“Clarifying the nature of the infinite”:

the development of metamathematics

and proof theory∗
Jeremy Avigad and Erich H. Reck
December 11, 2001

Abstract
We discuss the development of metamathematics in the Hilbert school,
and Hilbert’s proof-theoretic program in particular. We place this pro-
gram in a broader historical and philosophical context, especially with
respect to nineteenth century developments in mathematics and logic.
Finally, we show how these considerations help frame our understanding
of metamathematics and proof theory today.

Contents
1 Introduction 2

2 Nineteenth century developments 4

2.1 Developments in mathematics . . . . . . . . . . . . . . . . . . . . 5
2.2 Changing views of mathematics . . . . . . . . . . . . . . . . . . . 7
2.3 Developments in logic . . . . . . . . . . . . . . . . . . . . . . . . 12

3 Hilbert’s proof theory 15

3.1 Hilbert as mathematician . . . . . . . . . . . . . . . . . . . . . . 15
3.2 Logic and the Hilbert school, from 1899 to 1922 . . . . . . . . . . 17
3.3 The emergence of “Beweistheorie” . . . . . . . . . . . . . . . . . 21

4 Formalism and Hilbert’s program 23

4.1 Strong formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 A more tenable formalist viewpoint . . . . . . . . . . . . . . . . . 26
4.3 Clarifying the nature of the infinite . . . . . . . . . . . . . . . . . 29
∗ Carnegie Mellon Technical Report CMU-PHIL-120. We are grateful to Steve Awodey,
Paolo Mancosu, Colin McLarty, Gisbert Selke, Wilfried Sieg, and W. W. Tait for comments
and suggestions.

1
5 Proof theory after Hilbert 32
5.1 Proof theory in the 1930’s and 1940’s . . . . . . . . . . . . . . . . 32
5.2 Proof theory in the 1950’s . . . . . . . . . . . . . . . . . . . . . . 34

6 Proof theory today 37

6.1 Contemporary perspectives . . . . . . . . . . . . . . . . . . . . . 37
6.2 Proof theory and constructive mathematics . . . . . . . . . . . . 40
6.3 Contemporary goals . . . . . . . . . . . . . . . . . . . . . . . . . 41

1 Introduction
The phrase “mathematical logic” can be understood in two ways. On the one
hand, one can take the word “mathematical” to characterize the subject mat-
ter. Understood as such, “mathematical logic” refers to the general study of
the principles of mathematical reasoning, in much the same way as the phrase
“inductive logic” refers to the study of empirical reasoning, or “modal logic”
refers to the study of reasoning that involves modalities. On the other hand,
one can take the word “mathematical” to characterize the methods of the dis-
cipline in question. From this point of view, “mathematical logic” refers to
the general study of the principles of reasoning, mathematical or otherwise, us-
ing specifically mathematical methods and techniques. We are using the word
“metamathematics” in the title of this paper to refer to the intersection of these
two interpretations, amounting to the mathematical study of the principles of
mathematical reasoning. The additional phrase, “and proof theory,” then de-
notes a particular metamathematical approach, in which the aim is to model
mathematical practice by giving a formal account of the notion of proof.
This terminological clarification is not just legalistic hair-splitting. Logic in
general, as a separate field of study, finds its roots in Aristotle’s treatment of
syllogistic reasoning. The application of mathematical techniques to logic dates
back, perhaps, to Leibniz, and was revived and substantiated in the works of
Boole in the mid-nineteenth century. In contrast, Frege developed his logic
specifically with applications to the foundations of mathematics in mind. By
the end of the nineteenth century various aspects of logic in both senses had,
thus, been recognized and studied separately. But it wasn’t until the twentieth
century that there was a clear delineation of the field we have called metamath-
ematics; nor was there, before that time, a clear sense that the synthesis of the
two disciplines — “mathematical logic” in both senses above — might be a fruit-
ful one. This synthesis was set forth most distinctly by David Hilbert in a series
of papers and lectures in the 1920’s, forming the basis for his Beweistheorie.
Hilbert’s program, together with its associated formalist viewpoint, launched
the field of Proof Theory, and influenced the development of mathematical logic
for years to come.
Historical hindsight has, however, not been kind to Hilbert’s program. This
program is often viewed as a single-minded and ultimately inadequate response
to the well-known “foundational crisis” in mathematics early in the twentieth

2
century, and is usually contrasted with rival intuitionist and logicist programs.
In the preface to his Introduction to Metamathematics, Stephen Kleene summa-
rizes the early history of mathematical logic as follows:
Two successive eras of investigations of the foundations of mathe-
matics in the nineteenth century, culminating in the theory of sets
and the arithmetization of analysis, led around 1900 to a new crisis,
and a new era dominated by the programs of Russell and Whitehead,
Hilbert and of Brouwer.1
Conventional accounts of these developments tell a tale of epic proportions:
the nineteenth century search for certainty and secure foundations; the threats
posed by the discovery of logical and set-theoretic paradoxes; divergent re-
sponses to these threats, giving rise to heated ideological battles between the
various schools of thought; Hilbert’s proposal of securing mathematics by means
of consistency proofs; and Gödel’s final coup de grace with his incompleteness
theorems. Seen in this light, Hilbert’s program is unattractive in a number of
respects. For one thing, it appears to be too single-minded, narrowly concerned
with resolving a “crisis” that seems far less serious now, three quarters of a cen-
tury later. Moreover, from a philosophical point of view, strict formalism does
not provide a satisfying account of mathematical practice; few are willing to
characterize mathematics as an empty symbol game. Finally, and most point-
edly, from a mathematical point of view Gödel’s second incompleteness theorem
tells us that Hilbert’s program was, at worst, naive, and, at best, simply a fail-
ure.
In this paper we will argue that the conventional accounts are misleading,
and fail to do justice to the deeper motivations and wider goals of Hilbert’s
Beweistheorie. In particular, they overemphasize the role of the crisis, and pay
too little attention to the broader historical developments. Furthermore, the
usual way of contrasting Hilbert’s program with rival logicist and intuitionist
ones provides a misleading impression of the relationship of Hilbert’s views to
logicist and constructive viewpoints. This has led to narrow, distorted views not
only of Hilbert’s program, but also of research in metamathematics and proof
theory today.
We are, by no means, the first to argue against the conventional accounts
along these lines.2 As will become clear, in writing this essay we have drawn
on a good deal of contemporary scholarship, as well as historical sources. Our
goal has been to present this material in a way that helps put modern research
in metamathematics and proof theory in a broader historical and philosophi-
cal context. In addition, we try to correct some common misperceptions, and
establish the following claims.
First, we will argue that Hilbert’s proof theoretic program was not just a
response to the foundational crisis of the twentieth century. We will show that
1 Kleene[71], page v.
2 See,
for example, Feferman [33], Sieg [103], Sieg [105], Stein [110], the introductory notes
in volume 2 of Ewald [31], or the introductory notes to part 3 of Mancosu [83].

3
it addresses, more broadly, a deeper tension between two trends in mathematics
that began to diverge in the nineteenth century: general conceptual reasoning
about abstractly characterized mathematical structures, on the one hand, and
computationally explicit reasoning about symbolically represented objects, on
the other. We will show that one of the strengths of Hilbert’s program lies in
its ability to reconcile these two aspects of mathematics. Second, with its sharp
focus on deductive reasoning, syntax, and consistency proofs, Hilbert’s program
is often viewed as philosophically and methodologically narrow. But Hilbert was
sensitive to a broad range of mathematical issues, as well as to the diversity of
goals and methods in mathematical logic. We will argue that he deserves more
credit than is usually accorded him for clarifying and synthesizing the various
viewpoints. Third, Hilbert’s program is usually associated with a type of formal-
ism that relies on very restrictive ontological and epistemological assumptions.
We will argue that both Hilbert’s writings and the historical context make it
unlikely that Hilbert would have subscribed to such views, and show that far
weaker assumptions are needed to justify his metamathematical methodology.
We will thereby characterize a weak version of formalism that is still consistent
with Hilbert’s writings, yet is compatible with a range of broader philosophical
stances. Finally, we will show how these considerations are important to under-
standing current research in metamathematics and proof theory. We will discuss
the general mathematical, philosophical, and computational goals that are pur-
sued by contemporary proof theorists, and we will see that many of Hilbert’s
central insights are still viable and relevant today.

2 Nineteenth century developments

Conventional accounts of the “crisis of foundations” of the early twentieth cen-
tury portray the flurry of research in the foundations of mathematics as a direct
response to two related developments: the discovery of the logical and set-
theoretic paradoxes around the turn of the century, most prominently Russell’s
paradox, and the ensuing foundational challenges mounted by Henri Poincaré,
L. E. J. Brouwer, and others. We will argue that it is more revealing to view
the foundational developments as stemming from, and continuous with, broader
developments in the nineteenth century, thereby understanding the stances that
emerged in the twentieth century as evolving responses to tensions that were
present earlier.
In Section 2.1, we will briefly summarize some relevant changes to the math-
ematical landscape that took place in the nineteenth century. In Sections 2.2
and 2.3, we will discuss the ways these internal developments were accompanied
by evolving conceptions of the nature of mathematics, as well as of the nature
of mathematical logic.

4
2.1 Developments in mathematics
It is not unreasonable to trace the origins of much of contemporary mathematics
to the nineteenth century, “an age of deep qualitative transformations,”3 and
an age during which mathematics underwent, in the words of Howard Stein, “a
transformation so profound that it is not too much too call it a second birth
of the subject — its first birth having occurred among the ancient Greeks.”4
Though it is well beyond the scope of this paper to do justice to these events,
the following brief sketch of some of the major developments will help set the
stage for our subsequent discussion.
To begin with, during this period geometry moved well beyond the long dom-
inant Euclidean paradigm. Poncelet launched the field of projective geometry,
which deals with those properties of geometric figures that are invariant under
projections, and he saw that introducing “points at infinity” allowed for a uni-
form treatment of the conic sections. Early in the century Gauss, Lobachevsky,
and Bolyai came independently to the conclusion that one could not prove Eu-
clid’s fifth axiom from the other four, and they studied the consequences of its
negation; concrete interpretations of non-Euclidean geometries were later pro-
vided by Beltrami, Cayley, Klein, and Riemann. Moreover, in a sweeping work
presented in 1854 (but not published until 1868), Riemann brought even greater
generality to the term “geometry” by introducing the study of general geometric
manifolds of varying curvature.
Algebra was similarly transformed. The century saw the development of
Galois’ theory of equations; the rise of the British school of algebraists, includ-
ing Hamilton, Cayley, Boole, and Sylvester; and Grassmann’s work on algebras
generalizing Hamilton’s quaternions. Even more dramatic were the evolving
interactions of algebra with other branches of mathematics. In France and
Germany, algebraic methods in geometry took their place beside traditional
synthetic methods, a trend which reached a high point with Klein’s Erlanger
Programm of 1872. Having established a first foothold with Gauss, algebraic
methods also gained currency in number theory, in the work of mathematicians
like Kummer, Germain, and Dedekind. Niels Abel used algebraic methods in
his studies of the elliptic functions, and such methods in analysis flowered in
the hands of Jacobi, Dirichlet, Riemann, and Hamilton. Similarly, Boole was
respected for his algebraic treatment of integration, though, of course, he is best
known for his algebraic approach to logic. These developments are summed up
by the observation that “between the beginning and the end of the last cen-
tury. . . the subject matter and methods of algebra and its place in mathematics
changed beyond recognition,”5 and, even more pointedly, that “there was a
manifest tendency of algebraization of mathematics.”6
The nineteenth century also saw major changes in analysis, including the
3 In Chapter 2 of [76], “Algebra and algebraic number theory,” by I. G. Bashmakova and

A. N. Rudakov with the assistance of A. N. Parshin and E. I. Slavutin, page 35. (In general,
when the translation of a work is quoted, pages numbers refer to the translations.)
4 Stein [110], page 238.
5 Chapter 2 of [76], Page 35.
6 Ibid.

5
development of Fourier analysis, the study of elliptic functions and differen-
tial equations, and the rise of complex analysis. Analytic methods in number
theory, incipient in the work of Euler, were put to dramatic use by Dirichlet
and Riemann; for example, in Dirichlet’s proof that there are infinitely many
primes in any arithmetic progressions in which terms are relatively prime, and
in Riemann’s studies of the zeta function. Of course, there was the now fa-
miliar “rigorization” of analysis, carried out by Bolzano, Cauchy, Weierstrass,
Dirichlet, and others. This resulted in the modern epsilon-delta definitions of
limits, continuity, differentiation, and integration, which grounded calculus on
a foundation free of Leibniz’ infinitesimals and Newton’s fluxions and fluents.
Finally, there was a good deal of work of a foundational nature. Bolzano’s
1851 Paradoxes of the Infinite (published posthumously) can be seen as an early
call to buttress the foundations of mathematics. The rigorization of analysis was
one contribution to this general program, and was complemented by efforts from
Dedekind, Cantor, Kronecker, etc., to move the conception of the real numbers
from a geometric foundation to a new foundation based on the natural numbers.
In addition, modern logic was born with the work of Frege, Boole, and Peirce.
Towards the end of the nineteenth century, Frege and Dedekind concerned them-
selves with the foundations of arithmetic; Pasch addressed the foundations of
geometry; and Peano and the “Italian school” pushed for a symbolic and ax-
iomatic treatment of vast tracts of mathematics.
To help make sense of these developments, let us identify a few important
and interrelated trends. The first is a general shift towards abstraction, accom-
panied by a growing division between mathematics and the physical sciences.
Geometry evolved from the study of “space” to the study of general “spaces.”
Algebraic developments emphasized not the study of the natural or the real num-
bers, but rather, the study of general algebraic systems, with a clear emphasis on
the fact that such systems can carry multiple interpretations. Moreover, there
was an increasing focus on general descriptions and abstract characterizations of
mathematical objects. This can be seen, for example, in the characterization of
geometric spaces by groups of transformations, or by their intrinsic metric prop-
erties; and in the characterization of fields by groups of automorphisms. The
various axiomatic approaches to algebra, arithmetic, and analysis that emerged
towards the end of the century can also be understood in this way.
The new emphasis on abstract characterization brought with it a corre-
sponding de-emphasis on calculation. This second shift was, in fact, not just a
by-product of the new developments, but often a central goal. In a later ret-
rospective on Dirichlet’s work, Hermann Minkowski identifies and endorses this
goal when he extols “the other Dirichlet Principle: to conquer the problems
with a minimum of blind calculation, a maximum of clear-sighted thoughts.”7
Of course, the preference for abstract characterization over “blind calculation”
was not universal and uncontroversial; we will return to a discussion of this issue
later.
7 From Minkowski’s 1905 address in Göttingen, found in his Gesammelte Abhandlungen,

volume 2, Leipzig, 1911. This passage is also quoted in in Stein [110], page 241.

6
 A third notable trend is a growing confidence in dealing with the infinite.
From Archimedes’ method of exhaustion to the development of the calculus in
the seventeenth and eighteenth centuries, statements about the infinite were
viewed as statements about arbitrarily large quantities or limiting processes —
that is, about “potential” infinities — while references to the “actual” infinite
were treated with suspicion. Now, in contrast, completed infinitary structures
were taken at face value. In the case of geometry, G. H. Hardy has empha-
sized the importance of understanding points at infinity as objects in their own
right,8 and much has also been written about the shift from the concept of
a function as an analytic expression or rule to the concept of an “arbitrary”
function, often attributed to Dirichlet.9 Most centrally, by the 1880’s a number
of mathematical developments had contributed towards establishing a method-
ological framework for dealing with infinite totalities, including constructions of
the natural numbers by Dedekind and Frege; constructions of the real numbers
by Weierstrass, Dedekind, Cantor, and others; Dedekind’s development of the
theory of ideals in algebraic number theory; and Cantor’s investigations of the
transfinite.
The last trend we wish to highlight is a dramatically fruitful cross-fertilization
between the various branches of mathematics. We have already noted several
examples, such as the use of algebraic methods in geometry, number theory,
analysis, and logic; and the use of analytic methods in geometry and number
theory. With this increasing interaction came a growing interest in unification,
as mathematicians tried to view the various aspects of their subject as part of
a coherent whole. The desire for a general mathematical framework involved a
good deal of reflection, as mathematicians tried to identify the common features
of their various mathematical activities; and it involved an interest in rigor and
explicitness, as general methodological standards for mathematics emerged.

