*288*
*52*
*12MB*

*English*
*Pages 452*
*Year 1974*

A Metaphysics of

DISCARDED

Elementary Mathematics by Jeffrey Sicha

The University of Massachusetts Press Amherst

Copyright

(()

1974 by the

University of Massachusetts Press All Rights Reserved Library of Congress Catalog Card Number 73-79504 ISBN Number 0-87023-149-9 Printed in the United States of America

r:.· '--I

i)

(/_ 1/

c/

Contents

Preface

ix

Chapter 1:

Introduction

A.

Introduction

1

B.

Nominalism

2

c.

A Platonistic Objection

D.

Numerical Quantifiers and Counting

11

E.

Prospectus

14

Chapter 2:

8

Language, General Terms and Abstract Entities

A.

Introduction

19

B.

Moves, Pieces, and Positions in Chess and Language

C.

21

Abstract Entities, Abstract Singular Terms, and Sellarsian General Terms

37

D.

The Perspicuous Language, SL

45

E.

Sellarsian Quotes and SL

55

F.

Summary of the Main Stages of the Reconstruction

Chapter 3:

65

Quantification, Entailment, and Identity

A.

Quantification

(1)

69

The Rejection of Extended Semantic Formalism

(ii)

The Sellarsian Account of Quantification

(iii)

69

75

More About What the Sellarsian Account is Not

v

99

B.

Linguistic Rules, Entailments, and Necessary Statements

112

(i)

112

Linguistic Rules Again Truth-functionality and

(ii) (iii) (iv)

Entailment

114

Necessity and Entailment

117

Truth ex vi terminorum and 123

Eat ailment

127

Identity and Abstract Entities

c.

The Reconstruction of the Natural

Chapter 4:

Numbers 141

Preliminaries (i) The Goal and Plan of This

A.

(ii)

(iii) (iv)

B.

Chapter Numerical Quantifiers, Other Quantifiers, and Inference

143

DST Elimination

157

DSTs and Truth

160 165

Numbers as Universals (i)

Numerical Quantifiers and 165

Sellarsian General Terms (ii)

Number Universals and 'the 169

number of' (iii) (iv) ( v)

(vi) C.

141

Addition and Order for Number Universals Some Properties of 'plus'

177 and

Further Reconstructions

184

Hultiplication

192

Induction and Succession

206

Numbers as Individuals

240

(i)

Introduction

240

MML Distributive Singular Terms

241

(ii)

vi

Induction and Succession

(iii)

Number Individuals and 'the

(iv) ( v)

NUMBER of'

253

Addition for Number Individuals

256

Multiplication for Number In-

(vi)

261

dividuals Chapte~

5:

Topics Pertinent to the Reconstructions

A.

Logicism

B.

Nominalism and the Existence of

c.

265

Natural Numbers

269

Proof

281

(i) (ii) D.

Formalism

281

More on Proof and Deduction

301

The Identity of Statements, Predicates, and Numerical Quantifiers

Chapter 6: A.

Further Reconstruction

Counting

334

(i)

334

Introduction Some Conceptual Functions of

334

Numerals (iii)

Nonconceptual Functions of Numerals

(iv)

337

Counting and a Conceptual Function of Numerals

344

Numerousness and Object-Language 348

Logic (i) (ii) C.

320

Numerals, Numerical Words, and

(ii)

B.

246

Object-Language Numerical Connectives

348

Multiplication Again

355 367

Classes vii

( i) (ii) D.

The Reconstruction

367

The Russell Paradox

394

Ordered Pairs, Integers, and 398

Rationals Notes

415

Bibliography

424

Index

428

viii

,

I

. Jl

&llltits $3 I .

Preface Very few of the remarks I could make about this book would be helpful to those who have not yet read the book. Thus, out of the possible prefatory remarks, I

confine myself

to ones which are important because of a special feature of my book:

though it is neither a book on logic nor a book on

the philosophy of mathematics, it does treat some issues which are commonly thought to be within the province of logicians and philosophers of mathematics.

Moreover, some tech-

nical topics are in the book because, according to some substantial body of current opinion, my views would not be plausible without some discussion of these topics.

As a conse-

quence, the book does contain a fair bit of technical machinery.

For those who wish to avoid contemplating the crank-

turning that is a consequence of this machinery, I

provide a

guide for a small tour of the book. The itinerary of the small tour is:

chapter l

(section

lE contains a prospectus of the remainder of the book); chapter 2; sections 3A(i), 3A(ii) to the first paragraph of p. 80 and from the first paragraph of p. 85 to the end of the section,

3A(iii), 3B(i), any of 3B(ii), 3B(iii), and 3B(iv) that

strike the reader's fancy,

and 3C; sections 4A, 4B(i), 4B(ii),

4B(iii), 4C(i), 4C(ii), 4C(iv), and 4C(v); sections SA, SB, SC(i) to the bottom of p. 290, and SD; sections 6A and 6B(i). The itinerary of the small tour, I

must warn the reader,

does not exclude all technical passages; but those it does include are generally of a milder sort and are not more, I hope, than are needed to convince the reader that the technical machinery does indeed work when the crank is turned. Support for my work on this book came primarily from an Andrew Mellon Postdoctoral Fellowship at the University of Pittsburgh during the academic year 1967-68.

