - may thrive for a long time before anyone dunk:to awmanse
for all, and not separately for each application. This requires that it have its
own objects, and not borrow them from different physical theories in turn. it, experience suggests that all mathematical theories, » essentiallv
Frege argued, correctly, that the bare consistency of a theory’ does not suffice developed, are capable of axiomatisation, though often o y t , . .'
second order language. The failure of the logicist thests can there orbe
to warrant its applications, within mathematics or outside it; he concluded
that, to justify such applications, we must require the theory to be true. He localised in the justification of the axioms; a great part, at east, o -
admittedly did not envisage the possibility canvassed by Field of showing it to of mathematical theorems is the necessity of deductive consequen ejv
have a property- stronger than consistency, but weaker than (analytic) truth; we Frege are virtually the only two philosophers to have a resse w ■
may leave it to the final chapter to consider whether this is a genuine alternative. the most striking, and perplexing, fact about deductive reas"nin^a^, js t0
It is not inaccurate to express this by saying that mathematics must be capacity' for yielding new and often surprising knowle ge, e por
adapted for all possibilities. A less happy formulation is that it is concerned explain this capacity’ without undermining our perception o i wj,atever
with possible, not really existing, objects; and this has suggested a formulation it to be cogent, we must be allowed to be able to recognise
of mathematical theories using modal logic. The suggestion does not, however, renders the premisses of a deductive step true already rendersson _
go to resolve any genuine dilemma. The problem w-hich Frege failed to solve true; for it to be fruitful, we must be able to grasp the premisses and ac: ow -
was to specify definite truth-conditions for statements involving reference to edge them as true without perceiving the possibility ofdrawing *at c0^ If
an quantification over value-ranges, which required a determination of what Frege’s solution of this problem must be along the right genera^n *
v a ue ranges were to belong to the domain; our problem is to do the same deductive inference were not a creative process, proving eor involving
or, say, real numbers. If we follow Frege in deriving the existence of real mechanical activity; Frege sought to explain its creative <^mtterns there to
num ers from that of infinite sets of natural numbers, the problem reduces the recognition of patterns common to different thoug P for
e recognised, whose recognition was nevertheless not
/
 306 Frege: Philosophy ofMathematics
thoughts to be grasped. He was satisfied to restrict the recognition of such
patterns to a particular tvpe of case, namely the discernment of complex first-
or higher-order predicates within a sentence or of functional expressions within
a complex term; though undoubtedly important, this special case is surely
inadequate to bear the full weight of an explanation of the fruitfulness of CHAPTER 24
deductive reasoning. The discernment of common patterns that accounts for
its fruitfulness is not to be confined to patterns exhibited by individual propo­
sitions, but must relate also to sequences of propositions that make up a proof
or the description of an effective procedure. The proof or procedure usually The Problem ofMathematical Objects
does not require a unique ordering of the propositions, which may be
rearranged without destroying the validity of the proof or the effectiveness of
the procedure; perception of a pattern common to two such sequences will
normally require apprehension of the possibility' of such a rearrangement. This The necessary existence of mathematical objects
topic deserves detailed study by cases, which it has not received because
philosophers seem oddly uninterested in it, being content to accept that deduct­ The logicist thesis failed because of its inability to justify the existence of
ive reasoning is both cogent and astonishingly fruitful without bothering their mathematical objects, more particularly of systems of objects satisfying the
heads to explain how this can be so. There can be little doubt, however, that axioms of the theories of natural numbers and of real numbers. More precisely,
the general idea underlying Frege’s explanation of its fruitfulness must be Frege’s attempt to establish the thesis failed, even according to his more
correct; it is difficult to see how an explanation could be offered along any generous interpretation of it, because his application of the context principle
I other lines. failed to justify their existence. The problem is best thought about in connec­
■

tion with the necessary truth of mathematical statements. Their necessity is

enough to rule out our possession of a faculty' of mathematical intuition
/
conceived in analogy' with perception: if this were the source of our mathemat­
ical knowledge, the propositions of mathematics would be as contingent as
those of astronomy. The existence of a system of mathematical objects is like
the existence of God in this, that one may believe in it or disbelieve in it, but
one cannot intelligibly say that it exists but might not have done, or does not
exist but might have done. It differs from the existence of God in that God,
according to the theologians, is the most actual of all beings, whereas math­
ematical objects are non-actual. Hence, while the incoherence of the concep­
tion of God would show that he does not exist, the mere coherence of the
r conception is not enough to show that he does. The necessity of God s
existence derives, rather, from its being the condition for the existence ot
everything else, so that there is no prior condition of which we can say that,
“ « had not been fulfilled, God would not have existed. That is why, as
Quinas perceived, the necessity' of God’s existence does not entail that we
a? now it a priori. All that we know a priori is that, if God exists, then he
exists necessarily.
e ?y contrast, if we are able to know that a system ot mathematical objects
miph ’ iWe Can °nty know it a priori: it makes no sense to suppose that we
DoLk-r V k by some a posteriori means. It must therefore be from the
jp . r 1 lty our knowing its existence a priori that the necessity' of its existence
1Ves; and this entails that the coherence of the conception of the system is