2.2 Changing views of mathematics

The developments we have just discussed can be seen as changes that are internal
to mathematical practice. But they both precipitated and were facilitated by a
shifting understanding of the nature of mathematics itself, first, at the level of
informal views held by mathematicians as to the general goals and the proper
methods of their subject, and, later, at the level of more extended philosophical
reflection.
Before the nineteenth century, the subject of mathematics was commonly
divided into two branches, the science of number and the science of space.
Both of these were sometimes subsumed under the general characterization of
mathematics as the science of measure, magnitude, or quantity. Thus, in the
seventeenth century, Descartes observed that only those disciplines that involve
order or measure are regarded as “mathematical,” and suggested recognizing a
“pure” branch of the subject:
8 Hardy [53], Appendix IV, pages 502–3.
9 But see Monna [86] and Youschkevich [122] for a fuller discussion.

7
 It then follows that there must be a certain general science which
explains everything which can be asked about order and measure,
and which is concerned with no particular subject matter, and that
this very thing is called “pure mathematics” [mathesis universalis].10
In the next century, Euler wrote:
Mathematics, in general, is the science of quantity; or, the science
that investigates the means of measuring quantity.11
And Gauss maintained:
Mathematics really teaches general truths which concern the rela-
tions of magnitudes.12
Nineteenth century developments, however, strained these definitions. To
begin with, the multiplicity of geometric spaces, including higher dimensional
ones, that were now within the ambit of mathematical research indicated that
geometry could no longer be viewed as the study of physical space, and high-
lighted the fact mathematics need not have direct application to the physical
world. The acceptance of infinitary notions, like points at infinity, algebraic
ideals, and transfinite ordinals, also served to distance mathematics from direct
application to the physical world, and stretched the notions of “quantity” and
“magnitude” considerably.
Of course, the view of mathematics as an abstract science was not new in
and of itself. It was commonly accepted that mathematical abstractions could
be assigned multiple interpretations; for example, Eudoxus’ theory of propor-
tions could be applied to different kinds of magnitudes, including lengths, areas,
and volumes; various astronomical systems could be described as instances of,
say, circular or elliptic motion; and the methods of probability could be applied,
alternatively, to problems involving gambling, problems of insurance, and prob-
lems having to do with measurement subject to errors. But as mathematical
abstractions were applied increasingly to the study of other mathematical ob-
jects, like equations or geometric spaces, the abstractions themselves began to
take on a life of their own. What was sought after, therefore, was a general char-
acterization of mathematics that could accommodate the new developments.
One route to such a characterization involved altering the conception of the
subject matter of mathematics, i.e. revising the view as to what mathematics
is about. This is illustrated by a proposal due to Boole, in The Mathematical
Analysis of Logic (1847). Boole begins by acknowledging the traditional view
(here, “Analysis” refers to mathematical activities in general):

Thus the abstractions of mathematical Analysis, not less than the

ostensive diagrams of ancient Geometry, have encouraged the notion,
10 From “Rule four” of Descartes [27], page 17.
11 From “On Mathematics in General”, reprinted in [100], pages 37–41, source and translator
not named. The quoted passage is from p. 37.
12 Gauss [43], page 42 of the reprinting in [100].

8
 that Mathematics are essentially, as well as actually, the Science of
Magnitude.13
But he goes on to argue that this definition is too restrictive, and based on a
sample of mathematical activities that is too small. He proposes the following
redefinition:
We might justly assign it as the definitive character of a true Cal-
culus, that it is a method resting upon the employment of Symbols,
whose laws of combination are known and general, and whose re-
sults admit of a consistent interpretation. That to existing forms of
Analysis a quantitative interpretation is assigned, is the result of the
circumstances by which those forms were determined, and is not to
be construed into a universal condition of Analysis.14
That is to say, mathematics is to be viewed as the science of symbolic calculi,
which may or may not have quantitative interpretations. With this definition,
Boole stands firmly in the tradition of British algebraists, for whom an algebraic
system consisted of a collection of symbols together with explicit rules governing
their use. What is novel, however, is his adoption of a syntactic standpoint to
characterize mathematics in general, by proposing to define mathematics as the
science of such systems.
An alternative route to redefining mathematics is to focus on the abstract
concepts used in mathematical descriptions and characterizations, rather than
on their symbolic representations. This shift of emphasis suggests allowing
mathematicians the latitude to study these concepts, independent from imme-
diate questions as to their applicability in physics and other sciences. Something
of this sort underlies Georg Cantor’s famous characterization of mathematics,
from 1883:
Mathematics is in its development entirely free and is only bound
in the self-evident respect that its concepts must both be consistent
with each other and also stand in exact relationships, ordered by
definitions, to those concepts which have previously been introduced
and are already at hand and established.15
The context of this passage is a discussion of Cantor’s new theory of transfinite
numbers, but the same emphasis on freedom can be found in connection with
traditional branches of mathematics as well. Thus Richard Dedekind writes in
1889:
In speaking of arithmetic (algebra, analysis) as merely a part of logic
I mean to imply that I consider the number-concept entirely inde-
pendent of the notions or intuitions of space and time . . . . Numbers
are free creations of the human mind. . . 16
13 Boole [14], page 4 of the original edition.
14 Ibid.
15 Cantor [20], Section 8, paragraph 4.
16 Dedekind [25], pages 790–791.

9
After having put aside intuitions of space and time, Dedekind then characterizes
the system of natural numbers purely “logically,” and hence abstractly. His
earlier 1872 characterization of the real numbers [24] has a similar character.
What we have arrived at, then, are two new conceptions of mathematics,
both generalizing the narrower characterizations offered by Descartes, Euler,
and Gauss above. To be explicit: one might view mathematics as the science
of symbolic representation and calculation; or one might view mathematics as
the science of conceptual reasoning about structures, characterized abstractly.
These two conceptions are not necessarily incompatible, but they differ in im-
portant respects. In particular, at the very down-to-earth level of everyday
mathematical practice, they suggest different styles of research, involving dif-
ferent methods and goals.
The divergence becomes more pronounced when one considers the views of
Leopold Kronecker, for whom symbolic and algorithmic issues are central. In
“On the concept of number” (1887), he writes:
[A]ll the results of the profoundest mathematical research must in
the end be expressible in the simple forms of the properties of in-
tegers. But to let these forms appear simply, one needs above all
a suitable, surveyable manner of expression and representation for
the numbers themselves. The human spirit has been working on this
project persistently and laboriously since the greyest prehistory. . . .17
This focus on symbolic representation pervades Kronecker’s work.18 But he
does not just emphasize the general symbolic nature of mathematics, as Boole
did; he places issues of calculation at the core of any mathematical pursuit. The
corresponding methodological stance is nicely captured in the following passage,
written by his student Hensel:
He [Kronecker] believed that one can and must in this domain for-
mulate each definition in such a way that its applicability to a given
quantity can be assessed by means of a finite number of tests. Like-
wise that an existence proof for a quantity is to be regarded as
entirely rigorous only if it contains a method by which that quantity
can really be found.19
The gradual rise of the opposing viewpoint, with its emphasis on conceptual
reasoning and abstract characterization, is elegantly chronicled by Stein[110],
as part and parcel of what he refers to as the “second birth” of mathematics.
The following quote, from Dedekind, makes the difference of opinion very clear:
A theory based upon calculation would, as it seems to me, not offer
the highest degree of perfection; it is preferable, as in the modern
theory of functions, to seek to draw the demonstrations no longer
17 “Überden Zahlbegriff,” Kronecker [81], page 955.
18 See,for example, [29, 30, 85] and the introductory notes on Kronecker in volume 2 of [31].
19 Found in L. Kronecker, Vorlesungen über Zahlentheorie, K. Hensel ed., Teubner, Leipzig,

1901, page vi. Quoted in Stein [110], page 250, and Marion [85], page 191.

10
 from calculations, but directly from the characteristic fundamental
concepts, and to construct the theory in such a way that it will, on
the contrary, be in a position to predict the results of the calculation
(for example, the decomposable forms of a degree).20
In other words, from the Cantor-Dedekind point of view, abstract conceptual
investigation is to be preferred over calculation.
The significance of the opposition between Dedekind and Kronecker becomes
clearer when we consider how the two viewpoints result in very different ap-
proaches towards dealing with the infinite. On Kronecker’s view of mathemat-
ics, one is only permitted to deal with infinitary notions, such as that of a real
number or an ideal, via explicit symbolic representation; and one is then only
allowed to consider operations and relationships that are algorithmically pre-
sented. On the view expounded by Dedekind, infinitary mathematical objects
may be characterized abstractly and taken at face value, subject to constraints of
“determinateness” and “consistency.” This second — revolutionary and liberal
— conception is backed by strong rhetoric, especially from Cantor:
It is not necessary, I believe, to fear, as many do, that these principles
present any danger to science. For in the first place the designated
conditions, under which alone the freedom to form numbers can be
practised, are of such a kind as to allow only the narrowest scope
for discretion. Moreover, every mathematical concept carries within
itself the necessary corrective: if it is fruitless or unsuited to its pur-
pose, then that appears very soon through its uselessness, and it will
be abandoned for lack of success. But every superfluous constraint
on the urge to mathematical investigation seems to me to bring with
it a much greater danger, all the more serious because in fact abso-
lutely no justification for such constraints can be advanced from the
essence of the science — for the essence of mathematics lies precisely
in its freedom.21
It is important to be clear as to what is at stake in these debates. From a
contemporary point of view, it is tempting to read into this opposition general
disagreements about the ontology and epistemology of mathematics. In other
words, one may try to find in the associated writings precursors of the philosoph-
ical stances that emerged more clearly in the early twentieth century: logicism,
realism/platonism, formalism, finitism, intuitionism, etc. But although Cantor
indulges in some metaphysical speculation, and although some of Dedekind’s
writings suggest a structuralist point of view, Kronecker has almost nothing to
say about metaphysics; and all three writers are primarily concerned with the
mathematical implications of the divergent viewpoints. The effects that these
viewpoints have becomes evident when one compares the mathematical work of
Kronecker and Dedekind: both developed theories of algebraic ideals building
20 Dedekind, “Sur la théorie des Nombres entiers algébrique,” in his Werke, volume 3, page

296; quoted in Stein [110], page 245.

21 Cantor [20], Section 8, paragraph 5.

11
on the work of Dirichlet and Kummer, and both wrote on the foundations of
arithmetic; but despite the similarity of subject matter, their approaches stand
in stark opposition to one other.22
We are not suggesting that general ontological and epistemological consid-
erations played no role in the late nineteenth century debates, but, rather, that
such philosophical issues are often intertwined with mathematical ones. While
nineteenth century changes in mathematical practice were fueled by develop-
ments internal to the subject, coming to terms with these changes required the
more broadly philosophical task of characterizing mathematics and its goals
more generally; and such reflection influenced the informal viewpoints and atti-
tudes that guide choices of subject matter, problems, language, and methods.23
Finally, we want to emphasize that a real tension between abstraction and
calculation does not arise until one tries to come to terms with the infinite and
its proper role in mathematics. In the finitary realm, there is no sharp distinc-
tion between characterizing a mathematical structure uniquely and determining
what properties its elements have; indeed, this is essentially what one means by
calling a characterization “finitary.” A distinction only arises when one deals
with infinite structures. The difference is made clear by Turing’s later discovery
that the halting problem is unsolvable; this tells us that one can specify a subset
of the natural numbers (that is, describe a set of natural numbers via reference
to the infinite structure hN, 0, S, +, ×i) without having an algorithmic proce-
dure to determine whether or not a number is in this set. We have made this
distinction with the hindsight of a twentieth century analysis of computability
and definability. But even without the modern distinction, in the nineteenth
century it became clear that one could describe infinite sets and operations on
them without explicit reference to algorithmic representation. In the wake of the
Cantor-Dedekind revolution, these are the kinds of issues that mathematicians
had to come to terms with.

2.3 Developments in logic

Nineteenth century developments in logic, though perhaps not quite as dramatic
as the developments in mathematics, were still remarkable and substantive, eas-
ily refuting Kant’s earlier claim that the logic of his time was complete and in
no need of revision. In particular, the century brought the development of rela-
tional and full quantificational logic in the work of Frege; it brought innovative
formal and algebraic studies of a number of systems of logic by Boole, Peirce, and
their followers; and the end of the century brought axiomatic characterizations
of various mathematical structures and related studies of mathematical reason-
ing, in the hands of Pasch, Dedekind, Peano, Hilbert, etc. Just as nineteenth
22 See Edwards [28], Marion [85], and Stein [110].
23 One might view these developments as the evolution of two divergent “styles” of doing
mathematics, the first based on the understanding of mathematics as a science of abstract
description and reasoning, and the second based on the understanding of mathematics as a
science of symbolic representation and calculation. An analysis of this phenomenon, akin to
Ian Hacking’s analysis of “styles” in the sciences [52], may be illuminating.

12
century developments in mathematics altered common views of the subject, so
too did developments in logic require a revised assessment of the field, as well
as of its role in mathematics and philosophy.24
One thing that evolved were common views as to the relationship between
logic and mathematics. At the beginning of this essay, we described two ways of
interpreting the phrase “mathematical logic”; this distinction hinged on whether
one views logic as a a study of mathematical reasoning, or as a branch of math-
ematics. We can characterize this succinctly as the difference between studying
the “logic of mathematics” and studying the “mathematics of logic.” A sec-
ond, related issue is whether one’s primary goals in pursuing the subject are
philosophical or mathematical. From a philosophical point of view, one may
aim to clarify, characterize, or justify the basic principles of reasoning; in par-
ticular, one may wish to provide a philosophical foundation for mathematics.
From a mathematical point of view, the standards of success are likely to differ,
emphasizing intrinsic mathematical interest, interaction with other branches of
mathematics, and applicability to the sciences.
Various orientations towards logic can be found in the nineteenth century.
For example, Frege’s work is primarily, though not exclusively, motivated by
philosophical, foundational goals. In The Foundations of Arithmetic, his most
programmatic work, he writes with regard to the relationship between logic and
arithmetic:

The aim of proof is, in fact, not merely to place the truth of a
proposition beyond all doubt, but also to afford us insight into the
dependence of truths upon one another. The answers to the ques-
tions raised about the nature of arithmetic truths – are they a priori
or a posteriori? synthetic or analytic? – must lie in this same direc-
tion.25
In the background here are traditional philosophical questions as to the status
of various kinds of truths, such as those of arithmetic, especially in the context
of Kant’s philosophy of mathematics and of nineteenth century responses to it.
In contrast, Boole’s work represents a more mathematical perspective. In
the previous section, we considered Boole’s expansive definition of mathematics
as the general science of calculation. Now we can note that Boole’s explicit
reason for proposing this new definition is to argue for the inclusion of symbolic
logic among the branches of mathematics:
It is upon the foundation of this general principle, that I purpose to
establish a Calculus of Logic, and that I claim for it a place among
the acknowledged forms of Mathematical Analysis, regardless that
in its object and in its instruments it must at present stand alone.26
24 Much of the corresponding evolution is nicely and more completely chronicled by Peckhaus
[88].
25 Frege [41], pages 2–3.
26 Boole [14], page 4.

