REALISM IN MATHEMATICS

This page intentionally left blank

 REALISM IN
MATHEMATICS

PENELOPE MADDY

CLARENDON PRESS OXFORD

 This book has been printed digitally and produced in a standard specification
in order to ensure its continuing availability

OXTORD
UNIVERSITY PRESS

Great Clarendon Street, Oxford OX2 6DP

Oxford University Press is a department of the University of Oxford.
It furthers the University's objective of excellence in research, scholarship,
and education by publishing worldwide in
Oxford New York
Auckland Bangkok Buenos Aires Cape Town Chennai
Dar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata
Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi
Sao Paulo Shanghai Taipei Tokyo Toronto

Oxford is a registered trade mark of Oxford University Press

in the UK and in certain other countries
Published in the United States
by Oxford University Press Inc., New York
© Penelope Maddy 1990
The moral rights of the author have been asserted
Database right Oxford University Press (maker)
Reprinted 2003
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,
without the prior permission in writing of Oxford University Press,
or as expressly permitted by law, or under terms agreed with the appropriate
reprographics rights organization. Enquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the address above
You must not circulate this book in any other binding or cover
And you must impose this same condition on any acquirer
ISBN 0-19-824035-X
 For
Dick and Steve
This page intentionally left blank
 PREFACE

THE philosophy of mathematics is a borderline discipline, of

fundamental importance to both mathematics and philosophy.
Despite this, one finds surprisingly little co-operation between
philosophers and mathematicians engaged in its pursuit; more
often, widespread disregard and misunderstanding are broken only
by alarming pockets of outright antagonism. (The glib and
dismissive formalism of many mathematicians is offset by the
arrogance of those philosophers who suppose they can know what
mathematical objects are without knowing what mathematics says
they are.) This might not matter much in another age, but it does
today, when the most pressing foundational problems are unlikely
to be answered without a concerted co-operative effort. I have tried
in this book to do justice to the concerns of both parties, to present
the background, the issues, the proposed solutions on a neutral
ground where the two sides can meet for productive debate.
For this reason, I've aimed for a presentation accessible to both
non-philosophical mathematicians and non-mathematical philo-
sophers and, if I've succeeded, students and interested amateurs
should also be served. As far as I can judge, very little philosophical
training or background is presupposed here. Mathematical pre-
requisites are more difficult to avoid, owing to the relentlessly
cumulative nature of the discipline, but I've tried to keep them to a
minimum. Some familiarity with the calculus and its foundations
would be helpful, though surely not necessary. And the relevant set
theoretic concepts are referenced to Enderton's excellent intro-
ductory textbook (see his (1977)), for the benefit of those innocent
of that subject.

The central theme of the book is the delineation and defence of a

version of realism in mathematics called 'set theoretic realism'. In
this, my deep and obvious debt is to the writings of the great
mathematical realists of our day: Kurt Godel, W. V. O. Quine, and
Vlll PREFACE

Hilary Putnam (in the early 1970s). More personally, I have learned
most from John Burgess, Paul Benacerraf, Hartry Field, and Tony
Martin. After these, it would be impossible to mention everyone,
but I can't overlook the forceful criticisms of Charles Chihara, the
insightful comments of Anil Gupta, and the generous correspond-
ence, assistance, and advice of Philip Kitcher and Michael Resnik.
Most recently, Burgess, Field, Lila Luce, Colin McLarty, Martin,
Alan Nelson, Resnik, Stewart Shapiro, Mark Wilson, and Peter
Woodruff have all done me the service of reading and reacting to
drafts of various parts of the manuscript. (Naturally, the remaining
errors and oversights should be charged to my shortcomings rather
than to their negligence.) And finally, what I owe to my long-time
companion Steve Maddy is too complex and varied to be
summarized here. I am grateful to all these people and offer my
heartfelt thanks. Also to Angela Blackburn and Frances Morphy of
Oxford University Press.
Much of this book is based on a series of articles (Maddy 1980,
1981, 1984rf, 1988a,b, forthcoming a, b) the preparation of which
was supported, at various times, by the American Association of
University Women, the University of Notre Dame, the National
Endowment for the Humanities, the National Science Foundation,
and the University of Illinois at Chicago. The original publishers
kindly granted advance permission to reproduce material from
these pieces; in the end, only parts of (forthcoming a) (in chapter 5,
sections 1 and 2) and (forthcoming b) (in chapter 1, section 4)
actually survived, so I am particularly obliged to Kluwer Academic
Publishers and the Association for Symbolic Logic. Preparation of
the final draft was supported by National Science Foundation
Grant DIR-8807103, a University of California President's Re-
search Fellowship in the Humanities, and the University of
California at Irvine. The help of all these institutions is hereby
gratefully acknowledged.

Finally, I feel compelled to add a personal note on sexist language.

Some years ago, when I first introduced the ideas behind set theory
realism, constructions like 'the set theoretic realist thinks his
entities . . .' struck me as amusing, but since then I've discovered
that some readers and editors are legitimately disapproving of this
usage. Of the many alternatives available, I've chosen one that does
the least violence to the standard rhythm, that is, the use of 'she'
 PREFACE ix

and 'her' in place of 'he' and 'his' in neutral contexts. Some might
find this just as politically incorrect as the automatic use of the
masculine, but I sincerely doubt that phrasing like 'when the
mathematician proves a theorem, she . . .' makes anyone tend to
forget that some mathematicians are men. So I'll stick with this
policy. To those who find it distracting, I apologize; this is not, after
all, a political treatise. At least you have my reasons.
P.M.
Irvine, California
June 1989
This page intentionally left blank
 CONTENTS

1. Realism 1
1. Pre-theoretic realism 1
2. Realism in philosophy 5
3. Realism and truth 15
4. Realism in mathematics 20
2. Perception and Intuition 36
1. What is the question? 36
2. Perception 50
3. Intuition 67
4. Godelian Platonism 75
3. Numbers 81
1. What numbers could not be 81
2. Numbers as properties 86
3. Frege numbers 98
4. Axioms 107
1. Reals and sets of reals 107
2. Axiomatization 114
3. Open problems 125
4. Competing theories 132
5. The challenge 143
5. Monism and Beyond 150
1. Monism 150
2. Field's nominalism 159
3. Structuralism 170
4. Summary 177
References 182
Index 199
This page intentionally left blank
 1
REALISM

1, Pre-theoretic realism

Of the many odd and various things we believe, few are believed
more confidently than the truths of simple mathematics. When
asked for an example of a thoroughly dependable fact, many will
turn from common sense—'after all, they used to think humans
couldn't fly'—from science—'the sun has risen every day so far, but
it might fail us tomorrow'—to the security of arithmetic—'but 2
plus 2 is surely 4'.
Yet if mathematical facts are facts, they must be facts about
something; if mathematical truths are true, something must make
them true. Thus arises the first important question: what is
mathematics about? If 2 plus 2 is so definitely 4, what is it that
makes this so?
The guileless answer is that 2 + 2 — 4 is a fact about numbers,
that there are things called '2' and '4', and an operation called
'plus', and that the result of applying that operation to 2 and itself
is 4. '2 + 2 = 4' is true because the things it's about stand in the
relation it claims they do. This sort of thinking extends easily to
other parts of mathematics: geometry is the study of triangles and
spheres; it is the properties of these things that make the statements
of geometry true or false; and so on. A view of this sort is often
called 'realism'.
Mathematicians, though privy to a wider range of mathematical
truths than most of us, often incline to agree with unsullied
common sense on the nature of those truths. They see themselves
and their colleagues as investigators uncovering the properties of
various fascinating districts of mathematical reality: number
theorists study the integers, geometers study certain well-behaved
spaces, group theorists study groups, set theorists sets, and so on.
The very experience of doing mathematics is felt by many to
support this position:
2 REALISM

The main point in favor of the realistic approach to mathematics is the

instinctive certainty of most everybody who has ever tried to solve a
problem that he is thinking about 'real objects', whether they are sets,
numbers, or whatever . . . (Moschovakis (1980), 605)

Realism, then (at first approximation), is the view that math-

ematics is the science of numbers, sets, functions, etc., just as
physical science is the study of ordinary physical objects, astro-
nomical bodies, subatomic particles, and so on. That is, math-
ematics is about these things, and the way these things are is what
makes mathematical statements true or false. This seems a simple
and straightforward view. Why should anyone think otherwise?
Alas, when further questions are posed, as they must be,
embarrassments arise. What sort of things are numbers, sets,
functions, triangles, groups, spaces? Where are they? The standard
answer is that they are abstract objects, as opposed to the concrete
objects of physical science, and as such, that they are without
location in space and time. But this standard answer provokes
further, more troubling questions. Our current psychological
theory gives the beginnings of a convincing portrait of ourselves as
knowers, but it contains no chapter on how we might come to
know about things so irrevocably remote from our cognitive
machinery. Our knowledge of the physical world, enshrined in the
sciences to which realism compares mathematics, begins in simple
sense perception. But mathematicians don't, indeed can't, observe
their abstract objects in this sense. How, then, can we know any
mathematics; how can we even succeed in discussing this remote
mathematical realm?
Many mathematicians, faced with these awkward questions
about what mathematical things are and how we can know about
them, react by retreating from realism, denying that mathematical
statements are about anything, even denying that they are true: 'we
believe in the reality of mathematics, but of course when
philosophers attack us with their paradoxes we rush to hide behind
formalism and say "Mathematics is just a combination of meaning-
less symbols" . . .'- 1 This formalist position—that mathematics is
just a game with symbols—faces formidable obstacles of its own,
which I'll touch on below, but even without these, many math-
ematicians find it involving them in an uncomfortable form of
1
Dieudonne, as quoted in Davis and Hersh (1981), 321.
 PRE-THEORETIC REALISM 3

double-think. The same writer continues: 'Finally we are left in

peace to go back to our mathematics and do it as we have always
done, with the feeling each mathematician has that he is working
on something real' (Davis and Hersh (1981), 321). Two more
mathematicians summarize:
the typical working mathematician is a [realist] on weekdays and a
formalist on Sundays. That is, when he is doing mathematics he is
convinced that he is dealing with an objective reality whose properties he is
attempting to determine. But then, when challenged to give a philosophical
account of this reality, he finds it easiest to pretend that he does not believe
in it after all (Davis and Hersh (1981), 321)

Yet this occasional inauthenticity is perhaps less troubling to the

practising mathematician than the daunting requirements of a
legitimate realist philosophy:
Nevertheless, most attempts to turn these strong [realist] feelings into a
coherent foundation for mathematics invariably lead to vague discussions
of 'existence of abstract notions' which are quite repugnant to a
mathematician . . . Contrast with this the relative ease with which formal-
ism can be explained in a precise, elegant and self-consistent manner
and you will have the main reason why most mathematicians claim to be
formalists (when pressed} while they spend their working hours behaving
as if they were completely unabashed realists. (Moschovakis (1980), 605-6)
Mathematicians, after all, have their mathematics to do, and they
do it splendidly. Dispositionally suited to a subject in which
precisely stated theorems are conclusively proved, they might be
expected to prefer a simple and elegant, if ultimately unsatisfying,
philosophical position to one that demands the sort of metaphysical
and epistemological rough-and-tumble a full-blown realism would
require. And it makes no difference to their practice, as long as
double-think is acceptable.
But to the philosopher, double-think is not acceptable. If the very
experience of doing mathematics, and other factors, soon to be
discussed, favour realism, the philosopher of mathematics must
either produce a suitable philosophical version of that position, or
explain away, convincingly, its attractions. My goal here will be to
do the first, to develop and defend a version of the mathematician's
pre-philosophical attitude.

Rather than attempt to treat all of mathematics, to bring the project

4 REALISM

down to more manageable size, I'll concentrate here on the

mathematical theory of sets.2 I've made this choice for several
reasons, among them the fact that, in some sense, set theory forms a
foundation for the rest of mathematics. Technically, this means that
any object of mathematical study can be taken to be a set, and that
the standard, classical theorems about it can then be proved from
the axioms of set theory.3
Striking as this technical fact may be, the average algebraist or
geometer loses little time over set theory. But this doesn't mean that
set theory has no practical relevance to these subjects. When
mathematicians from a field outside set theory are unusually
frustrated by some recalcitrant open problem, the question arises
whether its solution might require some strong assumption
heretofore unfamiliar within that field. At this point, practitioners
fall back on the idea that the objects of their study are ultimately
sets and ask, within set theory, whether more esoteric axioms or
principles might be relevant. Given that the customary axioms of
set theory don't even settle all questions about sets,4 it might even
turn out that this particular open problem is unsolvable on the basis
of these most basic mathematical assumptions, that entirely new set
theoretic assumptions must be invoked. 5 In this sense, then, set
theory is the ultimate court of appeal on questions of what
mathematical things there are, that is to say, on what philosophers
call the 'ontology' of mathematics. 6
Philosophically, however, this ontological reduction of math-
ematics to set theory has sometimes been taken to have more
dramatic consequences, for example that the entire philosophical
foundation of any branch of mathematics is reducible to that of set
theory. In this sense, comparable to implausibly strong versions of
2
A set is a collection of objects. Among the many good introductions to the
mathematical theory of these simple entities, I recommend Enderton (1977).
3
See e.g. the reduction of arithmetic and real number theory to set theory in
Enderton (1977), chs. 4 and 5. There are some exceptions to the rule that all
mathematical objects can be thought of as sets—e.g. proper classes and large
categories—but 1 will ignore these cases for the time being.
4
Some details and philosophical consequences of this situation are the subject of
ch. 4.
s
Eklof and Mekler (forthcoming) give a survey of algebraic examples, and
Moschovakis (1980) does the same for parts of analysis.
6
In philosophical parlance, 'ontology', the study of what there is, is opposed to
'epistemology', the study of how we come to know what we do about the world. I
will use the word 'metaphysics' more or less as a synonym for 'ontology'.
 R E A L I S M IN P H I L O S O P H Y 5

the thesis that physics is basic to the natural sciences,7 I think the
claim that set theory is foundational cannot be correct. Even if the
objects of, say, algebra are ultimately sets, set theory itself does not
call attention to their algebraic properties, nor are its methods
suitable for approaching algebraic concerns. We shouldn't expect
the methodology or epistemology of algebra to be identical to that
of set theory any more than we expect the biologist's or the
botanist's basic notions and techniques to be identical to those of
the physicist. But again, this methodological independence of the
branches of mathematics from set theory does not mean there must
be mathematical entities other than sets any more than the
methodological independence of psychology or chemistry from
physics means there must be non-physical minds or chemistons.8
But little hangs on this assessment of the nature of set theory's
foundational role. Even if set theory is no more than one among
many branches of mathematics, it is deserving of philosophical
scrutiny. Indeed, even as one branch among many, contemporary
set theory is of special philosophical interest because it throws into
clear relief a difficult and important philosophical problem that
challenges many traditional attitudes toward mathematics in
general. I will raise this problem in Chapter 4.
Finally, it is impossible to divorce set theory from its attendant
disciplines of number theory and analysis. These two fields and
their relationship to the theory of sets will form a recurring theme
in what follows, especially in Chapters 3 and 4.

2. Realism in philosophy

So far, I've been using the key term 'realism' loosely, without clear
definition. This may do in pre-philosophical discussion, but from
7
This view is called 'physicalism'. I'll come back to it in ch. 5, sect. 1, below.
8
There was a time when the peculiarities of biological science led practitioners
to vitalism, the assumption that a living organism contains a non-physical
component or aspect for whose behaviour no physical account can be given.
Nowadays, this idea is discredited—simply because it proved scientifically sterile—
and, as far as I know, no one ever urged the acceptance of 'chemistons'. Today,
psychology is the special science that most often lays claim to a non-physical subject
matter, but as suggested in the text, it seems to me that a purely physical ontology is
compatible with the most extreme methodological independence. For discussion, see
Fodor (1975), 9-26.
6 REALISM

now on I will try to be more precise. This doesn't mean I'll succeed
in defining the term exactly, but at least I'll narrow the field
somewhat, I hope helpfully. Let me begin by reviewing some
traditional uses of the term in philosophy.

One of the most basic ontological debates in philosophy concerns

the existence of what common sense takes to be the fundamental
furniture of the world: stones and trees, tables and chairs, medium-
sized physical objects. Realism in this context, often called
'common-sense realism', affirms that these familiar macroscopic
things do in fact exist. But it is not enough for the realist to insist
that there are stones and trees and such like, for in this much the
idealist could agree, all the while assuming that a stone is a mental
construct of some sort, say a bundle of experiences. However, such
an idealist, like the Bishop Berkeley, will have serious trouble
agreeing with the realist that stones can exist without being
perceived.9 Thus the common-sense realist can state her position in
a way that rules out idealism by claiming that stones etc. exist, and
that their existence is non-mental, that they are as they are
independently of our ability to know about them, that their
existence is, in a word, 'objective'.
A more recent opponent of the common-sense realist uses a more
devious technique.10 The phenomenalist hopes to say exactly what
the realist says while systematically reinterpreting each and every
physical object claim into a statement about what she calls 'sense
data', or really, into statements about possible sense data. For
example, my overcoat exists in the closet though unperceived
because part of the translation of 'the overcoat is in the closet' is
something like 'if I were in the closet and the light were on, then I'd
have an overcoat-like experience', which is, presumably, true.
Physical objects are not taken to consist of ideas, as with Berkeley,
but physical object statements are taken to mean something other
than what we ordinarily take them to mean.
9
Berkeley's notorious solution was to suppose that God is perceiving the object
even when we aren't; indeed he uses this as a novel argument for the existence of
God. See, e.g. Berkeley (1713), 211-13, 230-1. It's worth noting, however, that in
earlier work, Berkeley (1710 §§ 3, 58—9) includes a 'counterfactual' analysis that
prefigures the Millian phenomenalism described in the next paragraph.
10
This idea took shape in Mill (1865), ch. 11 and its appendix, and was
developed in the form described here by the logical positivists. See Ayer (1946),
63-8.
 R E A L I S M IN P H I L O S O P H Y 7

This ambitious programme was a complete failure for a number

of nagging reasons, only the first of which is our seeming inability
to specify the required sense datum—the overcoat-like experience
—without reference to the overcoat itself.11 But whatever the
failings of phenomenalism, the attempt itself shows that the realist,
to state her position completely, must also rule out such unintended
misinterpretations of common-sense statements; she must insist
that these statements be taken 'at face value'. Because it is hard to
say exactly what this comes to, apart from repeating that it rules
out phenomenalism, realism is in some ways more difficult to state
than its particular rivals. In any case, we can be sure that common-
sense realism is opposed to both idealism and phenomenalism.
Our discussion so far has centred on the problem of stating
common-sense realism; we must now ask why we should believe it.
The failure of heroic philosophical alternatives like idealism and
phenomenalism is some reassurance, but we would like a positive
argument. Admittedly, we find it difficult not to believe in ordinary
physical objects; of his own philosophical scepticism, the great
David Hume writes:
since reason is incapable of dispelling these clouds, nature herself suffices to
that purpose . . . I dine, I play a game of backgammon, I converse, and am
merry with my friends; and when after three or four hours' amusement, I
would return to these speculations, they appear so cold, and strained, and
ridiculous, that I cannot find in my heart to enter into them any farther.
(Hume (1739), 548-9)
But even if common-sense realism is psychologically inevitable, we
should still ask after its justification.
The reply given by many contemporary philosophers is simply
that the existence of ordinary things provides the best account of
our experience of the world. In his landmark essay on ontology,
'On what there is', W. V. O. Quine puts the point this way:
By bringing together scattered sense events and treating them as
perceptions of one object, we reduce the complexity of our stream of
experience to a manageable conceptual simplicity. . .. we associate an earlier
and a later round sensum with the same so-called penny, or with two different
so-called pennies, in obedience to the demands of maximum simplicity in our
total world-picture. (Quine (1948), 17)

11
See Urmson (1956), ch. 10, for a survey of this and other difficulties.
8 REALISM

Now we can hardly be said to make an explicit inference from

purely experiential statements to physical object statements that
account for them, because (as noted in connection with phenomen-
alism) we have no independent language of experience. What
actually happens is a developing neurological mediation between
purely sensory inputs and our primitive beliefs about physical
objects.12 The justificatory inference comes later, when we argue
that the best explanation of our stubborn belief in physical objects
is that they do exist and that our beliefs about them are brought
about in various dependable ways, for example by light bouncing
off their surfaces on to our retinas etc. Thus the assumption of
objectively existing, medium-sized physical objects plays an indis-
pensable role in our best account of experience.
But, one might object, didn't the gods of Homer provide the
Greeks with an explanation of their experience? Here Quine points
to an important difference:
For my part I do, qua lay physicist, believe in physical objects and not in
Homer's gods; and I consider it a scientific error to believe otherwise. , , ,
The myth of physical objects is epistemologicaliy superior to most in that it
has proved more efficacious than other myths as a device for working a
manageable structure into the flux of experience. (Quine (1951), 44)
Physical objects, not Homer's gods, form part of our best scientific
theory of the world, and for that reason, our belief in the former,
but not the latter, is justified.
Notice, however, that this sort of answer will not satisfy the
philosophical sceptic who calls all our belief-forming techniques,
including those of natural science, into question. Rene Descartes,
for example, was well aware that science presupposed an objective
external world, but he wanted a justification for science itself. How
can we know, Descartes asked, that the scientific world-view is
correct? How do we know our senses aren't deceiving us? How do

12
For more on this, see ch. 2, sect. 2, below. Students of Quine may detect a
tension between my position in this paragraph and such Quinean remarks as 'From
among the various conceptual schemes best suited to these various pursuits, one—
the phenomenalistic—claims epistemological priority' ((1948), p. 19). Here and in
what follows, I will ignore this lingering trace of positivism in the master. In fact,
there is no phenomenalistic language or theory, and a good scientific explanation
must do more than accurately predict sense experiences. (Cf. Putnam (1971),
355-6.)
 REALISM IN P H I L O S O P H Y 9

we know we aren't dreaming? How do we know there is no Evil

Demon systematically making it appear to us as if the world is as
we think it is?13
These Cartesian challenges depend on a conception of epistemo-
logy as an a priori14 study of knowledge and justification, a study
above, beyond, outside, indeed prior to, natural science, a study
whose aim is to establish that science on a firm footing. One might
think that the justificatory practices of science itself are the best we
have, but classical epistemology appeals to higher canons of pure
reason. Unfortunately, its attempts to reconstruct natural science
on an a priori, philosophically justified foundation have all failed,
beginning with Descartes's own effort.15
In light of this history, Quine suggests a radically different
approach to epistemology. Our best understanding of the world,
after all, is our current scientific theory, so by what better canons
can we hope to judge our epistemological claims than by scientific
ones? The study of knowledge, then, becomes part of our scient-
ific study of the world, rather than an ill-defined, pre-scientific enter-
prise: 'Epistemology . . . simply falls into place as a chapter of
psychology and hence of natural science. It studies a natural
phenomenon, viz., a physical human subject' (Quine (19696), 82).
Standing within our own best theory of the world—what better
perspective could we have?—we ask how human subjects like
ourselves are able to form reliable beliefs about the world as our
theory tells us it is. This descriptive and explanatory project is
called 'epistemology naturalized'.
Thus science is used to justify science, but this circle is not vicious
once we give up the classical project of founding science on
something more dependable than itself. From naturalized perspect-
ive, there is no point of view prior or superior to that of natural
science, and Quine's argument for common-sense realism becomes
perfectly reasonable: the assumption of physical objects is part of
our best theory, and being part of our best theory is the best
justification a belief can have.
13
Descartes (1641), esp. Meditation One.
14
'A priori' means prior to experience, as opposed to 'a posteriori'.
15
Descartes argued that our perceptions are reliable because God is no deceiver.
Serious objections to this approach arose immediately; see his 'Objections and
replies', published as an appendix to his Meditations. A more recent failed effort to
found science is that of the positivists. See Quine's discussion (19696).
10 REALISM

A second form of philosophical realism concerns itself with the

more esoteric objects of science, with unobservable theoretical
entities like electrons, genes, and quarks. Here the scientific realist
asserts that our belief in such things is justified, at least to the extent
that theories involving them provide us with the best explanation
we have for the behaviour of observable objects. Once again,
however, it is not enough for the realist to say just this. While there
can be no idealist here, analogous to the Berkeleian, insisting that
electrons are just bundles of sensory experiences, there is a position
analogous to phenomenalism: instead of translating talk about
medium-sized physical objects into talk about what sensory
experiences would occur under what circumstances, the opera-
tionalist would have us translate talk of unobservable theoretical
entities into talk about how observables would behave under which
circumstances.16 Thus, for example, part of the translation of
'there's a quark here' might be 'if we set up a cloud chamber, we'd
get this kind of track' and 'if we prepared a photographic emulsion,
we'd see this kind of trace' and 'if we had a scintillation counter,
we'd get this type of signal'. This project failed as resoundingly as
phenomenalism, 17 and for some of the same reasons, but again, its
very existence shows that along with asserting the existence of
those unobservables presupposed by our best theory, the scientific
realist must also insist that this assertion be taken 'at face value'.
The scientific realist's most conspicuous opponent is the instru-
mentalist, who holds that unobservables are a mere 'useful fiction'
that helps us predict the behaviour of the observable. Thus the
instrumentalist denies just what the scientific realist asserts—that
there are electrons etc.—but continues to use the same theories the
realist does to predict the behaviour of observables. For a practising
scientist, instrumentalism would seem as dramatic a form of
double-think as the duplicitous mathematical formalism described
earlier. But there are worse problems than inauthenticity. For
example, the distinction between theoretical and observational
turns out to be devilishly hard to draw. 18 And even if it could be
16
The classic statement of operationalism is Bridgman (1927). It is criticized by
Hempel (1954), but the logical positivists found more subtle forms (e.g. Carnap
(1936/7)).
17
One of its many critics is Putnam (1962). Others are Maxwell (1962) and
Achinstein(1965).
18
See the papers cited in the previous footnote.
 REALISM IN P H I L O S O P H Y 11

drawn, it is unclear why the difference between being humanly

observable and not should have such profound metaphysical
consequences.19 Finally, the instrumentalistic scientist, happily
using the false premisses of theoretical science to derive purportedly
true conclusions about observables, provides us no explanation of
why a batch of false claims should be so dependable.
So, once again, as in the case of common-sense realism, the
failure of the opposition provides some negative support for
scientific realism. The positive argument is also analogous.20 After
the above-quoted defence of common sense realism, Quine con-
tinues:
Positing does not stop with macroscopic physical objects. Objects at the
atomic level are posited to make the laws of macroscopic objects, and
ultimately the laws of experience, simpler and more manageable.. .
Science is a continuation of common sense, and it continues the common-
sense expedient of swelling ontology to simplify theory. . . . Epistemologic-
ally these are myths on the same footing with physical objects and gods,
neither better nor worse except for the differences in the degree to
which they expedite our dealings with sense experiences. (Quine (1951),
44-5)
From the point of view of epistemology naturalized, what better
justification could we have to believe in the most well-confirmed
posits of our best scientific theory than the fact that they are the
most well-confirmed posits of our best scientific theory?

The final realist/anti-realist controversy I want to consider is in fact

the oldest debate in which the term 'realism' arises. We common-
sense realists all agree that there are many red things—red roses,
red houses, red sunsets—but the ancient question is whether or not
there is also, over and above this lot of particular red things, a
further thing they share, namely, redness. Such an additional
thing—redness—would be a universal. The most basic difference
between particulars and universals is that a universal can be present
('realized', 'instantiated', 'exemplified') in more than one place at a
19
See Devitt (1984), §§ 7.1, 7.6, 8.5.
20
The analogy is not perfect because with physical objects, unlike theoretical
entities, we believe in them from the start; we are never in the position of deciding
whether or not to begin believing in them. My point is that the form of the
justification for the belief, however it is arrived at, is the same in both cases.
12 REALISM

time, while a particular cannot.21 Plato originated the most

dramatic version of realism about universals in his spectacular
theory of Forms: Redness, Equality, Beauty, and so on, are perfect,
eternal, unchanging Forms; they exist outside of time and space; we
know them by means of the non-sensory intellect; ordinary physical
properties, perceived by the usual senses, are but pale and imperfect
copies.22 Aristotle, Plato's student, took direct aim at the more
bizarre elements of this view, and defended a more modest form of
realism, according to which universals exist only in those things
that exemplify them. 23 Their opponent is the nominalist, like John
Locke, who holds that there is nothing over and above particulars.24
One classical argument for realism about universals, the One
over Many, survives in the modern debate. David Armstrong, its
most vocal contemporary advocate, puts it like this: 'Its premise is
that many different particulars can all have what appears to be
the same nature' (Armstrong (1978), p. xiii); 'I would . . . draw the
conclusion that, as a result, there is a prima facie case for
postulating universals' (Armstrong (1980), 440-1). A similar
argument can be given in linguistic form, arguing for example that
a universal redness must exist since the predicate 'redness' is
meaningful. Most contemporary thinkers, including Armstrong,
reject this second form of the argument. 25 To see why, consider a
scientific universal like 'being at a temperature of 32 degrees
Fahrenheit'. Scientists tell us that this is the same universal, or
property, to use a more natural-sounding word, as 'having such-
and-such mean kinetic energy'.26 But these two predicates have
very different meanings. So properties are not meanings. They are
21
A particular need not be spatially or temporally continuous—my copy of
Principia Mathematica is in three volumes, on two different shelves—but even then,
it is only part of the particular that is present in each of the disparate locations. A
universal is understood to be fully present in each of its instances.
22
The standard reference for Plato's theory is his Republic, chs. 5-7. Wedberg
(1955), ch. 3, gives a useful summary. He cues the later Timaeus (37 D-38 A) as
defining the sense in which Forms are outside time. A metaphorical passage in
Phaedrus (247) declares their location as 'the heaven which is above the heavens',
indicating they are not spatial, and Aristotle's commentary confirms this: 'the Forms
are not outside, because they are nowhere' (Physics, 203 a 8).
23
See his Metaphysics, bk. 1, § 9 for criticisms of Plato, and Categories 2, for his
own view.
24
See Locke (1690), bk. 3, ch. 3,§ 1.
25
See Armstrong (1978), pt. 4. See also Putnam (1970), § 1.
26
Wilson (1985) argues that this frequently cited example is more complex than
philosophers ordinarily appreciate, but I don't think this important observation
affects the point at issue here.
 REALISM IN P H I L O S O P H Y 13

individuated by scientific tests, such as playing the same causal role,

rather than by synonymy of predicates.
But even Armstrong's preferred form of the One over Many
argument has been severely criticized from a number of different
perspectives.27 One stunningly simple counter-argument, Quine's,
goes like this. We want to say that Ted and Ed are white dogs. This
is supposed to commit us to the universal 'whiteness'. But for 'Ted
and Ed are white dogs' to be true, all that is required is that there be
a white dog named 'Ted' and a white dog named 'Ed'; no
'whiteness' or even 'dogness' is necessary. If there is more than this
to the One over Many argument, the realist owes an account of
what that is. If not, the realist needs some other support for the
existence of these universals.
Contemporary thinkers have proposed a more modern argu-
ment, modern in the sense that it partakes of the 'naturalizing'
tendency identified above in recent epistemology. Notice first that
the One over Many is presented as an a priori philosophical
argument for an ontological conclusion: there are universals. Now
epistemology naturalized, as described above, has renounced the
classical claim to a philosophical perspective superior to that of
natural science. As a result, the Cartesian demand for certainty
beyond the scientific was also rejected. From this new naturalized
perspective, the considered judgement of science is the best
justification we can have. Finally, we've seen that this shift in
epistemological thinking produces a corresponding shift in onto-
logical thinking; for example, despite philosophical qualms about
unobservable entities, we should admit they exist if our best science
tells us they do. The moral for the defender of universals is clear: to
show that there are universals, don't try to give a pre-scientific
philosophical argument; just show that our best scientific theory
cannot do without them.
Much of the current debate takes this form.28 The question at
issue is whether intensional entities like universals are needed, or
whether science can get by on extensional entities like sets.29 Most
27
See Quine (1948), Devitt (1980), Lewis (1983).
28
With Putnam (1970) and Wilson (1985; and forthcoming) on the positive,
Quine (1948; 19806) on the negative.
29
Universals are intensional because two of them can apply to the same
particulars without being identical, for example, 'human being' and 'featherless
biped'. Sets, by contrast, are extensional; two sets with the same members are
identical.
14 REALISM

will grant chat sets are more promiscuous than universals; random
elements can be gathered into a set even if they have no property in
common. Furthermore, science seems to need a distinction between
random collections and 'natural' ones; when we notice that all the
ravens we've examined are black, we conclude that all ravens
belong to the set of black things, not that all ravens belong to the set
of things that are either black or not examined by us. The open
question is whether the nominalist can deal with this distinction
between natural and unnatural collections without appealing to
universals.30

To summarize, then, to be a realist about medium-sized physical

objects, the theoretical posits of science, or universals, is to hold
that these entities exist, that they do so objectively—they are not
mental entities, and they have the properties they do independently
of our language, concepts, theories, and of our cognitive apparatus
in general—and to resist various efforts—phenomenalism, opera-
tionalism—to reinterpret these claims. And, in the naturalized
spirit, the realist assumes that the most strongly held of our current
theoretical beliefs are probably at least approximately correct
accounts of what these things are like. Beyond what's sketched
above, I will pause no further over arguments for or against these
three forms of philosophical realism, 31 but I will take something of
a stand.
Most of what follows will presuppose both common-sense and
scientific realism.32 Indeed, as will come out below, the debate
about the existence and nature of mathematical entities is almost
always posed by comparing them with medium-sized physical
objects and/or theoretical entities; the philosopher's temptation is
to embrace common-sense and scientific realism while rejecting
mathematical realism. 3 ' I want to remark, however, that I don't

30
Lewis (1983) provides a useful survey of the debate.
31
Devitt (1984) provides a useful compendium of arguments for common-
sense and scientific realism.
32
Some argue that common-sense and scientific realism are incompatible,
because physics reveals chat medium-sized physical objects are quite different from
our common-sense conception. But showing we are often wrong about stones and
tables is not the same as showing that these things don't exist. I see no problem in
allowing that science can correct common sense. Devitt (1984, § 5.10) sketches a
position of this sort.
33
See e.g. Putnam (19756), 74.
 REALISM AND TRUTH 15

think the rough-and-ready mathematical realism introduced in the

previous section stands or falls with these other realisms. If its
central tenet is that mathematics is as objective a science as
astronomy, physics, biology, etc., then this might remain true even
if those natural sciences turn out not to be as objective as the realist
thinks they are. In the long run, I'm much more interested in
blocking a serious metaphysical or epistemological disanalogy
between mathematics and natural science than I am in maintaining
a strict realism about either.34
Finally, because I think the issues involved are not nearly as clear,
I will remain officially neutral on universals. The question will arise
(in Chapter 2, section 4; Chapter 3, section 2; and Chapter 5,
sections 2 and 3) but I think nothing I say will hang on the ultimate
resolution of the metaphysical debate outlined above. The problem
in question is quite general; it is no more a problem for
mathematics than it is for the rest of science.35

3. Realism and truth

In recent years, many philosophers have come to think that realism

should be understood, not as a claim about what there is, but as a
claim about semantics.36 Whether one is a realist or not—about the
objects of common sense, theoretical entities, universals, or
mathematical objects—is said to depend on what one takes to be
the conditions for the truth or falsity of the corresponding
statements.
Now my own pre-philosophical statements about mathematical
realism do involve what sounds like a semantic element: I claimed
that mathematics is about numbers, sets, functions, etc., and that
the way these things are is what makes mathematical statements
true or false. This sort of talk can be read as espousing a

34
For example, my argument against Wittgenstein (Maddy (1986)) takes the
form: even if his general anti-realism is correct, still his strong maths/science
disanalogy need not be accepted.
35
Unless, of course, all mathematical entities are universals. A version of this
view is considered in ch. 5, sect. 3, below.
36
The influence here is Dummett's; see Dummett (1978), introd. and chs. 1, 10,
and 14, and (1977), ch. 7. Devitt (1984) gives a more complete discussion of the
relationships between realism and semantics, and his ch. 12 takes up Dummett's
position in particular.
16 REALISM

correspondence theory of truth, according to which the truth of a

sentence depends partly on the structure of the sentence, partly on
the relations between the parts of the sentence and extra-linguistic
reality,37 and partly on the nature of that extra-linguistic reality.
One definitive aspect of correspondence theories is that what it
takes for a sentence to be true might well transcend what we are
able to know.
Semantic anti-realists, by contrast, want to identify the truth
conditions of a sentence with something closer to our abilities to
know, with that which justifies the assertion of the sentence, with
some version of its 'verification conditions'. Notice that phenom-
enalism could be reinterpreted this way, as the claim that the truth
of 'my overcoat is in the closet' reduces to the truth of various
counterfactual conditionals like the one about what experiences I'd
have if I were in the closet with the light on. Moves to
verificationism are variously motivated—by the hope of avoiding
scepticism, by the desire to eliminate metaphysics, by a disbelief in
the objective reality of the entities in question, by attention to
purported facts of language learning, by scepticism about the
notion of correspondence truth itself. 38
Thus these semantic thinkers identify realism about a certain
range of entities with a correspondence theory of truth for
sentences concerning those entities, and likewise anti-realism with
verificationism. Given our previous characterization of realism as a
position on what there is, such an identification seems wrong-
headed. For example, an idealist like Berkeley could embrace a
correspondence theory; for him, the extra-linguistic reality that
makes ordinary physical object statements either true or false
consists of bundles of experiences. Being a correspondence theorist
doesn't make him a realist in our sense because his objects aren't
objective.
On the other hand, our realist thinks her entities do exist
objectively, which includes the belief that they exist and are as they
are independently of our abilities to know about them. She holds,
37
This formulation will have ro be modified in the special case of statements
explicitly about language, but I'll ignore this complication.
58
Examples of each, in order: the phenomenalist Mill (1865), chs. 1 and 11,
inspired by the idealist Berkeley (1710; 1713), (see the preface to Berkeley (1713) );
the positivists, for example Ayer (1946, ch. 1); the idealist Brouwer (1913; 1949),
see also Heyting (1931; 1966); the thoroughgoing verificationist Dummett (1975;
1977, ch. 7); and finally, Putnam (1977; 1980).
 R E A L I S M AND TRUTH 17

then, that there are, or at least may be, many truths about those
entities that are beyond the reach of even our most idealized
procedures of verification. Thus she could hardly be a verificationist.
So, even if holding a correspondence theory isn't the same thing
as being a realist, it might seem that realism requires a correspon-
dence theory. But this only follows on the assumption that these
two candidates for a theory of truth—correspondence and verifi-
cationism—exhaust the field. They do not.
A third type of truth theory is based on the simple observation
that 'so-and-so is true' says no more than 'so-and-so', in particular
that 'My overcoat is in the closet' is true if and only if my overcoat
is in the closet. This sort of theory has various names—redundancy
theory, disappearance theory, deflationary theory—but I'll call it a
disquotational theory. On this view, truth is a nothing more than a
simplifying linguistic device. In cases like that of my overcoat, it
does little more than stylistic work. When we use it in more
complex contexts—for example, if I claim that everything in the
Bible is true—it works as an abbreviating device that saves me a lot
of time. With that one sentence I'm asserting that in the beginning
God created the heavens and the earth, and now the earth was a
formless void . . . and God's spirit hovered over the water, and God
said, 'Let there be light', and there was light, and. . ., and so on
through the many sentences in the Bible.
In recent years, there has been considerable debate between
naturalized realists over correspondence versus disquotational
truth.39 The issue, of course, is whether or not the notion of
correspondence truth must figure at all in our best theory of the
world. For many purposes that seem to require a full-blown
correspondence notion, disquotational truth has turned out to do
the job as well. This is true even in philosophical contexts. For
example, I claimed earlier that mathematical statements are true or
false independently of our ability to know this. On the disquo-
tational theory, this would come to a long (indeed infinite)
conjunction: whether or not 2 + 2 = 4 is independent of whether
or not we can know which, and whether or not every non-empty
bounded set of reals has a least upper bound is independent of
whether not we can know which, and whether or not there is an
inaccessible cardinal is independent of whether or not we can know
39
See e.g. Field (1972), Grover et al. (1975), Leeds (1978), Devitt (1984), Field
(1986), and the references cited there.
18 REALISM

which, and . . ., etc. So the question is whether there are any jobs
for which a notion of correspondence truth is actually indispensable.
To see what hangs on this, consider: Alfred Tarski's celebrated
definition of correspondence truth reduces the problem of pro-
viding an account of that notion to the problem of providing an
account of the word-world connections, that is, an account of the
relation of reference that holds between a name and its bearer,
between a predicate and the objects that satisfy it.40 The disquo-
tational theory, on the other hand, includes a theory of reference as
unexciting as its theory of truth: 'Albert Einstein' refers to Albert
Einstein; 'gold' refers to gold. So what hangs on the debate between
correspondence and disquotational truth is the need for a substan-
tive theory of what it is by virtue of which my use of the name
'Albert Einstein' manages to pick out that certain historical
individual. This is no trivial matter; hence the lively debate over
whether or not a correspondence theory is really necessary.
I won't get into that debate here, because it would take us too far
afield and I certainly have nothing helpful to add, but I do want to
take note of where the hunt has led these investigators. It's perhaps
not surprising that where push seems to come to shove on the
question of truth and reference is in that portion of our theory of
the world that treats the activities of human beings.
Consider this case: Dr Jobe heals Isiah Thomas's ankle injury in
time for the big game. How do we explain this phenomenon? 41
As a first step, we notice that Dr Jobe has a vast number of true
beliefs about sports injuries, about the rigours of basketball, about
the physical and mental condition of basketball players, and about
Isiah in particular. What Dr Jobe thinks about these things is
usually correct. How, then, are we to explain his reliability on these
topics? In an effort to answer this question, we begin to detail such
things as the doctor's previous experience, his medical training, his
interactions with many basketball players, and his previous
interactions with Isiah himself.
Now the truth of the doctor's beliefs might be accounted for
disquotationally. But the correspondence theorist will point out
that among the sorts of connections between Jobe's beliefs and the
subject matter of those beliefs that we've been describing in order to
account for his reliability are just the sorts of connections that
40
For Tarski's theory, seeTarski (1933); for the reduction, see Field (1972).
41
Here 1 follow Field (1986), though the particular example is my own.
 R E A L I S M AND TRUTH 19

might well set up a robust referential connection between his uses

of the predicate 'basketball player' and basketball players, between
his use of 'Isiah Thomas' and the Piston guard.42 So the debate
between correspondence and disquotation comes down to the
question of whether or not such a theory of reference can or need be
constructed from these materials.
But notice, both parties to the debate agree that Jobe's reliability
needs explanation, and they agree on the sorts of facts that might
provide one. When a person is reliable on some subject—whether it
be Jobe on Isiah's ankle, or a geologist on Mount St Helens, or a
historian on the causes of the Industrial Revolution, or a ten-year-
old kid on his favourite rock star—that reliability needs an
explanation. We look for an account of how the person's cognitive
machinery is connected back to what she's reliable about, via what
she's read and the sources of that material, via her conversations
with others and their sources, via her observations of indicators or
instruments, and via her actual experience with her subject matter.
The disagreement is only about whether or not this welter of
material will produce a non-trivial theory of reference.
As I've said, I have no intention of taking sides on this last
question; I'm perfectly willing to let the participants reach their
own conclusions. The point that needs making here is this: even if
the disquotationalist succeeds in relieving the realist of her
dependence on correspondence truth, and thus on non-trivial
reference, the matter of explaining the various 'reliable connec-
tions' will remain. Thus, in cases where the need for a referential
connection seems to involve the mathematical realist in difficulties,
casting off the need for reference is not likely to help, because the
requirements of 'reliable connection' are almost certain to lead to
the same, or essentially similar, difficulties.43
In what follows, then, I will sometimes phrase both the
challenges to mathematical realism and my responses in terms of
42
What the correspondence theorist has in mind here is a version of the causal
theory of reference described in ch. 2, sect. 1, below.
43
As Field (1986) makes clear, from the point of view of the overall realist
project, a theory of reliable connection would probably be easier than one of full
reference; for example, reference is compositional—the reference of a whole is
thought to depend systematically on the reference of the parts—while a theory of
reliability might not require so much detailed structure. But the aspects of the theory
of reference that are thought to create difficulties for the mathematical realist are
those it seems to share with the theory of reliability, for example, concerns about
causation. See ch. 2, sect. 1, below.
20 REALISM

correspondence truth and reference, but here, again for the record, I
want to emphasize that this way of stating things is convenient but
not essential. Those realists who believe that a robust reference
relation is not needed in science, either because the disquotational-
ist is right, or for some other reason, are invited simply to recast the
discussions that follow in terms of 'reliable connection'.

One final remark on realism and truth. Some anti-realists, assuming

the realist is wedded to correspondence truth, have argued that
realism is unscientific because it requires a connection between
scientific theory and the world that reaches beyond the bounds of
science itself.44 Here the anti-realist attempts to saddle the realist
with the now-familiar unnaturalized standpoint, the point of view
that stands above, outside, or prior to, our best theories of the
world, and from which is posed the question: what connects our
theories to the world?
We've seen that in epistemology, the contemporary realist has
answered by rejecting the extra-scientific challenge itself, along
with the radical scepticism it engenders. The same goes for
semantics. There is no point of view prior to or superior to that of
natural science. What we want is a theory of how our language
works, a theory that will become a chapter of that very scientific
world-view. In order to arrive at this new chapter, it would be
madness to cast off the scientific knowledge collected so far.
Rather, we stand within our current best theory—what better
account do we have of the way the world is?—and ask for an
account of how our beliefs and our language connect up with the
world as that theory says it is. This may be the robust theory of
reference required by correspondence truth. If the disquotationalist
is right, it may be something less structured, an account of reliable
connection. But neither way is it something extra-scientific.