307
 24 The Problem of Mathematical Objects 309
308 Frege: Philosophy ofMathematics
not for demarcating mathematics from logic, but for singling out the
sufficient, in the light of the context principle, to justify the assertion of its criterion,
tical part of logic, since the far less complex kinds of deductive
existence. mathema'
' that do not require mathematical expression need no reference to
It is on this, and not on the contention that the notion of class belongs to reasoning
-1 objects. It is difficult to maintain that any more convincing account of
logic, conceived as tire science of deductive inference, that Frege’s claim that l0g*Ca neral nature of mathematics has ever been given.
arithmetical truths are analytic ultimately rests. By no means all abstract objects ^^apparent from this account how misguided it is to criticise Frege for
exist of necessity: the Equator does not, for one. Mathematical objects, when
ducing one mathematical theory, arithmetic, to another, set theory'. He would
genuine, do, because the truth-conditions for statements about them have
^ve had no objection to considering the notion of class as a mathematical
been fixed in such a way that no condition for their existence needs to be ha'e but would not have seen that as in any way conflicting with characterising
fulfilled; that is why Frege felt entitled to call them logical objects. He did
it^s a logical one. His reasons for regarding it as a logical notion, namely that
attempt to make clear his criterion for applying the epithet ‘logical’, namely to
what governs every realm of reality and every degree of reality - the merely c|ass cannot be considered as a whole made up out of its members, but must
thinkable as well as what in fact exists. He was doubtless at fault, however, be explained as the extension of a concept, were indeed sound: given his
for failing to make clear what, in general, he conceived as belonging among assumption that every concept has an extension (and every' function a value­
the fundamental logical laws. He indeed claimed Axiom V of Grundgesetze as range), they were cogent. His initial attempt at avoiding Russell’s contradiction
1 being among those fundamental laws, but we, accustomed to think of the laws retained this assumption (allowing the abstraction operator still to be applied
of logic as restricted to those governing deductive inference, misunderstand to any expression for a first-level function of one argument), resorting to the
his ground for doing so; we are not helped to understand him aright by the desperate expedient of denying that, to have the same extension, concepts (and
fact that this particular ‘law’ is self-contradictory, or by his viciously circular functions) needed to be co-extensive. When he discovered the inadequacy of
attempt to justify it. this solution, he rejected the notion of a class (of the extension of a concept)
as altogether spurious; had he taken a less hostile view of it, he would still
t The nature of mathematics presumably have denied it to be logical in character, in having proved to lack
the required generality'. For the failure of his solution indicated the impossi­
Frege’s conception of what belongs to logic was indeed more generous than bility of retaining the assumption that every concept has an extension; with
that which is natural to us; but the foregoing way of drawing the contrast, this assumption gone, it looks unlikely that all answers to the question, ‘How-
between universal applicability and relevance to deductive inference, distorts many objects fall under the concept F?’, can be explained in terms of the
his view'. It is not that logical objects were for him irrelevant to deductive extension of the concept F. The natural assumption, which Frege is very likely
inference; it was merely that he did not expect their relevance to it to be to have made, is that that question has an answer whenever the concept F is
apparent outside mathematics. For him, the whole point of mathematics lay definite and (unlike the concept red) is defined over a determinate domain or
in its applications. A mathematical theorem, on his view, encapsulates an entire
has a criterion of identity associated with it. On this assumption, the notion
deductive subroutine - perhaps a very complex one - which, once discovered,
does not need to be gone through again explicitly on future occasions; but it of cardinality has sufficient generality to be recognised as logical in character;
Russell’s paradox had shown, contrary' to first impression, that that of the
expresses it, not as a principle of inference, but as a proposition to which we
have given sense by fixing its truth-conditions, and which may therefore be extension of a concept does not. It may7 indeed be replied that it is only ^en
considered on its oto account, without an eye on its possible applications. On the concept F has an extension (determines a set) that the question, ow
this new, therefore, that part of mathematics which is independent of intuition many objects fall under it?’, has an answer, so that the two notions ha\e e
simply comprises all the complex deductive reasoning of which we are capable, same generality; but, even after he had recognised the inadequacy of his
m/ged °f aU that would restrict its application to particular realms of reality. solution of the contradiction, Frege is unlikely to have attained that paradoxical
( e might qualify' this as all such reasoning as involves only completely definite conclusion (which, indeed, is rejected when it is said that a class is proper
concepts, Frege himself believed genuine reasoning with imperfectly definite p en it has the same cardinality as the universe). It is not, of course, a
concepts to be impossible.) Geometry apart, mathematics therefore simply is did not make a grave mistake: only that to characterise the mistake as
°^C, n.° d*sdnct’on ’n principle can be drawn. Most of the deductive reasoning at of reducing a simpler mathematical theory7 to a more comp exone: «
which it in this way encapsulates requires, for its formulation, reference to sconceive both his objective and the distance by which he e s
as act o jects - mathematical or logical objects; we might use this as a attaining it.
The important claim Frege made is that there exists a method of charac
 24 The Problem of Mathematical Objects 311
310 Frege: Philosophy ofMathematics
xtem of objects demands much more than its merely not involving a
ing a system of mathematical objects which sen es to confer senses upon the 3 S adiction. It requires that we should have a clear grasp of the range of
statements of that mathematical theory of which the system is a model in the C°? vidual objects that the system comprises, and of the constitutive relations
light of which the context principle guarantees that we do make genuine u nveen them, enabling us to recognise the truth of fundamental axioms
reference to those objects. The existence of that system is therefore a priori and verning the mathematical theory which describes that system. For the claim
independent of intuition, and the axioms of the theory may rank, accordingly, as fhat a coherent conception of the system suffices to ensure its existence is
analytic. Frege believed that he could, by introducing value-ranges, thereby based upon the context principle, applied not to a circular procedure like
introduce all logical objects that would be required in mathematics; and he Frege’s own, but to some legitimate means of fixing the senses of statements
had a quite erroneous idea of how to give a coherent and presuppositionless
characterisation of the system of such value-ranges. These mistakes do not concerningit.
The possession of such a conception of a mathematical system - of an
invalidate the general claim; if it can be sustained, we have a highly plausible
account of the character of mathematics in general. intuitive model for the theory' that relates to it - is without question essential
for us to have a mathematical theory' at all, rather than a mere piece of
formalism; and it is this which tempts us to speak of mathematical intuition.
The existence of mathematical objects The term would not be altogether inappropriate, were it not hard to resist the
The conception of mathematical intuition as analogous to sense-perception is pressure to interpret it as denoting something analogous to sense-perception;
open to an evident objection. A physical complex apprehended by the senses that pressure makes it, too, a dangerous piece of terminology’. The danger lies
may prove to have properties not immediately apparent, just as a mathematical in its creating the impression that the grasp of an intuitive model for a theory
system may prove to have properties not apparent from our initial grasp of it. is unmediated by language: that wre perceive its structure by a direct intellectual
But, whereas those of the physical system need in no way be implicit in our apprehension. If it were so, it would be useless, because it could not be
means of identifying it, those of the mathematical system must be; this would conveyed to others: only a solipsistic mathematics could result from alleged
/ not be true if mathematical intuition were analogous to sense-perception. If intuitions of this kind. In fact, we have no such powers: we frame intuitive
/ the continuum hypothesis, say, is determinately true, that can only be because models by means of concepts common to us all, and the models have no more
it follows from principles not yet formulated by us, but already inchoately content, and are no more definite, than the verbal or symbolic descriptions by
present in our intuitive conception of the intended model of set theory. If that means of which they may be communicated.
conception were a kind of blurred perception, on the other hand, it might be But can an intuitive conception of a mathematical system be sufficiently
that it could be filled out, with equal faithfulness to our present grasp of it, sharp as to be self-justifying, so that the mere possession of that conception
however implicit, both so as to verify and to falsify the continuum hypothesis, warrants the assertion that such a system exists? It is a belief that it can that
which nevertheless possessed a determinate truth-value according to the way leads to the talk, so dear to Dedekind, of mathematical objects as the free
things happened in fact to be. Since this supposition is manifestly absurd, this creations of the human mind; but there is no such thing as the human mind,
path to justifying the existence of mathematical objects, without appeal to the only individual minds. The metaphor is dangerously psychologistie, tempting
context principle, is closed.
r Can Frege’s thesis that it is possible to justify a priori the existence of a
us to scrutinise the internal operations of our minds. A conception ot a
mathematical system — an intuitive model — cannot transcend the means -
system of mathematical objects be sustained in the face of his own failure to
produce an acceptable vindication of it? The thesis amounts to a claim that necessarily linguistic and symbolic means - by which one person can convey
«to another; it exists only in so far as it can be described. Frege would insist
the fact that a given conception of a system of mathematical objects is coherent
is enough to warrant asserting the existence of that system; that it is in at a system so conceived existed independently of being conceived. Saying
effect self-justifying. This is not intended as an admission that mathematical at has its danger, too - that of suggesting that something more is needed
existence is after all to be equated with consistency. The theory of negative r its existence than our having a clear conception of it: it is only a step from
types (derived from the theory of simple types by allowing negative and positive con of mathematical reality as contingent - a matter of w c

I integers to serve as type-indices) is obviously consistent if the theory of simple

types is, since any proof can be reinterpreted in the latter theory; but that fact
does not of itself suffice to justify our believing in the existence of a system
be S tuents of it God has chosen to endow with existence. It would certai y
it do °ngtO Say ^at t^le system existed in advance of our conceiving it, because
reear? nOt exist in time at all; but if we say that we created it, we have to
of sets so stratified. In the intended sense, the coherence of a conception of tt as having come into existence, and as not having existed previous y
r