13
Later, in the introduction to The Laws of Thought (1854), Boole supplements his
mathematical analysis with a discussion of general philosophical issues. In fact,
the further development of logic from Boole to Jevons, one of Boole’s students,
may be characterized as a further step in this general direction.27 Nevertheless,
taken as a whole, The Laws of Thought has more the flavor of an algebraic
monograph than that of a philosophical one.
A further dichotomy that began to appear in logic was an incipient form of
what we now refer to as the distinction between syntactic and semantic concep-
tions of mathematical reasoning. In Section 2.2, we distinguished two basic con-
ceptions of mathematics that emerged in the nineteenth century: mathematics
as the science of symbolic representation and calculation, versus mathematics
as the science of abstract characterization and reasoning. These orientations
have a bearing on one’s view of logic. If one chooses to emphasize the former,
then logic can be construed as the science of symbolic representations of logi-
cal notions, and the means of calculating with them. This viewpoint underlies
Leibniz’ notion of a calculus ratiocinator, and is made explicit later in Alfred
North Whitehead’s Universal Algebra of 1898. There Whitehead writes:
The ideal of mathematics should be to erect a calculus to facilitate
reasoning in connection with every province of thought, or of ex-
ternal experience, in which the succession of thoughts, or of events,
can be definitely ascertained and precisely stated. So that all se-
rious thought which is not philosophy, or inductive reasoning, or
imaginative literature, shall be mathematics developed by means of
a calculus.28
In contrast, if one focuses on abstract characterization, one might look to
logic to provide a formal means of describing mathematical objects and struc-
tures. Axiomatic characterizations of the natural numbers, the real numbers,
and Euclidean space are important in this respect, even without a specification
of the corresponding deductive rules. This view of the nature of mathematical
logic helps explain two historical facts that may seem odd from a contemporary
point of view: first, that Dedekind considered his characterization of the natu-
ral numbers to be a logical one, even in the absence of a deductive framework;
and second, that Peano and his Italian school at the end of the nineteenth cen-
tury set up elaborate symbolic frameworks for mathematics, while not feeling
compelled to specify a corresponding formal system or rules of inference.
The divergence between syntactic and semantic viewpoints persisted in the
twentieth century, as Huntington, Veblen, and others in the school of American
“postulate theorists” presented axiomatic characterizations of various mathe-
matical structures with a clear interest in semantic categoricity, but with only
an informal notion of semantic consequence and no attempt to develop a de-
ductive counterpart. In contrast, in Russell and Whitehead’s Principia Math-
ematica one finds the development of an explicit, elaborate deductive system
27 See Peckhaus [88].
28 Ibid., page viii.

14
for mathematics, with little or no concern for a mathematical analysis of the
underlying semantics.29
Thus, by the turn of the century a number of different dimensions or axes
had emerged along which one might try to characterize work in mathematical
logic, depending on the general goals and methods embodied in the work, and
the perceived relationship to mathematics. Logicians today are used to distin-
guishing between semantic and syntactic notions, like truth and provability, or
definability in a model and definability in a theory; and both philosophically
and mathematically motivated branches of logic are mature, well-developed dis-
ciplines. But these distinctions were not as clear around the turn of the century,
and a good deal of work was involved in sorting them out. In what follows, we
will discuss Hilbert’s role in this effort, and we will argue that he was uniquely
positioned to bring the various trends and points of view together in a com-
pelling and fruitful way.

3 Hilbert’s proof theory

3.1 Hilbert as mathematician
When Hilbert addressed the Second International Congress of Mathematicians
in Paris, in 1900, he was 38 years old, and was recognized as one of the foremost
mathematicians of his day, rivaled in stature only by Poincaré. He had already
achieved a stunning array of successes in a wide range of mathematical fields,
and was a brilliant proponent of the new, abstract methods in mathematics.
In a series of papers between 1888 and 1893, Hilbert solved what was known
as the “fundamental problem of invariant theory,” and proved what are now
known as Hilbert’s basis theorem and the Nullstellensatz. His elegant but non-
constructive proof of the basis theorem is said to have elicited the comment “this
is not mathematics, but theology!” from Paul Gordon, a central figure in the
study of invariants.30 In 1893, he began the Zahlbericht [57], a report on number
theory that brought together and simplified work in the subject from Gauss to
Dedekind and Kronecker, and lay a firm foundation for algebraic number theory.
In 1944, Hermann Weyl called it a “jewel of mathematical literature”: “Even
today, after almost fifty years, a study of this book is indispensable for anybody
who wishes to master the theory of algebraic numbers.”31 His Foundations of
Geometry 32 provided axiomatic developments of Euclidean and non-Euclidean
geometry, again providing a unifying framework for developments in the subject
since Euclid. His work on integral equations and the foundations of physics, and
his solution to Waring’s problem, still lay ahead.33
29 Russell and Whitehead were aware of Hilbert’s and Veblen’s work before writing the

Principia, and both are mentioned in the preface to Russell’s Principles of Mathematics [97],
first edition, page viii. See here also Scanlan [99] and Awodey and Reck [5].
30 Quoted in Blumenthal [13], page 394.
31 Weyl [120], page 626.
32 Grundlagen der Geometrie, Hilbert [58].
33 Reid [95] is a very readable biography of Hilbert. For further biographical information,

15
 Hilbert’s work illustrates many of the themes that we have portrayed as
characteristic of the nineteenth century revolution in mathematics: the shift
of emphasis from symbolic representation to abstract characterization; the use
of infinitary, nonconstructive methods; and the search for foundational unity.
Though well aware of the revolutionary character of his work, Hilbert did not
take Kroneckerian objections lightly, and was sensitive to constructive and al-
gorithmic issues throughout his career. In 1931, recalling his formative years as
a mathematician in Königsberg, he wrote:
In those days we young mathematicians, Privatdozenten and stu-
dents, played the game of transforming transfinite proofs of mathe-
matical theorems into finite terms, in accordance with Kronecker’s
paradigm. Kronecker only made the mistake of declaring the trans-
finite mode of inference to be inadmissible.34
In this passage, Hilbert seems to play down the importance of having finitary
proofs, in contrast to Kronecker, who held such proofs to be central to math-
ematics. But Kronecker’s influence on the young Hilbert was probably greater
than the elder Hilbert wanted to admit, as suggested by Hermann Weyl, in his
obituary of Hilbert:
When one inquires into the dominant influences acting upon Hilbert
in his formative years one is puzzled by the peculiarly ambivalent
character of his relationship to Kronecker: dependent on him, he
rebels against him. Kronecker’s work is undoubtedly of paramount
importance for Hilbert in his algebraic period. But the old gentle-
man in Berlin, so it seemed to Hilbert, used his power and authority
to stretch mathematics upon the Procrustean bed of arbitrary philo-
sophical principles and to suppress such developments as did not con-
form: Kronecker insisted that existence theorems should be proved
by explicit construction, in terms of integers, while Hilbert was an
early champion of Georg Cantor’s general set-theoretic ideas.. . . A
late echo of this old feud is the polemic against Brouwer’s intuition-
ism with which the sexagenarian Hilbert opens his first article on
“Neubegründung der Mathematik” (1922): Hilbert’s slashing blows
are aimed at Kronecker’s ghost whom he sees rising from the grave.
But inescapable ambivalence even here — while he fights him, he
follows him: reasoning along strictly intuitionistic lines is found nec-
essary by him to safeguard non-intuitionistic mathematics.35

In the heat of the Grundlagenstreit of the 1920’s, Hilbert grew increasingly

critical of Kronecker, Brouwer, and others, whom he saw as restricting the
freedom to exercise the new methods that he had done so much to develop. His
see, for example, Weyl [120], Blumenthal [13], and the introductory notes in volume 2 of Ewald
[31]. For more on Hilbert’s influential 1900 address in Paris, see [50].
34 Hilbert [66], paragraph 9.
35 Weyl [120], page 613.

16
writings and public lectures began to take on a more forceful and polemical tone,
resulting in many of the rhetorical flourishes that are now associated with the
era: “No one shall be able to drive us from the paradise that Cantor created for
us.”36 “[N]o one, though he speak with the tongues of angels, will keep people
from negating arbitrary assertions, forming partial judgements, or using the
principle of excluded middle.”37 “Forbidding a mathematician to make use of
the principle of excluded middle is like forbidding an astronomer his telescope
or a boxer the use of his fists.”38 At the same time, as indicated by Weyl
above, the metamathematical program that he developed from the late 1910s
on can be viewed as a means of grounding mathematics in a way that meets
even Kronecker’s and Brouwer’s strictures. Weyl comments on this tendency
in Hilbert’s work further when he writes: “With regard to what he accepts as
evident in this ‘metamathematical’ reasoning, Hilbert is more papal than the
pope, more exacting than either Kronecker or Brouwer.”39
We have portrayed Hilbert as a revolutionary figure, a powerful proponent of
the new methods in mathematics, yet, at the same time, sensitive to conservative
mathematical norms. In the coming pages we will trace the development of
his metamathematical program, and argue that one can view it not only as
a response to the foundational crisis and the Grundlagenstreit, but also as a
thoughtful means of diffusing some of the tensions raised by the new methods
in mathematics; and, further, as a way of bringing together a number of the
aspects of logic discussed in Section 2.3. It is a matter of historical speculation as
to whether or not the proof-theoretic program would have become as important
to Hilbert as it did had it not been for the foundational challenges of Brouwer,
Poincaré, and Weyl. But we would like to emphasize that at least some of the
issues that Hilbert’s program was designed to address were salient long before
the foundational crisis.

3.2 Logic and the Hilbert school, from 1899 to 1922

We have already noted that even Hilbert’s early work was foundational in many
respects. His Zahlbericht aimed as much to buttress the foundations of alge-
braic number theory as to extend the technical boundaries of the field, and The
Foundations of Geometry aimed to clarify the most basic geometric notions. Of
the twenty-three questions he posed in 1900, three of the ten discussed explicitly
in the lecture concern topics that are clearly of a foundational nature: Cantor’s
continuum problem, the consistency of arithmetic, and a mathematical treat-
ment of the axioms of physics. In addition, the negative solution to the tenth
problem, on the solvability of Diophantine equations, required foundational de-
velopments in the theory of computability for its ultimate solution. The years
to come will, of course, see Hilbert address logical and foundational issues more
systematically.
36 Hilbert [64], page 376.
37 Hilbert [64], page 379.
38 Quoted by Weyl in [120], page 639.
39 Ibid., page 641.

17
 In the last section, we described divergent viewpoints as to the nature and
purpose of mathematical logic. We would now like to argue that the develop-
ment of Hilbert’s program, and associated developments in the Hilbert school,
helped bring the various viewpoints together, and clarified them significantly.
To support this claim, we will trace the development of logic in the Hilbert
school, vis-à-vis these viewpoints, in the years preceding the mature presenta-
tion of Hilbert’s Beweistheorie. A much fuller account of these events is provided
by Wilfried Sieg in [105] and [106].
In much of Hilbert’s early foundational work it is not yet clear which un-
derstanding of the nature of logic he has in mind. Consider The Foundations
of Geometry. Hilbert’s axiomatization can be viewed as a foundation for ge-
ometric reasoning, a perfection of Euclid’s work, which had long been hailed
as a model of mathematical rigor. But it differs from the kind of foundation
offered by Frege’s earlier work on logic and arithmetic, and from that offered
by Russell and Whitehead in the Principia of 1910–1913, in an important way:
in contrast to these other works, Foundations does not address the underlying
deductive framework explicitly. In particular, while the geometric axioms are
stated clearly and precisely in Foundations, there is no formal description of the
logical language, nor any characterization of the appropriate rules of inference.
At the same time, much of Foundations suggests the use of the axiomatic
method as a mathematical tool, a way of exploring geometric structures and
comparing them with one another. This emphasis is discernible in the introduc-
tion:
Consider three distinct sets of objects. Let the objects of the first
set be called points and be denoted by A, B, C, . . . ; let the objects
of the second set be called lines and be denoted by a, b, c, . . . ; let
the objects of the third set be called planes and be denoted by α, β,
γ, . . . .40
The structural emphasis becomes even more prominent later in the book when
several interpretations of various subsets of the axioms are studied in detail. In
his 1900 essay, “The concept of number,”41 Hilbert provides a similar axiomatic
treatment of the real numbers (strongly influenced by Dedekind [24]).
In both Foundations of Geometry and “The concept of number” there is
an ambiguity as to whether Hilbert views his axioms primarily as providing a
semantic characterization or a deductive foundation. In the case of the real
numbers, he indicates that the axioms provide a categorical description of the
structure in mind: “We therefore recognize the agreement of our number-system
with the usual system of real numbers.”42 But with respect to both Euclidean
geometry and the real numbers he also alludes to deductive aspects of the ax-
iomatic method; for example, with respect to geometry, he requires that “the
system of axioms is adequate to prove all geometric propositions.”43 Support-
40 Hilbert [58], second page of the introduction.
41 “Über den Zahlbegriff,” Hilbert [59].
42 Ibid., paragraph 14.
43 Ibid., paragraph 3.

18
ing the deductive point of view, he adds that new geometric statements are
only valid “if one can derive them from the axioms by means of a finite num-
ber of logical inferences.”44 He does not, however, elaborate on the notion of
“derivation” or “logical inference.”45
The ambiguities we have just discussed do not detract from the importance
of the work; both Foundations and “The concept of number” are masterful, and
constitute a landmark in foundations of mathematics. But it does make clear
that there were conceptual issues that were still up in the air at the time.
A decisive shift is found in Hilbert’s 1904 Heidelberg lecture, “On the foun-
dations of logic and arithmetic,”46 which addresses the problem of providing a
consistency proof for arithmetic. The paper contains the first hint of moving
from the search for a semantic consistency proof (giving a model) to a syntactic
one, and a discussion of a rudimentary deductive system. But the discussion of
the deductive system is sketchy and inadequate, and there is no clear distinc-
tion between the mathematical and metamathematical methods. Hilbert does
allude to the circularity of justifying the fundamental laws of arithmetic using
general mathematical principles, and presents his new approach as a means of
avoiding this problem. But the approach left the door open for a strong attack
by Poincaré, who emphasized the circularity of using mathematical induction
to justify induction itself.47
Hilbert would not develop the conceptual wherewithal to formulate an ap-
propriate response to Poincaré’s criticism for more than a decade. After 1904, he
turned his attention to other matters. But he never turned his back on founda-
tional issues completely, and, indeed, gave lectures on mathematical logic often
during this period.48 Around 1917 he returned to a study of foundations in
earnest, and that year witnessed two developments of paramount importance.
The first is a renewed, more detailed exploration of foundational aspects
of the axiomatic method. In “Axiomatic thought,”49 a lecture presented to
the Swiss Mathematical Society, Hilbert discussed the unifying methodologi-
cal role it could play in mathematics and the sciences, providing a “framework
of concepts” and a “deepening of the foundations” (Tieferlegung der Funda-
mente) in any field of sustained inquiry. The talk is not a technical one, but
Hilbert discusses specific applications of axiomatics in the theory of invariants,
the foundations of geometry, number theory, algebraic geometry, classical me-
chanics, quantum mechanics, electrodynamics, and so on. What is crucial to
our account is that, in this lecture, Hilbert explicitly distinguishes between
mathematical and foundational uses of the axiomatic method. As mathemati-
cal uses, he cites showing that the axioms of geometry are consistent because
the Euclidean plane can be interpreted in the arithmetic of the real numbers;
or showing that a physical theory of radiation is consistent by interpreting its
44 Ibid.,paragraph 16.
45 See Awodey and Reck [5] for further discussion.
46 “Über die Grundlagen der Logik und der Arithmetic,” Hilbert [60].
47 Poincaré [90].
48 Sieg [105], page 8.
49 “Axiomatisches Denken,” Hilbert [61].

19
basic concepts in terms of real analysis. He continues:
The problem of the consistency of the axiom system for the real
numbers can likewise be reduced by the use of set-theoretic concepts
to the same problem for the integers; this is the merit of the theories
of the irrational numbers developed by Weierstrass and Dedekind.
In only two cases is this method of reduction to another special
domain of knowledge clearly not available, namely, when it is a mat-
ter of the axioms of the integers themselves, and when it is a matter
of the foundation of set theory; for here there is no other discipline
besides logic which it would be possible to evoke.50
Hilbert’s observation is that there is an important difference between justifying
a mathematical theory using mathematical methods, and justifying the math-
ematical methods themselves. At this stage, he had not yet developed the
metamathematical stance that would underlie his proof-theoretic program; in-
deed, his ensuing discussion mentions Russell’s work and suggests that he has
something more like the logicist program in mind. But the important point is
that he had begun to grapple with the issue.
The second development of key importance is that Hilbert began to draw a
clear line between syntactic and semantic notions in logic. These developments
are insightfully chronicled by Sieg [105, 106], which includes a discussion of
Hilbert’s lecture notes from the winter semester of 1917–1918, prepared with
the assistance of Paul Bernays. In these lectures, Hilbert notes that formal
languages and theories can be interpreted in a mathematical structure, and,
indeed, are usually designed with a particular interpretation in mind; yet he also
emphasizes the fact that the corresponding deductive systems can be studied
on their own, independent of any semantic interpretation. With respect to the
syntactic calculus, Hilbert stresses that the rules of inference should be sufficient
to capture all ordinary forms of mathematical argumentation, thus providing a
formal model of deductive reasoning in mathematics.
Having separated deductive syntax from its semantics, Hilbert and Bernays
were able to explore the relationship between the two. As Sieg points out,
the completeness of propositional logic is proved in a footnote to the lecture
notes, and a modern formulation (i.e. the assertion that every valid formula
is provable, and conversely) appears in Bernays’ Habilitationsschrift of 1918.51
Furthermore, the question as to whether or not first-order logic is complete is
explicitly presented in Hilbert and Ackermann’s logic text of 1928,52 which is
in large part based on the 1917/1918 lecture notes.
Our account so far has emphasized two developments in Hilbert’s thought
on logic: a new emphasis on syntactic, deductive notions of logical consequence,
and a shift to a foundational viewpoint regarding the role of logic with respect
to mathematics. This is not to say that Hilbert abandoned the semantic and
50 Hilbert [61], paragraphs 37–39.
51 See also the discussion in Zach [123].
52 Hilbert and Ackermann [68].