Subsequently,

money to defray secretarial costs came from Ford Foundation funds. ix

I owe the staff of the University of Massachusetts Press sincere thanks for their patience with me and their excellent work on this book.

In particular, I thank the editor, Janis

Bolster, and the typist, Nonny Burack.

The reviewers of the

book also deserve thanks for their helpful and penetrating comments. Professor William Kneale gave me much help on the parts of the material of the book which appeared in earlier form in my doctoral dissertation.

Bruce Aune and Jacqueline Thomason

read through several late drafts of the book and provided detailed comments which led to what I believe are great improvements.

In addition, I owe Bruce Aune, my friend and former

teacher, much more for his help and our philosophical talks over the years.

To Wilfrid Sellars I have a debt manifest

throughout the book; so great is this debt that only the whole book, rather than any brief words I might write here, can express it.

X

Chapter 1:

A.

Introduction

Introduction The appearance, roughly seventy years ago, of paradoxes

in the foundations of mathematics produced three main philosophical responses:

formalism, logicism, and intuitionism.

Each of these philosophies of mathematics has, it is claimed, run against criticisms that have necessitated abandoning them as they were originally conceived.

For example, it has been

said that Godel's proofs have written an end to the original program of formalism, that Brouwer's formulation of intuitionism is founded on a nearly unintelligible neo-Kantianism containing unpurged elements of psychologism, and that logicism has a variety of inadequacies ranging from problems with such principles as the axioms of reducibility and infinity to unacceptable platonism and obscurity on the nature and extent of logic.

This much of the recent history of the philosophy

of mathematics is familiar to most philosophers.

But, it

seems to me, the present situation in the philosophy of mathematics is far from clear to the philosopher dealing with general philosophical problems. There are, perhaps, several different reasons for this unclarity.

First, while Russell and Frege, Brouwer, and to a

lesser extent Hilbert discussed their philosophies of mathematics at length and framed their positions in the philosophy of mathematics in light of views they held in more general philosophical areas such as epistemology and metaphysics, many of the discussions of logicism, intuitionism, and formalism which have followed theirs have dealt primarily with technical matters in logic and mathematics and with the philosophy of mathematics little or not at all.

Moreover, many later

writers who do have philosophies of mathematics have not discussed relevant general philosophical problems at sufficient length; for example, few have expounded their position on the l

lB general problem of abstract entities in enough detail that one could determine how their views on this problem relate to their views on the nature of numbers.

Second, the post-World

War II philosophical climate has not been congenial to systematic philosophy.

And, of course, a writer must have some in-

clination toward systematic philosophy even to attempt an investigation of the relationship of his philosophy of mathematics to his general philosophical views. In part, this book is an attempt to rectify the situation just explained; in part, it is not. a survey of logicism, formalism,

It does not contain

and intuitionism, nor does it

try to elaborate these views and to sort and evaluate the criticisms laid against them.

But it is meant to be an ex-

ample of the derivation of views in the philosophy of mathematics from more general philosophical views.

Of course, the

book does not discuss everything that, in the long run, needs to be discussed; it does not even try to present a complete philosophy of mathematics.

The treatment of both general

philosophy and the philosophy of mathematics is limited to logical and metaphysical issues; epistemology is almost completely ignored.

Further, of the philosophy of mathematics,

that part which might be called the philosophy of elementary mathematics is discussed at length; of the philosophy of the real numbers and of other matters only glimpses are present.

B.

Nominalism To set the stage for the views I shall present in later

chapters, I shall now undertake a brief investigation of an issue that arises in the context of a contemporary philosophy of mathematics which is akin to traditional formalism and which has the merit of being firmly rooted in general philosophy. 2

lB Its basic tenets can be succinctly set forth:

Certain

sentences of mathematics are to be regarded "merely as strings of marks without meaning"; our understanding of these mathe,, matical sentences is given solely by the "syntactical or meta1 mathematical rules governing these marks." The sentences that are to be so regarded contain variables "calling for abstract entit"ies as values" and cannot be "translated into nom2 inalistic language." These sentences comprise the part of mathematics which is "platonistic"; by regarding them solely as strings of marks one avoids any "question of truth" while leaving oneself with a syntactical account of their useful3 ness as "convenient computational aids." One of the familiar notes struck by this view is the notion of a computational aid or device.