24 The Problem ofMathematical Objects 313

312 Frege: Philosophy ofMathematics
, ftOf-theoretic) sense. Merely? How does Field know, or why does he believe
A non-Fregean answer K to be consistent? Most people do, indeed; but then most people are not
4 minalists- If IF is consistent, then, being a first-order theory, it has a
The fact remains that it is extremely difficult to frame a clear description of denumerable model; but it is not from such a model that Field derives his
a mathematical system, as intuitively conceived, at least when it is fundamental belief in its consistency, since he has no reason to suppose it to exist save by
in being the source of general notions that vv e use in many other contexts, and
assuming the consistency of the theory. Our primordial reason for supposing
particularly when it is from them that our notion of a particular infinite
ZF to be consistent lies in our belief that we have an intuitive model for it,
cardinality is derived. Attempts to do so, at least for systems of cardinality
greater than the natural numbers, always have a certain cloudiness, and leave the cumulative hierarchy in which the sets of rank a + 1 comprise ‘all’ sets
some quite unconvinced that any sharp conception is being conveyed, while of elements of rank a (together with the elements of rank a, it is necessary to
satisfying others. Furthermore, it is notoriously hard to resolve such disagree­ add when we start with Urelemente). The constructible hierarchy yields a
ments over whether or not a given conception of such a system is so much as more restrained model; but, considered as an intuitive model, it requires that
coherent, let alone sharp, or even to see by what means they could be resolved. we have a grasp of the totality of ordinal numbers less than the first strongly
Why is this? Frege can give us no help at this point: in so much as discussing inaccessible one. Such a model is, from an ordinary' standpoint (not that of
the matter, we have had to leave him behind. We are trying to solve the traffickers in large cardinals), of enormously high cardinality: it is to the field
problem he failed to solve, in his spirit but in a different way; and the of real numbers as a skyscraper to a two-storey farmhouse. If we have a
attempt has simply led us into the presence of a range of familiar philosophical conception of such a structure, why should we jib at the system of real
disagreements which more resemble differences of taste than divergent rational numbers? If ever there were a case of a pointless reduction of (the conservative­
conclusions. Discovering the correct way out of this impasse is not relevant to ness of) a mathematical theory to (the consistency of) a more complex one, it
an exposition of Frege’s work, only to evaluating it. is to be found here, and not in Frege’s work.
Logicism, as represented first by Frege and then by Russell and Whitehead, Field indeed offers a reason for believing ZF to be consistent, namely
F failed because it combined three incompatible aims: to keep mathematics that ‘if it weren’t consistent someone would have probably discovered an
uncontaminated by empirical notions; to represent it as a science, that is, as inconsistency in it by now’.2 He refers to this as inductive knowledge.3 To
a body of truths, and not a mere auxiliary of other sciences; and to justify have an inductive basis for the conviction, however, it is not enough to observe
classical mathematics in its entirety. There are still those who wish to abandon that some theories have been discovered to be inconsistent in a relatively short
the first of these three aims, and revert to an empiricist conception of math­ time; it would be necessary' also to know, of some theories not discovered to be
ematics as a natural science like any other; but Frege’s arguments against such inconsistent within around three-quarters of a century, that they are consistent.
a conception were surely conclusive. Field recommends abandoning the second Without non-inductive knowledge of the consistency7 of some comparable
aim: mathematics, for him, is the servant, not the queen, of the sciences, and mathematical theories there can be no inductive knowledge of the consistency
should refrain from giving itself airs. His strategy for proving the conservative­ of any mathematical theory. Since Field claims no non-inductive knowledge
ness of a mathematical theory S over a physical theory T, formulated nominal­ of the consistency of any theory, he can have no knowledge of consistency
consistency7
istically, is first to prove a theorem that a model of T can be constructed in at all.
S, and then to construct a model of S in (an adaptation of) Zermelo-Fraenkel If the problem of mathematical objects is not to be solved by abandoning
set theory' ZF. The final step is to prove that, if ZF is consistent, so is cither of the first two aims, perhaps we need to abandon the third; and, in
ZF + T. Now if S is a second-order theory, we need the second-order version
Particular, the assumption, in which Frege had an unswerving faith, that, given
of ZF, which we must assume to be ‘semantically consistent’, i.e. to have a a^y domain of mathematical objects, quantification over it can be interpreted
model, we obtain conservativeness with respect to model-theoretic conse­
j assicaUy, so that statements formed by means of such quantification will be
quences. Field hopes, however, that first-order formulations of physical theor­ eterminately either true or false, and hence obey classical logic. His faith in
ies will be sufficient for the purposes of physics, and first-order versions of
is assumption constitutes his sole blindness to the fundamental problems of
mathematical theories sufficient for applications to physics. In this case, we
need consider only the usual first-order version of ZF, and shall obtain con- ® Philosophy of mathematics: he had at least the excuse that, when he was
‘r I?a’3'eness with respect to proof-theoretic consequences, a result which 2? his major works, no one had yet raised the question.
o ows merely from the consistency of ZF’,1 i.e. its consistency in the ordinary z ntum°nists deny the assumption for quantification over any infinite totality,
1 H. Field, Science without Numbers, Oxford, 1980, p. 19. 3 IbicL' pd’gg ea^'sm< Mathematics and Modality, Oxford, 1989, p. 232.
r

24 The Problem ofMathematical Objects 315

31 4 Frege: Philosophy ofMathematics
. t,rminate criteria of application and of identity; he saw no need for any prior
on the ground that it is impossible to complete an infinite process. Indepen­ d ascription of the domain. Precisely that is what we now take for granted
dently of any such general doubt, however, the assumption demands that the reunited. We know well enough what is needed for something to be
conception of the domain be completely definite: any haziness about what as rcH - -- ——i— ...
elements it does or does not contain must obviously vitiate the assumption. recognised as a f same set or ord.nal number as one given in an0^,fr, b " a
» the
ertain way is the same set or ordinal number as one given in another: but
Provided that there are some clear general principles concerning the condition
not think of that as allowing us to form statements qu3ntifrinz
for membership of the totality, and some means of identifying individual certainly do; or all ordinal numbers and to treat them as having determiiZ
elements with an indisputable claim to belong to it, both universal and existen­ over all sets
tial quantification over a hazily circumscribed totality' can have an intelligible truth-values.. In the mathematical realm, reality cannot be left to blow all
sense, in that they sometimes yield statements recognisably true or recognisably haziness away: we have to remove it ourselves by contriving adequate means
false. Such quantification cannot, however, be construed as invariably yielding of laying down just what elements the domain is to comprise.
statements with determinate truth-values. Statements involving it must be This does not apply only to concepts like set and ordinal number for which
regarded as making claims which their authors make justifiably if they are contradiction results from treating their extensions as forming determinate
capable of vindicating them. If the claim embodied in such a statement can domains of quantification, but to all means of specifying such a domain: the
be vindicated, the statement may be regarded as tru>a; if the claim is shown requirement of a prior specification of the domain, when interpreting a theory',
impossible to vindicate, the statement may be taken to be false; but, if neither formalised or unformalised, is general. The criterion of application of the
vindication nor refutation is forthcoming, it cannot be presumed to be either. concept real number, for example, might be said to be that whatever has a
A realist view of the external world involves assuming that universal or determinate relation of magnitude to any given rational is a real number, and
I . existential generalisation over a totality' given by an empirical concept, such as the criterion of identity that, if the real number x is greater than or less than