4. Realism in mathematics

Let me turn at last to realism in the philosophy of mathematics

proper. Most prominent in this context is a folkloric position called
'Platonism' by analogy with Plato's realism about universals. As is
44
See e.g. Putnam (1977), 125. Or, from a different point of view, Burgess
(forthcoming a).
 R E A L I S M IN MATHEMATICS 21

common with such venerable terms, it is applied to views of very

different sorts, most of them not particularly Platonic.45 Here I will
take it in a broad sense as simply synonymous with 'realism' as
applied to the subject matter of mathematics: mathematics is the
scientific study of objectively existing mathematical entities just as
physics is the study of physical entities. The statements of
mathematics are true or false depending on the properties of those
entities, independent of our ability, or lack thereof, to determine
which.
Traditionally, Platonism in the philosophy of mathematics has
been taken to involve somewhat more than this. Following some of
what Plato had to say about his Forms, many thinkers have
characterized mathematical entities as abstract—outside of
physical space, eternal and unchanging—and as existing necessarily—
regardless of the details of the contingent make-up of the physical
world. Knowledge of such entities is often thought to be a
priori—sense experience can tell us how things are, not how they
must be—and certain—as distinguished from fallible scientific
knowledge. I will call this constellation of opinions 'traditional
Platonism'.
Obviously, this uncompromising account of mathematical reality
makes the question of how we humans come to know the requisite
a priori certainties painfully acute. And the successful application
of mathematics to the physical world produces another mystery:
what do the inhabitants of the non-spatio-temporal mathematical
realm have to do with the ordinary physical things of the world we
live in? In his theory of Forms, Plato says that physical things
'participate' in the Forms, and he uses the fact of our knowledge of
the latter, via a sort of non-sensory apprehension, to argue that the
soul must pre-exist birth. 46 But our naturalized realist will hardly
buy this package.
Given these difficulties with traditional Platonism, it's not
surprising that various forms of mathematical anti-realism have
been proposed. I'll pause to consider a sampling of these views
before describing the two main schools of contemporary Platonism.
45
For example, though the term 'Platonism' suggests a realism about universals,
many Platonists regard mathematics as the science of peculiarly mathematical
particulars: numbers, functions, sets, etc. An exception is the structuralist approach
considered in ch. 5, sect. 3, below.
46
See his Phaedo 72 D-77 A.
22 REALISM

In the late 1600s, in response to a number of questions from

physical science, Sir Isaac Newton and Gottfried Wilhelm von
Leibniz simultaneously and independently invented the calculus.
Though the scientist's problems were solved, the new mathematical
methods were scandalously error-ridden and confused. Among the
most vociferous and perceptive critics was the idealist Berkeley, an
Anglican bishop who hoped to silence the atheists by showing their
treasured scientific thinking to be even less clear than theology. The
central point of contention was the notion of infinitesimals,
infinitely small amounts still not equal to zero, which Berkeley
ridiculed as 'the ghosts of departed quantities'. 47 Two centuries
later, Bolzano, Cauchy, and Weierstrass had replaced these ghosts
with the modern theory of limits. 48
This account of limits required a foundation of its own, which
Georg Cantor and Richard Dedekind provided in their theory of
real numbers, but these in turn reintroduced the idea of the
completed infinite into mathematics. No one had ever much liked
the seemingly paradoxical idea that a proper part of an infinite
thing could be in some sense as large as the whole—there are as
many even natural numbers as there are even and odd, there are as
many points on a one-inch line segment as on a two-inch line
segment—but the infinite sets introduced by Cantor and others
gave rise to outright contradictions, of which Bertrand Russell's is
the most famous: 49 consider the set of all sets that are not members
of themselves. It is self-membered if and only if it isn't. The opening
decades of this century saw the development of three great schools
of thought on the nature of mathematics, all of them designed to
deal in one way or another with the problem of the infinite.
The first of these is intuitionism, which dealt with the infinite by
rejecting it outright. The original version of this position, first
proposed by L. E. J. Brouwer, 50 was analogous to Berkeleian
4
" See Berkeley (1734), subtitled 'A Discourse Addressed to an Infidel Mathe-
matician. Wherein It is Examined Whether the Object, Principles, and Inferences of
the Modern Analysis are More Distinctly Conceived, or More Evidently Deduced,
than Religious Mysteries and Points of Faith'. The quotation is from p. 89.
48
For a more detailed description of the developments sketched in this paragraph
and the next, see Kline (1972), chs. 17, 40, 41, and 5 1, or Boyer (1949).
49
The paradox most directly associated with Cantor's work is Burali-Forti's
(1897). See Cantor's discussion (1899). Russell's primary target was Frege, as will be
noted below.
50
Brouwer (1913; 1949). Other, less opaque, expositions of this position are
Heyting(1931; 1966) and Troelstra (1969).
 R E A L I S M IN M A T H E M A T I C S 23

idealism: it takes the objects of mathematics to be mental

constructions rather than objective entities. The modern version,
defended by Michael Dummett,51 is a brand of verificationism: a
mathematical statement is said to be true if and only if it has been
constructively proved. Either way, a series of striking consequences
follow: statements that haven't been proved or disproved are
neither true nor false; completed infinite collections (like the set of
natural numbers) are illegitimate; much of infinitary mathematics
must either be rejected (higher set theory) or radically revised (real
number theory and the calculus).
These forms of intuitionism face many difficulties—e.g. does
each mathematician have a different mathematics depending on
what she's mentally constructed? how can we verify even state-
ments about large finite numbers? etc.—but its most serious
drawback is that it would curtail mathematics itself. My own
working assumption is that the philosopher's job is to give an
account of mathematics as it is practised, not to recommend
sweeping reform of the subject on philosophical grounds. The
theory of the real numbers, for example, is a fundamental
component of the calculus and higher analysis, and as such is far
more firmly supported than any philosophical theory of math-
ematical existence or knowledge. To sacrifice the former to preserve
the latter is just bad methodology.
A second anti-realist position is formalism, the popular school of
double-think mentioned above. The earliest versions of the view
that mathematics is a game with meaningless symbols played
heavily on a simple analogy between mathematical symbols and
chess pieces, between mathematics and chess, but even its advocates
were uncomfortably aware of the stark disanalogies:52
To be sure, there is an important difference between arithmetic and chess.
The rules of chess are arbitrary, the system of rules for arithmetic is such
that by means of simple axioms the numbers can be referred to perceptual
manifolds and can thus make [an] important contribution to our
knowledge of nature.
The Platonist Gottlob Frege launched a fierce assault on early
formalism, from many directions simultaneously, but the most

51
Dummett (1975; 1977).
52
Frege cites this quotation from Thomae in his critique of formalism: Frege
(1903), § 88.
24 REALISM

penetrating arose from just this point. It isn't hard to see how
various true statements of mathematics can help me determine how
many bricks it will take to cover the back patio, but how can a
meaningless string of symbols be any more relevant to the solution
of real world problems than an arbitrary arrangement of chess
pieces?
This is Frege's problem: what makes these meaningless strings of
symbols useful in applications? 53 Suppose, for example, that a
physicist tests a hypothesis by using mathematics to derive an
observational prediction. If the mathematical premiss involved is
just a meaningless string of symbols, what reason is there to take
that observation to be a consequence of the hypothesis? And if it is
not a consequence, it can hardly provide a fair test. In other words,
if mathematics isn't true, we need an explanation of why it is all
right to treat it as true when we use it in physical science.
The most famous version of formalism, the one expounded
during the period under consideration here, was David Hilbert's
programme.54 Hilbert, like Brouwer, felt that only finitary math-
ematics was truly meaningful, but he considered Cantor's theory of
sets 'one of the supreme achievements of purely intellectual human
activity' and promised, in a famous remark, that

No one shall drive us out of the paradise which Cantor has created for us.
(Hilbert (1926), 188,191)

Hilbert proposed to save infinitary mathematics by treating it

instrumentally—meaningless statements about the infinite are a
useful tool in deriving meaningful statements about the finite—but
he, unlike the scientific instrumentalists, was sensitive to the
question of how this practice could be justified. Hilbert's plan was
to give a metamathematical proof that the use of the meaningless
statements of infinitary mathematics to derive meaningful state-
ments of finitary mathematics would never produce incorrect
finitary results. The same line of thought might have applied to its
use in natural science as well, thus solving Frege's problem.
Hilbert's efforts to carry through on this project produced the rich

53
See Fregef 1903), § 91.
54
See Hilbert (1926; 1928).
 REALISM IN MATHEMATICS 25

new field of metamathematics, but Kurt Godel soon proved that its
cherished goal could not be reached.55
For all the simplicity of game formalism and the fame of Hilbert's
programme, many mathematicians, when they claim to be formal-
ists, actually have another idea in mind: mathematics isn't a science
with a peculiar subject matter; it is the logical study of what
conclusions follow from which premisses. Philosophers call this
position 'if-thenism'. Several prominent philosophers of math-
ematics have held this position at one time or another—Hilbert
(before his programme), Russell (before his logicism), and Hilary
Putnam (before his Platonism) 56 —but all ultimately rejected it. Let
me briefly indicate why.
A number of annoying difficulties plague the if-thenist: which
logical language is appropriate for the statement of premisses and
conclusions? which premisses are to be presupposed in cases like
number theory, where assumptions are usually left implicit? from
among the vast range of arbitrary possibilities, why do mathemat-
icians choose the particular axiom systems they do to study? what
were historical mathematicians doing before their subjects were
axiomatized? what are they doing when they propose new axioms?
and so on. But the question that seems to have scotched if-thenism
in the minds of Russell and Putnam was a version of Frege's
problem: how can the fact that one mathematical statement follows
from another be correctly used in our investigation of the physical
world? The general thrust of the if-thenist's reply seems to be that
the antecedent of a mathematical if-then statement is treated as an
idealization of some physical statement. The scientist then draws as
a conclusion the physical statement that is the unidealization of the
consequent.57
Notice that on this picture, the physical statements must be
entirely mathematics-free; the only mathematics involved is that
used in moving between them. Unfortunately, many of the
55
See Godel (1931). Enderton (1972), ch. 3, gives a readable presentation.
Detlefsen (1986) attempts to defend Hilbert's programme against the challenge of
Godel's theorem. Simpson (1988) and Feferman (1988) pursue partial or relativized
versions within the limitations of Godel's theorem.
56
See Resnik (1980), ch. 3, for discussion. There if-thenism is called 'deductiv-
ism'. See also Putnam (1979), p. xiii. Russell's logicism and Putnam's Platonism will
be considered below.
57
See Korner (1960), ch. 8. Cf. Putnam (1967b), 33.
26 REALISM

statements of physical science seem inextricably mathematical. To

quote Putnam, after his conversion:
one wants to say that the Law of Universal Gravitation makes an objective
statement about bodies—not just about sense data or meter readings.
What is the statement? It is just that bodies behave in such a way that the
quotient of two numbers associated with the bodies is equal to a third
number associated with the bodies. But how can such a statement have any
objective content at all if numbers and 'associations' (i.e. functions) are
alike mere fictions? It is like trying to maintain that God does not exist and
angels do not exist while maintaining at the very same time that it is an
objective fact that God has put an angel in charge of each star and the
angels in charge of each of a pair of binary stars were always created at the
same time! If talk of numbers and 'associations' between masses, etc. and
numbers is 'theology' (in the pejorative sense), then the Law of Universal
Gravitation is likewise theology. (Putnam (19756), 74-5)

In other words, the if-thenist account of applied mathematics

requires that natural science be wholly non-mathematical, but it
seems unlikely that science can be so purified. 58
The third and final anti-realist school of thought I want to
consider here is logicism, or really, the version of logicism advanced
by the logical positivists. Frege's original logicist programme aimed
to show that arithmetic is reducible to pure logic, that is, that its
objects—numbers—are logical objects and that its theorems can be
proved by logic aione.^ 9 This version of logicism is outright
Platonistic: arithmetic is the science of something objective (be-
cause logic is objective), that something objective consists of objects
(numbers), and our logical knowledge is a priori. If this project had
succeeded, the epistemological problems of Platonism would have
been reduced to those of logic, presumably a gain. But Frege's
project failed; his system was inconsistent. 60 Russell and White-
head took up the banner in their Principia Mathematics but were
forced to adopt fundamental assumptions no one accepted as
58
Hartry Field's ambitious attempt to do this will be considered in ch. 5, sect. 2,
below. See Field (1980; 1989).
59
See Frege( 1884).
60
The trouble was the original version of Russell's paradox. (See Russell's letter
to Frege, Russell (1902),) Frege's numbers were extensions of concepts. (See ch. 3
below.) Some concepts, like 'red', don't apply to their extensions, others, like
'infinite', do. Russell considered the extension of the concept 'doesn't apply to its
own extension'. If it applies to its own extension then it doesn't, and vice versa. This
contradiction was provable from Frege's fundamental assumptions. There have been
efforts to revive Frege's system; see e.g. Wright (1983) and Hodes (1984).
 R E A L I S M IN MATHEMATICS 27

purely logical.61 Eventually, Ernst Zermelo (aided by Mirimanoff,

Fraenkel, Skolem, and von Neumann) produced an axiom system
that showed how mathematics could be reduced to set theory,62 but
again, no one supposed that set theory enjoys the epistemological
transparency of pure logic.
Still, the idea that mathematics is just logic was not dead; it was
taken up by the positivists, especially Rudolf Carnap.63 For these
thinkers, however, there are no logical objects of any kind, and the
laws of logic and mathematics are true only by arbitrary conven-
tion. Thus mathematics is not, as the Platonist insists, an objective
science. The advantage of this counterintuitive view is that
mathematical knowledge is easily explicable; it arises from human
decisions. Question: Why are the axioms of Zermelo—Fraenkel
true? Answer: Because they are part of the language we've adopted
for using the word 'set'.
This conventionalist line of thought was subjected to a historic
series of objections by Carnap's student, W. V. O. Quine.64 The key
difficulty is that both mathematical and physical assumptions are
enshrined in Carnap's official language. How are we to separate the
conventionally adopted mathematical part of the language from the
factually true physical hypotheses? Quine argues that it isn't
enough to say that the scientific claims, not the mathematical ones,
are supported by empirical data:
The semblance of a difference in this respect is largely due to overemphasis
of departmental boundaries. For a self-contained theory which we can
check with experience includes, in point of fact, not only its various
theoretical hypotheses of so-called natural science but also such portions of
logic and mathematics as it makes use of. (Quine (1954), 367)
Mathematics is part of the theory we test against experience, and a
successful test supports the mathematics as much as the science.
Carnap makes several efforts to separate mathematics from
natural science, culminating in his distinction between analytic and
synthetic. Mathematical statements, he argues, are analytic, that is,
61
See Russell and Whitehead (1913).
62
Zermelo's first presentation is Zermelo (19086). See also Mirimanoff (1917d,
b), Fraenkel (1922), Skolem (1923), and von Neumann (1925). The standard
axioms are now called 'Zermelo-Fraenkel set theory' or ZFC (ZF when the axiom of
choice is omitted). See Enderton (1977), 271-2.
63
See Carnap (1937; 1950).
64
See Quine (1936; 1951; 1954).
28 REALISM

true by virtue of the meanings of the words involved (the logical

and mathematical vocabulary); scientific statements, on the other
hand, are synthetic, true by virtue of the way the world is. Quine
examines this distinction in great detail, investigating various
attempts at clear formulation, and concludes:
It is obvious that truth in general depends on both language and
extralinguistic fact. The statement 'Brutus killed Caesar' would be false if
the world had been different in certain ways, but it would also be false if
the word 'killed' happened rather to have the sense of 'begat'. Thus one is
tempted to suppose in general that the truth of a statement is somehow
analyzable into a linguistic component and a factual component. Given this
supposition, it next seems reasonable that in some statements the factual
component should be null; and these are the analytic statements. But, for
all its a priori reasonableness, a boundary between analytic and synthetic
statements simply has not been drawn. That there is such a distinction to be
drawn at all is an unempirical dogma of empiricists, a metaphysical article
of faith. (Quine (1951), 36-7)

Without a clear distinction between analytic and synthetic, Carnap's

anti-Platonist version of logicism fails.
I will leave the three great schools at this point. I don't claim to
have refuted either formalism or conventionalism, though I hope
the profound difficulties they face have been drawn clearly enough.
Intuitionism I reject on the grounds given above; I assume that the
job of the philosopher of mathematics is to describe and explain
mathematics, not to reform it.

Let me return now to Platonism, the view that mathematics is an

objective science. Platonism naturally conflicts with each of the
particular forms of anti-realism touched on here—with intuition-
ism on the objectivity of mathematical entities, with formalism on
the status of infinitary mathematics, with logicism on the need for
mathematical existence assumptions going beyond those of logic—
but the Platonist's traditional and purest opponent is the nominal-
ist, who simply holds that there are no mathematical entities. (The
term 'nominalism' has followed 'Platonism' in its migration from
the debate over universals into the debate over mathematical
entities.) Two forms of Platonism dominate contemporary debate.
The first of these derives from the work of Quine and Putnam
sketched above—their respective criticisms of conventionalism and
if-thenism—and the second is described by Godel as the philo-
 R E A L I S M IN M A T H E M A T I C S 29

sophical underpinning for his famous theorems.65 As Quine and

Putnam's writings have just been discussed, let me begin with them.
Quine's defence of mathematical realism follows directly on the
heels of the defences of common-sense and scientific realism
sketched above. On the naturalized approach, we judge what
entities there are by seeing what entities we need to produce the
most effective theory of the world. So far, these include medium-
sized physical objects and the theoretical entities of physical
science, and so far, the nominalist might well agree. But if we
pursue the question of mathematical ontology in the same spirit,
the nominalist seems cornered:
A platonistic ontology . . . is, from the point of view of a strictly physicai-
istic conceptual scheme, as much a myth as that physicalistic conceptual
scheme itself is for phenomenalism. This higher myth is a good and useful
one, in turn, in so far as it simplifies our account of physics. Since
mathematics is an integral part of this higher myth, the utility of this myth
for physical science is evident enough. (Quine (1948), 18)
If we countenance an ontology of physical objects and unobserv-
ables as part of our best theory of the world, how are we to avoid
countenancing mathematical entities on the same grounds? Carnap
suggested what Quine calls a 'double standard'66 in ontology,
according to which questions of mathematical existence are
linguistic and conventional and questions of physical existence are
scientific and real, but we've already seen that this effort fails.
We've also seen that Putnam takes the same thinking somewhat
further, emphasizing not only that mathematics simplifies physics,
but that physics can't even be formulated without mathematics:67
'mathematics and physics are integrated in such a way that it is not
possible to be a realist with respect to physical theory and a
nominalist with respect to mathematical theory' (Putnam (1975 b),
74). He concludes that talk about 68
mathematical entities is indispensable for science . . . therefore we should
65
See his letters to Wang, quoted in Wang (19746), 8-11, and Feferman's
discussion (19846).
66
Quine (1951), 45.
67
See the long quotation from Putnam (19756) above. A more complete account
appears in Putnam (1971), esp.§§ 5 and 7.
68
He really says 'quantification over', which derives from Quine's official
criterion of ontological commitment (1948), but I don't want to get into the debate
over that precise formulation.
30 REALISM

accept such [talk]; but this commits us to accepting the existence of the
mathematical entities in question. This type of argument stems, of course,
from Quine, who has for years stressed both the indispensability of [talk
about] mathematical entities and the intellectual dishonesty of denying the
existence of what one daily presupposes. (Putnam (1971), 347)

We are committed to the existence of mathematical objects because

they are indispensable to our best theory of the world and we
accept that theory.
The particular brand of Platonism that arises from these Quine/
Putnam indispensability arguments has some revolutionary features.
Recall that traditional Platonism takes mathematical knowledge to
be a priori, certain, and necessary. But, if our knowledge of
mathematical entities is justified by the role it plays in our
empirically supported scientific theory, that knowledge can hardly
be classified as a priori. 69 Furthermore, if we prefer to. alter our
scientific hypotheses rather than our mathematical ones when our
overall theory meets with disconfirmation, it is only because the
former can usually be adjusted with less perturbation to the theory
as a whole.70 Indeed, Putnam/} goes so far as to suggest that the
best solution to difficulties in quantum mechanics may well be to
alter our logical laws rather than any physical hypotheses. Thus the
position of mathematics as part of our best theory of the world
leaves it as liable to revision as any other part of that theory, at least
in principle, so mathematical knowledge is not certain. Finally, the
case of necessity is less clear, if only because Quine rejects such
modal notions out of hand, but the fact that our mathematics is
empirically confirmed in this world surely provides little support
for the claim that it is likely to be true in some other possible
circumstance. So Quine/Putnam Platonism stands at some consider-
able remove from the traditional variety.
But while disagreement with a venerable philosophical theory is
no clear demerit, disagreement with the realities of mathematical
practice is. First, notice that unapplied mathematics is completely
without justification on the Quine/Putnam model; it plays no
indispensable role in our best theory, so it need not be accepted:72
69
See Putnam (19756) for an explicit discussion of a posteriori methods in
mathematics. Kitcher (1983) attacks the idea that mathematics is a priori from a
different angle.
70
See Quine (1951), 43^.
71
Putnam (1968).
72
See also Putnam (1971), 346-7.
 R E A L I S M IN M A T H E M A T I C S 31

So much of mathematics as is wanted for use in empirical science is for me

on a par with the rest of science. Transfinite ramifications are on the same
footing insofar as they come of a simplificatory rounding out, but anything
further is on a par rather with uninterpreted systems. (Quine (1984), 788)
Now mathematicians are not apt to think that the justification for
their claims waits on the activities in the physics labs. Rather,
mathematicians have a whole range of justificatory practices of
their own, ranging from proofs and intuitive evidence, to plausibil-
ity arguments and defences in terms of consequences. From the
perspective of a pure indispensability defence, this is all just so
much talk; what matters is the application.
If this weren't enough to disqualify Quine/Putnamism as an
account of mathematics as it is practised, consider one last point. In
this picture of our scientific theorizing, mathematics enters only at
fairly theoretical levels. The most basic evidence takes the form of
non-mathematical observation sentences—e.g. 'this chunk of gold
is malleable'—and the initial levels of theory consist of non-
mathematical generalizations—"gold is a malleable metal'. Math-
ematics only enters the picture at the more theoretical levels—'gold
has atomic number 79'—so it is on an epistemic par with this
higher-level theory.73 But isn't it odd to think of '2 + 2 = 4' or 'the
union of the set of even numbers with the set of odd numbers is the
set of all numbers' as highly theoretical principles? In Charles
Parsons's phrase, Quine/Putnamism 'leaves unaccounted for pre-
cisely the obviousness of elementary mathematics'.74
By way of contrast, the Godelian brand of Platonism takes its
lead from the actual experience of doing mathematics, which he
takes to support Platonism as suggested in section 1 above. For
Godel, the most elementary axioms of set theory are obvious; in his
words, they 'force themselves upon us as being true'.75 He accounts
for this by positing a faculty of mathematical intuition that plays a
role in mathematics analogous to that of sense perception in the
physical sciences, so presumably the axioms force themselves upon
us as explanations of the intuitive data much as the assumption of
medium-sized physical objects forces itself upon us as an explana-
tion of our sensory experiences. To push this analogy, recall that
this style of argument for common-sense realism might have been
73
See Quine (1948), 18-19.
74
Parsons (1979/80), 151. See also Parsons (19836).
75
Godel (1947/64), 484.
32 REALISM

undercut if phenomenalists had succeeded in giving non-realistic

translations of our physical object statements. Similarly, Godel
notes that Russell's 'no-class' interpretation of Principia was an
effort to do the work of set theory, that is, to systematize all of
mathematics, without sets. Echoing the common-sense realist,
Godel takes the failure of Russell's project as support for his
mathematical realism:
This whole scheme of the no-class theory is of great interest as one of the
few examples, carried out in detail, of the tendency to eliminate
assumptions about the existence of objects outside the 'data' and to replace
them by constructions on the basis of these data.76 The result has been in
this case essentially negative . . . All this is only a verification of the view
defended above that logic and mathematics (just as physics) are built up on
axioms with a real content which cannot be 'explained away'. (Godel
(1944), 460-1)

He concludes that
the assumption of [sets] is quite as legitimate as the assumption of physical
bodies and there is quite as much reason to believe in their existence. They
are in the same sense necessary to obtain a satisfactory system of
mathematics as physical bodies are necessary for a satisfactory theory of
our sense perceptions . . . (Godel (1944), 456-7)

But this analogy of intuition with perception, of mathematical

realism with common-sense realism, is not the end of Godel's
elaboration of the mathematical realist's analogy between math-
ematics and natural science. Just as there are facts about physical
objects that aren't perceivable, there are facts about mathematical
objects that aren't intuitable. In both cases, our belief in such
'unobservable' facts is justified by their role in our theory, by their
explanatory power, their predictive success, their fruitful inter-
connections with other well-confirmed theories, and so on. In
Godel's words:
even disregarding the [intuitiveness] of some new axiom, and even in case it
has no [intuitiveness] at all, a probable decision about its truth is possible
also in another way, namely, inductively by studying its 'success'. . . .
There might exist axioms so abundant in their verifiable consequences,
76
In this passage, 'data' means 'logic without the assumption of the existence of
classes' (Godel (1944), 460 n. 22). Earlier in this same paper, Godel refers to arithmetic
as 'the domain of the kind of elementary indisputable evidence that may be most
fittingly compared with sense perception' (p. 449).
 REALISM IN MATHEMATICS 33

shedding so much light upon a whole field, and yielding such powerful
methods for solving problems . . . that, no matter whether or not they are
[intuitive], they would have to be accepted at least in the same sense as any
well-established physical theory. (Godel (1947/64), 477)
Quite a number of historical and contemporary justifications for set
theoretic hypotheses take this form, as will come out in Chapter 4.
Here the higher, less intuitive, levels are justified by their
consequences at lower, more intuitive, levels, just as physical
unobservables are justified by their ability to systematize our
experience of observables. At its more theoretical reaches, then,
GodePs mathematical realism is analogous to scientific realism.
Thus GodePs Platonistic epistemology is two-tiered: the simpler
concepts and axioms are justified intrinsically by their intuitiveness;
more theoretical hypotheses are justified extrinsically, by their
consequences. This second tier leads to departures from traditional
Platonism similar to Quine/Putnam's. Extrinsically justified hypo-
theses are not certain,77 and, given that Godel allows for
justification by fruitfulness in physics as well as in mathematics,78
they are not a priori either. But, in contrast with Quine/Putnam,
Godel gives full credit to purely mathematical forms of justification
—intuitive self-evidence, proofs, and extrinsic justifications within
mathematics—and the faculty of intuition does justice to the
obviousness of elementary mathematics.
Among GodePs staunchest critics is Charles Chihara.79 Even if
Godel has succeeded in showing that the case for the existence of
mathematical entities runs parallel to the case for the existence of
physical ones, Chihara argues that he has by no means shown that
the two cases are of the same strength, and thus, that he has not
established that there is as much reason to believe in the one as to
believe in the other.80 Furthermore, Chihara argues, the existence
of mathematical entities is not required to explain the experience of
mathematical intuition and agreement:
I believe it is at least as promising to look for a naturalistic explanation
based on the operations and structure of the internal systems of human
beings. (Chihara (1982), 218)

77
Godel (1944), 449.
78
Godel (1947/64), 485.
79
See Chihara (1973), ch. 2; (1982).
80
Chihara (1982), 213-14.
34 REALISM

. . . mathematicians, regarded as biological organisms, are basically quite

similar. (Chihara (1973), 80)
And finally, he questions whether Godel's intuition offers any
explanation at all: 81
the 'explanation' offered is so vague and imprecise as to be practically
worthless: all we are told about how the 'external objects' explain the
phenomena is that mathematicians are 'in some kind of contact' with these
objects. What empirical scientist would be impressed by an explanation
this flabby? (Chihara (1982), 217)
Now the Godelian Platonist is not entirely defenceless in the face
of this attack. For example, Mark Steiner82 points out that
Chihara's 'explanation' is likewise lacking in muscle tone: the
similarity of human beings as organisms can hardly explain their
agreement about mathematics when it is consistent with so much
disagreement on other subjects. Still, most observers tend to agree
that no appeal to purported human experiences of xs that underlie
our theory of xs can justify a belief in the existence of xs unless we
have some independent reason to think our theory of xs is true. 83
Thus the purported human dealings with witches that underlie our
theory of witches don't justify a belief in witches unless we have
some independent reason to think that our theory of witches is
actually correct.
But notice: we have recently rehearsed just such an independent
reason in the case of mathematics, namely, the indispensability
arguments of Quine and Putnam. Unless endorsing these commits
one to the view that there is no peculiarly mathematical form of
evidence—and I don't see why it should 84 —there is room for an
attractive compromise between Quine/Putnam and Godelian Platon-
ism. It goes like this: successful applications of mathematics give
us reason to believe that mathematics is a science, that much of it at
least approximates truth. Thus successful applications justify, in a
general way, the practice of mathematics. But, as we've seen, this
isn't enough to give an adequate account of mathematical practice,

K1
These remarks of Chihara's are actually addressed to a quotation from Kreisel,
but it is clear from the context that he thinks the same objection applies to Godel's
intuition.
82
Steiner (19756), 190.
8
-' See Steiner (19756), 190. For a similar sentiment, see Putnam (19756), 73-4.
84
Nor does Parsons (19836), 192-3.
 REALISM IN MATHEMATICS 35

of how and why it works. We still owe an account of the

obviousness of elementary mathematics, which Godel's intuition is
designed to provide, and an account of other purely mathematical
forms of evidence, like proof and various extrinsic methods. This
means we need to explain what intuition is and how it works; we
need to catalogue extrinsic methods and explain why they are
rational methods in the pursuit of truth.
From Quine/Putnam, this compromise takes the centrality of the
indispensability arguments; from Godel, it takes the recognition of
purely mathematical forms of evidence and the responsibility for
explaining them. Thus it averts a major difficulty with Quine/
Putnamism—its unfaithfulness to mathematical practice—and a
major difficulty with Godelism—its lack of a straightforward
argument for the truth of mathematics. But whatever its merits,
compromise Platonism does nothing to remedy the flabbiness of
Godel's account of intuition. And it is in this neighbourhood that
many contemporary objections to Platonism are concentrated.85

I opened this chapter with the hope of reinstating the mathemat-

ician's pre-philosophical realism, of devising a defensible refinement
of that attitude that remains true to the phenomenology of practice.
Along the way, I've sided with common-sense realism, scientific
realism, and philosophical naturalism, and seconded many of the
advances of Quine/Putnam and Godelian Platonism. It will come as
no surprise, then, that the position to be defended here is a version
of compromise Platonism. I'll call it 'set theoretic realism'.
Chapter 2 outlines a naturalistic epistemology for items located
on the lower tier of Godel's two-tiered epistemology, a replacement
for Godel's intuition. The ontological question of the relationship
between sets and other mathematical entities, particularly natural
and real numbers, is the subject of Chapter 3. Chapter 4 contains
some preliminary spadework on the problem of theoretical
justification, the second of Godel's two tiers. I argue that this ill-
understood problem is the most important open question of our
day, not only for set theoretic realism, but for many other
mathematical philosophies as well. Chapter 5 takes a final look at
set theoretic realism from physicalist and structuralist perspectives.
s5
See ch. 2, sect. 1, below.
 2
PERCEPTION AND
INTUITION

1. What is the question?

The general outlines of the epistemological challenge to Platonism
have already been hinted at, but I'd like now to place the problem
in the context of contemporary philosophy. The sense that there is a
problem goes back, as we've seen, to Plato himself, but the modern
form, the one exhaustively discussed in the contemporary litera-
ture, derives from Paul Benacerraf s 'Mathematical truth', which
appeared in the early seventies. 1 Since then, it has become
commonplace for scholarly writings on the philosophy of mathe-
matics to begin by dismissing Platonism on the basis of Benacerraf s
argument. Benacerraf himself draws no such dogmatic conclusion,
but his successors, even those with generally realistic leanings, have
scorned Platonism. 2
The Benacerrafian syllogism rests on two premisses. The second
is a traditional Platonistic account of the nature of mathematical
entities as abstract, in particular, as non-spatio-temporal. The first
premiss concerns the nature of human knowledge: what is it for me
to know something? It was originally suggested, again by Plato,3
that it is enough that I believe it, that my belief be justified, and that
the belief be true. Though Plato raised some objections of his own
to this 'justified, true belief account of knowledge, it wasn't until
1963 that Edmund Getder pointed out what is now considered its
fatal weakness.4
Suppose I see Dick driving a Hillman; suppose he offers me a ride
1
Benacerraf (1973).
2
The anti-Platomsms of Field (1980), Bonevac (1982), Gottlieb (1980), and
Hellman (1989), are all at least partly motivated by Benacerrafian considerations.
This style of argument is also noted with approval by Kitcher (1983), 59, and Resmk
(1981), 529, (1982), 95, and (forthcoming a, b}.
3
See his Theaetetus, 202 c.
4
Gettier(1963).
 WHAT IS THE Q U E S T I O N ? 37

to work in this car. On the basis of this experience, I come to

believe that Dick owns a Hillman. My belief that Dick owns a
Hillman is surely justified—he gave me a lift in one—and let us
further suppose that it is true—that Dick does indeed own a
Hillman. But—and here's the catch—he doesn't own this Hillman.
The Hillman Dick actually owns is in the shop, as it often is, and
this one, the one I saw, the one I rode in, was borrowed from Frank.
In this case, though I have a justified, true belief that he does, I can't
be said to know Dick owns a Hillman. For knowledge, there is
some further requirement.
Some years after Gettier's paper came a response from Alvin
Goldman,5 diagnosing the problem in cases like mine and Dick's,
and proposing a fourth clause in the definition of knowledge to cure
it. The difficulty, according to Goldman and many others who
largely agreed with him,6 is that Dick's Hillman was not the car
that caused me to believe as I did. For a justified, true belief to count
as knowledge, what makes the belief true must be appropriately 7
causally responsible for that belief. This idea, in its many versions,
is called the 'causal theory of knowledge'.
The two premisses, then, of our Benacerrafian argument are the
causal theory of knowledge and the abstractness of mathematical
objects. What makes '2 + 2 = 4' true is the nature of the abstract
entities 2 and 4 and the operation plus; for me to know that '2 +
2 = 4', those entities must play an appropriate causal role in the
generation of my belief. But how can entities that don't even inhabit
the physical universe take part in any causal interaction whatso-
ever? Surely to be abstract is also to be causally inert. Thus, if Platon-
ism is true, we can have no mathematical knowledge. Assuming that
we do have such knowledge, Platonism must be false.
This dramatic conclusion can be pushed further by recent
progress in the theory of reference.8 How does a name pick out a
thing? In this field, the classical theory is Frege's:9 a name is
associated with a description that uniquely identifies the thing the
name names; for example, 'Isiah Thomas' is associated with the
description 'best friend of Magic Johnson'. Numerous variations on
5
Goldman (1967).
6
See e.g. Skyrms (1967) or Barman (1973).
7
Suppose, by some neural fluke, my justified true belief is caused by my being hit
on the head by Dick's Hillman. This would be inappropriate,
8
See Lear (1977).
9
See Frege( 18926).
38 P E R C E P T I O N AND I N T U I T I O N

this idea have been proposed—that there are in fact many

descriptions associated with the name, that some of these might
even be false, that what counts is the truth of a sufficient number of
the more important ones, and so on 10 -—but the central descriptive
character of the referring relation remains.
In 1972, Saul Kripke 11 called this account into question. Suppose
I hear the name 'Einstein' used repeatedly in discussions to which I
am not very attentive, and I come to believe only one thing about
him, namely, that he invented the atom bomb. Of course, Einstein
didn't invent the atom bomb, but this is nevertheless the one and
only description I associate with the name. On the description
theory, my use of the name should refer to someone else, or to no
one, if no single person invented the bomb. That this isn't the case is
made clear by the reaction of my physicist friend who insists that
I'm dead wrong in my belief about Einstein. If the description
theory were correct, I'd have made a true statement about someone
else—the person who did invent the bomb—or a truth-valueless
statement about nobody—if no one person invented the bomb—
but in fact I made a false statement about Einstein himself.
Kripke and others 12 react to this problem by proposing a very
different picture of how we refer. My use of the name 'Einstein'
picks out Einstein, not by virtue of my knowledge of some uniquely
identifying description of the man, but because my usage is
borrowed from those I heard using it, theirs in turn borrowed from
their teachers or from books, the usage there borrowed from
someone else's, in a chain leading back, ultimately, to someone who
was in a position to dub the actual individual. Thus my use of
'Einstein' refers to Einstein, despite my ignorance, because it is part
of a network of borrowed usage that extends from me back to a
somewhat imaginary event called an 'initial baptism' in which
Einstein himself participated.
A similar story works for scientific general terms like 'gold': a
chain of communication leads back to an event in which the baptist
isolated some samples of the metal and declared that 'this and stuff
like it is gold'. Thus the scientific community was able to refer to

10
See e.g. Searle (1958), Strawson (1959), ch. 6.
11
Kripke (1972). For further discussion of the following and other examples, see
Devitt (1981), 13-20, and Salmon (1981), 23-32.
12
Most notably, Putnam (1975*), chs. 11-13. See also Devitc (1981) and
references cited there.
 WHAT IS THE QUESTION? 39

gold via a direct connection with samples even before it knew

enough about atomic weights to give a uniquely identifying
description. And successive, very different scientific theories can be
about the same things, because the referents are picked out by
chains leading back to the dubbing of samples, not by the very
different, often erroneous, descriptions the competing theories
espouse.
Of course not all general terms fit this picture. 'Bachelor', for
example, refers to whatever satisfies the description 'unmarried
male', not to things more or less like Marcel Proust in some yet-to-
be-discovered respect. The theory works, not when we have an
explicit description or definition in mind for which our term is an
abbreviation, but when we notice a similarity between various
things, dub these and things like them by some term, and then set
out to discover the underlying traits that make these things what
they are. Such groupings, common in science, are called 'natural',
as opposed to 'nominal', kinds.13
The kind consisting of all mathematical objects seems unlikely to
be nominal, because available descriptions tend to be blatantly
circular ones like 'what mathematicians study'. Rather, in picking
out the kind, we get our point across by examples: mathematical
objects are numbers, sets, functions, Hilbert spaces, and things like
that. But, some might argue, if all mathematics is reducible to set
theory, or if we simply restrict our attention to set theory, there is a
simple definition after all, namely, that a set is a thing that occurs in
the iterative hierarchy. 14
Two things scotch this suggestion. First, the definition in terms of
the iterative conception is still circular; we have to know what a set
is, indeed what an arbitrary subset is, before we can understand it.
Considering this problem, I suppose there's little need for a second
objection, but I want to point out that the iterative conception is a

13
These natural kinds are the natural collections mentioned in connection with
universals in ch. 1, sect 2, above. See Quine (I969d), and Ayers (1981) for very
different discussions of these ideas.
14
The iterative hierarchy is arranged in stages. The first stage consists of
whatever individuals we begin with. (In pure set theory, this is the empty set.) The
second stage consists of the subsets of the first; the third of the subsets of the first and
second; and so on. The first infinite stage, stage o>, consists of all the sets generated at
the finite levels. Stage co + 1 includes all subsets of stage <o. And so on. Enderton
(1977), 7-9, gives an informal introduction; for a more complete discussion, see
Boobs (1971) or Shoenfield (1977).
40 P E R C E P T I O N AND I N T U I T I O N

theory of what sets are in much the same sense as 'having atomic
number 79' is a theory of what gold is. Sets of natural numbers and
point sets were considered long before Zermelo proposed his 1908
axiomatization, and the iterative picture was first described only
years after that, in 1930.15 In both cases, we start with samples,
dub the kind, then go on to investigate the nature of that kind; in
both the cases, the natural kind as identified by samples pre-dated
the scientific theory and would survive its demise. If our theory of
the atomic structure of gold turned out to be incorrect, we could go
on to form another theory of the same stuff; similarly, theoretical
considerations might lead us to drop the claim that all sets occur in
the iterative hierarchy and to study non-well-founded sets as well.16
If the most basic mathematical kinds are natural rather than
nominal, the referential difficulty for Platonism arises when we
consider the nature of the required initial baptism. In the simplest
case, we imagine the baptist standing in front of a number of
samples of the natural kind in question and declaring that 'these
and things like them are . . .'. In the case of gold, there is a direct
causal interaction with the samples, that is, they are touched, seen,
perceived. Because the links in the chain of communication are
causal, and the direct connection between the baptist and the
samples is causal, 17 this theory is often called 'the causal theory of
reference'. Running parallel to Benacerraf's epistemological di-
lemma, we have two premisses—the causal theory of reference and
the abstractness, and hence causal inertness, of mathematical
objects—that lead to another unpalatable conclusion for the
Platonist: we can't refer to mathematical objects. If mathematical
reality is as the Platonist says it is, we are doomed not only to

15
See Zermelo (1908fr) for the first explicit statement of set theoretic axioms.
The iterative conception is described in Zermelo (1930), though some (see Wang
(1974a)) give prior credit to Mirimanoff (1917<2, b).
16
Indeed, the axiom of foundation, which restricts the range of set theoretic
study to the members of the iterative hierarchy, is often taken not as a truth, but as a
simplifying assumption. See Maddy (1988a), § 1.2. On the possibility of non-well-
founded sets, see Aczel (1988).
17
Kripke allows for baptism by description as well as by ostension—for
example, 'Gold is the stuff found in Fort Knox'—but an attempted descriptive
baptism of sets would necessarily run in a circle—'Sets are things like the set of my
two hands'. Besides which, Devitt has argued (1981, pp. 36—42) that when a
description is used referentially rather than attributively—as it is in a descriptive
baptism—a causal grounding is still required. (Kripke's own attitude is not
altogether clear. See Kim (1977), 615-17.)
 WHAT IS THE Q U E S T I O N ? 41

ignorance, but to silence as well. And again, assuming that we can

talk about mathematical objects, Platonism must be false.
These, then, are the serious epistemological challenges many
philosophers take to have sunk the Platonist's ship. Some have
replied that the causal theories are irrelevant to mathematics,
because they are theories of a posteriori, contingent knowledge and
mathematical knowledge is a priori and necessary, but this sort of
response is of no use to the compromise Platonist who follows
Quine in questioning these distinctions.18 Let me take a moment
now to examine the cogency of the causal arguments from this
perspective.

We are faced with two anti-Platonist arguments. These arguments

seem to depend on causal theories of knowledge and reference,
along with the traditional Platonistic account of the nature of
mathematical entities. Thus one pro-Platonist approach would be
to call the causal theories into question. In fact, this approach was
adopted early on, by Mark Steiner, in a paper that appeared in the
same year as Benacerraf's.19
Steiner's argument is in two parts. First, he finds fault with
several particular formulations of the causal theory and concludes
that even its best formulation is implausible. He then argues that a
suitably generalized 'causal theory' allows that a fact about
numbers, for example, might play a role in the causal explanation
of my belief in the corresponding axiom of number theory after all,
simply because the axioms of number theory and analysis will all
figure in any such explanation.20 He summarizes: 'the most
plausible version of the causal theory of knowledge admits
Platonism, and the version most antagonistic to Platonism is
implausible' (Steiner (197 5 a), 116).
Now there is room for rebuttal to both parts of Steiner's claim.
For example, the case Steiner uses to sink the 'best' version of the
18
See Wright (1983), § xi, and Hale (1987), 86-90, for the a priorist line of
thought, and Lewis (1986), 108—15, for the necessitarian. It might also be argued
that the causal condition isn't required for mathematical knowledge because there is
no such thing as a mathematical Gettier case, that is, a mathematical case in which
the subject has a justified true belief without knowledge. But such cases can arise, if
not for all compromise Platonists, at least for the set theoretic realist. See sect. 2
below.
19
See Steiner (1973), which later became ch. 4 of his (1975a).
20
This thinking rests on the Quinean idea that mathematics will be part of the
overall theory that causally explains my belief in the axioms of number theory.
42 P E R C E P T I O N AND I N T U I T I O N

causal theory involves inferential knowledge, but Benacerrafs

version of the causal theory doesn't require that inferential
knowledge meet the causal condition placed on more basic, non-
inferential knowledge. 21 On the other hand, again for example,
there is no guarantee that the particular axiom of number theory I
believe will in fact figure in the causal explanation of that belief,22
and even if it did, would it figure 'appropriately'? But most
interesting for our assessment of the problem is the strong reaction
of one reviewer:
it is a crime against the intellect to try to mask the problem of naturalizing
the epistemology of mathematics with philosophical razzle-dazzle. Super-
ficial worries about the intellectual hygiene of causal theories of knowledge
are irrelevant to and misleading from this problem, for the problem is not
so much about causality as about the very possibility of natural knowledge
of abstract objects. (Hart (1977), 125-6)

If the causal theory is not the problem, then attacking the causal
theory doesn't help. But if the causal theories of knowledge and
reference are removed from the premisses of the anti-Platonist
arguments, what will take their place?
The idea that the causal theories are not the problem gains
support from the historical facts: the causal theory of knowledge
has gradually lost favour in the years since the appearance of
Benacerrafs article, while the sentiment that there is a persuasive
Benacerraf-style argument against Platonism remains strong. One
might try to pin down this new argument by combing the
contemporary epistemological literature for a descendant of the
causal theory that could take its place in the first premiss. The best
candidate would be reliabilism: for my justified, true belief to count
as knowledge, it must be generated by a reliable process. Depending
on how this account is spelled out, it may or may not involve a
causal constraint strong enough to do the causal theory's job in the
argument against Platonism. 23
I won't go into the details of current reliabilist epistemology
because in fact I don't think the force of Benacerraf-style thinking
depends on this particular philosophical epistemology any more
than it depended on the causal theory. To the extent that these are
21
SeeBenacerraf(1973),413.
22
Hart makes this point (1977, p. 124).
23
In Maddy (1984a), I indicate that reliabilism might not do this job. Casullo
(forthcoming) gjves a more refined account that suggests otherwise.
 WHAT IS THE Q U E S T I O N ? 43

intended as a priori philosophical theories of what knowledge or

justification consists in, any broad sceptical conclusion based on
them—e.g. that mathematics is not a science—errs against the
tenets of epistemology naturalized. To the extent that reliabilism
and the rest are proposals for naturalized accounts of what
knowledge is, given the overwhelming evidence in favour of
mathematical knowledge, they will not last long as parts of our best
theory if they purport to rule it out.24 Indeed, for all we know, from
the naturalized perspective, the very notions of knowledge and/or
justification might be ultimately dispensable. But for all that, I think
a Benacerraf-style worry would remain.
To see this, recall the discussion of truth and reference in the
third section of Chapter 1. The current debate between supporters
of correspondence and disquotational truth suggests that even if the
correspondence theorist's substantive notions of truth and refer-
ence turn out to be scientifically dispensable, there will remain a
problem of explaining the reliability of an expert's beliefs about the
field of her expertise. Now I want to add a similar assessment of the
situation in the theory of knowledge. Even if reliabilism turns out
not to be the correct analysis of knowledge and justification,
indeed, even if knowledge and justification themselves turn out to
be dispensable notions, there will remain the problem of explaining
the undeniable fact of our expert's reliability. In particular, even
from a completely naturalized perspective, the Platonist still owes
us an explanation of how and why Solovay's beliefs about sets are
reliable indicators of ihe truth about sets.25
The nominalist Hartry Field, realizing that the causal theory of
knowledge is something 'almost no one believes any more',26 ends
up rephrasing the Benacerraf worry in very similar terms: ' . . .
Benacerraf s challenge . . . is to provide an account of the mechan-
isms that explain how our beliefs about these remote entities can so
well reflect the facts about them.' Field combines this with the
traditional Platonist's conception of mathematical entities:
The relevant facts about how the platonist conceives of mathematical
objects include their mind-independence and language-independence; the
fact that they bear no spatio-temporal relations to us; the fact that they do

24
See Burgess (1983), 101.
25
R. M, Solovay is one of our leading contemporary set theorists.
26
All the quotations in this paragraph come from Field (1989), 25-7.
44 P E R C E P T I O N AND I N T U I T I O N

not undergo any physical interactions (exchanges of energy-momentum

and the like) with us or anything we can observe; etc.
From these two premisses, he draws a guarded pro-nominalist
conclusion:
The idea is that if it appears in principle impossible to explain this, then
that tends to undermine the belief in mathematical entities, despite
whatever reason we might have for believing in them. . . . Like Benacerraf,
I refrain from making any sweeping assertion about the impossibility of the
required explanation. However, I am not at all optimistic about the
prospects of providing it.

Here we have a statement of the problem that makes no appeal to

theories of truth, reference, justification, or knowledge. It simply,
naturalistically, asks for an explanation of a purported fact.
Of course, there is more to it than this. If this new version of the
Benacerrafian syllogism is to share the anti-Platonistic moral of the
original, there must be strong reasons to suppose that the required
explanation will not be forthcoming. Field's formulation surely
implies this, and in my discussion of truth, I also suggested that
whatever difficulties the theory of reference was supposed to cause
the Platonist, the obstacles to explaining reliability would be
similar, perhaps identical. But on the surface, this new argument
hardly seems to share the knock-down conclusiveness of its
predecessor. Given that the first premiss, the replacement for the
causal requirement, includes no clause that stands in explicit
contradiction to the traditional Platonistic assumptions of the
second, we must ask why the task of providing the required
explanation should still seem so daunting.
Obviously, what we are up against here is another, less specific,
version of the same vague conviction that makes the causal theory
of knowledge so persuasive: in order to be dependable, the process
by which I come to believe claims about xs must ultimately be
responsive in some appropriate way to actual jcs. And, invoking the
second premiss, nothing can be responsive to non-spatio-temporal,
unchanging, acausal, unobservable Platonic entities. How, then,
can Solovay's reliability be anything more than a fluke? How can it
possibly be explained?
I won't try to make this vague conviction any more definite, nor
will I try to refute it. But I do want to point out it is not, by itself,
enough to cause a problem for traditional Platonism, Even if our
 WHAT IS THE Q U E S T I O N ? 45

most basic reliable beliefs, for example perceptual beliefs, are

directly conditioned by the objects of those beliefs, many other, less
basic beliefs are inferred from these. A physicist, for example,
needn't see, or causally interact with, a molecule of water on the
other side of the moon in order to reliably believe that it has a
certain structure; she need only have good reason to believe that
there is water on the other side of the moon and a well-established
theory of the structure of water molecules. In other words, any
reasonable theory of reliability will have to allow for various forms
of inference as reliable belief-forming mechanisms.
Why then couldn't Solovay's beliefs about sets also be reliably
inferred? Of course, many of them are, assuming deduction is
reliable, but what about the axioms from which these are deduced?
In fact, this possibility is rarely considered; underlying the
persuasiveness of all these Benacerraf-style arguments—from the
original one based on the causal theory of knowledge to the current
one based on the need for an explanation of reliability—is the
unspoken assumption that some mathematical beliefs are not
inferred. This omission is especially glaring given that one
prominent form of Platonism, namely Quine/Putnamism, does treat
all mathematics as inferred: mathematics is a collection of highly
theoretical hypotheses, justified by their indispensable role in
science. As this sort of hypothetical inference might well qualify as
a reliable process, Quine/Putnamism should at least be considered
as a possible reply to Benacerraf-style worries.27
But as I say, it often isn't,28 a fact for which I offer this
explanation. Mathematicians aren't the only ones swayed by the
pre-theoretic realism described at the beginning of Chapter 1.
Many of us tend to think of mathematics, not as a highly theoretical
adjunct to physical science, but as a science in its own right, with its
own subject matter and its own methods. On this view, math-
ematics is parallel to, not subservient to, natural science, so it is
natural to suppose that Platonistic epistemology should run parallel
to scientific epistemology, and from this it follows that some
mathematical beliefs should be basic and non-inferential, just as
27
Of course, as indicated in ch. 1, sect. 4, above, I don't think Quine/Putnamism
offers an acceptable account of mathematical knowledge (or the reliability of
mathematical beliefs), but my point here is that Benacerraf and those who cite him
rarely even consider it.
28
Field (1989), 28—30, is an exception. He rejects a Quine/Putnam solution for
some of the reasons rehearsed in ch. 1, sect. 4, above.
46 P E R C E P T I O N AND I N T U I T I O N

some scientific beliefs are. Furthermore, as the most fundamental

belief-forming mechanism in physical science is perception, the
corresponding faculty in mathematical science is expected to be
'perception-like', and hence, most likely causal.29 And here the
seeming impossibility arises.
But if this pre-theoretic science/mathematics analogy lies behind
the stubborn Benacerraf-style worries of some philosophers, I don't
want to suggest that it accounts for all. Even an advocate of a
Quine/Putnam Platonism, who holds that all mathematical beliefs
are gained by the reliable process of inference to the best
explanation, might worry all the same because she holds that all
explanations are ultimately causal. For another, the role of the
science/mathematics analogy might be played instead by a strong
form of physicalism that requires every legitimate entity to be part,
as Armstrong puts it, of 'a single, all-embracing spatio-temporal
system'.30 And there are doubtless other possibilities. Thus, I think
it's best to see these serious philosophical worries about Platonism
as a syndrome with more than one aetiology.
One last question: does the nagging Benacerraf-style problem for
Platonism constitute an argument in favour of nominalism? Of
course, many anti-Platonists have claimed that it does, but John
Burgess strenuously disputes this. 31 From a naturalized perspective,
epistemology is a descriptive and explanatory enterprise; its goal is
to describe and explain the belief-forming mechanisms of human
knowers. So, Burgess considers the actual practices of scientists.
Here we find well-confirmed affirmations of mathematical know-
ledge and no attributions of causal powers to mathematical entities.
Thus, a causal requirement on knowledge of the sort enshrined in
Benacerraf's first premiss simply doesn't turn up in the descriptive
phase of epistemology naturalized. He concludes that 'a causal
29
I'm assuming here that perceptual beliefs aren't inferential, in particular, that
they aren't inferred from sense data. This doesn't rule out 'inference to the best
explanation' accounts of perception (e.g. Harman (1973), ch. 11, or Gregory (1970;
1972)) because there the inferences are from states of the nervous system to beliefs,
and I'm reserving the word 'inference' for inferences from beliefs to beliefs.
Also, there are those who hold that perception needn't require causation (Kim
(1977)), but I see no need to quibble about this. The issue is whether or not
mathematical objects can participate in interactions suitably similar to the
participation of my hand in the formation of my perceptual belief that there is a
hand before me when I look at it in good light. See Grice (1961), and sect. 2 below.
30
Armstrong (1977), 149. I'll return to this idea in ch. 5, sect. 1, below.
31
See Burgess (forthcoming ft).
 WHAT IS THE Q U E S T I O N ? 47

criterion for knowledge [is] problematic, whether regarded as part

of a proposed analysis of the meaning of "know" or as part of a
proposed analysis of scientific standards of justification' (Burgess
(forthcoming b)). Presumably the same could be said for the require-
ments on reliability implicit in Field's version of the argument, and
even for the various other assumptions on which I have suggested
that Benacerraf-style worries might rest.
Field naturally sees the situation somewhat differently. He
doesn't claim that Benacerraf-style worries provide a conclusive
argument against Platonism; he recognizes that they do nothing to
disarm the otherwise powerful arguments for Platonism:
Of course, the reasons for believing in mathematical entities (in particular,
the indispensability arguments) still need to be addressed, but the role of
the Benacerrafian challenge (as I see it) is to raise the cost of thinking that
the postulation of mathematical entities is a proper solution, and to thereby
increase the motivation for showing that mathematics is not really
indispensable after all. (Field (1989), 26)

That is Field's plan: to undermine the indispensability arguments

by showing how mathematics could be useful in applications
without being true. This project requires that science be rewritten in
nominalistically acceptable terms, contra Putnam's claims, and that
the use of mathematics in nominalized science be justified nominal-
istically, without appeal even to Hilbert-style finitistic metamath-
ematics.
If this were possible,32 Field would be presenting an alternative
overall theory, according to which there is no mathematical
knowledge, but the use of mathematics in science is nevertheless
justified. The advantage of this picture over the Platonist's is that it
doesn't leave a puzzling open question in the psychological part of
our theory about how people come to have reliable beliefs about
Platonic entities. Thus I take it Field would be arguing, in good
naturalized form, that his overall theory of the world is better than
the Platonist's. At this point, then, the Benacerraf-style problem
would become an argument in favour of nominalism.
Even then, though, I suspect Burgess would doubt that Field's
theory is actually preferable. After all, it requires a revision in the
32
Numerous difficulties with Field's project have emerged. For a sampling, see
Malament (1982), Resnik (19850, b), Shapiro (19836), Burgess (1984), and
Detlefsen (1986), ch. 1. For my own discussion, see ch. 5, sect. 2, below.
48 P E R C E P T I O N AND I N T U I T I O N

standard canons of scientific practice to balance a worry generated

by a vague and surely less well-established psychological conviction
about the nature of reliable processes. Perhaps he would argue that
no gain in the psychological portion of our theory could justify the
rejection of otherwise effective scientific methods. At this point, the
form of Field's reply would depend on the actual details of his
proposed nominalist theory.
This is an important debate, one I won't attempt to resolve, but I
do want to point out that, despite appearances, it is irrelevant to my
project. If Burgess is right, his arguments would undermine the
effectiveness of Benacerraf-style worries as justifications for nom-
inalism. However welcome this conclusion might be, it does
nothing to take the compromise Platonist off the epistemological
hook. By rejecting pure Quine/Putnamism, by embracing some
version of Godel's science/mathematics parallelism, the com-
promise Platonist incurs the very real debt detailed in the last
chapter: within the bounds of epistemology naturalized, she owes a
descriptive and explanatory account of mathematical knowledge
(or mathematical reliability) that does justice to the actual practice
of mathematics, an account of both intuition and other peculiarly
mathematical justifications. And to provide such an account is to
meet the Benacerraf-style worries head on, whether or not they
constitute an effective argument for nominalism.

So, despite all this talk about the exact (or inexact!) nature of the
causal premiss to the Benacerrafian argument, it is not my plan to
attack the problem at that point. Instead, I intend to reject the
traditional Platonist's characterization of mathematical objects; I
will bring them into the world we know and into contact with our
familiar cognitive apparatus. An account of our 'perception-like'
connection will be the goal of this chapter. To give a taste of how it
will go, let me return to the (convenient, but inessential) language
of the causal theories.
Consider again the initial baptism of gold: the baptist stands in
front of an array of samples, looks at them, and declares that these
and things like them are gold. An analogous baptism of sets would
go like this: our baptist, at her desk, declares, These three things—
the paper weight, the globe, and the inkwell—taken together,
regardless of order, form a set', or 'The individual books on this
shelf, taken together, in no particular order, form a set.' In this way,
 WHAT IS THE QUESTION? 49

she isolates samples of the kind 'set', and the word then refers to the
kind of which these samples are members.
The obvious objection is that, while the gold-dubber causally
interacts with her samples, the set-dubber causally interacts only
with the members of her samples. Here the Platonist might
respond:33 the extent of the causal interactions of both the gold-
dubber and the set-dubber is something like light bouncing off
certain objects and bringing about some retinal changes. In the case
of the gold-dubber, strictly speaking, the interaction is actually with
the front surface of a time slice of the sample. In other words, the
thing whose kind we count the baptist as having dubbed is not the
thing with which the dubber has causally interacted; her interaction
is only with a fleeting aspect of the temporally extended sample.
Similarly, the set-dubber has only aspects of her sample sets within
her causal grasp. But, if the interaction of the gold-dubber with an
aspect of her samples is enough to allow her to pick out a sample
and dub a kind, why shouldn't the set-dubber's interaction with an
aspect of her samples accomplish those same things? The Platonist
could argue that the relation of element to set is no more
objectionable than the relation of fleeting aspect to temporally
extended object.34
If an argument of this sort could be filled in, then it seems the
Platonist might adopt a causal theory of reference after all. And
since the bare causal interactions described form the basis for a
perceptual connection between the baptist and her samples, a
causal theory of knowledge is within reach as well. Both causal
theories require more than the mere causal interaction—the
bouncing of light rays off objects and onto retinas—they require
that the baptist and the knower perceive a physical object. That
these two are not the same is clear from experiments on patients,
blind from birth, whose sense organs are restored to perfect
operating condition, but who cannot properly perceive the objects
around them.35 The Platonist's hope is that an account of what
33
That an argument of this sort might be available to the Platonist was first
suggested to me by John Burgess.
34
Of course, the set-dubber's interaction is also with mere fleeting aspects of the
members of her sets, so the relation between what she interacts with, and what's
kind she dubs, is the composition of the aspect/object and the element/set relations.
This added complexity could be eliminated by imagining that the set-dubber uses
sets of aspects, rather than sets of objects, as samples, but a more reasonable course
would be to assume that if both the relations in question are legitimate, then so is
)5
their composition. See Hebb (1949), ch. 2.
50 P E R C E P T I O N AND I N T U I T I O N

makes one pattern of sensory stimulation into a perception of a

physical object might also provide an account of what makes
another pattern of stimulation into a perception of a set of physical
objects.