Mathematics is lik-

ened to an apparatus such as the abacus which, by virtue of its material structure and composition, can help us make computations. theories.

Similar remarks have been made about scientific Thus there have been philosophers of science who

termed their position "instrumentalism" and maintained that the primary function of scientific theories is to systematize and manipulate meaningful empirical statements about what is observable.

To this end scientific theories have a

describable structure provided in part by ordinary syntax and logic and in part by mathematical syntax and contain "strings of marks" which stand in various syntactical relationships to other strings of marks.

Certain marks of this vast mean-

ingless apparatus of theory are connected with observation language expressions which are meaningful.

By moving from

the observation language to the theory, utilizing the syntactical and mathematical relationships in the theory,

and final-

ly returning to the observation language, we use the theory as an instrument.

The parallel thesis for mathematics is 3

lB that the language of mathematics aids us in systematizing and manipulating statements about quantities and measures in the material world. While this comparison between the instrumentalist position and the philosophy of mathematics indicated in the quotations may do a great deal to make us feel at home philosophically, it also raises several questions:

Are there state-

ments which stand to the apparatus of platonistic mathematics as statements about the observable stand to a scientific theory?

How are these statements to be delimited?

And can this

be done without recourse to mathematical notions? The answer to the first question is affirmative.

Some

mathematical formulae are nominalistically meaningful; this is attested by the fact that they can be ''translated into nominalistic language" where a nominalistic language is one whose mechanisms of reference permit reference only to concrete objects.4

Such formulae would provide the liaison between the

mathematical apparatus and otner nonmathematical statements about concrete objects in the most direct manner:

namely, by

being at once about concrete objects and yet in the apparatus of mathematics and hence connected by syntactical rules and logic to nominalistically uninterpretable formulae. The most obvious candidates for statements which are in some sense mathematical and which can nevertheless be translated into nominalistic language are what I shall call "numerical quantifier statements"; the simplest numerical quantifier statements are of the following forms: (l)

there is one K, there are two Ks, there are three Ks,

and so on.

Each of these statements is equivalent to a state-

ment containing only ordinary quantifiers, connectives, vari-

4

lB ables, and identity.

Thus

there are two Ks is equivalent to ( 2)

(Ey)(Ex)(Kx and Ky and not(x Kz, then x

y) and ( z) (if

= z or y = z)).

Provided that 'Kx' is true of concrete objects,

(2) is ac-

ceptable in a nominalistic language. This success suggests further nominalistic translations. Consider (3)

5 + 3

= 8.

Among the various interpretations of (3), there is one which is nominalistic and which has the advantage of using numerical quantifiers. (4)

Thus (3) becomes, roughly,

there are five things of a sort and there are three other things of the same or another sort if and only if there are eight things of some sort or other.

Let us, for the present, accept (4) as the basis of a nominalistic interpretation of

(3)

and thus ignore certain obvious

questions (such as the precise logical form of (4)).

What

has been done for (3) can also be done for ( 5)

4

X

3 = 12 •

Thus a roughly formulated nominalistic interpretation of (5) is (6)

there are four things of a sort for each of which there are three things of the same or another sort if and only if there are twelve things of the latter sort.

Once again let us ignore any deficiencies and press on in our 5

5

lB co~parison

of the instrumentalist view of scientific theories

and the nominalist philosophy of mathematics. Having come this far, the nominalist philosopher is in a position to make a bold stroke which has the dual effects of supplying a familiar sort of epistemology for his view of mathematics and completing the analogy with instrumentalism. He claims that statements such as (4) and (6) formulate truths 6 about correct counting. For example, if o~e counts that there are five entities of one sort and that there are three other entities of the same or different sort, then from (4) one knows that correctly counting all the entities would lead to the result that there are eight.

Further, he points out

that statements of the forms illustrated in (1) formulate the results of counting.

Hence, these basic types of nominalis-

tically interpretable mathematical statements are intimately connected with an activity that, in fundamental cases, enables us to gain knowledge about concrete objects.

As observing

and observation reports provide the epistemological foundation for scientific theories, so counting and statements of the types illustrated above anchor that great nominalistically uninterpretable part of mathematics in discourse about coDcrete objects and provide the link that allows the uninterpreted mathematical formulae to perform their function as instruments. Of course, this move would fail if couDting in some way involves numbers in a seDse of the word 'number' in which numbers are abstract entities.

This failure would be analo-

gous to the failure that the instrumentalist view of scientific theories would suffer if it should prove that the epistemological viability of observation concepts dep'ends on their being part of a total system of concepts which necessarily includes theoretical concepts. 6

But such a failure in

lB the present case seems unlikely, since there is a popular view of counting which fits the nominalist's needs exactly. This view of counting, suggested by the remarks of con.

temporary phllosophers,

7

can be roughly stated as follows:

counting is the activity of putting initial segments of serially ordered strings of entities in l-1 correspondence with the entities of some collection.