r
I
the concept of a star, does yield statements determinately true or false, provided
that the concept is definite. Truth-conditions differ from conditions for the
justifiability of a claim in that they obtain independently of the speaker’s
a rational number p if and only if the real number y is, respectively, greater
than or less than p, then x = y. This is quite adequate to explain what is
required of a specified mathematical entity' for us to recognise it as a real
knowledge or his capacity to perform some task (unless of course they are the number; but it does not suffice as a means of circumscribing a domain of
conditions for the truth of some statement about such matters). A concept is quantification, when such quantification is to yield statements with determinate
definite provided that it has a definite criterion of application - it is determinate truth-values. It does not do so, because it fails to determine the limits of
what has to hold good of an object for it to fall under the concept - and a acceptable specification of something to be acknowledged as a real number:
definite criterion of identity - it is determinate w'hat is to count as one and we still need a means of saying which real numbers the domain comprises.
the same such object. On a realist view, we do not need, in the empirical case,
The principal consequence of the set-theoretic paradoxes was that even
to be able to circumscribe the extension of the concept more closely in order platonists were compelled to allow that there are mathematical concepts whose
to be assured that generalisation with respect to it will yield statements with extensions form hazy totalities: the concept of an ordinal number, for example,
determinate truth-values, independent of the speaker’s warrant for making
them. We do not need to be able to say just what objects there are which fall n this regard, Cantor saw much more clearly than Frege: but even he was in
r
under the given concept: provided the concept is definite, reality will of itself error in regarding the distinction between consistent and inconsistent totalities
determine the truth or falsity of such statements. On this view, reality dispels as an absolute one. So to regard it is to provoke intolerable perplexity'. Consider
all haziness: we need do nothing further to eliminate it. at happens when someone is first introduced to the conception of transfimte
Frege was a resolute realist about mathematics, as about the external world; wT k numbers- A certain’ resistance has first to be overcome: to someone
but even he did not argue that mathematical reality will determine the truth­ that k&S been used to finite cardinals, and only to them, it seems obvious
values of mathematical statements, without any need for us to circumscribe at b ere Can On^ be finite cardinals. A cardinal number, for him, is arrived
e domain of quantification or to specify what objects belonged to it. He did imn^ C°untinSi and the very' definition of an infinite totality' is that it is
not argue in this way: but he may be accused of having in effect treated an a SSlb e t0 C0unt iL This is not a stuPid prejudice. The scholastics favoured
ma ematical concepts, in this regard, as analogous to empirical ones. For his the ^Ument t0 sb°w that the human race could not always have existed, on
use o e context principle to justify assuming that the domain of the individual of all tk^L tad, there would be no number that would be the num er
varia i,eS comPr’ses cardinal numbers, or value-ranges, required no more than must k e human bein?s there had ever been, whereas for every' concept there
at e re evant concept, of a cardinal number or of a value-range, have die nrpC a number which is that of the objects falling under it. All the same,
u ice is one that can be overcome: the beginner can be persua e a
 316 Frege: Philosophy of Mathematics 24 The Problem ofMathematical Objects 317
it makes sense, after all, to speak of the number of natural numbers. Once his -tension is to describe it as having an increasing sequence of extensions-
initial prejudice has been overcome, the next stage is to convince the beginner Mt eX1
hazy is hazy is the length of the sequence, which vanishes in the indiscernible
that there are distinct cardinal numbers: not all infinite totalities have as many
wl-
/tance The intuitive concept of ordinal number, like those of cardinal
members as each other. When he has become accustomed to this idea, he is mber and of set, is an indefinitely extensible one.5 Certain objects must be
extremelv likely to ask, ‘How many transfinite cardinals are there?’. How "“cognised outright as falling under such a concept: but what distinguishes it
should he be answered? He is very likely to be answered by being told, ‘You from all definite concepts is the principle of extendibility governing it. Russell’s
must not ask that question’. But why should he not? If it was, after all, all concept of a class not containing itself as a member is a prototypical example
right to ask, ‘How many numbers are there?, in the sense in which number’ of an indefinitely extensible concept: for, once we form a definite conception
meant ‘finite cardinal’, how can it be wrong to ask the same question when of a totality W of such classes, it is evident that W cannot, on pain of
‘number’ means ‘finite or transfinite cardinal’? A mere prohibition leaves the contradiction, be a member of itself, and thus the totality- consisting of all the
matter a mystery. It gives no help to say that there are some totalities so large members of W, together with W itself, is a more extensive totality than W of
that no number can be assigned to them. We can gain some grasp on the idea classes that are not members of themselves.
of a totality too big to be counted, even at the stage when we think that, if it The principle of extendibility constitutive of an indefinitely extensible con­
cannot be counted, it does not have a number; but, once we have accepted cept is independent of how lax or rigorous the requirement for having a
that totalities too big to be counted may yet have numbers, the idea of one definite conception of a totality is taken to be, although that will of course
too big even to have a number conveys nothing at all. And merely to say, ‘If affect which concepts are acknowledged to be indefinitely extensible. It is clear
you persist in talking about the number of all cardinal numbers, you will run that Frege’s error did not lie in considering the notion of the extension of a
into contradiction’, is to wield the big stick, not to offer an explanation. concept to be a logical one, for that it plainly is. Nor did it lie in his supposing
What the paradoxes revealed was not the existence of concepts with incon­ every definite concept to have an extension, since it must be allowed that even­
sistent extensions, but of what may be called indefinitely extensible concepts. concept defined over a definite totality determines a definite subtotality. We
The concept of an ordinal number is a prototypical example. The Burali-
may say that his mistake lay in supposing there to be a totality- containing the
Forti paradox ensures that no definite totality comprises everything intuitively
extension of every concept defined over it; more generally, it lay in his not
recognisable as an ordinal number, where a definite totality is one quantifi­
having the glimmering of a suspicion of the existence of indefinitely extensible
cation over which always yields a statement determinately true or false. For a
totality to be definite in this sense, we must have a clear grasp of what it concepts.
One reason why the philosophy of mathematics appears at present to be
comprises: but, if we have a clear grasp of any totality of ordinals, we thereby
have a conception of w’hat is intuitively an ordinal number greater than any becalmed is that we do not know how to accomplish the task at which Frege
member of that totality. Any definite totality of ordinals must therefore be so so lamentably failed, namely to characterise the domains of the fundamental
circumscribed as to forswear comprehensiveness, renouncing any claim to mathematical theories so as to convey what everyone, without preconceptions,
cover all that we might intuitively recognise as being an ordinal. It does not " ill acknowledge as a definite conception of the totality- in question: those who
follow that quantification over the intuitive totality of all ordinals is unintelli­ believe themselves already to have a firm grasp of such a totality are satisfied
gible. A universally quantified statement that would be true in any definite yith the available characterisations, while those who are sceptical of claims to
totality of ordinals must be admitted as true of all ordinals whatever, and there ave such a grasp reject them as question-begging or unacceptably vague. An !
is a plethora of such statements, beginning with ‘Every ordinal has a successor’. •mpasse is thus reached, and the choice degenerates into one between an act
Equally, any statement asserting the existence of an ordinal can be understood, ° aith and an avowal of disbelief, or even between expressions of divergent
wi out prior circumscription of the domain of quantification, as vindicated by tastes. Moreover, the impasse seems intrinsically impossible of resolution; for
the specification of an instance, no matter how large. Yet to suppose all undamental mathematical theories, such as the theory of natural numbers or
qnanh e statements of this kind to have a determinate truth-value would eory real numbers, are precisely those from which we initially em e
lead directly to contradiction by the route indicated by Burali-Forti.4
better than describing the intuitive concept of ordinal number as having a
iR j London Mathematical Society, series 2, vol. 4, 1906, pp.
if !»■ »preserve us from contradiction
quantifying over a full? □ assun?ptlons as before; but, when we do not conceive ourselves to c
quantuymg over a fully determ.nate totality, we shall have no motive to do so.
318 Frege: Philosophy ofMathematics 24 The Problem of Mathematical Objects 319
our conceptions of different infinite cardinalities, and hence no characterisation we have only a conception of finite totalities to go on. Admittedly, the lame
of their domains could in principle escape the accusation ot circularity. characterisation of the totality which supplies our usual ground for supposing
Now what is it for a totality to be infinite? More exactly, what is it for it to that we do have a definite conception of it does not always appeal to the notion
be intrinsically infinite, that is, for die very conception of that totality- to entail of completing the process of extension. The standard characterisation of the
its infinity? It is for us always to have a means of finding anodier element of totality' of natural numbers, as consisting of eterything attainable from 0 by
the totality, however many we have already identified; the new element will be reiterating the successor operation, does have this form; but the characteris­
characterised in terms of those previously identified. For a non-denumerable ation of the real numbers as comprising those corresponding to all cuts in the
totality like the real numbers, Cantor’s diagonal construction provides just rational line does not. The question is, however, whether there is any way of
such a means, tfiven any denumerable set of elements. A denumerable totality-, achieving a more precise characterisation of the highly unsurveyable totality of
likewise, is one for which we can find a further element, given any initial alt such cuts; it is only by appeal to a principle of extension that we convince
segment of it: the similarity' between Frege’s proof of the infinity of the ourselves that this cannot be done by any method of enumerating them.
sequence of natural numbers and the foregoing demonstration that the concept The requirements for characterising an indefinitely extensible concept are
class not a member of itself is indefinitely extensible can hardly escape notice. far less exigent than those for giving a description of a definite totality as one
We have a strong comiction that we do have a clear grasp of the totality of of which we have a clear grasp. A criterion of application (and a criterion of
natural numbers; but what we actually grasp with such clarity is the principle identity) are indeed required: it must be stated what, in general, is demanded
of extension by which, given any natural number, we can immediately cite one of something for us to recognise it as falling under the concept. As alreadv "
greater than it by 1. A concept whose extension is intrinsically infinite is thus noted, this asks for much less than a precise circumscription of a totality’;
a particular case of an indefinitely extensible one. Assuming its extension to wre choose to explain the concept real number in a Dedekindian manner (pre
constitute a definite totality - one of which we can form a sharp conception ably not the best choice) by saying that a real number is required to hi 4
and which forms a determinate domain of quantification - may not lead to determinate relations of magnitude to rationals, we say nothing about t
inconsistency; but it necessarily leads to our supposing that wre have provided manner in -which an object having such relations is to be specified, but simp
definite truth-conditions, independently of w-hether or not wre can prove them, leave any purported specification to be judged on its merits when it is offeree
for statements that cannot legitimately be so interpreted. The hypothesis that The concept requires a base of objects satisfying the criterion of application
the domains of the fundamental mathematical theories are given by what are and unquestionably well specified, and a principle of extendibility. The former
in fact indefinitely extensible concepts explains why we are at such a loss to is easily provided; the latter will be stated in terms of a definite totality of
supply uncontentious characterisations of their domains. objects falling under the concept, where it is again left to be judged, in any-
It springs to the lips to retort that the argument begs the question: it proposed case, whether we have such a definite totality’ or not. When the
depends, in the one case, on identifying totalities of w-hich we can form a concepts of natural number and of real number are regarded as indefinitely
definite conception with denumerable ones, and, in the other case, on identify­ extensible ones, our grasp of them is beyond question; it is only w-hen they
ing them with finite ones. It is, however, this reply that begs the question. masquerade as definite concepts that any’ attempt to characterise them beconr
These totalities are those from which w’e derive, respectively, our conception vague or circular.
of one of the cardinality of the continuum and our conception of an infinite This diagnosis breaks the impasse; but, of course, at a price. Quantificar
one. until we have a conception of the real numbers (or of the set of all sets over the objects falling under an indefinitely extensible concept obvio
of natural numbers), we have only a conception of denumerable totalities to does not yield statements with determinate truth-conditions, but only c
go on; and, until we have the conception of the totality of all natural numbers, embodying a claim to be able to cite an instance or an effective operation;
the logic governing such statements is not classical, but intuitionistic. Ado
of such a solution therefore entails a revision of mathematical pract
and a = h LJ” an enumera,ion si» “i, a}, ... of real numbers in an interval [a, b], with at - » accordance with constructivist principles. Such a tension w'ould hate
nda2 b, we can form sequences a'11, a121,... and b"\ V1' where «(’*” is the first element abhorrent to Frege: it is unclear that it would be less of a betrayal <
If SseTeouences te interYal.(«W’ and is the first’ element in the interval (««+,\ W-
the enumeration can l>dl,eir ast terms determine an interval within which no element o fundamental principles of his philosophy of mathematics than his own ev
Ko?s first cXt occr
SSof either detem,ine
’t e"eyumeration
of th? ; As Hallett an intervai
suchnotes, °r aisnumber
this proof their
similar <in c°mn,onto
principle
expedient of reducing arithmetic to geometry’.
ordinals). non-denumerability of the second number class (of the denumerable
320 Frege.' Philosophy oj Alathematics 24 The Problem ofMathematical Objects 321
Frege’s contribution to the philosophy of mathematics would have already to be embodied in that scientific theory. The question
w'ould then arise whether a version of it divested of that classical force (and
Frege’s attempt to justify the existence of mathematical objects was not simply thus of realistic metaphysical assumptions) would not be scientifically prefer­
a failure that left us where we were before: it left us with a precise range of able. These questions have scarcely been raised, let alone answered, by either
options. We cannot simply ignore the problem, but must choose betw een them. mathematicians, philosophers or physicists.
If we set aside intuition either of our mental creations or of the abstract realm, These speculations have taken us very far from Frege’s work. His failure to
there are only three. W e can maintain that we do have intuitive conceptions make any enquiry into the validity of classical logic, as applied to mathematical
of the real numbers, of Cantor’s second number-class, and perhaps even of a theories, is the one big lacuna - as opposed to the big error - in his philosophy
model for Zermelo-Fraenkel set theoiy, sufficiently determinate to confer of mathematics. It is one for which he can hardly be blamed. He can probably
senses on the propositions of the relevant theories which will warrant applying be reproached for his increasing inability to see through the errors and con­
to them the principle of bivalence. This heroic stance will validate an invocation fusions in others’ expositions of their ideas to the merit of those ideas them­
of the context principle just as Frege intended; but it is far from compelling. selves; considering the disappointments that disfigured his entire life, we can
It is futile simply to claim to have an intuition; it must be capable of being only regret, not blame. He left behind him a philosophy of arithmetic which
conveyed to others by being expressed in language or symbolism. No one he himself believed, for the last two decades of his life, to have been a total
denies that attempts to convey such intuitive models succeed in expressing failure, the only valuable part of his work, in his eyes, having been in formal
something; but the claim that they convey a conception of a domain of quanti­ and philosophical logic. That philosophy of arithmetic w'as, indeed, fatally
fication sufficiently definite to warrant attributing to statements involving flawed; but it had an incontestable clarity, so that, even wffiere it was mistaken,
quantification determinate truth-values is, to most, quite unconvincing. Alter­ it pointed very precisely to where the problems lay. But it did much more than
natively, we can side with the constructivists in admitting mathematical objects that. Frege’s polemic against formalism contained a definitive refutation of
without claiming to be able to circumscribe precisely in advance which such that deadening philosophical interpretation of mathematics. To important
objects are to be recognised; propositions concerning them must then be questions in the philosophy of mathematics, above all those concerning the
construed as obeying intuitionistic, not classical, logic. And, finally, we can application of mathematics, the fruitfulness of deductive reasoning and the
join with the nominalists in thinking that mathematics can dispense with objects nature of mathematical necessity, his work provided, if not full-dress answers,
altogether. The attempt actually to dispense with them within mathematics at least sketches of what must be the correct answers; later philosophers have
would involve a more far-reaching transformation of the subject as currently come nowhere near his partial success in answering those questions, and have
practised than a constructivist revolution. If a demonstration that dispensing frequently failed even to address them. Above all, Frege provided the most
with them would be in principle possible whenever mathematics was applied plausible general answer yet proposed to the fundamental question, ‘What is
within an extra-mathematical theory were capable of being given only piece­ mathematics?’, even if his answer cannot yet be unarguably vindicated. For all
meal, theoryr by theory', mathematics would lose its generality and its autonomy. his mistakes and omissions, he was the greatest philosopher of mathematics
If, for each mathematical theory, such a demonstration could be given in yet to have written.
advance for all physical theories satisfying certain general conditions, the
question would arise on what grounds this was preferable to the second,
constructivist, option. Investigation might reveal that a constructivist version
of a given mathematical theory was perfectly adequate for the applications
made of it within natural science. If so, then, for anyone who agrees with
Frege that it is applicability alone that raises mathematics from the rank of a
game to that of a science, a constructivist reformulation of the mathematical
theory would clearly be preferable to an indirect justification in terms of the
property of conservativeness. If, on the other hand, it proved that the classical
version of the mathematical theory had a substantial effect upon the scientific
theoiy, the question w'ould not yet be settled: for, on the hypothesis that
everything derivable by aid of the classical theory could in principle be derived
from the scientific theory alone, the classical force of the mathematical theory
 Bibliographical Note Index of Frege's Writings