2. Perception

The question is what bridges the gap between what is causally

interacted with and what is perceived, and the hope is that
something like what does the bridging in the case of physical object
perception can be seen to do the same job in the case of set
perception.36 Notice that this way of putting the problem already
assumes that we do in fact perceive physical objects, as opposed to
sense data, or percepts, or representations of some kind. In the
theory of perception, this is called 'direct realism'. Psychologists of
perception are generally direct realists in this sense,37 and though
some philosophers are still inclined to debate the issue, I'll rely
implicitly on the persuasive counter-arguments to be found in the
literature. 38
What we want here is a strong sense of 'perceives' that rules out
illusions; what is perceived in this sense is really there. But that isn't
all. We also insist that, for example, a hiker doesn't perceive a leaf-
dwelling insect on a bush she passes if the bug blends too perfectly
with its surroundings for her to distinguish it from them. In such a
case, even though light from the bug registers a pattern on the
hiker's retina, she gains no beliefs about it. It isn't enough,
however, to require that she gain beliefs in order to perceive, or
even that she gain beliefs visually, because it is possible to gain
beliefs, even visually, that clearly don't count as perceptual beliefs
about the bug; she could, for example, come to believe there is a
Ceylonese leaf insect on the bush by reading a sign. This belief isn't
a perceptual belief that there is a Ceylonese leaf insect on the bush
because it doesn't involve it looking to the hiker (in a phenomenal
or non-metaphorical sense) as if there is a bug on the bush. For
36
I will concentrate here on visual perception. Naturally the blind can know and
refer as well as the sighted, but I make the customary philosophical assumption that
what's true of vision can be adapted to the other senses.
37
For history, see Hebb (1980), 19. For agreement from various psychological
camps, see). Gibson (1950), 26-7, Neisser (1976), 16, and Gregory (1972), 220.
38
See Pitcher (1971), ch. 1, or the references cited in Machamer (1970), § II.
 PERCEPTION 51

perception, then, we require that the perceiver gain appropriate

perceptual beliefs.
But this still isn't enough. Consider an example analogous to
Gettier's: an illusionist arranges a system of mirrors so that it looks
to Steve as if there is a tree in front of him. Suppose that the actual
source of the image Steve sees is a tree behind him. Suppose, finally,
that there is in fact a tree in front of him, where he sees the illusion,
though that tree is hidden. Here Steve gains a true perceptual
belief—there is a tree in front of me—but it doesn't count as a
perception that there is a tree in front of him because its causal
genesis is faulty. Clearly, we must insist that whatever makes the
belief true—in this case, the tree in front of Steve—be responsible
for his belief in some appropriate way. Paul Grice's solution is to
insist that the thing perceived must play the same sort of role in the
causation of the perceiver's perceptual state as my hand plays in the
generation of my belief that there is a hand before me when I look
at it in good light.39 In sum, then, for Steve to perceive a tree before
him is for there to be a tree before him, for him to gain perceptual
beliefs, in particular that there is a tree before him, and for the tree
before him to play an appropriate causal role in the generation of
these perceptual beliefs.40
Notice that the content of a perceptual belief state is extremely
rich and varied. For Steve to acquire the perceptual belief that there
is a tree before him, he must also acquire a great variety of other
perceptual beliefs, depending on the occasion, such as, that the tree
is roughly so big, so far away, that it is in leaf, swaying in the
breeze, and so on.41 Such beliefs are non-inferential,42 and not
necessarily conscious or linguistic.43 When the various components
of a perceptual belief state arise as a body, on a given occasion, they
often influence each other non-inferentially, as, for example, a
belief about the identity of an object can influence perceptual beliefs
about its shape and size, and obviously, vice versa.
39
See Grice{ 1961).
40
This account is a crude version of Pitcher's (1971), ch. 2. It is similar to Arm-
strong's (1961), and to various psychological theories of perception as information
acquisition, like Gregory's (1972). Pitcher and Armstrong insist that belief content
exhausts the perceptual state, while Goldman and others aren't so sure. (See
Goldman (1977), § 6.) For our purposes, this issue is beside the point.
41
See Pitcher (1971), 87-9.
42
Except for the weak sense, noted earlier, in which they can be considered as
'inferred' from states of the nervous system.
43
See Armstrong (1973), chs. 2 and 3.
52 P E R C E P T I O N AND I N T U I T I O N

Beliefs in general are psychological states. I assume that a

person's behaviour gives good evidence for our hypotheses about
her psychological state, though I won't go so far as to assert, as
some philosophers would, that being in a certain psychological
state simply is behaving (or being disposed to behave) in certain
ways. If this behaviouristic position is incorrect, then behavioural
evidence is not conclusive, but it is still often the best available.
Sometimes, there might also be introspective evidence for or against
the claim that a person is in a given psychological state, and if there
is a correspondence (not necessarily the identity) between psycho-
logical states and brain states, as I suppose there is, then
neurophysiological evidence would also be relevant. 44
The question before us now is this: how do we manage to
perceive physical objects? Assuming 45 that to have a concept is to
have the capacity for beliefs of a certain sort, we can rephrase the
question: how do we come to have the concept of a physical object?
Psychologists and neuropsychologists have produced various re-
sults and theories to answer the question of how conceptual
elements enter human perceptual states. I'll review some of their
work here before returning to the philosophical issues involved.

There is considerable experimental evidence that the ability to

perceive a primitive distinction between a figure and its background
is inborn in humans and many laboratory animals. 46 The structure
of the retina is probably responsible for the presence of this
conceptual information in the human perceptual state, much as it is
in the frog. Warren McCulloch and his co-workers have isolated
various structures in the frog's retina which send impulses to the
frog's brain only under certain complex sets of conditions,
independent of the level of general illumination, for example, in the
presence of sharp boundaries between relatively light and relatively
dark patches, or dark areas with curved edges, or movement of
such edges. In fact, one fibre
responds best when a dark object, smaller than a receptive field, enters that
field, stops, and moves about intermittently thereafter. The response is not
affected if the lighting changes or if the background (say a picture of grass
44
Of course it will take some substantial progress in neuroscience before this is a
real possibility, but my point is that it isn't ruled out a priori.
45
With Armstrong (1973), ch. 5, § 1.
46
See Hebb (1949), 19-21.
 PERCEPTION 53

and flowers) is moving, and is not there if only the background, moving or
still, is in the field. Could one better describe a system for detecting an
accessible bug? (Lettvin et al (1959), 254)
As might be expected, the researchers came to think of these fibres
as 'bug-detectors', and the frog's behaviour certainly suggests that
this mechanism enables it to acquire perceptual beliefs about
nearby bugs. Similar mechanisms in humans are probably respons-
ible for perceptual beliefs concerning figure and background, and
perhaps some concerning distance and size.47
But beyond this fairly simple level, the evidence indicates that the
capacity to acquire perceptual beliefs of the familiar sort is not
present at birth.48 Psychologists talk of a phenomenon called
'identity' in perception. A figure is said to be seen with identity
when it appears similar to some other figures but not to others,
when it is seen as falling into some categories and not in others, when
it is easily recalled, recognized, or named. When I see a triangular
figure, for example, I automatically see it as more like other
triangles than like squares, I can recall it, recognize it, and call it
and other similar figures 'triangles'. In the terminology we've
adopted here, I acquire the perceptual belief that there is a triangle
before me. Experiments on newly sighted human patients who had
been blind from birth, and on chimpanzees raised in total darkness,
demonstrate that the capacity to acquire such a belief—what we've
also called having the concept of a triangle—is present only after
considerable perceptual experience. For example,
Investigators (of vision following operation for congenital cataract) are
unanimous in reporting that the perception of a square, circle, or triangle,
or of sphere or cube, is very poor. To see one of these as a whole object,
with distinctive characteristics immediately evident, is not possible for a
long period. The most intelligent and best-motivated patient has to seek
corners painstakingly even to distinguish a triangle from a circle . . . A patient
was trained to discriminate square from triangle over a period of 13 days,
and had learned so little in this time 'that he could not report their
form without counting corners one after another . . . And yet it seems that
the recognition process was beginning already to be automatic, so that
some day the judgement "square" would be given with simple vision,
47
See e.g. Bower (1966).
48
This has little to do with the philosophical controversy over innateness because
even their defenders admit that something sensory is needed to 'draw out' or
'awaken 1 innate ideas.
54 P E R C E P T I O N AND I N T U I T I O N

which would then easily lead to the belief that form was always
simultaneously given', {Hebb (1949), 28, 32)
Similar results were obtained with the chimpanzees.
Given that a capacity as simple as the ability to see a triangle as
more like another triangle than like a square is the product of
considerable sensory experience, it is to be expected that so
complex a talent as that of seeing a series of different patterns as
aspects of one thing—that is, as a sequence of views of one physical
object—is not present at birth. This expectation is substantiated by
the experiments of Jean Piaget and his colleagues.49 The child's
ability to acquire perceptual beliefs about physical objects, as
judged from behaviour, develops between the ages of one and
eighteen months. At the beginning of this period, the child's world
is a welter of isolated incidents. Then
the behavior of the child begins to be centered on objects; but to him there
is no objective reality—no general space or time, no permanence of objects.
There are only events—i.e. components of the child's own functioning.
When an object in his field of vision disappears, it ceases to exist. {Phillips
(1975), 28)
Some months later, objects begin to enjoy a sort of permanence:
there is a shift in the child's conceptualization from object reality
dependent on his own actions to object reality dependent on the surround.
The result is a kind of 'context-bound object permanence'. (Phillips (1975),
38)
At this stage, the object is associated with a particular location; the
child expects to find it there even when it is clearly hidden in a new
location. Similarly, we find that
the infant does not realize that a moving object is the same object when it
becomes stationary . . . [or] that a stationary object that begins to move is
still the same object after it starts moving. (Bower (1982), 206) 50
The ability to distinguish the object by features such as size, shape,
and colour, in addition to its location or trajectory, is a major
49
See Piaget (1937) and Phillips {1975).
50
Bower and his colleagues would dispute some details in the Piagetian account
of the development of the object concept, but the general idea of a non-trivial
developmental period is all that matters here. Even psychologists who play down the
role of learning in perception agree that a development of this general sort takes
place. SeeE. Gibson (1969),ch. 16.
 PERCEPTION 55

development, followed by the ability to distinguish objects that

share a common boundary and a stepwise improvement in
searching behaviours. By the end of the developmental period, the
child possesses our familiar concept of an independently existing
physical object and is fully capable of acquiring perceptual beliefs
about them.
Supposing then, as the evidence suggests we should, that the
concept of a triangle or of a physical object—that is, the ability to
acquire beliefs, including perceptual ones, about such things—is
acquired over some period of time, what can be said about how this
is accomplished? Presumably some change must take place within
the brain, some alteration in the neural structuring, that enables us
to see things with identity, that fills the gap between the pattern of
light on our retinas and the perceptual beliefs we acquire. To fix our
ideas, I'll sketch a particular neurophysiological theory of what
goes on in this development, a theory due to Donald Hebb.51 I
doubt that the philosophical morals I'll eventually draw actually
depend on the correctness of exactly this scientific theory, but in an
area of abstract epistemology, I find it reassuring, even useful, to
have at least one fairly specific example of how a naturalistic story
of our knowledge might go.
That part of the brain involved in perception, the visual cortex,
can be divided into four layers, and the pattern of retinal
stimulation is topologically equivalent to the pattern of activity in
the first of these only. After that, topological characteristics are not
preserved, and the neural connections between the inner layers, and
between these and the outer layer, are so complex as to seem
random. Excitation of a small part of the initial layer can stimulate
widely separated areas in all three inner layers; widely separated
parts of the initial layer can stimulate neighbouring cells in the
inner layers; inner areas can stimulate other inner areas and even
outer ones. For any one cell to fire and pass along its excitement to
the cells connected with it, it must be stimulated by many other cells
simultaneously. But any cell in the visual cortex is connected with
many, many others, so the firing of a number of cells in the first
layer will usually result in a convergence of sufficient stimulation
for firing on a number of cells in various other layers, if for
statistical reasons alone. Thus any visual stimulus creates a

51
See Hebb (1949), esp. chs. 4 and 5, and (1980), ch. 6.
56 P E R C E P T I O N AND INTUITION

veritable hum of activity throughout the visual cortex, and the

hums corresponding to different retinal stimuli are globally the
same. The question, then, is how all this activity is organized.
To answer it, Hebb makes a simple theoretical assumption,
namely, that if one cell repeatedly plays a role in the firing of
another, then a change takes place increasing the first cell's efficacy
in firing the second. He suggests that this phenomenon could be
produced by the growth of synaptic knobs, but for theoretical
purposes, any plausible mechanism will do. Now suppose that
someone unfamiliar with triangles fixes her gaze on the apex of a
given triangular figure. This generates a certain pattern of stimula-
tions which recurs as long as she stares at that corner, and occurs
again every time she looks at it. As a result of this repeated
exposure, groups of cells at various cortical levels are repeatedly
efficacious in firing one another (at convergences of various kinds).
As a result, it becomes easier for these interconnected cells to fire
one another, they become more interdependent, eventually forming
what Hebb calls a 'cell-assembly'. It will respond to the apex of any
similar triangle.
Analogous processes naturally result in assemblies for the base
angles of the triangle, but this leaves open the question of how the
triangle as a whole, as a unit, is perceived with identity. Behavioural
experiments suggest that acquiring the ability to recognize triangles
depends essentially on successive eye movements and fixations on
various parts of the figure. The eye is especially inclined to trace
lines because all the movement-inducing peripheral stimulations are
urging it in one direction. Thus angles are frequent fixation points,
lying as they do at the intersection of straight lines, and when the
eye is fixated on one angle, it is stimulated to move towards
another.
Because of the numerous interconnections between neurons in
various parts of the visual cortex, cells in one assembly happen by
chance to be connected with various cells in the other two. When
the motor stimulations described above induce repeated movement
from one angle to another, the neurons in these chance connections
become increasingly efficacious, and the cell-assemblies for the
three corners are integrated into a second-order cell-assembly. The
individual angle assemblies can still work independently, so actual
perception of a triangle involves what's called a 'phase sequence' of
excitations of the corner assemblies and the integrated assembly. If
 PERCEPTION 57

the corner assemblies are tf, b, and c, and the integrated assembly t,
then a phase sequence is something like a-b-t-a-c-t-c-t-b.52
Once an integrated cell-assembly of this sort has been formed,
looking at a triangle will cause it to reverberate for half a second or
more. This represents a considerable gain both in organization and
in duration over the random hum of activity brought about by the
same visual stimulation before the formation of the assembly. This
longer, repeatable trace should persist long enough to allow the
structural changes required for long-term memory. In other words,
the cell-assembly is what permits the subject to see a triangle with
identity, to acquire perceptual beliefs about it: it is a triangle-
detector in much the same sense as the fibre located by McCulloch
is a bug-detector; it provides the subject with her concept of triangle.53
The ability to perceive physical objects is not unlike the ability to
perceive triangular figures, though it is more complex. The trick is
to see a series of patterns as constituting views of a single thing. Just
as the ability to see triangles develops over time, through a
painstaking process of seeking out corners and comparing one
triangle with another, the ability to see continuing physical objects
develops over a period of experience with watching and manipu-
lating them. The close resemblance in the structure of the learning
processes suggests that what is involved in the physical object case
is just a more elaborate version of the cell-assembly, in particular,
the development of higher-order cell-assemblies which respond to a
52
This account is based on Hebb (1949), chs. 4 and 5. Hebb (1980) cites evidence
suggesting that the basic assemblies may respond to the sides of the triangle rather than
its angles (see pp. 89, 98—102). This modification doesn't affect the upshot of my
discussion. The 1980 work also summarizes new supports for Hebb's theory
discovered in the years since 1949 (see pp. 98-100), and develops various
applications of the view, including a fascinating account of scientific problem-
solving (pp. 119-21).
53
Philosophers of mind sometimes worry about a blanket objection Dennett has
raised to attempts to locate concepts in the neurons: 'Suppose your "grandmother"
neuron died; not only could you not say "grandmother", you couldn't see her if she
was standing right in front of you . . . you would have a complete cognitive blind
spot . . . Nothing remotely like that pathology is observed, of course, and neurons
malfunction or die with depressing regularity, so ... theories that require grandmother
neurons are in trouble' (1978, p. xiii). I don't think Hebb's theory requires anything so
specific as a grandmother neuron, but it does posit shape and other general
detectors. Hebb (1949) already contains a response to objections like Dennett's:
'The assembly is thought of as a system inherently involving some equipotentiality,
in the presence of alternate pathways each having the same function, so that brain
damage might remove some pathways without preventing the system from
functioning . . .' (p. 74).
58 P E R C E P T I O N AND INTUITION

series of aspects of a single object, and fire, as before, in a

complicated phase sequence.54 When an object stimulates a phase
sequence of such assemblies, it participates in the generation of the
subject's perceptual belief state in the appropriate causal way, in
the way my hand participates in the generation of my belief that
there is a hand before me when I look at it in good light.
Crudely put, human beings develop neural object-detectors
which allow them to perceive independent, continuing physical
objects. It is these complex cell-assemblies that bridge the gap
between what is interacted with and what is perceived. The object I
perceive on a given occasion, or more precisely, the front side of a
time slice of that object, is causally responsible only for the pattern
of retinal stimulations, while the unifying concept of a familiar
physical object is contributed by my physical object-detector. The
presence of the object-detector, in turn, is partly the result of the
structure of my brain at birth (conditioned by the evolutionary
pressures of the environment on my ancestors) and partly the result
of my early childhood experiences with physical objects. All this,
while undeniably complex, is still naturalistic, and causal.

Let me now return to the subject that inspired this detour into the
theory of perception in the first place, namely my claim that we can
and do perceive sets, and that our ability to do so develops in much
the same way as our ability to see physical objects. Consider the
following case: Steve needs two eggs for a certain recipe. The egg
carton he takes from the refrigerator feels ominously light. He
opens the carton and sees, to his relief, three eggs there. My claim is
that Steve has perceived a set of three eggs. By the account of
perception just canvassed, this requires that there be a set of three
eggs in the carton, that Steve acquire perceptual beliefs about it,
and that the set of eggs participate in the generation of these
perceptual beliefs in the same way that my hand participates in the
generation of my belief that there is a hand before me when I look
at it in good light.
This claim will doubtless elicit a clamour of objections, and even
if it doesn't, it needs elucidation, so let me begin at the beginning,
with the assertion that there is a set of eggs in the carton in front of
Steve. The simplest way to resist this idea is to deny that there are
sets. To this I reply with a version of the Quine/Putnam arguments

54
See Hebb (1980), 107,
 PERCEPTION 59

sketched towards the end of Chapter 1—mathematical entities are

indispensable for our best theory of the world—supplemented by
the observation that our best theory of mathematical ontology is
that (at least some)55 mathematical entities are sets.
A second mode of resistance is to join the traditional Platonist in
denying that sets have location in space or time. But notice: there is
no real obstacle to the position that the set of eggs comes into and
goes out of existence when they do, and that, spatially as well as
temporally, it is located exactly where they are. A set of higher
order, like the set consisting of the set of eggs and the set of Steve's
two hands, would again be located where its members are, that is,
where the set of eggs and the set of hands are, which is to say, where
the eggs and hands are. In this way, even an extremely complicated
set would have spatio-temporal location, as long as it has physical
things in its transitive closure.56 And any number of different sets
would be located in the same place; for example, the set of the set of
three eggs and the set of two hands is located in the same place as
the set of the set of two eggs and the set of the other egg and the two
hands.57 None of this is any more surprising than that fifty-two
cards can be located in the same place as a deck. In any case, I
hereby adopt this view as part of set theoretic realism. On some
terminological conventions, this means that sets no longer count as
'abstract'. So be it; I attach no importance to the term.58
More controversial is the second part of the claim that Steve sees
a set: the contention that he gains perceptual beliefs about the set of
eggs, in particular, that this set is three-membered. Let me break my
defence of this idea into two parts. First, and least controversially, I
contend that the numerical belief—there are three eggs in the
carton—is perceptual,59 that is, that it looks to Steve, in a non-
metaphorical sense, as if there are three eggs there. There is

55
As indicated in ch. 1, sect. 1, I'm not assuming that all mathematical entities
are sets, but our standard mathematical theories indispensably refer to sets of points,
sets of numbers, etc. See ch. 3 below for the relationship of the familiar natural and
real numbers to sets.
56
The transitive closure of a set consists of its members, the members of its
members, the members of the members of its members, and so on. For a formal
definition, see Enderton (1977), 178.
57
Set theorists will notice that this means there are proper class-many sets
located where any physical thing is. (They stack neatly!) Pure sets have no location.
For more on pure sets, see ch. 5, sect. 1, below.
58
See Katz (1981), 207 n. 29. Parsons uses the term 'quasi-concrete' for
mathematical objects of this sort. See his (1983(2), 26.
59
See Kim (1981) for a similar claim.
60 P E R C E P T I O N AND I N T U I T I O N

empirical evidence, based on reaction times, that such beliefs about

small numbers are non-inferential. 60 Furthermore, this belief about
the number of eggs can non-inferentialiy influence and be influenced
by other clearly perceptual beliefs acquired on this occasion; for
example, the welcome fact that there are enough eggs for the recipe
can make the eggs themselves look larger.61 This particular
perceptual belief about the number of eggs is thus part of a rich
collection of perceptual beliefs acquired on this occasion, beliefs
about the size and colour of the eggs, the fact that two eggs can be
selected from among the three in various ways, the locations of the
eggs in the nearly empty carton, and so on.
So far so good. Now let me take up the question of whether or
not this perceptual belief is a belief about a set. What is a numerical
belief about, after all? The easiest answer would be that Steve's
belief is about the eggs, the physical stuff there in the carton, but
Frege long ago demonstrated the inadequacy of that response.62
The trouble is that the physical stuff in the carton has no
determinate number property: it is three eggs, but many more
molecules, even more atoms, and only a quarter of a carton of eggs.
For a given mass of physical stuff, there is no predetermined way
that it must be divided up, and without this, there is no determinate
number property. So the physical stuff by itself cannot be three.
If not the physical stuff making up the eggs, then what is the
subject of a number property? Some would say, 'the eggs', meaning
by this the physical stuff as divided up by the property of being an
egg, what I'll call an 'aggregate1.63 Frege's answer is that a
numerical statement is a statement about a concept, for example,
the concept 'egg in the carton in front of Steve'. Others might
choose the extension of Frege's concept, that is, what is usually

60
See Kaufman etal. (1949).
61
For another example of such non-inferential influence, suppose a majority vote
from a three-membered panel will defeat your motion. When two panellists raise
their hands to vote no, your numerical belief that there are two of them (enough to
dash your hopes) may weli influence your perception of their facial expressions (how
malevolent they look!). If only one had voted negatively, she might only have looked
dense.
62
Frege (1884), §§22-3.
63
Some would say that 'there are three eggs in the carton' is properly analysed as
saying 'there is an x, there is a y, and there is a z, all distinct, such that x is an egg in
the carton, y is an egg in the carton, z is an egg in the carton, and anything that's an
egg in the carton is either x or y or z1.1 take this to be a variant of the aggregate view:
some physical stuff is divided up in a certain way.
 PERCEPTION 61

called the 'class' of eggs in the carton.64 And the set theoretic realist
opts for the set of eggs in the carton.
But on what grounds? If Steve is supposed to see a set of eggs,
shouldn't the set theoretic realist hold that he can see, for example,
that it is a set and not an aggregate? Attractive as this move might
seem, I think it is not correct. Notice that the various candidates for
the bearer of the number property—the set, the aggregate, the
concept, the class—have their most basic properties, any properties
that might count as perceptual, in common; for example, they all
have subcollections (e.g. the egg-stuff under a more exclusive
property), they are all capable of combination (e.g. the disjunction
of two concepts), and so on. These similarities are what makes
them all potential candidates for number-bearing. The properties
that separate them are theoretical properties, like extensionahty.
Asking Steve to look and see whether he's perceiving a set or an
aggregate is like asking him to look and see whether the egg is solid
or mainly an empty space littered with atoms.
What I'm getting at is this: the amount we know about things by
perception is very limited. About physical objects, for example, we
know little more than that they are, in Hebb's words, 'space-
occupying and sense-stimulating something[s]\65 Beyond that, the
bulk of our knowledge about them is theoretical: that they are
made up of atoms, of this and that sort, arranged in such-and-such
a way, and so on. The same goes for sets. What we perceive is
simply something with a number property, something that can be
combined with others of its ilk, and so on. Nailing down this
number-bearer's more esoteric properties is a theoretical matter.
So, to decide the case between sets, aggregates, concepts, classes,
and whatever else, we need to look, not to our perceptual
experiences, but to our overall theory of the world, and we must
ask which of these is best suited to playing the role of the most
fundamental mathematical entity. (Compare: deciding whether
Berkeleian bundles of God's experiences or the physicist's bundles
of atoms are best suited to playing the role of the most fundamental

64
A class differs from a set in that it is essentially dependent on a property, like
'being an egg in the carton in front of Steve'. Sets, by contrast, are generated
iteratively, by taking at each stage every subset of what's been generated before,
regardless of whether or not the members of that subset can be singled out by some
property. See ch. 3, sect 3, below.
65
Hebb (1980), 109. I'll consider this passage in more detail in the next section.
62 PERCEPTION AND I N T U I T I O N

physical entity.) On this score, sets win going away; they are
extremely simple and manageable entities that form the basis for a
surprisingly effective and efficient mathematical theory. In contrast,
properties, on which both aggregates and classes depend, are hard
to handle—no comparably flexible and complete theory is known 66
—and prone to paradox—for example, consider the property a
property has when it doesn't have itself. 67 And classes are little
better.68 The elementariness of the notion of set, its ease of
manipulation, and the immense success of set theory, both as a
foundation for other branches of mathematics and as a math-
ematical theory in its own right, all help to make the set of eggs the
most attractive candidate for the role of number-bearer.
I take all this to support the set theoretic realist's claim that the
bearers of number properties are sets, and thus, that Steve's
perceptual belief is a belief about a set. But before leaving this
point, I want to call attention to the contingency of this conclusion.
In its support, I depend on the idea that mathematical entities are
indispensable to physical science; if they weren't, there would be no
reason to include sets in our overall theory. Thus my preference for
sets is contingent on the way the world is, to the extent that our best
theory seems to demand them.
To appreciate the force of this fact, consider, for example, one of
the fundamental differences between sets and aggregates, namely,
that there are sets of higher rank—-sets of sets, sets of sets of sets,
and so on—while there are no aggregates of aggregates. If the
physical world were simpler, allowing for a simpler physical theory
with no continuous phenomena, then our overall theory might have
no need for real numbers, and consequently, for sets of higher
rank. 69 Since we're engaged in science fiction, we might imagine
that our perceptual experiences of discrete objects in this simpler
66
Bealer (1982), ch. 5, suggests a property theory that essentially mimics set
theory, but as Anderson (1987, p. 151) points out, the plausibility of some set
theoretic assumptions does not carry over to their property theoretic translations.
67
The related 'paradoxes of set theory' present no problem for sets on the
iterative conception. For example, the Russell set, the set of all non-self-membered
sets, cannot be formed at any stage, because new non-self-membered sets (all sets are
non-self-membered on the iterative picture) will be available at the next stage. See
Godel (1947/64), 474-5.
68
See ch. 3, sect. 3, below.
69
Reals, as standardly constructed, occur a few stages after w. See Enderton's
construction (1977, ch. 5). The relationship of continuous phenomena with sets of
higher rank will be considered in more detail in ch. 4 and ch. 5, sect. 2, below.
 PERCEPTION 63

world are exactly like our experiences of discrete objects in this

world. Still, in the simpler world, we might have no justification for
including higher ranks in our overall theory, and thus, much less
justification for taking our numerical perceptions to be perceptions
of sets rather than aggregates. So my claim that sets are the best
candidates for the bearers of number properties depends on the fact
that they are the best mathematical entities for the mathematical
theory this particular world—with its continuous phenomena—
requires.
Let's grant, then, that there is a set of eggs in the carton, and that
Steve gains the perceptual belief that this set is three-membered.70
For this to count as Steve's perceiving a set, only one further
condition must be satisfied: the set of eggs must participate, in an
appropriate causal way, in the generation of Steve's belief.
Appropriate participation is exemplified by the role of my hand in
the generation of my perceptual belief that there is a hand before
me, and that in turn comes down to my hand's stimulation of a
phase sequence of cell-assemblies. So our question is: could a set of
eggs do the same?

The behavioural evidence of Piaget and his colleagues suggests that

the ability to gain perceptual beliefs about sets develops in a series
of stages parallel to those for perceptual beliefs about physical
objects, though at a somewhat later age.71 At the beginning of this
period, a child may be able to classify objects into groups in a
consistent way—say, triangles with triangles and squares with
squares—but she does not correctly grasp the inclusion relation—
in a group of two black squares and five black circles, the child
thinks there are more round things than black. For the younger
child, a set ceases to exist when its subsets are attended to.
70
Steve needn't express his belief in this way to himself; implicit in the word 'set'
is a more refined theory than most people are aware of. When I say he gains a
perceptual belief about a set, I mean he gains a perceptual belief about a something
with a number property, which we theorists know to be a set. Analogously, when he
perceived the tree in front of him, he gained a perceptual belief with less theoretical
content than a botanist could provide. Nevertheless, what he perceived was the
botanist's tree.
71
Piaget and Szemiriska (1941) would put that age between seven and fourteen
years. See also Phillips (1975), ch. 4. More recent work suggests that these
developmental periods occur somewhat earlier, between two and a half and five
years. See Gelman (1977). Once again, the details are less important than the parallel
between development of the physical object and set concepts.
64 P E R C E P T I O N AND I N T U I T I O N

A similar confusion is observed in connection with the set's

number properties. The younger child imagines that the number of
elements in a set changes when it is rearranged, particularly when
its elements are moved closer together or further apart. For older
children, on the other hand, once a one-to-one correspondence
between two sets has been established, the belief in their equi-
numerosity cannot be shaken; indeed the very question seems silly
to them. Once a perceptual belief about a set is gained, the thought
that the set could change its number property when its elements are
moved about (barring mishap) appears preposterous.
So, just as the concept of an independent and continuing physical
object is acquired in stages, the concept of a set with inclusions
and a constant number property is itself gained over time, and
depends on experience with groups of objects. It should be noted
that the child's development of the set concept is not a linguistic
achievement. Of course, children are rarely taught the word 'set',
but they are taught number words, and it might be thought that
their early errors are primarily verbal, and that it is verbal
instruction that corrects them. The evidence is heavily against this
assumption:72
The ma/or point is that the development of the concept of number begins in
infancy, long before speech or formal instruction play any part. The infant
is forced to generate number concepts by the requirements of its everyday
activities—activities so commonplace that the fondest parent barely thinks
them worthy of comment. They are worth mentioning because these are
the simple beginnings from which the whole structure of mathematical
thinking takes root. (Bower (1982), 250)
One must expect that the set concept, like the physical object
concept, could be developed in the complete absence of language.
And how does the set concept develop? We have seen the
evidence that it develops over a period of time, like the object
concept, and that the determining factor in both these develop-
ments is repeated exposure to the sort of things in question. The
development of the object concept is brought about by the child's
experiences with various physical objects in her environment, and
the set concept by experiences with sets of physical objects, for
example by forming one-to-one correspondences between them, by
regrouping them to form salient subsets, and so on.73
72
See also Phillips (1975), 145.
73
These manipulations are crucial. Kitcher (1983, p. 103) has suggested that my
 PERCEPTION 65

Hebb's theory of the formation of the neural triangle-detector

made essential use of the behavioural evidence that development of
the ability to see triangles with identity requires repeated fixations
on corners of triangular figures, eye movements from one corner to
another, and even, in some cases, active seeking out of corners.
Because of the behavioral similarities between this process and that
leading to the development of the object concept, it is theorized that
an object-detector develops in a similar way, as a result of various
experiences with physical objects in the environment. Given the
evidence that the set concept requires a similar developmental
period involving repeated experience with sets in the environment
parallel to the required experiences with triangles and physical
objects, it seems reasonable to assume that these interactions with
sets of physical objects bring about structural changes in the brain
by some complex process resembling that suggested by Hebb,74 and
that the resulting neural 'set-detector' is what enables adults to
acquire perceptual beliefs about sets.
This assumption provides a solution to another difficulty. Recall
that many things—a mass of physical stuff, and many different
sets—occupy the same spatio-temporal location. These things also
produce the same retinal stimulation, but on one occasion the stuff
is seen, on another one set, on another a different set altogether,
and so on. We've all been amused by the psychologist's examples in
which we see a single picture first as an undifferentiated mass, then
as representing a definite number of distinct objects, or the child's
puzzle in which the homogeneous jungle foliage resolves itself into
a pack of ferocious beasts. Where a bookbinder sees a large set of
individual books (so many perhaps that she has no perceptual
access to the exact number), the encyclopaedia salesman sees three
account is inconsistent with the view of a group of perceptual theorists called the
'ecological realists'. He says that their 'general claim that the information which we
gain in perception concerns transformations of the sensory array caused by events in
which perceived objects participate seems to be at odds with the idea that we can
acquire perceptual information about unchanging abstract objects'. I merely point
out that the 'abstract' objects involved here are not unchanging; in particular, they
can be moved about for purposes of more ready classification and to display one-to-
one correspondences. The account in the text also seems to me in harmony with the
ecological realist's emphasis on invariants (the child learns that the numerable
collection is invariant under transpositions of its elements) and affordances (Steve's
three-membered set of eggs affords cake-making).
74
Hebb (1949) doesn't mention set perception, but he does consider the
perception of number properties of collections to be part of his theory in Hebb
(1980), 122-4.
66 PERCEPTION AND I N T U I T I O N

of his rival's product; where I see a set of four shoes, you might see
a set of two pairs. A microscopic image looks to me like an
unorganized mess of Jackson Pollock drips, while the biologist sees
three paramecia and an amoeba.
Hebb's theory provides a key to these phenomena. They involve
a change in perception on the part of a single subject (the child's
puzzle), a difference in perception between two roughly compar-
able" subjects (the bookbinder and the encyclopaedia salesman), and
a difference in perception between two subjects with different
training (me and the biologist). The last case is easiest: perceptual
development continues past childhood; the biologist acquired
further perceptual abilities (further cell-assemblies) during her
education in lab techniques. In the other two cases, preceding
experiences and neural activity (thoughts) influence the ease with
which a given pattern of stimulation will trigger a given cell-
assembly: the bookbinder and the salesman have different interests;
they notice different things. In the puzzle cases, some shift in our
attention causes a new cell-assembly to come into play quite
suddenly. If there were no such set assemblies, the phenomenon of
seeing different groupings with different number properties would
have to be explained by some additional organizing events
occurring after the initial stimulation of the ordinary physical
object assemblies. This approach is less true to the phenomenon. 75
On this account, then, when Steve looks in the egg carton, there
is a set of eggs there, he acquires perceptual belief about that set,
namely, that it has three members, 76 and the set of eggs participates
in the generation of his perceptual belief in the appropriate way,
that is, in the way my hand participates in the generation of my
belief that there is a hand before me when I look at it in good light,
which is, by some aspect of the object in question causally
interacting with the retina in such a way as to bring about the
stimulation of the appropriate detector. In the case of sets, just as in
the case of physical objects, it is the presence of a complex neural
development that bridges the gap between what is causally
interacted with and what is perceived.
75
See Hebb (1980), 83-7. Also Bruner (1957), 241-4.
76
He undoubtedly acquires other perceptual beliefs about the set of eggs at the
same time, e.g. that it has various two-element subsets. Notice also that, as
promised, Gettier-style cases are easily constructed; the same illusionist who set up
the tree example described earlier could arrange his mirrors to create the illusion of a
set where there is actually another, hidden, set.
 INTUITION 67

Thus, the Hebbian neurophysiological account of what bridges

the gap between what is causally interacted with and what is
perceived in the case of physical objects can also provide for the
perception of sets. I think this lends considerable credibility to the
set theoretic realist's claim that sets are perceivable, but, as
mentioned earlier, I don't intend to tie set theoretic realism to this
particular neurophysiological theory. Benacerraf-style worries are
based on a deep conviction that a certain kind of explanation
cannot be given; see Putnam's rhetorical outburst:
What neural process, after all, could be described as the perception of a
mathematical object? (Putnam (1980), 430)
My goal in this section has been to indicate that there is at least one
plausible answer to this question.

3. Intuition

So far, I have argued that sets, suitably understood, can be

perceived. While this may be enough to answer the Benacerraf-style
worries about compromise Platonism, it provides only the barest
beginning of an account of set theoretic knowledge. What is the
relation, for example, between our knowledge of particular facts
about particular sets of physical objects, and our knowledge of the
simplest set theoretic axioms? How, for example, do we come to
know that any two objects can be collected into a set with exactly
those two members, or that the members of any two sets can be
combined into a set that is their union? These general beliefs
underlie two of the most elementary set theoretic axioms—Pairing
and Union—and our epistemology must account for them.
Given our analogy between mathematics and natural science,
let's first ask for the source of comparable beliefs in the physical
sciences. Rudimentary accounts of elementary physical knowledge
most often begin with simple, enumerative induction: every swan in
my sample is white; therefore, I conclude, all swans are white. It
might be argued that our most basic general set theoretic beliefs are
justified in the same way, but this line is unconvincing.77 Do I test
77
The idea that mathematics is a simple inductive science goes back to Mill
(1843), bk. 2, chs. 5 and 6. Objections have been mounted by many writers,
including Frege (1884), §§9-10, Hempel (1945), and Kim (1981).
68 PERCEPTION AND I N T U I T I O N

to see whether or not the two sets of fingers on my right and left
hands can be combined to form a larger set of fingers? No, once I
am able to understand the question, the answer is obvious. Would I
be more sure that two sets can be combined if I successfully
combined a wide variety of sets containing different kinds of
objects? No,78 but the observation of white swans in a wide variety
of different environments does add support for the claim that they
are all white. Evidently, particular observations provide a very
different type of support for general hypotheses like 'any two sets
can be combined' than they do for general hypotheses like 'all
swans are white'.
Does this mean that set theory is dramatically different from
physical science, after all? Where there is an analogy, there must be
differences as well as similarities, but I think we've not yet reached
the point at which our comparison between mathematics and
natural science is of no further use. Rather, I think that we've filled
in the details incorrectly, that our primitive general set theoretic
beliefs actually correspond, not to simple enumerative inductions,
but to primitive general beliefs about physical objects that are no
more subject to simple inductive support than their set theoretic
counterparts.
Consider, for example, the child's hard-won beliefs that physical
objects exist independently of the human viewer, that they are
independent of their state of motion. Can these be tested by
observation of particular examples? Does the observation that a
teacup persists when it's moved across the table add support over
and above that provided by similar observations of coffee-cups?
Evidently not. These are primitive general beliefs about physical
objects that are not supported by simple enumerative induction. We
cannot check to see whether or not physical objects persist when no
one is observing them, but we believe it nevertheless, and beliefs of
this sort appear at the most elementary levels of our physical theory
of the world.
Hebb's analysis of neural operations provides one possible
account of their source. Recall that repeated viewings of a triangle's
apex lead to the development of a first-order cell-assembly that
78
In childhood, such manipulations with a variety of sets helped engender my
ability to see sets in the first place, but once I have this ability, my conviction that
two sets can be combined doesn't depend on my testing a variety of the sets I can
now see.
 INTUITION 69

responds to angles of like magnitude and orientation. Assemblies at

this level respond to particular contours in the visual field, simple
tastes, localized tactile pressures, and the like. The same mechan-
isms that produce assemblies at this simple level are capable of
producing higher-order assemblies as well, such as the integrated
triangle assembly. Similarly, repeated viewing of a single physical
object from one perspective produces a second-order assembly
which integrates the first-order assemblies for the contours of the
object's parts. And finally, manipulation of the object, or seeing it
in motion, permits the development of a third-order assembly
integrating the assemblies for the object's various perspectives. At
this point, perception of the object involves a complex phase
sequence of stimulations of all these assemblies.79
With these mechanisms in place, the subject is able to perceive an
independent, continuing physical object. But there is no reason to
suppose that neural developments come to an end here. Hebb
suggests:
Fundamental . . . is the generalized idea of a thing, an object, a space-
occupying and sense-stimulating something, as the activity of a higher-
order cell-assembly made up of neurons that are usually or always active in
the relatively small number of different situations of infancy when a visible,
tangible object attracts attention. Those neurons must be a small
proportion of the total number excited on any one of such occasions but
may still be a large number in absolute terms. The theoretical possibility of
such an assembly is clear and the psychological support for its existence is
also clear. (Hebb (1980), 109)

This fourth-order assembly would correspond to the general

concept of a physical object. It would be stimulated during the
phase sequence associated with perception of a particular physical
object when attention is drawn to its more general features.
A subject with such an assembly would automatically have
various general beliefs about the nature of the objects that stimulate
it; we might say that these beliefs are 'built into' the cell-assembly
much as three-sidedness is built into the triangle-detector in the
form of mechanisms stimulating eye movements from one corner to
another, or three-angledness in the form of the three first-order
corner components. Crudely put, the very structure of one's
79
See Hebb (1980), 107-8. Hebb cites work on hierarchies of neurons that
provides physiological support for the idea of higher-order assemblies.
70 P E R C E P T I O N AND I N T U I T I O N

triangle-detector guarantees that one will believe any triangle to be

three-sided. Similarly, anyone with a general physical object
assembly would believe that physical objects are 'space-occupying
and sense-stimulating', to use Hebb's examples, or observation-
and trajectory-independent, to use examples mentioned earlier.
These are primitive, very general beliefs about the nature of
whatever stimulates the appropriate higher-order assembly. I call
them 'intuitive beliefs1.
What goes for physical objects should also go for sets: the
development of higher-order cell-assemblies responsive to particu-
lar sets gives rise to an even higher-order assembly corresponding to
the general concept of set. The structure of this general set assembly
is then responsible for various intuitive beliefs about sets, for
example that they have number properties, that those number
properties don't change when the elements are moved (barring
mishap), that they have various subsets, that they can be combined,
and so on. And these intuitions underlie the most basic axioms of
our scientific theory of sets.80
I've suggested that the very structure of one's general physical
object assembly gives one some intuitive beliefs about physical
objects, for example that objects can look different from different
points of view or that they don't cease to be when we cease to see
them, and that one's general set assembly gives one intuitive beliefs
about sets, for example that any two objects can be collected into a
set. Of course, it's deceptive to describe these beliefs in this way,
because they are not, in fact, linguistic. A child of two, for example,
is perfectly capable of perceiving particular physical objects, and
thus has or will soon have intuitive beliefs about physical objects in
general, but she lacks the vocabulary to express them as I have. In
short, such beliefs are accessible to those who lack the linguistic
terms, but not to those who lack the concept. 81 When a term for the
concept is introduced, linguistic expressions of intuitive beliefs will
naturally seem obvious, too obvious to benefit from simple
enumeration.
Granting, then, that there are such primitive, general beliefs
about physical objects and about sets, we should inquire into their
epistemological status. It's already been noted that they are not
80
Hebb (1980, esp. pp. 122-4) discusses mathematical cases as part of his
theory.
81
See Hebb (1980), 108-9.
 INTUITION 71

inductively supported, but this should not be taken to imply that

they are, in any sense, infallible. The first and most obvious source
of potential error is in the uncertain transition from intuitive belief
to linguistic formulation. Because all but the most severely disabled
eventually attain the requisite cell-assemblies, widespread agree-
ment about an attempted linguistic formulation constitutes one of
its best claims to correctness. On this account, then, it is legitimate
to suspect the claim of a single scientist as to the intuitiveness of a
certain principle if few others share this opinion. 82
A second source of potential error is the distinct possibility that
the intuitive belief itself is false.83 We might be radically mistaken
in the concepts we form; perhaps stimulation by aspects of things
causes us to form assemblies which embody features very different
from those of the things themselves; I suppose it is possible that
physical objects do in fact disappear when no one is watching, or
that sets actually don't have subsets, hard as it is for us to imagine
such things. Some intuitive beliefs have in fact been falsified by the
progress of science, for example the belief that, at any given
moment, a physical object is in a certain location and moving at a
certain speed, or that every property determines a set of things with
that property.84 Thus, in scientific contexts, intuitive beliefs must
be tested like any other hypothesis, and like any other hypothesis,
they can be overthrown.
82
For example, the beliefs often described as Godel's 'intuitions' about various
consequences of the continuum hypothesis {1947/64, pp. 479-80) are not widely
shared. I should also note that these judgements of Godel's are more esoteric than
the primitive propositions I've been characterizing as linguistic formulations of
intuitive beliefs. In many such cases—when advanced ideas are called 'intuitive'—
what is really at work is a hunch or conjecture based on mathematical experience, a
theoretical judgement of the sort a natural scientist makes when her familiarity with
the field suggests that this, not that, is the sort of theory likely to work. For the set
theoretic realist, this counts as a theoretical, rather than an intuitive, justification. I'll
come back to Godel's case briefly in ch. 4, sect.3, below.
83
Or 'incorrect' in some sense. There is some difficulty with classifying intuitive
beliefs as true or false because they are non-linguistic, and probably non-
propositional. Still, they can be classified as correct or incorrect, as tending towards
success or failure in their behaviour-guiding function. (See Goldman (1977), 276.)
I'll equivocate a bit and continue to use the words 'true' and 'false'.
84
The set theoretic principle that every property determines a set of things that
have that property is called 'unlimited comprehension'. It is false because it would
allow the formation of Russell's set, which then leads to paradox. It has been
replaced in axiomatic set theory by Zermelo's Separation Axiom, which asserts only
that every property determines the set of things in a previously given set with that
property; within a fixed set, we're allowed to 'separate' the elements with that
property from the rest. See Enderton (1977), 4-6, 20-1.
72 P E R C E P T I O N AND I N T U I T I O N

This highlights the central epistemological question: does the

intuitiveness of a belief count in its favour? The extent to which a
claim strikes us as obvious, and the degree of community agreement
on this degree of obviousness, both constitute evidence that the
claim is a good linguistic version of a primitive intuitive belief.
We've already seen that intuitive beliefs themselves can be false,
that intuitiveness is not conclusive evidence, that it can be
outweighed by opposing theoretical evidence, but we now ask
whether it provides any support at all for the truth of the claim.
There is no doubt that such evidence has been counted in favour of
various axioms in the history of set theory, but is there any
rationale for this practice?
In line with the earlier discussion of Gettier's problem and the
causal theory, we might rephrase our question as: does a true
intuitive belief count as knowledge? This rendition is deceptive
because we've already granted that intuitive support is not enough
in itself to fully justify a claim, but I think this formulation can still
be used to focus the main problem. Recall that Goldman, in
response to Gettier, added a causal requirement to the traditional,
justified-true-belief account of knowledge, and that this require-
ment is seen by many as fatally damaging to Platonism. But, our
intuitive beliefs are products of our cell-assemblies, and the
processes responsible for generating those—a combination of
evolutionary pressures on our ancestors that determine our initial
brain formation and the sum of our childhood interactions with
physical objects and sets—are causal. They are suitably causal
because it is the corresponding general facts about the environment
that both exert the evolutionary pressure and provide the childhood
interactions. The problem now comes not from Goldman's
addition, but from one of the traditional requirements, namely, that
the belief be justified: Steve's belief in the Pairing Axiom may be
intuitive—the axiom may seem obvious to him and to us—but
what if he can offer no justification beyond those feelings?
Turning this question into an outright objection to intuitive
evidence would require the assumption that a justification must
always take the form of a convincing series of reasons available to
the knower. In contemporary epistemology, this is called 'internal-
ism',85 The 'externalist', by contrast, insists that a belief can be
85
This approach goes back to Descartes (1641). For the contemporary debate
between internalists and externalists, see Bonjour (1980) and Goldman (1980).
 INTUITION 73

justified even though the knower is ignorant of that justification.

Consider, for example, Steve's perceptual belief that there is a tree
in front of him. It is generated by a suitable, causal process, but let's
suppose that Steve has no knowledge of optics, retinas, or brain
function, that he can produce no reasons for his belief. Does this
mean that Steve doesn't know there is a tree in front of him? Or, to
take a more exotic example from Goldman,86 consider the case of
the professional chicken-sexer. The man looks at a chick and comes
to believe that it is male, but he has no awareness of the process by
means of which he comes to this judgement. Given that he
invariably proves right in his classifications, we are inclined to say
he knows the chick to be a male. But he can offer no reasons, no
arguments, no explicit justifications.
I side here with the externalist, rejecting the demand that Steve be
prepared to justify his belief in the Pairing Axiom. On this view, it is
enough that the causal process that generates the belief be 'reliable',
that is, the sort of process that generally leads to true beliefs. This is
true in the perceptual case, in the chicken-sexing case, and, if our
assumptions are correct, in the intuitive case as well. Thus the
strength of Steve's conviction that Pairing is obviously true, along
with the prevalence of similar convictions in others, supports the
claim that Pairing is a good linguistic rendering of an intuitive
belief, and the fact that a belief is intuitive lends prima-facie
support to the claim that it is true. If Pairing is in fact true, and if
further theoretical support is forthcoming—for example, evidence
that the axiom is consistent, that it produces theorems of the sort
expected, and so on—then it seems Steve's belief can amount to
knowledge.
One peculiarity of intuitive evidence should be noted. The
acquisition of intuitive beliefs doesn't depend on any particular
experience, that is, any sufficiently rich course of experience will
produce the required cell-assemblies:
In the case of visual perception, what assemblies develop and become the
basis of perception is fully dependent on an innate property of the
organism, the reflex responsiveness of the eye muscles, as well as on the
innately determined structure of the striate and peristriate cortex, which—
according to the theory—is what makes the formation of cell-assemblies

86
Goldman (1975), 114. He also defends externalism in Goldman (1976 and
1979).
74 P E R C E P T I O N AND I N T U I T I O N

inevitable, given exposure to a normal visual environment. (Hebb (1980),

105)

So, though experience is needed to form the concepts, once the

concepts are in place, no further experience is needed to produce
intuitive beliefs. This means that in so far as intuitive beliefs are
supported by their being intuitive, that support is what's called
'impurely a priori1. Notice, however, that it doesn't follow that
even these primitive mathematical beliefs are a priori. Without the
corroboration of suitable theoretical supports, no intuitive belief
can count as more than mere conjecture.
I should also emphasize that the particular mathematical
intuitions discussed here are by no means the end of the story.
Because of my interest in set perception, I've concentrated on the
intuitions involved in the perception of small sets of medium-sized
physical objects, but I don't want to suggest that these discrete
intuitions, the ones that underlie number theory and the beginnings
of set theory, are the only intuitions relevant to a complete account
of set theoretic knowledge. To begin with, part of perceiving a
physical object is perceiving it as existing in external space, that is,
perceiving a boundary between it and the space surrounding it. The
concept of this space and the ability to perceive a boundary or
shape develop along with the concept of physical object itself, as a
result of the child's interactions with the environment. 87 As in the
case of the triangle-detector, neurological correlates of boundaries
are partly constituted by motor stimulations for eye-movements
along edges of a figure; perceived shapes are closely related to acts
of tracing.
Thereafter, just as children gradually develop the ideas of
inclusion, collection, and numeration leading to the discrete set
concept, they also develop parallel ideas based on part/whole and
enclosure relations rather than inclusion, proximity and distance
rather than collection, measurement rather than number. 88 These
developments, beginning, as their set theoretic counterparts do, in
perception and action, lead eventually to the perception of lines—
edges, intersections of planes, trajectories—as continuous struc-
tures. By the age of ten or twelve, the child expresses intuitive

^ See Piaget (1937), or Phillips (1975), chs. 2-$.

S8
See Piaget and Inhelder (1948).
 GODELIAN PLATONISM 75

beliefs about geometric figures that reveal a primitive notion of

continuity.
These intuitions play a role in our systematic thinking in
geometry and analysis that is analogous to the role of intuitions of
discrete collections in arithmetic and that of intuitions of physical
objects in natural science. In part, for example, they lead to the
obviousness of density, and even the Dedekind-style continuity
axiom.89 The ability of set theoretic methods to provide a
consistent rendering of our confused intuitive beliefs about the
relation of the line to its smallest parts is one of its greatest
achievements, and one of its strongest supports.90
Finally, let me repeat: I'm not suggesting that all, or even most,
epistemic support for our theory of sets is intuitive. In many cases,
set theoretic methodology has more in common with the natural
scientist's hypothesis formation and testing than with the caricature
of the mathematician writing down a few obvious truths and
proceeding to draw logical consequences. As the science/mathematics
analogy would indicate, our set theoretic hypotheses demand
theoretical or extrinsic support, that is, support, as in natural
science, in terms of verifiable consequences, lack of disconfirma-
tion, breadth and explanatory power, intertheoretic connections,
simplicity, elegance, and so on. A preliminary description of the
important role of such non-intuitive, non-demonstrative justifica-
tions in modern set theory will be sketched in Chapter 4.