20
mathematical viewpoints; rather, with a more mature understanding of the
conceptual issues at hand, by the 1920’s Hilbert was able make the necessary
distinctions, recognize the uses and merits of the various stances, and study
the relationships between them. These developments mark a turning point in
logic. Without recognizing a sharp distinction between syntax and semantics,
the problem of demonstrating the completeness of first-order deductive systems
could not even be posed, much less solved by Gödel in 1929. And without rec-
ognizing the distinction between mathematical and foundational roles of logic,
Hilbert’s metamathematical program of the 1920’s, which involved using math-
ematical methods to prove the consistency of mathematical reasoning, would
have been methodologically incoherent.

3.3 The emergence of “Beweistheorie”

In [105] and [106], Sieg goes on to provide an illuminating account of the step-
by-step transition of the Hilbert school from the viewpoint of “Axiomatisches
Denken” to the full development of the proof-theoretic program. First, there
was an attempt to justify classical mathematics on Russell’s logicist grounds,
ending with a rejection of the problematic axiom of reducibility; then a brief
flirtation with constructivism, quickly dismissed in light of the centrality of the
law of the excluded middle to mathematics. By 1922, Hilbert had arrived at the
program of justifying classical mathematics by means of a consistency proof,
presented in “The new grounding of mathematics. First report.”53 Finally, a
more mature statement of the program, with a more refined treatment of the
metamathematical stance, appeared in “The logical foundations of mathemat-
ics” of 1923.54
The essence of Hilbert’s metamathematical program is well known, involving
the formal study of mathematical methods, represented by formal systems of
deduction, using mathematical methods. Kleene summarizes Hilbert’s contri-
bution in that respect:
To Hilbert is due now, first, the emphasis that strict formalization of
a theory involves the total abstraction from the meaning, the result
being called a formal system or formalism (or sometimes called a
formal theory or formal mathematics); and second, his method of
making the formal system as a whole the object of a mathematical
study called metamathematics or proof theory.55
But Hilbert was also careful to distinguish the portion of mathematics treated
as an object of study, via formal representation in a deductive system, from
the mathematics used to study it. This distinction is crucial to understanding
Hilbert’s program, and it is one we will return to in Section 4.
Here we would like to point out that aside from the distinction between math-
ematics and metamathematics, the use of mathematical techniques to study
53 “Neubegründung der Mathematik. Erste Mitteilung,” Hilbert [62].
54 “Die logischen Grundlagen der Mathematik,” Hilbert [63].
55 Kleene [71], pages 61–62.

21
mathematical methods of reasoning was in and of itself somewhat radical. To
illustrate this point, consider Paolo Mancosu’s discussion of the work of Hein-
rich Behmann, a student of Hilbert’s. Mancosu writes of a talk on Russell
and Whitehead’s Principia Mathematica, given by Behmann in the Göttingen
colloquium in 1914:

Behmann began the lecture by stressing the fact that the term
“Mathematische Logik” is ambiguous and can be used to charac-
terize two different traditions. The first tradition consists in a gen-
eral construction of logic by mathematical means (“Mathematik der
Logik”) and is associated with the names of Boole, Schröder, and
partly Peano. The second tradition (“Logik der Mathematik”) an-
alyzes the role played by logic in the construction of mathematics.
Behmann mentions Bolzano, Frege, and Russell as representative of
this direction of work. Although PM belongs squarely in the second
tradition, it also accounts for results developed in the first tradition.
Thus, for Behmann, PM is the first unified account of these two
traditions.56
It is likely that this perceived synthesis contributed to the interest in Russell’s
work that we find in Hilbert during this time.57 One should note, however, that
Russell’s synthesis is at best only partial, and, in particular, does not consider
deductive systems as syntactic mathematical objects in their own right. Rus-
sell, like Frege, never considered a formal deductive system independent of its
intended mathematical interpretation, and neither Russell nor Frege recognized
the importance of exploring the possibility of multiple interpretations.
Today it is hard to appreciate the novelty of Hilbert’s approach. Modern set
theorists are fully comfortable with the notion that one may view the Zermelo-
Fraenkel axioms as a foundation for mathematics, and yet study various models
of these axioms using ordinary mathematical methods. But in 1922 this notion
would have seemed at least a little bit odd, a strange partnership between two
quite distinct research programs: the first, of providing an axiomatic foundation
for mathematics, in the tradition of Frege, Russell-Whitehead, and Zermelo;
the second, studying logic as a branch of algebra or as a mathematical tool,
in the tradition of Boole and Peirce. It required someone of Hilbert’s breadth
as a mathematician to realize that combining these two traditions would be
fruitful; and it required someone of Hilbert’s abilities to carry out the synthesis
successfully. In the next section, we will discuss this synthesis in more detail.
56 Mancosu [84], page 306.
57 See also the discussion in Mancosu [82].

22
4 Formalism and Hilbert’s program
4.1 Strong formalism
Hilbert’s program of justifying abstract mathematics via finitary consistency
proofs relies on the implicit assumption that the relevant portions of mathematic
are adequately modeled by the formal deductive systems. One might read into
this stance the view that classical mathematics is nothing more than a system of
axioms and rules, in the sense that it is devoid of any content or characteristics
that cannot be explained purely in terms of a formal system representing it.
Such extreme views were criticized well before the development of Hilbert’s
Beweistheorie. For example, Brouwer writes in his 1912 paper,“Intuitionism
and Formalism”:
The viewpoint of the formalist must lead to the conviction that if
other symbolic formulas should be substituted for the ones that now
represent the fundamental mathematical relations and the mathematical-
logic laws, the absence of the sensation of delight, called “conscious-
ness of legitimacy,” which might be the result of such substitution
would not in the least invalidate its mathematical exactness. To the
philosopher or anthropologist, but not to the mathematician, be-
long the task of investigating why certain systems of symbolic logic
rather than others may be effectively projected upon nature. Not
to the mathematician, but to the psychologist, belongs the task of
explaining why we believe in certain systems of symbolic logic and
not in others, in particular why we are averse to the so-called con-
tradictory systems in which the negative as well as the positive of
certain propositions are valid.58
In light of Hilbert’s writings from the 1920’s, the formalism critiqued by Brouwer
may appear to be a reasonable characterization of Hilbert’s later views. For ex-
ample, in “On the infinite,”59 Hilbert seems to be quite explicit in denying much
of abstract mathematics any “content” or “meaning.” First, he distinguishes
between formulas which communicate “contentual” finitary propositions, and
more general formulas, that refer to the “ideal objects of the theory.”
But since ideal propositions, namely, the formulas insofar as they do
not express finitary assertions, do not mean anything in themselves,
the logical operations cannot be applied to them in a contentual
way, as they are to the finitary propositions. Hence it is necessary
to formalize the logical operations and also the mathematical proofs
themselves.. . . 60
He soon adds:
58 Brouwer [16], page 736.
59 “Über das Unendliche,” Hilbert [64].
60 Hilbert [64], page 381.

23
 [W]e will be consistent in our course if we now divest the logical
signs, too, of all meaning, just as we did the mathematical ones,
and declare that the formulas of the logical calculus do not mean
anything in themselves either, but are ideal propositions. . . 61
It was rhetoric like this that led Russell to quip, years later:
The formalists are like a watchmaker who is so absorbed in making
his watches look pretty that he has forgotten their purpose of telling
the time, and has therefore omitted to insert any works.62
When it comes to questions of meaning in mathematics, however, there are
a number of factors that weigh against taking Hilbert’s rhetoric at face value.
The first, and most obvious, is that it belies Hilbert’s passionate commitment
to the subject. The formalist that Brouwer caricatures holds that mathematics
is nothing more than a symbol game, and the task of justifying a particular
choice of game is outside the mathematician’s scope. But it is clear that, to
Hilbert, mathematics is much more than a symbol game, with methods and
rules of inference that are by no means arbitrary; the issue of whether or not
to allow the use of the law of the excluded middle, for example, is of central
mathematical importance. He writes, angrily, in response to the foundational
challenges:

What Weyl and Brouwer do amounts in principle to following the

erstwhile path of Kronecker: they seek to ground mathematics by
throwing overboard all phenomena that make them uneasy and by
establishing a dictatorship of prohibitions à la Kronecker. But this
means to dismember and mutilate our science, and if we follow such
reformers, we run the danger of losing a large number of our most
valuable treasures.63
Hilbert is critical of any attempts to restrict mathematical practice on purely
philosophical or ideological grounds, and clearly feels that the task of determin-
ing the proper practice of mathematics lies within the mathematical community.
Moreover, Hilbert has too much respect for intuitive thinking, thoughtful
reflection, and the role of the mind in mathematical and scientific activity to
see mathematics as a mere game with symbols. Thus he observes:
In our science it is always and only the reflecting mind [der überlegende
Geist], not the applied force of the formula, that is the condition of
a successful result.64

Elsewhere, Hilbert makes it clear that such a view is not incompatible with a
formal viewpoint, since to proceed axiomatically means “nothing else than to
61 Ibid.
62 From the preface to the second edition of Russell [97], page vi.
63 Hilbert [62], paragraph 10.
64 Hilbert in a letter to Minkowski, quoted in Stein [110], page 242.

24
think with consciousness.”65 In 1927 he is even more forceful in indicating that
the type of formalism he has in mind does not deny the importance of intuitive
thought:
The formula game that Brouwer so deprecates has, besides its math-
ematical value, an important general philosophical significance. For
this formula is carried out according to certain definite rules, in
which the technique of our thinking is expressed. These rules form
a closed system that can be discovered and definitively stated. The
fundamental idea of my proof theory is none other than to describe
the activity of our understanding, to make a protocol of the rules
according to which our thinking actually proceeds.66
In other words, the formal systems under consideration are not empty symbol
games; we are interested in them precisely because they reflect our intuitive or
logical understanding.
How, then, can we account for Hilbert’s more extreme proclamations, most
salient in the articles “The new grounding,” “Logical foundations,” and “On
the infinite”? It is helpful, in this regard, to keep the historical context in mind.
We have already noted that Hilbert’s Beweistheorie, put forth for the first time
in “The new grounding,” was conceptually quite novel. In particular, it relied,
first, on the observation that one can separate syntax and semantics, and hence
study uninterpreted deductive systems from a purely syntactic point of view; and
second, on the observation that one can distinguish between the mathematical
methods used in the metatheory and the mathematical methods modeled in the
formal system under investigation. The force of these two observations is that
one may use very weak, restricted, syntactic methods to study formal systems
that embody much more powerful and abstract forms of reasoning. In addressing
a community of mathematicians not used to making such distinctions, this point
is a hard one to make; it is not surprising that Hilbert relies on rhetorical
flourishes to drive the point home.
But we also have to place the three articles mentioned above in the context
of the Grundlagenstreit, which raged in full force in the early 1920’s.67 It is
worth noting that, for example, “Axiomatic thought,” written in 1917, makes
no mention of the foundational challenges by Brouwer and Poincaré, nor any
mention of the paradoxes. But soon after, Brouwer’s call to refound contempo-
rary mathematics without use of the law of the excluded middle became more
pointed, and in 1920 Weyl, Hilbert’s best student, announced, “I now abandon
my own attempt and join Brouwer.”68 By 1927, in “Intuitionistic reflections
on formalism,”69 Brouwer is openly hostile and sarcastic towards the Hilbert
program. The scale of the confrontation is further illustrated by Hilbert’s per-
sonal battles with Brouwer, culminating, in 1928, with the removal of Brouwer
65 Hilbert [62], paragraph 14.
66 Hilbert [65], page 475. This passage is also quoted in Sieg [102], page 171.
67 The collection of writings in Mancosu [83] provides a sense of the intensity of the debate.
68 Weyl [118], page 56 in the original, page 98 in translation; also quoted in [116], page 480.
69 Brouwer [17].

25
from his editorship of the Mathematische Annalen. Amidst the polemical ex-
changes, it is unsurprising, though regrettable, to find substantial developments
overshadowed by extreme rhetoric. It is interesting to observe that Hilbert’s
tone grows somewhat softer in his writings of the late 20’s and early 30’s, and
his rhetoric becomes less belligerent.

4.2 A more tenable formalist viewpoint

As a philosophy of mathematics, formalism is usually understood to imply an
ontological stance in which abstract mathematical objects are denied any real
existence, as well as an epistemological stance in which mathematical knowledge
and justification is to be explained solely in terms of formal derivability. But
we have seen that Hilbert resisted such a strong identification of mathematical
practice with formal deduction. In this section we will argue that the philosoph-
ical assumptions needed to justify Hilbert’s metamathematical program are far
weaker, and are largely independent of strong metaphysical claims.
To start with, Hilbert’s strategy of justifying abstract mathematics using
consistency proofs only requires the acceptance that informal mathematical
proof in various mathematical domains is adequately modeled by formal de-
ductive systems, in the sense that an informal proof of a theorem indicates the
existence of a formal derivation of the theorem’s formal symbolic representation.
With this assumption, proving the formal system consistent using uncontrover-
sial, finitary methods shows that the informal practice in non-contradictory. In
other words, a finitary consistency proof provides a precise sense in which one
has justified the abstract methods using more concrete ones. Furthermore, as
Hilbert pointed out, the consistency of a formal axiomatic system for mathe-
matics implies that any universal arithmetic statement that is derivable in the
system is, in fact, true. Thus, a finitary consistency proof serves to transform
any proof of a universal number-theoretic statement in the formal system in
question into a finitary proof of the same statement. Once again, this serves to
justify the abstract methods with respect to the more concrete ones; questions
as to what mathematical objects “really” are need not enter the debate.
What is central, then, to Hilbert’s metamathematical program is the clear
separation of the mathematical framework being investigated from the math-
ematical framework used to investigate it. To bring out the philosophical sig-
nificance of such a separation, it is helpful to compare Hilbert’s position with
that developed by Rudolf Carnap in the late 1920’s and early 1930s, in con-
nection with scientific procedure more generally.70 Carnap’s “syntactic” view
of the sciences (which was, at least partly, influenced by Hilbert) can serve to
illuminate further this central aspect of the view of mathematics provided by
Hilbert’s formalism.
From Carnap’s point of view it is meaningful to speak of various “frame-
works” in which science, including mathematics, is carried out. Carnap is care-
ful to distinguish between questions that are “internal” to a given framework
70 See Carnap [21] and, with respect to the distinction between “internal” and “external”

questions, especially Carnap [22].