The following books and articles, though not mentioned in the text, will be found of ‘Antwort auf die Ferienplauderei des §§ 18-83 48
Herm Thomae’ (1906) 82n. § 19 64, 73, 224n., 242n., 261n.
relevance: ‘Ausfiihrungen iiber Sinn und §20 73
Alberto Coffa, ‘Kant, Bolzano and the Emergence of Logicism’, The Journal of Philo­ Bedeutung’ (1892-5) 92n. §21 73,74
Begriffsschrifi (1879) 2, 8, 11-12, 26, 39, §§21-8 74
sophy, vol. 79, 1982, pp. 679-89. §§ 21-54 73
Bob Hale, Abstract Objects, Oxford, 1987. 41, 42, 68-9, 120, 122, 173, 175
‘Booles rechnende Logik und die § 22 75,76
Philip Kitcher, ‘Frege, Dedekind and the Philosophy of Mathematics’, in L. Haaparanta §§ 22-5 74
& J. Hintikka (eds.), Frege Synthesized, Dordrecht, 1986, pp. 299-343. Begriffsschrift’ (1881) 3, 38-9,
AAV. Moore and Andrew Rein, 'Grundgesetze, Section 10’, in Frege Synthesized, 65 §§ 22-8 74
pp. 375-84, and ‘Frege’s Permutation Argument’, Notre Dame Journal ofFormal Logic, Die Grundlagen der Arithmetik (1884) xi, §23 75 “
vol. 28, 1987, pp. 51-4. 1-4, 7, 8, 11, 12, 18, 111, 121, §24 43, 74
Michael Resnik, Frege and the Philosophy ofMathematics, Ithaca, New York, and London, 122, 159, 166, 176, 180, 181, §25 76
1986, and ‘Frege’s Proof of Referentiality’, in Frege Synthesized, pp. 177—95. 183, 192-8, 205-7, 297 §26 77, 78-9, 80-1
Peter Schroeder-Heister, ‘A Model-theoretic Reconstruction of Frege’s Permutation Introduction 10-22, 118, 180-1 §27 65, 66, 77, 81
Argument’, Notre Dame Journal ofFormal Logic, vol. 28, 1987, pp. 69-79. §§1-2 11 §28 82
Christian Thiel, ‘V.ahrheitswert und Wertverlauf: zu Freges Argumentation im § 10 §2
. ’ 47~ §29 83
der “Grundgesetze der Arithmetik”’, in M. Schim (ed.), Studien zu Frege/Essays on § 3 3, 23-5, 32, 57, 126, 129 §§ 29-33 82
Frege, vol. I, Stuttgart, 1976, pp. 287-99. §4 31, 48 §§29-44 50, 73, 82
§5 55-6, 67 § 30 83
Published, or seen by me, too late for me to comment on in this book were Franz von §§5-17 47,55 §31 83
Kutschera, Gottlob Frege, Berlin and New York, 1989, and Mary Tiles, Mathematics and §§5-44 55 §32 83
the Image of Reason, London and New York, 1991. §§ 5-54 242 § 33 83
§ 6 56-7 §§33-44 20, 144
The best German-language edition of the Grundlagen is that edited and annotated
by Christian Thiel, Hamburg, 1986. §7 58 §34 40,‘ 83-4
§§7-11 58 §35 86
§8 59 §38 82, 87
§ 9 59,256,257 §39 86
§ 10 61-3, 72 §40 86-7
§12 63, 66-7, 69 §41 87
§13 69 §44 87
§ 14 45, 69-70 §45 84
§15 70 §46 73, 74, 81, 87-8, 89, 91, 93,
193
§ 16 59, 70
§17 36, 59, 70-1 §47 73, 88, 93, 193
§48 84, 88-9
§ 18 71, 72
§§ 18-28 72 §51 93
§§ 18-44 48
§53 65, 91