4, Godelian Platonism

In this chapter, I've sketched the epistemological beginnings of set

theoretic realism, a version of compromise Platonism. The differ-
ences between this view and Quine/Putnam Platonism should be
clear enough from the discussion in the final section of Chapter 1,
but its relationship to Godelian Platonism has been less explicitly
89
Density is the claim that between any two points there is another. Dedekind's
axiom is a bit more complex. If the points on a line are divided into two (non-empty)
groups, and the points of one group are all to the left of all points in the other, then
there is either a right-most point in one group, or a left-most point in the other. In
other words, if you cut a line in two, there's always a point at which you cut it. See
Dedekind(1872).
90
See ch. 3, sect. 1, below.
76 P E R C E P T I O N AND I N T U I T I O N

drawn. I'll conclude this chapter by further detailing my consider-

able debt to Godel and by locating our major disagreement.91
Godel's Platonism rests on an analogy between mathematics and
natural science, an analogy he traces back to Russell.92 Mathemat-
ical things are taken to be as objective as the objects of the natural
sciences: '[Sets] may . . . also be conceived as real objects . . .
existing independently of our definitions and constructions' {Godel
(1944), 456). The next stage of the analogy is epistemological: 'The
analogy between mathematics and a natural science . . . compares
the axioms of logic and mathematics with the laws of nature and
logical evidence with sense perception . . .' {Godel (1944), 449);
'we do have something like a perception also of the objects of set
theory . . .' (Godel (1947/64), 483-4). The problem, for Godel as
for the set theoretic realist, is to explicate this perception-like
connection.
To flesh out his proposal, Godel considers the details of our
ordinary perceptual experience and concludes that
in the case of physical experience, we form our ideas . . . of . . . objects on
the basis of something else which is immediately given, . . . That something
besides the sensations actually is immediately given follows . . , from the
fact that . . . our ideas referring to physical objects contain constituents
qualitatively different from sensations or mere combinations of sensations,
e.g., the idea of object itself . . . (Godel (1947/64), 484)

Here the set theoretic realist agrees, and proposes that the source of
this extra constituent, what bridges the gap between retinal
stimulation and perception, is the neural cell-assembly. She agrees
also that these conceptual elements of perceptual experience
'represent an aspect of objective reality, but, as opposed to the
sensations, their presence in us may be due to another kind of
relationship between ourselves and reality' (Godel (1947/64), 484).
For the set theoretic realist, this 'other kind of relationship' is the
complex causal process that produces the cell-assembly, namely,
91
There are others. First, as was emphasized in ch. 1, sect. 4, any compromise
Platonist will place more weight than Godel does on the indispensability arguments.
Second, as remarked in sect. 3 above, Godel's 'intuitions' cover more esoteric cases
than the set theoretic realist's, cases involving what 1 would take to be theoretical
rather than intuitive evidence. In fact, Godel's text doesn't explicitly count these as
intuitions, so on this point I may be disagreeing more with his readers than with
Godel himself.
92
See Godel (1944), 449. For Russell's views, see Russell (1906), (1907), and
(1919), esp. p. 169.
 GODELIAN PLATONISM 77

the evolutionary pressures on our ancestors and our childhood

experiences with objects,
Now what of mathematics? Godel goes on to suggest that
'Evidently the "given" underlying mathematics is closely related to
the abstract elements contained in our empirical ideas' (Godel
(1947/64), 484). Again the set theoretic realist agrees, understand-
ing the 'given1 here to mean the intuitive beliefs that underlie the
simplest set theoretic axioms. These intuitions are 'closely related'
to the cell-assemblies responsible for the 'abstract elements1 of our
perceptual beliefs: the object and set concepts. In such cases,
indeed, 'the axioms force themselves upon us as being true'.93 This
sort of obviousness is evidence for the intuitive basis of the axiom in
question.
Finally, Godel and I agree that not all axioms can be justified on
intuitive grounds, that the science/mathematics analogy should be
extended one step further, to the level of scientific hypotheses:
the axioms need not necessarily be evident in themselves, but rather their
justification lies (exactly as in physics) in the fact that they make it possible
for these 'sense perceptions' to be deduced . . . (Godel (1944), 449)
besides mathematical intuition, there exists another (though only probable)
criterion of the truth of mathematical axioms, namely their fruitfulness in
mathematics and, one may add, possibly also in physics. (Godel (1947/64),
485).
He even supplies a list of various particular forms these justifica-
tions might take:
Success here means fruitfulness in consequences, in particular in 'verifiable'
consequences, i.e., consequences demonstrable without the new axiom,
whose proofs with the help of the new axiom, however, are considerably
simpler and easier to discover, and make it possible to contract into one
proof many different proofs. The axioms for the system of real numbers,
rejected by the intuitionists, have in this sense been verified to some extent,
owing to the fact that analytical number theory frequently allows one to
prove number-theoretical theorems which, in a more cumbersome way,
can subsequently be verified by elementary methods. (Godel (1947/64),
477)
And in a passage quoted earlier:
There might exist axioms so abundant in their verifiable consequences,
93
Godel (1947/64), 484.
78 P E R C E P T I O N AND I N T U I T I O N

shedding so much light upon a whole field, and yielding such powerful
methods for solving problems . . . that, no matter whether or not they are
[intuitive], they would have to be accepted at least in the same sense as any
well-established physical theory. (Godel (1947/64), 477)

Examples are described in Chapter 4 answering to each of these

forms of justification and others like them.
This quick survey of Codel's writings surely suggests that set
theoretic realism should be understood as a further development
along the same lines; indeed it was the passage about the
relationship between 'the "given" underlying mathematics' and 'the
abstract elements contained in our empirical ideas' that set me on
this road in the first place. Still, Chihara has rightly pointed out that
my quotations and references are highly selective, sometimes
deceptively so.94 In particular, I neglected to site Godel's assertion
that 'the objects of transfinite set theory . . . clearly do not belong to
the physical world and even their indirect connection with physical
experience is very loose, . ." (Goclel (1947/64), 483). And, in
featuring Godel's claim that 'we do have something like a
perception also of the objects of set theory', I omitted the qualifier
'despite their remoteness from sense experience'.95 Godel's insist-
ence on the traditional Platonistic characterization of mathematical
objects as non-spatio-temporal clearly disqualifies set theoretic
realism as a straightforward development of his thinking.
Of course, my motivation for bringing sets into the physical
world and for tying mathematical intuition so closely to ordinary
perception is naturalism; set theoretic realism seems to me the most
promising approach for bringing mathematical ontology and
epistemology into line with our overall scientific world-view.
Godel, by contrast, not only characterizes sets as traditional
Platonistic entities, he also goes on to postulate a realm of non-
physical, non-spatial, mentalistic monads, as well. It is worth
asking why.
Consider the following problem: the human mind is finite and
94
See Chihara (1982), written partly in reply to Maddy (1980). {John Burgess
once warned me that I wasn't distinguishing clearly enough between my own ideas
and Godel's, but this wise counsel fell on deaf ears.) In the course of arguing that my
view is not Godel's, Chihara raises two difficulties for set theoretic realism. The
first, concerning the status of the continuum hypothesis, will be touched on in
ch. 4, sect. 3, below. The second, concerning singletons, was the inspiration for
ch, 5, seer. 1.
45
Both quotations in this sentence comes from Godel (1947/641, 483-4.
 G O D E L I A N PLATONISM 79

the set theoretic hierarchy is infinite. Presumably any contact

between my mind and the iterative hierarchy can involve at most
finitely much of the latter structure. But in that case, I might just as
well be related to any one of a host of another structures that agree
with the standard hierarchy only on the minuscule finite portion
I've managed to grasp. According to Benacerraf, Godel felt that his
monads were immune to this finiteness problem, that they could
somehow gain unambiguous access to the full hierarchy.96 Thus his
monadology would succeed where naturalism purportedly fails.

The question of how our finite minds make contact with the infinite
is given a very general formulation by Kripke in his interpretation
of Ludwig Wittgenstein.97 In its strongest form, this argument, the
so-called 'rule-following argument', applies not only to the case of
the mathematical infinite, but also to any rule with an indefinite
number of potential applications. Consider, for example, our
ordinary use of the word 'triangle'. Kripke's Wittgenstein98 argues
that all our training, our past usages, our mental images, our stated
and unstated intentions, our associations, and so on, cannot
predetermine whether it is correct or incorrect to call a given figure
a triangle. This is because, for example, these constraints can all be
perversely interpreted to conform to the set of things that are
triangular up to now or square after now as completely as they do
to the set of triangular things. So nothing we've associated with the
word 'triangle' can predetermine whether 'triangle' should now be
applied to this figure A rather than that D. The argument doesn't
depend on the fact that there are indefinitely many triangles; it can
also be applied to show that nothing predetermines whether this
person or that is correctly referred to as 'Kripke', because
everything I've associated with the name 'Kripke' can now be
interpreted as applying as well to a being consisting of Kripke's
space-time worm up till now and Putnam's afterwards.
These conclusions may seem outrageous, and indeed they should,
but I will leave the interested reader to more complete expositions
96
Benacerraf made this observation in an oral presentation that became
Benacerraf (1985). As far as I know, the details of Go'del's monadology and how it is
to overcome this problem are as scant as his account of intuition.
97
See Kripke (1982). For Wittgenstein's views, see Wittgenstein (1953) and the
posthumous (1978).
98
I put it this way because there is some debate over whether or not Kripke's
argument is really Wittgenstein's.
80 P E R C E P T I O N AND I N T U I T I O N

of this fascinating paradox. All I want to do here is to indicate—in

light of Godel's motivations—what I take to be the main
ingredients of a naturalistic solution." It begins, of course, with the
idea that my linguistic training sets up a neural connection between
the word 'triangle' and my triangle-detector. This is not a complete
reply, however, because of the inevitable physical limitations on my
neural processors. The range of the term 'triangle' cannot be
identified with the set of things that stimulate my detector because
my detector is sometimes wrong, and even if it weren't, it couldn't
respond to triangles too small, too large, or too far away.
At this point, the realist must appeal to the objective fact that
triangles are more like one another than like squares, that is, to the
fact that there is an objective difference, noted in the second section
of Chapter 1, between random and natural collections.100 Our best
theory of what my detector responds to involves not the sceptic's
random collections, but those collections science takes to be the
natural ones. Triangles, then, are those things belonging to the
natural collection that includes most of the things that stimulate my
detector. This distinction between natural and random may be
difficult to pin down; it may be a matter of degree rather than all or
nothing, it may ultimately require an ontological commitment to
properties or universals, but it is a crucial part of the naturalist's
world view, and it serves to rule out the gerrymandered interpreta-
tions on which the sceptic's arguments depend. 101
I have tried in this section to clarify the nature and the origin of my
central disagreement with Godel. This should not be allowed to
obscure my obvious debt to his thinking, both in the formulation of
mathematical realism in Chapter 1 above and in the account of
mathematical perception and intuition offered in this chapter. His
influence will again be prominent in Chapter 4, when 1 take up the
subject of theoretical evidence and justification, but before turning
to that more esoteric epistemological study, I want to touch on
another Benacerrafian worry about Platonism, an ontological one
this time.
99
For a fuller account of both the paradox and its naturalistic solution, see
Maddy (1984b).
100
And between natural and unnatural individuals, to deal with the 'Kripke' case.
101
Putnam's (1977) and (1980) present a puzzle in some ways similar to Kripke's
version of the rule-following argument. In reply to Putnam, Merrill (1980) and
Lewis (1983; 1984) use natural kinds much as Maddy (1984b) does in reply to
Wittgenstein.
 3
NUMBERS

1. What numbers could not be

The widely cited epistemological challenge to Platonism has been

my focus so far, and I will return to epistemology in the next
chapter to take up an equally important, though less talked-of
aspect of mathematical knowledge. But before doing that, I would
be negligent not to discuss a celebrated ontological challenge to
Platonism which some observers consider nearly as daunting as the
epistemological.1 The set theoretic realist's answer is implicit in
what has already been said, but it will take some effort to see this
and to draw out the details.
Let me frame the problem with another touch of simplified
history. The emergence of set theory as a foundational theory can
be traced to two separate lines of development that eventually
converged.2 The first of these began with concern over the
foundations of the calculus.3 The much- and deservedly-criticized
infinitesimals were replaced by Weierstrass's theory of limits, which
in turn depended on a theory of the real numbers proposed in
different forms by Cantor and Dedekind.4 These accounts are
thoroughly set theoretic; Dedekind's, for example, makes essential
use of infinite sets of rational numbers. By these means, he clarifies
the notion of continuity, defines the real numbers, and proves that
they are, in fact, continuous. Thus, by identifying real numbers
with certain sets (called 'Dedekind cuts'), Dedekind obtains a rich
1
Among those bothered are Benacerraf (1965), Jubien (1977), Kitcher (1978;
1983, p. 104), Field (1980, p. 126, n. 66; 1989, pp. 20-5), and Resnik (1981, p. 529).
2
I'll concentrate here on two foundational motivations. Cantor's more purely
mathematical interests will be discussed in the next chapter.
3
This was touched on in ch. 1, sect. 4, above. The resulting worry about the
infinite produced the three great schools of thought in the philosophy of
mathematics that flourished in the early decades of our century.
4
Cantor's theory appears in Cantor (1872). See Dauben (1979), ch. 2, for
discussion. Dedekind's account, developed somewhat earlier, first appeared in
Dedekind (1872). Enderton (1977), ch. 5, presents Dedekind's more elegant
formulation.
82 NUMBERS

and explanatory theory of their nature and behaviour that puts the
calculus and higher analysis on a consistent foundation. To this
day, set theory provides our best account of mathematical analysis,
which in turn plays a central role in our most successful physical
theories. This achievement is one of the strongest theoretical
supports for the mathematical theory of sets; it plays an indispens-
able role in our best theory of the world.
Meanwhile, from a more philosophical orientation, Frege was
concerned to provide a foundation for ordinary arithmetic. 5 He
was scandalized by the lack of understanding, even among
mathematicians, of the fundamental concepts of their subject, in
particular, the concept of a natural number. Frege's aim was to
show that arithmetic is in fact a branch of logic, but in doing so he
made use of extensions of concepts, that is, often infinite
collections. This project failed, as has been noted, 6 but its core
has been incorporated into modern set theory. As with the reals,
the natural numbers are identified with certain sets, and all their
basic properties, once assumed as axioms, become provable, explic-
able.7 Thus modern set theory also provides our best account
of the natural numbers, another key ingredient in our overall
theorizing.
Here we have a scientific success story. It depends, however, on
identifying numbers with sets. In the opening pages of Chapter 1, I
suggested that the philosophy and foundations ot set theory are of
interest regardless of whether or not all mathematical objects are
properly taken to be sets, but in this case, where the theoretical
justification of the subject, indeed the very motivation for the
subject, depends so centrally on these particular identifications,
philosophical questions about the nature and propriety of set
theoretic reduction cannot be put aside. It is this question—of the
relationship between sets and numbers—that I want to consider in
this chapter.

5
See Frege (1884), introd,
6
See ch. 1, sect. 4, above.
I don't mean to suggest that all proofs are also explanations, but some clearly
are. For example, the set theoretic account of natural numbers tells us why
multiplication is commutative: because the cross product A X B is equinumerous
with B x A. (See Drake (1974), 52.) This same fact can be proved from the Peano
axioms, but the proof requires a series of clever lemmas and sheds little light on why
the theorem is true. (See Enderton (1977), 81-2.) Sreiner (1978) makes an effort to
characterize mathematical explanation.
 WHAT N U M B E R S C O U L D NOT BE 83

To see the source of the problem, let's return for the moment to
Frege and reconsider his central ontological query: what is a
number?
Take a numerical statement like 'there are two boys playing in
the garden'. I've already reviewed8 Frege's observation that the
subject of this numerical ascription cannot be the mere physical
stuff that makes up the boys, because that physical stuff could be
divided into units in various different ways—two boys, twenty boy-
parts (heads, torsos, arms, hands, legs, and feet), millions of cells,
even more molecules, still more atoms, and so on. What has
changed, Frege argues, from the occasion when we judge that the
stuff is two boys to that when we judge it to be millions of atoms is
the substitution of the concept 'atom in this mass of physical stuff
for the concept 'boy in this mass of physical stuff. This leads him to
the conclusion that 'the content of a statement of number is an
assertion about a concept'.9
But to identify the content of a statement of number is not yet to
identify the number itself. Frege's analysis suggests that number is
something various concepts can share, for example 'boy in the
garden' and 'hand on my keyboard'. In Fregean terms, this is to say
that two is a concept under which other concepts can fall, a second-
order concept, as it were. But, however straightforward this might
sound, Frege insists, for reasons I'll consider later, that a number is
a thing, a 'self-subsistent object',10 rather than a concept. So,
instead of saying that two is the concept that applies to any concept
equinumerous 11 with 'boy in the garden', he identifies the number
with the extension of that concept, that is, with the collection of all
concepts equinumerous with 'boy in the garden'.
The trouble with this account is that two turns out to be a very
large collection indeed. In a set theory like Frege's, with the (false)
principle of unlimited comprehension, we could form this set, but
alas, we could also form the paradoxical Russell set. When this
inconsistent naive set theory is replaced by the contemporary
8
In ch. 2, sect. 2, above.
9
Frege (1884), § 46.
10
Frege (1884), § 55.
1
' This sounds circular, defining the number two in terms of equality in number,
but in fact, Frege defines 'equinumerous' without reference to numbers. Two
concepts are equinumerous if there is a one-to-one correspondence between the
things falling under the first and the things falling under the second. See Frege
(1884), § 63, or Enderton (1977), 128-9.
84 NUMBERS

axiomatic version based on the iterative conception, two, as defined

here, no longer exists. To see this, notice that new two-membered
sets are formed at each stage, so there is no stage at which the set of
all two-membered sets is formed.
Thus, in contemporary set theory, two is customarily identified
not with the (non-existent) collection of all two-membered sets, but
with some particularly convenient example of a two-membered set.
Standardly, the natural numbers are taken to be the set of finite von
Neumann ordinals, that is, 0 for 0, {0} for 1, (0, {0} } for two,
{0, {0}, (0, {0}}} for 3, and so on. 12 Other, similar,
identifications would do as well, for example Zermelo's 0 for 0,
{0} for 1, { {0} } for 2, { { {0} } } for 3, and so on.13 And it is this
fact that generates an ontological question about numbers.

Once again, the version of this problem most exhaustively cited and
discussed in the contemporary philosophical literature derives from
a paper by Paul Benacerraf, this time 'What numbers could not
be'.14 Benacerraf asks us to consider the education of two
hypothetical young children, Ernie (for Ernst Zermelo) and Johnny
(for John von Neumann).
Ernie and Johnny are both brought up on set theory. When the
time comes to learn arithmetic, Ernie is told, to his delight, that he
already knows about the numbers; they are 0 (called 'zero'), {0}
(called 'one'), (0, {0}} (called 'two'), and so on. His teachers
define the operations of addition and multiplication on these sets,
and when all the relabelling is done, Ernie counts and does
arithmetic just like his schoolmates. Johnny's story is exactly the
same, except that he is told that the Zermelo ordinals are the
numbers. He also counts and does arithmetic in agreement with his
schoolmates, and with Ernie. The boys enjoy doing sums together,
learning about primes, searching for perfect numbers, and so on.
But Ernie and Johnny are curious little boys; they want to know
everything they can about these wonderful things, the numbers. In
12
See e.g. Enderton (1977), 68. Von Neumann's proposal is contained in his
(1923), 347.
0
See Zermelo (19086), 205. In fact, there are reasons why Zermelo's version
isn't as good as von Neumann's. For example, von Neumann's account works just as
well for infinite numbers as for finite, and its 'less than' relation is extremely simple:
membership.
14
Benacerraf (1965). See also Parsons (1965). The argument is further developed
bvKitcher(1978).
 WHAT N U M B E R S C O U L D NOT BE 85

the process, Ernie discovers the surprising fact that one is a member
of three. In fact, he generalizes, if n is bigger than m, then m is a
member of n. Filled with enthusiasm, he brings this fact to the
attention of his favourite playmate. But here, sadly, the budding
mathematical collaboration breaks down. Johnny not only fails to
share Ernie's enthusiasm, he declares the prized theorem to be
outright false! He won't even admit that three has three members!
According to Benacerraf, the moral of this sad story—or one of
them—goes like this: if numbers are sets, then they must be some
particular sets. Any choice of particular sets will exhibit properties
that go beyond what ordinary arithmetic tells us about the
numbers. (Ordinary arithmetic is mute on the subjects Ernie and
Johnny debate. Their classmates are puzzled by the very questions
these boys take so much to heart.) If one of these particular choices
is the correct one, that is, if one sequence of sets really is the
numbers, then there ought to be arguments that tell us which
sequence that is.15 (This doesn't seem to be the sort of question that
requires some further, deep number theoretic theorem.) But there
are no such arguments. Therefore, numbers are not sets.
Friends of numbers might be prepared to fall back on the
position that while numbers aren't sets, still they are objects of
some other kind. Suppose, then, that we've identified some
sequence of objects suitable for counting and arithmetic, and we
claim that the fourth of these (we started from zero) is the number
three. Benacerraf argues that this object plays the role of three in
our sequence by virtue of its relations to the other members of the
sequence, but if it is to be singled out, independently of the
sequence, as this object or that, it must have some additional
properties. And, he continues, these additional properties will be
superfluous to the object's numerical functioning, in the same sense
that the properties Ernie and Johnny debated were superfluous,
which leads to the question: why should three have these
superfluous properties and not some others? If this object really is
three, there should be arguments to show that these superfluous
properties are the correct ones, but there are no such arguments.
Therefore, numbers are not objects at all.
15
The metaphysician wants more here than the previously remarked arguments
from convenience. The fact that von Neumann ordinals are more convenient than
the Zermelo ordinals, and hence standard in contemporary set theory, doesn't give
us any further reason to think that they really are the numbers.
86 NUMBERS

This second conclusion is less firmly supported than the first; the
second argument leaves room for the position that numbers are the
sort of objects whose non-superfluous properties are all the
properties they have. 16 If these non-sets are connected closely
enough with sets, perhaps even the explanatory virtues of the set
theoretic reduction can be preserved without the actual identifi-
cation of numbers with sets. Thus, for example, Cantor suggests
that natural numbers are separate entities 'abstracted' from
equinumerous sets, and Dedekind that the reals are 'associated'
with the corresponding cuts. This sort of move obviously flaunts
ontological economy—it's inefficient to overburden our theory
with more things than we need to make it work effectively—but
worse than that, it requires an account of the sort of 'abstraction' or
'association' involved. Neither Cantor nor Dedekind provide this.17
Of course, if we take Benacerraf's argument that natural
numbers aren't sets to be persuasive, as I think we should, an
analogous line of thought shows that the reals can't be sets either.
We could tell a story of Georgie (for Georg Cantor) and Rich (for
Richard Dedekind}, one of whom learns that the reals are Dedekind
cuts and the other of whom that they are Cantor's fundamental
sequences.18 The rest of the story follows as before, and the
conclusion: real numbers aren't sets. 19
If numbers aren't sets after all, the story of scientific success
recited at the beginning of this section is called into question. If its
illumination of the theories of natural and real numbers is to count
as evidence for the theory of sets, we need to understand the true
nature of the ontological relationship between numbers and sets. If
not identity, then what?

2. Numbers as properties

Assuming that numbers aren't sets, the set theoretic realist faces the
16
Steiner (1975tf), 88-92, suggests a move of this sort in his reply to Benacerraf,
but only in an epistemological sense: 'we accept mathematical objects, contra
Benacerraf, but we agree that the only things to know about these objects of any
value are their relationships with other things' (p. 134).
17
For a scorching attack on Cantor's notion of abstraction, see Frege (1979),
68-71.
18
These depend on an idea that goes back to Cauchy. See Enderton (1977), 112.
19
The same goes for other set theoretic reductions. See e.g. Kitcher (1978) on
ordered pairs. The problem for ordered pairs carries over to functions, understood
as sets of ordered pairs, and so on.
 N U M B E R S AS P R O P E R T I E S 87

prospect of adding a new type of entity to her ontology and an

extra epicycle to her epistemology. And, to preserve the explanat-
ory force of the standard set theoretic reduction, she must also
describe a relationship between sets and numbers that makes the
behaviour of, for example, the von Neumann ordinals somehow
relevant to our understanding of numbers. Finally, the efforts to
naturalize the epistemology of set theory will be wasted if the
account of numbers isn't also naturalistic.
Let's begin then with the suggestion that the epistemology for
numbers should be as similar as possible to that given for sets. In
that case, numbers must also be located in space-time. Where, then,
is the number ten?
The easy answer is: ten is located where the set of my fingers is
located, in motion over the keys of my word processor. But if this is
right, then ten is also located where the starting line-up of any
American League baseball team is located, and on the Times best-
seller list, and many other places. Now a set of physical objects can
have a discontinuous location—the set of Angel starting baseball
players is located in left field, right field, second base, and even in
the dugout, with the designated hitter—but only part of the set is
located in each of these places, while the number ten is fully present
in each and every ten-element set. By traditional criteria, this makes
the set of baseball players a particular and the number ten a
universal.20
In terms of the science/mathematics analogy, then, the idea is
this: set theory is the study of sets and their properties, of which
number is one, just as physics is the study of physical objects and
their properties, of which length (for example) is one. To see how
this works in a bit more detail, consider the formal features of the
quantity 'length'.21
First, objects with a given physical quantity, like length, are
comparable; they form a linear ordering with respect to that
quantity. For example, there is a simple linear ordering of medium-
sized, easily movable objects that goes like this: A is shorter than B
if one end of B extends beyond the end of A when they are laid side
by side with the other ends coincident. A method that also works
for stationary objects might go like this: A is shorter than B if a
string that lies straight with one end at each end of A won't reach
both ends of B. Obviously, more sophisticated tests will extend the
20
See ch. 1, sect. 2, above.
21
Here I follow Ellis (1966).
88 NUMBERS

linear ordering further. At this point, a scale of measurement can be

assigned, as long as it agrees with the established linear ordering,
that is, as long as the ordering of the numerical assignments agrees
with the linear ordering. This is not difficult: comparison with any
fair ruler measures length in yards. And finally, the same quantity
can be measured on different scales; length is detected by metre
sticks as well.
Now compare the case of number properties. Sets of easily
movable objects are directly comparable with respect to number; A
is less numerous than B if each member of A can be set beside a
unique member of B in such a way that there are elements of B left
over. The linear ordering of simple sets this produces can be
extended to all sets using the mathematical idea of a one-to-one
correspondence: A is less numerous that B if there is a one-to-one
correspondence between the members of A and a proper subset of
the members of B. Scale can be assigned, for example, by
comparison with the English number words, that is, by counting in
English.22 Or, more elaborately, in terms of one-to-one correspond-
ence with the set of von Neumann ordinals.
The only disanalogy is that there is no room for measuring sets
on different scales, because there is no room for an arbitrary choice
of unit: sets come with their elements already individuated. This is
how sets avoid Frege's objection to physical masses as bearers of
number properties. While it is arbitrary to assign two to the
physical mass that makes up the boys playing in the yard, there is
no arbitrariness in assigning two to the set of boys playing in the
yard. 23
This suggestion—that numbers are properties of sets, analogous
to physical properties, and in particular, to physical quantities—
meets our epistemological desiderata with admirable economy; the
naturalistic epistemology previously described for set theory needs
no elaboration. Just as the perception of physical objects includes
the perception of their properties, so the perception of sets does the
same. Indeed, our account of set perception developed from the
observation that we perceive their number properties; all that is
new here is the further claim that these number properties are the
numbers.
22
Compare Benacerraf (1965), 292: 'The central idea is that [the sequence of
.number words] is a sort of yardstick which we use to measure sets.'
23
Yourgrau (1985) disagrees, but for a cogent reply, see Menzel (1988).
 N U M B E R S AS P R O P E R T I E S 89

Knowledge of numbers is knowledge of sets, because numbers

are properties of sets. Conversely, knowledge of sets presupposes
knowledge of number; for example, Piaget's studies indicate that
subset relations cannot be properly perceived before number
properties.24 From this perspective, arithmetic is part, perhaps the
most important part, of the theory of hereditarily finite sets.25
Neither arithmetic nor this finite set theory enjoys epistemological
priority; the two theories arise together. Arguments that arithmetic
should not be reduced to set theory because set theory is less certain
than arithmetic miss the fact that it makes little sense to separate
the epistemological basis of arithmetic from that of finite set
theory.26 Higher set theory is admittedly less certain than the
theory of hereditarily finite sets, but this is irrelevant.
Furthermore, the significance of the set theoretic reductions is
now clear. The von Neumann ordinals are nothing more than a
measuring rod against which sets are compared for numerical size.
We learn about numbers by learning about the von Neumann
ordinals because they form a canonical sequence that exemplifies
the properties that numbers are. The choice between the von
Neumann and the Zermelo ordinals is no more than the choice
between two different rulers that both measure in metres. The
debate between Ernie and Johnny is like an argument over whether
an inch is wooden or metal.

Some version of the identification of the natural numbers with

properties of sets is considered by both Frege and Benacerraf, and
both writers ultimately reject positions of this sort. Benacerraf s
conclusion is drawn with little strong conviction—he says only that
'"seventeen" need not be considered a predicate of [sets]'27—but

24
See Piaget and Szemiriska (1941), ch. 7.
25
A set is hereditarily finite if it is finite, and its members are finite, and the
members of its members are finite, and so on, (See Enderton (1977), 256.) It might
seem that arithmetic only treats finite sets of physical objects, but the most natural
way of understanding such examples as 'here are three pairs of shoes' involves a set
of three two-membered sets. The restriction to hereditarily finite sets, rather than to
finite sets simpliciter, rules out such things as the singleton containing the set of all
von Neumann ordinals, hardly the sort of thing covered by ordinary arithmetic.
26
Notice, for example, that recursion theory can be developed with equal
naturalness from numbers or from hereditarily finite sets. An argument of the sort
considered in the text appears in Steiner (1975a), ch. 2. Sentiments related to my
own are voiced by Parsons (1965), 173.
27
Benacerraf (1965), 284.
90 NUMBERS

Frege's is harder to evaluate. I will briefly consider a few of their

reasons.
The most conspicuously cited grounds for their opposition to the
property view are grammatical: comparison of the role of number
words with ordinary adjectives and predicates,28 the use of number
words with the definite article, their immunity to pluralization, 29
and so on. Of course, grammar is no infallible guide to the actual
structure of the world, so such evidence must be taken with a grain
of salt. Frege admits as much when he dismisses the contrary
grammatical evidence:
our concern here is to arrive at a concept of number usable for the purposes
of science; we should not, therefore, be deterred by the fact that in the
language of everyday life number appears also in attributive constructions.
(Frege (1884), §57)

Benacerraf also ignores strong grammatical evidence, the evidence

that number words are names, when he later denies that numbers
are objects.30
In fact, Frege's commitment to the objecthood of numbers seems
to waver at the very moment when he presents his own definition.
He writes: 31 'the number which belongs to the concept F is the
extension of the concept "[equinumerous with] the concept F" '. To
the word 'extension1 he appends a footnote: 'I believe that for
"extension of the concept" we could write simply "concept"/ This
surely sounds like a suggestion that numbers could be concepts
rather than objects. Frege immediately rejects his proposal: 'But
this would be open to the two objections . . .'. The first is the
inconclusive grammatical considerations. The second, considerably
more interesting, is 'that concepts can have identical extensions
without themselves coinciding'. This suggests the fact that concepts
are intensional rather than extensional. But whatever the force of
28
Benacerraf (1965), 282-4; Frege (1884), § 57.
29
Frege (1884), §38.
30
In defence of his own view, Frege admits that it leads to some unusual ways of
speaking—e.g. that a number is 'wider or less wide than the extension of some other
concept'—but insists that there is nothing 'to prevent us speaking in this way' (1884,
§69). Benacerraf also speaks somewhat oddly when he says that 'any object
[including Laurence Olivier] can play the role of 3' (1965, p. 291). Unless it is argued
that these oddities are semantic rather than grammatical, a distinction notoriously-
hard to draw, these are further examples of both writers' willingness to ignore
grammatical evidence when need be.
31
All the quotations in this paragraph come from Frege (1884), §68.
 N U M B E R S AS PROPERTIES 91

this objection, Frege does not regard it as decisive. He concludes: 'I

am, as it happens, convinced that both these objections can be met;
but to do this would take us too far afield for present purposes.'
Thus the proposal is rejected on grounds of convenience rather than
principle.
It may seem obvious that Frege is leaving open the possibility that
numbers are concepts, and depending on the relationship between
concepts and properties, perhaps the possibility of a property view
as well, but reading Frege is never a simple matter. He holds, for
example, that 'the concept horse' must refer to an object, rather
than a concept,32 so it can be argued that this tantalizing footnote
suggests no more than that 'the concept "equinumerous with F"'
actually refers to the extension of the concept 'equinumerous with
F'. On this reading, 'concept' could be substituted for 'extension of
the concept' in his original definition because the two actually refer
to the same thing, which isn't, by the way, a concept at all. And the
reason the concepts 'human being' and 'featherless biped' aren't
identical isn't that these expressions have different meanings or
stand for different properties, but that identity, and difference, are
relations between objects, not between concepts.33
I have nothing to contribute to the debate over what Frege
actually had in mind here, but leaving Frege himself behind, I'd like
to examine the idea that numbers might be concepts and the extent
to which the intensionality of concepts stands in the way of such a
view. To see what's at stake here, consider a simple arithmetical
identity: 2 = S(S(0)). 3 4 If Frege's definition had read 'concept' in
place of 'extension of concept', 2 would be the concept 'equinumer-
ous with the concept "identical with 0 or 1"', and the successor of
the successor of 0 would be:
the concept 'equinumerous with the concept "member of the series of
natural numbers ending with S(0}" '
which is
the concept 'equinumerous with the concept "member of the series of
natural numbers ending with the concept 'equinumerous with the series of
natural numbers ending with 0'"'.
32
See Frege (1892^), 45.
33
This line of interpretation appears in Resnik (1965). For other relevant
discussions of Frege, see Hodes (1984) and Luce (1988).
34
'S1 here means 'the successor of.
92 NUMBERS

Now there's no doubt that the extension of 2, so defined, is the

same as the extension of S(S(0)), so defined—each involves being
equinumerous with a concept under which two things fall—but if
coextensive concepts can nevertheless differ, our simple arithmeti-
cal identity is in jeopardy. This is a clear difficulty for the concept
view.
Now let's see how this works out on the property view proposed
here. If 2 is the number property of, for example, the von Neumann
ordinal (0, {0}}, then to have the property 2 is to be equinum-
erous with this set. When successor is defined for von Neumann
ordinals, S(S(0)) turns out to be the same set as {0, {0} } itself, so
being equinumerous with either one is the same as being equinum-
erous with the other.35 The same goes for Zermelo's version. But
unless we're able to affirm, for example, the identity
the property 'equinumerous with {0, {0}, (0, {0} } }'
= 5

the property 'equinumerous with {0, {0} { {0} } }'

a new version of the old Benacerraf problem will arise: which of
these is 3? Which properties really are the numbers? Those defined
in terms of equinumerosity with particular von Neumann ordinals
or those defined in terms of equinumerosity with initial segments of
the Zermelo ordinals? This is the analogous problem for the
property view.
Now let's compare the two problems. A Fregean concept is
closely connected to a predicate; indeed, it is the referent of a
predicate. Leaving Frege's views on identity aside, let us ask when
two predicates are the same. We might insist this is only so when
they are typographically identical, but a more flexible notion allows
for trivial grammatical transformations, for example, that 'is Sam's
only friend' is the same predicate as 'is the only friend of Sam'. On
the other hand, anyone would admit that 'is a featherless biped' is a
different predicate from 'is human'. This leaves intermediate cases
like 'is a bachelor' and 'is an unmarried male'. If we think these two
phrases mean the same thing, because 'bachelor' and 'unmarried
male' mean the same thing, then a natural account of identity
between predicates equates it with synonymy.

'5 The successor of x is x u {x}.

 NUMBERS AS P R O P E R T I E S 93

Now what of properties? We might simply specify that two

properties are the same just when the predicates that pick them out
are synonymous, but we've already seen36 that this approach
doesn't square with our ordinary scientific thinking. Recall that the
same length property can be measured on various yardsticks; the
property 'measures three inches on this yardstick' is the same as
'measures three inches on that yardstick'. Furthermore, length can
be measured in metres as well as yards, so the same property can be
expressed by a predicate mentioning yards and another mentioning
metres. Yet I think no one would suggest that all these predicates
are synonymous. For more dramatic examples, we can turn to
ordinary scientific identities like that of 'temperature1 with 'mean
molecular motion'. These could hardly rest on sameness of
meaning.
For scientific properties, then, synonymy is too strong a
condition. Somewhere between predicates—individuated by same-
ness of meaning—and sets—individuated by sameness of member-
ship—there is an intermediate category of properties.37 One
suggestion current among philosophers of science is that the
appropriate mode of individuation might be specifiable in terms,
not of coextensiveness, but of law-like coextensiveness.38 Thus
'temperature* = 'mean molecular motion' follows from the laws of
physics. By contrast, 'featherless biped' = 'human' is true by the
accident of what species happen to exist in our world; no physical
law would be violated if there were bipedal fish.
If something along these lines can be made to work for physical
properties, our analogy suggests a similar course for mathematical
properties, and in particular, for number. Consider again the
predicates 'equinumerous with (0, {0}, {0, {0}}}' and 'equi-
numerous with (0, {0}, {{0}}}'. They aren't synonymous, but
they are coextensive, so they are different predicates that determine
the same set. But our real concern is whether they express the same
36
In ch. 1, sect. 2, above.
37
In the course of his attack on Catnap's distinction between analytic and
synthetic—see ch. 1, sect. 4, above—Quine casts serious doubt on the notion of
synonymy as well. I don't mean to differ with Quine here. My point is simply that
scientific properties aren't individuated by synonymy even if concepts are. If
synonymy is a bankrupt concept, so much the worse for concepts; my concern is
with properties.
38
See Putnam (1970), 321.
94 NUMBERS

scientific property; for that they must be coextensive by law rather

than by accident. Would a law of mathematics be violated if these
two failed to be coextensive? Of course! Their coextensiveness is
provable from the axioms of set theory. Thus the neo-Benacerrafian
dilemma for property theory—which properties are really the
numbers?—dissolves; understood as scientific properties rather
than sets or predicates, the von Neumann-style numbers and the
Zermelo-style numbers are in fact identical.

I think the part of this story aimed most directly at the problem of
multiple reductions for arithmetic carries over to the analogous
problem for the real numbers. That is: what makes one set theoretic
version of the reals preferable to the others? Answer: nothing; each
version serves to detect and measure the same underlying properties.
But when it comes to the question that inspired Benacerraf's
discussion—what are the natural numbers? or in this case, what are
the real numbers?—the answer is, perhaps not surprisingly, a bit
more complex.
In fact, I think 'what are the real numbers?' is not as directly
analogous to 'what are the natural numbers?' as it at first seems. To
see this, compare Dedekind's project with Frege's. Frege was faced
with a firmly entrenched linguistic practice which strongly favoured
the view that number words were names and numbers were objects,
and his job was to clarify the nature of those objects. In Dedekind's
case, on the other hand, there was no pre-existing systematic use of
real numbers; that was the problem! What pre-existed in this case
was the intuitive, geometric line, and Dedekind's job was to
produce a system of numbers that would mimic its properties,
particularly continuity. So the question Dedekind faced was not
'what are the real numbers?', analogous to Frege's 'what are the
natural numbers?', but rather, 'what is continuity?'
The answer Dedekind gave was: continuity is what Dedekind
cuts have. And, as the Benacerraf-style argument points out, so do
Cantor's fundamental sequences, and other set theoretic versions of
the reals. So there is after all a single underlying property that all set
theoretic versions of the reals serve to detect, a single property
shared by all the particular disparate phenomena they are used to
measure, namely continuity. Thus, if there is a proper answer to
'what are the reals?', an answer that runs parallel to our answer to
'what are the naturals?', that is, parallel to 'that which the various
 N U M B E R S AS P R O P E R T I E S 95

set theoretic versions of the naturals serve to detect', then that

proper answer is: the real numbers are the property of continuity.
But this sounds odd, and I think the reason it does is clear from
what's already been said. The reals, unlike the naturals, are not
what needs to be accounted for. Continuity is what needs
explication, and the various set theoretic versions of the reals do
that. But our intuitive ontology is not the reals and the various set
theoretic versions, as it is the naturals and their various set theoretic
versions; rather we have the phenomenon of continuity and we
have the set theoretic reals, which all explicate that property by
exemplifying it. But there aren't any pre-theoretic reals to be
identified with anything.

Of all the many mathematical things whose relationship to sets

might be questioned, I've considered only numbers, natural and
real, because they are most intimately connected with the founda-
tions and justification of set theory itself. For set theoretic
purposes, the only other essential is the notion of function, for
which I propose a similar treatment. A function is a relation
between sets, and a relation is just a two-placed version of a
property.39 Just as von Neumann ordinals give us a way of
detecting the number property of a set—we ask whether or not the
set is equinumerous with the appropriate ordinal—the set theoretic
version of a function should give us a way of detecting whether or
not two given sets stand in the appropriate relation of argument to
value.
In fact, we do this by identifying the function with a set of
ordered pairs.40 Given two sets, x and y, we can tell if they stand in
the functional relation by asking whether, in our chosen set of
pairs, x is the first member of some ordered pair of which y is the
second member. Naturally, we could do the same job with many
other particular sets—e.g. a set of pairs of pairs (e.g. with ((0, jc),
({0}, y)) in place of (x, y)) or a closely related set of ordered triples
(e.g. with (0, y, x) in place of (x, y))—and similarly, we could use
various substitutes for the standard set theoretic version of the
39
This is an oversimplification because the mathematical notion of a function
shifted over the centuries from a rule-like relation to an arbitrary mapping. (See ch.
4.) Still, even the most general mapping can be detected by a set, as described in the
text.
40
See Enderton (1977), ch. 3.
96 NUMBERS

ordered pair. 41 All that matters, here as with numbers, is that the
set theoretic version give us a convenient way of detecting the
relation, the function, that we're interested in.

I've argued that numbers are properties of sets, that elementary

arithmetic is the study of the number properties of hereditarily
finite sets, that our knowledge of arithmetical facts is of a piece with
our knowledge of these finite sets,42 and suggested similar accounts
for real numbers and for functions. This leaves open the metaphysi-
cal question of whether or not properties (and relations) should be
included as a separate category in the set theoretic realist's
ontology. And this, obviously, is just a special case of the age-old
debate over universals. 43
Recall that in our naturalized context, this question reduces to
that of whether or not our best overall theory of the world requires
us to speak of properties in addition to ordinary physical objects
(common-sense realism), various unobservables (scientific realism),
and sets (set theoretic realism). The question is just as pressing for
physical properties as it is for mathematical ones: should physical
properties (like being gold) or physical quantities (like being an inch
long) be included, along with physical objects, in the ontology of
the natural sciences? Putnam, for one, says yes. For example, he
argues, a scientist may conjecture that 'there is a single property,
not yet discovered, which is responsible for such-and-such',44 a
statement for which Putnam sees no property-free translation.
Nominalistic philosophers of science may either doubt the central-
ity of these locutions, or disagree about the prospects for
translation. Lewis, for example, suggests that the nominalist might
treat indispensable statements about properties as asserting the
existence of what I've called natural collections or kinds. The idea
of naturalness might be taken as primitive, or it might be parsed in
terms of various objective similarities between things, without
appeal to universals.45
41
The standard version nowadays is the Kuratowski ordered pair (x, y) which is
just the set {{x}, {x, y}}, but there are many other possibilities. See Enderton
(1977), 35-8.
42
I don't mean to suggest that we can't count hereditarily infinite sets, but again,
I deny that this is part of the elementary arithmetic Frege hoped to reduce to pure
logic.
43
See ch. 1, sect. 2, above.
44
Putnam (1970), 316.
45
Lewis doesn't advocate either of these views, but he does argue that they are
live possibilities. See Lewis (1983), 347-8.
 NUMBERS AS PROPERTIES 97

Without pretending to resolve this issue, let me consider an

analogous question for our chosen branch of mathematics: should
natural numbers, as well as sets, be included in the ontology of the
theory of sets? This question can be taken in two senses. First, if
we're asking whether there is anything of mathematical significance
that can't be said without explicit reference to number properties,
then I think the answer is no. This is exactly the moral of the set
theoretic reductions: everything we wanted out of numbers can be
got out of von Neumann ordinals (or Zermelo ordinals, or . . .}. To
say that 2 < 3 is to say that if x is equinumerous with the (0, {0} }
and y is equinumerous with {0, {0}, {0, {0}}}, then x is
equinumerous with a proper subset of y. To say that 2 + 2 — 4 is to
say that if two disjoint sets x and y are equinumerous with (0,
{0} }, then their union is equinumerous with (0, {0}, (0, {0} },
{0, {0}, (0, {0}}}}- 'Every natural number has a successor' is
'if x is a von Neumann ordinal, the union of x and {x} is a von
Neumann ordinal'. '2 is prime' says 'if x is equinumerous with (0,
{0} }, then there are not two sets of cardinality less than 2 but
greater than 1 whose cross product is equinumerous with x\ And
so on. In practice, these locutions are often simplified even further,
but the fact that number theorists have no problem operating inside
set theory demonstrates that nothing mathematically important is
sacrificed by such translations. The same goes for real numbers and
functions.
But it isn't enough simply to do arithmetic and mathematics; our
overall theory of the world must also contain a chapter that tells
what we are doing and why it works the way it does. This is the
descriptive and explanatory theory of our practice required by
epistemology naturalized, just the sort of theory, in fact, that we're
now trying to construct. Within that theory, explanations of
knowledge and reference (or reliability), of puzzles like Wittgen-
stein's,46 of the notion of law-like coextensiveness, and so on,
might well appeal to objective similarities between individual
objects: two triangles are more alike than a triangle and a square;
two samples of gold are more similar than either one is to a sample
of aluminium; ordered pairs of the form (x, x + 2) are more similar
to each other than to a pair of the form (y, y + 1). It remains an
open question—for natural science just as for mathematics—
whether this distinction between natural and unnatural collections
requires a full-blown realism about universals.
46
See ch, 2, sect. 4, above.
98 NUMBERS

To summarize: for the set theoretic realist, sets have number

properties in the same sense that physical objects have length. The
further question of the ontological status of these properties is
again thoroughly analogous; the mathematical and physical sciences
are facing the same metaphysical question. The difference is that
the mathematical sciences can offer an answer to part of that
question; the set theoretic reductions of arithmetic show that all
that is mathematically important about numbers can be said using
only sets, while Putnam and others still debate the analogous
question in physical science. As for the second question, the
metascientific question of what needs to be said in our theory of our
respective sciences, the issue is common to both: how are natural
kinds, or perhaps objective similarity to be treated? Thus, I take the
problem to be a general one, not at all special to the philosophy of
mathematics, and that is why I feel justified in leaving it unresolved
here.

3. Frege numbers

The previous section on numbers as properties more or less

completes what I have to say here on my own view of numbers, but
I'm not quite ready to leave the topic entirely. So far, in evaluating
the effort to identify numbers with sets, I've concentrated on von
Neumann and Zermelo ordinals, giving little consideration to Frege
numbers: for example, the Frege number three is the extension of
the concept 'equinumerous with the series of natural numbers
ending with 2'. I'd like to pause a moment over Frege's idea, not
because I intend to modify the theory of numbers already proposed,
but because I think attention to Frege numbers will cast some
helpful light on what our theory of sets is and is not.
To focus the issue, let me reconsider Benacerraf s argument one
last time. We hear of Ernie and Johnny, who learn the von
Neumann and Zermelo ordinals, respectively. Each claims that his
sets are the natural numbers. Either version will do: there are no
conclusive arguments to decide between them. Assuming there is no
third candidate for which there are decisive arguments, the
conclusion follows that numbers are not sets.
But is there another candidate? Benacerraf mentions a Frege-style
proposal, that the number three is the set of all three-element sets. If
numbers are actually properties of sets, as I've suggested, Benacerraf
 FREGE N U M B E R S 99

fears this would count as a compelling argument in favour of this

set theoretic account over von Neumann's and Zermelo's; thus he is
motivated to reject the property view, even on slender grammatical
grounds. Only the conviction that numbers are not properties of
sets allows him to put the Frege-style option aside, and only then is
he satisfied that he need look no further: 47 'There is little need to
examine all the possibilities in detail, once the traditionally favored
one of Frege . . . has been seen not to be uniquely suitable'
(Benacerraf (1965), 284).
There is a missing premiss in this reasoning, something along the
lines that if a property is to be treated set theoretically, then it
should be identified with the set of things that exemplify it. There is
something to this; the set theorist (usually presupposing von
Neumann's numbers) speaks not of 'evenness', but of the set of
even numbers, not of 'primehood', but of the set of primes. In terms
of this methodology, Frege-style sets are the natural choice.
But, though I've embraced the property theory, I've advocated a
different line on how these properties ought to be treated within set
theory, namely, in terms of comparisons with a standard measuring
rod like the von Neumann ordinals. While this approach allows us
to say everything we want to say, it leaves us without anything in
set theory to properly call 'the numbers'.48 I think this is as it
should be. I've argued that numbers are really properties, in the
sense that the theory of numbers, arithmetic, is a theory of number
properties. If 'numbers' are 'that which forms the subject matter of
arithmetic', then numbers are properties. But if 'numbers' are 'the
referents of number words', then there are no numbers. 'One',
'two', 'three', and so on, may enjoy the superficial grammar of
names, but they are really just another measuring rod like the
Zermelo ordinals.49 We should resist the urge to find referents for

47
Notice that Benacerraf doesn't rule out a candidate for identification as the
numbers simply because it has superfluous properties; the Frege-style sets have
those. Rather he rules out candidates (with superfluous properties) for which there
are no good arguments.
48
Granted the usual conventions, the set theorist will call {0, {0} } 'two', but on
on my view, this is iike calling the standard metre in Paris 'the metre'.
49
Compare Benacerraf (1965), 292: 'Questions of identification of the referents
of number words should be dismissed as misguided in just the way that a question
about the referents of the parts of a ruler would be seen as misguided.' Despite
Benacerraf's rejection of the property view, there are many points of contact
between the position advocated here and that sketched in the final pages of his
paper. I won't try to sort out our agreements and disagreements.
100 NUMBERS

these words, and the temptation to treat number properties as sets

is just one version of this misguided impulse.
Still, Frege numbers are of interest, even if it is wrong to say that
they are the natural numbers. 50 The extension of 'equinumerous
with . . .' is a collection; set theory is our theory of collections; it's
natural to ask how this particular collection fits in. Asking this
question will tell us something about the nature of sets.

One of the strongest arguments against the claim that Benacerraf's

Frege-style sets are a suitable set theoretic version of the Frege
numbers is that they don't exist.^ We've seen that this is true
because, for example, new three-element sets are formed at every
stage of the iterative construction, so there is no stage at which the
set of all three-element sets is formed. This collection is too big to
be a set; in standard terminology, it is a 'proper class'. Other
collections, like the collection of all sets, are also proper classes.
Thus the common wisdom is that the set of all three-membered sets
would be a good candidate for the role of a Frege number were it
not for the unfortunate fact that it is too big to exist. But 1 think this
underestimates the drawbacks of the set of all three-membered sets.
To see this, consider: ordinary arithmetic is used to count
physical objects, perhaps sets of these (like the two pairs of shoes),
perhaps occasionally even sets of sets of these, and so on, but never
sets of more than finite rank. This is the content of my insistence
that arithmetic is part of the theory of hereditarily finite sets.
Eventually, of course, we go beyond ordinary arithmetic; we try to
number things like the sets of equinumerous sets of real numbers,
and then we are doing higher set theory. 52 Just for the moment, I'd
like to suggest that this expansion of our arithmetical theory is non-
trivial, that—just as the move from ordinary distances to astro-
nomical ones requires a change in our notion of space from
Euclidean to non-Euclidean—when we demand that our numbers
count more complicated, infinitary things, we are asking for more
complicated numbers."13 On this picture, there are Frege-style sets
50
Readers of Maddy (1981) and (1983) will recognize a change of heart here. I
have Michael Resnik to thank for the insight that the very complexity of the theory
of the 1983 article suggests it is not a theory of ordinary numbers,
51
See Benacerraf (1965), 284.
52
Cantor's continuum hypothesis, to he discussed in the next chapter, says that
there are exactly two sets of equinumerous infinite sets of reals.
' • Notice, these new numbers are not more complicated in that they are infinite—
I'm still talking about finite numbers—they are just more complicated in that the
finite sets they number can have infinite sets in their transitive closures.
 FREGE N U M B E R S 101

that correspond to the Frege numbers of ordinary arithmetic. The

trouble arises only when we try to expand our arithmetic to count
objects of higher set theory, but perhaps the trouble is due to the
vagaries of higher set theory and not to the simpler numbers of
ordinary arithmetic, with which we began.
Now suppose this separation of infinitary versus non-mfinitary
arithmetic can be maintained. If the only problem with Frege-style
sets as surrogates for Frege numbers were their size, we could then
identify the 'small Frege numbers', the Frege numbers of ordinary
arithmetic, with the corresponding Frege-style sets; that is, the
small Frege number three would be the set of all three-membered
hereditarily finite sets. What I want to claim is that even here the
Frege-style sets are unsuitable.
The small Frege numbers count physical objects (and sets of
these, etc.), and presumably the physical objects in our world
fluctuate: new animals are born; old stars explode; trees are
converted into tables and chairs. Correlated with these shifts in the
population of physical objects are shifts in the population of sets:
when the cub is born, it is a member of sets that didn't exist before;
when the old star explodes, sets also vanish; sets of trees give way
to new sets of tables and chairs. Finally, these fluctuations in the
population of sets bring about fluctuations in the population of
three-membered, hereditarily finite sets, so the set of all these, our
Frege-style candidate for the small Frege number three, is now this
set, now that. This is not to say that the Frege-style set changes its
membership from time to time—sets are fully determined by their
members, so they can't do that—rather, the proposal that the small
Frege number three be identified with the Frege-style set of all three-
element, hereditarily finite sets is the proposal that it be identified
with one set after another, with different sets at different times.54
This is not satisfactory. If we really thought Frege numbers were
the numbers, we might well insist that whatever three is, it is the
same thing today as it was yesterday; if we thought the number
words had referents, they should have the same referents today as
tomorrow. But even if we reject (as I have) the theses that numbers
are objects and that number words have referents, I think we
should still object to this account of the Frege numbers themselves.
When we asked after the collection of all three-membered sets, we
didn't want now this collection, now that, depending on the
54
This argument is an adaptation of Hambourger's (1977), with temporal
considerations replacing modal ones.
102 NUMBERS

vagaries of physical existence. What we had in mind was a

collection whose membership is allowed to change from time to
time while it stays the same, that is, the collection which collects, at
any given time, the three-element sets in existence at that time. Such
a collection, even if it is small enough, is not a set.
Frege numbers, then, are collections that are not sets. Large Frege
numbers are too large to be sets, so they are proper classes. Small
Frege numbers, though small enough to be sets, are individuated
differently, so they are classes, too, without being proper. Those
inclined to insist on referents for the number words will need a
theory of sets and classes. Even those immune to this temptation
will find the distinction between these two types of collections
important to the clarity of their theory of sets. And there remains
the possibility that classes might do important work in set theory
itself.55 Let me conclude this section, then, with a brief look at the
distinction between sets and classes.

