Mathematics and Reality

Mathematics and Reality
Author(s): Stewart Shapiro

Source: Philosophy of Science, Vol. 50, No. 4 (Dec., 1983), pp. 523-548
Published by: The University of Chicago Press on behalf of the Philosophy of Science Association
Stable URL: http://www.jstor.org/stable/187555 .
Accessed: 19/10/2014 15:14
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
The University of Chicago Press and Philosophy of Science Association are collaborating with JSTOR to
digitize, preserve and extend access to Philosophy of Science.
http://www.jstor.org
This content downloaded from 140.254.87.149 on Sun, 19 Oct 2014 15:14:13 PM

All use subject to JSTOR Terms and Conditions
Philosophy of Science
December, 1983
MATHEMATICS AND REALITY*
STEWART SHAPIROt
Department of Philosophy
Ohio State University of Newark
The subject of this paper is the philosophical problem of accounting for the
relationship between mathematics and non-mathematical reality. The first sec-
tion, devoted to the importance of the problem, suggests that many of the rea-
sons for engaging in philosophy at all make an account of the relationship be-
tween mathematics and reality a priority, not only in philosophy of mathematics
and philosophy of science, but also in general epistemology/metaphysics. This
is followed by a (rather brief) survey of the major, traditional philosophies of
mathematics indicating how each is prepared to deal with the present problem.
It is shown that (the standardformulations of) some views seem to deny outright
that there is a relationship between mathematics and any non-mathematical real-
ity; such philosophies are clearly unacceptable. Other views leave the relation-
ship rather mysterious and, thus, are incomplete at best. The final, more spec-
ulative section provides the direction of a positive account. A structuralist
philosophy of mathematics is outlined and it is proposed that mathematics ap-
plies to reality though the discovery of mathematical structures underlying the
non-mathematical universe.
Introduction. It can hardly be denied that one must master a consider-

able amount of mathematics in order to comprehend the non-mathemat-
ical universe scientifically. Consider, for example, the amount of math-
ematics presupposed by any branch of natural or social science. A basic
assumptionof this article is that there is a relationship between the subject
matter of mathematics (whatever it is) and the subject matter of science
*Received May 1983.

tI would like to thank Mike Resnik, John Corcoran, Robert Kraut, Mark Steiner, and
Bill Lycan for many useful comments on previous versions of this paper and for encour-
agement to continue the project. I would also like to thank, as groups, the members of
the Kenyon Symposium, the Ohio State University Philosophy Colloquium, the Buffalo
Logic Colloquium, and the Hebrew University Philosophy Colloquium for devoting a ses-
sion to this project and for providing many thoughtful and useful comments.
Philosophyof Science, 50 (1983) pp. 523-548.

Copyright? 1983 by the Philosophyof ScienceAssociation.
523

524 STEWART SHAPIRO
(whatever it is) and, in particular, that it is no accident that mathematics

applies to reality. My purpose is meta-philosophical: to discuss the prob-
lem of accounting for this relationship.
Section I deals with the importance of this problem for philosophy. A
description of the subject matter of mathematics is among the main prob-
lems in the philosophy of mathematics, and a description of the scientific
world is among the main problems in epistemology/metaphysics in gen-
eral and philosophy of science in particular. It is suggested that many of
the reasons for engaging in philosophy at all make an account of the
relationshipbetween mathematicsand reality a priorityin the above fields.
Any world view which does not provide such an account is incomplete
at best.
Section II contains a survey of the major traditional philosophies of
mathematics, indicating how each is prepared to deal with the present
problem. It is shown that the standardformulations of some views seem
to deny outright that there is a relationship between mathematics and any
non-mathematical reality. Such philosophies are, for this reason, unac-
ceptable. Other views may not be prima facie incompatible with a rela-
tionship to reality, but they leave the relationship rathermysterious. Such
philosophies are, for this reason, incomplete. This section is rather brief
since it amounts to a coordinated criticism of four much maligned views.
The final, more speculative section provides the direction of an account
of the indicated relationship. I sketch a view of mathematics as the study
of patterns or structures and propose that mathematics applies to science
through the discovery of mathematical structures underlying the non-
mathematicaluniverse. The article closes with a comparison of this view
with the one presented in HartryField's provocative Science without num-
bers (1980).
A friend once told me that, during an experiment in a physics course,

he noticed a phenomenon (outside the scope of the lesson) that puzzled
him. When he asked for an explanation, the lab instructor wrote an in-
tegral-termon the blackboard and said that the phenomenon occurred be-
cause the function defined by the integral has a zero at a particularvalue.
My friend noted to me that he became even more puzzled that the oc-
currence of a zero in an integral function should count as an explanation
of a physical event, but, in the context of a physics course, he could not
press the matter further. On his behalf, I would like to press the matter
further.
This example indicates that a scientific explanation of a physical event
often amounts to no more than a mathematical "description". Indeed,
much of the theoretical work in science consists of constructing (or dis-

MATHEMATICS AND REALITY 525
covering) mathematical models of physical phenomena. Many scientific

(and engineering) problems are described as tasks of finding the right
differential equation, the right formula, the right function, or the right
integral to be associated with a class of phenomena.
Strictly speaking, a mathematical description, model, structure, theory,
or whatever, cannot serve as an explanation of a non-mathematical event
without an account of the relationship between mathematics per se and
scientific reality per se. Without such an account, it is not clear how
"scientificexplanations"succeed in explaining anything. That is, one cannot
begin to account for how science contributes to knowledge without an
account of what mathematical-scientific activity has to do with the reality
of which science contributes knowledge. This problem becomes partic-
ularly important in the context of philosophy of science and philosophy
of mathematics.
For nearly every field of study, there is a branch of philosophy, called
the "philosophy of" that field. Although present concern is limited to
science and mathematics, it will be useful to make a few brief remarks
concerning the philosophy of an arbitraryfield X.
Since the main purpose of a given field of study is to contribute to
knowledge, the philosophy of X is, at least in part, a branch of episte-
mology. Its purpose is to provide an account of the goals, methodology,
and subject matter of X. In particular, a philosopher of X tries to describe
the activity of X and show how that activity accomplishes its goals-that
is, how the activity of X contributes to knowledge of the subject matter
of X.
Thus conceived, the philosophy of X is not understood as a field iso-
lated from the practice of X. Rather, the philosophy of X is engaged in
by people who care about X in order to describe and account for its ac-
tivity, its successes and failures, and its importance. A practicer of X, or
an Xist, who adopts such a philosophy should gain something thereby.
The adopted philosophy should orient the Xist to his work by providing
a clear account of what he is trying to accomplish and how his practice
contributes to this.
In extremely broad terms, the goal of science is to understand various
aspects (e.g., physical, chemical, behavioral) of the world. Pythagoreans
aside, this world is prima facie non-mathematical. On the other hand, the
methodology of science includes considerable mathematics. An account
of how scientific methodology accomplishes its goals, then, should in-
volve an account of how mathematics can contribute to understandingthe
prima facie non-mathematical scientific world.
The problem can occur on several levels. It may begin when one won-
ders how it is possible for a particular mathematical fact to serve as an
explanation of a non-mathematical event. This happened, for example,