322 323
324 Frege: Philosophy ofMathematics Index of Frege’s Writings 325
§ 54 94, 170 §87 3 § 30 210, 215n., 217 § 66 125-6,244
§ 55 100,102,103,107,111,120, § 88 23, 36, 37, 41 §31 159,210,214,215,216,218, §§ 68-9 263
131, 146, 190, 225 § 89 63, 65 219 §§ 68-85 263
§§ 55-61 102,111,115 § 90 70 § 32 184,210,237-8 §70 264,268
§ 56 101, 102, 105-6, 108, 131 §§90-1 12 §37 122n. §§ 70-6 263,268,270
§§ 56-7 108, 109, 111, 131, 133 § 91 36, 69 § 38 8n., 122n. §71 268
‘ -
§§ 56-61 101 §§ 92-103 87 §40 122n. §72 241n., 268-9
§ 57 102, 108,115,121,133,170 §§ 92-104 247 §§ 40-6 8n. §74 225
§§ 58-61 109-10 §96 197 §41 122n.
- §75 263, 273, 274
§ 60 110,117-18,119,125,126, §97 248 §42 121n., 122n. §76 265, 276
180, 189 §100 178 §43 122n. §77 263, 265
§61 110,111 § 101 178 §44 123n. §78 266
§ 62 3-4,95,111,112,117,118, § 102 178, 247 §45 122n. §§ 79-81 265
155, 159, 165, 181, 183, 223-4 § 103 242n. §46 121n., 122n. §81 266
§§62-7 125 § 104 248, 26In. §54 8n., 287n. §82 266
§§ 62-9 111,117,119,121,167, § 106 119, 181, 200 §65 123n. §83 266, 267
180 §§106-7 119 §69 123n. §84 266
§ 63 112,114,119,127-8,142 §§106-8 119 §95 13, 123n. §85 263
§§ 63-5 117,188,189,204 §§ 106-9 181 §97 123n.
§§ 63-7 123, 134, 155, 187 §98 123n. §§ 86-137 252
‘Einleitung in die Logik’ (1906) 5 § 87 253, 258
§§ 63-8, 162 ‘Erkenntnisquellen der Mathematik und §101 123n.
§§ 63-9 40, 181 §103 13, 123n. § 88 253,258
der mathematischen §§ 89-92 255,257
-5 64 32-3, 115, 116, 125, 128, 142, Naturwissenschaften’ §105 123n.
168 § 107 123n. § 91 60,256,259-60
(1924-1925)44, 186n., 296 § 92 258,259
4-7 116 Function und Begriff (1891) 2, 7, 170-1 § 113 124n.
>4-8 117 § 119 124n. § 93 253
Crundgesetze der Arithmetik (1893,
113,114,115,128,141,142, § 121 124n. §§ 93-119 253
1903) xi, 4-6, 26, 71, 92, 121, §§ 121-36 252
180 122, 159, 160, 168, 176, 180, theorem 32 123n.
§ 66 126, 157 193, 195-6, 197-8, 205-6, 207 theorem 49 123n. § 124 252
§§ 66-7 155, 188 Part I 4, 7, 158, 241, 297 theorem 89 13 § 131 252
§§66-9 117 Part II 1, 4, 7, 8, 121, 241-2 theorem 90 123n. §137 60, 257-8
§67 160 Part III 1, 4, 7-8, 11, 59, 60-1, 73, theorem 94 123n. §138 249
§ 68 33, 112, 116, 120, 159, 166 241-51, 296 theorem 97 123n. § 139 249
§69 177 Vol. I 187,241 §140 248
theorem 107 123n.
§ 70 39, 65, 89, 90, 146 §§140—7 247
§§70-2 119,142
Preface 7n., 8n., 47-8,49,67,81,181 theorem 108 13, 123n.
Introduction 3, 7n., 8n., 32, 70 § 144 290
§§71-2 143 theorem 110 123n.
§3
""210 § 145 248
§71-83 131 theorem 113 123n.
§5 211 §§ 146-7 249
§§72 119,120 theorem 117 123n.
§6 211 § 149 243n.
§73 123 theorem 122 123n.
§7 211 theorem 145 124n.
§ 153 241n.
§74 74, 120 § 156 261
§75 123 §8 134, 205 theorem 155 124n.
§9 8n., 210 §§ 156-9 242, 261
§76 120 theorem 157 124n. § 157 64, 246,261
§77 120 § 10 159, 210, 212, 216, 303n. Vol. II 241
§11 158 §§33-44 47 § 158 261
§78 13, 123, 132 § 159 262,283
§79 12, 120, 124 §20 210, 218 §§ 55-65 244
§21 7n. §§ 55-67 241 n. § 160 279
§80 69 §§ 160-4 242, 261, 277
§81 120 §22 139 §57 244
§25 217-18 §58 241n., 244 §161 277
§§ 82-3 124
§29 210, 214, 215 §60 245 § 162 278
§ 83 12, 120, 123, 132 § 163 280
§§29-31 214 §61 245
§65 241n. § 164 241, 262, 283-4, 285, 290
326 Frege: Philosophy of Mathematics
§175 286, 288 Grossenbegrifies griinden’
§213 289 (1874) 63, 68, 278n., 279
§216 288, 289 review of Cantor (1892) 293n,
§217 288 review of Cohen, Das Prinzip der
§218 289 In/initesimal-Methode und seine
§244 290 Gesehichte (1885) 125, 194
§245 241, 290 review of Husserl, Philosophic der
Appendix 4, 6, 103, 169, 171, 241, Arithmetik (1894) 19-20, 31-2
290 85, 86, 92n„ 95
theorem 635 289 review of Schroder, Ibrlesungen iiber die
theorem 637 289, 290
theorem 641 289
Algebra der Logik (1895) 9In.,
92n. General Index
theorem 689 290 ‘Siebzehn Kemsatze zur Logik’
‘Kurze Ubersicht meiner logischen (c. 1876?) 183
Lehren' (1906) 151n. ‘Uber Begriff und Gegenstand’ (1892) 7,
letter to Hbnigswald (1925) 188n. 8, 35, 90, 98n., 186 a priori/a posteriori 3, 7, 23-8, 58, 255, Axiom V of Grundgesetze 4-6, 170-1,
letter to Anton Marty (1882) 90 ‘Uber den Begriff der Zahl’ 292, 307-8, 310 176, 210-11, 213,219, 226, 233m,
letter to Peano (1896) 241n. (1891-2) 82n. abstract objects 181—2, 191, 223-40, 308
‘Uber die Grundlagen der Geometric’ 297-304 Axiom VI of Grundgesetze 158
letters to Russell (1902—4) 93n., 159,
216, 270n. (1903,1906)5 dependent 239, 300 axioms 305
Uber die Zahlen des Herm H. Schubert abstraction Ayers, Michael 113
letter to Zsigmondy (1918 or 1919) 6
‘Logik’ (1880s) 25,' 49-50 (1899) 82n., 241n. logical 167-8, 176, 180, 225, 268, 291
‘Uber formale Theorien der Arithmetik’ psychological 20, 40, 50—2, 83—5, Baker, Gordon 67, 84, 85, 89
‘Logik in der Mathematik’ (1914) 33, Bartlett, James 217
150-1, 153, 243n. (1886) 8, 43, 94, 295 143-6, 167-8, 250
‘Uber Schoenflies’ (1906) 5 abstraction operator 121, 133, 159, 170, Baumann, JJ. 83
‘Logische Untersuchungen’ (1918, Bell, David 19-21, 95-8
1923) 5, 6 ‘Uber Sinn und Bedeutung’ (1892) 7, 32 188-9, 205, 210, 213, 216, 217,
‘Was ist eine Function?’ (1904) 5, 81, 219, 221, 223, 226, 309 Benac, Theodore J. 88n., 260n.
‘Der Gedanke’ (1918) 78, 225, 239 241n. actual (n?!rWif/z)/non-actual 80-1, 133, Benacerraf, Paulxii, 52-4, 179, 296
‘Rechnungsmethoden, die sich auf eine ‘Was kann ich als Ergebnis meiner Arbeit 166, 181-3, 187, 204, 207-8, 209, Berkeley, George 76
Erweiterung des ansehen?’ (1906) 5 225, 239, 297, 307 Bolzano, Bernard 28-30, 47, 65-6, 67-8,
Adeleke, Samson viiin., 242n., 288n. 70, 150-1, 154, 223
adjectival versus substantival uses of Boolean algebra 295
number-words 73, 99-102, 108-9, Bostock, David xii
119, 121, 143, 146 Brouwer, Luitzen Egbertus Jan 77, 292
Aleph-Null 7, 95-6, 133, 227, 293 Burali-Forti paradox 316
analytic/synthetic 3, 12, 23-46, 58, 63,
68-70, 298-9, 302-3, 308, 310 Cantor, Georg xii, 8, 50, 51, 61, 82, 142,
analytic judgements extend our 229n, 243, 246, 247, 250-1, 252,
knowledge 41-2, 151 258, 262, 263-76, 280, 284, 293,
Anscombe, Elizabeth 294n. 315, 318, 320
application of mathematics 7, 48, 59-61, cardinal equivalence 114-15, 119-20,
252, 255-61, 292-301, 312-13, 128-31, 143-52, 176
320-1 cardinal numbers 63-4, 73, 130, 133,
application operator 217 139, 155-66, 167-8, 172, 176, 179,
Aquinas, St. Thomas 30, 307 187, 190, 199, 206, 226, 246, 261,
archimedean law 282, 289-90 303, 314-16, 317
Aristotle 112 transfinite 241, 316
associative law for addition 56, 58 cardinality operator 113-15, 116-19,
Austin, John 23, 48, 70, 88 121, 122-3, 125-30, 133, 134,
axiom of choice 149 137-9, 143, 145-8, 155-61, 166,
i 167, 170, 176-7, 180-1, 187-8,
axiom of infinity 132, 302
axiom of reducibility 302 200-2, 204-6, 226
327
328 Frege: Philosophy ofMathematics General Index 329
214, 220-1, 226, 232-3, 248, geometry xi, xii, 25, 33, 45, 68, 69, 116, Jourdain, Philip 242n.
Cauchy, Augustin Louis 68 223-4, 261-2, 273, 292, 308,319 Julius Caesar problem 157, 159-61, 180,
Cauchy condition for convergence 252 314-15, 319
cumulative hierarchy 233, 313 projective xii, 255 187, 189, 209-11, 213-14, 225,
Cellucci, Carlo 34n. Riemannian 293 226
chess 253-5 Currie, Gregory 33, 116n,, 242n.
Czuber, E. 24In. Godel, Kurt 254, 301 justification 25
choice sequences 235 Goldbach’s conjecture 62
Cimino, Marcello viii Goodman, Nelson 163 Kaal, Hans 93n.
classes 91-3, 121, 159, 168, 187-8, 200, Davidson, Donald 113, 21 In. Grassmann, Hermann Gunther 56-7 Kac, Mark 293n.
224-5, 233, 248-9, 308-9 de Moivre’s theorem 178, 248 Kambartel, Friedrich 93n.
Dedekind, Richard xi, xii, 8, 11, 12-13, group 279, 282, 294
classical logic 319, 321 Kant, Immanuel 2, 3, 7, 12, 23, 24-5,
Coffa, Alberto 322 47-54, 57, 61, 72, 77, 147, 243, 28-9, 36, 41, 48, 55-6, 63-7, 81,
246, 249-51, 262, 267, 270, 274, Haaparanta, L. 242n., 322