26
and questions that are “external” to it. Such a distinction is implicit in Hilbert’s
approach. On the “internal” level of mathematics, we are free to refer to ab-
stract mathematical objects and structures, including infinitary ones. On the
metamathematical level, however, we view mathematics “externally” — and
from that point of view, mathematics is characterized by the fact that we can
represent the language and rules of inference in a concrete, symbolic way. Inter-
nal to the framework, one is free to carry out mathematics unfettered by worries
about meaning. Externally, one has a vantage point from which to evaluate the
framework, in terms of overall usefulness, applicability, and, indeed, consistency.
The important thing is to recognize that the two viewpoints are quite different
in character, and can be kept separate.
How does Hilbert use this separation? For one thing, he notes that the
formalist viewpoint provides a useful perspective from which to try to evaluate
classical mathematics, from an external point of view. In particular, it allows
one to pose the question of consistency in a clear, precise, and rigorous fashion.
But we also see that, for Hilbert, this is not the only criterion that one can appeal
to in justifying abstract mathematics externally; Hilbert is equally forceful in
appealing to the beauty and elegance of the subject, and its fruitfulness and
applicability to the sciences.
More pointedly, Hilbert’s goal is to put down intuitionistic challenges on two
counts. First, he appeals to pragmatic and aesthetic considerations, which he
takes to weigh strongly in favor of a classical framework over an intuitionistic
one. Second, he aims to justify the classical framework using intuitionistically
acceptable methods. The effectiveness of this twofold line of attack is evidenced
by Weyl’s comments on Hilbert’s 1927 lecture, spoken only seven years after
Weyl had taken up the banner of intuitionism:
[I] am very glad to confirm, there is nothing that separates me from
Hilbert in the epistemological appraisal of the new situation thus
created. He asserted, first of all, that the passage through ideal
propositions is a legitimate formal device when real propositions are
proved; this even the strictest intuitionist must acknowledge.71
We can extract from these considerations a more reasonable formalist stance,
which we will call “weak” formalism, to distinguish it from the “strong” for-
malism lampooned by Russell and Brouwer. This stance rests simply on the
acceptance of the fact that from an external point of view, a characteristic fea-
ture of mathematical practice is that it involves a level of rigor and precision
that makes mathematical proofs amenable to formalization. Indeed, a large
part of what makes a proof “mathematical,” from this point of view, is that the
assumptions can be stated clearly, the language can be modeled symbolically,
and the rules of inference can be described explicitly. Furthermore, the weak
formalist holds that this, together with the consistency of the system under con-
sideration, is enough to justify the proof as “mathematics.” In contrast to the
strong formalist, the weak formalist does not maintain that mathematics has no
71 Weyl [119], page 483.

27
further “meaning.” But he or she does maintain that questions about meaning
in mathematics are of an entirely different character from mathematical ones;
they are external to the framework, not internal to it.
Hilbert’s further observation is that even some external questions can be
posed in a manner that is clear and precise enough to make them amenable to
mathematical investigation; the question of the consistency of the framework
is one such. Other external questions are less amenable to mathematization,
such as those involving aesthetic and pragmatic issues. Hilbert does bring these
issues into play as well, but he makes no efforts to treat them mathematically.
Hilbert’s point of view can therefore be summarized as follows: one should keep
internal mathematical questions and external questions as to the propriety of
the mathematical framework separate; when external questions can be turned
into mathematical ones, do so; but for the results to be philosophically informa-
tive, one has to distinguish between the mathematical methods that are used
externally from the ones that are being studied.
A significant by-product of the separation between internal and external
questions is the relegation of issues of meaning to the external evaluation of the
framework. Seen as such, Hilbert’s formalism is at its core a way of insisting
that the ordinary practice of mathematics is not dependent either on empirical
facts and knowledge or on particular metaphysical assumptions. With respect
to the independence from empirical questions, compare the following remark by
Hans Freudenthal about Hilbert’s axiomatic treatment of geometry:
“Consider three distinct sets of objects. . . ” – thus the umbilical
cord between reality and geometry has been cut. Geometry has
become pure mathematics, and the question whether and how it can
be applied to reality is answered just as for any other branch of
mathematics.72
With respect to the independence from metaphysics, we can, once more, turn
to Weyl’s 1927 conciliatory lecture. In this lecture he laments the failure of
the intuitionist program to provide an immediate phenomenological meaning to
mathematics, but seems to accept that the overall evaluation of the appropriate-
ness of a particular mathematical framework must appeal to external standards:

What “truth” or objectivity can be ascribed to this theoretic con-

struction of the world, which presses far beyond the given, is a pro-
found philosophical question. It is closely connected with the further
question: what impels us to take as a basis precisely the particular
axiom system developed by Hilbert? Consistency is indeed a neces-
sary but not a sufficient condition for this. For the time being we
probably cannot answer this question except by asserting our belief
in the reasonableness of history, which brought these structures forth
in a living process of intellectual development.. . . 73
72 Freudenthal [42], page 111, translated by Erich H. Reck.
73 Weyl [119], page 484.

28
The independence of mathematical practice from answering such questions is
expressed very nicely by Howard Stein in the following remark:
[Hilbert’s] point is, I think, this rather: that the mathematical logos
has no responsibility to any imposed standard of meaning: not to a
Kantian or Brouwerian “intuition,” not to finite or effective decid-
ability, not to anyone’s metaphysical standards for “ontology”; its
sole “formal” or “legal” responsibility is to be consistent (of course,
it has also what one might call a “moral” or “aesthetic” responsibil-
ity: to be useful, or interesting, or beautiful; but to this it cannot
be constrained — poetry is not produced through censorship).74

We have characterized our “weak formalism” as a fairly minimal stance.

One occasionally finds objections to a stronger formalist claim that mathemati-
cal proof can be identified with proof in a particular formal system, such as, say,
Zermelo-Fraenkel set theory. Indeed, Gödel’s incompleteness theorems seem to
weigh strongly against such a stance. But the weak formalist only maintains
that particular mathematical theories can be represented in formal frameworks,
not that all such theories can be represented in a single overarching framework.
One might also object to the claim that the formalizability of mathematical
proof is the defining characteristic of mathematical practice, or even that it is
the most important one. However, our weak formalist stance only assumes that
this is one important characteristic, and that, therefore, the modeling of math-
ematics with formal deductive systems is informative from an external point
view. Evaluations may vary as to the weight one places on formal characteriza-
tions, as mathematics certainly exhibits other important characteristics as well.
But it seems that any satisfactory account of mathematical practice will have
to take this formal nature into consideration, and so is likely to be compatible
with a weak formalist point of view.

4.3 Clarifying the nature of the infinite

In the introduction to this essay, we claimed that it is impossible to fully ap-
preciate Hilbert’s program without looking beyond the narrow concerns of the
Grundlagenstreit. And, in Section 2, we discussed tensions that emerged from
the nineteenth century transformation of mathematics, stemming from divergent
views of the subject as a science of abstract characterization and reasoning, on
the one hand, and as a science of symbolic representation and calculation, on
the other. Let us now see what a weak formalist viewpoint has to offer towards
resolving this tension.
In Section 2.2, we quoted Kronecker’s insistence that mathematical objects
be defined in terms of the integers, for which “one needs above all a suitable,
surveyable manner of expression and representation.” Cantor summarizes such
a viewpoint, as follows:
74 Stein [110], page 255, emphasis in the original.

29
 But as for the irrationals, they ought in pure mathematics to receive
a purely formal meaning, in that they as it were only serve as in-
struments to fix properties of groups of integers and to ascribe these
properties in a simple and uniform manner. According to this opin-
ion, the true material of analysis is exclusively formed from the finite
integers, and all the truths which have been found in arithmetic and
analysis, or whose discovery is hoped for, ought to be conceived as
relations of the finite integers to each other; infinitesimal analysis
and with it the theory of functions are held to be legitimate only to
the extent that their propositions can be interpreted as provable by
laws governing finite integers.75
While Cantor disagrees with this viewpoint, he is not entirely dismissive of it.
Later in the same work he returns to it in connection with the problem of human
understanding of the infinite:
The finiteness of human understanding is often adduced as the rea-
son why only finite numbers can be thought;. . . But if it turns out
that the understanding can also in a determinate sense define and
distinguish infinite, that is, superfinite numbers, then either the
words “finite understanding” must be given an extended meaning
(from which that conclusion can then no longer be drawn); or the
predicate “infinite” must be in certain respects conceded to human
understanding. The latter is in my opinion the only correct course.76
Cantor repeatedly suggests that a key factor in our understanding of the infinite
is that infinitary objects stand in definite relationships to one another, “bound
to each other and to the finite numbers by fixed laws.”77
Apparently, we are faced with an epistemological dilemma: Cantor holds
that, under suitable circumstances, we can have knowledge of infinitary math-
ematical objects, while Kronecker maintains that mathematical knowledge is
restricted to objects that have finite representations. We find Hilbert sensi-
tive to both sides of the debate. On the one hand, he is clearly attentive to
Kroneckerian demands for “surveyability” and “representability.” In “The New
Grounding” he writes:

[A]s a precondition for the application of logical inferences and for

the activation of logical operations, something must already be given
in representation: certain extra-logical discrete objects, which exist
intuitively as immediate experience before all thought. If logical
inference is to be certain, then these objects must be capable of being
completely surveyed in all their parts, and their presentation, their
difference, their succession (like the objects themselves) must exist
for us immediately, intuitively, as something that cannot be reduced
75 Cantor [20], paragraph 6 of section 4.
76 Ibid., paragraph 4 of section 5.
77 Ibid., paragraph 1 of section 6.

30
 to something else. Because I take this standpoint, the objects of
number theory are for me — in direct contrast to Dedekind and
Frege — the signs themselves.. . . 78
Hilbert returns to this requirement of “surveyability” and finite representability
often during the 1920’s. On the other hand, he recognizes that the limitation to
finitary objects is insufficient for a full development of abstract mathematics.
[W]e cannot conceive of the whole of mathematics in such a way.
Already when we cross over into the higher arithmetic and algebra
– for example, if we wish to make assertions about infinitely many
numbers or functions – the contentual procedure breaks down.79
Hilbert’s concern with “higher arithmetic and algebra” is, of course, part of his
continued defense of abstract mathematics, in the tradition of Dedekind and
Cantor.
For Hilbert, the formal, metamathematical stance provides a means of ob-
taining the best of both worlds.
[W]e can achieve an analogous point of view if we move to a higher
level of contemplation, from which the axioms, formulae, and proofs
of the mathematical theory are themselves the objects of a con-
tentual investigation. But for this purpose the usual contentual ideas
of the mathematical theory must be replaced by formulae and rules,
and imitated by formalisms.. . . In this way the contentual thoughts
(which of course we can never do wholly without or eliminate) are
removed elsewhere — to a higher plane, as it were; and at the same
time it becomes possible to draw a sharp and systematic distinction
in mathematics between the formulae and the formal proofs on the
one hand, and the contentual ideas on the other.80
Note that in this passage, Hilbert does not go so far as to say that our thoughts
of the infinite are not “contentual.” Indeed, he indicates that the content of
such thoughts cannot be removed, only moved “to a higher plane.” Meanwhile,
the formal axioms and rules lend concrete representation to the abstract ideas.
For Hilbert, these formal representations provide the key to resolving the
epistemological dilemma. They do so by supplying the “surveyable manner of
expression” demanded by Kronecker and the “fixed laws” required by Cantor,
while, at the same time, supporting the use of abstract infinitary objects in
modern mathematics. Thus, within the formal system, one is free to pursue
Cantorian set theory; from a metamathematical point of view, formal proofs
provide the theory with a concrete content. As a result, mathematicians can
contemplate infinitary notions, and still meet Kronecker’s epistemic demand for
finite representability: we can know a theorem of abstract mathematics because
we can grasp the finitely presented proof. This resolution is not exactly what
78 Hilbert [62], paragraph 25.
79 Ibid., paragraph 32.
80 Ibid., paragraph 33.

31
Kronecker had in mind, since it is the elements of the formal system rather than
the mathematical notions themselves that are presented concretely. But such a
shift of focus from the objects of the theory to the formal framework is entirely
consistent with the metamathematical stance.
It is important to note that this use of formalism does not preclude the
possibility that there are other epistemological issues that should be addressed.
Our weak formalist may, for example, inquire as to what justifies the partic-
ular choice of a mathematical framework, or its use in scientific inquiry. And
nothing that Hilbert says denies the validity of such pursuits; indeed, it is here
that he appeals to external, pragmatic factors, including fruitfulness and appli-
cability to the sciences. Thus, when we view Hilbert’s proof-theoretic program
as a response to the particular issues raised by the Cantor-Kronecker debate
rather than a comprehensive philosophy of mathematics, the metamathemati-
cal stance is seen to resolve a central epistemological problem without placing
undue restrictions on the range of broader philosophical options.

5 Proof theory after Hilbert

5.1 Proof theory in the 1930’s and 1940’s
What became of proof theory, after Hilbert? Of course, Gödel’s incompleteness
theorems dealt a serious blow to Hilbert’s plans of using finitary consistency
proofs to rid the world, “once and for all, of the question of the foundations
of mathematics as such.”81 By 1930, Heyting had presented a formalization
of intuitionistic number theory, and it was not long before Gödel and Gentzen
had observed independently that a simple translation can be used to interpret
classical first-order arithmetic in its intuitionistic version. This made it clear
that intuitionistic reasoning is not identical with finitary reasoning, since it was
commonly accepted that classical first-order arithmetic went beyond finitary
means. Reflecting on these developments years later, Bernays notes that these
results came as a surprise, since they were “against the prevailing views at that
time.”82
Describing the ensuing development of proof theory, Bernays writes:
An enlarging of the methods of proof theory was therefore suggested:
instead of a restriction to finitist methods of reasoning, it was re-
quired only that the arguments be of a constructive character, al-
lowing us to deal with more general forms of inference.83
In other words, proof theorists began to pursue a “modified Hilbert’s program,”
which involved justifying classical methods on constructive grounds, or relative
to constructive methods. Examples of this include Gerhard Gentzen’s two con-
sistency proofs for arithmetic using transfinite induction up to ε0 , published in
81 Hilbert [66], paragraph 39.
82 Bernays [10], page 502. See also von Neumann [117], especially pages 61–62, where the
“intuitionistic” and “finitistic” are used interchangeably.
83 Ibid.

32
1936 and 1938, respectively;84 and Gödel’s Dialectica interpretation of arith-
metic in a quantifier-free theory of higher-type functionals, first presented in
1941, though not published until 1958.85
As the raging foundational battles began to die down, narrow questions of
consistency alone began to seem less pressing, and proof theorists began to
justify their work in terms of broader epistemological goals. As early as 1928,
Hilbert wrote:
[E]ven if one were not satisfied with consistency and had further
scruples, he would at least have to acknowledge the significance of
the consistency proof as a general method of obtaining finitary proofs
from proofs of general theorems — say of the character of Fermat’s
theorem — that are carried out by means of the ε-function.86
Here Hilbert is referring to his epsilon-substitution method, the technical means
by which he hoped to obtain a consistency proof. In 1930, in “The philosophy
of mathematics and Hilbert’s proof theory,” Bernays writes:
[T]he current discussion about the foundations of mathematics does
not have its origins in a predicament of mathematics itself. Math-
ematics is in a completely satisfactory state of methodological cer-
tainty.. . .
The problematic, the difficulties, and the differences of opinion
begin rather at the point where one inquires not simply about the
mathematical facts, but rather about the grounds of knowledge and
the delimitation of mathematics. These questions of a philosophical
nature have received a certain urgency since the transformation the
methodological approach to mathematics experienced at the end of
the nineteenth century.
The characteristic moments of this transformation are: the ad-
vance of the concept of set, which aided the rigorous grounding of the
infinitesimal calculus, and further the rise of existential axiomatics,
that is, the method of development of a mathematical discipline as
the theory of a system of things with determinate operations whose
properties constitute the content of the axioms. In addition to this
we have, as the result of the two aforementioned moments, the es-
tablishment of a closer connection between mathematics and logic.
This development confronted the philosophy of mathematics with
a completely new situation and entirely new insights and prob-
lems.. . . The debate concerning the difficulties caused by the role of
the infinite in mathematics stands in the foreground in the present
stage of this discussion.87
84 Gentzen [45] and [46].
85 Gödel [49].
86 Hilbert [65], page 474, also quoted in Delzell [26], page 113.
87 “Die Philosophie der Mathematik und die Hilbertsche Beweistheorie,” Bernays [9], pages

234–235.