526 STEWART SHAPIRO
when my friend was puzzled that a physical event could be explained by

the fact that a certain function has a zero at a given value. In this case,
a reply might consist of a detailed description of the relevant scientific
theory which associates a certain class of functions with a class of phys-
ical phenomena.' A question on a different level could then be raised as
to what a class of mathematical objects (such as functions, infinite sets
of ordered pairs) can have to do with physical phenomena. Here, the
inquiry concerns the relevance of the given mathematical-scientific the-
ory. A possible reply to the second question would be to point out that
similar uses of mathematics have an important role in scientific meth-
odology. When this-the entire enterprise-is questioned, one might note
the vast success of the methodology in predictingand controllingthe world.
The last reply clearly explains why one might engage in mathematical/
scientific research and (up to standardHumean considerations) provides
the assurance that the methodology will continue to predict and control.
Such a statement, however, should not satisfy the epistemologist/philos-
opher of science, whose job is to account for scientific explanation and,
in particular, to show how and why the mathematical methodology suc-
ceeds in predicting and controlling the non-mathematical world.
In the philosophy of mathematics, the situation is not quite as straight-
forward. Since most (in fact, nearly all) mathematics is not directly aimed
at understandingthe non-mathematical world, it may be possible to ac-
count for the activity and goals of at least most mathematics without even
mentioning the possibility of a relationship with non-mathematicalreality.
This is the case, for example, with the traditional philosophies of math-
ematics discussed in the next section. If any such philosophy is consid-
ered complete, however, it must presuppose a sharp distinction, a dis-
tinction in kind, between pure mathematics and what is called "applied
mathematics"-mathematics aimed at the non-mathematical world. The
latter would constitute a separate subject.
There is an ambiguity in the phrase "applied mathematics". One sense
of the term refers to particularbranches of mathematics, such as the the-
ories of applied differential equations, numerical analysis, and mathe-
matical biology. Such branches often form important parts of scientific
theories. The other sense of "applied mathematics" concerns the informal
use of mathematical concepts and methods in either science, engineering,
or everyday life. This use involves such activities as calculating physical
constants and balancing checkbooks.
Within mathematics itself, the fact of the matter is that any distinction
between "pure" branches and "applied" branches is artificial. If one ex-
'It might be suggested to the questioner that if he wants to fully understand the reply,
he should take a few courses.

amines the practice, writings, and textbooks of both pure and applied
mathematics, one finds more similarities than differences concerning such
basic aspects as aims, techniques, logic, and even subject matter. More-
over, even though the day to day work of the pure mathematician is not
directly aimed at understanding non-mathematical reality, the branches
of so-called applied mathematics do not have a monopoly on relevance
to the scientific world. As Nicolas Goodman puts it:
. . .most branches of mathematics cast light fairly directly on some

part of nature. Geometry concerns space. Probability theory teaches
us about randomprocesses. Group theory illuminates symmetry. Logic
describes rational inference. Many parts of analysis were created to
study particularphysical processes and are still indispensable for the
study of those processes. . . . It is a practical reality that our best
theorems give information about the concrete world (1979, p. 550).
It might be added that many branches of applied mathematics, such as

the theory of applied differential equations, are straightforward sub-
branches of areas in pure mathematics and, occasionally, applications are
discovered in areas of pure mathematics, such as algebra or complex anal-
ysis. I conclude that there is no significant or philosophically illuminating
distinctionto be made between branchesof pure mathematicsand branches
of applied mathematics.
Of course, the informal use of mathematical concepts and methods does
not always bear such a close resemblance to the methodology of pure
mathematics. The scientist, for example, often relies on estimation pro-
cedures, such as so-called "fudge factors" and informal curve fitting.
Moreover, much of the reasoning is more or less inductive. Even this
mathematics, however, is rooted in pure mathematics, usually arithmetic
or real analysis. The mathematical concepts involved include natural
numbers or real numbers, number-theoretic functions, and related nu-
merical constructions. The estimation procedures are thought of as ex-
pedients which should ultimately be dispensable (at least in principle). I
conclude that it would be a mistake to adopt a philosophy which divorces
even this use of mathematics from the work of the pure mathematician.
Finally, a philosophy of mathematics should not only provide an in-
ternally coherent account of mathematical goals and practice, it should
also account for the importance of mathematics. That is, a philosophy of
mathematics should explain not only what the mathematician does, but
also why he does it, and, in particular, how mathematical knowledge is
related to general knowledge. This can be accomplished only by an ac-
count of mathematics which does not obscure or deny the connections
between mathematics and science.

528 STEWART SHAPIRO
II
In this section, attention is turned to some of the major, traditional
philosophies of mathematics: formalism, logicism, platonism, and intu-
itionism. For reasons of brevity, only the standard articulations are con-
sidered. It is shown that none provides even the direction of an account
of the relationship between mathematics and scientific reality, and, more-
over, some even seem to imply that there is no such relationship. I ignore
the various strengths and weaknesses of the views that do not relate to
our present concern. My discussion of Hartry Field's Science without
numbers is postponed until the addendum to the final section.
1. For present purposes, formalism is easiest to deal with, at least in
its extreme versions. This philosophy holds that mathematics consists of
no more than the manipulation of characters according to rules. That is,
the formulas of mathematics are considered to be strings of meaningless
characters, not genuine symbols (which symbolize something). A theo-
rem of, say, arithmetic, does not represent a fact about the natural num-
bers (or anything else for that matter). Rather, such a theorem is the result
of a series of manipulations according to the rules of arithmetic. More-
over, the rules of the various branches of mathematics are often consid-
ered to be arbitrarilychosen. From this point of view, a branch of math-
ematics might be thoughtof as a game played with typographicalcharacters,
a proof of a theorem in a branch being a correct play of the game.2
If this account of mathematics is correct, then any systematic corre-
spondence or relationship between the theorems of a branch of mathe-
matics and facts about the world must be a fortuitous coincidence. A
mathematician(or scientist) who studies and proves theorems in, say, real
analysis, cannot have a reasonable hope that such theorems will shed light
on any reality beyond the rules of real analysis, just as a chess player
who studies and proves theorems about endgames cannot expect his re-
sults to shed light on any reality beyond the rules of chess. It seems as
though an extreme, traditional formalist must actually deny a systematic
and regular connection between mathematics and non-mathematical (that
is, non-formal) reality.3
2Differentformalistic views are advocated in Cohen (1971), Curry (1958), and Robinson
(1965). See Resnik (1980, Chapter 2) for a discussion of the historical development of
formalism.
3A similar objection was raised by Frege:
An arithmetic with no thought as its content will also be without possibility of ap-
plication. Why can no application be made of a configuration of chess pieces? Ob-
viously because it expresses no thought. . . . Why can arithmetic equations be ap-
plied? Only because they express thoughts. How could we possibly apply an equation
which expresssed nothing and was nothing more than a group of figures, to be trans-

It is possible, of course, for someone to simply reject the claim that

the choice of rules for a branch of mathematics is always arbitrarywithout
rejecting the main tenets of formalism. One may hold, for example, that
in some cases, at least, the rules for charactermanipulation are explicitly
or implicitly chosen to be in accord with a scientific theory. Such a for-
malist might claim that a branch of mathematics, such as real analysis,
is applied to, say, physics, through a mapping of the language of real
analysis into the language of physics. The mapping, of course, should
assign theorems and correct inferences of physics to axioms and correct
inferences of real analysis.4
This formulation of formalism has the virtue of not claiming that sys-
tematic connections between mathematics and reality are impossible, but
it does not begin to explain the connection. Moreover, it presupposes a
sharpdistinctionbetween pure mathematicsand applied mathematics:Pure
mathematics would consist of those branches whose rules are indeed ar-
bitrary; applied mathematics of those whose languages are mapped to a
language of science or, in other words, those whose rules correspond with
some of the rules of a science language. As concluded in the previous
section, however, such a distinction is artificial. It ignores the intimate
connections and similaritiesbetween branches of mathematics. If the rules
of the branches of applied mathematics are intimately connected with the
languages of science, then so are the rules of pure mathematics. It is the
job of philosophy to explain these connections, not merely to note that
they can or must exist. Formalism per se does not and perhaps cannot
do this.
2. The slogan of logicism is "mathematicsis logic". There are, in fact,
two views that go by this name, one of which we call "translation lo-
gicism" and the other "postulate logicism". Although the two views are
somewhat related, the postulate version is more compatible with connec-
tions to science.
(i) Translation logicism is the view that when mathematical terms are
given proper definitions, mathematical statements are themselves true or
false solely in virtue of their forms. In other words, mathematical truths,
properly understood, are logical truths.5
formed into another group of figures in accordance with certain rules? (1903, Section
91)
See Resnik (1980, pp. 62-63) for a discussion of this passage.
4This account of mathematics and science is perhaps compatible with certain philoso-
phies of language that analyze meaning, at least in part, in terms of possible inference.
See, for example, Harman (1975) and Dummett (1973).
'Related views were held by Ayer (1946) and Hempel (1945) who believed that math-
ematical truths, properly understood, are analytic. Slight modifications of the following
remarks apply to these views.