At the beginning of this chapter, I sketched two foundational

worries that contributed to the development of the modern theory
of sets. The historical path from the troubles with the calculus,
through Cantor and Dedekind's work on the real numbers, to
Zermelo's axiomatization and beyond, has been a mainly math-
ematical development whose vicissitudes eventually led to a well-
articulated picture of collections formed sequentially, from previously
given elements, in a hierarchy of stages: the iterative conception of
set. At each stage, new collections are formed with complete
freedom, without concern for any method of construction. Finite
combinatorics tell us that there is a unique subcollection of a finite
collection for every way of saying yes or no to each individual
element. Carrying this notion into the infinite, subcollections are
'combinatorially' determined, one for every possible way of
selecting elements, regardless of whether there is a specifiable rule
for these selections. This is the mathematical notion of a collection:
a collection formed combinatorially, in a series of stages that make
up the iterative hierarchy. 56
55
For example, in non-demonstrative arguments for new set theoretic hypo-
theses. See Maddy (1988#) for some examples.
56
This picture appears in Zermelo (1930). The combinatorial idea is explicit in
Bernays (1935). As indicated above, Enderton (1977), 7-9, Boolos (1971), and
Shoenfield (1977) give modern presentations. Of course, the temporal and
constructive imagery is only metaphorical; sets are understood as objective entities,
existing in their own right.
 FREGE N U M B E R S 103

The second historical trail is the one that begins with Frege's
logicism and continues through Russell and Whitehead's effort to
salvage some aspects of that idea.57 Here the original goal was to
found arithmetic, and the collections involved are extensions of things
more or less like predicates.58 This is a very different conception: the
entire universe is simply divided into two piles, depending on whether
or not each thing satisfies the given predicate. This contrasts twice
with the iterative picture: we are free to collect absolutely any things,
regardless of whether they are all available at some stage, but we are
not free to collect combinatorially, without recourse to a rule of any
kind. The Cantorian, mathematical, collection is a set; the Fregean,
logical, collection is a class.59
Collections of these two types differ in several other ways, one of
which we've already noted: classes can be larger than sets. Thus, for
example, there is a class of all sets—the extension of the predicate
'is a set'—but it is not a set, because there is no stage at which it can
be formed. Such a class is a proper class, because it is too large to be
coextensive with any set. On the other hand, because classes can
only be formed when there are suitable predicates, there may well
be more sets of real numbers than there are classes of reals. This
depends, of course, on the details of our theory of predicates, but it
is hard to imagine how the existence of a predicate for every set at
each stage could be guaranteed without somehow presupposing the
combinatorial notion of set theory. 60
We've also seen that a class can change its membership, while a
set cannot. The collection of things with a certain property can vary
in membership from time to time, but as long as it is identified as
the extension of the appropriate predicate, it remains the same
class. A set, on the other hand, is completely identified by its
membership, so the collection of things with a certain property is
now one set, now another.
57
In fact, Russell and Whitehead's position (1913) is a hybrid between the two
notions considered here: extensions, which depend on predicates, are formed in
stages. A similar amalgam is proposed by Keith Devlin in support of the axiom
candidate V = L (see ch. 4, sect. 4, below). Godel (1944), 464, traces the ideas behind
this axiom to Russell.
58
Here I'm including Frege's concepts, Russell's prepositional functions, my
scientific properties, etc, I won't distinguish between these in what follows.
59
I discuss this contrast in more detail, with more historical considerations, in
Maddy (1983). See also Parsons (1974<z; 1977), and Martin, 'Sets versus classes'
(unpublished).
60
Notice that this counts against efforts, remarked on earlier, to mimic set theory
in property theory.
104 NUMBERS

A novel and striking difference is one that shows up in the

structure of the membership relation. A class can be a member of
another class—the class of von Neumann ordinals is a member of
the class of infinite collections—just as sets are members of one
another. But the class of von Neumann ordinals isn't the only
member of the class of infinite collections—there are infinitely
many infinite collections—so the class of infinite collections has the
distinction of being self-membered. Of course, no set can be self-
membered, because all its members must be formed at stages before
that at which it is formed.
And finally, classes, unlike sets, lead to paradox. We've seen that
some collections are self-membered—e.g. the class of all infinite
collections—while others are not—e.g. any set. 'Being non-self-
membered' seems a perfectly unobjectionable predicate, satisfied by
some collections, not satisfied by others, so there ought to be a class
that is its extension. But this is Russell's paradox in much the same
form as he first presented it to Frege:61 is the class of non-self-
membered collections self-membered or not? And again, on the
iterative conception, all sets are non-self-membered, so the Russellian
set, the set of all non-self-membered sets, is the set of all sets. But
there is no such set, because there is no stage at which it could be
formed, and thus, no paradox.
Most contemporary set theory is done without explicit mention
of classes. Theories that do attempt to encompass both sorts of
collections generally separate sets from proper classes on the basis
of size, without taking into account the fundamental differences
between sets and small classes. Sometimes classes are not allowed
as members of any further collections, which avoids paradox, but
seems restrictive and artificial; 62 other times they are allowed
membership in further collections, but these collections are formed
in stages, disallowing self-membership, and are determined com-
61
Again, see his letter to Frege, Russell (1902).
62
One system like this is von Neumann—Bernays—Godel (see von Neumann
(1925), Bernays (1937), Godel (1940)), in which there is a class for every first-order
formula with quantifiers ranging only over sets. This allows the familiar Zermelo-
Fraenkel axioms to be condensed into a finite list, but has little effect beyond this
metamathematical simplification. Morse—Kelley (see Kelley (1955) and Morse
(1965)) is a stronger system, allowing quantification over classes in formulas that
determine classes, but it still disallows classes as members and has the added
difficulty that its classes look like little more than another layer of sets that was
somehow left out of the hierarchy. See Drake (1974), 16-17, Fraenkel, Bar-Hillel,
and Levy (1973), ch. 2, §7.
 FREGE NUMBERS 105

binatorially, which raises serious questions about how these

additional layers of non-set collections really differ from sets.63
A parallel situation arises in the theory of truth, 64 with sentences
like 'Everything I've ever said is false.' If everything else I've ever
said is false, then this statement is paradoxical. Russell gives us a
predicate that can't have an extension; here we have a sentence that
can't have a truth value. Parallel to the class theoretic solution that
disallows classes as members, we could insist that the question of
truth not be raised for sentences involving the notion of truth; this
escapes paradox, but again it is too restrictive and artificial.
Another solution, parallel to regimenting classes into stages,
requires that the notion of truth be typed: truthx applies only to
statements that don't involve the notion of truth at all; truth 2 only
applies to statements involving truth 1? and so on. Thus 'false* in
'Everything I've ever said is false' is not-true« for some «, and
'everything' only ranges over statements involving truth^-x. In both
cases—class theory and truth theory—the diagnosis is that we can't
survey some entire category—all classes, all statements—but only
one level or type at a time, and in both cases, the restrictions square
poorly with the pre-theoretic notions.
Thus the paradoxes of truth and class theory lead to unpalatable
hierarchies in which statements can't refer to themselves and classes
can't be self-membered. An alternative available in both cases is to
allow gaps, that is, statements that are neither true nor false, classes
of which some items are neither members nor non-members.
Kripke showed how such a system would go for truth theory,65 and
I have proposed an analogous theory of classes.66
To a certain extent, this approach produces the desired results:
the class of infinite collections is self-membered; the class of non-
self-membered collections is neither self-membered nor non-self-
membered. Frege numbers, large and small, can be defined in
various ways, and the small ones at least are fairly well behaved.
Still, the context of three-valued logic—true, false, and neither—
63
For example, Ackermann (1956) seems to add several layers of combinatori-
ally determined 'classes'. (See Fraenkel, Bar-Hille!, and Levy (1973), ch. 2, §7.7.)
Reinhardt (1974), in a descendant of Ackermann's system, reinforces this picture by
explicitly assuming the axiom of foundation for classes. (Foundation disallows self-
membership, among other class-like pathologies. See Enderton (1977), 206.)
64
For further discussion of the parallel, see Parsons (19741?).
65
See Kripke (1975).
66
InMaddy{1983).
106 NUMBERS

makes the system awkward, so I think the jury must still be

considered out on the question of a workable theory of classes in
general and of Frege numbers in particular. 67

67
Maddy (1984c) contains some modifications and developments of the system,
as well as answers to one of the open questions of Maddy (1983): a fixed point
theorem of William Tait implies that the construction does not reach a fixed point.
Robert Flagg has since demonstrated what Tait conjectured, namely, that if the
construction is carried out over a standard model of ZF, then the fixed point will be
the first admissible ordinal greater than the least upper bound of the ordinals of the
ground model. Maddy (1984c) also contains some sobering information on the
prospects for axiomatization. See Feferman (1984a) for a compendium of related
systems.
 4
AXIOMS

1. Reals and sets of reals

The epistemology of compromise Platonism follows Godel's in

being two-tiered: the most primitive truths are intuitively given,
obvious; the more theoretical hypotheses are justified extrinsically,
by their consequences, by their ability to systematize and explain
lower-level theory, and so on. In Chapter 2, I sketched the set
theoretic realist's version of intuition, a neurologically based
phenomenon that produces firmly held elementary beliefs and
provides them some preliminary level of justification. The re-
mainder of our evidence for the principles we choose as axioms
must come from theoretical sources; the time has come to look into
the structure of this second type of mathematical justification and to
confront the questions it raises.
Because extrinsic arguments involve more advanced levels of
mathematical theorizing, they bring us face to face with more
esoteric set theoretic matters than have heretofore been relevant. As
this is obviously unavoidable, I beg the indulgence of my non-
mathematical reader. Indeed, I hope she might come away with a
greater appreciation for why Hilbert and others refuse to be budged
from Cantor's paradise.
The story begins with the mathematical concerns that led Cantor
there in the first place.1 I'll turn to axiomatics in the next section.

Cantor's dissertation was in number theory, but when he arrived at

his first university teaching job in 1869, one of his senior colleagues
posed him a problem in analysis: 2 consider functions from reals to
reals that can be represented by infinite trigonometric series; are
1
In ch. 3, sect. 1, above, I put greater emphasis on Dedekind's and Frege's
foundationai motivations. Cantor's, as we'll see, were more strictly mathematical.
2
In mathematics, 'analysis' means the study of real and complex functions,
which includes the calculus and its foundations and extensions. In this discussion of
Cantor's career and its antecedents, I follow Dauben (1979).
108 AXIOMS

such representations unique? Eduard Heine, Cantor's colleague,

had proved uniqueness under certain special circumstances, but a
general solution eluded him, just as it had his fellow analysts
Dirichlet, Lipschitz, and Riemann. Within months, Cantor had
obtained the desired result: if a function f(x) is given by a
trigonometric series that converges for every value of x, then that
representation is unique.
But Cantor didn't stop there. Typical of related work in the
theory of functions was the generalization of such theorems by
allowing a number of 'exceptional points', for example points at
which the series is not required to converge. Heine's partial results
had allowed finitely many exceptions, but Cantor, inspired by the
work of Hankel, hoped to accommodate infinitely many. The
method depended crucially on the distribution of the exceptional
points. Everyone knew that infinitely many points contained in a
bounded interval will accumulate around at least one point.3
Cantor could see how to deal with infinitely many exceptional
points if they had exactly one accumulation point; with a little
effort, he could see how to allow for finitely many accumulation
points; indeed, his method would still work on an infinite set of
accumulation points as long as that set had only one accumulation
point of its own, or for that matter, finitely many accumulation
points of its own; and so on. The outlines of the generalization he
had in mind were there, but Cantor needed a way to formulate his
most lenient condition on exceptional points with precision.
He quickly realized there was no hope of defining such a
complicated set of points without an accurate theory of the real
numbers themselves; it was this problem that led him to his account
of reals in terms of fundamental sequences. He was then able to
formulate the requirement on exceptional points appropriate for
the generalization of his uniqueness theorem. But what an odd set
of points it was: infinite, and quite complex, yet still somehow
small enough, or well-behaved enough, in relationship to all the
reals, to do no damage! This apparently got Cantor to wondering
how continuous sets like the reals relate to seemingly smaller,
discrete infinite sets like the natural numbers.
Meanwhile, their shared interest in the theory of real numbers
had brought Dedekind and Cantor into correspondence. In a letter
3
An accumulation point of a set of points has points of that set arbitrarily close
to it. for example, 1 is an accumulation point of • '/z, 2/% -V^ 4/5, . . . ;.
 REALS AND SETS OF REALS 109

to his friend, Cantor raised a nagging question: can the real

numbers be brought into one-to-one correspondence with the
naturals? Admittedly, there seem to be more reals than naturals,
but then, there seem to be more rationals that naturals, too, and
Cantor had shown that there were not.4 He wrote:5
At first glance one might say no, it is not possible, for [the set of natural
numbers] consists of discrete parts while [the set of real numbers] builds a
continuum; but nothing is won by this objection, and as much as I am
inclined to the opinion that [the set of naturals] and [the set of reals] permit
no such unique correspondence, I cannot find the reason, and while I attach
great importance to it, the reason may be a very simple one.

Dedekind had no easy answer. Cantor replied:

I raised the question because I have considered it for a number of years and
have always found myself doubting whether the difficulty it gave me was
subjective or whether it was due to the subject itself. Since you write that
you are also in no position to answer it, I may assume the latter.

Shortly thereafter, Cantor found his theorem: the correspondence is

impossible; there are, in this sense, more real numbers than there
are naturals. 6
If there are so very many points on the line, how many might
there be in a plane, in a three-, or four-, or w-dimensional space?
This next question to Dedekind took longer to answer, perhaps
because Cantor so firmly expected spaces of higher dimension to
have more points. When he finally produced a one-to-one corres-
pondence between the line and the plane, he was moved to remark:
'I see it, but I don't believe it!'7 But there it was. Every infinite set he
had considered so far had either the cardinality of the reals or the
cardinality of the naturals.
This discovery served to refocus attention right where it had
started: on sets of reals. Cantor wrote:8
And now that we have proved, for a very rich and extensive field of [sets],
the property of being capable of correspondence with the points of a
4
See Enderton (1977), 130.
5
This quotation and the next are drawn from Dauben (1979), 49-50.
6
See Enderton (1977), 132. A set is 'countable' if it is no larger than the
naturals; otherwise it is 'uncountable'.
7
Dauben (1979), 55.
8
This and the next quotation from Cantor (1878) are translated by Jourdain in
the introduction to his edition of Cantor (1895/7), 45.
110 AXIOMS

continuous straight line . . . the question arises . . . Into how many and what
classes (if we say that [sets] of the same or different [size] are grouped in the
same or different classes respectively) do [infinite sets of reals] fall?
Cantor had an opinion:
By a process of induction, into the further description of which we will not
enter here, we are led to the theorem that the number of classes is two . . .
This conjecture, that every infinite set of reals is either countable or of
the cardinality of the continuum, has come to be called Cantor's
'continuum hypothesis' (CH).
Cantor's best effort in the direction of a proof of this conjecture
involved the notion of a perfect set. A closed set of reals is one that
contains all its accumulation points; a perfect set is closed, and
every one of its points is an accumulation point.9 Perfection played
a key role in Cantor's analysis of continuity itself, and eventually he
was able to show that every non-empty perfect set is equinumerous
with the continuum. After a false start pointed out by Ivar
Bendixson, Cantor proved that every closed set of reals can be
decomposed into a countable set and a {possibly empty) perfect set,
which implies that the continuum hypothesis is true for closed sets.
(From now on, I'll assume that perfect sets are non-empty by
stipulation.) At this point, Cantor was optimistic about the
possibility of generalizing this result, called 'the Cantor—Bendixson
theorem':10 'In future paragraphs it will be proven that this remark-
able theorem has a further validity even for [sets of reals] which are
not closed . . .'. Of course, a complete generalization would constitute
a proof of Cantor's continuum hypothesis in its entirety.

Characterizing sets of exceptions or singularities wasn't the only

problem that brought analysts of this period up against questions
about sets of real numbers. From the eighteenth century on, a wide
range of considerations—from vibrating strings and heat flow to
the foundations of the calculus—consistently pushed mathema-
ticians from narrower to more inclusive notions of function.1
9
For example, the set consisting of the points between 0 and 1 inclusive is
perfect. The set consisting of those points plus the single point 2 is closed but not
perfect.
10
Cantor (1883), 244, translated by Dauben (1979), 118. For other discussions
of the history of the Cantor—Bendixson result, see Moore (1982), 34-5, and Hallett
(1984), 90-2.
11
This development is traced by Kline (1972), 335-40, 403-6, 505-7, 677-9,
949-54.
 REALS AND SETS OF REALS 111

What began in Galileo's time with curves and continuous motions

developed to Euler's combinations of parts of different curves, then
to a series of accounts in terms of ever-widening class of
expressions that could legitimately define a function, until math-
ematicians were finally faced with the idea of a purely arbitrary
function as absolutely any correspondence between reals and reals
regardless of how it might or might not be expressed by
mathematical operations. Odd and seemingly pathological ex-
amples proliferated: functions that pass through every value
between a and b without being continuous,12 continuous functions
that aren't different} able,13 and even Dirichlet's 'shotgun' function
(zero on rationals, one on irrationals) which was nowhere
continuous, without either derivative or integral.14
Doubts circulated about the soundness of this extremely general
notion, and by 1900 there was considerable controversy about the
proper extent of the function concept.15 In an effort to get a
responsible handle on the vast range of non-continuous functions,
the French analysts Rene Baire, Emile Borel, and Henri Lebesgue
set out to give a systematic classification, Lebesgue's version made
use of Borel's earlier hierarchy of sets of reals. The simplest sets of
reals are the closed sets, mentioned earlier, and their complements,
the open sets.16 The union of two closed sets is closed, but the
union of countably many closed sets may be open17 or worse.18
This 'or worse' gives rise to the Borel hierarchy:19
S? = the open sets
111 — the closed sets
^2 = countable unions of closed sets
12
Due to Darboux. See Kline (1972), 952.
13
By Riemann, Cellerier, and Weierstrass between 1854 and 1875. See Kline
(1972), 955-6.
14
In 1829. See Kline (1972), 950.
15
See Monna( 1972).
16
Equivalently, a set is open if it contains on open interval ({x a < x < b } )
around each of its points.
17
For example, the union of the closed intervals [l/n, («—!)/«] is the open
interval (0, 1). (By notational convention, square brackets indicate that the
endpoints are included, round brackets that they are not.)
For example, the union of the closed intervals [0, (»—!}/«] is neither closed nor
open.
19
The Borel sets were introduced by Borel (1898) as sets resulting from the
closed sets using complement and countable intersection. Lebesgue (1905) intro-
duced a hierarchy, though not this one, which is due to Hausdorff (1919). It
continues to generate new sets of reals until the first uncountable ordinal. (For limit
ordinals \, ^ = countable unions of sets in the FI^s for a a < \.) For the basic theory
of Borel sets, see Kuratowski (1966), §§5, 6, and 30, or Moschovakis (1980), ch. 1.
112 AXIOMS

11" = complements of S? sets

.
.
.
2a + i — countable unions of Da sets
Ila + i = complements of 2« + i sets
A« = sets that are both S" and 11"
Borel sets = the union of the £^s
Lebesgue's hierarchy of 'Borel functions 1 was defined in terms of
these simple sets of reals, and he proved it equivalent to another
hierarchy given earlier by Baire,20
The Borel sets, despite their complexity, turn out to be fairly well
behaved. Consider, for example, the perfect subset property: a
collection of sets of reals is said to have the perfect subset property
if every infinite set in the collection is either countable or contains a
perfect subset. 21 Thus the Cantor-Bendixson theorem says that the
closed sets of reals have the perfect subset property. Paul
Alexandroff (1916) made good on Cantor's hunch that this result
could be generalized by extending it to include all Borel sets.
Another property of interest to analysts is separability: two sets A
and B are separated by a set C if A is a subset of C and B is disjoint
from C. Waclaw Sierpiriski (1924) proved that disjoint Borel sets in
n° can be separated by a Borel set in A°, and gave applications of
this fact to the general theory of functions.
Lebesgue was also instrumental in the isolation of a second
important collection of sets of reals. The impetus this time came
from the theory of integration, another field filled with perplexities
at the time. Among the numerous pathological functions under
scrutiny were seemingly unobjectionable examples that turned out
not to be integrable using the state-of-the-art Riemann integral. To
extend the concept of integration, Lebesgue needed a new gauge of
the size of a set of reals, not a generalization of number, like
Cantor's cardinality, but a generalization of length that could take
that concept from intervals to more complex sets. For this purpose,

20
To see the connection, notice that a function f is continuous if it has the
following property: whenever A is an open set, f~l[K\ = (x f(x) € A} is also open.
Lebesgue defined a function g to be S" iff g~ l[A] is £„ whenever A is S°. Thus the S?
functions are the continuous ones. Baire's version of this hierarchy appears in Baire
(1899).
21
Recall I'm assuming perfect sets are non-empty.
 REALS AND SETS OF REALS 113

he developed the notion of Lebesgue measure,22 and even

Dirichlet's function became integrable. Since closed sets are
Lebesgue measurable, and the collection of Lebesgue measurable
sets is closed under complement and countable union, it follows
that Borel sets have another nice property: they are all Lebesgue
measurable.
The next step in this development, the last one I'll touch on here,
was also precipitated by Lebesgue, but this time by an uncharacter-
istic error. One of his analyses of Baire functions included a 'trivial'
lemma that the projection of a Borel set is Borel.23 This isn't true,
but the slip went unobserved for a decade, until Mikhail Suslin, a
young student in Moscow, burst into his professor's office with the
news. Together, student and professor, Nikolai Luzin, established
the elementary properties of these 'analytic' sets,24 which led, some
years later, to the introduction of a new hierarchy of sets of reals:25
SQ = the open sets
ITo = the complements of So sets
S} = the projections of HQ sets
II} = the complements of S} sets
.
.
.
S^+i = the projections of II* sets
LI*
+ 1 = the complements of S^+i sets
.
.
.
A* = sets that are both S^ and II*
projective sets = the union of the S^
The relationship between Borel and projective sets was cinched by
Suslin, who showed that the Borel sets are the A} sets.
Despite their added complexity, projective sets inherit some of

22
In Lebesgue (1902). For an elementary exposition, see Williamson (1962).
23
Instead of sets of points on the line, think of sets of points in the plane;
Borelness and so on can be defined for these sets just as easily. Then the projection of
a Borel set in the plane is, so to speak, the shadow it casts on the x-axis.
24
See Suslin (1917) and Luzin (1917).
25
Introduced by Luzin (1925) and Sierpiriski (1925). For the classical theory of
projective sets, see Kuratowski (1966), §§38—9, or Moschovakis (1980), §§!E,
ch. 2, and parts of ch. 3.
114 AXIOMS

the regularity properties of Borel sets. Lebesgue measurability and

the perfect subset property for analytic or £j sets were immediately
established by Luzin and Suslin and separability came a few years
later.26 Separability was extended to Hi by P. Novikov in his
(1935), but after that, the best efforts stalled: separability remained
open past Ilj, measurability beyond S}and H\, and for the perfect
subset property, as Luzin remarked: 27 There remains here only one
important gap: we do not know if every uncountable complement
of an analytic set [i.e. every uncountable ll\ set] has the
[cardinality] of the continuum.' There the matter stood for some
years.

Thus, for all the thoroughly admirable foundational goals pursued

by Frege and Dedekind, the deepest contributions to the modern
mathematical theory of sets, those of Georg Cantor, were inspired
almost exclusively by mathematical concerns, particularly concerns
arising from analysis. I have tried in this section to give a hint of
how Cantorian set theory grew out of the study of real functions
and to sketch in the sorts of questions that arose naturally in that
context. Let me turn now to how and why this naive mathematical
theory came to be axiomatized.

2. Axiomatization

To understand the impulse that led to the axiomatization of set

theory, we must return to Cantor and his continuum problem.
Around the time he was making his optimistic prediction, quoted
earlier, about generalizing the Cantor-Bendixson theorem, another
letter to Dedekind contains the first hint of what was to become a
corner-stone of his theory of infinite numbers, namely, the concept
of well-ordering.28
Again the catalyst was his earlier work on sets of singular points.
In order to describe the process of taking accumulation points of
sets of accumulation points of sets of accumulation points, etc., he
had proceeded as follows from a given point set A:
26
In Luzin (1927).
27
Luzin (1925), as translated by Hallett (1984), 108.
28
See Moore (1982), 40-1. The letter in question is dated 5 Nov. 1882.
 AXIOMATIZATION 115

AO = A
A! = {x x is an accumulation point of A0}
.
.
.
A n+1 = {x | x is an accumulation point of An}
.
.
.
A^ — the intersection of the AMs
AUJ+I = {x x is an accumulation point of Aw}
and so on. Now, in the early 1880s, his interest shifted from the
derived sets themselves to the subscripts. Why shouldn't there be
to + co after all the o> + m, and (to + co) +1 after that? Here is a
sequence that carries into the transfinite, and after any batch of
entries, there is always a next.29
Considered in terms of cardinality, this transfinite sequence of
ordinals yields a wonderful bonus. The collection of all countable
ordinals is not itself countable. In fact, as Cantor had shown by the
late 1890s,30 the cardinality of this set is the very next infinite
cardinality after that of the natural numbers. Thus, he called the
latter K 0 and the former Kj. And the set of ordinals of cardinality K!
has cardinality K 2 , and so on, so the infinite sequence of ordinals
yields an infinite sequence of cardinal numbers as well. Cantor
extended the arithmetic operations—plus, times, exponentiation—
from the finite numbers into the infinite, and showed that the
cardinality of the continuum is 2 . By this point, then, the
continuum hypothesis had become: 2 ° = Kj. But, for all Cantor's
efforts, it remained unproved.
Then, in 1904, at the Third International Congress in Heidel-
berg, came a shock; Julius Konig read a paper that purportedly
showed the continuum hypothesis to be false. In particular, Konig
argued that the continuum could not be well-ordered at all, and
ipso facto, that it could not be put in one-to-one correspondence
29
Technically, a well-ordering of a set A is an ordering in which every non-empty
subset of A has a least member. This produces the idea in the text: as we run through
the elements of A, at any point there is a least member of the elements we haven't
listed yet. See Enderton (1977), 172-3.
30
See his last major work, Cantor (1895/7).
116 AXIOMS

with the well-ordering X j . A contemporary report describes the

scene this way: 31
For Cantor to claim that every set can be well-ordered and, in particular,
that the continuum has the second [infinite] cardinality was a kind of
dogma that was part and parcel of what he knew and believed in set theory.
Consequently Konig's address, which culminated in the proposition that
the continuum could not be an aleph (hence could not be well-ordered
either), had a stunning effect, especially since its presentation was
extremely elaborate and precise.

Apparently Cantor was less upset by the proof itself, of which he

was sceptical, than he was by what he saw as his public humiliation
before his colleagues.32 His scepticism, at least, was well taken; one
of Konig's assumptions, a 'theorem' of Felix Bernstein, turned out
to be incorrect. This was pointed out to the members of the
Congress, on the day directly after the presentation of Konig's
proof, by Ernst Zermelo.
But a doubt remained. As early as 1883, just after the above-
described tetter to Dedekind in which the notion is introduced,
Cantor wrote that: 33
The concept of a well-ordered set turns out to be essential to the entire
theory of point-sets. It is always possible to bring any well-defined set into
the form of a well-ordered set ... this law of thought appears to me to be
fundamental, rich in consequences, and particularly marvelous for its
general validity . . .
By 1895, he realized the need for a proof of this fundamental
principle, and set out to find one. His letter to Dedekind of 1899
contains one attempt.34 Now, though Konig's attack had been
unsuccessful, it brought home the possibility that not only the
continuum hypothesis, but the well-ordering principle itself might
one day be overthrown.
31
Schoenflies, as quoted in van Heijenoorr (ed.) (1967), 192.
32
See Dauben (1979), 247-50, 283, for discussion of this episode.
33
Cantor (1883), as translated by Moore (1982), 42. For an analysis of Cantor's
thinking on well-ordering, see Hallett (1984), § 3.5.
34
In fact, Cantor's argument is very close to the one usually given today (see e.g.
Drake (1974), 56); it goes through with minor modifications once Zermelo's axiom
has been identified. Apparently it was the (unnecessary) use of proper classes—his
'inconsistent multiplicities'—that disturbed Cantor. In any case, neither he nor
others (including Hilbert) who saw the argument at the time were convinced by it. In
1903, when it was rediscovered by Jourdain, Cantor refused have his version
published. See Moore (1982), §§1.6 and 2.1 for details.
 AXIOMATIZATION 117

Before 1904 was over, a proof was presented, not by Cantor, but
again by Zermelo,35 The surprisingly short argument depends on a
novel assumption:
. . . that even for an infinite totality of sets there are always mappings that
associate with every set one of its elements . . .
By way of defending this principle of choice—the function
'chooses' one element from each set—Zermelo admits that
This logical principle cannot, to be sure, be reduced to a still simpler
one.. .
but in its favour:
it is applied without hesitation everywhere in mathematical deduction . . .
As an illustration, he cites the seemingly obvious fact that a set can't
be divided into more non-empty disjoint parts than it has members,
a fact whose proof also depends on choice.
In this unassuming manner, Zermelo proposes that mathemat-
icians accept a principle that is simple, that was so obvious to
previous theorists that they used it unconsciously, and that is
necessary to prove various important and natural results (like the
well-ordering principle, the theorem that all infinite cardinalities
are alephs, and the partition principle just noted). He could hardly
have been prepared for the storm of controversy that ensued.
Gregory Moore's extraordinary history of this period traces the
complex reaction through its independent manifestations in France,
Germany, Hungary, England, Italy, Holland, and the United
States.36 Choice and a welter of other set theoretic principles,
including well-ordering and unlimited comprehension, were sud-
denly up for grabs, embraced here, denied there. It was in defence
of his principle of choice and his proof of well-ordering that
Zermelo was driven to axiomatize the practice of set theory.

Zermelo's defence of the principle of choice and his axiomatization

of set theory appear in two papers written within days of one

35
Zermelo (1904). This short paper began as a letter to Hilbert one month after
the Congress and was published later that year. All the quotations in this paragraph
come from p. 141.
36
Moore (1982), ch, 2. My brief account of the controversy over choice draws
heavily on Moore's work.
118 AXIOMS

another in the summer of 1907.37 The first of these contains a

spirited response to critics of what was now his axiom of choice. He
admits again, as he had in 1904, that the principle has not been
proved, but points out: 'even in mathematics unprovability . . . is in
no way equivalent to nonvalidity, since, after all, not everything can
be proved, but every proof in turn presupposes unproved prin-
ciples' (Zermelo (1908#), 187). Thus, even his opponents must rely
on unproved axioms. How, then, are these axioms justified?—'by
pointing out that the principles are intuitively evident and necessary
for science. . .' (Zermelo (1908#), 187). From our set theoretic
realist's perspective on mathematical evidence, Zermelo is recogniz-
ing both intrinsic supports-—in terms of 'intuitive evidence'—and
extrinsic supports—in terms of the role of the axiom in overall
scientific theorizing. He proposes to apply these criteria to choice.
To confirm the intuitiveness of his axiom, Zermelo cites
historical evidence:
That this axiom, even though it was never formulated in textbook style,
has frequently been used, and successfully at that, in the most diverse fields
of mathematics . . . is an indisputable fact, . . Such an extensive use of a
principle can be explained only by its self-evidence . . . (Zermelo (1908a), 187)

Cantor, Dedekind, and the other early set theorists had passed over
numerous uses of choice in various forms without comment,
usually without noticing it themselves.38 Unconscious applications
of the principle can also be found in analysis, particularly in the
work of Baire, Borel, and Lebesgue touched on in the previous
section.39 This surely constitutes some evidence for its obviousness,
and hence, for its intuitiveness, though the initial protest against
Zermelo's linguistic formulation remains to be explained (see
p. 123 below). Zermelo anticipates the objection that evidence for
intuitiveness should not be counted as evidence for truth:
No matter if this self-evidence is to a certain degree subjective—it is surely
a necessary source of mathematical principles . . . and Peano's 40 assertion
that it has nothing to do with mathematics fails to do justice to manifest
facts. (Zermelo (1908a), 187)
37
These are Zermelo (I908d) and (19086).
38
For examples, see Moore (1982), § 1.4.
39
See Moore (1982), §§ 1.7 and 4.1. For the role of the axiom in the work of
Suslin and Luzin, see §§ 3.6 and 4.1.
40
Peano was among the critics of Zermelo's axiom. See Moore (1982), §2.8.
 AXIOMATIZATION 119

Here he appeals to the undeniable fact that intuitiveness is often

taken, in practice, as evidence for truth. This isn't enough (as
observed in the third section of Chapter 2), but the additional
considerations cited there can be brought to bear.
Turning to extrinsic supports, Zermelo gives his version of the
role of set theory in our overall theory:
Set theory is that branch of mathematics whose task is to investigate
mathematically the fundamental notions 'number', 'order', and 'function',
taking them in their pristine, simple form, and to develop thereby the
logical foundations of all arithmetic and analysis; thus it constitutes an
indispensable component of the science of mathematics. (Zermelo (19086J,
200)
Set theory is essential to mathematics, especially arithmetic and
analysis. If we add to this the Quine/Putnam-style claim that
mathematics is essential to our theory of the world, we have an
indispensability argument for set theory. The question then
becomes, what version of set theory is essential to mathematics, and
in particular, does that version include choice? Zermelo argues that
it does:
the question that can be objectively decided, whether the principle is
necessary for science [by which Zermelo means the science of mathemat-
ics], I should now like to submit to judgement by presenting a number of
elementary and fundamental theorems and problems that, in my opinion,
could not be dealt with at all without the principle of choice. (Zermelo
(1908*), 187-8)
He goes on to list a series of theorems from set theory and analysis,
from which he concludes:
Now so long as the relatively simple problems mentioned here remain
inaccessible to Peano's [choiceless] expedients, and so long as, on the other
hand, the principle of choice cannot be definitely refuted, no one has the
right to prevent the representatives of productive science from continuing
to use this 'hypothesis'—as one may call it for all I care—and developing
its consequences to the greatest extent, especially since any possible
contradiction inherent in a given point of view can be discovered only in
that way . . . principles must be judged from the point of view of science,
and not science from the point of view of principles fixed once and for all.
(Zermelo (1908*), 189)
If choice produces a better, more effective theory than choiceless
120 AXIOMS

mathematics, Zermelo counsels that we opt for choice and jettison

any unscientific prejudice that stands in our way.
Zermelo's contention that mathematics without choice would be
'an artificially mutilated science'41 was substantially confirmed in
the years following. The implicit uses of choice in analysis were
uncovered gradually, and it is now clear that the theory of real
functions and Borel and projective sets sketched in the previous
section would change significantly without a weak version of choice
called 'dependent choice' and would collapse completely without
the even weaker 'countable choice'.42 The theory of transfinite
cardinal numbers, on the other hand, could hardly survive without
the full axiom of choice.43 And, to cite just one more example, the
importance of the principle in algebra is dramatically demonstrated
in the history of Bartel van der Waerden's classic textbook. The first
edition, published in 1930, included the axiom and its already
considerable store of algebraic consequences.44 The book itself
stimulated further productive research along these lines, but Dutch
opponents convinced van der Waerden to omit choice in his second
edition and return to more familiar methods. Algebraists were
appalled by this 'mutilated' version of their discipline, and the
axiom and its consequences were reinstated by popular demand in
the third edition in 1950.45
In the case of the axiom of choice, then, we have our first
example of an extrinsic defence of a set theoretic hypothesis,
beginning with a straightforward indispensability argument: our
best theory of the world requires arithmetic and analysis, and our
best theory of arithmetic and analysis requires set theory with at
least the axiom of dependent choice. Beyond this pure Quine/
Putnamism, the compromise Platonist finds the sort of intra-
mathematical arguments that Code! anticipates.46 First, as
41
Zermelo (1908^), 189.
42
The principle of countable choice says that there is a choice function for any
countable collection of non-empty sets; dependent choice says that these choices can
be made in such a way that each depends on the previous one. See Moore (1982),
103 and 325, and Moschovakis (1980), 423 and 445, for discussions of the role of
these principles.
43
For example, the principle that for any two cardinal numbers K and \, either
K = \ o r K < X o r X < K i s equivalent to the full axiom of choice. See Moore (1982),
§4.3, and 330-1, for more.
44
For a discussion of these, see Moore (1982), §3.5.
45
See Moore (1982), §4-5.
46
In Godel (1947/64), 477. All the quotations in this paragraph come from that
location.
 AXIOMATIZATION 121

Sierpiriski noted,47 it has a number of 'verifiable consequences',

that is, 'consequences demonstrable without the new axiom, whose
proofs with the help of the new axiom, however, are considerably
simpler and easier to discover . . .' Second, it yields a 'powerful method'
for solving pre-existing open problems, for example the well-
ordering question. And finally, it systematizes and greatly simplifies
the entire theory of transfinite cardinal numbers, 'shedding so
much light upon a whole field . . . that. . . [it] would have to be
accepted at least in the same sense as any well-established physical
theory'.
These are not, however, the only arguments mathematicians have
given in favour of the axiom of choice. To trace the source of the
other main line of defence, let me return to the two concepts of
collection discussed in the final section of Chapter 3. The
mathematical notion, originating in Cantor's thinking about sets of
pre-existing points, eventually developed into the full iterative
conception of Zermelo (1930). The logical notion, beginning with
Frege's extension of a concept, now takes a number of different
forms depending on exactly what sort of entity provides the
principle of selection, but all these have in common the idea of
dividing absolutely everything into two groups according to some
sort of rule.
In the early 1900s, these two notions had not yet been
distinguished, and this ambiguity is what produced the deepest
division over Zermelo's principle. One of the great ironies of this
entire historical episode is that the strongest negative reaction to the
axiom came from the very group of French analysts—Baire, Borel,
and Lebesgue—who unwittingly used it with great frequency and
whose work provides part of the basic indispensability argument
for at least dependent choice.48 Yet it is not hard to see how this
happened. The efforts of these analysts were originally motivated
by their doubts about the extremely general notion of function
proposed by Dirichlet and Riemann; instead they concentrated on
developing their hierarchies of functions definable by acceptable
mathematical means. The conflict here is between the notion of
functions as completely arbitrary correspondences, one for each
possible combination of pairs of reals, and the notion of functions
as transformations, determined by some sort of definition or rule.
47
In Sierpiriski (1918). For discussion, see Moore (1982), §4.1.
48
See Moore (1982), §§ 1.7, 2.3, and 4.1.
122 AXIOMS

Transferred into the realm of sets, this is just the contrast between
the mathematical and the logical notions of collection.
The true shape of this conflict emerged in the aftermath of
Zermelo's first proof, in a series of letters between the three analysts
and their opponent, Jacques Hadamard. 49 There Lebesgue writes:
to define a set M is to name a property P which is possessed by certain
elements of a previously defined set N and which characterizes, by
definition, the elements of M. . . . The question comes down to this, which
is hardly new: Can one prove the existence of a mathematical object
without defining it? . . . I believe that we can only build solidly by granting
that it is impossible to demonstrate the existence of an object without
defining it. (Baire et al. (1905), 314)
Hadamard's position was the opposite:
. . . Zermelo provides no method to carry out effectively the operation
which he mentions, and it remains doubtful that anyone will be able to
supply such a method in the future. Undoubtedly, it would have been more
interesting to resolve the problem in this manner. But the question posed in
this way (the effective determination of the desired correspondence) is
nonetheless completely distinct from the one that we are examining (does
such a correspondence exist?). , .. Can one prove the existence of a mathe-
matical object without defining it? I answer... in the affirmative. . . . the
existence . . . is a fact like any other . . . (Baire etal. (1905), 312, 317)

For Lebesgue and the rest, the existence of a choice function, or to

put it more simply, a choice set,50 depends on our having a rule
with which to determine what is in the set and what isn't. For
Hadamard, what rules we have is irrelevant, purely psychological;
a set either exists or it doesn't.
Lebesgue's point of view is obviously most plausible on the
logical notion of collection, the notion in fact suggested in the
above quotation. In contrast, the mathematical notion, according
to which, for any given things, there is a set consisting of any
combination of those things, aligns with Hadamard's thinking.
Given a collection of non-empty, disjoint sets formed from some
batch of things, a choice set will be among the combinations of
those original elements. From this point of view, Zermelo's axiom
becomes obvious. During the debate over choice, the notion of
49
Baire etal. (1905).
50
The axiom can be rephrased to say: for any collection of non-empty, disjoint
sets, there is a set that contains exactly one element from each of them.
 AXIOMATIZATION 123

collection in question was still ambiguous, which is why some

found the principle obvious and others found it preposterous.
Indeed, the very depth of the conviction on both sides suggests
that both notions of collection enjoy some intuitive backing. This
would also explain why many opponents of choice continued to use
it unawares even after the principle had been isolated and the
controversy was joined; recall that intuitions, properly so called,
are common to (nearly) all. On this theory, then, Zermelo's
historical evidence does support the intuitiveness of the principle,
but his linguistic formulation met with protest because the word
'set' therein wasn't always connected up with the appropriate pre-
linguistic intuitive beliefs.
Fleshing this idea out requires a few more speculative wrinkles to
the perceptual/intuitive story told in Chapter 2, but let me indicate
how it might go. We learn to perceive 'something with a number
property'. This concept develops in the course of extensive
childhood experience with manipulating and rearranging medium-
sized physical objects, so it is inextricably linked to finite
combinatorial notions, for example the idea that in any order, there
are still ten pennies, or that absolutely any proper subcollection of
the pennies will number fewer than ten. This underlies the
mathematical, combinatorial notion. On the other hand, on most
occasions of counting, the child counts things collected under some
umbrella: the boys in the garden, the pennies in my pocket. Even
seemingly random collections are for the most part spatially
circumscribed: the things in this box. This aspect of numerical
experience underlies the logical notion. Thus the intuitive concept
of 'that which has a number' contains elements of both notions.
Only later do we realize that these two notions can come apart.
In finite cases, it might be argued that any combinatorial possibility
is also determined by a property—the property 'being this thing or
that thing or . . . ' through a finite list of the members—though this
approach will be problematic, for example on accounts of scientific
properties as 'natural' collections. But infinite cases raise more
pressing doubts. Is there such a rule for determining membership in
each and every combinatorially determined subset of the natural
numbers? It isn't obvious that there is. Yet the combinatorial ideal
suggests that each and every such subset can be counted, just as
finite collections can be counted even when no non-trivial member-
ship rule is available. This clash of intuitions—between 'every
124 AXIOMS

collection is collected by some property' and 'any collection, with a

common property or not, can be counted'—leads to confusion.
Eventually, theoretical considerations take over. Sets rather than
classes make the most workable and fruitful mathematical entities,
and functions understood as arbitrary mappings provide an
important flexibility that narrower notions of function cannot
equal. By these means, we are led to the conclusion that being
collected under one umbrella was an accidental feature of the
'things with number properties' of our childhood experience, and
not an essential feature of all 'things with number properties'. Not
surprisingly, then, it is to such theoretical facts that Hadamard
ultimately appeals:
From the invention of the infinitesimal calculus to the present, it seems to
me, the essential progress in mathematics has resulted from successively
annexing notions which, for the Greeks or the Renaissance geometers or
the predecessors of Riemann, were 'outside mathematics' because it was
impossible to describe them. (Baireetal. (1905), 318)
In the years since this controversy raged, the mathematical
notion has been developed and accepted, because of its effective-
ness, and the Hadamard position has prevailed. Thus D. A. Martin,
one of our leading contemporary set theorists, writes:
much of the traditional concern about the axiom of choice is probably
based on a confusion between sets and definable properties. In many cases
it appears unlikely that one can define a choice function for a particular
collection of sets. But this is entirely unrelated to the question of whether a
choice function exists. Once this kind of confusion is avoided, the axiom of
choice appears as one of the least problematic of the set theoretic axioms.
(Martin, 'Sets versus classes' (unpublished), 1-2)
This last pro-choice argument, then, is that objections to choice are
based on the wrong notion of collection. It depends on both the
intuitive evidence for choice assuming the mathematical notion and
the theoretical evidence that the mathematical notion is the correct
one.

I have concentrated on the axiom of choice because its fascinating

history provides the clearest illustration of the interplay between
intuitive and theoretical supports for set theoretic hypotheses, but a
similar analysis can be given for each of the currently accepted
 OPEN PROBLEMS 125

axioms of Zermelo-Fraenkel set theory. I won't do this here,51

because my goal is illustrative rather than exhaustive, but let me at
least indicate the extreme variety of that list. Pairing—for any two
things there is a set with exactly those members—and Union—any
sets can be combined into a set with all their members as
members—have been cited earlier52 as examples of nearly un-
adorned intuitions. In contrast, the axiom of infinity—there is an
infinite set—proclaims the bold and revolutionary hypothesis that
led Cantor into his paradise and the rest of us with him. There is
nothing obvious about it, but it launched modern mathematics, and
the success and fruitfulness of that endeavour provides its purely
theoretical justification.

3. Open problems

By the mid-1930s, the fundamental assumptions underlying set

theoretic practice had been codified into a simple axiomatic system,
ZFC, which was strong enough to imply the known theorems of
classical number theory and analysis and pre-axiomatic set theory.
The well-orderability of the reals was provable within this theory,
leaving the continuum hypothesis in Cantor's preferred form: Kj =
2 °. But the question remained open, as did those raised by Suslin
and Luzin: is there an uncountable 11} set with no perfect subset?
Are all 2l and H\ sets Lebesgue measurable? Can disjoint 2] or II]
sets be separated?
Developments in logic during the twenties and early thirties,
especially Godel's completeness and incompleteness theorems of
1930 and 1931,53 raised the possibility of a new sort of proof—a
proof of unprovability—and it was in this direction that progress
on these open problems was first made. The most dramatic result of
Godel's work in the late thirties was his demonstration that the
51
Maddy (1988a) contains a summary of many of these arguments.
52
In Ch. 2, sect. 3, above.
53
Godel (1930; 1931). The completeness theorem establishes that every logically
valid formula is provable. (See Enderton (1972), ch. 2, for a now-standard proof
using the alternative methods of Henkin (1949).) The first incompleteness theorem
shows that there are sentences expressible in the language of ZFC that aren't
provable or disprovable from ZFC. The second incompleteness theorem gives an
example of such a sentence, namely, the one expressing the consistency of ZFC. (See
Enderton (1972), ch. 3.)
126 AXIOMS

continuum hypothesis can not be disproved from the axioms of

ZFC.54 At the same time, Godel noted some consequences for
analysis which were finally proved some time later by John
Addison:55 in ZFC, it cannot be proved that all uncountable Fl{ sets
have perfect subsets or that all A] sets are Lebesgue measurable. On
the question of separation, Addison used similar methods to show
that ZFC does not imply either the separability of X] sets or the
non-separability of FI* sets.56
The possibility remained that the axioms of ZFC would be
enough to establish the continuum hypothesis as true, but Godel for
one did not expect this. This conjecture was based partly on
evidence that ZFC is too weak to decide the question at all, and
partly on his strong conviction that the continuum hypothesis is in
fact false: 'certain facts (not known at Cantor's time) . . . seem to
indicate that Cantor's conjecture will turn out to be wrong . . .'
(1947/64), p. 479). These facts consist of 'highly implausible
consequences of the continuum hypothesis', of which Godel lists
several. Thus we find Godel arguing against Cantor's conjecture on
extrinsic grounds, in terms of its purportedly undesirable con-
sequences.
The question is, what makes these particular consequences
undesirable? Many of them depend on the idea that sets of reals
which are large in number should not also be small in measure-
theoretic terms. There might well be some intuitive belief lurking
behind this vague principle, but if so, it is extremely undependable:
an elementary theorem of measure theory shows that there are
uncountable sets of Lebesgue measure zero; the standard example
goes back to Cantor. Given that the suggested principle has been
discredited, it might be argued that Godel is relying on some other,
more dependable intuition. If so, we should expect most other set
theorists to share his views, but they don't. Martin, for example,

54
Unless ZFC is inconsistent; anything can be both proved and disproved in an
inconsistent system. (Terminology: to disprove is to prove the negation, so a proof
that CH cannot be disproved is a proof that its negation cannot be proved.) See
Godel (1940).
55
In Addison (1959). Other partial results by Mostowsla and Kuratowski were
destroyed during the war. See Addison's paper for the complicated history of these
theorems. Moschovakis (1980), § 5A, presents Addison's results.
56
In Addison (1958).
 OPEN P R O B L E M S 127

writes:57 'While Godel's intuitions should never be taken lightly, it

is ... hard for some of us to see why the examples Godel cites are
implausible at all' (Martin (1976), 87). The final, most likely
possibility is that Godel is relying, not on intuition, but on his
mathematical experience, exercising the sort of theoretical judge-
ment that produces the natural scientist's hunch that a theory of
this sort rather than that is the kind that ought to work.
Unfortunately, Godel's efforts to pin down his ideas have since
proved unsuccessful.58
But whatever his reasons, Godel was correct in his prediction
that the continuum hypothesis would be shown not to follow from
ZFC. This result was finally established by Paul Cohen in 1963.59
Solovay then extended Cohen's method to treat questions from
analysis: ZFC cannot disprove that all uncountable Fl] sets have
perfect subsets or that all £] an d H] sets are Lebesgue measur-
able.60 This inconclusive picture was completed by Leo Harring-
ton,61 who used Cohen's method to show that ZFC does not yield
the separation property for X] or Fl]. Thus these various problems
are open in an entirely new sense of the word; provably, they
cannot be decided on the basis of the standard assumptions of set
theory.

Many have been provoked to philosophical extremes by the

thought of questions this open. The if-thenist, for example, simply
declares all such problems to be solved; what we wanted to know,
after all, was whether or not the continuum hypothesis and the rest
follow from ZFC. But whatever machinations might be available to
her opponents, the Platonist's position is clear. In Godel's words:
It is to be noted, however, that on the basis of the point of view here
57
See also Martin and Solovay (1970), 177. Moore (forthcoming) attributes a
similar lack of agreement to Cohen. An exception to this rule is Nyikos
(forthcoming), who agrees with Godel that at least one of these examples is
extremely implausible.
58
For an account of these efforts, see Moore (forthcoming), § 6.
59
The published version is Cohen (1966).
60
See Solovay (1970). Solovay's result actually depends on the relatively weak
additional assumption that the existence of an inaccessible cardinal (see the next
section) cannot be refuted in ZFC. Moschovakis writes, 'In the present context this is
surely a reasonable assumption' (1980), p. 284), and I know of no dissent from this
view.
61
See Moschovakis (1980), 284.
128 AXIOMS

adopted, a proof of the undecidability of Cantor's conjecture from the

accepted axioms of set theory . . . would by no means solve the problem.
For if the meanings of the primitive terms of set theory [based on the
iterative conception] are accepted as sound, it follows that the set-
theoretical concepts and theorems describe some well-determined reality,
in which Cantor's conjecture must be either true or false. Hence its
undecidability from the axioms being assumed today can only mean that
these axioms do not contain a complete description of that reality. (Godel
(1947/64), 476)
For the set theoretic realist, the world consists of physical objects,
sets of these, sets of physical objects and sets, and so on, through
the transfinite levels of the iterative hierarchy. There is a fact of the
matter about the cardinality of the set of Dedekind cuts—a set of
pairs of sets of sets of pairs of von Neumann ordinals—and it is the
set theorist's job to discover it.62
And how might this elusive fact be ascertained? Godel's answer,
and the set theoretic realist's, is that we need to find new axioms,
axioms we can justify just as Zermelo's axioms were justified, by a
combination of intrinsic and extrinsic considerations.63 This
approach is seconded by our current set theorists. For example,
Martin writes: 'Although the ZFC axioms are insufficient to settle
CH, there is nothing sacred about these axioms, and one might
hope to find further axioms . . .' (Martin (1976), 84). Considerable
work has been done on this project in recent years, some of which
will be described in the next section.
In another attack on Godel's ideas, Chihara takes this situation
as evidence against Platonism.64 Godel's and Cohen's results show
that both ZFC + CH and ZFC + not-CH are consistent theories,
perhaps equally worthy of investigation. 65 Chihara concludes that
For Godel, . , . the proliferation of set theories poses the thorny problem of
62
Of course, set theory is as fallible as any other science, and it could turn out
that the continuum question is based on faulty presuppositions, but there is no
conclusive reason to believe this now.
63
See G6del{ 1947/64), 476-7.
64
See Chihara (1973), 63-5. All quotations in this paragraph and the next
come from this location.
65
In fact, neither of these theories is much studied, as such, because neither CH
nor not-CH is considered a viable axiom candidate. Neither is intuitive and neither is
sufficiently fruitful to merit acceptance on extrinsic grounds, though CH is better off
than not-CH in this regard: Sierpiriski (1934) derives eighty-two propositions from
CH, none of which is known to be settled by not-CH (see Martin and Solovay
(1970), 143).
 OPEN P R O B L E M S 129

determining which of the many set theories [one for each possible
cardinality of the continuum] is the one that most truly describes the real
world of sets. (Chihara (1973), 65)
Now no Platonist would deny that the continuum hypothesis poses
a 'thorny problem'—it has engaged many of the best set theoretic
minds since Cantor's—but it isn't immediately clear why this casts
doubt on Platonism. After all, scientific realism leaves us with many
thorny problems, from the shape of the universe to the existence of
a free quark, and no one counts this as evidence that there is no
objective physical world.
On the Platonist's view, there is a real and extremely difficult
problem about the cardinality of the continuum. Chihara seems to
hold that this fact counts in favour of alternative philosophies of set
theory for which the continuum problem presents no such
challenge. For example, he suggests as a 'reasonable option' his
'mythological Platonism', which takes the continuum hypothesis to
be analogous to 'Hamlet's nose is three inches long', that is, to be
neither true nor false. But how reasonable would it be for the
physicist to solve the question of the free quark by adopting a
philosophy of physics according to which it is no longer a problem?
In fact, thorny problems are the life-blood of science, its motivator,
and set theory is no different from the rest.