33
This passage is striking: as early as 1930, Bernays declares that mathematics
is in a “completely satisfactory state of methodological certainty”! According
to Bernays, the goal is not to justify mathematics; rather, it is to come to
philosophical terms with the nineteenth century transformation of mathematics,
particularly with respect to modern methods of abstract characterization and
reasoning, and with respect to modern methods of dealing with the infinite.
Beyond proving consistency, one might look to proof theory for a means of
understanding abstract, infinitary methods in concrete, constructive, or even
finitary terms. In addition to Hilbert and Bernays, Gentzen hints at this more
general program at the end of his first consistency proof,88 and expands on
it in “The concept of infinity in mathematics.”89 Like Weyl, he distinguishes
between “actualist” and “constructive” interpretations of the infinite, the former
providing the abstract, but epistemologically problematic, basis for set theory.
He notes that the set theoretic paradoxes point to the need to delimit the
allowable modes of inference in abstract mathematics, and continues:
I must say that to me the clearest and most consequential delim-
itation seems to be that given by the principle of interpreting the
infinite constructively.
We should nevertheless be reluctant to discard the extensive non-
constructive part of analysis which has, among other things, cer-
tainly stood the test in a variety of applications in physics. Hilbert
sees his proof theory as a means of resolving these difficulties. The
theory is intended to clarify as far as possible the mutual relation-
ship between the two interpretations of the infinite by means of a
purely mathematical investigation.90
Here again we see a shift of focus from establishing consistency to “ascribing
a constructive sense” to a priori non-constructive methods, and clarifying the
relationship between the two interpretations of the infinite.

5.2 Proof theory in the 1950’s

The trend of expanding beyond consistency issues in proof theory continued
through the 1950’s. Comparing developments in this decade to those of the
1920’s and 1930’s, Charles Delzell wrote in 1996:
Back then, the threat of an inconsistency (say, in arithmetic) seemed
plausible enough to justify a defensive build-up (though critics could
call it “Cold War hysteria”). By 1950, however, the trepidation in
foundations had largely died down, and the time was ripe to look for
possible peaceful, commercial uses of the formidable -theorems.91
88 Gentzen [45].
89 “Der Unendlichkeitsbegriff in der Mathematik,” Gentzen [44].
90 Ibid., page 227.
91 Delzell [26], page 114.

34
By 1958, Kreisel was openly critical of Hilbert’s initial focus on consistency. To
take its place, he proposed
. . . a different general program which does not seem to suffer the
defects of the consistency program: To determine the construc-
tive (recursive) content or the constructive equivalent of the non-
constructive concepts and theorems used in mathematics, particu-
larly arithmetic and analysis.92
By finding constructive interpretations of classical methods, according to Kreisel,
“one not only understands ‘ordinary’ mathematics better, but obtains new the-
orems.”93 In later years, he states this shift of emphasis more forcefully. For
example, he writes, with Takeuti, in 1974:
It would be simply absurd to judge proof-theoretic work mainly,
let alone solely, by its relevance to Hilbert’s original programme,
because this programme is itself of dubious relevance.94
And, looking back on his career in 1988:
Like many others. . . I was repelled by Hilbert’s exaggerated claims
for consistency as a sufficient condition for mathematical validity
or some kind of existence. But unlike most others I was not only
attracted by the logical wit of consistency proofs. . . but also by the
so to speak philosophical question of making explicit the additional
knowledge provided by those proofs (over and above consistency
itself).95
General assessments of the success of Kreisel’s various proof-theoretic programs
can be found in [87], but few will deny that his influence has been widespread.
An account of the development of proof theory after the 1950’s would sup-
plement a discussion of Kreisel’s work with a discussion of work by major figures
like Solomon Feferman, William Tait, Kurt Schütte, and the Munich school of
ordinal analysis. Moreover, we have not discussed a number of important de-
velopments in proof theory in the 1930’s and 1940’s, most notably the work of
Herbrand;96 and we have given short shrift to the related study of intuitionism,
and the metamathematics of constructive theories, stemming from the work of
Brouwer’s student Arend Heyting. But space does not allow us to trace the
development of proof theory up through the present day, and so, for the most
part, we will leave off our historical account here.97 Before closing this section,
however, we wish to consider one more person who was influential in furthering
a formal, metamathematical point of view, namely, Stephen Kleene.
92 Kreisel [78], page 155. The emphasis is Kreisel’s.
93 Ibid.
94 Kreisel and Takeuti, Dissertationes Math. 118, page 35. Quoted in Arai [1].
95 Kreisel [80], page 395.
96 See Herbrand [55].
97 See footnote 102 for some papers and surveys that might help fill the gap.

35
 We began this essay with a quote from Kleene’s Introduction to Metamathe-
matics, the first edition of which appeared in 1952. A generation of logicians was
introduced to metamathematics by Kleene’s book, in much the same way that
the previous generation was introduced to the subject by Hilbert and Bernays’
Grundlagen der Mathematik.98 In the introductory chapters, Kleene is thorough
in acknowledging the influence of Hilbert and Bernays, and his metamathemat-
ical methodology is set forth clearly in Chapter III: the formalization of math-
ematics, the separation of theory and metatheory, and the use of only finitary
methods in the metatheory. There is only one respect in which Kleene distances
himself from a narrow construal of the formalist program:
We note in advance that metamathematics will be found to provide
a rigorous mathematical technique for investigating a great variety
of foundation problems for mathematics and logic, among which the
consistency problem is only one. For example, metamathematical
methods are applied now in studies of systematizations of mathe-
matics arising from the logicistic and intuitionistic schools, as well
as from Hilbert’s.. . . Our aim in the rest of this book is not to reach
a verdict supporting or rejecting the formalistic viewpoint in any
preassigned version; but to see what the metamathematical method
consists in, and to learn some of the things that have been discovered
in pursuing it.99
In other words, Kleene resists yoking metamathematics to the narrow goal of
assuring consistency, preferring to leave the goals broad and open-ended.
In fact, hints of such a broad view can already be found in Hilbert’s “Ax-
iomatic thought,” delivered in 1917, just before the eruption of the Grundla-
genstreit:
When we consider the matter more closely we soon recognize that the
question of the consistency of the integers and of sets is not one that
stands alone, but that it belongs to a vast domain of difficult epis-
temological questions which have a specifically mathematical tint:
for example (to characterize this domain of questions briefly) the
problem of the solvability in principle of every metamathematical
question, the problem of the subsequent checkability of the results
of a mathematical investigation, the question of a criterion of sim-
plicity for mathematical proofs, the question of the relationship be-
tween content and formalism in mathematics and logic, and finally
the decidability of a mathematical question in a finite number of
operations.100
In other words, Hilbert foresaw metamathematics as providing a mathematical
framework for studying a wide range of philosophical issues. In the next section,
98 Hilbert and Bernays [69].
99 Kleene [71], page 59.
100 Hilbert [61], paragraph 41.

36
we will consider some of the ways in which contemporary proof theory has borne
out his more general hopes.

6 Proof theory today

6.1 Contemporary perspectives
Today, the phrase “proof theory” means different things to different people.
At the end of The Problems of Philosophy, Bertrand Russell notes that when
the basic concepts and questions driving a line of philosophical inquiry can be
stated in a clear and precise fashion, the subject ceases to be called philosophy,
and becomes a science.101 In that sense, proof theory, like other branches of
logic that were born of philosophical need, has evolved into a branch of math-
ematics in its own right. Seen as such, proof theory is the general study of
deductive systems (mathematical or otherwise) and their properties, with inter-
nally motivated questions and research programs that are evaluated by ordinary
mathematical standards. In a branch often described as structural proof theory,
one studies structural properties of deductive systems, particularly with respect
to transformations (or “reductions”) of proofs; this is relevant to computational
pursuits like automated deduction, where choices of representation are central.
In the related branch known as proof complexity, one is interested in issues of
efficiency, with respect to various measures of proof length; from that point of
view deductive systems are thought of as being roughly equivalent when there
are, say, polynomial-time computable transformations between them.
In this section, we will focus on what might be called metamathematical (or
traditional ) proof theory, the branch which is, perhaps, most closely linked to
Hilbert’s original use of the term. In this branch, formal deductive systems are
used to explore the more general notion of mathematical provability, and the
particular choice of deductive system is only relevant as a means to this more
general end. Of course, the boundaries between the various subdisciplines are
not sharp, and ideas and methods flow freely between them. So the metamath-
ematical branch of proof theory should be viewed as embodying one perspective
on the subject, one that informs, and is informed by, the others.
In discussing this metamathematical branch, we will try to convey a sense
of the subject’s goals and methods, its interaction with other branches of proof
theory, and its interaction with mathematics, philosophy, and computer science.
A full-scale assessment of even just this one branch of proof theory, however,
lies well beyond the scope of this paper. Such an assessment would have to
address the writings of a number of contemporary logicians and proof-theorists,
as well as the writings of a number of philosophers that have been influenced
by a proof-theoretic point of view.102
101 Russell [96], page 155.
102 For a small and scattered sample of views on proof theory, see: articles by Feferman
[34, 35, 38, 39] and most of the essays collected in [36]; Girard [48]; Kohlenbach [72]; Kreisel
[78, 79, 80] (see also the collection of essays on Kreisel in [87]); Rathjen [92]; Sieg [101, 103,
104, 107]; Simpson [108] and the introduction to [109]; and Tait [111, 112]. See also the recent

37
 In Section 5.1, we saw that, since the 1920’s, work in metamathematical
proof theory has broadened its focus, from the specific goal of finding consis-
tency proofs to the more general goal of understanding abstract mathematical
reasoning in concrete terms. Modern proof theorists generally categorize for-
mal mathematical frameworks as either classical or constructive. (The word
“intuitionistic” is usually used interchangeably with “constructive.”) Often the
distinction hinges simply on whether or not the underlying logic includes the law
of the excluded middle. For example, Peano arithmetic, a classical theory, con-
sists of the consequences of a certain set of axioms in classical first-order logic;
while Heyting arithmetic, a constructive theory, consists of the intuitionistic con-
sequences of the same set of axioms. But in other cases the choice of language
and axioms reflects more general methodological differences. Roughly speaking,
the axioms of a “classically justified” theory are viewed as characterizing some
portion of the mathematical universe, while the axioms of a “constructively jus-
tified” theory are viewed as characterizing mathematical constructions. So, for
example, the axioms of Zermelo-Fraenkel set theory describe a universe of sets,
while the rules of Martin-Löf type theory offer a general framework for reasoning
about certain constructions, which have a clear computational interpretation.
For the purpose of studying aspects of classical mathematics, one needs to
have appropriate formal frameworks. Metamathematical proof theory has much
to offer in this respect, and we now know that a good deal of mathematics
can be formalized in theories that are, from a set-theoretic perspective, quite
weak. There is a long tradition, from Hilbert and Bernays, Weyl, and Heyting,
through Takeuti, to the present day, of representing tracts of mathematics in
such theories. More recently, the school of Reverse Mathematics, founded by
Harvey Friedman and Stephen Simpson, has calibrated the strength of a large
number of central mathematical theorems in terms of five weak subsystems of
second-order arithmetic.103 Similarly, proof theorists have studied fragments
of set theory, subsystems and extensions of first-order arithmetic, theories of
higher-order arithmetic, and Feferman’s theories of “explicit mathematics.”
Furthermore, we now have a robust understanding of such formal theories,
from a number of perspectives. To start with, we have a clear sense of the
foundational stances on which they are based. For example, there are a num-
ber of interesting theories that have the same strength as primitive recursive
arithmetic, which many take to embody a reasonable interpretation of Hilbert
and Bernays’ informal notion of finitary proof; and Feferman, among others,
has helped clarify the types of theories that are justified on the basis of a “pred-
icative” stance arising from the early work of Weyl. We also have a clear un-
derstanding of the relationship between these theories and constructive ones,
ranging from Heyting’s formalization of first-order arithmetic to theories of in-
tuitionistic inductive definitions and constructive type theories. In fact, one can
collection of essays [54]; and the online discussion [124], including a number of relevant postings
by Harvey Friedman. These writings are, for the most part, reflections on proof theory; see
[1, 19, 37, 113] and the references cited there for a survey of mathematical developments,
providing a sense of the state of the subject today.
103 See Simpson [109].

38
often find direct interpretations of the classical theories in their constructive
counterparts.104
A particular metamathematical goal of contemporary research is to explore
various measures of a theory’s mathematical “strength.” One such measure is
its consistency strength: a theory T1 is said to be stronger than T2 , in this sense,
if T1 proves that T2 is consistent. Another approach is to characterize a theory’s
computational strength. To understand this second notion, suppose a theory T
proves a statement of the form ∀x∃yR(x, y), where x and y range over the natu-
ral numbers, and R is a computationally decidable predicate. This tells us that
a computer program that, on input x, systematically searches for a y satisfying
R(x, y) always succeeds in finding one. Now suppose that f is a function that,
on input x, returns a value that is easily computed from the least y satisfying
R(x, y).105 Then f is a computable function, and we can say that the theory, T ,
proves that f is totally defined on the natural numbers. A simple diagonaliza-
tion shows that no effectively axiomatized theory can prove the totality of every
computable function in this way, so this suggests using the set of computable
functions that the theory can prove to be total as a measure of its strength.
A third, similar, measure, arises naturally in the study of ordinal analysis: one
can gauge the strength of a theory by the computable orderings that it proves
to be well ordered. In each case, the underlying intuition is that the more in-
finitary axioms one adds to a theory, the more consequences one obtains “down
below.” These measures may not be the only way of characterizing a theory’s
strength, and do not address more general issues of expressiveness, simplicity,
naturality, or fruitfulness. But they do provide at least one way of exploring
a theory’s capabilities, and can be viewed as “clarifying the nature of the in-
finite” by characterizing the computational and combinatorial consequences of
these axioms.
Overall, while we have not entirely fulfilled Gentzen’s goal of obtaining a
constructive interpretation of the full-fledged set-theoretic conception of classi-
cal mathematics, we do have detailed constructive analyses of fragments that
are close to mathematical practice. For example, with respect to the theo-
ries described above, we typically have exact characterizations of the types of
computable functions that one can prove to be total, in terms of recursion
along systems of ordinal notations, or in terms of natural computational and
type-theoretic constructs. In sum, we now know that a good deal of ordinary
mathematics can be formalized in theories that are considerably weaker than set
theory; and we have a thorough understanding of such theories from a construc-
tive point of view. In light of these developments, the schism between classical
and constructive viewpoints that seemed so divisive in the 1920’s now seems
rather benign.
Proof theorists are often branded as reactionaries by the larger logic com-
munity as a result of the emphasis on studying restricted formal frameworks for
mathematics. This stems from the viewpoint that proof theorists advocate the
104 See,
for example, Avigad [4].
105 For
example, R(x, y) may assert that y codes a halting computation of a particular Turing
machine on input x, and f may return the result of such a computation.

39
restriction of mathematical methods, as a means of guaranteeing consistency.
But few proof theorists doubt the consistency of Zermelo-Fraenkel set theory,
and even fewer maintain that such doubts should influence mathematical prac-
tice. The community of proof theorists does take it to be interesting that so
much of ordinary mathematics can be carried out in weak theories,106 but this
by no means implies that abstract methods should be barred from mathematics,
or that they are not fruitful, or even that they are less desirable than elemen-
tary ones. Indeed, some proof theorists, like Kreisel and Friedman, have argued
that metamathematical methods may help us discover mathematical theorems
whose proofs require stronger methods, like large cardinal hypotheses.

6.2 Proof theory and constructive mathematics

At this point, it is worthwhile to consider the relationship between proof theory
and constructive mathematics, since the two subjects share many of the same
concerns. A constructivist’s perspective on the tension between abstract and
concrete views of mathematics is nicely illustrated by the following quotation,
from the introduction to Errett Bishop’s Foundations of Constructive Analysis
(1967):
It appears then that there are certain mathematical statements that
are merely evocative, which make assertions without empirical valid-
ity. There are also mathematical statements of immediate empirical
validity, which say that certain performable operations will produce
certain observable results, for instance the theorem that every posi-
tive integer is the sum of four squares. Mathematics is a mixture of
the real and the ideal, sometimes one, sometimes the other, often so
presented that it is hard to tell which is which. The realistic com-
ponent of mathematics — the desire for pragmatic interpretation —
supplies the control which determines the course of development and
keeps mathematics from lapsing into meaningless formalism. The
idealistic component permits simplifications and opens possibilities
which would otherwise be closed. The methods of proof and the
objects of investigation have been idealized to form a game, but the
actual conduct of the game is ultimately motivated by pragmatic
106 Indeed, Harvey Friedman has conjectured that every number-theoretic theorem appear-

ing in the Annals of Mathematics from 1950 to the present day can be proved in elementary
arithmetic, a fragment of first-order arithmetic so weak that it cannot prove the totality of
iterated exponentiation (see “FOM: grand conjectures,” April 16, 1999, on the online forum
[124]). Some might judge this conjecture whimsical in its generality; but it is rendered some-
what more plausible by the fact that a good deal of algebra and analysis can be carried out
in conservative extensions of elementary arithmetic (see, for example, [74, 109] or the lists
of references in [3, 40]). Friedman’s conjecture emphasizes the importance of distinguishing
between proofs that are hard and proofs that require strong ontological or set-theoretic as-
sumptions. For that reason, it seems to us that an interesting metamathematical question is
as to whether Fermat’s last theorem can be proved in elementary arithmetic, and, if so, how
long such a proof must be.