530 STEWART SHAPIRO
Like the extreme formalism, translation logicism directly denies the

possibility of a systematic relationship between the theorems of a branch
of mathematics and areas of scientific reality. The view holds that each
mathematical theorem is, in fact, a logical tiuth. The descriptions of in-
teresting physical phenomena, however, are never true in virtue of their
forms and, thus, cannot be substantially explained by logical truths. For
example, if the mathematical fact that a particular function has a zero at
a value is indeed a logical truth, then it cannot represent, account for, or
substantially explain the physical phenomenon that puzzled my friend.6
I conclude that translationlogicism, as a philosophy of mathematics, must
be rejected.
(ii) Postulate logicism is the view that mathematicsconsists of no more
than a study of the logical consequences of uninterpretedsets of axioms.
Thus conceived, a branch of mathematics does not have a pre-determined
subject matter, like the natural numbers, Euclidian space, or the set-the-
oretic universe. On the contrary, all primitive (non-logical) terms of
mathematicsare taken to be without interpretationand, thus, without fixed
meaning.
According to this version of logicism, then, a branch of mathematics
can be "about" any interpretationof its primitive terms that satisfies all
of its axioms. It follows that a systematic correspondence between the
theorems of a branch of mathematics and facts of reality can be estab-
lished through an interpretationof the language in non-mathematical real-
ity. Indeed, it is clear that some branches of mathematics, such as Eu-
clidian geometry, were developed with an interpretationof the primitive
terms in mind. In those branches at least, postulate logicism does succeed
in explaining the connections between mathematics and reality. If the
axioms as interpretedare true, then the theorems are true under the same
interpretation.On this view, then, the role of mathematics in science is
to uncover logical connections between certain scientific statements.
Like formalism, this version of logicism presupposes a sharp distinc-
tion between applied mathematics and pure mathematics with applied
mathematics consisting of those branches whose primitive terms are in-
terpreted,pure mathematics of those whose primitive terms are not. Once
6Glymour(1980) suggests that some physical phenomena are properly described by log-
ical truths once certain theoretical identifications are made. The example he provides is
rather simple: The explanation that the temperature of a gas and its mean kinetic energy
have concomitant values is that temperature is mean kinetic energy and, thus, the phe-
nomenon is an instance of the law of identity. Of course, even if Glymour's remark is
correct, the explanation consists of the theoretical identifications, not the law of identity.
An attempt to rescue translation logicism along these lines could consist of a claim that
whenever a mathematical fact is used to explain a physical phenomenon, a series of the-
oretical identifications actually reduces the phenomenon to a logical truth. This attempted
rescue is not pursued further here.

again, such a distinction artificially ignores the importantconnections be-

tween branches of mathematics.
Second, the present view only accounts for applications involving
branches of mathematics that have straightforwardinterpretationsof their
languages. The application of mathematics, however, is more pervasive
than this. Some fields, such as non-Euclidian geometry and abstract al-
gebra, do not have immediate, straightforwardinterpretations, but, as is
well-known, do shed light on many parts of reality (concerning group
theory, see Budden 1972). In fact, both of these fields were well-devel-
oped branches of pure mathematics before the applications were discov-
ered. To a postulate logicist, the primitive terms of these branches could
only be considered arbitrary, at least at first. The variety and depth of
the applications must be a great coincidence. That is, if this view of math-
ematics is correct, then the founders of branches of mathematics seem to
have an uncanny ability to choose axiom systems which indirectly shed
light on parts of reality.
Finally, the present view accounts for an application of a branch of
mathematicsonly if all of its primitive terms are interpretedin a way that
satisfies all of its axioms. There are some branches, however, that have
found applications even though some of the primitive terms are not in-
terpretedin non-mathematical reality. A prime example is complex anal-
ysis. The use of the term "imaginary number" itself indicates that no
straightforwardinterpretationof these mathematical objects is forthcom-
ing, yet the theory is useful in physics and engineering. A related phe-
nomenon is the occasional use of an "uninterpreted"theory, such as set
theory, to solve problems undecidable in a weaker theory, such as arith-
metic. Some of these results have applications in recursive function the-
ory and thus in computability. One would be hard put to find an inter-
pretation of the primitive terms of set theory in the non-mathematical
reality that relates to computability. At any rate, postulate logicism does
not account for these applications.
3. Platonism is the view that the subject matter of mathematics is an
immaterialand non-mental realm or universe that exists independently of
the physical world. According to this view, mathematical assertions are
taken literally as statements about this mathematical universe. The as-
sertion that there are infinitely many prime numbers, for example, is
understood as expressing a fact about the independent domain of natural
numbers, that among these numbers, infinitely many are prime. One of
the clearest and most influential statements of mathematical platonism is
found in Godel (1964).
A relationship between mathematics thus conceived and physical real-
ity amounts to a relationship between the mathematical universe and the
physical universe. The platonist, however, seems to believe that the math-

532 STEWART SHAPIRO
ematical world has an existence that is separate from and independent of

the physical world. Of course, the independentexistence of the two worlds
does not entail that they are unrelated. Thus, platonism does not, of itself,
imply that mathematical knowledge is irrelevant to knowledge about the
physical world. However, as it stands, platonism does not explain this
relationship and, in fact, leaves it mysterious. That is, to hold that math-
ematics is about a non-physical universe, to emphasize the independence
of this universe from the material world, and to leave it at that is to ignore
and even to obscure one of the most important aspects of mathematics,
its importancein scientifically understandingthe non-mathematicalworld.
Godel himself seems to recognize the existence and importance of the
connections with physical reality. As a platonist, he holds that a propo-
sition independent of the axioms of a branch of mathematics may never-
theless have a truth value determined by the mathematical reality under
study. Godel claims that in such a case a probabilistic "criterionof truth"
of a proposition consists of its "fruitfulness in mathematics and, one may
add, possibly also in physics" (1964, p. 272, emphasis mine). Also, some
well-known authors, such as Putnam (1971) and Quine, go so far as to
take the usefulness of a branch of mathematics in science as evidence that
the theorems of that branch are true of the independent domain of math-
ematical objects. It is clear, however, that fruitfulness in science could
not even be a probabilistic criterion for mathematical truth if the math-
ematical universe were unrelated to the reality studied by physics.
This problem for mathematical platonism may be construed as a mod-
em version of the Platonic problem of participation. In the present con-
text, there are actually two problems. The first is to articulate the rela-
tionship between the mathematical universe and the physical universe; the
second is to show how knowledge concerning the former contributes to
knowledge concerning the latter.7My conclusion is that mathematical pla-
7Theepistemological problem concerning platonism is of special concern here. Generally

speaking, the problem is to show what access the mathematician has to information about
the mathematical universe and, thus, to show how mathematical knowledge is possible.
A common response, suggested by Godel (1964), is to postulate a faculty of perception
or "intuition" of the mathematical world. It is claimed that mathematical intuition vis-a-
vis the mathematical world is analogous to sense perception vis-a-vis the physical world-
each faculty leads to the formulation of theories about an objective world. The main datum
of the present paper is that most of our theories about the physical world are formulated
in mathematical terms. Thus, the onus on a Godelian platonist is not only to account for
the ontological relationship between the mathematical world and the physical world, but
also to account for the epistemic relationship between mathematical intuition and sense
perception and, thus, to show how theories built on the former contribute to theories built
on the latter.
Godel (1964) seems at least partially aware of the problem. In a much quoted passage,
he suggests that "the 'given' underlying mathematics is closely related to the abstract ele-
ments contained in our empirical ideas" (p. 272). This indicates that the present gap in
platonism might be filled through a phenomonological study of mathematical experience,

tonism stands in serious need of further articulation in both areas.