In the most familiar sorts

of cases, counting is no more than tokening (e.g., uttering) a string of numerals or numerical words while, say, pointing at the entities in some collection (e.g., books on a table). There are good reasons to think that such an activity need not be understood as in any way involving numbers as abstract entities.

First, according to this view of counting, it is

not even true that one must count with numerals or numerical words though in fact this is commonly the case. any string of serially ordered expressions.

One could use

Thus taking ad-

vantage of the historically produced serial order of the letters of the alphabet from 'a' to 'j'

inclusive, we could count

with the following serially ordered collection of expressions ( 7)

b, c, d, •

•

•

I

bac, •

•

•

I

e, f, g, h,

ca, •

•

•

•

I

... ,

•

•

da,

I

i,

... ,

•

bba, bbb, bbc, •

•

•

I

ba, bb, be, bd,

j

I

•

•

I

•

•

•

•

I

•

•

I

baa, bab,

bca,

•

•

•

I

bda,

caa, ••..

Learning such a serially ordered collection as (7)

can be ac-

complished by the sort of repetitive training that children do in fact undergo in learning our common numerical words. Clearly, the nominalist says, such training in no way involves or presupposes numbers as abstract entities; or else how could children who are obviously without mathematical concepts learn to count?

Second, the notion of 1-1 correspondence which ap-

pears in the characterization of counting can be defined in purely logical terms and hence need not involve anything 7

::::..:--. "'··

.

.

'

...

lC which might be taken to be mathematical abstract entities. Thus the nominalist's account of methematics comes to rest heavily on an apparently plausible view of counting. Counting, as he presents it, is the activity which provides the epistemological foundation of those mathematical statements like "there are three cows (in the field)" which can be interpreted nominalistically and yet which are part of the On the other hand, counting is not a mathematical apparatus. mathematical activity in any sense of the word 'mathematical' which would provide solace to platonistic philosophers.

C.

A Platonistic Objection The nominalist's view of mathematics is, of course, not

acceptable to philosophers with so-called platonistic turns of mind, and it is not difficult to locate a point at which the platonist might strike.

Let us consider the following,

perhaps not entirely clear, yet forceful,

line of thought.

The most obvious weakness, the platonist claims, is in the nominalistic treatment of counting which has covertly assumed in the characterization of counting a mathematical fact which is not translatable into a nominalistically acceptable fact about concrete objects.

We must remember, the platonist

insists, that for the nominalist counting is an activity in the material world; part of this activity is the actual production of concrete objects (e.g., sounds) which are put in 1-l correspondence with other concrete objects.

It is true

of natural numbers, as working mathematicians commonly understand them, that if one correlates 1-l the objects of some collection with the natural numbers less than or equal to 5, then there are five objects in the collection that is being counted (note that the natural numbers are assumed to begin with l).

However, this truth rests on a mathematical fact: 8

lC

for any natural number n the number of natural numbers less than or equal to n is n. But, the platonist contends, there is no analogous truth about concrete objects (e.g., sounds).

For example, from the

fact that someone has produced, say, a token of the numerical word 'five' it does not follow that he has produced five tokens altogether, each of which is a token of one and only one of the numerical words which precede the numerical word 'five' in the conventional sequence of numerical words. the platonist concludes,

Therefore,

a correct characterization of count-

ing must, whatever else it contains, speak of the 1-l correlation of objects with initial segments of the natural numbers where the natural numbers cannot be concrete objects, like sounds, which might or might not be produced correctly. While there is something to be said for the platonist's objection, his argument does not warrant the conclusion he draws.

We can agree that linguistic objects and other con-

crete objects are not similar, in an important way, to the natural numbers.

That is, simply from the fact that someone

has produced a token of the numerical word 'five' it does not even follow that four other tokens have been produced, much less that the four other tokens are tokens of different numerical words, those numerical words which precede 'five' in the conventional serial order.

Such considerations do indeed

show that the nominalistic view of counting is inadequate in the form in which I have presented it.

What the nominalist's

view of counting explains in its present form is a method by which we can find out, if we proceed correctly, that there are as many entities in one group as there are in another. In order to be able to draw any conclusion of the form "there is(are)

Ks," we must have an additional statement telling

how many entities one of the groups contains. 9

Obviously what

lC we must have is a statement saying how many tokens have been produced by the person who is counting.

Of course, it fol-

lows that we cannot require that our knowledge of such facts about the production of tokens always be obtained by counting. Such a requirement would lead to a regress.

However, know-

ledge that, for example, Jones has produced, while counting, one and only one token of each numerical word beginning with the numerical word 'one'

and ending with the numerical word

'five' can rest on facts about the previous linguistic training of Jones, the present absence of factors which might disrupt the abilities he acquired through training,

and so on.