Cohen, Hermann 125 Hacker, Peter 67, 84, 85, 89 91, 98, 111
282, 283, 295-6, 311 Kitcher, Philip 11, 322
completeness of a formalisation of Hale, Bob 322
logic 30, 230 Dedekind cut 249-50, 281, 283-90, 319 Kossak, E. 142
deductive inference, fruitful 36, 41, Hallett, Michael 50n., 318n.
completeness of ordering 282, 288-9, Hankel, H. 63, 243, 247, 249, 262n., Kreisel, Georg xii, 235
305-6, 321 Kripke, Saul 30
293, 302 277n.
definition 23 see also contextual definition Kutschera, Franz von 242n., 322
completion rates viii condition for correct 30-5, 143, 150-4 Heine, E. 243, 249, 252-3
concepts 65, 66, 67, 88-93, 96-8 fruitful 14-15, 21, 34, 41 Helme, Mark vii
indefinitely extensible 316-19 Leibniz, Gottfried Wilhelm 56, 58, 70
piecemeal 244-6 Helmholtz, Hermann Ludwig Ferdinand Leibniz’s law 112, 128, 141-2
numbers ascribed to 74, 88, 94 requiring proof of admissibility 23, 57, von 60, 258
concept-formation 21, 39-40, 175 linguistic turn 111-12, 181
72, 126, 129 Hermes, Hans 93n. Locke, John 83
congruence relation 129-30 Desargues’s theorem 255 Hilbert, David xi, xii, 5, 26, 229, 233n.,
conservative extension 297-8, 312, 320 logic 12
description operator 2, 127, 147, 158, 292, 296, 303 characterised by universality of
consistency xi, 45, 178, 219, 222, 216, 219, 245 Hilbert space 293n. application 24, 43-6, 224, 308
229-30, 247-8, 261, 298, 304, 310, domain of quantification 175-6, 205-7, Hintikka, Jaakko 242n., 322 logical constants 203, 209, 257
312-13 219-22, 228-9, 232-5, 238, 304, Hobbes, Thomas 83 logical objects 133, 159, 166, 187, 197,
consistency proof 215-19, 221-2, 229 313-19 Holder, Otto 280, 281-3, 285, 287, 290 209, 224-5, 233, 249, 301-2, 304,
constructive mathematics 312-13 Donahue, Charles vii Hume, David 83, 142, 281 308, 310
content 13-16 Husserl, Edmund xi, 14, 19-21, 31-2, logicism 12, 224-5, 299, 301-5, 312
context principle 3, 21-2, 110, 111-12, equivalence classes 167, 180, 225, 233 50, 53, 82, 92n., 95-8, 141-52, 154 Lotze, Hermann 24, 183
117- 19, 125, 155-6, 180-240, 297, equivalence relation 128—9, 162—6, 167
300,308,310-11,314,320 Euclid 12, 82, 282-3, 290-1 Illigens, Eberhard 264—5, 268, 270, magnitude 63-4, 270, 272, 304
as governing reference 183-99, 210, Euler’s function 139 273-4 mathematical existence 229
230, 238 extensions of concepts see classes inconsistency 10, 133, 188, 194, 208, mathematical logic 12
as governing sense 183-4, 202-4, 209 209, 222, 223 see also Russell’s mathematical objects 225, 231, 233-5,
generalised 210-13, 215, 220, 230, feature-placing predicates 162 contradiction 239, 249, 292, 297-302, 304-5,
238 Field, Hartryxii, 297-300, 304, 312-13 induction, empirical 25, 51-2 307-21
contextual definition 125-31, 155-61, finitism 234 induction, mathematical 12, 69, 72, McGuinness, Brian 93n.
165, 169, 170, 175, 180-1, 188, finitistic statements 292, 303 120-1, 124 mean value theorem 67, 223
189-99, 200, 230, 236, 238, 239, infinitesimals 194 measurement theory 300
244
formalism 247, 249, 251-62, 252-62, metamathematics 255
294, 321 infinity 318
continuous function nowhere infinity of the natural number-series 124, Mill, John Stuart 42, 48, 58-60, 62, 75,
Frege, Gottlob x, xi-xiii 76, 81, 96, 258
differentiable 68, 223 early and middle periods 2-4, 15, 21, 132-3, 138, 206, 226-7, 303, 305,
continuum hypothesis 302, 310 318 modal logic 304-5
168, 170, 173, 180, 183, 191-3, Moore, Adrian 322
contradiction 133, 159, 208, 217, 225, 195-6, 198, 205-7 intuition 3, 7, 12, 44-5, 63-71, 116,
311, 315—16, 318 see also Russell’s 223-5, 232, 300, 301, 307, 308, Moore, G.E. 32, 143
late period 6-7
contradiction Furth, Montgomery 216n. 310-11, 320
intuitionism 227, 235, 313-14, 319-20 natural numbers as finite cardinals 51,
criterion for synonymy 171-2
53, 61, 274, 293
criterion of identity 3-4, 113-14, Gabriel, Gottfried 93n. necessity, epistemic and ontic 28-30,
118- 19, 127-8, 130, 135, 137-8, Gauss, Karl Friedrich 243, 278-9 Jevons, W.S. 70, 86-7 307-8
161-6, 167, 188-9, 200-1, 204 Gentzen, Gerhard 26, 254 Jordan-Holder theorem 281n.
 General Index 331
330 Frege: Philosophy ofMathematics 219-22, 223-6, 233, 239,'249, 261,
structure 295 286, 291, 304, 309-10, 314
negative types 310 261-2, 263-91, 296, 298-300, subjcctive/objective 20, 21, 65-7, 73
304-5, 313, 317-20 van Heijenoort, Jean 233n.
Neumann, Peter vii-viii, 242n., 288n. 76-80, 81, 181 Vcraart, Albert 93n., 242n.
Newton, Isaac 73, 243, 261 recognition statements 204, 209
nominalism 182, 207—8, 226, 236, 297, recursion equations 51, 57, 72 Veronese, G. 280, 282
299, 304, 320 reduction 224 Tarski, Alfred 217 von Wright, Georg Henrik 294n.
reductionism 191-9 theory of types 132, 310
non-standard analysis 194 Thiel, Christian 93n., 322
numbers as objects 3, 65, 67, 108-10, reference 32, 191-9, 200-1, 205, 208, Waismann, Friedrich 88, 148-51,
210, 214-15, 230-3, 235, 236-9 Thomae, Johannes 83, 241n., 243, 249, 259-61
114, 131-40, 190-1
numbers as sets of featureless units 20, Rein, Andrew 322 252-3 Wang, Hao 49n.
Resnik, Michael 322 thoughts 15, 78, 193, 202, 225, 237, 239 Weber, Heinrich 279n.
50, 82, 86-7, 144-6
numerical equivalence see cardinal Rhees, Rush 294n. Tiles, Mary 322 Weierstrass, Karl Theodor Wilhelm 68,
equivalence Russell, Bertrand 10, 26, 33, 49, 51-2, Troelstra, Anne 235 243-4, 246, 250-1, 263
numerically definite quantifiers 100, 103, 127, 132, 149, 189, 216, 225, 239, two-sorted theory 134—8 Whitehead, Alfred North 225, 242n.,
146—7 242n., 244, 246, 247, 249-50, 267, 278, 280, 302, 312
271, 278, 280, 287n„ 296, 302, Wittgenstein, Ludwig ix, x, 93, 182,
303, 312, 317 Ulam, Stanislaw 293n.
one-one correlation 51, 114, 119, 130, understanding 13, 15-16, 202-4, 207 242n., 260, 293^
142-5, 148-9, 157-8 axiom of infinity 42, 132 units see numbers as sets of featureless Wright, Crispin xii, 123, 160-2, 184-96,
ordinal numbers 229n., 234, 293, 313, Russell’s contradiction 4-6, 10, 103, 159, 198-9, 205-8, 224, 226, 232,236,
315,316-17 168, 216, 219, 241, 290, 298, 309 units
unsaturatedness 89-90, 185 239, 292
strongly inaccessible 313
original equivalence 155, 157, 160, 164, Sambin, G. 34n. Zermelo, Ernst 263n.
166, 170, 179, 189, 201, 204, 209, value-ranges 2, 121, 132, 133, 159,
saturation see unsaturatedness 170-1, 187-9, 197, 199, 210-14, Zsigmondy, Karl 6
226-7
Schirn, Matthias xi, 230n., 242n.
Schoenflies, A. 5
paradox of analysis 32, 143 Scholz, Heinrich 242n.
Peacocke, Christopher 171 Schroder, E. 83, 92n., 142
Peano arithmetic 254 Schroder-Bernstein theorem 130
Peano axioms 12-13, 49, 120-1, 123 Schroeder-Heister, Peter 322
Peano, Giuseppe 12, 241n., 244 Schubert, H. 241n.
performance indicators ix
Picardi, Eva 34 second-order quantification 217-19
platonism 301-5 sense 16, 34, 193, 195, 196-7, 201, UNIVERSITA’ DEGLI STUDI I
plural subjects 75, 93, 96 202-4, 207, 209, 210, 225, 236-7, L’AQUILA
294
positival class 287
dependence of 202-4, 209 25.09.96 103^1 |
positive class 241, 288
postulation 22, 178, 229, 239, 247-9 sense/reference distinction 2, 8, 15-17,
primitive root 248 66-7, 78, 183, 191-2
Pringsheim, Alfred 241n., 268-9, 271 set theory 12, 223, 224, 233-4, 301-2, I bTbuot eg a
309 JFAGOL'Ri’ LfTTERE E EMSHMI
psychologism 13-21, 31-2
Putnam, Hilary 153, 21 In. New Foundations 230, 234
von Neumann-Bemays 168
quantitative domain 278-81, 283-91 Zermelo-Fraenkel 234, 291, 312-13,
320
quantity 64, 73, 261-2, 268-91, 299-300
see also magnitude Shanker, Stuart 67
Quine, Willard Van Orman 29, 134, 185, Simons, Peter M. 242n.
230 sortal concept 162-3
Stolz, Otto 142, 243, 247, 249, 277n.
strategy of analysis
ratio 274, 276, 282-3, 290-1, 299-300, adjectival 99-110, 115, 121, 131-2
305
substantival 99, 115, 121
real numbers 73, 160, 165, 178, 227, Strawson, Sir Peter 162
233, 241-2, 246, 250, 252-3, 258, structuralism 295-7, 300
i • '((^fjjiiiiil^