4. Mathematical intuitionism is a philosophy that accepts a faculty of
mathematical intuition, but rejects the platonist claim that this intuition
is a "perception"of an objective realm. In particular, the view holds that
mathematical objects, such as numbers, are constructed by the mathe-
matician and, as far as mathematics is concerned, have no existence apart
from his mind. For an intuitionist, mathematical assertions report mental
constructions performed on mathematical objects, objects which are
themselves viewed as mental constructs.
One straightforwardway to account for the applications of mathematics
thus conceived to science would be to postulate a relationship between
materialreality and the portion of the human mind that does mathematics.
In particular,one may attempt to articulate and defend the possibility that
the structureof a mind performing constructions on its own mathematical
constructs is similar, analogous, or related to the structure of physical
operations on material reality.
The intuitionists themselves, however, do not suggest such an account.
In fact, by the assertions that mathematics is subjective and that philos-
ophy of mathematics should not require metaphysical speculation, they
seem to eschew the possibility of a relationship to material reality. Thus,
they would probably hold the success of this program to be unlikely at
best. Moreover, if the program involves ontological assertions about
mathematical objects, it will be in outright conflict with the basic moti-
vation behind intuitionism.
A second possibility, perhaps, would be for an intuitionist to combine
his views on mathematics with a Kantian philosophy of science. Suppose
one holds, following Kant, that the human mind plays an active role in
perceiving and understanding the world, or, in other words, that what
one perceives and understands is as much a function of the human mind
itself as it is a function of "inputs" from external reality. A person who
holds such a view would hold that a scientific theory is not merely a
theory of an objective world as such, but rather a theory about a world
as perceived and understood by the human mind.
This conception of science seems to be at least in the same spirit as
the intuitionistic claim that mathematics is a study of certain features of
the human mind (see Heyting 1956, p. 10). An intuitionist who accepts
this view of science might attempt to account for the application of math-
ematics by discovering a relationship between the mental activity of math-
ematical construction and the mental activity of perception and subjective
ordinary perception, and science. Some preliminary and speculative suggestions for this
(based on some physiological data and work from the Piaget school) are found in Maddy
(1980).

534 STEWART SHAPIRO
understanding.8That is, an intuitionist might claim that either mathe-

matical constructions themselves, or closely related mental constructions,
are involved, not necessarily in the objective world itself, but in the hu-
man activity of understanding the world. I will not pursue the matter
furtherhere.
III
The time has come to abandon the systematic criticism series and to
make some positive suggestions. I propose here a structuralistic philos-
ophy of mathematics which, I believe, provides a more promising account
of the relationship between mathematics and scientific reality.
1. Mathematical structuralism agrees with platonism and intuitionism
that there is a subject matter of mathematics, but it holds that this subject
matter consists of patterns or structures and not collections of mathe-
matical objects. Arithmetic, for example, is not understood as the study
of a particular set consisting of the natural numbers, but rather as the
study of the natural-number-structure,the structureor pattern of any sys-
tem that has an infinite sequence of objects with an initial object and a
successor relation (or operation). The natural number structure is exem-
plified, for example, in the arabic numerals, the expressions of the Ro-
man alphabet in lexical order, an infinite sequence of strokes II I. I
and an infinite sequence of distinct moments in time. More precisely, the
natural number structure is the structure common to these systems. For
a second example, group theory is the study of the structures9common
to all systems of objects that have an associative binary operation, an
identity object, and inverses for each object. Common group structures
are exemplified in various systems of symmetries and systems of per-
mutations. Similarly, Euclidian geometry is the study of Euclidian space
structure, topology the study of topological structures, etc.
One lesson we've all learned from Plato is that to give a list of a certain
kind of thing is not to say what the kind is. I thus do not pretend to have
defined what a structure is by invoking the word "pattern"and giving a
8As indicated in the previous note, Gbdel hinted at such a relationship in the context of
platonism and mathematical perception.
9There is an ambiguity in the word "structure". The various senses depend on whether
or not a (single) structureis taken to be unique up to isomorphism. Here, I prefer the sense
in which "A and B have the same structure"entails "A and B are isomorphic". This sense
of the term simplifies the discussion of the ontological issues and the application of math-
ematics to science. There is, however, a sense in which one says, for example, that there
is a structure common to all groups (perhaps called "the group structure") even though,
of course, not all groups are isomorphic. The phrases "partial structure" and "incomplete
structure"have been used for the latter sense (see Barbut 1970).

few examples, but perhaps, by ostention, I've pointed in a familiar di-

rection. A few heuristic comments are in order.10
One way to grasp a particular pattern is to consider a collection (or
collections) of objects arranged in a certain way and to ignore those fea-
tures of the objects that do not relate to the arrangement. That is, one
begins with a system of objects and focuses on the arrangementitself. A
second, more abstract, way to grasp a patterninvolves a direct description
of the relations among the positions or places of the pattern (and thus
avoids reference to particularobjects and systems of objects). To take an
example, one can grasp the pattern of a baseball defense either by going
to a game (or several games) and focusing on the arrangement of the
players (ignoring such features as the height, weight, and hair color of
the various players) or by understandinga direct description of the pattern
("there are four infielders arranged in such and such a way, three out-
fielders, etc. ").1
A mathematical structure can, perhaps, be similarly construed as the
form of a possible system of related objects, ignoring the features of the
objects that are not relevant to the interrelations. The structure is thus
completely described in terms of the interrelations.
A typical beginning of a mathematics text consists of the announcement
that certain mathematical objects, such as real numbers, are to be studied.
In some cases at least, the only thing we are told about-these objects is
that there are certain relations among them and/or operations on them.
We may be informed, for example, that they have a dense linear order,
or that there is an associative and commutative operation of addition on
them. One easily gets the impression that the objects themselves don't
matter;the relations and operations are what we study. The idea that the
subject matter of a given branch of mathematics is a certain structure (or
class of structures)is a straightforwardinterpretationof this observation. 12
The writers who have articulated and defended structuralist philoso-
'?See Resnik (1981) and (1982) for a more complete account of structures and struc-
turalism.
"A third (perhaps intermediate) way to grasp a pattern is through a description of it as
subpatternor other variant of another. For example, an infield is a subpattern of a defense
and a "left-handed shift" is a variant of an infield in which the shortstop plays to the right
of second base.
"20f course, many, indeed most, mathematics texts have less abstract beginnings. A
book may begin, for example, by defining a particularclass of mathematical objects, such
as the class of analytic, complex valued functions or the class of partial recursive functions.
Alternately, it may be announced that the book concerns objects, such as sequences of
real valued complex functionals, which are "constructed" from (familiar) mathematical
objects. I suggest that these works can also be given structuralist interpretations. If one
understands the base theory (that is, arithmetic, real analysis, etc.) as the theory of a
structure, then the given text is seen as the theory of either a substructure of the base
theory structure or a second, perhaps richer, mathematical structure determined from that
of the base theory. For reasons of brevity, this account is not elaborated further.