If we are to look towards new axioms for a solution to the

continuum problem, it is worth asking in which direction that
solution might be expected. Opinion on this matter is divided, but
opinion there is. No one pretends to have anything resembling
conclusive evidence for any alternative, but a brief look at the range
of considerations offered will give the flavour of the debate.
Cantor, of course, held the continuum hypothesis to be true, and
occasionally, during his many efforts, even believed that he had
proved it.66 One of the forces behind his strong conviction may
have been his confidence that the partial solution contained in the
Cantor-Bendixson theorem could be generalized.67 To a certain
extent, we've seen that this confidence was well placed: Cantor—
Bendixson was extended, most dramatically by Suslin to the SJ
sets. Thus, despite the fact that there are far fewer Si sets than there
66
For examples, see Moore (1982), 42-5, and Hallett (1984), 92.
67
Cantor had other reasons, too. See Maddy (1988a), 490—2, and the references
cited there.
130 AXIOMS

are arbitrary sets of reals,68 this partial result affirming the

continuum hypothesis for a wide range of sample sets might be
taken as confirmation for the hypothesis in general.
Unfortunately, whatever confirming evidence Suslin's result may
have promised, it is severely undercut by the details of the argument
itself. What he actually shows is that every uncountable X j set has a
perfect subset. The trouble is that some uncountable sets don't have
perfect subsets, so there is actually no hope of generalizing Suslin's
result on 2} sets to all sets of reals. In Martin's words: 'Thus, while
our simple [£{] sets have the cardinalities required by CH, this is so
because they have an atypical property, the perfect subset prop-
erty' (Martin (1976), 88). In Cantor's defence, it should be noted
that most of the sets he was familiar with were S{ at worst, and
Bernstein's theorem on uncountable sets without perfect subsets
didn't appear until 1908.
If Cantor's reasons for believing the continuum hypothesis are
ultimately unpersuasive, GodePs reasons for disbelieving it have
also drawn few converts. By contrast, the sentiments of another
major player in this drama proceed along lines that many seem to
find more plausible. Cohen's thinking depends on a contrast
between two ways in which larger cardinals can be generated from
smaller ones.
One method builds up from below. Cantor's original procedure
for building ever larger ordinals depended on three principles of
generation.69 The first allows the passage from one number to its
successor, from 2 to 3, from 3 to 4. There is no largest number in
this series, so Cantor's second principle generates their 'limit', the
next number after them all, that is, co. From here, the first principle
yields co -f 1, co + 2, etc., and the second, to + to. And so on. But all
ordinals generated by these processes are still countable. The third
principle tells us that after all ordinals of a certain cardinality, there
is a next, in this case, CO]. This ordinal is uncountable; its cardinal
number is NI.
These methods produce a sequence of infinite cardinal num-
bers—K0, K!, K2, N3, etc.—but they cannot take us beyond this
point. To build up further from below, we need the axiom of
replacement: given any set, if each of its elements is replaced by
something else, the result is still a set. Thus if (0, 1, 2, 3, . . . } is a
69
There are 2*° 2,} sets of reals and 22 " sets of reals altogether.
69
Cantor (1883). See Dauben (1979), 96-9, or Hallett (1984), §2.1, for
discussion.
 OPEN P R O B L E M S 131

set, as the axiom of infinity guarantees, and 0 is replaced by the set

of finite ordinals, 1 by the set of countable ordinals, 2 by the set of
ordinals of size K l 5 and so on, we have the set whose elements have
the cardinalities K0, K l 5 K2, and so on. If we take the union of all
these sets, as the union axiom says we can, the result is a set whose
cardinality is the next largest after K0, X l 5 N2, etc. This is K w .
Obviously, this process can be continued.
The second way of generating larger cardinalities is very
different. Beginning again with the set {0, 1, 2, 3, . . .}, this time we
take the set of its subsets, as allowed by the power set axiom. One
of Cantor's most beautiful theorems shows that the power set of
any set has a larger cardinality than the set itself;70 in this case, that
larger cardinality is 2 , also the cardinal of the set of reals.71
Taking the power set of this power set yields a set of cardinality
22 °, and so on.
Cohen's idea is simply that power set is stronger than any
principle for building up from below.72 In his book establishing the
unprovability of the continuum hypothesis, he writes:
A point of view which the author feels may eventually come to be accepted
is that CH is obviously false. . . . Nj is the set of countable ordinals and this
is merely a special and the simplest way of generating a higher cardinal.
The set [of subsets of the natural numbers] is, in contrast, generated by a
totally new and more powerful principle, namely the Power Set Axiom. It is
unreasonable to expect that any description of a larger cardinal which
attempts to build up that cardinal from ideas deriving from the
Replacement Axiom can even reach [a set of size 2 ]. Thus [2 °] is greater
than Kn, Xw, Xa, where a. = XM, etc. This point of view regards [the
power set of the set of natural numbers] as an incredibly rich set given to us
by one bold new axiom, which can never be approached by any piecemeal
process of construction. (Cohen (1966), 151)

This line of thought harmonizes with various others insisting that

the continuum hypothesis places an artificial and unwarranted
restriction on the number of reals.73 Most such thinkers—they
include Martin 74 —feel that the continuum is likely to be quite large
compared with K t . This sets them apart from Godel, who, while
70
Cantor (1891). See Enderton (1977), 132-3.
71
See Enderton (1977), 149.
72
Dauben (1979), 269, traces this way of thinking to Baire.
73
See Maddy (1988a), § n.3.4. This paper gives a more complete and detailed list
of the various arguments for and against Cantor's hypothesis.
74
See e.g. his Trojective sets and cardinal numbers' (unpublished), 2. Some of
Martin's views are reported in Maddy (1988a), § v.4.
132 AXIOMS

rejecting Cantor's hypothesis, seemed to lean towards a relatively

small continuum of size X2.75

So the axiomatization of set theory produced a range of problems

more open than had previously been possible, that is, problems
neither provable nor disprovable from the accepted assumptions in
the field. The most famous of these is Cantor's continuum
hypothesis, but various others appeared among the natural
questions asked by analysts in the twenties and thirties. From the
Platonist's perspective, there is good reason to believe that these
questions nevertheless have unambiguous answers; some even
proffer opinions about what those answers might be. The hope is
for new, strongly supported axioms that will resolve these difficult
issues.

4. Competing theories

In the current set theoretic landscape, two opposing theoretical

approaches dominate efforts to solve the profoundly open prob-
lems described above. My goal in this section is to give an overview
of each, with special attention to perceived strengths and weak-
nesses. In the final section of this chapter, I'll turn to the
philosophical questions raised by this controversy.

The most complete and concise of these two theories stems from
Godel's proof that the continuum hypothesis cannot be disproved
from the axioms of ZFC. To show this, Godel described a
simplified world of sets in which all the axioms, and hence all
consequences of the axioms, are true. But in this world, K j = 2 ',
so the negation of the continuum hypothesis cannot be proved from
ZFC.76
This simplified world of Godel's is arranged in a hierarchy of
stages, just like the standard iterative hierarchy, and there is one
stage for every ordinal number, again in imitation of the standard
picture. The difference is that at any given stage, instead of forming
all possible subsets of what has been given so far, in Godel's world
See Moore (forthcoming), esp. § 6.
76
See Drake (1974), ch. 5, or jech (1978), §§ 12-13, for textbook presentations
of this argument.
 COMPETING THEORIES 133

one adds only those subsets explicitly definable by predicative77

formulas. This procedure yields the constructible universe, called L,
and the claim that the constructible universe is the real universe,
written V = L, is the axiom of constructibility. Godel's work shows
that adding the axiom of constructibility to ZFC can't introduce
any contradictions that weren't already present in ZFC alone.
So one live theoretical option is just that: add V = L to the
assumptions of ZFC. This move produces a theory so powerful that
the axiom of choice is no longer needed; it can be proved. In fact,
not only is every set well-orderable—a condition equivalent to
choice78—but the entire universe can be arranged in a giant well-
ordering—a condition called 'global choice'. This holds in L
because the relevant formulas can be well-ordered at each stage,
which produces in turn a well-ordering of the sets introduced at
that stage. The well-orderings of the stages are then strung together
to produce a well-ordering of the entire constructible universe. The
axiom also has striking effects in the area of cardinal arithmetic; it
implies not only the continuum hypothesis, but the generalized
continuum hypothesis as well, that is, that N a + i = 2Kot, for all
ordinals a.
The open questions from analysis are also decided in ZFC + V =
L. Addison's unprovability results, like Godel's, are established by
showing that the negations of the propositions in question follow
from the axiom of constructibility. In the constructible universe,
there is an uncountable Flj set with no perfect subset, a non-
Lebesgue-measurable A2 set, and separation holds for FI^, Fl^,
Us, and so on. All these follow from the existence of a
particularly simple well-ordering of the reals; as a subset of the
plane, it is A£.
So V = L answers all our outstanding questions. Indeed, further
elaborations, largely due to Ronald Jensen, settle nearly all
important set theoretic questions, and some from other branches of
mathematics as well.79 Here is an axiom that clearly provides

77
That is, formulas that only refer to sets formed at previous stages.
78
SeeEnderton (1977), 151-4, 196-7, 199.
79
Devlin (1977) makes the case for V = L's efficacy both inside and outside set
theory. The importance of Jensen's contribution to constructibility theory comes out
in the historical notes to Devlin's compendium (1984) and in the introduction:
'without his work there would have been practically nothing to write about!' (p.
viii).
134 AXIOMS

'powerful methods for solving problems'. 80 It is also a safe

assumption; as remarked above, it engenders no contradictions that
wouldn't already follow from ZFC alone. There are even those who
find it a 'natural' assumption, beginning with Godel, who intro-
duced it this way:
The proposition [V = L] added as a new axiom seems to give a natural
completion of the axioms of set theory, in so far as it determines the vague
notion of an arbitrary infinite set in a definite way. (Godel (1938), 557)

Keith Devlin goes even further, claiming constructibility to be

'closely bound up with what we mean by "set" ', but he defends
this by identifying sets with extensions of properties rather than
combinatorially defined collections.81
Despite the remark just quoted, Godel soon came to reject V =
L, and despite its strengths, constructibility today has more
detractors than supporters. The most fundamental reason for
resistance to the axiom is implicit in my crude sketch: instead of
forming all possible subsets of what has been given so far, one adds
only those subsets explicitly definable by predicative formulas. This
requirement on subsets clearly violates the combinatorial idea that
every possible collection be formed, regardless of whether there is a
rule for determining which previously given items are members and
which are not. Devlin, to bolster his case, concocts a clever
compromise between the logical and the mathematical notions of
collection—he proposes that extensions of properties be formed in
stages,82 thus avoiding inconsistency—but the combinatorial idea
he rejects stands at the end of a clear historical trend in
mathematics, from functions and collections determined by rules
towards functions and collections determined arbitrarily. Most set
theorists have adopted the full iterative conception and thus find
the axiom of constructibility an artificial restriction. For example,
Moschovakis writes:
The key argument against accepting V = L ... is that the axiom of construct-
ibility appears to restrict unduly the notion of arbitrary s e t . . . there is no

80
Godel (1947/64), 477.
81
See ch. 3, sect. 3, above for this distinction. The quotation is from Devlin
(1977), p. iv. The relevant notion of set is sketched on pp. 13-18. Fraenkel,
Bar-Hillel, and Levy (1973), 108-9, also cite a range of considerations in favour of
V = L.
82
See Devlin (1977), 27-8.
 COMPETING T H E O R I E S 135

a priori reason why every subset. .. should be definable . .. (Moschovakis

(1980), 610)
And Godel: '[V = L] states a minimum property. Note that only a
maximum property would seem to harmonize with the concept of
set. . .' (Godel (1947/64), 479). Many others express opinions along
these lines.83
Further arguments against the axiom of constructibility focus on
its consequences. Of course, anything that might count as a reason
for disbelieving the continuum hypothesis would likewise count
against V = L; it may be that Godel's change of mind was partially
motivated in this way. Though anti-constructibility arguments of
this form are sometimes offered, more common and concrete
objections involve the consequences of V = L for the occurrence of
so-called pathologies low down in the projective hierarchy, among
the fairly simple sets of reals. For example, the axiom of choice
implies that the real numbers can be well-ordered, but there is every
reason to suppose that the simplest such well-ordering (as a subset
of the plane) is an extremely complex set. Provably, it cannot be as
simple as £{, but in the presence of the axiom of constructibility, it
is A|. This means that, beginning from a set too simple to be a well-
ordering of the reals, such a well-ordering can be obtained by one
application of complementation and one application of projection,
a prospect which seems highly unlikely to many. Indeed many set
theorists feel that such a choice-generated oddity should not appear
anywhere among the projective sets. The same sort of thinking
applies to the uncountable II j set with no perfect subset and the Al
non-Lebesgue-measurable set; though choice guarantees that such
sets exist, they should not be so simple.84
The second live theory begins from a style of axiom for which
Godel had high hopes, namely the large cardinal axiom. The first of
these, the axiom of inaccessible cardinals,
roughly speaking, means nothing else but that the totality of sets
obtainable by use of the procedures of formation of sets expressed in the
other axioms forms again a set (and, therefore, a new basis for further
applications of these procedures) . . . {Godel (1947/64), 476)
83
See e.g. Drake (1974), 131, Scott (1977), p. xii, and Wang (I974a), 547.
84
Opinions of this sort can be found in Martin (1977), 811, (1976), 88, and
'Projective sets and cardinal members', p. 2; Moschovakis (1980), 276; and Wang
(1974a),547.
136 AXIOMS

The idea is that the two main operations for generating new sets
from old postulated by ZFC—replacement and power set—are not
enough to exhaust all the ordinals.85 Obviously this thinking is ripe
for generalization. Such axioms, according to Godel, show that
ZFC 'can be supplemented without arbitrariness by new axioms
which only unfold the content of the [iterative conception of
set] . . .' (Godel (1947/64), 477). He proposed large cardinal axioms
as a cure for the sort of open problems we've been considering:
It is not impossible that. . . some completeness theorem would hold which
would say that every proposition expressible in set theory is decidable from
the present axioms plus some true assertion about the largeness of the
universe of all sets. (Godel (1946), 85)
By the sixties, large cardinals of ever-increasing size were a boom
industry.
The most striking application of inaccessibles themselves is
Solovay's theorem, mentioned above, applying Cohen's method to
questions in analysis. This important work presupposes the
irrefutability, if not the existence, of an inaccessible cardinal. Other
small large cardinals have consequences for Borel sets,86 but the
most dramatically effective large cardinal axiom postulates the
existence of a much larger measurable cardinal. 87 To give just a
modest hint of its size, there are inaccessibly many inaccessible
cardinals smaller than the first measurable.
The strongest arguments for the assumption of measurable
cardinals are extrinsic ones, most notably Dana Scott's discovery
that it implies V ^L. 88 Thus the strong considerations against the
axiom of constructibility all return as extrinsic supports for the
existence of a measurable cardinal. And, once again, Solovay
coaxed out results in analysis. He showed that in addition to
refuting constructibiiity, a measurable cardinal yields the preferred
results for projective sets: every uncountable ^\ set nas a perfect
subset, every Si and FI^ set is Lebesgue measurable, and there is no
85
Inaccessibles were introduced in Zermelo (1930) and in Sierpiriski and Tarski
(1930). For the sort of argument in their favour considered here, see also Drake
(1974), 267, and Wang (1974a), 554, For arguments of other types, see Maddy
(1988a), 501-4.
86
See the detailed and innovative work of Harvey Friedman, described in
Harrington et al. (1985).
87
Measurable cardinals were first proposed by Ulam (1930). See Drake (1974),
chs. 6 and 8, or Jech (1978), ch. 5, for textbook discussions.
88
Scott (1961).
 C O M P E T I N G THEORIES 137
A] well-ordering of the reals.89 But for all the good news, there is
also bad. The most conspicuous open problem remains so; Solovay
and Levy's application of Cohen's method shows that measurable
cardinals cannot decide the continuum hypothesis.90
Then, in the late sixties, the appearance of a short paper by David
Blackwell91 produced a surge of interest in hypotheses of a
completely different sort, hypotheses given in game-theoretic terms.
The notion of an infinite game was first introduced by the Polish
school in the thirties. Such a game can be based on any set A of real
numbers between 0 and 1. Imagine two tireless players who takes
turns choosing digits. When they are done (!), they will have
constructed an infinite sequence, which can be taken as a decimal
expansion, which represents a real number, r. If r is in A, the first
player wins; otherwise the second player wins. The game A is
determined if there is a winning strategy for one player or the other.
Blackwell used the known fact that all open games92 are deter-
mined to give a new and elegant proof of Luzin's theorem on the
separability of S} sets.
This came as a welcome surprise to Moschovakis, Addison, and
Martin, all of whom were engaged in efforts to go beyond Novikov's
theorem and extend the separation property to higher levels of the
projective hierarchy. From the researches of the early analysts, it
was known that separation holds for sets at the circled levels:

In the constructible universe, Addison had shown that the pattern

extends on the IT-side:

89
Solovay (1969).
90
Levy and Solovay (1967).
91
Blackwell (1967).
92
That is, games whose set of wins for the first player is an open subset of [0,1].
138 AXIOMS

Even for disbelievers in V = L, this meant that ZFC can't disprove

separation for ILj sets, but for all that was known at the time, it
remained possible that ZFC could prove Fl] sets separable. Still,
few expected this. A new hypothesis was needed.
In this context, then, BlackwelPs proof focused considerable
attention on determinacy assumptions. £3 sets were known to be
determined, a result which Martin later extended to all Borel sets.93
On the other hand, the axiom of choice implies the existence of a
non-determined set,94 But again, it seems natural to insist that such
a choice-generated oddity not appear among the relatively simple
sets, for example among the projective sets. Thus protective
determinacy—the assumption that all projective sets of reals are
determined—was proposed by a wide range of researchers.95
Projective determinacy, like the axiom of measurable cardinals, is
supported by many of the considerations thought to count against
V = L. For example, it extends the results obtainable from a
measurable cardinal by guaranteeing not only that every uncount-
able S| set has a perfect subset, but that every uncountable
projective set has a perfect subset; not only that all 2j and FI^ setsa
are Lebesgue measurable, but that all projective sets are Lebesgue
measurable; not only that there is no A] well-ordering of the reals,
but that there is no projective well-ordering of the reals.96 Many
agree with Martin that these are 'pleasing consequences about the
behavior of projective sets'.97 And the separation question, the
problem that inspired this renewed interest in determinacy, was
solved by Addison and Moschovakis and independently by
Martin;98 the initial zigzag pattern continues for the length of the
projective hierarchy:
93
The result for £3 is in Davis (1964), Martin's in (1975). See Moschovakis
(1980), §6r, for a proof of Martin's theorem for the finite levels of the Borel
hierarchy. Martin (1985) gives a simplified proof for all Borel sets.
94
Gale and Stewart (1953). See Moschovakis (1980), 293.
95
First Solovay and Takeuti, independently, then Addison, Martin, and
Moschovakis. (See Addison and Moschovakis (1968), 708-9, Moschovakis (1970),
31, and (1980), 422, 605, and 610-11, Martin (1976), 90, (1977), 814, and
'Projective sets and cardinal numbers', p. 8.) In fact, most of these propose a
stronger axiom candidate, quasi-projective determinacy, but I won't go into the
exact definition here. See Maddy (1988#), 737. Determinacy assumptions were
introduced by Mycielski and Steinhaus (1962).
96
Lebesgue measurability appears in Mycielski and Swierczkowski (1964), and
the perfect subset property in Davis (1964). The non-existence of a projective well-
ordering follows because a well-ordering is not Lebesgue measurable.
97
Martin (1976), p. 90.
98
Addison and Moschovakis (1968) and Martin (1968).
 COMPETING THEORIES 139

This pattern is considered more natural than the one generated by

V = L, if only because AAAAAAAAAAA is a more natural
continuation of A than A "
Other extrinsic evidence for projective determinacy is found in its
strong intertheoretic connections with the axiom of measurable
cardinals,100 and in the naturalness of the new game-theoretic proofs:
One [reason for believing projective determinacy] is the naturalness of the
proofs from determinacy—in each instance where we prove a property of
Il3 (say from [the determinacy of A] sets]), the same argument gives a new
proof of the same (known) property for H\ . . . Thus the new results appear
to be natural generalizations of known results and their proofs shed new
light on classical [analysis], (This is not the case with the proofs from V =
L which all depend on the [A^>] well-ordering of [the reals] and shed no light
on nj.) (Moschovakis (1980), 610)

Moschovakis's thick book contains various beautiful and per-

suasive examples.101
But perhaps the most striking feature of determinacy hypotheses,
what makes this a particularly fascinating case for the philosopher,
is that all arguments given in its favour from the mid-sixties until
the mid-eighties are extrinsic. Determinacy supporters were quite
explicit on this point:
No one claims direct intuitions . . . either for or against determinacy
hypotheses . . . (Moschovakis (1980), 610)
There is no a priori evidence for [projective determinacy] . . . (Martin
(1976), 90)
Is [projective determinacy] true? It is certainly not self-evident. (Martin
(1977), 813)

For twenty years, while extrinsic arguments of the sort outlined

here developed rapidly, there was no change in the lack of intrinsic
support. And yet projective determinacy was still considered a
viable axiom candidate.

"9 See Moschovakis (1970), 33-4; Martin (1977), 806, 811, and 'Projectice sets
and cardinal numbers', p. 8; Wang (1974«), 547, 553^. For other reasons, see
Maddy(1988<z), §v.l.
lu lcn
° SeeMaddy (19880), §v.2. See also Maddy (1988), §v.3.
140 AXIOMS

The best hope for something more than purely extrinsic evidence
lay in the possibility of deriving determinacy hypotheses from
suitable large cardinal axioms. 102 Martin showed, early on, that the
determinacy of S| sets is implied by the existence of a measurable
cardinal:
Some set theorists consider large cardinal axioms self-evident, or at least as
following from a priori principles implied by the concept of set. [The
determinacy of 2} sets] follows from large cardinal axioms. It is possible
that [projective determinacy] itself follows from large cardinal axioms, but
this remains unproved. (Martin (1977), 813)
One way to increase the evidence for [projective determinacy] would be to
prove it from large cardinal axioms . . . (Martin, 'Projective sets and cardinal
numbers' (unpublished), 8)
Unfortunately, attempts to extend Martin's result made use of
cardinals so large that even the most enthusiastic large cardinal
theorists were concerned for their consistency.103
Before turning to the developments of the mid-eighties, we
should pause to ask why deriving determinacy hypotheses from
large cardinal assumptions is viewed as providing non-extrinsic
support. Obviously no such theorem can provide direct intrinsic
support for a determinacy hypothesis; this a hypothesis either has
or lacks on its own. What happens in such a case is that the
determinacy hypothesis proved inherits the existing supports for
the large cardinal assumption from which it is proved. Thus
Martin's theorem places the power of arguments for measurable
cardinals squarely behind the determinacy of 2] sets. But the
arguments given above for measurable cardinals were all ex-
trinsic!104 Where is the intrinsic evidence supposed to come from?
The answer is that there are various other arguments for
measurable cardinals that don't depend on consequences. In fact,
however, I think these arguments rest on ideas that are not happily
classified as either intuitive or extrinsic, ideas Martin referred to

102
Moschovakis (1970), 31, notes Solovay's conjecture that this might he possible,
and cites Martin's result as confirming evidence.
103
Martin's theorem on measurable cardinals and determinacy appears in
Martin (1970). Martin (1980) extends the result to ii sets, and Woodin took this
line even further. See Maddy (1988ii), §vi, and Martin and Steel (1989) for
discussion of these developments and the very large large cardinals involved.
104
In fact, its implication of the well-supported X] determinacy is also counted
as extrinsic evidence for measurable cardinals.
 COMPETING THEORIES 141

above as 'a priori principles implied by the concept of set'. One

example has already been given in favour of inaccessibles: the idea
that the iterative hierarchy is too large and complex to be exhausted
by the simple operations of replacement and power set. The more
general idea that motivates all large cardinals is simply that the
universe goes on through as many transfinite stages as it can.
Various large large cardinals are also defended on grounds arising
from the idea that the hierarchy of stages is so complex and rich
that it must contain stages that resemble one another in certain
ways.105 For want of a better term, I call these 'rules of thumb'.
The rules of thumb underlying large cardinal axioms are clearly
rooted in the iterative conception, which is drawn, I've suggested,
from intuition. Why, then, do I avoid attributing intuitive status
also to these rules of thumb? The answer is simple: because I think
they extend beyond anything that could plausibly be traced to an
underlying perceptual, neurological foundation. That sets are
formed from previously given things, that they are formed
combinatorially, with no concern for rules of formation—these
ideas might well have intuitive backing, and there is no doubt they
figure centrally in the iterative conception. But if, as I've suggested,
the support for the assumption of an infinite stage is purely
extrinsic, following from the immense success of modern infinitary
mathematics, then part of the standard iterative conception, the
part that drives the hierarchy into the infinite and insists that it go
on as long as possible, that part is based on the developing
methodology of set theory itself, not on simple intuition.
On the other extreme, why shouldn't these principles be counted
as extrinsic? Natural science has its own principles of similar
generality, for example Maxwell's principle that a law of nature
should be valid at all points in space and time.106 Such principles
are indirectly subject to extrinsic support—if they consistently led
to ineffective theories, they would eventually be dropped—but this
proves nothing; even intuitions must be confirmed by conse-
quences. An unavoidably central aspect of the appeal of these rules
of thumb—be they scientific or mathematical—is that they 'seem

105
See Maddy (l9SSa), § vi.2. Though the arguments for large large cardinals are
similar is some structural and philosophical respects to the arguments for small large
cardinals, they cannot be said to be equally convincing. Again, Maddy (1988a)
provides details.
106
Wilson (1979) discusses this example.
142 AXIOMS

right', even if this seeming is unlikely to enjoy a strictly intuitive

basis. If we confine ourselves to an unbiased description of practice,
I think we must admit that rules of thumb fall somewhere between
the intuitive and the extrinsic.
Leaving aside for now the problem of elucidating and defending
the efficacy of this new category of purported evidence, let me pick
up the thread of the story of determinacy. Large cardinals are
thought to enjoy certain non-extrinsic justifications in terms of
various rules of thumb. Thus, if projective determinacy could be
proved from such an axiom, its purely extrinsic defence would be
enriched by support from this other source. This, then, was the
goal.
What happened in the mid-eighties was a fulfilment of this hope.
Work of Martin, John Steel, and Hugh Woodin107 established that
projective determinacy 108 can be proved from the existence of a
supercompact cardinal. Though supercompacts are much larger
than measurables—there are measurably many measurable cardi-
nals below the first supercompact—they can be viewed as a
generalization of that notion.109 Thus the theory ZFC + SC (there
is a supercompact cardinal) enjoys all the extrinsic supports of
projective determinacy plus any non-extrinsic, rule-of-thumb evi-
dence available for large cardinals in general and for supercompacts
in particular. 110 It solves all the outstanding open problems from
analysis, and it does so in ways that many find natural. In many
respects, then, ZFC + SC presents an attractive alternative to
ZFC + V = L.
But what about the continuum hypothesis? If it is true in L, a
minimal environment, perhaps it is false in a maximized world of
large cardinals. There is some evidence in that direction in results
proved from the full (false) axiom of determinacy: if every set is
determined, then the reals can be mapped onto N%2, Kw + i, Nj"*0) + l5
and beyond. In the presence of the axiom of choice, this would
imply the existence of sets of reals of these cardinalities, badly
falsifying the continuum hypothesis, but full determinacy also
107
See Martin and Steel (1988), Woodin (1988), and Martin and Steel (1989).
These papers also contain useful historical information.
108
Indeed more. The full quasi-projective determinacy, mentioned in n. 95
above, is provable from this large cardinal assumption.
109
See e.g. Solovay, Reinhardt, and Kanamori (1978), §2. For a textbook
discussion of supercompactness, seejech (1978), 407—13.
110
Forasurvevof the latter, see Maddv (L988j), §§vi.l and vi.2.
 THE C H A L L E N G E 143

implies that choice is false. Indeed, in that strange world of full

determinacy, every uncountable set of reals has a perfect subset, so
the continuum hypothesis is true in Cantor's original form: all
infinite sets of reals have the cardinality of the naturals or the
cardinality of the reals themselves. But in Cantor's favoured form—
Kj = 2 °—it remains false; the continuum isn't of size Ka for any
a because it can't be well-ordered.
It isn't clear what all this madness means for the real world if
there is a supercompact cardinal and the axiom of choice is true.
Supporters of ZFC + SC disagree over their expectations for the
size of the continuum,111 not because some support restrictive
principle like V = L, but because they disagree over whether Kj = 2
or its opposite is the least restrictive.112 (Notice, it is possible to
read the equation as restricting the number of reals, but also as
maximizing the number of countable ordinals.) Efforts are under way
to extend the theory to decide the question one way or the other.
I have described two theories, two extensions of ZFC, that cannot
both be true. Each theory answers at least the open questions of
Luzin and Suslin, and one even decides the size of the continuum.
Each enjoys an array of extrinsic supports, supplemented to varying
degrees by intuitive and rule-of-thumb evidence, a small portion of
which has been described in this summary. The philosophical open
question is: on what rational grounds can one choose between these
two theories?

5. The challenge

We've seen that set theory arose in response to both foundational

and purely mathematical concerns, that it developed as a branch of
analysis and as a study in its own right, and that both pursuits
produced natural open questions. In the confusion surrounding the
paradoxes and Zermelo's controversial proof of the well-ordering
theorem, the informal practice was axiomatized, which develop-
ment had two consequences of concern to us here: the role of
111
For example, Foreman (1986) proposes 'generic large cardinal' axioms that
imply CH. Martin, on the other hand, has conjectures about the relationship
between the cardinals in the world of full determinacy and the real world that imply
the CH is badly false. See Maddy (1988a), § v.4.
112
See Maddy (\988a), §§ n.3.4 and n.3.11.
144 AXIOMS

extrinsic argument in mathematics was crystallized as never before,

and it became possible to show that various natural open problems
in set theory and analysis were open in a new and stronger sense.
The hope of answering these extremely open questions led to the
emergence of two competing theories that answer all or most of
them and do so in very different ways. The difficult problem facing
the set theorists of our day is how to adjudicate between these, that
is, how to determine, within the limits of our cognitive capacities,
which is more likely to be true.
Proof is obviously the most common source of evidence in
mathematics, but even proof must begin from axioms that are not
themselves proved. In many circles, the preferred account of our
knowledge of axioms is that they are somehow self-evident; in the
words of Roderick Chisholm, following Leibniz, 'once you under-
stand it, you see that it is true'. 113 But our brief survey in this
chapter shows that even the accepted axioms of ZFC do not enjoy
this status, let alone the more controversial axiom candidates like
projective determinacy. A new account of our knowledge of axioms
and of the evidential role of non-demonstrative mathematical
arguments in general is clearly needed.
For the set theoretic realist, non-demonstrative arguments come
in two varieties—intuitive and extrinsic—and examination of cases
reveals what is probably at least one more—those based on rules of
thumb. In Chapter 2, I argued that intuitive support is prima-facie
evidence for truth. Perhaps it is not surprising that most of the
available intuitive evidence is marshalled in support of the accepted
axioms of ZFC. In the dispute between V = L and SC, I've
suggested that V = L labours at a modest intuitive disadvantage,
because it rejects the combinatorial component of set theoretic
intuition. But, given that intuitive evidence is never conclusive, that
it needs supplementation by extrinsic supports and can be (indeed
has been) overthrown by theoretical counter-evidence, this alone
can hardly settle the question in favour of SC.
What's needed is a dependable method for comparing the
strength of the non-intuitive, non-demonstrative arguments rele-
vant to this case, some of which were touched on in the previous
section. But before we can answer the question of which axiom
candidate is supported by better such arguments, we must face the
prior question of whether these arguments carry any weight at all,
"•' See Chisholm (1977), 40.
 THE CHALLENGE 145

and if so, why. We need to explain how, why, and to what extent
such arguments count as evidence for the truth of their conclusions.
Only then can we determine which among them constitute the
better evidence.
A prerequisite for this inquiry is an appreciation for the rich
variety of extrinsic supports offered by set theorists. The discussion
in this chapter alone reveals a wide range of types:

1. Verifiable consequences. For example, Sierpifiski's demon-

stration that many theorems proved from the axiom of choice can
also be proved, though the proofs are more complicated, without it.
Another case, to which attention was not explicitly drawn, provides
support for measurable cardinals. Martin (1970) derived the
determinacy of Borel sets (indeed 2} sets) from the existence of a
measurable cardinal. Five years later, in his (1975), this result was
'verified', that is, proved from ZFC alone.
2. Powerful new methods for solving pre-existing open problems.
The axiom of choice settled the open question of whether or not the
reals could be well-ordered. V = L and SC both provide methods
for solving the long-standing open problems in analysis, and V = L
even decides the continuum hypothesis.
3. Simplifying and systematizing theory. The axiom of choice
brings order into the chaos of transfinite arithmetic.
4. Implying previous conjectures. The existence of a measurable
cardinal implies that V =£L, and a supercompact rules out
projective well-orderings of the reals. Both these had been
previously conjectured.
5. Implying 'natural' results. The zigzag pattern of separation
properties in the projective hierarchy generated by projective
determinacy (and hence by SC) is considered more natural than the
Fl-side pattern of V = L.
6. Strong intertheoretic connections. The detailed intertheoretic
connections between determinacy and measurable cardinals, be-
yond Martin's result on 2} determinacy, are too complicated to
summarize here, but a simpler sort of example involves the
extension of known patterns from one theory to the next. For
example, in ZFC, every uncountable 2} set has a perfect subset; in
the presence of a measurable cardinal, this extends to Si sets, and
with projective determinacy, to £3 sets and beyond. Thus the three
theories seem to be pulling in the same direction.
146 AXIOMS

7. Providing new insight into old theorems. Projective deter-

minacy allows many classical properties of 11} sets to be extended
to II1. Along the way, this procedure often provides a new and
simpler proof from ZFC of the classical theorem for Il{.
The range of rules of thumb is hardly less bewildering than this
array of extrinsic justifications. I've mentioned a handful mar-
shalled in favour of large cardinals, but there are many more.114
And even this rough-and-ready classification neglects the role of
conjectures—like the non-existence of projective well-orderings of
the reals—and other judgements of plausibility—like that against
choice-generated oddities at low levels of the projective hierarchy.
So even as description, leaving aside explanation, my account is far
from complete. I'm sorry to say that I won't complete it here; filling
in the details of structure of non-demonstrative, non-intuitive
arguments and evaluating their cogency is a subject for another
book, a book I unfortunately don't know how to write. After
drawing attention to the problem, for now I can do little more than
highlight its importance and encourage a concerted investigation.
Let me conclude with a few words about how this last might best
proceed.

There is an obvious similarity between this project and the central

business of philosophers of science: giving an account of the
confirmation of scientific theories. Indeed the very descriptions
of the styles of extrinsic justification—verifiable consequences,
simplifying and systematizing theory, strong intertheoretic
connections—suggest that the analogy is a powerful one. It would
seem that the compromise Platonist's science/mathematics analogy
stands to gain further detail from these distinctive parallels between
scientific and mathematical modes of theoretical justification.
To this line of thought, some will reply that the analogy is
superficial, that natural scientists, not mathematicians, test their
theories by experiment. The immediate response is that mathema-
ticians do use experiments. Let me quote from Martin: 115
I think that there has been more of what I might call subjecting
determinacy hypotheses [to] experiment than is suggested by what I and
other writers have said in print. An example: the first theorem in
114
See Maddy( 1988*).
m
Personal communication, Sept. 1984.
 THE C H A L L E N G E 147

determinacy I proved was that [the full axiom of determinacy] implies that
every set of degrees of unsolvability contains or is disjoint from what is
now called a 'cone'.116 When I discovered the two-line proof of this, I was
very excited. I was sure that, with a few minutes' thought, I would be able
to find a set of degrees which was a counterexample. Thus I would refute
[the full axiom of determinacy] and surely even [projective determinacy],
and probably even Borel determinacy. I started going through various
simple sets of degrees I knew about, checking each one out. I was surprised
to discover that I could always find—by elementary recursion-theoretic
means—the cone whose existence determinacy predicted . . . the effect on
me was much as that on a physicist when a theory predicts a new kind of
particle and such particles are then observed.
Here a mathematical experiment is undertaken quite explicitly.
Other cases can be culled from material earlier in this chapter; for
example, investigation of the consequences of V = L and SC for
low-level projective sets can be viewed as tests.
But, opponents of the analogy might continue, these are
experiments of a very different sort from those found in physics; no
accelerators are involved, no observations of instruments or
computer outputs, no cloud chambers, etc. Of course, this can
hardly be denied, but what needs to be appreciated here is the
extent to which scientific methodology varies even from one natural
science to another. Martin's experiments may use paper and pencil
and depend on previous results in recursion theory rather than
using an electron microscope and depending on previous results in
subatomic physics, but the botanist's experiments are different
from both: in another era, she took a field trip and brought back
hand drawings for comparison with previously gathered and
classified samples. We don't expect a study of the methodology
appropriate to physics to tell us all we want to know about how
botanists, biologists, psychologists, astronomers, or geologists
formulate and test their theories, so why should we expect
mathematical science to conform to confirmation techniques drawn
from some other science? The answer is that we shouldn't. If Martin's
experiments are different from the physicist's, this should come as no
surprise and shouldn't (by itself) count against their efficacy.
Respect for the variation between the sciences also undercuts the
opposite, overly quick, reaction to the analogy between mathem-
atics and natural science. Some, citing the similarities noted above,
116
This theorem is the crucial lemma in Martin (1968).
148 AXIOMS

might be inclined to conclude that the epistemological issues for

this aspect of mathematical knowledge can be identified with the
corresponding problems in the philosophy of science in general.
The motive for such a move would be much like that of the logicists
decades earlier; the problem of mathematical epistemology is
reduced to another—the epistemology of logic or the epistemology
of science in general—which is presumed to be an easier target.
What I've been suggesting is that this won't work either, that the
idiosyncrasies of mathematical theorizing require individual atten-
tion. A complete theory of the methods of physics (or psychology or
biology), even if there were such a thing, would not be enough.
So the theory of mathematical theory formation and confir-
mation we're after will exploit parallels with the various natural
sciences while attending to the unique aspects of mathematical
methodology. And, if it is to do the ambitious job set for it here, it
cannot rest content with pure description. When science is
functioning smoothly, it may be enough to describe its methods,
but in the case of contemporary set theory, even the practitioners
aren't sure how the various non-demonstrative arguments should
be evaluated. Thus Godel writes that the recalcitrance of the
continuum problem
may be due to purely mathematical difficulties; it seems, however . . . that
there are also deeper reasons involved and that a complete solution . . . can
be obtained only by a more profound analysis (than mathematics is
accustomed to giving) of the meanings of the terms . . . and of the axioms
underlying their use. (Godel (1947/64), 473)

I suggest that the 'more profound analysis' required is the very

investigation I'm urging here. What's needed is not just a
description of non-demonstrative arguments, but an account of
why and when they are reliable, an account that should help set
theorists make a rational choice between competing axiom candi-
dates.
Finally, this fundamental problem, the problem of describing,
explaining, and evaluating non-demonstrative arguments in math-
ematics, is a central challenge for the set theoretic realist, but it is
not hers alone. Of course, any compromise Platonist will face it, but
so will others. Anyone who holds that there is (most likely) a fact of
the matter about the size of the continuum (or the open questions of
Luzin and Suslin) must admit that one or the other of V = L and SC
 THE CHALLENGE 149

is false. This leaves the problem of adjudicating between these two

theories, and unless some alterative means is found, this in turn
requires confronting and evaluating the non-demonstrative argu-
ments for and against them. Actually, the range of Platonist and
anti-Platonist philosophical positions for which this challenge
remains a real one is quite broad, as will become clearer in my final
chapter, but for all that, it has been almost universally ignored. The
goal of this chapter will have been served if my portrait of the
challenge itself is vivid and compelling enough that the reader now
sees it as her own.
 5
MONISM AND BEYOND

The main outlines of set theoretic realism are now in place. Its
epistemology divides loosely into two parts, as befits a version of
compromise Platonism: the intuitive (Chapter 2) and the theoretical
(Chapter 4). Benacerraf's ontological puzzle is met by agreeing that
numbers aren't objects, but insisting none the less on a close
connection between numbers and sets, namely, that numbers are
properties of sets (Chapter 3). Finally, though it is far from solved,
the most serious open problem has at least been formulated with
some care (Chapter 4).
Only a few bits of tidying-up remain for this closing chapter.
First comes a final look at the ontology of set theoretic realism, this
time with closer attention to the relationship between the math-
ematical and the physical. This is followed by two sections
comparing and contrasting set theoretic realism with other con-
temporary positions, with an eye to illuminating some surprising
convergences. I conclude, in section 4, with a summary and a look
to the future.

1. Monism

Let me begin with a forceful objection Chihara once aimed at the

very heart of set theoretic realism, that is, at my claim that 'sets of
physical objects . . . have location in time and space and can be
literally perceived by the senses'.1 In the course of lampooning this
position, he writes:
Imagine that I am sitting at my desk. Its surface has been cleared of
everything except an apple. Now according to Maddy, we can literally
perceive on the desk, in addition to the apple, the set of apples on my desk
(which happens to be a unit set). It is claimed that this set has a location in
space, the exact spot where the apple is. Supposedly, it also came into
1
Chihara (1982), 223.
 MONISM 151

existence at a particular time (when the apple did), and will go out of
existence at a particular time (when the apple does). Obviously, Maddy
thinks this set can be moved about in space. Now if we can perceive this set
with the sense of sight, then what does it look like? Evidently, it looks
exactly like the apple itself. After all, I don't see anything at that exact
region in space that looks different from the apple. One wonders how this
object is to be distinguished (perceptually) from the apple, since it has
exactly the same shape and color. Perhaps it feels different. Let's touch it.
But I can't feel anything there other than the apple. Evidently, this strange
entity feels no different from the apple. How about its smell or taste?
Again, it would seem that the set must be identical in smell and taste to the
apple. So it looks, feels, smells, and tastes exactly like the apple and is
located in exactly the same spot and at exactly the same time—yet it is a
distinct entity! One would think that an entity with these properties would
be of interest to the physicist. Furthermore, essentially the same reasons
Maddy gives for maintaining that these sets can be perceived by the senses
can also be given for claiming that a set of such sets can be perceived. Thus,
we should be able to see the set whose only member is the unit set described
above; we should be able to perceive the unordered pair consisting of the
apple and the above unit set, and so on indefinitely. Presumably, all these
different entities would look, feel, smell, and taste exactly alike. (Chihara
(1982), 223^4)
Chihara's only explicit conclusion is the undoubtedly sound one
that my view differs from Godel's,2 but the sense that this passage
raises an important question for the set theoretic realist can hardly
be avoided.
On the face of it, the question is one already considered in the
second section of Chapter 2: how is it, on a given occasion, that we
see a physical mass rather than a set, or one set rather than another,
when all produce the same retinal stimulation? The answer there
was that differences in training, or interests, or attention, could
produce different cell-assemblies in different individuals, or facili-
tate the activation of one cell-assembly rather than another within a
single individual. And the activation of different cell-assemblies,
even given the same retinal stimulation, produces a difference in the
purely phenomenological look of the scene.
In Chihara's case, then, the difference between the physical mass
that makes up the apple and the singleton containing the apple is
that the latter has an unambiguous number property—one—while
the former is one apple, many cells, more molecules, even more
2
See ch. 2, sect, 4, above.
152 M O N I S M AND B E Y O N D

atoms, and so on. What makes the example unsettling is that in this
case, the singleton is so conspicuous that we rarely see the physical
mass or any of the other sets. A topologist, heroically immersed in
her work, might see left and right apple-halves, but it seems
unlikely that even an infant would see a physical mass undifferen-
tiated as a unit from its background—that is, something with no
number property at all—and the normal adult almost invariably
sees the single apple. Indeed the very question of how the physical
mass differs from the singleton can be asked in such a way as to beg
it: what's the difference between a single object and its unit set? A
'single object' already has an unambiguous number property! This
doesn't mean that the physical mass doesn't differ from the
singleton—it does—but once the physical mass is individuated,
separated from its surroundings and seen as an isolated thing, that
difference seems to evaporate.
So, while the set theoretic realist has a ready answer to one
question—what distinguishes a physical mass from a set?—
Chihara is asking another—what distinguishes an individuated
physical object from its unit set? The answer to this new question
cannot be that one has an unambiguous number property and the
other doesn't, because both the single object and the singleton have
the same number property: one. If there is a difference, it must lie
somewhere else, and Chihara's remarks pointedly suggest that it is
not perceptual.
The set theoretic realist's first option, in response to this
situation, is to insist that there is an unperceivable difference. It
wouldn't be the first such difference; we aren't very good at seeing
the difference between gold and fool's gold, and that between water
and heavy water is completely invisible. In science, unperceivable
differences are detected by more sensitive instruments or implied by
well-supported theory. Perhaps theoretical arguments could be
found in mathematics, or more likely, in our theory of mathemat-
ics, in the philosophy of mathematics, for distinguishing individual
things from their unit sets. These might have to do with inviolable
differences between concrete and abstract, or between mathemati-
cal and physical.
The set theoretic realist's other option is simply to deny that there
is any such difference at all, perceivable or otherwise, that is, to
identify individuals with their singletons. This is not to say that
every singleton is identical with its sole member; there is every
 MONISM 153

reason to distinguish between {{0, 1, 2, 3,. . .}} and {0, 1, 2,

3,. . .}, starting with the fact that one is finite and the other
infinite. Rather, we take it that the physical objects, x, the
individuals from which the generation of the iterative hierarchy
begins, are such that x = {x}. After that, the axiom of extension-
ality3 guarantees that sets formed at later stages will be distinct
from their singletons. And, again, this option does not suggest that
the physical mass of apple-stuff is identical with the singleton apple.
Here there is a real difference in the determinacy of number
property. All that's being denied is that the individual apple is
distinct from its unit set.
I think both these options are open to the set theoretic realist,
that is to say, both are fully consistent with the tenets of that
position as described in previous chapters. And neither of them, as
far as I can see, does any damage of the sort implied by Chihara's
rhetoric. In particular, neither option gives up the claim to a real,
perceptual difference between a set and an undifferentiated physical
mass, between three apples and the unindividuated stuff that makes
them up. Godel, considering an argument from Russell, remarks:
Russell adduces . . . against the extensional view of [sets] . . . the existence
o f . . . the unit [sets], which would have to be identical with their single
elements. But it seems to me that these arguments could, if anything, at
most prove that. . . the unit [sets] (as distinct from their only element) are
fictions . . . not that all [sets] are fictions. (Godel (1944), 459)
The same reply could be given to Chihara.
But even if both options are open, even if a set theoretic realist is
free to follow either, my own preference is for the second. The only
motivation I see for insisting on an unperceivable difference is the
desire to maintain a strict dualism between the mathematical and
the physical, and I feel no such desire. The remainder of this section
will thus presuppose the identification of individual with singleton,
but let me insist, one last time, that a set theoretic realist reluctant
to take this turn should feel no obligation to do so.4
3
Two sets are the same if and only if they have the same members. See
Enderton(1977),2, 17.
4
Quine (1969a), §4) advocates the identification of object with singleton for
the purpose of simplifying his formal theory. So doing requires a very minor
modification of Zermelo-Fraenkel set theory and the additional assumption that the
individuals, the elements from which the iterative hierarchy begins, form a set with
at least two members. See my Thysicalistic Platonism' (forthcoming), especially the
appendix.
154 M O N I S M AND B E Y O N D

While Chihara's example is aimed at one specific aspect of set

theoretic realism, other anti-Platonists may be bothered by a more
general worry arising from versions of physicalism. 5 Physicalism
began in the hands of the positivists as a very strong thesis about
the reducibility of all sciences to the vocabulary of fundamental
physics.6 In this form, it foundered on the methodological
independence of the special sciences/ but the crude idea that
physics is somehow basic retains its appeal. A sophisticated version
appears in the writings of the contemporary physicalist Hartry
Field.
Field describes the doctrine as a well-supported methodological
principle:
. . . physicalism [is] the doctrine that chemical facts, biological facts, psycho-
logical facts, and semantical facts, are all explicable (in principle) in
terms of physical facts. The doctrine of physicalism functions as a high-
level empirical hypothesis, a hypothesis that no small number of
experiments can force us to give up. It functions, in other words, in much
the same way as the doctrine of mechanism (that all facts are explicable in
terms of mechanical facts) once functioned . . . (Field (1972), 357)

Mechanism was rejected when Maxwell's theory of electromagnet-

ism could not be explained in mechanical terms. Field concludes
that
Mechanism has been empirically refuted; its heir is physicalism, which
allows as 'basic' not only facts about mechanics, but facts about other
branches of physics as well. I believe that physicists a hundred years ago
were justified in accepting mechanism, and that, similarly, physicalism
should be accepted until we have convincing evidence that there is a realm
of phenomena it leaves out of account. (Field (1972), 357)

He gives this example of how the physicalistic principle functions:

Suppose, for instance, that a certain woman has two sons, one hemophilic
and one not. Then, according to standard genetic accounts of hemophilia,
the ovum from which one of these sons was produced must have contained
a gene for hemophilia, and the ovum from which the other son was
produced must not have contained such a gene. But now the doctrine of
physicalism tells us that there must have been a physical difference between
5
I mentioned this possibility in ch. 2, sect. 1, above but postponed discussion
until now.
6
See Carnap (1934).
7
See e.g. Fodor (1975), 9-26.
 MONISM 155

the two ova that explains why the first son had hemophilia and the second
one didn't... We should not rest content with a special biological predicate
'has-a-hemophilic-gene'—rather we should look for non-biological facts
(chemical facts; and ultimately, physical facts) that underlie the correct
application of this predicate, (Field (1972), 358)
Field does not require that the biological predicate 'has a
haemophilic gene' be translated into the basic vocabulary of physics
or even chemistry, or that all biological laws be derivable from the
laws of physics or chemistry, or that any rational biological
methodology be identical with that of physics or chemistry. He only
insists that there be a chemical and ultimately physical explanation
of why this particular ovum has a haemophilia gene and that one
doesn't.8
Now presumably a physical explanation is one that involves only
physical things, physical facts. If physical things and physical facts
are just those spoken about in physics, then the indispensability
arguments make it hard to see why physicalism presents a problem
for any version of Platonism; according to Quine and Putnam,
physics speaks constantly and essentially of things mathematical. If
being part of physics were all there were to being physicalistically
acceptable, mathematical things would get into the physicalist's
ontology on the ground floor, even ahead of the chemical, the
geological, the astronomical, the biological, and so on.
Obviously, for those who take physicalism to raise a problem for
Platonism, 'being physical' comes to more than 'being mentioned in
physics'. To see what this something more might be, consider once
again Field's version of the epistemological problem for Platonism:
what raises the really serious epistemological problems is not merely the
postulation of causally inaccessible entities; rather, it is the postulation of
entities that are causally inaccessible and can't fall within our field of vision
and do not bear any other physical relation to us that could possibly
explain how we can have reliable information about them. (Field (1982),
69)
In contrasting mathematical entities with his own space-time
regions, Field clarifies the 'physical relations' he has in mind:
there are quite unproblematic physical relations, viz., spatial relations,
between ourselves and space-time regions, and this gives us epistemological
8
In his 'Physicalism' (forthcoming), Field differentiates his view from various
stronger and weaker versions of physicalism.
156 M O N I S M AND B E Y O N D

access to space-time regions. For instance, because of their spatial relations

to us, certain space-time regions can fall within our field of vision. (Field
(1982), 68)

To be epistemologically accessible, to be acceptable on Field's

world-view, an entity need not be causally efficacious, but it must at
least be spatio-temporally located. Assuming that acceptable
entities, for Field, are physical, this suggests that what it takes to be
physical, over and above being talked about in physics, is location
in space and time, and preferably in the causal nexus.9
On this reading, physicalism creates obvious problems for
traditional Platonism, with its acausal, non-spatio-temporal ob-
jects. The set theoretic realist, by contrast, begins in a much
stronger position: her sets of medium-sized physical objects can
'fall within our field of vision'. Of course, not all sets actually do
fall within our field of vision, but, to paraphrase Field, 'this raises
no more epistemological problems for [sets] than it raises for, say,
tigers' (p. 68). If all that's required for physicalistic acceptability is
spatio-temporal location, the set theoretic realist's impure sets are
unimpeachable.
But what about pure sets? What about the empty set, the set of
von Neumann ordinals, and its power set? Those unafflicted by
physicalistic scruples are free to respond that we gain knowledge of
the pure sets by theoretical inference from our elementary
perceptual and intuitive knowledge of impure sets, but physicalists
will complain that this reply does nothing to solve the problem of
spatio-temporal location. Does this mean that such a physicalist
must reject set theoretic realism after all? This follows only if set
theoretic realism is irrevocably committed to pure sets, and I
contend that it is not.
In fact, the pure sets aren't really needed. The set theoretic realist
who would simultaneously embrace physicalism can take the
subject matter of set theoretic science to be the radically impure
hierarchy generated from the set of physical individuals by the
usual power set operation, except that the empty set is omitted at
each stage. On this picture, each set, no matter how exalted in rank,
is located where the physical stuff in its transitive closure is located.
The theory of this structure differs only trivially from that of the
9
Cf. Armstrong (1977).
 MONISM 157

usual hierarchy with individuals.10 It can serve all the same

purposes.
So the set theoretic realist can locate all the sets she needs in space
and time. Still, a more physicalistically satisfying ontology would
be not only spatio-temporally located, but causally efficacious as
well. But notice: if sets are indeed perceivable, as I've argued, then
they must play the same role in the generation of my perceptual
beliefs about them as, say, my hand plays in the generation of my
perceptual belief that there is a hand before me when I look at it in
good light, a role which is, presumably, causal. Or, to use a non-
psychological example, suppose you deposit three quarters in a
soft-drink machine and a soda drops out. Which properties of that
which you deposited are causally responsible for the emergence of
the Pepsi? Well, the weight of the physical mass of metal, its shape,
and also the number property: three. (The machine counts
somehow.) From the set theoretic realist's perspective, that which
has a number property, that is to say, a set, is causally efficacious.
I conclude that set theoretic realism is consistent with physical-
ism. Once again, this is not to say that every set theoretic realist
must be a physicalist; non-physicalists may prefer to retain the
standard version of set theoretic realism with its pure sets. My
point is that the position can be physicalized without significant
trauma.