40
 considerations.107
The views expressed here resonate with the weak formalist stance that we pre-
sented in Section 4.2. In particular, the last sentence acknowledges the formal
character of mathematics, without denying the importance of external, prag-
matic considerations.
Bishop’s work has arguably had as much effect on the tenor and tone of
modern constructivism as Brouwer’s. Bishop’s Foundations, and later his book
Constructive Analysis (co-written with his student Bridges), show that one can
refashion parts of classical mathematics constructively, by restricting general
definitions and restating them in constructive terms, modifying statements of
classical theorems to render them constructively valid, and exercising additional
care to ensure that proofs are direct, i.e. avoid the use of the law of the excluded
middle.
On the surface, there are crucial differences between Bishop’s program and
metamathematical proof theory. Bishop-style constructivism involves rework-
ing classical mathematics, in constructive terms; the Hilbert-style metamathe-
matical approach is to maintain a classical mathematical framework, and use
constructive methods in the metatheory. But in practice, the divisions are not
sharp, and in many ways the two approaches have grown closer to each other.
On the proof-theoretic side, formalizing classical mathematics in restricted the-
ories typically involves adapting mathematics accordingly, for example, focusing
on mathematical objects that can be coded as sets of natural numbers. Ideas
from constructive mathematics are often useful in this regard. On the construc-
tive side, in recent years there has been a good deal of interest in studying the
metamathematical properties of formal systems that suffice to carry out the
constructive development of mathematics, and endowing these systems with ex-
plicit computational semantics.108 So now the two approaches share much in
common, in trying to bridge the gap between abstract mathematics and concrete
computation: both are concerned with developing mathematics using methods
for which we have a constructive or computational understanding, in such a way
that general classical intuitions are maintained.

6.3 Contemporary goals

We have characterized a general goal of metamathematical proof theory as that
of modeling classical mathematical reasoning with formal systems, and study-
ing those formal systems from a constructive point of view. Of course, various
research programs harbor more specific goals, with orientations towards math-
ematics, philosophy, and computer science. In this section, we would like to
complete our discussion by mentioning some of these more specific goals and
their relation to Hilbert’s original hopes.
107 Bishop [11], page viii.
108 For more information on constructivism, see, for example, Bishop and Bridges [12], Beeson
[7], Troelstra and van Dalen [114], or the recent survey, Bridges and Reeves [15].

41
 A good deal of work in proof theory is devoted to foundational purposes, in
the sense of providing a formal representation and analysis of the basic concepts
and assumptions used in mathematics. Such formalizations bring mathemati-
cal benefits on their own, unifying disparate branches of the subject, clarifying
concepts, and raising new mathematical questions.109 Note, however, that in
contrast to set theory, where the emphasis is on finding a framework broad and
powerful enough to characterize as much of mathematics as possible in one fell
swoop, proof theorists are typically more interested in studying local founda-
tions for mathematics, i.e. systems that minimally suffice to capture various
fragments of mathematical practice. For example, a proof theorist might be in-
terested in determining the weakest theory in which one can develop a workable
theory of real analysis, in the hopes that interesting constructive or computa-
tional information can be extracted from such a formal study. We might define
“recursive analysis” to be the fragment of real analysis that can be carried out in
RCA0 , the weakest of the five major subsystems of second-order arithmetic con-
sidered by the Reverse Mathematics school; a proof theorist is then interested
in knowing which theorems of analysis can be carried out in this theory. The
metamathematical analysis yields additional information: for example, anything
provable in RCA0 remains true when second-order quantifiers are assumed to
range over computable sets; and any “computational” consequence of RCA0 is
witnessed by a primitive recursive function.
As Hilbert observed repeatedly, the formalization of mathematics also makes
independence results possible. The Gödel-Cohen proofs of the independence of
the axiom of choice and the continuum hypothesis from the axioms of Zermelo-
Fraenkel set theory are a striking example, but there are others, both lower
down as well as higher up in the lineup of mathematical theories. For example,
recursive counterexamples to theorems in analysis tell us that these theorems
are independent from RCA0 . Recently Razborov [93], building on work by
Razborov and Rudich [94], considered the issue of the independence of lower
bounds in circuit complexity from weak fragments of arithmetic. The possi-
bility of obtaining natural combinatorial statements that are independent of
interesting mathematical theories is also quite striking. The independence of
the Paris-Harrington theorem from Peano arithmetic, and Friedman’s proof of
the independence of a finitary version of Kruskal’s theorem from theories of
predicative mathematics, are well known. The program of obtaining natural
combinatorial independences from much stronger theories is actively being pur-
sued by Friedman.110
There is continuing work towards characterizing the strength of formal the-
ories in terms of the traditional measures described above. Powerful compu-
tational schemata are used to characterize the strength of increasingly strong
theories; in the other direction, more restrictive classes of functions, like the
polynomial-time computable ones, have been characterized in terms of weak
fragments of arithmetic. Modern ordinal analysis draws heavily on intuitions
109 See Feferman [32] for a nice discussion of the various benefits of formalization.
110 See, for example, Friedman, “Boolean relation theory”, March 10, 2000 posting to [124].

42
of the infinite that arise from the set-theoretic study of large cardinals, and we
now seem to be close to having an ordinal characterization of the strength of full
second-order arithmetic, which formed the backdrop for Hilbert and Bernay’s
formalization of analysis.111 Other proof-theorists are pursuing Kreisel’s goal of
extracting useful mathematical information from classical proofs; for example,
Ulrich Kohlenbach has applied proof-theoretic methods to extract bounds on
rates of convergence of algorithms arising in the study of numerical approxima-
tions.112
We have already noted that the formal modeling of mathematical practice
can be philosophically informative, by clarifying the types of reasoning that
are justified under various philosophical stances and allowing us to explore the
strengths and limitations of the corresponding methods. In recent years, many
philosophers have turned their attention to issues that complement traditional
foundational ones, seeking accounts of the growth and development of mathe-
matical concepts, the structure of mathematical theories, mathematical expla-
nation, and the interaction between pure and applied branches of the subject.113
Issues like these can be fruitfully explored from a number of perspectives, and
it is to be expected that formal metamathematical methods will add clarity and
insight.
Finally, let us consider some aspects of proof theory that are more closely re-
lated to computer science. Some of the most vibrant areas of proof theory today
address computational concerns: metamathematical analyses of type theory are
central to the theory of functional programming languages, and the methods
of structural proof theory are central in automated deduction, logic program-
ming, and the design of expert systems.114 As hardware and software becomes
more complex, there has been increasing interest in formal methods of verifi-
cation, yielding further possibilities for proof-theoretic applications. There are
also striking relationships between the field of computational complexity and
the field of proof complexity.115 More broadly, a general interest in the storage,
manipulation, and communication of mathematical knowledge requires a solid
foundational understanding of mathematical language and proof.
Mathematics is a vital human pursuit, one we put great stock in; metamath-
ematical proof-theory is nothing more than a sustained, rigorous study of this
vast field of inquiry, with a particular focus. The general goal of the subject re-
mains that of studying the nature of mathematical proof in various domains, in a
way that is philosophically insightful, mathematically interesting, and computa-
tionally informative. What the various branches of the subject have in common
is a focus on symbolic representations of mathematical language, inference, and
proof. These concerns did not originate with Hilbert, but his Beweistheorie did
111 For surveys of recent developments in ordinal analysis, see, for example, Pohlers [89] and

Rathjen [92].
112 See, for example, [74, 72], for an overview of the methods, and [73, 75] for applications.
113 See, for example, Breger and Grosholz [51].
114 See, for example, Constable [23], Jäger and Stärk [70], and Troelstra and Schwichtenberg

[113].
115 See Beame and Pitassi [6], Buss [18], Krajıček [77], Pudlák [91], or Urquhart [115] for a

general overviews.

43
play an important part in making them central to modern mathematical logic.
In that sense, Hilbert’s program has been an enormous success.

References
[1] Toshiyasu Arai. Review of the Handbook of Proof Theory. Bulletin of
Symbolic Logic, 6:463–477, 2001.
[2] William Aspray and Phillip Kitcher, editors. History and Philosophy of
Modern Mathematics. University of Minnesota, Minneapolis, 1988.
[3] Jeremy Avigad. Weak theories of nonstandard arithmetic and analysis.
To appear in Stephen Simpson, editor, Reverse Mathematics 2001.
[4] Jeremy Avigad. Interpreting classical theories in constructive ones. Jour-
nal of Symbolic Logic, 65:1785–1812, 2000.

[5] Steve Awodey and Erich H. Reck. Completeness and categoricity: 19th
century axiomatics to 21st century semantics. Preprint.
[6] Paul Beame and Toniann Pitassi. Propositional proof complexity: past,
present, and future. Bulletin of the European Association for Theoretical
Computer Science, 65:66–89, 1998.
[7] Michael Beeson. Foundations of Constructive Mathematics. Springer,
Berlin, 1985.
[8] Paul Benacerraf and Hilary Putnam, editors. Philosophy of Mathematics:
Selected Readings. Second edition, Cambridge University Press, Cam-
bridge, 1983.
[9] Paul Bernays. Die Philosophie der Mathematik und die Hilbertsche
Beweistheorie. Blätter für deutsche Philosophie, 4:326–367, 1930–1931.
Translated as “The Philosophy of Mathematics and Hilbert’s Proof The-
ory” by Paolo Mancosu in [83], pages 234–265.

[10] Paul Bernays. Hilbert, David. In Paul Edwards, editor, The Encyclopedia
of Philosophy. MacMillan, the Free Press, and Collier MacMillan, New
York, 1967, pages 496–504.
[11] Errett Bishop. Foundations of Constructive Analysis. McGraw-Hill, New
York, 1967.
[12] Errett Bishop and Douglas Bridges. Constructive Analysis. Springer,
Berlin, 1985.
[13] Otto Blumenthal. Lebensgeschichte. In volume 3 of Hilbert’s Gesammelte
Abhandlungen [67], page 388–429.

44
[14] George Boole. The Mathematical Analysis of Logic. Barclay and Macmil-
lan, Cambridge, 1847. Reprinted in [31], volume 1, pages 451–509.
[15] Douglas Bridges and Steve Reeves. Constructive mathematics in theory
and programming practice. Philosophia Mathematica, 7:65–104, 1999.
[16] L. E. J. Brouwer. Intuitionism and formalism, 1912. Brouwer’s inaugural
address at the University of Amsterdam. An English translation by Arnold
Dresden appears in the Bulletin of the American Mathematical Society,
20:81–96, 1913–1914, and is reprinted in Ronald Callinger, editor, Classics
of Mathematics, Prentice Hall, Englewood, 1982, pages 734–740.
[17] L. E. J. Brouwer. Intuitionistische Betrachtungen über den Formalismus.
Koninklijke Akademie van wetenshappen te Amsterdam, Proceedings of
the section of sciences 31, 374–379, 1928. Translated as “Intuitionistic
reflections on formalism” by Stefan Bauer-Mengelberg in [116], pages 490–
492.
[18] Samuel Buss. Propositional proof complexity — an introduction. In Ulrich
Berger and Helmut Schwichtenberg, editors, Computational Proof Theory,
pages 127–178. Springer, Berlin, 1997.
[19] Samuel Buss, editor. The Handbook of Proof Theory. North-Holland,
Amsterdam, 1998.
[20] Georg Cantor. Grundlagen einer allgemeinen Mannigfaltigkeitslehre.
Ein mathematisch-philosophischer Versuch in der Lehre des Unendlichen.
Teubner, Leipzig, 1883. Translated by William Ewald as “Foundations of
a general theory of manifolds: a mathematico-philosophical investigation
into the theory of the infinite” in [31], volume 2, pages 878–920.
[21] Rudolf Carnap. Logische Syntax der Sprache. Springer, Wien, 1934. Trans-
lated as The Logical Syntax of Language, Routledge and Kegan Paul, Lon-
don, 1937.
[22] Rudolf Carnap. Empiricism, semantics, and ontology. Revue Interna-
tionale de Philosophie, 4:20–40, 1950. Reprinted in [8], pages 241–257.
[23] Robert Constable. Types in logic, mathematics, and programming. In
Buss [19], pages 683–786.
[24] Richard Dedekind. Stetigkeit und irrationale Zahlen. Vieweg, Braun-
schweig, 1872. Translated by W. W. Beman as “Continuity and irrational
numbers” in Essays on the theory of numbers, Open Court, Chicago, 1901;
reprinted by Dover, New York, 1963. The Beman translation is reprinted,
with corrections by William Ewald, in [31], volume 2, pages 765-779.
[25] Richard Dedekind. Was sind und was sollen die Zahlen. Vieweg, Braun-
schweig, 1888. The second edition, with a new preface, was published in
1893, and is translated by W. W. Beman as “The nature and meaning of

45
 numbers” in Essays on the theory of numbers, Open Court, Chicago, 1901;
reprinted by Dover, New York, 1963. The Beman translation is reprinted,
with corrections by William Ewald, in [31], volume 2, pages 787-833.
[26] Charles Delzell. Kreisel’s unwinding of Artin’s proof. In Odifreddi [87],
pages 113–246.
[27] René Descartes. Rules for the Direction of the Mind. Edited and translated
by L. Lafleur. Liberal Arts Press, 1961.
[28] Harold Edwards. The genesis of ideal theory. Archive for History of Exact
Sciences, 23:321–378, 1980.
[29] Harold Edwards. Kronecker’s place in history. In Aspray and Kitcher [2],
pages 139–144.
[30] Harold Edwards. Kronecker’s views on the foundations of mathematics. In
D. E. Rowe and J. McCleary, editors, The History of Modern Mathematics,
pages 67–77. Academic Press, 1989.
[31] William Ewald, editor. From Kant to Hilbert: A Source Book in the
Foundations of Mathematics. Clarendon Press, Oxford, 1996. Volumes 1
and 2.
[32] Solomon Feferman. Working foundations. Synthese, 62:229–254, 1985. An
expanded and updated version appears in [36], pages 105–124.
[33] Solomon Feferman. Infinity in mathematics: Is Cantor necessary? In
G. Toraldo di Francia, editor, L’infinito nella scienza, pages 151–209.
Istituto della Enciclopedia Italiana, Rome, 1987. Reprinted in Feferman
[36], pages 28–73 and 229–248.
[34] Solomon Feferman. Proof theory: a personal report. Appendix to Gaisi
Takeuti, Proof Theory, second edition, North-Holland, Amsterdam, 1987,
pages 447–485, 1987.
[35] Solomon Feferman. Hilbert’s program relativized: proof-theoretical and
foundational reductions. Journal of Symbolic Logic, 53:364–384, 1988.

[36] Solomon Feferman. In the Light of Logic. Oxford University Press, New
York, 1998.
[37] Solomon Feferman. Highlights in proof theory. In Hendricks et al. [54],
pages 11–31.

[38] Solomon Feferman. Relationships between constructive, predicative, and

classical systems of analysis. In Hendricks et al. [54], pages 221–236.
[39] Solomon Feferman. Does reductive proof theory have a viable rationale?
Erkenntnis, 53:63–96, 2000.