536 STEWART SHAPIRO
phies of mathematics seem to have widely varied concerns and interests.

Bourbaki (1950) and Barbut (1970) propose that the study of structure is
what unifies the various branches of mathematics. Mark Wilson (1981)
suggests that structuralism can help solve common puzzles concerning
the ontological presuppositions of equivalent mathematical theories. De-
dekind (1888) and Benacerraf (1965) propose accounts of arithmetic sim-
ilar to the above in order to define the natural numbers. Finally, Steiner
(1975) and Resnik (1975), (1981), and (1982) elaborate structuralistphi-
losophies of mathematics while attempting to account for mathematical
knowledge in the context of the problems with platonism. See Nutter
(1980) for a comprehensive analysis of mathematical structuralism and
its defenses.
2. Before proceeding to the applications of mathematics, a few brief
remarks on the ontological presuppositions of structuralism may be in
order. Structuralismholds that mathematical "objects", such as numbers,
sets, and points are actually places within structures. There are thus two
ontological issues, the status of the structures themselves and the status
of the so-called "mathematical objects" or places within each structure.
Without getting clear on the meaning of "existence", there is little need
to make a bold statement concerning structures and thus to offend some
ontological sensibilities. 3 It should be pointed out, however, that a given
structure is abstract in the sense that it can have more than one exem-
plification. Thus, we seem to have a special case of the problem con-
cerning the status of universals: A structure is a universal and a system
of objects exemplifying it is an instance. Second, the properties of a given
structureare independent of the mathematician. From the present point
of view, for example, Godel's incompleteness theorem shows that the
naturalnumber structurehas properties which are not consequences of its
(first-order) description. Third, the common criterion for commitment,
"3Ido not deny (nor do I apologize for) the fact that my own discourse, the language
of this paper, takes the word "structure" as a common noun and (informally) quantifies
over structures. Thus, for example, I speak freely of "the natural number structure", "the
set-theoretic structure", etc., and I entertain statements like "there is a structure satisfying
the axioms of arithmetic" and "all structures can be embedded in the set-theoretic hier-
archy". Of course, there are those who conclude from this alone that I am ontologically
committed to structures. There is perhaps a sense of "commitment" in which this is correct,
but accordingly, anyone who uses a universal term as a noun or (informally) quantifies
over a universal term is committed to the respective universal. Thus, for example, state-
ments like "there are five distinct colors on this vase" and "there is a quality of this sofa
that makes it unattractive" commit the speaker to colors and aesthetic qualities. Similarly,
one might hold that any mathematician who uses standard discourse (that is, such state-
ments as "every prime number greater than two is odd") is committed to the existence of
mathematical objects. In other words, in the sense of commitment under consideration, all
classical mathematicians are platonists. Resnik (1980, Chapter 5) uses the term "meth-
odological platonist" to refer to such mathematicians. To extend the phrase, I admit to
being a methodological platonist viv-a-vis structures.

due to Quine, is to be within the range of a variable (in an accepted

theory). Mathematical logic and, in particular, model theory can perhaps
be construed as a theory that quantifies over structures. To say that a
class of sentences is satisfiable, for example, may be to say that there is
a structurethat satisfies it. It is more common, however, to construe the
variables of model theory as ranging over sets. Sets are currently taken
to be places within the set-theoretic-hierarchy-structure.Thus, like any
other branch of mathematics, model theory is seen as the study of a par-
ticular structure. To complicate matters a bit, however, I suggest that the
thesis that mathematics can be reduced to set theory amounts to a claim
that any given mathematical structure(except that of set theory itself) can
be modeled in the set-theoretic structure and, moreover, that the latter
captures the relevant relationships between structures.14
There seems to be some disagreement among structuralistsconcerning
the status of mathematical "objects"-places within a given structure. As
is well-known, Benacerraf holds that "three", for example, cannot be
considered an object because there seems to be no way to determine the
truth values of such identity statements as "3 = {{{f}}}". Steiner, on the
other hand, remarks that:
we accept mathematical objects, contra Benacerraf, but we agree that
the only things to know about them of any value are their relation-
ships with other things. (This is the mark of abstract objects.) (1975,
p. 134)
One gets the impression that the disagreement here is over the meaning
of the word "object". In his early (1975), Resnik states that mathematical
entities, such as natural numbers, cannot be "identified absolutely, but
only relative to other 'objects' in the structure in which they occur" (p.
35). In a footnote, he remarks that the word "object" is placed in quotes
because
it is no longer clear whether numbers, sets, etc. should count as ob-

jects. I think that we may need to move towards a notion of an object
of a theory (1975, p. 35).
This remark suggests that the very notion of "object" is to be made rel-
ative to a theory.'" Numbers are objects relative to arithmetic, but they
are not objects relative to set theory. This conclusion is amplified in Res-
nik's later (1981):
14The ontological relevance of the reduction of mathematical structures to places within

the set-theoretic hierarchy is discussed in some detail in Resnik (1981).
"5Arelated idea is found in Kraut (1980), which deals with similar ontological/linguistic
issues in a broader context.

538 STEWART SHAPIRO
Natural number theory . . . deals with a certain pattern; it has the

means to raise and answer questions concerning the identity of var-
ious numbers, but it cannot evenformulate the question as to whether
the numberone is r{44].
On the otherhand, within numbertheory,
identity is absolute.... [A mathematical] theory speaks only of ele-
ments of a certain structure and has no means to identify or distin-
guish these from elements of another structure (1981, p. 537).
3. My account of the relationship between mathematics and science
begins with the suggestion that the contents of the non-mathematical uni-
verse exhibit underlying mathematical structuresin their interrelations and
interactions. For example, it might be claimed that a mathematical struc-
ture similar to the inverse-square variation of real numbers is exemplified
in the mutual attraction of physical objects. In general, physical laws
expressed in mathematical terms can be construed as proposals that a
certainmathematically defined structureis exemplified in a particulararea
of physical reality.
On this account, the problem of the relationship between mathematics
and reality is a special case of the problem of the instantiation of uni-
versals. Mathematics is to reality as universal is to instantiated particular.
As above, the "universal" here refers to a pattern or structure;the "par-
ticular" refers not to an individual object, but to a system of related ob-
jects. More specifically, then, mathematics is to reality as pattern is to
patterned.16
Chomsky seems to hold a structuralist view in that he believes that
human linguistic performance exemplifies mathematical structures of
transformationgrammar, structures innate to the mind. Piaget (1968) de-
scribes his conception of psychological development in terms compatible
with this account of mathematical application. Shapiro (1981) proposes
an account of computability (and Church's thesis) in structuralistterms.
For similar descriptions and accounts of other sciences, see the anthology
edited by Michael Lane (1970).
A first attempt to clarify the present thesis would be to claim that sci-
ence proceeds by discovering exemplifications of mathematical structures
among observable physical objects. In fact, this accounts for at least some
16Resnik (1982) suggests that a structuralist understanding of mathematics provides the

direction of a solution to the epistemological problem with platonism (see note 7 above).
It is proposed that, in some cases at least, learning a mathematical structure amounts to
pattern recognition. A sketch of the various levels of abstraction involved and a careful
discussion of ontological commitment is provided.
Of course, there are many philosophical problems connected with patterns, pattern ex-
emplification, and patternrecognition. Neither Resnik's work nor mine is designed to solve
all of these problems. At worst, our work can be seen as reducing certain problems con-
cerning mathematics to problems concerning patterns.

applications of mathematics. I'll treat one example in some detail.