From this vantage point, we can reconsider the platonist's view of counting and note that it too is inadequate and for much the same reason.

The platonist suggests that count-

ing is the activity of producing 1-l correspondences between entities of some group and an initial segment of the natural numbers understood as abstract entities.

But in many cases

of counting the entities to be counted are concrete (e.g., ordinary physical objects).

Establishing a 1-l correspond-

ence between concrete entities and the natural numbers which are,

according to the platonist, not concrete can only be

done by the mediation of yet other concrete objects.

Tokens

of numerals and number words are the obvious candidates. They are concrete and nevertheless stand in a special 1-l relation to natural numbers that other concrete objects do not: namely, they are, according to the platonist, the names of natural numbers.

Thus the 1-l correspondence between the ob-

jects to be counted and the natural numbers is a product of two 1-l correspondences, one between the numerals or number words and these objects and one between the numerals or number words and the natural numbers. leaves a question to be answered: 10

But this scheme also Has the person who is

lD counting produced the names of an initial segment of the natural numbers?

He could, for instance, have omitted one.

Thus even on the platonist's view of counting we must,

in or-

der to gain knowledge of "how many" by counting, know such facts as, for example, that there were five tokens of names of natural numbers which were produced by someone who was counting, and each of them was different from all the others, and they were tokens of names of natural numbers in an initial segment of the natural numbers. Given this criticism of the platonistic view, the nominalist is in a position to claim that he need not forsake his nominalism, but simply amend his view of counting by recognizing the importance of numerical quantifier statements about linquistic entities.

Though these numerical quantifier statements

are essential to the activity of counting in a way in which he had not previously envisaged, he is not, he claims, driven to any form of platonism since, like all numerical quantifier statements about concrete objects, those about linguistic entities are nominalistically interpretable.

With this amendment

the nominalist appears to be entitled to claim that he has shored up his original view without changing it significantly.

D.

Numerical Quantifiers and Counting One of the original strengths of the nominalist's phil-

osophy of mathematics as here presented is that the whole of mathematics appears to rest on counting, an activity which in no way involves abstract entities.

Counting, the nominalist

maintains, is an activity which requires the production of natural objects (usually linguistic) and which, though it presupposes a great deal of "know-how," no more utilizes natural numbers in any platonistic sense than other natural activities requiring "know-how" (e.g., swimming). 11

And although

lD

a reflective understanding of counting necessitates understanding the notion of 1-l correspondence, this notion is definable in purely logical terms.

Thus the nominalist seems

to be able to assure us that the fundamental epistemological means by which mathematics comes in contact with the empirical world is nominalistically unexceptionable. It is no coincidence that philosophers who have nominalistic inclinations and whose remarks suggest they subscribe to the nominalist view of counting are tempted to think that any serially ordered sequence of entities whatsoever could be called natural numbers if, indeed, any entities whatever need be called natural numbers. 8

After all, these philosophers

have thought, all that is required for counting is a serially ordered string of entities.

Since these serially ordered

strings of entities are the only entities figuring essentially in the activity of counting and since counting is the fundamental activity producing nominalistically interpretable mathematical knowledge about concrete objects, we are on solid ground in suggesting that these strings of serially ordered entities are the natural numbers. However, as my reply to the platonistic objection demonstrates, matters are not so simple.

My reply concedes to the

platonist that for a person who is counting to arrive at an answer to a "how many" question by counting, he must not only bring about a 1-l correspondence, but he must also know a numerical quantifier statement about the entities he is putting in 1-l correspondence with the objects he is counting.

It is

not just that no one who literally does not understand numerical quantifiers can ever obtain answers (true or false) to "how many" questions by counting, since this point also follows from the fact that the results of counting are expressed by numerical quantifier statements.

The point is rather that

"know-how" with serially ordered strings of entities is not 12

lD all that is essential to the activity of counting Cas distinct from the statement of results).

And this fact must

surely raise a doubt about the line of thought that leads to the suggestion that any serially ordered string of entities would do as the natural numbers.

This line of thought seems

plausible because it presents the serial orderedness of strings of entities such as numerals as the sole feature essential to counting.

Now, however, we know that the very activity of

counting requires an understanding of numerical quantifiers. At this juncture it would be well to remember a point that the nominalist cites in demonstrating the freedom of counting from a platonistic tinge:

any string of serially

ordered entities and not only the familiar one, such as the numerals and numerical words, can be used in counting.

Thus

even (7), the string of serially ordered letters of the alphabet, can be used in counting, though these letters seem to have hardly any connection with existence and difference, the two features which are essential to "numerousness."