In this section, I've suggested two minor alterations in the set

theoretic realist's ontology: the identification of physical objects
with their singletons and the elimination of pure sets. Together,
these moves produce a powerfully symbiotic picture of the
relationship between the mathematical and the physical: every
physical thing is already mathematical, and every mathematical
thing is based in the physical. In place of the customary dualism of

10
With two individuals, x and y, a version of the ordinals can be constructed
without pure sets—x, {x, y}, {x, y, {x, y] }, and so on—and the various axioms and
theorems can be tinkered with. Practically speaking, however, it is probably best to
keep the empty set as a notational convenience. Godel, for example, suggests it could
be treated as a fiction 'introduced to simplify the calculus like points at infinity in
geometry' (1944, p. 459). Fraenkel, Bar-Hillel, and Levy say the empty set is
introduced for 'reasons of convenience and simplicity, and can be regarded as a mere
notational convention' (1973, p. 24), and even Zermelo calls it 'fictitious' (19086,
p. 202). See my 'Physicalistic Platonism' (forthcoming), appendix.
158 M O N I S M AND B E Y O N D

mathematical and physical, this pared-down set theoretic realism

offers a version of monism.11
To appreciate just how closely the two are intertwined on this
view, try to separate them. A purely mathematical world would be
empty. What would a purely physical world be like? As soon as
there are number properties, there are sets that bear them, so a
world without mathematical things would have to be a world
without any things, a completely amorphous mass: the Blob. To
add even the structuring into individual physical objects is to admit
singletons, to broach the mathematical. The only way to confine
ourselves to the purely physical is to refrain from any differen-
tiation whatsoever.
Perhaps such a world is possible, but it clearly isn't our world,
with its objects, kinds, patterns, and structures of so many, widely
varied sorts. In place of the old picture—physical reality here and
now, mathematical reality nowhere and nowhen—set theoretic
monism offers a spatio-temporal reality inseparably physical and
mathematical. Physics and mathematics, on this new picture, are
two sciences, along with chemistry, biology, psychology, and the
rest, that study aspects of this reality. Each science has its own
vocabulary and laws, its own techniques and methods, but this
doesn't mean that the world itself is divided into the physical, the
mathematical, the chemical, the biological, the psychological, and
so on. Rather, everything is ultimately physico-mathematical or
mathematico-physical.
Some will note that, strictly speaking, this view is more
Aristotelian than Platonistic. They are right, in the sense that
Aristotle's forms depend on physical instantiations, while Plato's
are transcendent. I retain the term 'Platonism' here, not for its
allusion to Plato, but because it has become standard in the
philosophy of mathematics for any position that includes the
objective existence of mathematical entities. Others will point out
that singletons don't deserve the special status my presentation has
awarded them; for the monist, two objects (as opposed to the
undifferentiated mass of physical stuff that makes them up) are
already a doubleton, as well. This is also correct. The singleton case
is unique only in its psychological impact: it makes us realize just
how little it takes for mathematics to intrude. But for all this, I hope
11
For those who think of Platonism as a form of theology, this is a version of
pantheism.
 FIELD'S NOMINALISM 159
the general features of set theoretic monism are clear enough. I
leave it as an option for any set theoretic realist to whom it might
appeal.

2. Field's nominalism

Despite my efforts to turn away objections to Platonism based on

epistemology, on the possibility of multiple reductions, and on
purely physicalistic concerns, Field remains unconvinced. Instead of
seeking out new objections to counter, a more fruitful strategy at
this stage might be to have a look at Field's nominalistic alternative
to Platonism. Field is not the only nominalist on the contemporary
scene, but his version has the distinction of being non-revisionist:12
he sets out to show how we can go on using classical mathematics
in science exactly as we have been, but without admitting the
existence of mathematical things, and he attempts this without
short-changing the Quine/Putnam indispensability arguments. Here
is a form of nominalism that should give the Platonist pause.
Field and I agree that the indispensability arguments provide the
best evidence for mathematics as a whole. Moving beyond Quine/
Putnamism into Godelian territory, we also agree that there are
other possible forms of mathematical evidence:
if we assume that there is at least one body of pure mathematical assertions
that includes existential claims and that is true . . . then we are assuming
that there are mathematical entities. From this we can conclude that there
must be some body of facts about these entities, and that not all facts about
these entities are likely to be relevant to known applications to the physical
world; it is then plausible to argue that considerations other than
applications to the physical world, for example, considerations of
simplicity and coherence within mathematics, are grounds for accepting
some proposed mathematics axioms as true and rejecting others as false.
(Field (1980), 4)
Finally, we agree that the Benacerraf-style worry is a real one, that
the Platonist owes a descriptive and explanatory account of our
knowledge of (or reliability about) mathematical facts, and I can
12
Recall from ch. 1, sect. 4, that the nominalist claims there are no mathematical
entities. Chihara (1973, ch. 5) is also a nominalist, but he proposes an abbreviated
and reinterpreted mathematics that may or may not be adequate for scientific
purposes.
160 M O N I S M AND B E Y O N D

extend this accord by accepting for now the further stipulation that
this account must satisfy the physicalist.
Against this shared backdrop, two possible strategies stand out:
take mathematical statements to be (mostly) true and meet the
epistemological challenge head on, or take mathematical state-
ments (at least the existential ones) to be false and explain why
these falsehoods are so useful in applications. Field chooses the
second, which requires him to somehow circumvent the indispens-
ability arguments. What I want to suggest here is that this arduous
undertaking does not win him the advantages he hopes for, indeed
that it doesn't exempt him from what I take to be the most serious
challenge to Platonism.

The best approach to Field's methods is to return to the thinking of

the traditional Platonist: mathematics, if true, is necessarily true,
that is, it is true regardless of the contingent details of the physical
world, true in all possible worlds. Let M be our necessarily true
mathematical theory. Now suppose that N is a consistent theory of
what the physical world might be like, and further suppose that N
is nominalistic, that it makes no reference to mathematical entities.
Then N 4- M must also be consistent; otherwise, the truth of N
would imply the falsehood of M, and M couldn't be true in a
possible world where N is true, contradicting the necessity of M.
So, if M is our true mathematical theory, it must be consistent with
any consistent nominalistic theory N.
Now suppose that A is a nominalistic statement implied by N +
M. Then N + not-A 4- M is inconsistent. But we've just finished
arguing that M must be consistent with any nominalistic theory
that is itself consistent, so it follows that N + not-A is also
inconsistent. By elementary logic, this means that N implies A. So
we've shown, from the perspective of the traditional Platonist, that
if a nominalistic statement follows from a nominalistic theory plus
mathematics, then that same nominalistic statement follows from
the nominalistic theory alone. In technical terms, mathematics is
conservative over nominalistic physical science.13

The traditional Platonist's argument doesn't work for the set theoretic realist
because she doesn't take mathematics to be necessary (see ch. 2, sect. 2, above). The
set theoretic monist also denies that there is such a thing as nominalized physics,
because physical objects are already mathematical, which severely undercuts the
potential significance of any conservativeness claim.
 FIELD'S NOMINALISM 161
Of course, as a nominalist, Field rejects the assumption that
mathematics is true at all, let alone necessarily true, but he agrees
with the conclusion that good mathematical theories should be
conservative:
it would be extremely surprising if it were to be discovered that standard
mathematics implied that there are at least 106 non-mathematical objects
in the universe, or that the Paris Commune was defeated; and were such a
discovery to be made, all but the most unregenerate rationalist would take
this as showing that standard mathematics needed revision. Good
mathematics is conservative; a discovery that accepted mathematics isn't
conservative would be a discovery that it isn't good. (Field (1980), 13)
This conservativeness is a boon to the nominalist:
even someone who doesn't believe in mathematical entities is free to use
mathematical existence-assertions in a certain limited context: he can use
them freely in deducing nominalistically-stated consequences from nominal-
istically-stated premises. And he can do this not because he thinks those
intervening premises are true, but because he knows that they preserve
truth among nominalistically-stated claims. (Field (1980), 14)
So this is the beginning of Field's answer to the indispensability
argument: he admits that mathematics is used in science to derive
physical claims from other physical claims, but insists that we can
believe these results without believing the mathematics employed to
be true. We need only believe it is conservative, that whatever it
implies is already implied by the physical theory itself.
But this is only part of the story, for the role of mathematics in
deriving one physical statement from another is only one of its roles
in physical science. As Putnam has argued, many of the physical
statements themselves make essential appeal to mathematical
entities. In other words, the nominalistically stated physical claims
discussed so far don't cover most of physical science. Field is fully
sensitive to this point. He concludes only that 'once such a
nominalistic axiom system is available, the nominalist is free to use
any mathematics he likes for deducing consequences, as long as the
mathematics he uses [is conservative]' (Field (1980), 14). So the
answer to the indispensability arguments comes in two parts. First
it must be shown that physical theories can be stated without the
use of mathematics, then classical mathematics must be shown to
be conservative over those restated physical theories. That ac-
complished, the scientist can use whatever mathematics she likes in
162 M O N I S M AND BEYOND

deriving nominalistic consequences from nominalistic theory,

because any such consequence derived using mathematics is already
implied by the nominalistic theory alone.
For concreteness, let me rehearse a simplified but I hope
illustrative example of this strategy. We ordinarily use real numbers
to measure distances in our theory of space. Thus, by a standard
indispensability argument, any confirmation of our theory of space
also confirms the existence of real numbers. But, Field argues, the
use of the reals in this context is actually dispensable after all. To
show this, he must first reformulate our standard theory of space
without talk of distances. There is a way to do this, codified by
Hilbert; 14 rather than assigning locations and distances to points, it
makes use of comparative predicates like 'between' and 'con-
gruent'. This theory is nominalistic because it deals only with points
and regions of space and not with numbers. Call it H.
Now suppose we'd like to establish some nominalistic claim A
about space. To apply real number theory, we first move to a larger
theory, S, that combines H with some set theory. In S, we can prove
that there is a function from pairs of points to real numbers 15 that
does all the right things: for example, the segment between x and y
is congruent to the segment between x' and y' in H if and only if the
real number assigned to (x, y) is the same as the real number
assigned to ( x ' , y ' ) . In fact, we can show that the space itself is
isomorphic to the set of ordered triples of real numbers. In this rich
context, we translate A into an equivalent statement A' that talks
about distance and hence about real numbers, and we proceed to
prove A'. Because A and A' are equivalent, this also establishes A in
the theory S. But S, being a good mathematical theory, is
conservative over H, so A is also implied by H alone. And that was
what we wanted to show in the first place. But we needn't assume
the truth of S to do it, only its conservativeness. 16
Field extends this technique to cover applications of classical
14
See Hilbert (1899).
15
Whichever version of the reals we select for this measuring job.
16
Students of Field's theory will realize that I'm confining my attention to the
second-order version of his view. (See Shapiro (1983fc) and Field (1985) for
discussion of this distinction.) I do this because it is the second-order version that
offers the full use of classical mathematics without an ontology of mathematical
entities—the first-order version offers something less—and because I think our
understanding of space (and other mathematical notions) is essentially second-order.
(See Shapiro (1985) on this last point.) For an introductory account of the
differences between first- and second-order logic, see Enderton (19 7 2), ch. 4.
 FIELD'S NOMINALISM 163
mathematics in Newton's gravitational theory, but some comment-
ators17 doubt that it can be adapted to other parts of physics, in
particular to quantum mechanics. Because his efforts to date are (at
best) only partial, Field admits that the indispensability arguments
retain some force:
At present of course we do not know in detail how to eliminate
mathematical entities from every scientific explanation we accept; con-
sequently, I think that our inductive methodology does at present give us
some justification for believing in mathematical entities. But. . . justifi-
cation is not an all or nothing affair. . . . what we must do is make a bet on
how best to achieve a satisfactory overall view of the place of mathematics
in the world, . . . my tentative bet is that we would do better to try to show
that the explanatory role of mathematical entities is not what it
superficially appears to be; and the most convincing way to do that would
be to show that there are some fairly general strategies that can be
employed to purge theories of all reference to mathematical entities. (Field
(1989), 17-18)
Weighing what he sees as the epistemological and ontological
drawbacks of Platonism against the indispensability arguments,
Field wagers that mathematics can be shown to be dispensable,
after all. This is his project.

When this version of nominalism is compared with traditional

Platonism, some observers18 argue that its space-time points and
regions are abstract, and thus as susceptible as numbers to
epistemological challenge. From the physicalistic point of view
sketched in the last section, this can hardly be true. Space-time
points and regions have location, and some such regions 'fall within
our field of vision'. As I've indicated, the same is true of the set
theoretic monist's impure sets, so at this crude level, nominalism
and monism are on a par, and both are preferable to traditional
Platonism.
Like the traditional Platonist, Field is also faced with the
accusation that his entities are causally inert. The argument runs
that it is the objects in space-time, not space-time itself, that are
causally efficacious. Field responds to this charge along two
different lines. First, he suggests that physical objects be identified
17
e.g. Malament (1982).
18
e.g. Resnik(1985a),
164 M O N I S M AND B E Y O N D

with the space-time regions they occupy.19 In this way, at least the
occupied areas of space-time enter into the causal nexus. A stronger
argument involves the claim that
a field theory is most naturally construed as a theory that ascribes causal
properties . . . to space-time points. (Field (1982), 70)
In electromagnetic theory for instance, the behavior of matter is causally
explained by the electromagnetic field values at unoccupied regions of
space-time . . . (Field (1980), 114)
Given the omnipresence of fields, this observation brings causal
powers to all points of space-time. So once again, the nominalist
and the monist are on equal footing.
To further the comparison, we must consider the respective
ontologies more closely. The nominalist's world consists of space-
time regions; ordinary and theoretical physical objects are identi-
fied with the space-time regions they occupy, and points can be
identified with partless regions. The monist's world consists of sets;
physical objects are singletons among these. The contrast between
the two views becomes smaller when we realize that a thorough
account of the monist's discrete physical objects will involve the
study of perceptual continua as well: object boundaries, trajectories
of movement, etc.20 Thus spatial point sets join the monist's
ontology. So both nominalism and monism embrace discrete
objects and continua: for the former, both are species of space-time
regions; for the latter, both are species of sets.
Finally, what of that old point of contention, the numbers? The
nominalist, of course, eschews them. For the monist, as we've
seen,21 the question of the existence of numbers is a special case of
the age-old problem of the existence of universals. Now Field sides
with old-fashioned nominalism against universals as well as his
modern variety against mathematical entities,22 and I see no reason
why the monist's commitment to numbers need be any stronger
than Field's commitment to the properties of his regions. In so far
as old-fashioned nominalism is tenable, the monist can agree with
Field that numbers don't exist.

So far, then, the nominalist and the monist are not as far apart as
19
Field (1982), 70.
20
See ch. 2, sect. 3, above.
21
In ch. 3, sect. 2, above.
22
See Field (1980), 35, 55-6, and (1982), 70.
 FIELD'S NOMINALISM 165
rhetoric would suggest—where the nominalist sees a space-time
region containing the stuff of three apples, the monist sees a set of
three apples—but a dramatic difference soon emerges. These space-
time regions are the end of the ontological story for the nominalist,
but the monist's world, in addition to sets of apples, also contains
sets of sets of apples, sets of sets of sets of apples, and so on.
Physicalistically speaking, these sets of higher rank are no clear
liability; they have location and they can (at least in principle) be
causally efficacious. If physicalism doesn't rule them out, we should
ask what motivates the monist to include them and what the
nominalist hopes to gain by abstaining.
Part of the motivation for an escalation of ranks lies in arithmetic
itself. Two pairs of shoes are naturally viewed as a set of two sets; a
series of sets of ever-increasing rank, analogous to the von
Neumann ordinals, does good service as a measuring device for the
number properties of finite sets; this set theoretic context allows us
to prove the Peano axioms and to provide a simple and explanatory
theory that encompasses and explains various well-entrenched
generalizations, for example that the union of two disjoint two-
membered sets has four members. By contrast, the nominalist's
position here is much like that of the aggregate theorist considered
back in Chapter 2: a statement of number concerns a mass of
physical stuff together with a predicate. On the one hand, it isn't
clear that a smooth and flexible arithmetic can be established on
this basis, but, on the other, I doubt that a truly persuasive case of
the postulation of infinite ranks can be based on arithmetic alone.23
Let me turn, then, to the continua which inhabit both the
nominalist's and the monist's universes. Both theorists hypothesize
that these satisfy the nominalist's Hilbert-style axioms, and, given
that they are physical entities, the nominalist should be as interested
as the monist in answering further questions about them.24 A
number of deep questions can be asked about physical structures of
this complexity,25 and it is the consequences of asking them that I
now want to explore.
23
I consider this possibility at greater length, but no more conclusively, in my
'Physicalistic Platonism' (forthcoming), § 7.
24
Field's remarks about mathematics quoted at the beginning of this section
strongly suggest that as soon as entities are admitted into the nominalist's ontology,
all facts about them are worthy of investigation.
25
Many observers have remarked that these questions include the continuum
hypothesis. See Resnik (19856), 198. I'll consider other questions here, for reasons
that will become obvious.
166 M O N I S M AND B E Y O N D

Recall (from section 4 of Chapter 4) that being determined is a

property of sets of reals. This property can be defined mathemat-
ically in fairly simple terms; its nominalistic counterpart should be
expressible in Field's nominalistic theory.26 Now suppose that our
nominalist observes that a great number of the simple regions he
comes across are in fact determined. As long as he considers
nothing more complicated than the regions corresponding to what
the monist would think of as countable unions and intersections
generated from open sets,27 he will encounter no set that is not
determined. As a scientifically minded inquirer, he will want an
explanation of this fact.
The monist has an explanation for the corresponding fact about
the world of sets in the form of Martin's theorem that all Borel sets
are determined.28 This explanation involves, however, an inescap-
able commitment to infinite ranks; Harvey Friedman has shown
that the theorem requires them.29 So we imagine our scientifically
minded monist hypothesizing the axioms of infinity and replace-
ment, expanding her ontology accordingly, in order to gain an
explanatory theory of the behaviour of Borel sets.30
Where does this leave our nominalist? Presumably, he'd also like
to explain the behaviour of his Borel regions; surely he wants a
theory of his continuum that explains as much as the monist's. It
might be thought that the nominalist will have to break dowyn and
postulate some aggregate-theoretic analogue of the monist's higher
ranks in order to gain a counterpart to the theorem on Borel
regions. But to think this is to forget the role of conservativeness.
Recall that if ZFC, which includes replacement, is conservative over
the nominalist's theory H, then whatever ZFC can prove is already
true in the nominalist's world. In particular, if ZFC is conservative
over H, then all Borel regions are determined, regardless of whether
or not there really are higher ranks of any kind.
So, in order to show that Borel regions are determined, the

26
Saying a set is determined involves quantification over real numbers. Methods
of Shapiro (1983fc) and Resnik (19856) show how to simulate such quantification in
Field's system using space-time points.
27
Field (1980, p. 63) describes the nominalistic version of open sets. Countable
unions and intersections generate the Borel sets.
28
Martin (1975; 1985).
29
Friedman (1971).
30
Of course, this isn't the only reason for assuming infinity or replacement, but
I'm simplifying here. For more, see Maddy (1988a), § 1.8.
 FIELD'S NOMINALISM 167
nominalist need only establish the conservativeness of ZFC over the
theory H. In his book, Field gives a set theoretic proof of this fact in
a theory slightly stronger than ZFC itself, but this is a metamath-
ematical argument in terms of models that is obviously unavailable
to the nominalist. In fact, conservativeness itself is usually defined
metamathematically,31 so it is unclear that there is even a legitimate
nominalistic version of the bare claim that ZFC is conservative.
Field's reply is that the conservativeness of ZFC, nominalistically
stated, comes to a claim about logical possibility, where logical
possibility is a primitive notion not defined, as it is classically, in
metalogical terms. What the nominalist needs to know, then, is
(slightly more than) 32 that ZFC is logically possible in this sense.
How does he know this? Well, presumably the monist also claims
to know it, or something very like it, and Field suggests that
it is no more problematic for a nominalist to claim that a belief that [ZFC is
logically possible] is reasonable than it is for a platonist to make this claim.
(Field (1985), 140)
Thus, for example, both the nominalist and the monist can cite the
fact that no one has yet derived a contradiction from these
axioms.33
This situation can be clarified if we consider one more hypo-
thetical example, returning this time to the concerns of the classical
analysts. Let's see what happens if the nominalist examines his
regions and asks his nominalistic version of the question: is this
region measurable? 34 Once again, the simplest regions he considers
will be measurable. Indeed, this time he can go beyond the Borel
regions to the £j regions and still not encounter anything non-
measurable. But what will happen if he goes further? The
complements of the 2} regions, the FI{ regions, will still be
measurable, but what about their projections, the 21 regions, and
beyond?
Of course, the monist has considered these questions in
Platonistic terms, and we've seen that her theory can answer them
31
M is conservative over N if and only if, for any nominalistic assertion A, if A is
true in all models of M, then A is true in all models of N.
32
See Field (1985), 139^0.
33
See Field (1984), 88,124.
34
Field mentions the possibility of non-measurable regions (1980, p. 144, n. 26).
And again, measurability is a property that can be stated mathematically using only
quantification over reals, and thus, should be expressible in Field's system.
168 M O N I S M AND B E Y O N D

definitively.35 What makes this case different from the last is that
this time the monist has an embarrassment of riches: two
competing theories that answer these questions in different ways;
V = L answers that there is a Aj non-measurable set, while SC
implies that all projective sets are measurable. In other words, as
we've seen, the monist is placed in a position analogous to that of
the natural scientist who must chose between two competing
theories, and the reaction of the set theoretic community is much as
this analogy would suggest: theorists disagree, they offer competing
evidence, and further evidence is sought to decide the matter. I've
suggested that describing and explaining the justificatory power of
this sort of evidence is the main open problem for contemporary
Platonism.
Where does this leave the nominalist? In order to see which of his
regions is measurable, he needn't worry over which of the
platonist's two opposing theories is true; he need only know which
one is conservative.36 In this connection, Field says much what he
did before: 37
any reason that a platonist offers for believing that it is [ZFC + SC] rather
than [ZFC + V = L] that is true [or vice versa] can be taken over by a
nominalist to argue with just as much force that it is [ZFC + SC] rather
than [ZFC + V = L] that is possible. (Field (1985), 140)

In other words, the nominalist is faced with exactly the same

bewildering array of argument and counter-argument, evidence and
counter-evidence, as the monist. Which means that in order to
answer his physical question about the measurability of his regions,
the nominalist must face a carbon copy of the most difficult
epistemological open question that confronts the monist: what
makes these arguments good or bad? Thus, despite his noble effort
to refrain from committing himself to higher ranks, or anything like
them, the nominalist has not thereby saved himself from the
monist's most difficult epistemological challenge.

35
See ch. 4, sect. 4, above.
36
I say 'which one' because at most one can be conservative. This is because part
of conservativeness is second-order semantic consistency, and all standard models of
set theory agree on the structure of the reals and their subsets. Of course, both
theories might fail to be conservative.
37
Field actually speaks, not of ZFC + V = L and ZFC + SC, but in general, of
two set theories M and M* which are related as these two are.
 FIELD'S NOMINALISM 169
Two conclusions follow. The first involves Field's wager that
dispensing with mathematics in science will produce a better overall
theory than Platonism can provide. I've argued that physicalism
gives us no reason to prefer Field's nominalism to the monistic
version of set theoretic realism. Further, the elementary epistem-
ology of monism and this nominalism are comparable; at the most
fundamental level, sets or regions 'fall within our field of vision'.
Finally, both the monist and the nominalist are faced with a
difficult epistemological question at the theoretical level. I conclude
that Platonism need be no more problematic than Field's nominal-
ism, and thus, that there is nothing to balance the disadvantages of
nominalism—e.g. the need to rewrite science—when the overall
merits of the two theories are compared. On these terms, monism
should prevail.
Before turning to my second conclusion, let me pause to draw a
moral from this one. We've seen that the mathematical theory of
sets has its metaphysical as well as its historical roots in the theory
of the continuum, in the calculus and higher analysis. The Platonist
takes the ontology of this theory at face value. The nominalist tries
to abstain from this as a way of avoiding some difficult philosoph-
ical questions. I think the moral of this discussion is that anyone,
including the nominalist, who embraces the full continuum and
analysis in some form or another, will end up, sooner or later,
meeting those same difficult philosophical problems, perhaps
lightly disguised, but stubbornly undiminished. This lesson falls
neatly in line with the realistic urgings of Frege, Quine, and
Putnam. 38
My second conclusion transcends the pro-Platonist propaganda
that is the main theme of this book. Towards the end of Chapter 4,
I suggested that set theoretic realism in particular, and compromise
Platonism in general, are not the only positions faced with the
difficult problem of assessing and explaining the rationality of the
non-demonstrative arguments for and against V = L and SC. The
discussion in this section shows, quite surprisingly, that Field's
nominalism provides a fresh example; in order to answer what he
counts as physical questions, questions about space-time-regions,
he must deal with this same open problem. A problem common to
nominalist and Platonist is likely to be an extremely fundamental
38
See ch. 1, sect. 4, above.
170 M O N I S M AND B E Y O N D

one, deserving the attention of philosophers from a wide range of

persuasions. We'll meet with two more in the next section.

3. Structuralism

Before closing, I'd like to touch briefly on one other conspicuous

position in the philosophy of mathematics of recent years, namely
structuralism. While lip-service to the general idea behind this view
is fairly common—mathematics is about structures, not objects—
there is no complete and definitive statement of structuralist
orthodoxy comparable to Field's writings on his own version of
nominalism. Michael Resnik and Stewart Shapiro offer the most
comprehensive contemporary statements, 39 but this work is still in
progress, so a thorough comparison between structuralism and set
theoretic realism will have to wait till another day. My more
modest goal here is simply to sketch the main outlines of a
structural approach to mathematics and to suggest that it differs
less from set theoretic realism than partisan rhetoric would
indicate.

Though structuralist thinking goes back at least to Dedekind, 40

modern versions are inspired by considerations akin to those in
Chapter 3 above. The Platonist claims that mathematics is about
objects, but, as Benacerrafian meditation on multiple reductions
indicates, we seem to know nothing about these objects other than
that they are related to one another in certain ways. If mathematical
objects have distinguishing features over and above these, those
properties are hidden and presumably unimportant to the math-
ematician. How, for example, are we to say which particular
objects are the natural numbers?
39
See Resnik (1975; 1981; 1982), Shapiro (1983<z; 1985; forthcoming). (Parsons
(forthcoming) is another valuable resource.) One major difference between these
two is Resnik's preference for first-order formulations, e.g. number theory is the
study of any structure—there are many—that satisfies the first-order Peano axioms.
This smacks to me of if-thenism. Shapiro, by contrast, thinks of arithmetic as the
study of a single particular structure, the one specified by the second-order axioms.
Here I side with Shapiro. My thanks go to both writers for their help with this
section.
40
See Parsons (forthcoming), §3. On the recurrence of structuralism, Parsons
remarks, 'its tendency to be revived after attempts run into serious difficulties shows
that the ontologica! intuition behind it exerts a powerful attraction' (p. 8).
 STRUCTURALISM 171

The structuralist replies by rejecting the presuppositions of the

question:
In mathematics . . . we do not have objects with an 'internal' composition
arranged in structures, we have only structures. The objects of mathemat-
ics, that is, the entities which our mathematical constants and quantifiers
denote, are structureless points or positions in structures. As positions in
structures, they have no identity or features outside of a structure. (Resnik
(1981), 530. See also Shapiro (1983*), 534)
Arithmetic, for example, is not the study of certain objects, the
numbers, but the study of the natural number structure, an endless
sequence of featureless positions satisfying certain conditions. One
instantiation of that structure is the von Neumann ordinals,
another is the Zermelo ordinals. But sets themselves are also
positions in a structure, so the multiple reductions of number
theory to set theory just show that the natural number structure
occurs many times within the set theoretic hierarchy structure.
Some structures are physically instantiated: for example, the
substructure of the natural number structure consisting of its first
three positions is instantiated by the apples on the table. On the
other hand, many patterns of higher mathematics—e.g. the
iterative hierarchy structure—presumably are not physically re-
alized. Between these extremes, some applications of mathematics
in science come to the postulation of enough theoretical physical
entities to exemplify the relevant structure:
the claim that actual space exemplifies the structure of Euclidean geometry
involves an assertion that there is a continuum of space points. . . . science . . .
proceed [s] by discovering mathematical structures exemplified in material
reality, but the discovery is often indirect and involves the postulation of
theoretical entities. (Shapiro (1983a), 540)
In sum: universals are the subject matter of mathematics; some but
not all of these universals are physically instantiated.
Already, certain points of contact between structuralism and set
theoretical realism are obvious: both solve the multiple reductions
problem by exchanging objects for universals.41 Indeed much of
their talk is strikingly similar. In the three-apple case, the set
41
The structures themselves are universals (Shapiro (1983a), 536) or, in Resnik's
terminology, 'patterns'. The positions in these structures count as 'objects' for both,
but notice that these objects are structure-dependent: they have no properties apart
from the relations they bear to other positions in the same structure.
172 M O N I S M AND B E Y O N D

theoretic realist says there is a set on the table that is equinumerous

with {4>, {<)>}, {cj>, {4>}}}. The structuralist says there is a physical
arrangement on the table that instantiates the same pattern as {cf>,
{4>}, {(£>, {4>}}} under the successor relation. Both are claiming that
a physical mass has a certain organization. One calls that
organization forming a set equinumerous with {ct>, {4>}, {cj>, {4>}}},
the other calls it instantiating the same pattern as {4>, {(f>}, {4>,
{cj>}}}. At this point, I think it's fair to wonder if any real
significance attaches to this difference in description, 42 but I won't
undertake to answer that question here.
There are also agreements in epistemological thinking. For the
structuralist, various claims about a pattern of dots on a piece of
paper
are simply obvious to anyone who has sufficient mathematical experience
to understand them and who attends to the diagram. . . . they are in a sense
read off the drawing. So long as we are taking our perceptual faculties for
granted, they need no further justification. . . . [they] continue to hold when
talk of dots is replaced by talk of a sequence of squares, stars, a row of houses,
a stack of coins, etc. . . . These additional assertions are as evident or almost
as evident as the original ones. We have thus arrived at knowledge of an
abstract pattern or structure. (Resnik (1975), 34)

In place of the non-spatio-temporal, causally inert mathematical

entities of traditional Platonism, the structuralist substitutes per-
ceivable arrangements of things.43 The existence of infinite patterns
and facts about them are then justified theoretically:44
If [our theory of the infinite structure] turns out to be highly coherent and
confirmed by our knowledge of the finite patterns from which it arose, then
our belief in the existence of the pattern is justified. (Resnik (1975), 36-7)

42
The parallel is just as striking for real numbers: I say the space-time points
have the property of continuity, which can be detected using various set theoretic
constructions; Shapiro says they exemplify the 'structure of Euclidean geometry'
(Shapiro (1983a), 540), which is also exemplified by various set theoretic
constructions.
43
There has been some evolution in Resnik's thinking here. The account in
Resnik (1975) suggests that we perceive the pattern itself; in Resnik (1982), we see
the physical things and abstract the pattern. In Resnik (forthcoming a), this
abstractionist epistemology is abandoned altogether in favour of a yet-to-be-
developed 'postulational' view. Shapiro (1983a, p. 535) sticks with the abstrac-
tionist mode.
44
Shapiro's account of our knowledge of infinite structures reads somewhat
differently. See Shapiro (forthcoming).
 STRUCTURALISM 173

Thus the structuralist's epistemology parallels the two-tiered

account of the set theoretic realist. At the most elementary level,
both theorists turn to perceptual knowledge—of sets or patterns—
and after that, to theoretical knowledge, justified by its coherence
and its consequences for lower-level theory.

So far, then, the structuralist and the set theoretic realist are in
broad ontological and epistemological agreement: they meet the
problem of multiple reductions of number theory with a move from
numbers as objects to numbers as universals and the epistemologi-
cal problem for traditional Platonism with a two-tiered epistem-
ology of perceptual and theoretical justification. This only covers
the natural numbers, but both advocate the same sort of ex-
change—objects for universals—for the reals, and presumably for
other traditional mathematical objects that can be thought of as
universals, multiply instantiated in the iterative hierarchy. Where
the two part company, then, is in their view of the set theoretic
universe itself. For the set theoretic realist, this 'structure' consists
of real objects, the sets; these are the bedrock, the things that
instantiate the various mathematical universals. For the structural-
ist, it is just one more structure, made up of featureless points in
certain relations.
Though this surely sounds like a substantive disagreement, its
true significance is difficult to assess. To see this, consider the
structuralist's account of the interconnections between branches of
mathematics. In order to explain, for example, how the study of the
natural number structure can be advanced by study of the real
number structure (in analytic number theory) or the iterative
hierarchy structure (in reductions of number theory to set theory),
the structuralist must speak of one structure being 'contained' or
'modelled' in another. For such purposes and others—e.g. for
posing the question of whether or not V = L—the structuralist
must speak of several structures at once and of the relations
between them. This in turn requires an overarching 'structure
theory'.
Of course, set theory can provide such a background theory; all
structures can be taken to be sets (or proper classes of the least
problematic kind), as can the functions and relations between them.
This is the set theoretic realist's position. But the thoroughgoing
structuralist would insist on a yet-to-be-described structure theory
174 M O N I S M AND B E Y O N D

strong enough to encompass all structures, including the iterative

hierarchy structure. The set theoretic partisan might wonder what
such an all-encompassing structure theory would be like, and what
could make it preferable to the more familiar theory of sets, but in
fact, the prior question is: what would make these two theories
different? Shapiro concludes that
In a sense, the theories [set theory and a comprehensive structure theory]
are notational variants of each other. (Shapiro (forthcoming))
If this is so, the purported difference between set theory as pattern
and set theory as bedrock begins to elude us, along with that
between structuralism and set theoretic realism.
Just for the record, I'd like to mention here two considerations
that incline me to resist, at least for now, the characterization of set
theoretic realism as a form of structuralism. The first is an
epistemological disanalogy between arithmetic—a case for which
even the set theoretic realist adopts a structuralist approach—and
set theory—the case still open to debate. Structuralism for the
natural numbers is so appealing partly because our understanding
of arithmetic doesn't depend on which instantiation of the number
structure we choose to study. For the purposes of simple perceptual
access, as Resnik notes, a pattern of dots will do, as will a sequence
of squares, stars, houses, coins, etc. The structuralist might say the
same for set theory, that it matters not whether we begin from an
array of dots, coins, or whatever, as long as they instantiate the
initial stages of the iterative hierarchy pattern. For example, Mark
Steiner, another thinker with strong structuralist tendencies,45
writes:
One imagines or looks at material bodies, and then diverts one's attention
from their concrete spatial arrangement.. . . This is how one might become
familiar with the standard model of ZF set theory—by abstracting from
dots on a blackboard arranged in a certain way. Thus one arrives at an
intuition of the structure of ZF sets. {Steiner (1975tf), 134-5)
But I think it is not, in fact, the properties of such physical arrays
that give us access to the simplest of set theoretic truths. Experience
with any endless row might lead us to think that every number has
45
See Steiner (1975#), 134. As noted above (ch. 3, sect. 1), Steiner's structuralism
is only epistemic: numbers are objects, but the only things worth knowing about
them are their relations to other numbers.
 STRUCTURALISM 175

a successor, but it is experience with sets themselves that produces

the intuitive belief that any two things can be collected into a set or
that a set will have the same number of elements even after it has
been rearranged. In other words, though any instantiation of the
natural number structure can give us access to information about
that structure, our information about the set theoretic hierarchy
structure comes from our experience with one particular instanti-
ation.46 Thus one motivation for the move to structuralism in the
case of number theory is undercut in the case of set theory.
My second source of concern about the assimilation of set
theoretic realism to structuralism arises out of the simple question
of what set theory is about. The set theoretic realist answers that set
theory is the study of the iterative hierarchy with physical objects as
ur-elements; the set theoretic monist takes physical objects them-
selves to be sets and eschews pure sets altogether. The trouble with
these answers from the structuralist perspective is that some of the
'positions', in particular the ur-elements, have properties beyond
those they have solely by virtue of their relations with other
positions in the structure. The purely relational structure arising
from the iterative hierarchy with ur-elements would make no
distinction between the position occupied by this apple and the
position occupied by this orange, between the position occupied by
the set of the apple and the orange and the position occupied by the
set of the apple and this baseball, distinctions the set theoretic
realist will certainly want to preserve. Thus the structuralist bent on
assimilating set theoretic realism by claiming that the iterative
hierarchy with ur-elements is itself a purely relational structure will
have to move to a larger, containing pattern, from whose point of
view the baseball and the fruits are just positions with only
relational properties.47 My worry is how to square this with the
naturalist's common-sense realism.48
But whatever the upshot of these inconclusive speculations about
whether set theoretic realism should or shouldn't be considered a
version of structuralism, my main goal here is to call attention to
yet another point of agreement. Notice that in the pure iterative
46
Parsons (forthcoming), §9, makes a related point about the epistemological
importance of recognizing non-relational features of sets.
47
Resnik (personal communication) has suggested this move, and there are hints
of it in Shapiro (forthcoming).
48
See ch. 1, sect. 2, above.
176 M O N I S M AND B E Y O N D

hierarchy structure, the continuum hypothesis is either true or false,

the projective sets either do or don't include a non-measurable set
or an uncountable set without a perfect subset. Thus, only one of
SC and V — L can be true there. So again, as in the previous section,
the non-partisan conclusion is that structuralists, as well as
compromise Platonists and Fieldian nominalists, will have to face
the difficult problem of assessing the rationality of arguments for
and against the various theoretical hypotheses that might answer
these open questions.

There is a variation on structuralism according to which math-

ematics is the study not of structures but of possible structures.
Rather than investigating sets (compromise Platonism) or the set
theoretic pattern (structuralism), the modalist investigates what
would be the case if there were a set theoretic hierarchy of the sort
the Platonist describes.49 '2 + 2 - 4 ' translates to 'if there were a
natural number structure, 2 plus 2 would equal 4 in that structure'.
This view has obvious if-thenist elements, and it suffers from many
of the same difficulties. 50 Notice also that the modalist's actual
world is purely physical; all mathematical things exist (if at all) in
some other possible world. Even if the extreme monism of section 1
above is rejected, the pro-Platonist arguments of Quine and Putnam
suggest that such a separation of the physical from the mathemat-
ical is not feasible. Thus the modalist, like Field, must find a way to
defuse the indispensability arguments.51
Epistemologically, the modalist owes an account of modal
knowledge that has not been forthcoming. One might think all
that's needed is an explanation of the logical implication from, say,
the Peano axioms to 2 + 2 = 4, but there is more; the modal
translation will not work properly unless the Peano axioms are
jointly possible. 52 In set theory, the corresponding requirement is
that the iterative hierarchy be possible, and in this possible

49
This sort of translation is suggested in Putnam (\967a), though he doesn't
espouse modalism. Following Putnam's method, Hellman (1989) does.
50
For a partial list of these, see ch. 1, sect. 4, above. For more, see Maddy
(forthcoming b) or Resnik (1980), ch. 3.
51
See Hellman (1989), ch. 3, and Field (1988), §§6-7, for an assessment of the
modalist's prospects.
52
If no such structure is even possible, 2 + 2 = 4 and 2 + 2 = 5 are both true,
along with everything else. Again, the logic involved must be second-order if we are
to speak of a unique natural number structure.
 SUMMARY 177

structure, the continuum hypothesis is either true or false, the

projective sets either do or don't include a well-ordering of the
reals, and so on. Thus, the modalist faces a question analogous to
Field's—which of V = L or SC is conservative?—namely, which of
these patterns is possible? I bring up modalism here primarily to
point out that it is among the many positions, both Platonistic and
non-Platonistic, that face not just the difficult question of whether
or not a supercompact cardinal exists (here or in another possible
world), but the prior and perhaps more difficult problem of how
one might rationally answer such a question.

4. Summary

Realism about a given branch of inquiry is the contention that its

subject matter exists objectively, that various efforts to reinterpret
its claims should be resisted, and that most of its well-supported
hypotheses are at least approximately true. I've endorsed common-
sense realism about medium-sized physical objects on the grounds
that the best explanation of why it seems to us that there is an
objective world of such objects is that there is an objective world of
such objects that is responsible for our beliefs. This explanation
takes place, not within a priori philosophy, but within our scientific
theory of the world and ourselves as cognizers; this is naturalism.
I've also adopted scientific realism about the theoretical entities of
natural science, because these unobservable things play a role in our
best theory of the world. Similar reasoning cites the central role of
classical mathematics in both the statement and the development of
natural science as evidence for mathematical realism or Platonism.
These are the pro-Platonist indispensability arguments of Quine
and Putnam.
Quine/Putnam Platonism differs from the traditional variety over
the purported a priority, certainty, and necessity of mathematical
truth. As a complete theory of mathematical knowledge, it also
differs from the practice of mathematics itself: it fails to account for
unapplied mathematics and for the obviousness of elementary
mathematics; it ignores the actual justificatory practices of math-
ematicians. Godel's version of Platonism, by contrast, presents an
appealing two-tiered account of justification within mathematics—
intuitive and theoretical—but fails to support the scientific status of
178 M O N I S M AND B E Y O N D

mathematics as a whole and rests its account of elementary

knowledge on an unpersuasive notion of mathematical intuition.
Nevertheless, Godelian Platonism stands with Quine/Putnamism in
opposition to the traditional variety.
I've proposed a compromise between these two modern versions
of Platonism. From Quine/Putnamism, it takes the indispensability
arguments as supports for the (approximate) truth of classical
mathematics. From Godel, it takes the two-tiered analysis of
mathematical justification. But to provide a complete picture,
compromise Platonism owes a replacement for Godel's intuition; in
deference to naturalism, this replacement must be scientifically
feasible. The leading theme of this book has been the development
and defence of set theoretic realism, a version of compromise
Platonism designed to fill in this outline.
It has long been thought that Godel's intuition, his epistemologi-
cal bridge between the objects of mathematical knowledge and the
mathematical knower, cannot be developed naturalistically.
Benacerraf's classical statement of this worry (1973) depends on
the then-popular causal theories of knowledge and reference, but
I've argued that neither these nor a particularly robust notion of
truth are essential to posing the problem. What matters is that the
beliefs of mathematicians are reliable indicators of facts about
mathematical things; this fact calls out for a naturalistic explan-
ation. From this point> various forces—among them the conviction
that mathematics is a legitimate science analogous to the physical
sciences—lead to the conviction that at least part of this explan-
ation must involve a perception-like connection between object
known and knower. Add to this the traditional Platonist's
characterization of mathematical entities as non-spatio-temporal
and acausal, and it's easy to see why a naturalistic account is often
considered impossible.
The set theoretic realist meets this problem by admitting sets of
physical objects to the physical world, giving them spatio-temporal
location where the physical stuff that makes up their members (and
the members of their members, etc.) is located. These impure sets
then prove appealing candidates for the subjects of perceptual
numerical beliefs, and psychological and speculative neurological
considerations give scientific support to the view that they are
directly perceived. Thus part of the set theoretic realist's per-
ception-like connection is just perception itself. An accompanying
 SUMMARY 179

neurological phenomenon furnishes a rudimentary intuitive faculty

whose products—intuitive beliefs—provide fallible but prima-facie
justifications for the most elementary general assumptions of set
theory.
This account depends essentially on the close relationship
between numerical beliefs and beliefs about sets, which raises the
familiar ontological question of whether numbers simply are sets.
Part of the scientific support for set theory rests on the foundation it
provides for number theory and analysis, and this foundational
theory is standardly expressed by identifying the natural and real
numbers with certain sets. But, as Benacerraf has pointed out, this
identification is ultimately unsatisfying because it can be done with
equal ease in several different ways; this is the problem of multiple
reductions. If there is nothing to decide between the von Neumann
and the Zermelo ordinals when identifying the natural numbers
with sets, how can either sequence of sets claim to actually be the
numbers? The set theoretic realist's answer, implicit in the account
of set perception, is that neither sequence is the numbers, that
numbers are properties of sets which either sequence is equally
well equipped to measure. The same line of response works for the
real numbers when they are understood as detectors for the
property of continuity.
If the first tier of Godel's epistemological theory can be ascribed
to the set theoretic realist's perception and intuition, there remains
the problem of describing and accounting for the rationality of
reasoning at the theoretical level. In set theory, despite traces of the
traditional Platonistic view that axioms are obvious or self-evident,
theoretical defences for axiom candidates can be found even in
Zermelo's first axiomatization, and they figure prominently in the
search for new hypotheses that will decide natural analytic and set
theoretic questions left open by the currently accepted axioms of
ZFC. The problem of assessing the rationality of various non-
demonstrative arguments for and against new set theoretic hypo-
theses becomes more acute as set theorists devise alternative,
conflicting, theories. The first step in helping adjudicate such
disputes is a descriptive catalogue of the evidence offered by each
side. A modest contribution to that project is all that has been
attempted here. The next step, the evaluation of this evidence, is a
daunting undertaking, but I've argued that the set theoretic realist
faces this challenge in the distinguished company of thinkers
180 M O N I S M AND B E Y O N D

representing a wide range of competing mathematical philosophies,

structuralism, modalism, and a version of nominalism among them.
Finally, for the benefit of those with physicalistic leanings, I've
sketched set theoretic monism, a minor variation on set theoretic
realism. For the monist, all sets have physical grounding and spatio-
temporal location, and all physical objects are sets. These
manoevres produce a radical 'one-worldism'—a reality at once
mathematical and physical—that should appeal to philosophers of
this stripe.

In sum, then, I certainly do not claim to have shown that my

version of Platonism raises no difficult philosophical problems. At
best, at best, I have shown how to replace the two prominent
Benacerraf-style objections to traditional Platonism with a new
open question about the justification of theoretical hypotheses in
set theory. But whatever the complexities of this new problem, I
think this trade amounts to progress. In the defence of mathemat-
ical realism, the new problem enjoys a clear advantage over its
predecessors: nothing on its face is likely to inspire one of those
nagging a priori arguments against the very possibility of Platon-
ism. On the contrary, the questions it raises—questions of
rationality—are standard in the philosophy of all sciences, and
there is no obvious reason why they should be any less tractable in
mathematics than they are in physics or physiology.
But there is more to be said for this new problem than that it may
lighten the perceived burden on the defender of Platonism. I attach
considerable importance to the fact that it arises also for adherents
of alternative philosophical positions; this suggests that it taps into
a fundamental issue insensitive to minor variations in philosophical
fashion. And beyond this, there is the alluring possibility that
philosophical progress on questions of mathematical rationality
could make a real contribution to mathematics itself, especially to
the current search for new axioms. Thus, once again, I recommend
pursuit of this new problem even to philosophers blissfully
uninvolved in the debate over Platonism.