Crowell and Fox (1963) wrote a mathematics book about knots. In the
Introduction,they discuss the problem of using mathematicsto study these
physical objects:
Definition of a Knot: Almost everyone is familiar with the simplest
of the common knots, e.g., the overhand knot . . . and the figure-
eight knot.. . . A little experimenting with a piece of rope will con-
vince anyone that these two knots are different: one cannot be trans-
formed into the other without . . . "tying" or "untying". Neverthe-
less, failure to change the figure-eight into the overhand by hours of
patient twisting is no proof that it can't be done. The problem that
we shall consider is the problem of showing mathematically that these
two knots . . . are distinct from one another.
Mathematics never proves anything about anything except mathe-
matics, and a piece of rope is a physical object and not a mathe-
matical one. So before worrying about proofs, we must have a math-
ematical definition of what a knot is. . . . This problem . . . arises
whenever one applies mathematics to a physical situation. The def-
initions should define mathematical objects that approximate the
physical objects under consideration as closeiy as possible (1963, p.
3).
When these authors suggest that some mathematical objects can resem-
ble or "approximate"physical objects such as pieces of rope, they, of
course, do not mean that mathematical objects can have properties like
solidity, flexibility, and extension. Rather, they mean that the possible
relationships and interconnections of pieces of rope formed into knots can
be described or "modeled" in the relationships of a mathematical struc-
ture, a topological space in particular. I suggest that it is a straightforward
articulationof this view that rope formed into knots exemplifies the math-
ematical structuredescribed in the book. The point is to prove theorems
about knots. If Crowell and Fox as interpreted here are correct in their
assumptions, then it would follow that particular mathematical theorems
represent facts about knots. In particular, a mathematical theorem that
one particulartopological configuration cannot be transformedinto a sec-
ond by specified mathematical operations corresponds to a fact that one
knot cannot be transformed into another without "tying" or "untying".
It is easily seen that this first account of mathematical application is
much too simple. The claim that a structure is exemplified in physical
reality amounts to a claim that there are entities of some sort that answer
to the places within the structure. Many of the structuresstudied by math-
ematicians and used in science have an infinite number of places and,
thus, there are not enough observable physical objects to exemplify them.

540 STEWART SHAPIRO
The application of an infinite structure amounts to the postulation of in-

finitely many theoretical entities. For example, the claim that actual space
exemplifies the structureof Euclidian geometry involves as assertion that
there is a continuum of space points. The status of such entities need not
concern us here.'7 I only assert that space points are theoretical entities
called for by the use of Euclidian geometry in physical theory.
According to this view, then, science does proceed by discovering
mathematicalstructuresexemplified in material reality, but the discovery
is often indirect and involves the postulation of theoretical entities. The
situation might best be described as scientific theories incorporating
mathematical structures.'8Another example follows.
Turing's (1936) argument for a version of Church's thesis seems to
presuppose a view of this sort insofar as he considers the computational
process itself subject to mathematical study. In the argument, he makes
some rather straightforwardclaims that the possible instructions and ma-
terials of a human following an algorithm have properties exemplified by
certain mathematical structures. Let us look in detail at one small part of
his argument, his "proof" that a language used in computation must have
a finite alphabet. 9 It is assumed at the outset that each symbol can be
understood (or "defined") as a set of points in a unit square of standard
metric space (thus presupposing that there is an upper bound on the size
of a single symbol). It is further assumed that the set of points corre-
sponding to each symbol is measurable. Turing then defines a rather nat-
ural metric among the set of "symbols" thus construed. He points out
that with the induced topology, the "symbols" themselves form a con-
ditionally compact space. It then follows, mathematically, that "if we
were to allow an infinity of symbols, then there would be symbols dif-
fering to an arbitrarilysmall extent". This, together with the further as-
sumptionthat there is some limit to the human ability to distinguish among
tokens, entails that an alphabet of computation must be finite.
A structuralistinterpretationof this argumentis that Turing first claimed,
noticed, or postulated that the system of relevant relationships between
7HartryField (1980, pp. 31-33) argues that such entities as space-time points should
be construed as physical, not abstract entities because, among other reasons, contingent
properties of individual points are essential parts of causal explanations of physical phe-
nomena. On the other hand, some of the reviews of Field's work take issue with this by
arguing that space-time points are abstract. See, for example, the reviews by M. Resnik
in Nous and D. Malament in the Journal of Philosophy.
8Igal Kvart (private communication) has suggested that this account of the application
of mathematics can be reconciled with the traditional platonist view concerning a universe
of mathematical objects. Instead of speaking of mathematical structures underlying phys-
ical reality, a platonist can speak of isomorphisms between systems of mathematical objects
and systems of physical objects.
9Although the mathematics involved in this example is particularly clear and straight-
forward, it is curious that Turing felt that this needed to be proved.

possible tokens of symbols exemplifies the structure of measurable sets

in standard metric space. He went on to define a new metric on these
sets of points and proved mathematically that with this metric, the pos-
sible "symbols" exemplify one of the conditionally compact space struc-
tures. The final conclusion follows from a well-known theorem about
such structures, together with the assumption on human perceptual abil-
ities. I suggest that this interpretationis straightforward.
Consider the following objection to the present view: According to
structuralism,naturalnumbers are places within the naturalnumber struc-
ture. This structure is infinite. This seems to imply that whenever one
uses numbers or arithmetic in everyday life, such as counting tomatoes
or balancing a checkbook, one is presupposing an infinite structure. It
seems absurd to claim that when a child learns to count his toes, he is
actually applying an infinite structure to reality.
In reply, it should be noted that for virtually all applications of arith-
metic, one does not require the entire natural number structure, but only
various finite structureswhich can be seen as "partof" the naturalnumber
structure.For example, when a child learns to count to twenty and, thus,
learns to count small collections, he is learning a certain finite pattern,
the structure of a twenty-element-sequence. This is sufficient for many
of his purposes. He may even learn to enrich this structure with addition
and multiplication (as partial functions). Later, he will learn that this
structureis (or can be seen as) part of a larger one with, say, ninety-nine
places. Eventually, he learns that this latter structure is part of an even
larger one, a structure with infinitely many places. At this point, he is
beginning to grasp the natural number structure. In sum, a complete the-
ory of counting does indeed involve an infinite structure, but any given
instance of counting, or any given use of arithmetic, can be explained
without the complete theory.20 A finite "counting" structure is suffi-
cient.2'
4. From the present perspective, the main advantage of structuralism
is that it provides a more holistic view of mathematics and science, and
thus accounts for the rich interplay between the fields. It is suggested that
much of the theoretical work in science consists of "incorporating"well-
developed mathematical structures and, simultaneously, postulating the-
oretical entities. There is also traffic in the other direction. As pointed
20Noticethat the (full) naturalnumber structureis exemplified by the interrelationsamong

the finite counting structures. That is, the natural number structure can be construed as
the structure of other structures.
21It might be noted that according to Turnbull (1978), Plato himself may have held a
ratherextreme structuralism. Turnbull suggests that Platonic Forms are properly interpreted
as "principles of structure", with the Form of X being the mathematical structure of X.
This work relies on an interpretation of the later dialogues, including the geometric de-
scription of physical reality in the Timaeus.

542 STEWART SHAPIRO
out above, many branches of analysis were developed to study physical

processes. The present interpretationof this phenomenon is that the in-
dicated structures were discovered through experimental and theoretical
work in physics: The scientists provided mathematicaldescriptionsof (what
were taken to be) the structures underlying an area of physical reality.
Once defined, the structures can be studied by physicists and mathema-
ticians alike as any mathematical structures are studied.22
A question can arise at this point concerning the status of the adjective
"4mathematical" in the phrase "mathematicalstructure". One may wonder
what (if anything) distinguishes mathematical structures from other kinds
of structures. My view is that, extensionally speaking, there is no dif-
ference, or, at any rate, no philosophically illuminating difference. Vir-
tually any structure can be a mathematical structure if mathematicians
(qua mathematicians) study it (qua structure).
The difference lies more in the way that structures are presented and
studied. Mathematical structures are described abstractly-independent
of what the structuresmay be structuresof-and studied deductively. For
example, one way to present a structure mathematically is through ax-
ioms: One directly describes the relationships within the structure by
quantifying over its positions. A second way is to describe it as a well-
defined extension or substructureof another (purely described) structure.
In either case, the mathematical description does not involve anything
beyond structuresthemselves.
Along similar lines, Resnik (1982) makes a distinction between the
"pure theory" of a structure and an "applied version" of it:
Take the case of linguistics. Let us imagine that by using the ab-
straction process . . . a grammarian arrives at a complex structure
which he calls English. Now suppose that it later turns out that the
English corpus fails in significant ways to instantiate this pattern, and
that many of the claims which our linguist made concerning his struc-
ture will be falsified. Derisively, linguists name the structure Ten-
glish. Nonetheless, much of our linguist's knowledge about Tenglish
qua pattern stands; for he has managed to describe some pattern and
to discover some of its properties. Similarly, . . . we know much
about Euclidian space, despite its failure to be instantiatedphysically.
A pure theory of a pattern is a deductively developed or developable
theory which is based upon axioms which purportto characterize the
pattern in question. Its assertions do not extend to claims concerning
whether, where or how the pattern is instantiated, and they are true
of that pattern regardless of its applicability. . . An applied theory
22See Polya (1954) and (1977) for a wealth of examples of the rich interplay between
mathematics and physics.