Reflect-

ing on this point in conjunction with the remarks of the last paragraph and noting that, because of their definitions, numerical quantifiers are essentially connected with "numerousness," one might suggest that if anything were to be called natural numbers, it should be numerical quantifiers and not just any string of serially ordered entities. 9 Certainly many philosophers would agree that numerical quantifiers are essential to counting and even that it is numerical quantifiers, rather than just any serially ordered entities, which capture the basic features of "numerousness" (i.e., existence and difference).

Such agreement notwith-

standing, these same philosophers would be little inclined to accept what must appear to them a very odd suggestion: that numerical quantifiers are natural numbers.

viz.,

Moreover,

for some, such a suggestion would seem far too restrictive l3

lE and without the faintest hope of providing an account of the commonly accepted features of the system of natural numbers. Others, fearing the spectre of platonism, will wonder whether a treatment of the identity of numerical quantifiers (a treatment that surely must be provided) will not smuggle in platonism by having to appeal, in the long run, to a traditional account of synonymy.

Yet others, having a deep-rooted dis-

trust of logicism, will note that it is too important to be a coincidence that the expressions offered as candidates for natural numbers are definable solely in terms of the usual quantifiers, connectives, and identity. Certainly, even more substantial philosophical doubts could be formulated.

Despite all this, this suggestion about

numerical quantifiers is the one I propose to elaborate.

In

the course of developing this suggestion, I shall take up matters which bear on the philosophical themes of this chapter. The road will be long, not simply because the problems are difficult, but also because the avowed aim of this book is to set forth a piece of systematic philosophy which, as far as it is possible, derives a philosophy of elementary mathematics from more general philosophical theses.

For this reason,

I spend the next section describing the strategy and topics of the remaining chapters and the goal of the book.

E.

Prospectys The main aim of this book, as the title suggests, is to

provide a metaphysics of elementary mathematics. book has vey little to say about epistemology.

Thus the It also has

little to say about what branches of ordinary, intuitive mathematics should be counted as "elementary."

I do not know

that it is even possible to find an explanation of 'elementary' which would be satisfied by all and only those branches 14

lE of mathematics which fit some intuitive conception of what elementary mathematics is.

But, it seems to me, such botan-

izing would, if possible, be almost without interest.

So,

for the purposes of this book, elementary mathematics is taken to be the arithmetic of the natural numbers, integers, and rationals, and that part of the theory of classes which eschews the more exotic principles of that subject and which

' treats of the basic principles of the existence and identity of classes and the operations on classes.

What exactly must

be accomplished in the discussion of each of these pieces of mathematics will be explained when they are taken up. A secondary aim of the book, as I have already said, is that it should be a piece of "systematic" philosophy.

There

are, of course, a variety of senses of the word 'systematic' in which the book is not systematic.

For example,

a book

could be said to be systematic if the topics and problems treated in the book are characterized at the outset in such a way that one is able to determine the contents of the book from this characterization (plus, perhaps, a few collateral premises about the nature and practice of philosophy).

This

book does not even attempt to be systematic in this sense. There are two systematic features of this book.

First,

the philosophy of elementary mathematics presented is, for the most part, derived from much more general philosophical views.

The main example of this in the book is the derivation

of my account of the natural numbers from a more general account of abstract entities. Second, I attempt to forestall a number of possible objections by discussing certain topics not immediately related to those topics which must be discussed to accomplish the main aim of the book.

My criterion for including a topic of

this sort is the likelihood of a substantial group of philosophers thinking (rightly or wrongly) that my position on the 15

lE

central topics of the book would, in whole or in part, be vitiated if this subsidiary topic were not discussed.

Of course,

there is no one group of philosophers to whom all the subsidiary

to~ics

would seem important.

Not all of these subsidi-

ary topics can be considered in the detail they deserve.

But

I hope that on any subsidiary topic I have said enough to make my whole enterprise credible to a reader who thinks that subsidiary topic crucial. see~

A few subsidiary topics, those which

to me most likely to cause confusion about the central

topics of the book, are given considerable space. Since the purpose of this introductory chapter is both to introduce numerical quantifiers with an example of their importance and to illustrate a case in which my view avoids both a nominalistic doctrine and a platonistic objection, I wish, before turning to a short description of the highlights of the coming chapters, to return to the points of this chapter by making some very general remarks about my account of elementary mathematics.

This account,

and the more general

account of abstract entities from which it is derived, I prefer to describe as a "mild" nominalism.

My choice of the word

'mild' will become clearer as the book progresses.

But what

can be said now, at the stage at which the reader knows almost nothing about the view, is that my account takes the language of elementary mathematics as a full-fledged part of language and does not try to relegate it to second-class standing as an "instrument."

Nor do I worry about any strict standard of

nominalistic acceptability such as the one described in section B.

More than that, I accept that the statements of ele-

mentary mathematics are about numbers; according to the view I propose, it is, properly understood, true to say that natural numbers exist. However, these admissions must be tempered by the observation that, on the view I propose, the language of mathemat16

lE ics is a very special sort of discourse and talk about numbers is very special talk.