Sito

" 0., „ iiiMrit loxojiiftn fti.1 ©©hj<p^Sfdft)ftp to^to- ^!h4^

(cxo»;ninifi»*i^?ft’9>in ft@ ©tJ-l|l' W’'F«l©rfOiiii^toj IW^'> ©1 1^^7<K:
Oito© i0feto

w fe fe; ■&
<©(?$©© [o)@0©ft to to @®W© 1®©^® l^ft^O';
WwJgWiMSSto ®OW5tl @©fc » ?01 |©7©totoify' 5Kte<sl '$$ ©Hi
©tenWf ©111] ©Oft'WW] lto©WjW ©@® ~'^^j ©ffi ©
©gJ ft© @§ffFf@§fi(? Oife Wt
ft© WSOto©0 ^W0§tto
0©@© itoexM, to toft ©ft IW wtte 'fttewto ft© ©
©QfflJW S®21 lefeWlgaWi ©ft togjlfe ©WStolto ©£■ BSpiW
(oifc^feft© itow©© to ©ft to to' ©SffltoW *'
iwtoi©©ife W© te to itein^gxwOto ©©©©Mi 'Maw )© © w?s@>:
©©[©©jftoitoto©.
^J&S) O©^©t©i'
WfWfW’OW *W iWOBS1
''0©©fi©©©toJ^ ©or) do®(^r«Q ©@©fc, @ @tO?Mw B' JMI
otetowf© ©ft to stotot ft.w/’tow o®@ato(K© ®ro$
^©©oft I^Iliq^Oub
WtoQK ©iljr^Wi to iWogii !©W©@©?mM'B’
I-ton) |©®l©Mto to <t ftggJSSi: @ totote) /^'ii)@lW ilto8&©
W 1MW.L
_(w <p)ft toftrria© ©(OgiiXfe
InlttelAV Iw^jggV
w© fe @go iw®KM’is@@es-
'W ‘W^W1 IMBO®
irWHfft( /<D> losi.rWnj iBWIK^ I I