of a pattern [involves] claims stating how the pattern is instantiated

. . . (1982, p. 101).
With this terminology, mathematical structures are those that are pre-
sented and studied throughpure theories. Mathematicsper se is the "pure"
study of patterns.
To return to the above examples, Crowell and Fox's work is aimed at
understandingknots, Turing's work at understandinghuman computation
ability. Both of these subject matters are prima facie non-mathematical.
Thus it would seem that both works belong to science, not mathematics.
Yet the works are classified as mathematics,taughtin mathematicscourses,
etc. A straightforwardreason for the classification is that all but a few
pages of each work are devoted to the study and development of the
respective structures as such. The connections with reality, such as the
claim that tokens exemplify the structure of measurable sets of points,
are taken to be obvious-there is little or no empirical confirmation. In
short, for the most part, knot theory and computability are pure theories
of particularstructures.The relationship to reality may be what originated
the inquiry into these structures, but this relationship is of no concern in
the study itself.
On my view, the determination of what counts as mathematics and
what counts as science is, at least in part, a pseudo-problem. One expects
border-line cases.
5. A second, related advantage of structuralismis that it accounts for
the interconnections among the various branches of mathematics. Con-
sider, for example, the occasional use of real analysis to prove theorems
about naturalnumbers. This is possible because the naturalnumber struc-
ture is "contained" in the real number structure as a part. Thus it is no
accident that a study of the real number structure can shed light on the
naturalnumberstructure.The resultingnumber-theoretictheorems are often
not consequences of standard (first-order) arithmetic. Similarly, the fact
that the natural number structure and the real number structure (or, for
that matter, virtually any structurestudied in mathematics) can be "mod-
eled" in the set-theoretic hierarchy explains the occasional use of set the-
ory to prove theorems in arithmetic and analysis.
I propose, then, that branches of pure mathematics and branches of
applied mathematics have a common subject matter, structures. The in-
terplay between them is a result of the modeling of one structure within
anotheror, in other words, the use of one structureto study another struc-
ture. The difference between pure mathematics and applied mathematics
is, at least in part, an accident of history. Applied mathematics studies
structureswhich have traditionally formed part of particularscientific the-
ories. Thus, unlike some of the previous philosophies of mathematics,
structuralismdoes not presuppose a sharp distinction between branches

544 STEWART SHAPIRO
of pure mathematics and branches of applied mathematics.

I would suggest, in conclusion, that there is something like a contin-
uum with, say, set theory at one end, applied mathematics and theoretical
science toward the middle, and experimental science at the other end.23
6. Addendum. In a recent, provocative essay (1980), HartryField pre-
sents an account of mathematical application to accord with his overall
antiplatonist views.24 Although he probably would not agree, I believe
that his basic idea accords with structuralism.
There are two problems that Field attempts to resolve. The first is to
indicate a method of providing a "nominalistic" formulation of each ac-
cepted scientific theory, a formulation that does not refer to, or have
variables ranging over, abstractentities. The second is to account for how
the addition of a mathematical theory to a nominalistic theory can be
useful without presupposing that the mathematical assertions are true or,
indeed, that there is a subject matter for mathematics. I briefly discuss
each aspect.
For the purpose of formulating acceptable theories, Field develops in
detail a nominalistic version of Newtonian gravitational theory, a theory
that is to serve as a paradigm for other theories and branches of science.
Field's theory postulates, and has variables ranging over, both spacetime
points and regions, the latter regarded as classes (or "Goodmanian sums")
of space-time points.25 The axioms are formulated in terms of certain
primitive properties and relations among space-time points. Examples of
these include 'y Bet xz' interpreted as "x, y, and z are colinear and y is
between x and z" and 'xy S-Cong zw' interpreted as "x and y are simul-
taneous as are z and w and the distance between x and y is equal to the
distance between z and w".
The axioms of this theory entail that there are uncountably many space-
time points and, in particular, that the space-time points form a four di-
mensional continuum. Field calls these "structuralproperties" of space-
time points. He admits that his view "builds into physical space all the
complexity and structure that the platonist builds into the real number
system", but he adds:
23Acertain section of the Hebrew University of Jerusalem consists of three buildings in

a row. The rightmost is labeled "Mathematics", the middle "Applied Mathematics and
Theoretical Physics", and the leftmost "Physics". This is a definite improvement over the
usual state of affairs, but I would prefer one long building.
24Fieldholds a "nominalistic" philosophy which rejects the existence of all "abstract
entities", such as numbers and members of the set-theoretic hierarchy. Thus, statements
of mathematics are regarded as fictions. Concerning acceptable theories, Field follows
Nelson Goodman (1972) in allowing variables to range over "nominalistically existing"
objects and non-empty classes or "sums" thereof, but not over such items as classes of
classes.
25Accordingto Field, this commitment to space-time points does not violate nominalism.
As indicated in note 17 above, he takes such points to be theoretical physical entities on
a par with, say, electrons.

. . . the nominalistic objection to using real numbers was not on the

grounds of their uncountability or of the structuralassumptions . . .
typically made about them. Rather, the objection was to their ab-
stractness: even postulating one real number would have been a vi-
olation of nominalism as I'm conceiving it. Conversely, postulating
uncountably many physical entities . . . is not an objection to nom-
inalism; nor does it become any more objectionable when one pos-
tulates that these physical entities obey structural assumptions anal-
ogous to the ones that platonists postulate for the real numbers (1980,
p. 31).
This view has at least something in common with the above structur-
alist account of mathematical application. To "build structure in" seems
to be what I call "exemplifying" or "incorporating"a structure. In the
case of Newtonian gravitational theory, Field's "structuralassumptions"
amount to a claim that the class of space-time points exemplifies the
structureof R4 under what may be called the relation of "linear between-
ness". (This important substructureof R4 is similar to the structure stud-
ied in Euclidian geometry.) In these terms, Field's nominalism seems to
involve an assumption that every structure actually exemplified in phys-
ical reality can be completely described (in terms of primitive relations)
in a language that only has variables ranging over places within the struc-
ture and nonempty classes of such places. This quarantees that all the-
oretical entities postulated by a given branch of science-the entities an-
swering to the places within the structure-satisfy Field's nominalistic
scruples.
Of course, Field would not accept this description of his program. He
would certainly balk at my frank talk of structures(ratherthan "structural
assumptions" or "structuralproperties"). As indicated above, however,
I do not wish to make any bold ontological assertions (either way) con-
cerning structures themselves. My main disagreement with Field seems
to be in the philosophy of mathematics. He sees statements of, say, real
analysis as statements about entities, called "real numbers", whose ex-
istence is dubious. I, on the other hand, see statements of real analysis
as statements about a structure. The "objects" of the theory, the real num-
bers, are no more than the places in the structure. To put it another way,
on my view, the so-called "structural"properties of the real numbers are
all there is to say about them.26
I turn to Field's views on the use of mathematics within science. The
major claim is that mathematical theories are conservative over nominal-
26Accordingto my view, one cannot make sense of the phrase, "even postulating one
real number" in the passage quoted above. One cannot "postulate" one place within a
structurewithout invoking the structureas a whole. Similarly, one cannot speak of a "child"
without presupposing a "parent", or a "shortstop" without presupposing an "infield".