On this point nothing much can be said

in advance of actually presenting the view.

But the general

strategy followed in the remainder of the book is adherence to a course between strong nominalisms which hold that mathematical statements either must be construed as ordinary statements about· concrete, material objects or else must be classified as something less than ordinary meaningful discourse and those views which have been called "platonistic" and which maintain that mathematical statements are perfectly good parts of discourse that speak of those very special nonlinguistic, nonnatural objects, abstract entities. Of course, the preceding sections have given only one illustration of the nominalist-platonist controversy and have not characterized nominalism or platonism.

So, it is open to

any reader, if he wishes, to characterize these two opposing positions in such a way that there is no course to steer between them or to locate them in philosophical waters different from those where I locate them.

I do not wish to argue about

what to call my view; as I said, I prefer "mild" nominalism. If the reader understands the view and how it differs from other views on the subject, I will be content to let him call it what he likes. The structure of the book is built according to the strategy I have described.

Chapter 2 discusses a philosophy

of language which eventually leads to an account of abstract entities.

The source of the doctrines in this chapter is the

writings of Professor Wilfrid Sellars.

The most important

issues in part A of chapter 3 are those surrounding variables and quantification.

These issues cannot be avoided because

of the now very common philosophical inclination to connect ontological commitment with the employment of the mechanism 17

lE of quantification.

To philosophers with this inclination the

view of abstract entities sketched in chapter 2 could not be what it claimed to be.

Part B of chapter 3 gradually moves

into the topic (only partially discussed there)

of identity

and the entities involved in the account of chapter 2.

This

topic is taken up again later and the discussion of part B, chapter 3, provides the basis for an account of identity and natural numbers. Chapter 4 derives from the material of chapter 2 an account of the natural numbers.

After chapter 4, a variety of

obvious questions remains to be answered:

Can the account of

chapter 4 be extended, in some consistent manner, to encompass the integers and the rationals and the elementary parts of the theory of classes which have commonly been used in class-theoretic accounts of the natural numbers? count of chapter 4 a form of logicism?

Is the ac-

Is a formalist account

of theorem and proof suitable to the account of chapter 4? What view of counting is available to my view?

All these

questions and other that arise because of special features of the material of chapter 4 are taken up in chapters 5 and 6. A final warning, before I turn to the business of chapter 2, is in order.

This book does not attempt to introduce

mathematical or logical novelties.

There are no contribu-

tions in it to mathematics or to mathematical logic as sucho Those who look for such things will be disappointed. novelties there are, if there are any,

What

are philosophical,

though they are directed toward matters which involve mathematics and logic.

18

Chapter 2:

A.

Language, General Terms,

and Abstract Entities

Introduction Recent philosophy has made much of an analogy between

games and language.

This analogy has not, I think, always

been carefully drawn nor have the grounds for thinking that there is an analogy of an appropriate sort been well argued. Therefore, since I

intend to make extensive use of the exam-

ple of chess, I shall try to forestall any misunderstanding by setting out in the next section that analogy between games and language which is important for the work of this chapter. Essential to the analogy is a thesis I shall assume without argument:

in specifiable respects, both language and games 1 are activities governed by rules. Hence, before I can indicate the extent of the analogy between games and language, I must discuss which rules I have in mind when I say "rules of language" and the respects in which language is so governed. The topic of rules is vast and difficult and I shall do no more than present a few basic distinctions.

My treatment

follows very closely that of Professor Wilfrid Sellars, to whose works the reader can turn for a more extensive discus.

slon.

2

I shall say that a rule is something which permits, obliges, or forbids, not particular instances of repeatable activities, but the repeatable activities themselves. rules are general. linguistic activity.

Thus

There are a variety of rules which govern Some of them govern actions in the

proper sense of this word, the sense which, following Professor Sellars, I shall call the "conduct" sense. Thus there are rules pertaining to promising. 3 But there are other rules which treat of linguistic actions only in that weak sense of 'act' in which anything spoken of by a verb in the active voice is an action.

Thus a rule of this second sort, roughly 19

2A formulated,

is

(l)

in standard conditions, one may respond to red objects by tokening this is red.

Another example, once again only roughly formulated, (2)

is

one may respond to a tokening of this is red with a tokening of this is extended.

Closely related to these rules are ones which do govern actions in the conduct sense of 'act'. (3)

Thus there is the rule

one ought to bring it about that in standard conditions, language users are disposed to respond to red objects by tokening this is red.

Similarly, there is the rule (4)

one aught to bring it about that language users are disposed to respond to tokenings of this is red with tokenings of this is extended.

Such rules as (3) and (4) govern the actions of those sophisticated language users who teach the language to novices. I shall not be much concerned with the perpetuation of language through teaching.