46
[40] A. M. Fernandes and Fernando Ferreira. Groundwork for weak analysis.
To appear in the Journal of Symbolic Logic.
[41] Gottlob Frege. Die Grundlagen der Arithmetik, eine logisch-
mathematische Untersuchung über den Begriff der Zahl. Koebner, Bres-
lau, 1884. Translated as The Foundations of Arithmetic, Northwestern
University Press, Evanston, 1980.
[42] Hans Freudenthal. Zur Geschichte der Grundlagen der Geometrie. Nieuw
Archief voor Wiskunde, V(4):105–142, 1957. In part, a review of the 8th
edition of Hilbert [58].
[43] Carl Friedrich Gauss. Foundations of mathematics. Translated by G.
W. Dunnington in G. W. Dunnington, editor, Inaugural Lecture on As-
tronomy and Papers on the Foundations of Mathematics. Louisiana State
University Press, Baton Rouge, 1937. Reprinted in [100], pages 42–45.
[44] Gerhard Gentzen. Der Unendlichkeitsbegriff in der Mathematik.
Semester-Berichte, Münster, 9th Semester, Winter, 1936–1937. Trans-
lated as “The concept of infinity in mathematics” in [47], pages 223–233.
[45] Gerhard Gentzen. Die Widerspruchsfreiheit der reinen Zahlentheorie.
Mathematische Annalen, 112:493–465, 1936. Translated as “The consis-
tency of elementary number theory” in [47], pages 132–213.
[46] Gerhard Gentzen. Neue Fassung des Widerspruchsfreiheitsbeweises für
die reine Zahlentheorie. Forschungen zur Logik und zur Grundlegung der
exakten Wissenschaften, New Series 4:19–44, 1938. Translated as “New
Version of the Consistency Proof for Elementary Number Theory” in [47],
pages 252–286.
[47] Gerhard Gentzen. Collected Works. North-Holland, Amsterdam, 1969.
Edited by M. E. Szabo.
[48] Jean-Yves Girard. On the meaning of logical rules I: syntax versus seman-
tics. In Ulrich Berger and Helmut Schwichtenberg, editors, Computational
proof theory, pages 215–273. Springer, Berlin, 1997.
[49] Kurt Gödel. Über eine bisher noch nicht benützte Erweiterung des finiten
Standpunktes. Dialectica, 12:280–287, 1958. Reproduced with English
translation in Feferman et al., eds., Kurt Gödel: Collected Works, volume
2, Oxford University Press, New York, 1990, pages 241–251.
[50] Jeremy Gray. The Hilbert Challenge. Oxford University Press, Oxford,
2000.
[51] Emily Grosholz and Herbert Breger. The Growth of Mathematical Knowl-
edge. Kluwer Academic Publishers, Dordrecht, 2000.

47
[52] Ian Hacking. “Style” for historians and philosophers. Studies in History
and Philosophy of Science, 23:1–20, 1992.
[53] G. H. Hardy. A Course of Pure Mathematics. Cambridge University Press,
Cambridge, tenth edition, 2000.
[54] Vincent Hendricks, Stig Andur Pederson, and Klaus Frovin Jorgensen,
editors. Proof Theory: History and Philosophical Significance. Kluwer
Academic Publishers, Dordrecht, 2000.
[55] Jacques Herbrand. Recherches sur la théorie de la démonstration. PhD
thesis, University of Paris, 1930. Herbrand’s thesis. An English translation
appears in [56], and Chapter 5 is reprinted in [116].
[56] Jacques Herbrand. Collected Works. Harvard University Press, Cam-
bridge, 1971. Warren Goldfarb, editor.
[57] David Hilbert. Die Theorie der algebraischen Zahlkörper. Jahresbericht
der Deutschen Mathematiker-Vereinigung, 4:175–546, 1897. Often referred
to as the Zahlbericht. Reprinted in [67].
[58] David Hilbert. Grundlagen der Geometrie. In Festschrift zur Feier der
Enthüllung des Gauss-Weber Denkmals in Göttingen. Teubner, Leipzig,
1899. Translated by Leo Unger as Foundations of Geometry, Open Court,
1971. Ninth printing, 1997.
[59] David Hilbert. Über den Zahlbegriff. Jahresbericht der Deutschen
Mathematiker-Vereinigung, pages 180–194, 1900. Translated by William
Ewald as “On the concept of number” in [31], volume 2, pages 1089–1095.
[60] David Hilbert. Über die Grundlagen der Logik und der Arithmetic. Ver-
handlungen des dritten internationalen Mathematiker-Kongresses in Hei-
delberg vom 8. bis 13. August 1904, pages 174–185, 1905. Translated by
Beverly Woodward as “On the foundations of logic and arithmetic” in
[116], pages 129–138.
[61] David Hilbert. Axiomatisches Denken. Mathematische Annalen, 78:405–
415, 1918. Delivered to the Swiss Mathematical Society in Zürich on
September 11, 1917. Translated by William Ewald as “Axiomatic thought”
in [31], volume 2, pages 1105–1115.
[62] David Hilbert. Neubegründung der Mathematik. Erste Mitteilung. Ab-
handlungen aus dem mathematischen Seminar der Hamburgischen Uni-
versität, 1:157–177, 1922. Translated by William Ewald as “The new
grounding of mathematics. First report.” in [31], volume 2, pages 1115–
1134.
[63] David Hilbert. Die logischen Grundlagen der Mathematik. Mathematis-
che Annalen, 88:151–165, 1923. Presented to the Deutsche Naturforscher
Gesellschaft in Leipzig in September, 1922. Translated by William Ewald

48
 as “The logical foundations of mathematics” in [31], volume 2, pages 1134–
1148.
[64] David Hilbert. Über das Unendliche. Mathematische Annalen, 95:161–
190, 1926. Delivered to the Westphalian Mathematical Society on June
4, 1925. Translated as “On the infinite” by Stefan Bauer-Mengelberg in
[116], pages 367–392.
[65] David Hilbert. Die Grundlagen der Arithmetik. Abhandlungen aus
dem mathematischen Seminar der Hamburgischen Universität, 6:65–85,
1928. An address delivered at the Hamburg Mathematical Seminar in
1927. Translated as “The foundations of arithmetic” by Stephan Bauer-
Mengelberg and Dagfinn Føllesdal in [116], pages 464–479.
[66] David Hilbert. Die Grundlegung der Mathematik. Mathematische An-
nalen, 104:485–494, 1931. Translated as “The grounding of elementary
number theory” by William Ewald in [31], volume 2, pages 1148–1157.
[67] David Hilbert. Gesammelte Abhandlungen. Springer, Berlin, 1932–35.
Three volumes.
[68] David Hilbert and Wilhelm Ackermann. Grundzüge der theoretischen
Logik. Springer, Berlin, 1928. The second German edition was published
in 1938, and has been translated as Principles of Mathematical Logic by
Lewis Hammond, George Leckie, and F. Steinhardt, edited and with notes
by Robert Luce. AMS Chelsea Publishing, Providence, 1991.
[69] David Hilbert and Paul Bernays. Grundlagen der Mathematik. Springer,
Berlin, first volume, 1934, second volume, 1939.
[70] Gerhard Jäger and Robert Stärk. A proof-theoretic framework for logic
programming. In Buss [19], pages 639–682.
[71] Stephen Kleene. Introduction to Metamathematics. North-Holland, Ams-
terdam, fourth reprint, 1964 edition, 1952.
[72] Ulrich Kohlenbach. Foundational and mathematical uses of higher types.
Preprint.
[73] Ulrich Kohlenbach. New effective moduli of uniqueness and uniform a-
priori estimates for constants of strong unicity by logical analysis of known
proofs in best approximation theory. Numerical Functional Analysis and
Optimization 14:581–606, 1993.
[74] Ulrich Kohlenbach. Mathematically strong subsystems of analysis with
low rate of growth of provably recursive functionals. Archive for Mathe-
matical Logic, 36:31–71, 1996.
[75] Ulrich Kohlenbach. A quantitative version of a theorem due to Borwein-
Reich-Shafrir. Numerical Functional Analysis and Optimization 12:641–
656, 2001.

49
[76] A. N. Kolmogorov and A. P. Yushkevich, editors. Mathematics of the
19th Century: Mathematical Logic, Algebra, Number Theory, Probability
Theory. Birkhäuser, Basel, 1992. English translation of Matematika XIX
veka: Mateticheskaya logika, algebra, teoriya chisel, teoriya veroyatnostei,
Izdatel’stvo “Nauka,” Moskva, 1978.
[77] Jan Krajı́ček. Bounded Arithmetic, Propositional Logic, and Complexity
Theory. Cambridge University Press, Cambridge, 1995.
[78] Georg Kreisel. Mathematical significance of consistency proofs. Journal
of Symbolic Logic, 23:155–182, 1958.
[79] Georg Kreisel. A survey of proof theory. Journal of Symbolic Logic,
33:321–388, 1968.
[80] Georg Kreisel. Proof theory: some personal recollections. Appendix to
Gaisi Takeuti, Proof Theory, second edition, North-Holland, Amsterdam,
1987, pages 395–405.
[81] Leopold Kronecker. Über den Zahlbegriff. In Philosophische Aufsätze,
Eduard Zeller zu seinem fünfzigjährigen Doctorjubiläum gewidmet, pages
261–274. Fues, Leipzig, 1887. Translated as “On the concept of number”
by William Ewald in [31], volume 2, pages 947–955.
[82] Paolo Mancosu. The Russellian influence on Hilbert and his school.
Preprint.
[83] Paolo Mancosu, editor. From Brouwer to Hilbert: the Debate on the Foun-
dations of Mathematics in the 1920’s. Oxford University Press, Oxford,
1998.
[84] Paolo Mancosu. Between Russell and Hilbert: Behmann on the founda-
tions of mathematics. Bulletin of Symbolic Logic, 5:303–330, 1999.
[85] Mathieu Marion. Kronecker’s “safe haven of real mathematics.” In
M. Marion and R. S. Cohen, editors, Québec Studies in the Philosophy
of Science I, pages 189–215. Kluwer Academic Publishers, The Nether-
lands, 1995.
[86] A. F. Monna. The concept of function in the 19th and 20th centuries,
in particular with regard to the discussions between Baire, Borel, and
Lebesgue. Archive for the history of exact sciences, 9:57–84, 1972.
[87] Piergiorgio Odifreddi, editor. Kreiseliana: About and Around Georg
Kreisel. A K Peters, Wellesley, Massachusetts, 1996.
[88] Volker Peckhaus. 19th century logic between philosophy and mathematics.
Bulletin of Symbolic Logic, 5:433–430, 1999.
[89] Wolfram Pohlers. Subsystems of set theory and second order number
theory. In Buss [19], pages 209–335.

50
 [90] Henri Poincaré. Les mathémathique et la logique. Revue de métaphysique
et de morale, 13:815–35, 1905. Translated as “Mathematics and logic: II”
by William Ewald (incorporating a translation by George Halsted) in [31],
volume 2, pages 1021–1038.
[91] Pavel Pudlák. The lengths of proofs. In Buss [19], pages 547–637.
[92] Michael Rathjen. The realm of ordinal analysis. In S. Barry Cooper
and John Truss, editors, Sets and Proofs (Leeds, 1997), pages 219–279.
Cambridge University Press, Cambridge, 1999.
[93] Alexander Razborov. Unprovability of lower bounds on circuit size in
certain fragments of bounded arithmetic. Izvestiya of the RAN, 59:201–
222, 1995.
[94] Alexander Razborov and Stephen Rudich. Natural proofs. 26th Annual
Symposium on the Theory of Computing, 1994.

[95] Constance Reid. Hilbert. Springer, Berlin, 1970.

[96] Bertrand Russell. The Problems of Philosophy. Home University Library,
1912. Reprinted by Oxford University Press, Oxford, 1959.
[97] Bertrand Russell. The Principles of Mathematics. Allen and Unwin, sec-
ond edition, 1937. The first edition was published in 1903.
[98] Bertrand Russell and Alfred North Whitehead. Principia Mathematica.
Cambridge University Press, first volume, 1910; second volume, 1912;
third volume, 1913.
[99] Michael Scanlan. Who were the American postulate theorists? Journal
of Symbolic Logic, 56:981–1002, 1991.
[100] W. Shirmacher, editor. German Essays on Science in the 19th Century.
Continuum, New York, 1996.
[101] Wilfried Sieg. Foundations for analysis and proof theory. Synthese, 60:159–
200, 1984.
[102] Wilfried Sieg. Reflections on Hilbert’s program. In Wilfried Sieg, editor,
Acting and Reflecting: The Interdisciplinary Turn in Philosophy, pages
171–182. Kluwer Academic Publishers, Dordrecht, 1990.
[103] Wilfried Sieg. Relative consistency and accessible domains. Synthese,
84:259–297, 1990.
[104] Wilfried Sieg. Aspects of mathematical experience. In E. Agazzi and
G. Darvas, editors, Philosophy of Mathematics Today, pages 195–217.
Kluwer Academic Publishers, 1997.

51
[105] Wilfried Sieg. Hilbert’s programs: 1917–1922. Bulletin of Symbolic Logic,
5:1–44, 1999.
[106] Wilfried Sieg. Towards finitist proof theory. In Hendricks et al. [54], pages
95–114.
[107] Wilfried Sieg. Beyond Hilbert’s reach? preprint.
[108] Stephen Simpson. Partial realizations of Hilbert’s program. Journal of
Symbolic Logic, 53:349–363, 1988.
[109] Stephen Simpson. Subsystems of Second-Order Arithmetic. Springer,
Berlin, 1998.
[110] Howard Stein. Logos, logic, and logistiké. In William Aspray and Phillip
Kitcher, editors, History and Philosophy of Modern Mathematics, pages
238–259. University of Minnesota, 1988.
[111] William Tait. Finitism. Journal of Philosophy, 78:524–546, 1981.
[112] William Tait. Truth and proof: the platonism of mathematics. Synthese,
69:341–370, 1986. Reproduced in W. D. Hart, editor, The philosophy of
mathematics, Oxford University Press, Oxford, 1996, pages 142–167.
[113] A. S. Troelstra and Helmut Schwichtenberg. Basic Proof Theory. Cam-
bridge University Press, Cambridge, second edition, 2000.
[114] A. S. Troelstra and Dirk van Dalen. Constructivism in Mathematics: An
Introduction, volume 1. North-Holland, Amsterdam, 1988.
[115] Alasdair Urquhart. The complexity of propositional proofs. Bulletin of
Symbolic Logic, 1:425–467, 1995.
[116] Jean van Heijenoort. From Frege to Gödel: A sourcebook in mathematical
logic, 1879-1931. Harvard University Press, Camridge, 1967. Republished
by toExcel, San Jose, 2000.
[117] Johann von Neumann. Die formalistische Grundlegung der Mathematik.
Erkenntnis, volume 2. Translated as “The formalist foundations of math-
ematics” by E. Putnam and G. J. Massey in [8], 61–65.
[118] Hermann Weyl. Über die neue Grundlagenkrise der Mathematik. Mathe-
matische Zeitschrift, 10:39–79, 1921. A lecture delivered in Zürich in 1920.
Translated as “On the new foundational crisis of mathematics” by Benito
Müller in [83], pages 86–118.
[119] Hermann Weyl. Diskussionsbemerkungen zu dem zweiten Hilbertschen
Vortrag über die Grundlagen der Mathematik. Abhandlungen aus dem
Mathematische Seminar der Hamburgischen Universität, 6:86–88, 1928.
Comments presented after Hilbert’s lecture [65]. Translated as “Comments
on Hilbert’s second lecture on the foundations of mathematics” by Stefan
Bauer-Mengelberg and Dagfinn Føllesdal in [116], page 482.

52
[120] Hermann Weyl. David Hilbert and his mathematical work. Bulletin of
the American Mathematical Society, pages 612–654, 1944.
[121] Alfred North Whitehead. A Treatise on Universal Algebra. Hafner, New
York, 1960. Reprinting of the Cambridge University Press edition, 1898.
[122] A. P. Youschkevitch. The concept of function up to the middle of the 19th
century. Archive for the History of the Exact Sciences, 16:37–85, 1973.
[123] Richard Zach. Completeness before Post: Bernays, Hilbert, and the de-
velopment of propositional logic. Bulletin of Symbolic Logic, 5:331–366,
1999.
[124] Foundations of mathematics online discussion forum. On the world wide
web, http://www.math.psu.edu/simpson/fom.

53