546 STEWART SHAPIRO
istic theories within science in that there is no nominalistic assertion of

the science that can be derived in a combined theory that cannot be de-
rived in the nominalistic theory alone. Thus, according to Field, math-
ematics can shorten derivations, but in principle it is dispensible.
Avoiding a technical device Field employs (whose purpose is of no
concern here), let N be a nominalistic theory within science and S a math-
ematical theory such as set theory. According to Field, the way mathe-
matics is usually applied is that in the combined theory N + S each nom-
inalistic statementA is proved to be equivalentto a statementA' formulated
in the language of the mathematical theory S. The statement A' is called
an abstract counterpart to A. If the mathematical theory S is conservative
over the nominalistic theory N, then this process is acceptable to a nom-
inalistically-minded scientist. In this case, the use of mathematics (in N
+ S) is theoretically superfluous; hence it is ontologically harmless.27
In Field's version of Newtonian gravitational theory, the equivalences
are developed as follows: First, one proves in the combined theory that
there is a structure-preserving-function, a homomorphism, from the points
of space-time to R4. That is, one proves that there is a function f from
the class of space-time points to the set of quadruples of real numbers
such that, for example, if (a,b) is congruent to (c,d) then the distance in
R4 between f(a) and f(b) is equal to the distance in R4 between f(c) and
f(d). On the basis of this homomorphism, one can "translate"any state-
ment about space-time points into an equivalent statement about real num-
bers. The scientist is then free to use features of real analysis which are
not available in the nominalistic theory of space-time. Finally, some of
the conclusions in the combined theory can be translatedback to nominal-
istic conclusions in the language of space-time.
A structuralisthas a rather straightforwardinterpretationof this work.
Recall that Field's space-time points exemplify a certain mathematical
structure, a substructureof R4. The work concerning the "combined the-
ory" corresponds to a series of mathematical theorems relating this struc-
ture to the full structureof R4. The main result (concerning the existence
of the homomorphismf) is a well-known mathematical theorem that the
structureof R4 under linear betweenness can be extended to full R4 through
the introduction of a reference frame and a metric. Thus construed, the
use of mathematics in nominalistic physics is but one example of the
phenomenon, discussed above, of embedding or modeling a simple math-
ematical structure in a richer structure and using the latter to shed light
on the former.
27Shapiro(1983) contains a theorem to the effect that the relevant mathematical theories
are not conservative over the nominalistic physics that Field develops. In particular, it is
shown that there is a nominalistic assertion (in the language of N) which is a theorem of
the combined theory (N + S), but not a theorem of the nominalistic physics alone.

It might be pointed out in closing that Field's claims concerning math-

ematical application and his program are of considerable interest to a
structuralist, even one with platonistic leanings. If Field's beliefs con-
cerning science are correct, then they determine the kinds of mathematical
structuresactually exemplified in physical reality and, moreover, distin-
guish these from the structures used by mathematicians and scientists as
conceptual aids. Even if Field's basic assumptions are incorrect, his pro-
gram should shed light on these matters.
REFERENCES
Ayer, A. (1946), Language, Truth, and Logic. New York: Dover publications.
Barbut, M. (1970), "On the meaning of the word 'structure' in mathematics" in Lane,
pp. 367-388.
Benacerraf, P. (1965), "What numbers could not be", Philosophical Review 74: 47-73.
Bourbaki, N. (1950), "The architecture of mathematics", American Mathematical Monthly
57: 221-232.
Budden, F. J. (1972), The Fascination of Groups. Cambridge, England: Cambridge Uni-
versity Press.
Cohen, P. (1971), "Comments on the foundations of set theory" in D. Scott (ed.), Axi-
omatic Set Theory. Providence, Rhode Island: American Mathematical Society, pp.
9-15.
Crowell, R. and Fox, R. (1963), Introduction to Knot Theory. Boston: Ginn and Company.
Curry, H. (1958), Outlines of a Formalist Philosophy of Mathematics. Amsterdam: North
Holland Publishing Company.
Dedekind, R. (1888), "The nature and meaning of numbers" in R. Dedekind, Essays on
the Theory of Numbers, W. W. Beman (ed.) (1963). New York: Dover Press, pp.
31-115.
Dummett, M. (1973), "The philosophical basis of intuitionistic logic", reprinted in M.
Dummett, Truth and Other Enigmas (1978). Cambridge, Massachusetts: Harvard
University Press, pp. 215-247.
Field, H. (1980), Science Without Numbers. Princeton: Princeton University Press.
Frege, G. (1903), Grundgesetze der Arithmetic, Volume 2.
Glymour, C. (1980), Theory and Evidence. Princeton: Princeton University Press.
Godel, K. (1964), "What is Cantor's continuum problem", in P. Benacerraf and H. Put-
nam (eds.), Philosophy of Mathematics. Englewood Cliffs, New Jersey: Prentice-
Hall, pp. 258-273.
Goodman, Nelson (1972), Problems and Projects. Indianapolis: Bobbs-Merill.
Goodman, Nicolas D. (1979), "Mathematics as an objective science", American Mathe-
matical Monthly 88: 540-551.
Harman, G. (1975), "Meaning and semantics", in Semantics and Philosophy. New York:
NYU Press, pp. 1-16.
Hempel, C. (1945), "On the natureof mathematicaltruth",American Mathematical Monthly
52: 543-556.
Heyting, A. (1956), Intuitionism, An Introduction. Amsterdam: North Holland Publishing
Company.
Kraut, R. (1980), "Indiscernibility and ontology", Synthese 44: 113-135.
Lane, M. (ed.) (1970), Introduction to Structuralism. New York: Basic Books.
Maddy, P. (1980), "Perception and mathematical intuition", The Philosophical Review 89:
163-196.
Nutter, J. T. (1980), Structuralism in the philosophy of mathematics, Ph.D. Dissertation,
SUNY/Buffalo.
Piaget, J. (1968), Le Structuralisme. Paris: Presses Universitaires de France.
Polya, G. (1954), Mathematics and Plausible Reasoning. Princeton: Princeton University
Press.

548 STEWARTSHAPIRO
Polya, G. (1977), Mathematical Methods in Science. Washington, D.C.: Mathematical

Association of America.
Putnam, H. (1971), Philosophy of Logic. New York: Harper Torchbooks.
Resnik, M. (1975), "Mathematical knowledge and pattern cognition", Canadian Journal
of Philosophy 5: 25-39.
Resnik, M. (1980), Frege and The Philosophy of Mathematics. Ithaca, New York: Cornell
University Press.
Resnik, M. (1981), "Mathematics as a science of patterns: Ontology and reference", Nous
15: 529-550.
Resnik, M. (1982), "Mathematics as a science of patterns: Epistemology", Nous 16: 95-
105.
Robinson, A. (1965), "Formalism" in Y. Bar-Hillel (ed.), Logic, Methodology and Phi-
losophy of Science. Amsterdam: North Holland Publishing Company, pp. 228-246.
Shapiro, S. (1981), "Understanding Church's thesis", Journal of Philosophical Logic 10:
353-365.
Shapiro, S. (1983), "Conservativeness and incompleteness", Journal of Philosophy 80:
521-531.
Steiner, M. (1975), Mathematical Knowledge. Ithaca, New York: Cornell University Press.
Turing, A. (1936), "On computable numbers, with an application to the Entscheidungs-
problem", reprinted in M. Davis (ed.), The Undecidable (1965). Hewlett, New York:
The Raven Press, pp. 116-153.
Turnbull, R. (1978), "Knowledge of the forms in the later platonic dialogues", Proceed-
ings and Addresses of the American Philosophical Association 51: 735-758.
Wilson, M. (1981), "The double standard in ontology", Philosophical Studies 39: 409-
427.


Mathematics and Reality

Uploaded by

Copyright:

Available Formats

Mathematics and Reality

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mathematics and Reality

Uploaded by

Copyright:

Available Formats

Mathematics and Reality

Author(s): Stewart Shapiro

This content downloaded from 140.254.87.149 on Sun, 19 Oct 2014 15:14:13 PM