Mathematicians often think of themselves as scientists, exploring

the intricacies of mathematical reality; and, for good reason, they
are especially inclined towards such views in the absence of
philosophers, I have tried to show that, contrary to popular
 SUMMARY 181

philosophical opinion, something close to the mathematician's

natural attitude is defensible. Theories of mathematical knowledge
tend either to trivialize it as conventional or purely formal or even
false, or to glamorize it as perfect, a priori, and certain, but set
theoretic realism aims to treat it as no more nor less than the science
it is, and to be fair, all at once, to the mathematician who produces
the knowledge, the scientist who uses it, and the cognitive scientist
who must explain it. I propose it, then, as another step—after
Codel, Quine, and Putnam—on the long road towards math-
ematics naturalized.
 REFERENCES

ACHINSTEIN, P. (1965), The problem of theoretical terms', repr. in Brody

(ed.) (1970), 234-50.
ACKERMANN, W. (1956), 'Zur Axiomatik der Mengenlehre', Math-
ematiscke Annalen, 131, pp. 336-45.
ACZEL, P. (1988), Non-Well-founded Sets, Center for the Study of
Language and Information, Lecture Notes, no. 14.
ADDISON, J. W. (1958), 'Separation principles in the hierarchies of classical
and effective descriptive set theory', Fundamenta Mathematicae, 46,
pp. 123-35.
(1959), 'Some consequences of the axiom of constructibility', Funda-
menta Mathematicae, 46, pp. 337—57.
and MOSCHOVAKIS, Y. N. (1968), 'Some consequences of the axiom
of definable determinateness', Proceedings of the National Academy of
Sciences (U. S. A.), 59, pp. 708-12.
ALEXANDROFF, P. (1916) 'Sur la puissance des ensembles mesurables B',
Comptes rendus de I'Academie des Sciences de Paris, 162, pp. 232—5.
ANDERSON, C. A. (1987), 'Review of Bealer's Quality and Concept',
Journal of Philosophical Logic, 16, pp. 115—64.
ARISTOTLE (1952), The Works of Aristotle Translated into English, 12
vols., ed. W. D. Ross (Oxford: Oxford University Press).
Categories, in his (1952).
Metaphysics, in his (1952).
Physics, in his (1952).
ARMSTRONG, D. (1961), Perception and the Physical World (London:
Routledge and Kegan Paul).
(1973), Belief, Truth and Knowledge (Cambridge: Cambridge Uni-
versity Press).
(1977), 'Naturalism, materialism and first philosophy', repr. in his
(1981), 149-65.
(1978), Universals and Scientific Realism (Cambridge: Cambridge
University Press).
(1980), 'Against "ostrich nominalism".', Pacific Philosophical Quar-
terly, 16, pp. 440-9.
(1981), The Nature of Mind (Ithaca, NY: Cornell University Press).
 REFERENCES 183

AYER, A. J. (1946), Language, Truth, and Logic, 2nd edn. (New York:
Dover, 1952).
AYERS, M. R. (1981), 'Locke versus Aristotle on natural kinds', Journal of
Philosophy, 78, pp. 247-72.
BAIRE, R. (1899). 'Sur les fonctions de variables reelles', Annali di
matematica pura ed applicata, 3, pp. 1-122.
BOREL, E., HADAMARD, J., and LEBESGUE, H. (1905), 'Five letters on
set theory', repr. in Moore (1982), 311-20.
BARWISE, J. (ed.) (1977), The Handbook of Mathematical Logic
(Amsterdam: North Holland).
BEALER, G. (1982), Quality and Concept (Oxford: Oxford University Press).
BENACERRAF, P. (1965), 'What numbers could not be', repr. in Benacerraf
and Putnam (eds.) (1983), 272-94.
(1973), 'Mathematical truth', repr. in Benacerraf and Putnam (eds.)
(1983), 403-20.
(1985), 'Comments on Maddy and Tymoczko', in Kitcher (ed.)
(1985), 476-85.
and PUTNAM, H. (eds.) (1983), Philosophy of Mathematics, 2nd edn.
(Cambridge: Cambridge University Press).
BERKELEY, G. (1710), The Principles of Human Knowledge, in his (1957).
(1713), Three Dialogues between Hylas and Philonous, in his (1957).
(1734), The Analyst, in his (1957).
(1957), The Works of George Berkeley, Bishop ofCloyne, 9 vols., ed.
A. Luce and T. Jessop (London: Thomas Nelson and Sons).
BERNAYS, P. (1935), 'On platonism in mathematics', repr. in Benacerraf
and Putnam (eds.) (1983), 258-71.
(1937), 'A system of axiomatic set theory, I', Journal of Symbolic
Logic, 2, pp. 65-77.
BLACKWELL, D. (1967), 'Infinite games and analytic sets', Proceedings of
the National Academy of Sciences (U.S.A.), 58, pp. 1836-7.
BONEVAC, D. A. (1982), Reduction in the Abstract Sciences (Indianapolis,
Ind.: Hackett).
BONJOUR, L. (1980), 'Externalist theories of empirical knowledge',
Midwest Studies in Philosophy, 5 (Minneapolis: University of Min-
nesota Press), 53—73.
BOOLOS, G. (1971), The iterative conception of set', repr. in Benacerraf
and Putnam (eds.) (1983), 486-502.
BOREL, E. (1898), Lemons sur la theorie des fonctions (Paris: Gauthier-
Villars).
BOWER, T. G. R. (1966), 'The visual world of infants', Scientific American,
215, no. 6, pp. 80-92.
(1982), Development in Infancy, 2nd edn. (San Francisco: W. H.
Freeman and Company).
184 REFERENCES

BOYER, C. B. (1949), The History of the Calculus and its Conceptual

Development, (New York: Dover, 1959). (Original title: The Concepts
of the Calculus.)
BRIDGMAN, P. W. (1927), The Logic of Modem Physics (New York:
Macmillan).
BRODY, B. (ed.) (1970), Readings in the Philosophy of Science (Englewood
Cliffs, NJ: Prentice Hall).
BROUWER, L. E. J. (1913), 'Intuitionism and formalism', repr. in Benacerraf
and Putnam (eds.) (1983), 77-89.
(1949), 'Consciousness, philosophy, and mathematics', repr. in
Benacerraf and Putnam (eds.) (1983), 90-6.
BRUNER, J. (1957), 'On perceptual readiness', repr. in R. Harper et al.
(eds.), The Cognitive Processes (Englewood Cliffs, NJ: Prentice Hall,
1964), 225-56.
BuRALi-FoRTi, C. (1897), 'A question on transfinite numbers', repr. in van
Heijenoort (ed.) (1967), 104-12.
BURGESS, J. P. (1983). 'Why I am not a nominalist', Notre Dame Journal of
Formal Logic, 24, pp. 93-105.
(1984), 'Synthetic mechanics', Journal of Philosophical Logic, 13,
pp. 379-95.
(forthcoming a), 'Synthetic physics and nominalist realism', to appear
in C. W. Savage and P. Erlich (eds.), The Nature and Function of
Measurement.
(forthcoming b), 'Epistemology and nominalism', to appear in A.
Irvine (ed.), Physicalism in Mathematics.
CANTOR, G. (1872), 'Uber die Ausdehnung eines Satzes aus der Theorie der
trigonometrischen Reihen', Mathematische Annalen, 5, pp. 123—32.
(1878), 'Ein Beitrag zur Mannigfaltigkeitslehre', Journal fur die reine
und angewandte Mathematik, 84, pp. 242—58.
(1883), Grundlagen einer allgemeinen Mannigfaltigkeitslehre (Leip-
zig: B. G. Teubner).
(1891), 'Uber eine elementare Frage der Mannigfaltigkeitslehre',
Jahresbericht der Deutschen Mathematiker-Vereinigung, 1, pp. 75—8.
(1895/7), Contributions to the Founding of the Theory of Transfinite
Numbers, ed. P. E. B. Jourdain (Chicago: Open Court, 1915).
(1899), 'Letter to Dedekind', repr. in van Heijenoort (ed.) (1967),
113-17.
CARNAP, R. (1934), The Unity of Science (London: Kegan Paul, Trench,
Trubner and Company).
(1936/7), 'Testability and meaning', Philosophy of Science, 3, pp.
428-68, and 4, pp. 1^0.
(1937), Logical Syntax of Language (London: Routledge and Kegan
Paul).
 REFERENCES 185

(1950), 'Empiricism, semantics, and ontology', repr. in Benacerraf

and Putnam (eds.) (1983), 241-57.
CASULLO, A. (forthcoming) 'Causality, reliabilism, and mathematical
knowledge', to appear.
CHIHARA, C. (1973), Ontology and the Vicious-Circle Principle (Ithaca,
NY: Cornell University Press).
(1982), 'A Godelian thesis regarding mathematical objects: Do they
exist? And can we perceive them?', Philosophical Review, 91, pp. 211-
27.
CHISHOLM, R. (1977), Theory of Knowledge, 2nd edn. (Englewood Cliffs,
NJ: Prentice Hall).
COHEN, P. J. (1966), Set Theory and the Continuum Hypothesis (New
York: W. A. Benjamin).
DAUBEN, J. W. (1979), Georg Cantor (Cambridge, Mass.: Harvard
University Press).
DAVIS, M. (1964), 'Infinite games of perfect information', Annals of
Mathematics Studies, 52, pp. 85-101.
DAVIS, P. J., and HERSH, R. (1981), The Mathematical Experience
(Boston: Birkhauser).
DEDEKIND, R. (1872), 'Continuity and irrational numbers', in Essays on
the Theory of Numbers (La Salle, III: Open Court, 1901), pp. 1-27.
DENNETT, D. C. (1978), Brainstorms (Bradford Books).
DESCARTES, R. (1641), Meditations on First Philosophy, 2nd edn., in his
(1967).
(1967), Philosophical Works of Descartes, 2 vols., ed. E. S. Haldane
and G. R. T. Ross (Cambridge: Cambridge University Press).
DETLEFSEN, M. (1986), Hilbert's Program (Dordrecht: Reidel).
DEVITT, M. (1980), '"Ostrich nominalism" or "mirage realism"?', Pacific
Philosophical Quarterly, 61, pp. 433-9.
(1981), Designation (New York: Columbia University Press).
(1984), Realism and Truth (Princeton, NJ: Princeton University
Press).
DEVLIN, K. (1977), The Axiom of Constructibility (Berlin: Springer-
Verlag).
(1984), Constructibility (Berlin: Springer-Verlag).
DRAKE, F. (1974), Set Theory: An Introduction to Large Cardinals
(Amsterdam: North Holland).
DUMMETT, M. (1975), 'The philosophical basis of intuitionist logic', repr.
in his (1978), ch. 14, and in Benacerraf and Putnam (eds.) (1983), 97-
129.
(1977), Elements of Intuitionism (Oxford: Oxford University Press).
(1978), Truth and Other Enigmas (Cambridge, Mass.: Harvard
University Press).
186 REFERENCES

EKLOF, P. C, and MEKLER, A. H. (forthcoming], Almost Free Modules:

Set-Theoretic Methods, forthcoming from North Holland Publishers,
Mathematical Library Series.
ELLIS, B, (1966), Basic Concepts of Measurement (Cambridge: Cambridge
University Press).
ENDERTON, H. (1972), A Mathematical Introduction to Logic (New York:
Academic Press).
(1977), Elements of Set Theory (New York: Academic Press).
FEFERMAN, S. (1984tf), 'Toward useful type-free theories, P, Journal of
Symbolic Logic, 49, pp. 75-111.
(1984&), 'Kurt Godel: Conviction and caution', repr. in S. G, Shanker
(ed.), Godel's Theorem in Focus (London: Croom Helm, 1988),
96-114.
(1988), 'Hilbert's program relativized: Proof-theoretical and founda-
tional reductions', Journal of Symbolic Logic, 53, pp. 364—84.
FIELD, H. (1972), 'Tarski's theory of truth', Journal of Philosophy, 69, pp.
347-75.
(1980), Science without Numbers (Princeton, NJ: Princeton Uni-
versity Press).
(1982), 'Realism and anti-realism about mathematics', repr. in his
(1989), 53-78.
(1984), 'Is mathematical knowledge just logical knowledge?', repr. in
his (1989), 79-124.
(1985), 'On conservativeness and incompleteness', repr. in his (1989),
125-46.
(1986), The deflationary conception of truth', in G, MacDonald and
C. Wright (eds.), Fact, Science and Value (Oxford: Basil Blackweli), 55-
117.
(1988), 'Realism, mathematics, and modality', repr. in his (1989),
227-81.
(1989), Realism, Mathematics, and Modality (Oxford: Basil Black-
well).
(forthcoming), 'Physicalism'.
FODOR, J. A. (1975), The Language of Thought (New York: Thomas Y.
Crowell).
FOREMAN, M. (1986), 'Potent axioms', Transactions of the American
Mathematical Society, 294, pp. 1-28.
FRAENKEL, A. A. (1922), 'Zu den Grundlagen der Cantor-Zermeloschen
Mengenlehre', Mathematische Annalen, 86, pp. 230—7.
BAR-HILLEL, Y., and LEVY, A. (197'3), Foundations of Set Theory, 2nd
rev. edn. (Amsterdam: North Holland).
FREGE, G. (1884), The Foundations of Arithmetic, 2nd rev. edn. (Evanston,
111: Northwestern University Press, 1968).
(1892a), 'On concept and object', in his (1970), 42-55.
 REFERENCES 187

(18926), 'On sense and reference', in his (1970), 56-78.

(1903), Grundgesetze der Arithmetik, vol. ii. Relevant sections are
reprinted in his (1970), 182-233.
(1970), Translations from the Philosophical Writings of Gottlob
Frege, ed. P. Geach and M. Black (Oxford: Basil Blackwell).
(1979), Posthumous Writings, ed. H. Hermes, F. Kambartel, and F.
Kaulbach (Chicago: University of Chicago Press).
FRIEDMAN, H. (1971), 'Higher set theory and mathematical practice',
Annals of Mathematical Logic, 2, pp. 325-57.
GALE, D., and STEWART, F. M. (1953), 'Infinite games with perfect
information', Annals of Mathematics Studies, 28, pp. 245—66.
GELMAN, R. (1977), 'How young children reason about small numbers', in
N. Castellan, D. Pisoni, and G. Potts (eds.), Cognitive Theory, ii
(Hillsdale, NJ: Lawrence Erlbaum Associates), 219-38.
GETTIER, E. (1963), 'Is justified true belief knowledge?', Analysis, 23, pp.
121-3.
GIBSON, E. (1969), Principles of Perceptual Learning and Development
(New York: Appleton-Century-Crofts).
GIBSON, J. J. (1950), The Perception of the Visual World (Boston:
Houghton Mifflin),
GODEL, K. (1930), 'The completeness of the axioms of the functional
calculus of logic', repr. in van Heijenoort (ed.) (1967), 582—91.
(1931), 'On formally undecidable propositions of Principia Math-
ematica and related systems, I', repr. in van Heijenoort (ed.) (1967),
596-616.
(1938), 'The consistency of the axiom of choice and of the generalized
continuum hypothesis', Proceedings of the National Academy of
Sciences (U.S.A.), 24, pp. 556—7.
(1940), The Consistency of the Continuum Hypothesis (Princeton:
Princeton University Press).
(1944), 'Russell's mathematical logic', repr. in Benacerraf and
Putnam (eds.) (1983), 447-69.
(1946), 'Remarks before the Princeton Bicentennial Conference on
problems in mathematics', in M. Davis (ed.), The Undecidable (New
York: Raven Press, 1965), 84-8.
(1947/64), 'What is Cantor's continuum problem?', repr. in
Benacerraf and Putnam (eds.) (1983), 470-85.
GOLDMAN, A. (1967), 'A causal theory of knowing', Journal of Philo-
sophy, 64, pp. 357-72.
(1975), 'Innate knowledge', in S. Stich (ed.), Innate Ideas (Berkeley:
University of California Press), 111—20.
(1976), 'Discrimination and perceptual knowledge', Journal of Philo-
sophy, 73, pp. 771-91.
(1977), 'Perceptual objects', Synthese, 35, pp. 257-84.
188 REFERENCES

GOLDMAN, A. (1979), 'What is justified belief?', in G. Pappas (ed.),

Justification and Knowledge (Amsterdam: Reidel), 1-23.
(1980), The internalist conception of justification', Midwest
Studies in Philosophy, 5 (Minneapolis: University of Minnesota Press),
27-51.
GOTTLIEB, D. (1980), Ontological Economy (Oxford: Oxford University
Press).
GREGORY, R. L. (1970), The Intelligent Eye (New York: McGraw-Hill).
(1972), Eye and Brain, 2nd edn. (New York: McGraw-Hill).
GRICE, P. (1961), 'The causal theory of perception', repr. in R. Swartz
(ed.), Perceiving, Sensing and Knowing (Berkeley: University of
California Press), 438-72.
GROVER, D., CAMP, J., and BELNAP, N. (1975), 'A prosentential theory of
truth', Philosophical Studies, 27, pp. 73-125.
HALE, B. (1987). Abstract Objects (Oxford: Basil Blackwell).
HALLETT, M. (1984), Cantonan Set Theory and Limitation of Size
(Oxford: Oxford University Press).
HAMBOURGER, R. (1977), 'A difficulty with the Frege—Russell definition of
number', Journal of Philosophy, 74, pp. 409-14.
HARMAN, G. (1973), Thought (Princeton: Princeton University Press).
HARRINGTON, L. A., MORLEY, M. D., SCEDROV, A., and SIMPSON, S. G.
(eds.) (1985), Harvey Friedman's Research on the foundations of
Mathematics (Amsterdam: North Holland).
HART, W. H. (1977), 'Review of Steiner's Mathematical Knowledge',
journal of Philosophy, 74, pp. 118-29.
HAUSDORFF, F. (1919), 'Uber halbstetige Funktionen und deren Verall-
gemeinerung, Mathematische Zeitschrift, 5, pp. 292—309.
HEBB, D. O. (1949), The Organization of Behavior (New York: John
Wiley and Sons).
(1980), Essay on Mind (Hillsdale, NJ: Lawrence Erlbaum Associates).
HELLMAN, G. (1989), Mathematics Without Numbers (Oxford: Oxford
University Press).
HEMPEL, C. G. (1945), 'On the nature of mathematical truth', repr. in
Benacerraf and Putnam (eds.) (1983), 377-93.
(1954), 'A logical appraisal of operationalism', repr. in his (1965),
123-33.
• (1965), Aspects of Scientific Explanation (New York: The Free Press).
HENKIN, L. (1949), 'The completeness of the first-order functional
calculus', Journal of Symbolic Logic, 14, pp. 159—66.
HEYTING, A. (1931), 'The intuitionist foundations of mathematics', repr. in
Benacerraf and Putnam (eds.) (1983), 52-61.
(1966), Intuitionism: An Introduction, 2nd rev. edn. (Amsterdam:
North Holland).
 REFERENCES 189

HILBERT, D. (1899), Foundations of Geometry (La Salle, 111.: Open Court,

1971).
(1926), 'On the infinite', repr. in Benacerraf and Putnam (eds.)
(1983), 183-201, and in van Heijenoort (ed.) (1967), 367-92.
• (1928), 'The foundations of mathematics', repr. in van Heijenoort
(ed.) (1967), 464-79.
HODES, H. (1984), 'Logicism and the ontological commitments of
arithmetic', Journal of Philosophy, 81, pp. 123—49.
HUME, D. (1739), A Treatise of Human Nature, vol. i, in his (1886).
(1886), Philosophical Works, 4 vols., ed. T. H. Green and T. H.
Grose (London).
JECH, T. (1978), Set Theory (New York: Academic Press).
JUBIEN, M. (1977), 'Ontology and mathematical truth', Nous, 11, pp. 133-
50.
KATZ, J. J. (1981), Language and Other Abstract Objects (Totowa, NJ:
Rowman and Littlefield).
KAUFMAN, E. L, LORD, M. W., REESE, T. W., and VOLKMANN, J. (1949),
'The discrimination of visual number', American Journal of Psychology,
62, pp. 498-525.
KELLEY, J. L. (1955), General Topology (Princeton, NJ: van Nostrand).
KIM, J. (1977), 'Perception and reference without causality', Journal of
Philosophy, 74, pp. 606-20.
(1981), 'The role of perception in a priori knowledge', Philosophical
Studies, 40, pp. 339-54.
KITCHER, P. (1978), 'The plight of the Platonist', No«s, 12, pp. 119-36.
(1983), The Nature of Mathematical Knowledge (New York: Oxford
University Press).
-(ed.) (1985), PSA 1984, ii (East Lansing: Philosophy of Science
Association)
KLINE, M. (1972), Mathematical Thought from Ancient to Modern Times
(New York: Oxford University Press).
KORNER, S. (1960), The Philosophy of Mathematics (London: Hutchinson
University Library).
KRIPKE, S. (1972), 'Naming and necessity', in D. Davidson and G. Harman
(eds.), Semantics of Natural Language (Dordrecht: Reidel), 253—355,
763-9.
(1975), 'Outline of a theory of truth', Journal of Philosophy, 72,
pp. 690-716.
• (1982), Wittgenstein on Rules and Private Language (Cambridge,
Mass.: Harvard University Press).
KURATOWSKI, K. (1966), Topology, i (New York: Academic Press).
LEAR, J. (1977), 'Sets and semantics', Journal of Philosophy, 74, pp. 86—
102.
190 REFERENCES

LEBESGUE, H. (1902), 'Integrate, longueur, aire', Annali di matematica

pura ed applicata, 7, pp. 231-359.
— (1905), 'Sur les fonctions representables analytiquement, Journal de
mathematiques pures et appliquees, 60, pp. 139—216.
LEEDS, S. (1978), Theories of reference and truth', Erkenntnis, 13, pp.
111-29.
LETTVIN, J. Y., MATURANA, H. R., MCCULLOCH, W. S., and PITTS, W. H.
(1959), 'What the frog's eye tells the frog's brain', repr. in W. S.
McCulloch, Embodiments of Mind (Cambridge, Mass.: MIT Press,
1965), 230-55.
LEVY, A., and SOLOVAY, R. M. (1967), 'Measurable cardinals and the
continuum hypothesis', Israel Journal of Mathematics, 5, pp. 234—48.
LEWIS, D. (1983), 'New work for a theory of universals', Australian
Journal of Philosophy, 61, pp. 343-77.
(1984), 'Putnam's paradox', Australian Journal of Philosophy, 62,
pp. 221-36.
(1986), On the Plurality of Worlds (Oxford: Basil Blackwell).
LOCKE, J. (1690), An Essay Concerning Human Understanding (New
York: Dover, 1959).
LUCE, L. (1988), 'Frege on cardinality', Philosophy and Phenomenological
Research, 48, pp. 415-34.
LUZIN, N. (1917), 'Sur la classification de M. Baire', Comptes rendus de
I'Academie des Sciences de Paris, 164, pp. 91—4.
(1925), 'Sur les ensembles projectifs de M. Henri Lebesgue', Comptes
rendus de I'Academie des Sciences de Paris, 180, pp. 1572-4.
(1927), 'Sur les ensembles analytiques', Fundamenta Mathematicae,
10, pp. 1-95.
MACHAMER, P. (1970), 'Recent work on perception', American Philo-
sophical Quarterly, 7, pp. 1-22.
MADDY, P. (1980), 'Perception and mathematical intuition', Philosophical
Review, 89, pp. 163-96.
(1981), 'Sets and numbers', Nous, 15, pp. 494-511.
(1983), 'Proper classes', Journal of Symbolic Logic, 48, pp. 113—39.
(1984a), 'Mathematical epistemology: what is the question?', Monist,
67, pp. 46-55.
(19846), 'How the causal theorist follows a rule', Midwest Studies in
Philosophy, 9 (Minneapolis: University of Minnesota Press), 457-77.
(1984c), 'Informal notes on proper classes', unpublished notes.
(1986), 'Mathematical alchemy', British Journal for the Philosophy
of Science, 37, pp. 279-314.
(1988a), 'Believing the axioms', Journal of Symbolic Logic, 53, pp.
481-511,736-64.
(19886), 'Mathematical realism', Midwest Studies in Philosophy, 12
(Minneapolis: University of Minnesota Press), 275-85.
 REFERENCES 191

(forthcoming a), 'Physicalistic Platonism', to appear in A. Irvine (ed.),

Physicalism in Mathematics.
(forthcoming b), The roots of contemporary Platonism', to appear in
the Journal of Symbolic Logic.
MALAMENT, D. (1982), 'Review of Field's Science Without Numbers',
Journal of Philosophy, 79, pp. 523-34.
MARTIN, D. A. (1968), 'The axiom of determinateness and reduction
principles in the analytical hierarchy', Bulletin of the American
Mathematical Society, 74, pp. 687—9.
(1970), .'Measurable cardinals and analytic games', Fundamenta
Mathematicae, 66, pp. 287-91.
(1975), 'Borel determinacy', Annals of Mathematics, 102, pp. 363-71.
(1976), 'Hilbert's first problem: The continuum hypothesis', Proceed-
ings of Symposia in Pure Mathematics, 28, (Providence, RI: American
Mathematical Society), 81-92.
(1977), 'Descriptive set theory: projective sets', in Barwise (ed.)
(1977), 783-815.
(1980), 'Infinite games', Proceedings of the International Congress of
Mathematicians (Helsinki, 1978), pp. 269-73.
(1985), 'A purely inductive proof of Borel determinacy', Proceedings
of Symposia in Pure Mathematics, 42 (Providence, RI: American
Mathematical Society), 303-8.
'Projective sets and cardinal numbers', unpublished photocopy.
'Sets versus classes', unpublished photocopy.
and SOLOVAY, R. M. (1970), 'Internal Cohen extensions', Annals of
Mathematical Logic, 2, pp. 143—78.
and STEEL, J. (1988), 'Projective determinacy', Proceedings of the
National Academy of Sciences (U.S.A.), 85, pp. 6582-6.
(1989), 'A proof of projective determinacy', Journal of the
American Mathematical Society, 2, pp. 71—125.
MAXWELL, G. (1962), 'The ontological status of theoretical entities', repr.
in Brody (ed.) (1970), 224-33.
MENZEL, C. (1988), 'Frege numbers and the relativity argument',
Canadian Journal of Philosophy, 18, pp. 87-98.
MERRILL, G. H. (1980), 'The model-theoretic argument against realism',
Philosophy of Science, 47, pp. 69—81.
MILL,]. S. (1843), A System of Logic, in his (1963/88), vols. vii and viii.
(1865), An Examination of Sir William Hamilton's Philosophy, in his
(1963/88), vol. ix.
(1963/88), The Collected Works of John Stuart Mill, 29 vols., ed.
J. M. Robson and J. Stillinger (Toronto: University of Toronto Press).
MIRIMANOFF, D. (1917#), 'Les Antinomies de Russell et de Burali-Forti et
le probleme fondamental de la theorie des ensembles', L'Enseignement
mathematique, 19, pp. 37—52.
192 REFERENCES

MIRIMANOFF, D. (\9\7b), 'Remarques sur la theorie des ensembles et les

antinomies Cantoriennes, I', L'Enseignement mathematique, 19,
pp. 209-17.
MONNA, A. F. (1972), 'The concept of function in the 19th and 20th
centuries', Archive for History of Exact Sciences, 9, pp. 57—84.
MOORE, G. H. (1982), Zermelo's Axiom of Choice (New York: Springer-
Verlag).
(forthcoming), 'Introductory note to 1947 and 1964', The Collected
Works of Kurt Godel, vol. ii, forthcoming from Oxford University Press.
MORSE, A. (1965), A Theory of Sets (New York: Academic Press).
MOSCHOVAKIS, Y. N. (1970), 'Determinacy and prewellorderings of the
continuum', in Y. Bar-Hillel (ed.), Mathematical Logic and Foundations
of Set Theory (Amsterdam: North Holland), 24-62.
(1980), Descriptive Set Theory (Amsterdam: North Holland).
MYCIELSKI, J., and STEINHAUS, H. (1962), 'A mathematical axiom
contradicting the axiom of choice', Bulletin de I'Academie Polonaise des
Sciences, 10, pp. 1-3.
and SWIERCZKOWSKI, S. (1964), 'On the Lebesgue measurability
and the axiom of determinateness', Fundamenta Mathematicae, 54,
pp. 67-71.
NEISSER, U. (1976), Cognition and Reality (San Francisco: W. H. Freeman
and Company).
NOVIKOV, P. (1935), 'Sur la separabilite des ensembles projectifs de
seconde classe', Fundamenta Mathematicae, 25, pp. 459—66.
NYIKOS, P. (forthcoming), 'Testimony on large cardinals and set-theoretic
consistency results'.
PARSONS, C. (1965), 'Frege's theory of number', repr. in his (1983a), 150-
75.
(1974d), 'Sets and classes', repr. in his (1983d), 209-20.
(1974ft), The liar paradox', repr. in his (1983d), 221-67.
(1977), 'What is the iterative conception of set?', repr. in his (1983<z),
268-97, and in Benacerraf and Putnam (eds.) (1983), 503-29.
(1979/80), 'Mathematical intuition', Proceedings of the Aristotelian
Society, SO, pp. 145-68.
(1983d), Mathematics in Philosophy (Ithaca, NY: Cornell University
Press).
(1983ft), 'Quine on the philosophy of mathematics', in his (1983d),
176-205.
(forthcoming), 'The structuralist view of mathematical objects', to
appear in Synthese.
PHILLIPS, J. L. (1975), The Origins of Intellect: Piaget's Theory, 2nd edn.
(San Francisco: W. H. Freeman and Company).
PIAGET, J. (1937), The Construction of Reality in the Child (New York:
Basic Books, 1954).
 REFERENCES 193

and INHELDER, B. (1948), The Child's Conception of Space (New

York: W. W. Norton, 1967).
and SZEMINSKA, A. (1941), The Child's Conception of Number (New
York: Humanities Press, 1952).
PITCHER, G. (1971), A Theory of Perception (Princeton, NJ: Princeton
University Press).
PLATO (1871), The Dialogues of Plato, 5 vols., ed. B. Jowett (Oxford:
Oxford University Press).
Phaedo, in his (1871).
Phaedrus, in his (1871).
Republic, in his (1871).
Theaetetus, in his (1871).
Timaeus, in his (1871).
PUTNAM, H. (1962), 'What theories are not', repr. in his (1979), 215-27.
(19 67 a) ^ 'Mathematics without foundations', repr. in his (1979), 43-
59, and in Benacerraf and Putnam (eds.) (1983), 295-311.
(19676), The thesis that mathematics is logic", repr. in his (1979),
12-42.
(1968), 'The logic of quantum mechanics', repr. in his (1979), 174—
97.
(1970), 'On properties', repr. in his (1979), 305-22.
(1971), 'Philosophy of logic', repr. in his (1979), 323-57.
(1975a), Mind, Language and Matter (Philosophical Papers, 2)
(Cambridge: Cambridge University Press).
(19756), 'What is mathematical truth?', repr. in his (1979), 60-78.
(1977), 'Realism and reason', repr. in his (1978), 123-38.
(1978), Meaning and the Moral Sciences (Boston: Routledge and
KeganPaul).
(1979), Mathematics, Matter and Method (Philosophical Papers, 1,
2nd edn.) (Cambridge: Cambridge University Press).
(1980), 'Models and reality', repr. in Benacerraf and Putnam (eds.)
(1983), 421-44.
QUINE, W. V. O. (1936), 'Truth by convention', repr. in Benacerraf and
Putnam (eds.) (1983), 329-54.
(1948), 'On what there is', repr. in his (1980a), 1-19.
(1951), 'Two dogmas of empiricism', repr. in his (1980a), 20—46.
(1954), 'Carnap and logical truth', repr. in Benacerraf and Putnam
(eds). (1983), 355-76.
(1969a), Set Theory and its Logic, rev. edn. (Cambridge, Mass.:
Harvard University Press).
(19696), 'Epistemology naturalized', in his (1969c), 69-90.
(1969c), Ontological Relativity (New York: Columbia University
Press).
(1969d), 'Natural kinds', in his (1969c), 114-38.
194 REFERENCES

QUINE, W. V. O. (1980a), Prom a Logical Point of View, 2nd edn., rev.

(Cambridge, Mass.: Harvard University Press).
(1980£>), 'Soft impeachment disowned', Pacific Philosophical Quar-
terly, 61, pp. 450-1.
(1984), 'Review of Parsons's Mathematics in Philosophy', journal of
Philosophy, 81, pp. 783-94.
REINHARDT, W. N. (1974), 'Set existence principles of Shoenfield,
Ackermann, and Powell', Fundamenta Mathematicae, 84, pp. 5—34.
RESNIK, M. (1965), 'Frege's theory of incomplete entities', Philosophy of
Science, 32, pp. 329^1.
(1975), 'Mathematical knowledge and pattern cognition', Canadian
Journal of Philosophy, 5, pp. 25-39.
(1980), Frege and the Philosophy of Mathematics (Ithaca, NY:
Cornell University Press).
(1981), 'Mathematics as a science of patterns: Ontology and refer-
ence', Nous, 15, pp. 529-50.
(1982), 'Mathematics as a science of patterns: Epistemology', Nous,
16, pp. 95-105.
— (1985#), 'How nominalist is Hartry Field's nominalism?', Philosophi-
cal Studies, 47, pp. 163-81.
(19856), 'Ontology and logic: Remarks on Hartry Field's anti-
platomst philosophy of mathematics', History and Philosophy of Logic,
6, pp. 191-209.
(forthcoming a], 'A naturalized epistemoiogy for a Platonist math-
ematical ontology', to appear in Philosophica.
(forthcoming b), 'Beliefs about mathematical objects', to appear in A.
Irvine (ed.), Physicalism in Mathematics.
RUSSELL, B. (1902), 'Letter to Frege', repr. in van Heijenoort (ed.) (1967),
124-5.
(1906), 'On "insolubilia" and their solution by symbolic logic', repr.
in his (1973), 190-214.
(1907), 'The regressive method of discovering the premises of
mathematics', repr. in his (1973), 272-83.
(1919), Introduction to Mathematical Philosophy (London: Allen
and Urwin).
(1973), Essays in Analysis, ed. D, Lacky (London: Allen and
Unwin).
and WHITEHEAD, A. N. (1913) Principia Mathematica, 3
vols. (Cambridge: Cambridge University Press),
SALMON, N. (1981), Reference and Essence (Princeton, NJ: Princeton
University Press).
SCOTT, D. (1961), 'Measurable cardinals and constructible sets', Bulletin
de I'Academie Polonaise des Sciences, 7, pp. 145—9.
(1977), 'Foreword', in J. L. Bell, Boolean-Valued Models and
 REFERENCES 195

Independence Proofs in Set Theory (Oxford: Oxford University Press),

pp. xi-xviii.
SEARLE, J. (1958), 'Proper names', Mind, 67, pp. 166-73.
SHAPIRO, S. (1983a), 'Mathematics and reality', Philosophy of Science, 50,
pp. 523-48.
(1983&), 'Conservativeness and incompleteness', Journal of Philo-
sophy, 80, pp. 521-31.
(1985), 'Second-order languages and mathematical practice', Journal
of Symbolic Logic, 50, pp. 714^2.
(forthcoming), 'Structure and ontology', to appear in Philosophical
Topics.
SHOENFIELD, J. R. (1977), 'Axioms of set theory', in Barwise (ed.) (1977),
321-44.
SIERPINSKI, W. (1918), 'L'Axiome de M. Zermelo et son role dans la
theorie des ensembles et 1'analyse', Bulletin de I'Academie des Sciences
de Cracovie, pp. 97-152.
(1924), 'Sur une propriete des ensembles ambigus', Fundamenta
Mathematicae, 6, pp. 1—5.
(1925), 'Sur une classe d'ensembles', Fundamenta Mathematicae, 7,
pp. 237-43.
(1934), Hypothese du continu (Warsaw: Garasiriski).
and TARSKI, A. (1930), 'Sur une propriete caracteristique des
nombres inaccessibles', Fundamenta Mathematicae, 15, pp. 292—
300.
SIMPSON, S. (1988), 'Partial realizations of Hilbert's program', Journal of
Symbolic Logic, 53, pp. 349-63.
SKOLEM, T. (1923), 'Some remarks on axiomatized set theory', repr. in van
Heijenoort (ed.) (1967), 290-301.
SKYRMS, B. (1967), 'An explication of "X knows that p"', Journal of
Philosophy, 64, pp. 373-89.
SOLOVAY, R. M. (1969), 'On the cardinality of 21 sets of reals', in J.
Bulloff, T. Holyoke, and S. Hahn (eds.), Foundations of Mathematics
(Berlin: Springer-Verlag), 58-73.
(1970), 'A model of set theory in which every set of reals is Lebesgue
measurable', Annals of Mathematics, 92, pp. 1-56.
REINHARDT, W. N., and KANAMORI, A. (1978), 'Strong axioms of
infinity and elementary embeddings', Annals of Mathematical Logic, 13,
pp. 73-116.
STEINER, M. (1973), 'Platonism and the causal theory of knowledge',
Journal of Philosophy, 70, pp. 57—66.
(1975a), Mathematical Knowledge (Ithaca, NY: Cornell University
Press).
(1975£>), 'Review of Chihara's Ontology and the Vicious Circle
Principle', Journal of Philosophy 72, pp. 184—96.
196 REFERENCES

STEINER, M. (1978), 'Mathematical explanation', Philosophical Studies, 34,

pp. 135-51.
STRAWSON, P. (1959), Individuals (London: Methuen).
SUSLIN, M. (1917), 'Sur une definition des ensembles mesurables B sans
nombres transfinis', Comptes rendus de I'Academie des Sciences de
Pans, 164, pp. 88-91.
TARSKI, A. (1933), 'The concept of truth in formalized languages', repr. in
his Logic, Semantics, and Metamathematics, 2nd edn. (Indianapolis,
Ind.: Hackett, 1983), 152-278.
TROELSTRA, A. S. (1969), Principles of Intuitionism (Berlin: Springer-
Verlag).
ULAM, S. (1930), 'Zur Masstheorie in der allgemeinen Mengenlehre',
Fundamenta Mathematicae, 16, pp. 140-50.
URMSON, J. O. (1956), Philosophical Analysis (Oxford: Oxford University
Press).
VAN HEIJENOORT, J. (ed.) (1967), From Frege to Godel (Cambridge, Mass.:
Harvard University Press).
VON NEUMANN, J. (1923), 'On the introduction of transfinite numbers',
repr. in van Heijenoort (ed.) (1967), 346-54.
(1925), 'An axiomatization of set theory', repr. in van Heijenoort
(ed.) (1967), 393-413.
WANG, H. (1974a), 'The concept of set', ch. 6 of his (19746), repr. in
Benacerraf and Putnam (eds.) (1983), 530-70.
(1974fc), From Mathematics to Philosophy (London: Routledge and
Kegan Paul).
WEDBERG, A. (1955), Plato's Philosophy of Mathematics (Stockholm:
Almqvist and Wiksell).
WILLIAMSON, J. H. (1962), Lebesgue Integration (New York: Holt,
Rinehart, and Winston).
WILSON, M. (1979), 'Maxwell's condition—Goodman's problem', British
journal for the Philosophy of Science, 30, pp. 107—23.
(1985), 'What is this thing called "pain"?—the philosophy of science
behind the contemporary debate', Pacific Philosophical Quarterly, 66,
pp. 227-67.
• (forthcoming), 'Honorable intensions', in S. Wagner and R. Warner
(eds.) Notes Against a Program.
WITTGENSTEIN, L. (1953), Philosophical Investigations (New York:
Macmillan).
(1978), Remarks on the Foundations of Mathematics rev. edn., eds.
G. H. von Wright, R. Rhees, and G. E. M. Anscombe (Cambridge,
Mass.: MIT Press).
WOODIN, H. (1988), 'Supercompact cardinals, sets of reals, and weakly
homogeneous trees', Proceedings of the National Academy of Sciences
(U.S.A.), 85, pp. 6587-91.
 REFERENCES 197

WRIGHT, C. (1983), Frege's Conception of Numbers as Objects (Aberdeen:

Aberdeen University Press).
YOURGRAU, P. (1985), 'Sets, aggregates, and numbers', Canadian Journal
of Philosophy, 15, pp. 581-92.
ZERMELO, E. (1904), 'Proof that every set can be well-ordered', repr. in van
Heijenoort (ed.) (1967), 139-41.
(1908(3), 'A new proof of the possibility of a well-ordering', repr. in
van Heijenoort (ed,) (1967), 183-98.
(1908b), 'Investigations in the foundations of set theory, I', repr. in
van Heijenoort (ed.) (1967), 199-215.
(1930), 'fiber Grenzzahlen und Mengenbereiche', Fundamenta Math-
ematicae., 16, pp. 29-47.
This page intentionally left blank
 INDEX

abstract objects 2,21,36-7,40,59, union axiom; unlimited

152,163,172 comprehension; Zermelo—Fraenkel
accumulation points 108,114-15 axioms (ZFC).
Achinstein, P. 10 n.
Ackermann, W. 105 n. Baire.R. 111-12,118, 121,131 n.
Aczel, P. 40 n. Bealer, G. 62 n.
Addison.J. 126,133,137-8 Benacerraf, P. 36, 42, 43-4,45 n., 79,
Alexandroff, P. 112 81 n., 84-6, 88 n., 89-90,98-100,
analysis 107 n. 178,179
analytic vs. synthetic 27-8 Bendixson, I. 110
Anderson, C. 62 n. Berkeley, G. 6,16,22
a posteriori vs. a priori 9, 30, 33, 41, Bernays, P. 102 n., 104 n.
74, 177 Blackwell,D. 137-8
a priori, see a posteriori vs. a priori Bonevac, D. 36 n.
Aristotle 12, 158 Bonjour, L. 72 n.
Armstrong, D. 12-13, 46, 51 n., 52 n., Boolos, G. 39 n., 102 n.
156 n. Borel,E. 111,118,121
Ayer, A. 6 n., 16 n. Borel sets 111-14, 120,136, 138, 145,
Ayers, M. 39 n. 166,167
axiom of choice 117-24, 133,135, Bower, T. 53 n., 54, 64
138,142-3,145,146 Boyer, C. 22 n.
axiom of constructibility (V = L) Bridgman, P. 10 n.
103 n., 132-5,136,137-9,142-3, Brouwer, L. 16 n., 22
144-5, 147-9, 168-9, 173, 176, Bruner, J. 66 n.
177 Burali-Forti, C. 22 n.
axiom of extensionality 153 Burgess, J. 20 n., 43 n., 46-8, 49 n.,
axiom of foundation 40 n., 105 n. 78 n.
axiom of infinity 125,131,141,166
axiom of replacement 130—1, 136, 141, Cantor, G. 22,24, 81, 86, 102,107-
166 10,114-17,118,121,125,126,
axioms 4,27,31-3,40,67,70,72-3, 129-30,131 n.
77-8,102,107,114,117-18,125, Cantor—Bendixson theorem 110,112,
128-9,132,136,143-4,148, 114,129
179-80 Cantor's theorem 109, 131
see also axiom of choice; axiom of cardinal numbers 115,120-1,130-1,
constructibility (V = L); axiom of 145
extentionality; axiom of Carnap, R. 10 n., 27-8,29,154 n.
foundation; axiom of infinity; Casullo, A. 42 n.
axiom of replacement; large causal theory of knowledge 37, 41-2,
cardinal axioms; pairing axiom; 44,49, 72,178
power set axiom; projective causal theory of reference 38—41,42,
determinacy; separation axiom; 48-9, 178
200 INDEX

cell-assemblies 56-8, 66, 68-70, 73-4, 83 n., 84 n., 86 n., 89 n., 95 n.,
151 96 n., 102 n., 105 n., 109 n.,
certain vs. fallible 21,30,33,71, 115 n., 125 n., 131 n., 133 n.,
128 n., 177, 179 153 n., 162 n.
Chihara, C. 33-4, 78, 128-9, 150-4, epistemology 4 n.
159 n. see also causal theory of knowledge;
Chisholm, R. 144 epistemology naturalized; justified
classes 4 n., 61-2, 100, 102-6, 116 n., true belief; reliabilism
121-4, 134, 173 epistemology naturalized 9,11,
closed set 110 13,20,35,43,46-8,97
Cohen, P. 127, 130-1 equinumerous 83 n.
contingent vs. necessary 21, 30, 41, extensional vs. intensional 13 n., 61,
62-3,160-1,177 90-2
continuity 74-5, 81, 94-5, 110, 164, externalism vs. internalism 72—3
165,169,171,179 extrinsic justifications 33, 35, 71 n.,
see also real numbers 72-5, 76 n., 77-8, 80, 82, 107,
continuum hypothesis (CH) 100 n., 118-21, 124-5, 126-7, 128, 135,
110, 114-16, 125-32, 133, 135, 136-7, 138-«, 144-9, 150, 168,
142-3, 145,148, 165 n., 176, 177 172-3,176, 177,179-80
conventionalism 27—8, 29
correspondence theory of truth 16—20 fallible, see certain vs. fallible
countable 109 n. Feferman, S. 25 n., 29 n., 106 n.
countable choice 120 Field, H. 17 n., 18 n., 19 n., 26 n.,
36n.,43^,45n.,47-8, 81 n.,
DaubenJ. 81 n., 107 n., 109 n., 154-6, 159-69, 176-7
110 n., 116 n., 130 n., 131 n. Flagg, R. 106n.
Davis, M. 138 n. Fodor,J. 5 n., 154 n.
Davis, P, and Hersh, R. 2 n., 3 Foreman, M. 143 n.
Dedekind, R. 22,75,81-2,86,94, formalism 2-3, 10, 23-6, 28
102,108-9,114,118, 170 see a/so Hilbert's programme; if-
Dennett, D. 57 n. themsm
dependent choice 120,121 Fraenkel, A. 27
Descartes, R. 8-9, 72 n. Fraenkel, A., Bar-Hillel, Y., and Levy,
descriptive theory of reference 37—8 A. 104 n., 105 n., 134 n., 157 n.
detectors 53,57,58,65,80 Frege, G. 22 n., 23-4, 25, 26, 37, 60,
see also cell-assemblies 67 n., 82-3, 86 n., 88, 89-92, 94,
determinacy 137-40, 142-3, 145-7, 103, 104,114, 121, 169
166 Freidman, H. 136 n., 166
see also projective determinacy function 95-6,97,110-12,114,120,
Detlefsen, M. 25 n., 47 n. 121-2, 124
Devitt, M. 11 n., 13 n., 14 n., 15 n.,
17 n., 38 n.,40n. Gale, D. 138 n.
Devlin, K. 103 n., 133 n., 134 Gelman, R. 63 n.
Dieudonne, J. 2 Gettier, E. 36-7, 41 n., 51, 66 n., 72
disquotational theory of truth 17—20 Gibson, E. 54 n.
Drake, F. 82 n., 104 n., 116 n., 132 n., Gibson, J. 50 n.
135 n., 136 n. Godel, K. 25, 28-9, 31-5, 62 n., 71 n.,
Dummett, M. 15 n., 16n.,23 75-80, 103 n., 104 n., 120-1,
125-9, 130-2, 134-6, 148, 153,
Eklof, P., and Mekler, A. 4 n. 157 n., 177-9,181
Ellis, B. 87 n. Goldman, A. 37, 51 n., 71 n., 72 n., 73
Enderton, H. 4 n., 25 n., 27 n., 39 n., Gottlieb, D. 36 n.
59 n.,62n., 71 n., 81 n., 82 n., Gregory, R. 46 n., 50 n., 51 n.
 INDEX 201
Grice.P. 46 n., 51 Kelley,]. 104 n.
Grover, D. 17 n. Kim,J. 40n.,46n., 59 n., 67 n.
Kitcher, P. 30 n., 36 n., 64 n., 81 n.,
Hadamard,J. 122,124 84 n., 86 n.
Hale, B. 41 n. Kline, M. 22 n., 110 n., Ill n.
Hallett,M. 110 n., 114 n., 116 n., K6nig,J. 115-16
129 n., 130 n. Korner, S. 25 n.
Hambourger, R. 101 n. Kripke,S. 38,40 n., 79-80,105
Harman, G. 37 n., 46 n. Kuratowski, K. 111 n., 113 n.
Harrington, L. 127
Hart,W. 42 large cardinal axioms 135-7, 140-2,
Hausdorff, F. Ill n. 143 n., 146
Hebb, D. 49 n., 50 n., 52 n., 53-4, 55- see also inaccessible cardinals;
8,61,65-7,68-70,73-4 measurable cardinals;
Hellman, G. 36 n., 176 n. supercompact cardinals (SC)
Hempel, C. 10 n., 67 n. Lear, J. 37 n.
Henkin, L. 125 n. Lebesgue,H. 111-13,118,121-2
Heyting, A. 16 n., 22 n. Lebesgue measure 113,114,125—7,
Hilbert, D. 24-5, 116 n., 117 n., 162 133,135,136,138,167-8,176
Hilbert's programme 24—5 Leeds, S. 17n.
Hodes.H. 26 n., 9 I n . Levy, A. 137
Hume, D. 7 see also Fraenkel, A., Bar-Hillel, Y.,
and Levy, A.
idealism 6, 7, 16, 22-3 Lewis, D. 13 n., 14 n., 41 n., 80 n., 96
if-thenism 25-6, 127, 170 n., 176 Locke, J. 12
inaccessible cardinals 127 n., 135—6 logical notion of collection, see classes
indispensability arguments for logical positivism 6 n., 9 n., 10 n.,
mathematical entities 29—31,34— 16 n., 26-8,154
5,47, 58-9, 62-3, 76 n., 82,119- logicism 26-8, 148
20,121,155,159-63,176,177-8 Luce,L. 91 n.
infinity 22-4,78-9,81,82,115,172 Luzin,N. 113-14,118 n., 125,137
instrumentalism 10—11,24
intrinsic justifications 33, 72—5,118— McCulloch,W. 52-3
19,128,139-43 Machamer, P. 50 n.
see also extrinsic justifications; Malament, D. 47 n., 163 n.
intuition Martin, D. 103 n., 124, 126-7, 128,
intuition 31-5, 70-5, 76 n., 77, 78, 130,131,135 n., 137-42, 143 n.,
107,118-19,123-5,126-7,139- 145,146-7,166
43,144,150,174-5,177-9 mathematical notion of collection, see
intuitionism 22-3, 28 iterative conception of set
iterative conception of set 39—40, maths/science analogy, see science/
61 n., 62 n., 84, 100,102^, 123- mathematics analogy
4,128,132-3,134,141,153,156- Maxwell, G. 10 n.
7, 171, 173-7 measurable cardinals 136—7, 138,139,
140, 142, 145
Jech,T. 132 n., 136 n., 142 n. Menzel, C. 88 n.
Jensen, R. 133 Merrill, G. 80 n.
Jourdain, P. 109 n., 116n. metaphysics 4 n.
Jubien, M. 81 n. Mill,]. 6n., 16 n., 67 n.
justified, true belief 36-7, 41 n., 72 Miramanoff, D. 27, 40 n.
modalism 176-7, 180
Katz,J. 59 n. monism, set theoretic 158-9,163-9,
Kaufman, E. 60 n. 175,180
202 INDEX

Monna, A. Ill n. 155, 158, 159, 160, 163, 167-9,

Moore, G. 110 n., 114 n., 116 n., 117, 170, 177, 180
118 n., 120 n., 121 n., 127 n., compromise 34-5, 41, 48, 75, 76 n.,
129 n., 132 n. 107, 120, 146, 148, 150, 169, 176,
Morse, A. 104 n. 178
Moschovakis, Y. 2, 3, 4 n., Ill n., Fregean 26
113 n., 120 n., 126 n., 127 n., Godelian 28-9, 31-5, 75-9, 107,
134-5, 137-9, 140 n. 128, 159,177-8
MycielskiJ. 138 n. mythological 129
Quine/Putnam 28-31,33,34-5,45,
natural collections or kinds 14, 39-40, 46,75,119-20,159,177-8
80, 96-8 traditional 21,30,33,36-7,40-1,
natural numbers 35, 82-94, 97-100, 43-4,48, 59, 78,156, 160,163,
164, 165, 170-3, 174-5,176, 179 172, 177-8,179,180
naturalism 9, 13, 14, 17, 20, 21, 29, see also realism, mathematical;
35, 42-3, 46-8, 58, 78-80, 87, 88, realism, set theoretic
96,175,177-8,181 power set axiom 131,136,141
see also epistemology naturalized projective determinacy 138^0, 142,
necessary, see contingent vs. necessary 145-6
Neisser, U. 50 n. projective sets 113-14, 120, 135, 136-
nominalism 12, 28-9, 43-4,46-8, 96, 9,145-6,147,168,176,177
159-69,176,180 proper classes, see classes
Novikov, P. 114, 137 properties 12-13, 93-4, 96-8, 103 n.,
Nyikos,P. 127 n. 123-4
see also universals
pure sets 59 n., 156-7
objectivity 6, 14, 177 Putnam, H. 8 n., 10 n., 12 n., 13 n.,
ontology 4 14 n., 16 n., 20 n., 25-6, 28-30,
open set 111 34, 38 n., 47, 67, 80 n., 93 n., 96,
operationalism 10, 14 155, 161,169, 176 n., 177, 181
ordinal numbers 115, 130, 157 n. see also Platonism, Quine/Putnam

pairing axiom 67,125 Quine,W. 7-9,11,13,27-8,28-31,

paradoxes 22, 26 n., 62, 71 n., 83, 104, 34, 39 n., 41,93 n., 153 n., 155,
105 169,177,181
Parsons, C. 31, 34 n., 59 n., 84 n., see also Platonism, Quine/Putnam
89 n., 103 n., 105 n., 170 n.,
175 n. realism
perception 31,46,49-67,74,76-7, common sense 6—9, 14, 31—2, 35,
123, 150-3,157, 172-3,178-9 175,177
perfect set 110 mathematical 14-15, 20-1, 177; see
perfect subset property 112, 114, 125— also Platonism; realism, set
7, 130,133,135, 136-7,138,143, theoretic
145,176 pre-theoretic 1-3,15-16,35,45-6,
phenomenalism 6-7, 8, 10, 14, 16, 32 180-1
Phillips,]. 54, 63 n., 64 n., 74 n. scientific 10-11,14,35,129,177
physicalism 5, 35, 46, 154-7,159, 160, set theoretic 35,59, 61, 67, 71 n.,
163,165,169,180 75-8, 81, 86-7, 96,107,118,128,
PiagetJ. 54, 63,74 n., 89 144, 150-4, 156-9, 160 n., 169,
Pitcher, G. 50n.,51 n. 170—5, 178—81; see also monism,
Plato 12,20-1,36,158 set theoretic
Platomsm 20-1,26,28-35,36-7,40- about universals 11-14, 15, 28, 96-
1,42,46-8,81,127, 129,132, 8,164
 INDEX 203
and truth 15-20 structuralism 21 n., 35,170-7,180
real numbers 22,23,35,62,81-2,86, supercompact cardinals (SC) 142-3,
94-5,97,108-13,162,166 n., 144,145,147,148-9,168,169,
167 n., 168 n., 172 n., 173,179 176,177
reference 18-20,37-41 Suslin, M. 113-14,118 n., 125,129-
see also causal theory of reference; 30
descriptive theory of reference Swierczkowski, S. 138 n.
Reinhardt,W. 105 n. synthetic, see analytic vs. synthetic
see also Solovay, R., Reinhardt, W.,
and Kanamori, A. Tait,W. 106 n.
reliabilism 42—3 Takeuti, G. 138 n.
reliability 18-20,43-5, 73,148,159, Tarski,A. 18, 136 n.
178 theoretical justifications, see extrinsic
Resnik, M. 25 n., 36 n., 47 n., 81 n., justifications
91 n., 100 n., 163 n., 165 n., Thomae, J. 23
166 n., 170-3, 175 n., 176 n. transitive closure 59
rules of thumb 141-2,143,146 Troelstra, A. 22 n.
Russell, B. 22,25,26-7,32,76,103, truth 15-20, 43,105, 178
104 see also correspondence theory of
truth; disquotational theory of
Salmon, N. 38 n. truth; verificationism
Schoenflies, A. 116
science/mathematics analogy 2,15, 31-
3,45-6,67-8,75,76-8, 87-8, 93, Ulam, S. 136 n.
98,128 n., 129,146-8,168,178, uncountable 109 n.
181 union axiom 67,125,131
Scott, D. 135 n., 136 unit sets, see singletons
Searle,J. 38 n. universal* 11-14, 21 n., 87, 96-8,164,
separability 112,114,125-7,133, 171,173
137-9,145 see also properties
separation axiom 71 n. unlimited comprehension 71 n., 83,
set theory as a foundation for 117
mathematics 4-5, 81-2, 89, 102- Urmson, J. 7 n.
3,114,119,179
Shapiro, S. 47 n., 162 n., 166 n., 170- van der Waerden, B. 120
3,174,175 n. van Heijenoort, J. 116n.
Shoenfield,]. 39 n., 102 n. verificationism 16-17,23
Sierpiriski, W. 112,113 n., 121, 128 n., von Neumann,]. 27, 84, 104 n.
136 n., 145
Simpson, S. 25 n.
singletons 150-3, 158 Wang, H. 29 n., 40 n., 135 n., 136 n.,
Skolem, T. 27 139 n.
Skyrms, B. 37 n. Wedberg,A. 12 n.
Solovay, R. 127,128 n., 136,137, well-ordering 114-17,121, 125, 133,
138 n., 140 n. 135,137,138,139,143,145,146,
Solovay, R., Reinhardt, W., and 177
Kanamori, A. 142 n. Williamson,]. 113 n.
Steel, J. 140 n., 142 Wilson, M. 12 n., 13 n., 141 n.
Steiner,M. 34,41-2, 82 n., 86 n., Wittgenstein, L. 15 n., 79-80, 97
89 n., 174 Woodin,H. 140 n., 142
Steinhaus, H. 138 n. Wright, C. 26 n., 41 n.
Stewart, F. 138 n.
Strawson, P. 38 n. Yourgrau, P. 88 n.
204 INDEX

Zermelo, E. 27, 40, 84, 102, 116-23, 104 n., 125-8, 132-4, 136, 137-8,
136 n., 157 n., 179 142-3,144,145,146,153 n.,
Zermelo-Fraenkel axioms (ZFC) 27 n., 166-8, 179