Thus all the rules I shall formu-

late will be ones like (l)

and (2) which, again following Pro-

fessor Sellars, I shall call "rules of criticism" (as. opposed 4 to (3) and (4) which are "rules of conduct") . Further, concerning rules of criticism my main concern is with rules like ( 2) •

And,

as

~

first approximation of what I shall require

20

2B

later, I shall write (2) and rules like (2) in the following fashion:

(5)

one may infer tokens of 'this is extended' from tokens of 'this is red',

or

( 6)

tokens of 'this is extended' may be inferred from tokens of 'this is red'

Thus inference in the sense in which it is introduced in (5) and (6) is not an action in the conduct sense, though it is in the weak sense an act. It is at this point that an investigation of the analogy between language and games furthers our understanding of rules of criticism and language.

As the heart of this investiga-

tion, I answer the question: positions of language?

What are the moves, pieces, and

Through answering this question and

related questions, we acquire the resources with which to pursue a deeper investigation of language and to provide an account of abstract entities.

B.

Moves, Pieces, and Positions in Chess and Language What sorts of linguistic resources are needed in order

to be able to speak of moves, pieces, and positions in chess? The linguistic resources needed to be able to speak about the pieces of chess are those of ordinary language (e.g., English) plus the general terms 'pawn',

'bishop', etc.

The moves in

chess are those of moving the pieces, or counters, of the game and they can be spoken of by using exactly the same resources used in talking about the pieces.

To talk of the positions

we need the same resources plus such general terms as 'row', 'column', etc.

With this vocabulary, we can not only make re-

marks about particular moves, pieces, and positions of chess, but we can also formulate the rules of chess. 21

For instance

2B there is a rule that one may move a pawn only one square forward in a column in any move after the first except when capturing another piece.

This rule governs a player's manipula-

tion of any counter that is a pawn.

Derivatively, we can

speak of this rule as being about the pawn itself (with reference to the players suppressed, as in the rendering, "The pawn may be moved only one square.

.").

Let us extend this game terminology to language by deciding what in language are analogous to moves, pieces, and positions.

The most natural decision would seem to be that

the analogs of moves are inferrings.

However, the word 'in-

fer' was introduced above so that inferrings are responses to items of behavior with other items of behavior. these items of behavior "tokenings."

Let us call

Given the decision to

say that inferrings are the moves of language, it seems that the tokenings ought to be the pieces.

But while one can

straightforwardly say that chess players move pieces, it is not very clear what would be meant by saying that language users move tokenings.

The solution is that we have overlooked

the fact that chess players move pieces "from position to position."

Clearly it fits our intuitions to say that inferrings

are moves from one position to another.

Thus the tokenings

qua items of behavior having a place in the rule-governed structure of linguistic behavior mark the positions of language as the units of the chess players' board mark the positions of chess pieces; or, as we shall see, it is more accurate to say that a tokening marks a position in virtue of being of a certain type which marks a position in the rule-governed structure. What is it, then, the language user.

in language that moves?

The answer is

By inferring, the language user "moves

from position to position."

The tokening which is inferred

marks the position which the language user has come to occupy. 22

2B His position is specified by the rule-governed structure in terms of the types of inferences he is permitted to make to and from a type he has tokened.

My later discussion of types

will help to clarify this way of drawing the analogy. The terminology thus far established leaves several issues open.

Consider the question:

occupy more.than one position?

Does the language user

Both answers to this question

are, apparently, acceptable. On the one hand, one could say yes.

Presumably, then,

the language user occupies all those positions which are marked by types which he is presently disposed to token. Further, in most inferences the language user moves from several

position~

to another position.

In addition, by moving

from a position or positions the language user does not necessarily "abandon" the positions he moves from (though, of course, he might abandon them).

A person may still be, and

usually is, disposed to token a type that he has used as a premise for an inference.

Thus, in this respect, linguistic

activity is unlike chess because the chess player does "2.bandon," in some sense of 'abandon', his position as he moves a piece. On the other hand, the question could also be answered no.

Then I should have to alter my earlier terminology a lit-

tle to say that types do not mark positions, but rather mark "points" or "nodes" in a position.

The position the language

user occupies would be the total array of "points" which are marked by types the language user is disposed to token.

This

way of talking about linguistic activity is analogous to talking about the chess player as having pieces being in

~

position.

~

position or all his

The position, in this way of

talking, is understood to be the total array of the player's pieces on his board; all the pieces, at one time, are in the same position, though each is at a different "point" in the 0. fH) C) l.J ~

..; .j 23

.~

~

1 ~i

2B In the case of linguistic activity, all inferences position. are from one position, viz., the position the language user occupies at that time.

In addition, each inference, in adding

one new "point," insures that the language user "abandons" those "points" which are marked by the types which are the premises of his inference. No doubt these issues, and many others, could be profitably discussed at length.

But such discus: