THE ETHNOMETHODOLOGICAL FOUNDATIONS OF

MATHEMATICS
Studies in ethnomethodology

Edited by Harold Garfinkel

Professor of Sociology
University of California, Los Angeles
Theclhnomclhod~o~c~
foundations of mathematics

Eric Livingston

Routledge & Kegan Paul

London, Boston and Henley
Dedicated to my parents
Herbert and Rosetta Livingston

First published in 1986

by Routledge & Kegan Paul plc
14 Leicester Square, London WC2H 7PH, England
9 Park Street, Boston, Mass. 02108, USA and
Broadway House, Newtown Road,
Henley on Thames, Oxon RG9 lEN, England

Set in Times, 10 on 11 pt
by Hope Services of Abingdon
and printed in Great Britain
by Billing & Sons Ltd., Worcester

Copyright© Eric Livingston 1986

No part of this book may be reproduced in

any form without permission from the publisher,
except for the quotation of brief passages
in criticism

Library of Congress Cataloging in Publication Data

Livingston, Eric.
The ethnomethodological foundations of mathematics.
(Studies in ethnomethodology)
Bibliography: p.
1. Mathematics -Philosophy. 2. Ethnomethodology.
I. Title. II. Series.
QA8.4.L48 1985 511.3 85-1898
British Library CIP data also available
ISBN 0-7102-0335-7
Contents

Preface ix

Acknowledgments xii

Introduction
A Non-Technical Introduction to Ethnomethodological
Investigations of the Foundations ofMathematics
through the Use of a Theorem ofEuclidean Geometry
A Guide to the Reading of this Book 15

Part I 21

Introduction. The Phenomenon: The Existence of Classical

Studies of Mathematicians' Work 23

A Review of the Classical Representation of

Mathematicians' Work as Formal Logistic Systems 25

2 An Introduction to Godel's Incompleteness Theorems:

Their Metamathematicallnterpretation Contrasted with
the Proposal to Study Their Natural Accountability in
and as the Lived-Work of Their Proofs 31

Part II A Descriptive Analysis of the Work of Proving

Gooel's First Incompleteness Theorem 37

3 GOdel Numbering and Related Topics: Background Materials

for a Proof of Godel's Theorem 39
v
CONTENTS

4 The Double-Diagonalization/'Proof: Features of the Closing

Argument of a Proof of Go del's Theorem as Lived-Work 45

5 A Technical Lemma: A Lemma Used in the Proof of Godel's

Theorem; Its Origins as a Technical Residue of the Work of
Proving Go del's Theorem within that Self-Same Work 51

6 Primitive Recursive Functions and Relations: An Initial

Discussion of the Irremediable Connection between a Prover's
Use of the Abbreviatory Practices/Practical Techniques of
Working with Primitive Recursive Functions and Relations
and the Natural Accountability of a Proof of GOdel's Theorem 57

7 A Schedule of Proofs: An Extended Analysis of the Lived-Work

of Producing the Body of a Proof of Godel's Theorem 65

A A Schedule of Proofs 65
B A Schedule of Proofs as Lived-Work 69

1 As it is used in developing a schedule of proofs, a Godel

numbering is not an abstractly defined correspondence
between the symbols, expressions, and sequences of
expressions of P and the natural numbers; it is a
technique of proving 69
2 The schedule of proofs has a 'directed' character: it
leads to and is organized so as to permit, the
construction of G as a primitive recursive relation 76
3 The selection and arrangement of these-particular
propositions as composing this-particular,
intrinsically sequentialized order of proving
is the situated achievement of the work of
producing that schedule of proofs 80
(a)' Six themes concerning the lived-work of producing a
schedule of proofs 81
(b )The construction of a schedule of proofs so as to
provide an apparatus within itself for the analysis of
the work of its own construction 100
(c) The correspondence between the propositions of a
schedule of proofs and the syntax of formal number
theory as an achievement of the schedule of proofs
itself 117
(d)A review of the work of generalizing a schedule of
proofs so as to elucidate the character of the
development and organization of a schedule as a
radical problem, for the prover, in the production
of social order 125

vi
 CONTENTS

4 The work of providing a consistent notation for a schedule

of proofs articulates that schedule as one coherent
object 137

8 A Structure of Proving 149

A The Characterization Problem: The Problem of
Specifying What Identifies a Proof of Code/'s Theorem
as a Naturally Accountable Proof ofJust That Theorem;
The Texture of the Characterization Problem and the
Constraints on Its Solution; The Characterization
Problem as the Foundational Problem 149
B Generalizing the Proof of Code/'s Theorem (As a Means of
Gaining Technical Access to the Characterization
Problem) 154
C A Structure ofProving: The Availability to a Prover of the
Proof of Code/'s Theorem as a Structure of Practices; The
Proof as the Pair The-Proof/The-Practices-of-Proving-to-
Which- That-Proofis-Irremediably-Tied i 69

Part III Conclusion 173

9 Summary and Directions for Further Study 175

A Classical Studies ofMathematical Practice: A Review of
the Book's Argument 175
B Prospectus: Mathematicians' Work as Structure Building 177

Appendix 179

The Use of Ethnomethodological Investigations of Mathematicians'

Work for Reformulating the Problem of the Relationship between
Mathematics and Theoretical Physics as a Real-World Researchable
Problem in the Production of Social Order 181

Notes 190

Bibliography 237

vii
Preface

The early Greeks were both amazed and perplexed by mathematical

proofs. On one hand, the objects of geometry were made available
and described, and their properties were established, through the use
of drawn figures. Yet the Greeks recognized that the geometric objects
themselves had a curious, unexplicated relationship to their depiction.
They were further puzzled by the fact that the mathematical propo-
sition was demonstrated not as a matter of rhetorical argumentation,
merely to the satisfaction of mathematicians immediately present.
The mathematical theorem was proved as something necessarily true,
a fact anonymous as to its authorship, available for endless inspection,
established for all time - and this as a required feature of the actual
demonstration itself. When it was proved that the 'field' of construct-
able lengths contained incommensurable elements, the Greeks were
unable to turn away from the evidentness of the mathematical demon-
stration even though it went against their deepest philosophical
commitments. It was said that Pythagoras, on proving this fact, com-
mitted suicide by drowning.
At the turn of this century great interest was again shown in the
origins of mathematical truth and in the nature of mathematical proofs.
This was stimulated, in part, by circumstances similar to the Pythagorean
proof of the existence of irrational numbers: developments in set
theory had led to the proved existence of a continuous, 'space-filling'
curve; even more spectacularly (but received with much greater sus-
picion), Zermelo proved that any set could be well-ordered. As a
particular case, it follows that there exists some partial ordering of the
real numbers such that every nonempty subset of them has a first
element. Zermelo's proof, however, did not show what that partial
order is.
At the same time as these developments, elegant proofs using set-
theoretic methods - such as the proof that the transcendental numbers
ix
PREFACE
are more numerous than the algebraic ones - were given. But it was
also shown, by ingenious but elementary reasoning, that the unexpli-
cated notion of a 'set' led to contradictions, the most famous of these
being Russell's paradox. Hilbert's solution of the 'invariant problem'
demonstrated the power of abstract methods but, for some, raised
questions as to the sense in which the problem had actually been
solved. The burgeoning development of the new geometries had
invigorated philosophers' consideration of the empirical status of
mathematical truth. And Hilbert's researches in geometry again showed
the strength of the axiomatic method but exhibited as well flaws in
the reasoning of Euclid's Elements.
One consequence of these origins of early twentieth-century investi-
gations of the foundations of mathematics was regrettable. The
predominating interest of those studies, at least in the received view,
was to demonstrate that the methods used by practising mathematicians
and the results established through ordinary mathematical practice
were themselves free from criticism. The attempt was made to construct
indubitable foundations for mathematical practice. In consequence,
interest in the original question - what made up the evident and
transcendental character of mathematical proofs? - shifted to the
problem of demonstrating the incorrigibility of those same proofs.
During this time, mathematical research continued unabated, little
affected by the investigations directed specifically to mathematical
foundations. Moreover, it was mathematicians' daily production of
ordinary, naturally accountable proofs that supplied not only the
promised object but the basis of the logician's research. Although
foundational studies did have an indirect effect on mathematical
practice, a consequence of their most celebrated achievement, Godel's
incompleteness theorems (themselves proved in the style of ordinary
mathematics), was to hasten the incorporation of mathematical logic
into mainstream mathematical research, not to alter mathematical
practice. The living foundations of mathematics - and, as a particular
case, the origins of the adequacy of GOdel's own proofs - remained
untouched and unexamined.
These remarks set in contrast the direction taken in the present
work. This book is a study of the foundations of mathematics, but in
the original sense. It is a study of the genetic origins of mathematical
rigor, examining the proofs of ordinary mathematics and investigating
how the adequacy of such proofs, for the purposes of everyday mathe-
matical inquiry, is practically obtained. The book formulates and, in a
certain sense, solves the problem of the foundations of mathematics as a
problem in the local production of social order. It does this not by
reviewing received philosophies of mathematics, not by proffering a
theory of social action, not through an historical or cultural analysis,
but by rediscovering and exhibiting the naturally accountable mathe-

X
 PREFACE

matical proof, in its identifying detail for mathematicians, as a social

achievement. This book asks and attempts to describe how mathe-
maticians, in .and as their daily work, as the sine qua non of their work
for and in the company of mathematicians, produce the naturally
accountable proofs of ordinary mathematics which, as the witnessed,
local achievement of the work of their demonstration, therein exhibit
both the transcendental character of that work and of the objects
those proofs describe.
Eric livingston
May 1984

xi
Acknowledgments

This book could not have been written without the assistance, super-
vision and support of Harold Garfmkel. His influence has been pervasive
in formulating the problem of mathematical foundations as a problem
in the production of social order and in carrying out the research. His
suggestions provided critical directions for further investigation, and his
encouragement was steadfast.
In many different ways I am indebted to many other people.
Emanuel Schegloff and Melvin Seeman provided, each in his own way,
penetrating criticisms and sustaining intellectual and emotional support.
Herbert Enderton and Louis Narens showed interest in my research and
assisted me in learning the necessary mathematics. Conversations with
Professor Enderton, and his close attention to the mathematical details
of my argument, were invaluable; Professor Narens's approval of the
dire<.;tion of my investigations - if not, perhaps, his agreement with
their results - was a source of intellectual sustenance. I thank as well
Jeffery Alexander, Donald Babbitt, Robert Blattner, Phillip Bonacich,
Henry A. Dye, Rod Harrison, Rudolph De Sapio, V.S. Varadarajan
and Robert Westman. One of my brothers, Charles livingston, himself
a professional mathematician, gave unstintingly of his time to further
my studies and was supportive of them over their entire course.
Over the course of my undergraduate and graduate studies I was
fortunate in being part of a local culture of students and faculty engaged
in studies of naturally organized ordinary activities. I take this occasion
to express my indebtedness to them, and I thank, in particular, Melinda
Baccus, Stacy Burns, Trent Eglin, Richard Fauman, Harold Garfinkel,
George Girton, Gail Jefferson, Ken Liberman, Michael Lynch, Douglas
MacBeth, Ken Morrison, Christopher Pack, Anita Pomerantz, Britt
Robillard, Friedrich Schrecker, the late Harvey Sacks, Emanuel
Schegloff, Dave Sudnow and Larry Wieder.
I thank my friends not already mentioned and, in particular, Paul
xii
 ACKNOWLEDGMENTS

Colomy and Maureen McConaghy. I thank Mary Takami, the staff of

the Department of Sociology, UCLA, and the Department of Mathe-
matics, UCLA (for unrestricted use of the Department of Mathematics'
Reading Room). I thank Diane Wells, Librarian for the Department of
Philosophy's Reading Room, for her assistance in locating historical
materials. Special thanks go to Ralph Edwards, Ted Regier and Ralph
Edwards Productions for their encouragement and unsolicited support.
Readers of this book will notice my debt, particularly in la:ter
chapters, to a book by Joel W. Robbin, Mathematical Logic: A First
Course. I am grateful to Benjamin/Cummings Publishing Company for
granting me permission to draw on it as I have. I am grateful as well
to Harcourt Brace Jovanovich, Inc. for granting me permission to
reproduce the illustration that appears on page 145.
An Exxon Research Fellowship in the Program in Science, Techno!"
ogy, and Society at M.I.T. provided support for continuing my studies
of the discovering sciences. During my fellowship the final version of
this book was prepared, as well as the arrangements for the book's
publication. I gratefully acknowledge this support and the hospitality
of the STS faculty, students and staff.
Finally, I am indebted to Peter Hopkins, past editor-director of
Routledge & Kegan Paul, for his willingness to publish this book, and,
above all else, to my parents, my dearest friend Michelle Arens, and
my brothers, Lewis and Charles, for things that go without saying.

xiii
Introduction

A Non-Technical Introduction to Ethnomethodological

Investigations of the Foundations of Mathematics through
the Use of a Theorem of Euclidean Geometry•

I want to advance two recommendations for the study of mathe-

maticians' work. The first one is this: by returning the mathematical
object to its origins within mathematical practice, it is possible to
investigate the natural analyzability and natural accountability of
mathematicians' work in and as the inspectable, real world practices
of professional mathematicians. The second recommendation is that
those investigations then offer increasing, and technical, access to the
natural analyzability and natural accountability of mathematicians'
work as a problem, for mathematical theorem provers, in the local
production of social order.
The aim of my talk is to indicate the grounds for these recommen-
dations in ongoing, ethnomethodological studies of mathematicians'
work and, in doing so, to offer those studies as a distinctive and vital
alternative to classical studies of that work. 1
To this end, I am going to begin my talk in a somewhat unusual
manner: I am going to begin by proving a theorem of Euclidean

* This chapter is the text of a talk delivered at the annual meeting of the
American Sociological Association in San Francisco in September 1982. It is
intended to serve as a non-technical introduction to the remainder of the book.
In writing this chapter, I benefited from the extensive and generous help of
several people: Charles Livingston spent hours reviewing and helping me develop
the mathematics of the paper; Emanuel Schegloff read numerous drafts of the
paper, and his suggestions and criticisms were invaluable in articulating and
organizing the paper's argument. I am indebted as well to Herbert Enderton for
many stimulating conversations and to Michelle Arens, Paul Colomy, Rod
Harrison and Anita Pomerantz for their critical suggestions. Responsibility for
the material in this chapter is, of course, my own.
INTRODUCTION

geometry. This proof, in that it will recall for us the appearance of

mathematical argumentation, will be useful for sketching the current
state of studies about mathematicians' work and, then, later, as a means
of indicating what some features of the lived-work of mathematics
actually look like.
The theorem that I want to prove is this: First, given a circle with
center C and an angle a inscribed in it, the measure of that angle's
intercepted arc fhere } 2 is

defined to be the measure of the angle (3 that that arc is said to subtend.

The theorem that I am going to prove is that the measure of angle a is

1/2 the measure of angle (3, or more properly, that the measure of an
inscribed angle is 1/2 the measure of its intercepted arc.
The proof of this theorem goes like this: first, let us distinguish
three possible cases: the case where the center of the circle lies on one
of the edges of the angle:

the case where the center of the circle is interior to the angle:

and the case where the center of the circle is exterior to the angle:
2
 INTRODUCTION

Let us prove the theorem for {this case} E:J first.

By drawing in the subtended angle and labeling the angles a and J3,
what we wish to show is that the measure of a is 1/2 the measure of /3.

But note that {this line} and {this line} are radii of the circle and that,
therefore, they have equal length.

This means that fthis trianglef is an isosceles triangle and that, therefore,
fthis angle} must equal {this angle} so {this angle} is a, also.

Now note. that since {this} is a triangle, a plus a plus {this angle} must
equal 180°, and since J3 plus Hhis anglef make a straight line, J3 plus
{this angle} must also equal 180°. But then the measure of (3 must
equal the measure of a plus the measure of a or that
2 m(a) = m(fj)
or that
1
m(a) =- m((j)
2
We have finished with case one.
3
INTRODUCTION

If we turn to fthis case} ® next, we see, by drawing in the

diameter of the circle through the vertex of the angle

and the subtended angle

that if we can show that the measure of fa} is 1/2 the measure of
fj3} and if the measure of fy} is 1/2 the measure of f8}

then the measure of the sum of a and r will be 1/2 the sum of (3 and 5,
or that the measure of the inscribed angle will be 1/2 the measure of
the intersected arc. But by our first case, we see that the measure of a
is 1/2 the measure of {3, and similarly that the measure of r is 1/2 the
measure of 8, so we have finished with case two.
Case three is similar to the last case but a little more difficult to
see. Let us first draw in the diameter and the subtended angle.

If we label {this angle} a and {this angle} {3, and correspondingly,

fthis angle} a' and fthis angle} (J',
4
 INTRODUCTION

then we see that the incribed angle is a - (l and that the subtended
angle is a' - (l'. Then
f'the measure of the inscribed angle a- (l equals'}
m(a ,-- (l) = m(a)- m((l)

= t m(a')- ~ m((l') +'again, by case one'i

1 (a
=2m I
-(j ' )

{'equals the measure of the sub tended angle a'- (l'.'t

Thus, for this case as well, the measure of the inscribed angle is 1/2
the measure of its intercepted arc, and we are done.
Now, before undertaking an analysis of some of the work of this
proof, I want to use it to sketch briefly the central concerns of previous
studies of mathematicians' work.
Throughout its entire history, the study of mathematicians' work
has been directly to the solution of three problems: the problem of
explicating the notion of a rigorous proof; the problem of characteriz-
ing the mathematical object; and the problem of understanding the
nature of mathematical discovery.
Let me briefly elaborate on these problems. For provers, a math-
ematical proof (like the one I just gave) is not a form of rhetorical
argumentation; its cogency - whether or not, or in what sense, it
actually is the case - is not exhibited as being dependent on the par-
ticulars of its authorship or its audience; on the contrary, a proof
exhibits, as its witnessible achievement, the demonstrable adequacy
and practical objectivity of its reasoning. For provers, a theorem is not
true because what it asserts has merely been persuasively argued - the
measure of an inscribed angle is not 1/2 the measure of its intercepted
arc because I am a good debater - it is true because its proof exhibits
it to be so. The problem of explicating the notion of a rigorous proof
is that of determining what kind of a thing a proof is that this should
be so.
Next, consider the objects that concern a geometer. They are not,
for example, the circles and angles that he actually draws on the board,
nor are his theorems, in any obvious sense, empirical or empirically-
verifiable propositions. Mathematical objects - whether circles and
5
INTRODUCTION

angles or infinite-dimensional Hilbert spaces and unitary transformations

- appear as being disengaged from, and are, as a prover's accomplish-
ment, exhibited as being disengaged from, the actual material circum-
stances of a prover's examination of them. They are circles qua circles,
angles qua angles, and their properties are the properties of circles qua
circles and angles qua angles.
The problem of characterizing the mathematical object is the problem
of determining what kind of objects a mathematician refers to when he
proves theorems about them; what kind of objects are circles qua
circles and angles qua angles.
And, finally, let us consider how a theory of the mathematical
object affects our appreciation of the nature of mathematical discovery.
I will give two examples. First, if the mathematical object is con-
sidered to be an ideal, Platonic object - if the circles and angles of
geometry are understood as representing ideal circles and ideal angles
- then a theorem of mathematics, like our theorem about inscribed
angles, is a description of such objects and a mathematical discovery,
in some sense, is an intuition of the transcendental properties of those
ideal objects.
If, however, mathematical objects, like circles and angles, are under-
stood as being entirely defined by the structural relationships holding
between the symbolic expressions of a purely formal, logistic system,
then the achievedly-disengaged objects - like circles and angles - are
historical artifacts used for their suggestiveness and for their embedded-
ness in the practical ways we have of reasoning about them; and a
mathematical discovery is a discovery concerning the structure of
logical inference which makes its appearance through, but is only
incidentally related to, the practical objects of current mathematical
interest.
Thus, the problem of understanding the nature of mathematical
discovery is seen to be bound up with the problem of characterizing
the mathematical object, as, in fact, all three of these problems are
tied to one another.
We have, thus, briefly reviewed- in an obviously sketchy manner-
the classical problems that have animated studies of mathematicians'
work. Now the first thing that I want to call to your attention is a
fact massively recognized by practising mathematicians: none of the
studies of mathematical rigor, of the nature of the mathematical
objects, or of mathematical discovery - whether done by mathemati-
cians reflecting on their own practice, or by philosophers, historians
or sociologists - none of these studies are, for the practising math-
ematician, either instructive in or consequential for his own work
practices at the mathematical work-site.
At the work-site, the mathematician does not encounter the problem
of mathematical rigor as 'The Problem of Mathematical Rigor,' but as
6
 INTRODUCTION

the endlessly diverse, practical problem of finding and organizing, in

their exhibiting and problem-specific material detail, coherent,
accountable, mathematical arguments.
At the work-site, the mathematician is not attentive to theories of
the mathematical object- Platonic, logistic, or otherwise - his attention
is directed to fmding and elucidating the properties, for example, of
circles qua circles and angles qua angles - their ontology, for practical
purposes, being assured in their availability for further precise, rigorous
description.
And at the work-site, the mathematician is not interested in a theory
of mathematical discovery; he is interested in making mathematical
discoveries.
Of course, to those interested in the study of mathematicians' work,
these complaints are all too familiar. What can be gained by parading
them out?
Let us note the following, curious situation: on every occasion
when practising mathematicians come to address seriously, for and
among colleagues, the accountability of a piece of mathematical
argumentation, and on every occasion when an accountable mathemati-
cal argument is discovered, both the specific character of the problem-
at-hand and the specific character of that problem's solution arise from
within the situated, lived-work of doing professional mathematics -
that is, as a work-site phenomenon. But, in contrast, on every occasion
when the nature of mathematical activity is posed as its own topic for
mathematical, philosophical, historical, or sociological reflection
and analysis, the mathematical object is first disengaged from the
local work to which its natural accountability and analyzability are
integrally tied.
We come, then, to my first recommendation: in order to re-address
the classical problems in the study of mathematical practice in a
manner that is located within that practice, the mathematical object .
must first be returned to its origins within that practice itself, and by
so returning the mathematical object to those origins, the natural
analyzability and accountapility of mathematicians' work is capable of
being investigated in and as the inspectable, real-world practices of
professional mathematicians.
As a way of indicating the basis of this recommendation in current
ethnomethodological studies, I want now to return to our proof that
the measure of an inscribed angle is J/2 the measure of its inter~
cepted arc and - given the limitations of time - point out some of the
more obvious features of the lived-work of that proofs natural
analyzability.
First, let us review the argument that I gave for the case where the
center of the circle lay on one of the edges of the angle.

7
INTRODUCTION

Here, I argued that since fthis line} and fthis line} are radii, they have
equal length, fthis triangle} is an isosceles triangle and, therefore, that
both {these angles} are o:. But, then, since fthis angle} plus fthis
angle} plus fthis angle} measure 180 °, and since fthis angle} and
{this angle} measure 180 °, the measure of {3 must be twice that of the
measure of o:, or that the measure of o: must be 1/2 the measure of {3.

I Now, the first point that I want to make is this: nowhere in the
course of giving this proof did I define the notions of a circle, a triangle,
an angle, the measure of an angle, an inscribed angle, or an intercepted
arc; and, in fact, nowhere in the proof did it seem that we needed a
'theory' of those objects. 3
The point to be made is this: in order for us to prove our theorem,
we did not need to articulate such definitions. All the properties of
angles and measures that we needed were exhibited over the course of
giving the proof itself, and, moreover, as fellow provers, what you held
me to was the fact that all such properties would be so exhibited. 4
As just one example, consider the fact that in drawing the inscribed
angle

and then labeling it o:;

8
 INTRODUCTION

we are showing that that angle, whatever it is, is related to fthis area}
between the two drawn chords of the circle. In this way, then, the
angles of the drawing become seeably definite; and that this is so then
provides for the adequacy of partitioning the inscribed angles into
classes determined by the relationship of the angles to the center of
the circle: that is, {pointing to the cases drawn on the board} the case
where the center lies on one of the edges of the angle, is in the interior
of the angle, or is exterior to the angle.
Similar remarks could be made for the defmition of the measure
of an angle, or an intercepted arc, of triangles, and, in fact, of all the
objects of pointed relevance for our proof.
In the course of giving this illustration, another point is illustrated
as well - namely, that the figure is temporally drawn and that it takes
on its properties by being tied to the embodied action of its depiction.
One sees this property of the proof no more in evidence than when
trying to write up the proof on paper. The 'this's' of the drawing-
ordered and timed with the embodied pointings, paced writings and
accompanying talk - must be removed by locally developed notational
devices adequate to the particular problems-at-hand.
Consider for example, labeling the points of the figure

e9
necessary for the proof of our first case. We need just these:

~
Qv
On seeing the figure so labeled, however, the prover sees that fthis
point} is not labeled. In order to disengage the labeling from its occa-
sioned circumstantiality, a prover will label 'all' the points; moreover,

figus
if the prover is going to identify the angle a in the fashion of L PQC,
then he will want to be able to label {3 in the same way, thereby remov-
ing the need for denoting the angles a and (1 Moreover, when the
J>'Over comes to JabeJ the

9
INTRODUCTION

he does so as follows:

~let C be the center of the circle and let the inscribed angle meet the
circle at points P, Q, and R'} thereby 'tracing out' and 'showing again'
the inscribed angle, with the center of the circle labeled in such a way
so as to disengage it from the angle and being seen as being so disengaged.
Here, of course, a prover will not see this ordering of the work as
being essential to the adequacy of the representation: a prover envisions
the temporal drawing in terms of the projected gestalt that it is working
toward, and he will see beyond another prover's mistakes by rearrang-
ing the ordering of the drawing, for himself, so that it is a proper order.
It is the availability of such work that provides the inessential
character of a drawing's temporal construction.
But, on the other hand, if a prover does enough things incorrectly -
if his writings are not paced with his talk, if he organizes his material
presentation improperly - then the naturally analyzable mathematical
object will not be exhibited. And that this is so points to the fact that
provers work in such a way that this will not be the case. Moreover,
from within the production of the naturally analyzable object, the
mathematician will use that natural analyzability to argue that the local
work of its production had nothing to do with that object's analyzable
properties.
Let me draw this set of materials together: if we consider the pri-
mordial setting of mathematics to be those occasions when mathemati-
cians, in the presence of one another, work in such a way so as to
exhibit to each other the recognizable adequacy of their work, then one
of the things that we have seen is that mathematicians work in such a
way so as to disengage the mathematical object from the situated work
that makes it available and, therein, to disengage that object from the
situated work that gives it its naturally accountable properties. Thus,
the first major point is that the naturally accountable mathematical
object is the local achievement of mathematical provers.

II Let me offer a second collection of observations, these pointing to

the work of achieving the objective, transcendental character of the
reasoning of our proof.

First, let us look at the following drawing for the proof of our
first case:
10
 INTRODUCTION

One thing to note about this picture is that there is nothing about the
drawing itself that guarantees that these are all the necessary lines
needed for the proof. That these are all the necessary lines depends,
in fact, on the surrounding course of argumentation from within which
the necessary details of the figure take on their relevance specifically
for the particular argument being given.

2 A second observation is this: in beginning to prove our theorem,

I offered three cases

as being the necessary ones to consider for that proof. Now, without
even examining the local work of providing oneself with a scheme of
reasoning such that these are established as a disjoint yet exhaustive
set of cases, we, as provers, can see that -a priori -there is no particu-
lar significance in the fact that inscribed angles can be partitioned in
this way - there are other ways of partitioning these angles and,
associated with those partitionings, other ways of proving our
theorem. 5 The point is that when a prover, as myself, offers these three
cases as partitioning inscribed angles into equivalence classes, he is
pointing to, and other provers are waiting to see, the course of proving
that then follows that exhibits that partition as a partition adequate to
the proof of the theorem. Moreover, in that a prover has written the
cases as first fthis onet, then fthis onet, and then fthis onet,

other provers come to look for that temporal course of drawing to be

realized in the ensuing proof itself. And in our proof, not only is this order-
ing preserved, but it reflects a dependency between the cases themselves.
11
INTRODUCTION

3 As one final observation concerning the exhibited reasoning of the

proof, we should note that we have proved our theorem for each of the
three classes of inscribed angles by proving it for an arbitrary represen-
tative of each class. In the first case, fhere}

the proof for this picture stood as the proof for all possible inscribed
angles in which the center of the circle lay on one of its edges - that is,
for all angles like fthis} and fthis} as well:

Now the thing that gives the arbitrariness of the angles and, thus,
that gives the generality of the proof for this class of angles, is the fact
that the method of proving exhibits it as an arbitrary angle. That is, the
same method of proving our theorem for fthis angle} -

by drawing the subtended angle

and reasoning as we did - applies to any other inscribed angle in which

the center of the circle lies on one of its edges. Thus, the arbitrariness
of the angle - the character of the particular angle as a representative
of a class of such angles - is tied to the course of reasoning that exhibits
that angle as such an arbitrary one.
12
 INTRODUCTION

In summary, then, these brief observations point to the local, lived-

work of producing the practically accountable, practically objective
reasoning of our proof.

III The last major point that I wish to make has to do with the larger
structural organization of our proof.
Consider the relationship between how a prover comes to find such a
proof and the final, materially-presented proof of it. As an example,
a prover may; summarily speaking, come to find the relevant cases for
proving our theorem about inscribed angles in the following manner:
by first reviewing a picture such as fthisf

a prover may realize, after trying other schemes, that by drawing the
subtended angle

and the diameter through the vertex of the inscribed angle

that if he could prove the theorem for fthis case}-

then it would follow for fthe 'original' case}-

13
INTRODUCTION

Seeing this, and looking at the two cases and searching for an extract-
able method of classifying them that gives them as two possible cases
- like that concerning the relationship of the inscribed angle to the
center of the circle - that method then offers both the existence of a
third possibility - namely, where the center of the circles is exterior to
the angle

- and, therein, offers those three cases as making up all the possible
ones. 6
Now, when a prover comes to write up the proof, he will not detail
this work, but will, as I did in proving the theorem earlier, simply offer
the three cases as the necessary cases for the proof.
If we put this observation together with the other aspects of the
lived-work of proving this theorem that I have detailed, we come to the
following point: a prover, in the course of working out a proof, extracts
from the lived-work of that proof, the accountable structure of that
work - that is, he extracts the specifically remarkable features of the
presented proof, and does so against the background of practices that
both provide for that structure and that, simultaneously, that structure
makes available. A proof, then, is not the disengaged, material argument,
but it is always tied to the lived-work of that theorem's particular
proving; a proof is this inextricable pairing of the proof and the associ-
ated practices of its proving.
Having gone through this review of our proof, I now want to return
to my original argument.
Consider that the primordial origins of mathematics, that the genera-
tive and sustaining life of the discipline, lies in the coming together of
provers who, in the presence of each other, in exhibiting the demons-
trable adequacy of a mathematical line of argumentation, exhibit -
as their witnessible and recognizable achievement - the practical
objectivity of that argument, therein, their presence to it as mathemati-
cal provers, and therein, simultaneously, the adequacy of that argument
for anyone whomsoever. Consider that mathematics, as a professional
14
 INTRODUCTION

discipline, is sustained and renewed, and that it evolves, is revitalized,

and is taught on the occasions, and critically on the occasions, when
provers come together and do, for and among each other, the recog-
nizably adequate work of doing recognizably adequate mathematics.
Then, let us make one last observation concerning classical studies of
mathematicians' work: by addressing the problems of explicating the
notion of a rigorous proof, of chamcterizing the mathematical object,
and of understanding the nature of mathematical discovery, classical
studies of these problems have sought to explain, and, thereby, have
indirectly pointed to and depended on, the great regularities of mathe-
matical pmctice - namely, that mathematicians regularly produce
naturally accountable mathematical proofs; that mathematicians, as
their situated accomplishment, endlessly demonstrate the independence
of mathematical objects from the local work through which those objects
take on their naturally analyzable properties; that mathematicians make
recognizable and teachable mathematical discoveries; and that their
proofs and discoveries concern, particularly, the naturally accountable
objects given by mathematical practice.
We can see, then, in that the originating intention of those studies is
to account for the witnessible regularities of mathematical practice,
that those studies are, in fact, properly conceived of as theories of
social structure and practical action - as theories of how mathemati-
cians, in their daily work, as their daily, specifically- and identifiably-
mathematical work, come to produce, maintain, transform and exhibit
the observable and identifying orderliness of their practice.
In conclusion, then, we come to the second of my recommend-
ations concerning the study of mathematicians' work. It is this: by
seeing beyond the classical studies of mathematicians' work to the
produced and witnessible orderliness of mathematical practice that
those studies indirectly attempt to explain - that is, by affiliating
ourselves with their originating intentions and not with their received
histories - and by returning the mathematical object to its origins
within the mathematical work-site, it is then possible to take up the
study of mathematicians' work not as a means of elaborating familiar
and well-known topics, but, instead, as the investigation of the natural
analyzability and natural accountability of mathematical practice; to do
so as a problem, for provers, in the irremediably local production of
social order, and to do so by investigating, as discoverable matters,
the natural analyzability and accountability of mathematical practice,
in its identifying detail at the work-site, in and as the real-world
practices of professional mathematicians.

A Guide to the Reading of this Book

On every occasion when the professional mathematician comes to
15
INTRODUCTION

address seriously, in the company of colleagues, the adequacy, the

rigor, the accountability of some piece of mathematics, and on every
occasion when an accountable mathematical argument is discovered,
both the specific character of the problem-at-hand and that problem's
solution arise in the midst of and are made available from within the
situated, lived-work of doing professional mathematics. Yet on every
occasion in which the origin of mathematical rigor is posed as its own
topic for mathematical, philosophical, historical or sociological reflec-
tion and analysis, that rigor is disengaged by the analyst from the local
work that, in every particular case, provides for its achievement. In this
book I raise the question: what constitutes mathematical rigor from
within the settings in which that rigor lives as the circumstances of
doing professional mathematics? What I propose, and make available
as a real-worldly researchable matter, is that mathematical rigor consists
of the local work of producing and exhibiting, for and among math-
ematicians, a 'followable' - and, therein, a naturally accountable -
line of mathematical argumentation. If the problem of the foundations
of mathematics is understood as the problem of discovering the origins
of and analyzing mathematical rigor, then this book can be summarized
by saying that it formulates that problem as a problem in the production
of social order and solves it in the sense that that formulation is shown
to permit the investigation of mathematical foundations in and as the
real-world practices of mathematicians engaged in doing professional
mathematics.
The book begins with a short introductory chapter giving the neces-
sary definitions that allow me to advance and make precise the claim
that conventional studies of the foundations of mathematics are
'classical' studies. In that these studies have themselves become a prac-
ticeable mathematical discipline, their legitimacy and authority rest
on the fact that the mathematical reasoning that occurs in them is
itself mathematically rigorous. The claim that they are 'classical' studies
is the claim that the local, order-productive work of an 'ordinary'
mathematical argument's natural accountability is irremediably hidden
within them in and as the self-same practices of producing naturally
accountable mathematics. Thus, the characterization of conventional
studies as classical studies is tied to the solution of the foundation
problem itself.
As a first step in developing and explicating this claim, and as a means
of introducing the reader to the methods employed in foundational
studies, Chapter 1 provides a contrast between a very simple mathemati-
cal proof and the representation of that proof in a formal logistic
system. This contrast is used both to illustrate the inadequacy of such a
formal system as a literal description of mathematical practice and to
point out the 'reasoner's work' that surrounds even a simple proof and
recovers that proof as a naturally accountable mathematical object.
16
 INTRODUCTION

The inadequacy of logistic sytems for descriptively representing

mathematical practice is, of course, recognized by mathematicians and
mathematical logicians· alike; the claim that foundational studies
actually concern - and are even primordial to - ordinary mathematical
practices rests not on such systems' literal descriptiveness, but on the
construction of those systems so as to model supposedly essential
features of mathematical practice. Hilbert is generally credited with the
idea that the proofs of the consistency of such systems would supply
the foundations for those systems, for the mathematical theories that
they represent, and, by implication, for the ordinary mathematical
practices used in reasoning about those theories. The attempt to
develop such systems and to then prove their consistency is generally
referred to as the 'Hilbert program.'
It is against the background of this program that the interpretation
of Go del's celebrated incompleteness results, published in 1931, is most
often presented. Preceding Godel's discoveries, one -if not the central
- problem of the Hilbert program was to demonstrate the consistency
of a formal system 'representing' number theory. Godel's theorems
can be summarized by saying that they showed the impossibility of
proving, from within the formal system itself, the consistency of any
'reasonable' formal system strong enough to 'contain' number theory.
Thus, Godel's theorems point to the seeming paradox that, on one
hand, the discovery, demonstration and adjunction of rigorous math-
ematics make up the daily work circumstances of mathematicians and
mathematical logicians alike and, on the other, that the independence
of mathematical rigor from the lived-work of proving cannot be demon-
strated by mathematical methods.
Chapter 2 begins by reviewing this material. The proofs of Godel's
theorems, however, are not proofs given within a formal system, but
are 'informal' proofs of everyday mathematics about formal systems.
The question is thereby raised: what is it that makes up the rigor of
their proofs as proofs of ordinary mathematics? What is it; indepen-
dently of their metamathematical interpretation, that constitutes the
natural accountability of the proofs of G6del's theorems as 'informal'
proofs of everyday mathematics? Thus, Chapter 2 provides an introduc-
tion to Godel's incompleteness theorems and then offers the contrast
between the metamathematical interpretation of them and the proposal
to study the natural accountability of them in and as the lived-work
of their proofs. Part II of the book - divided into chapters 'G6del
Numbering and Related Topics,' 'The Double-Diagonalization/"Proof','
'A Technical Lemma,' 'A Schedule of Proofs,' and 'A Structure of
Proving' - provides an answer to that question by descriptively analyzing
the work of proving one of those theorems. Part II is a case study of the
origins of mathematical rigor in the work practices of mathematicians.
In 'G6del Numbering and Related Topics,' I introduce the basic
17
INTRODUCTION

mathematical apparatus needed for the proof of Godel's first incom-

pleteness theorem - a Godel numbering, the definitions of primitive
recursive functions and relations, and the definitions of representability
and numeralwise expressibility. Although most of the material in this
chapter only lays the groundwork for my later analyses, the discussion
does point to the various topics that gain importance over the remainder
of Part II.
In 'The Double-Diagonalization/"Proof' ,' the concluding part of the
proof of Godel's theorem is reviewed and then analyzed through a list
of fifteen descriptive 'remarks.' These remarks elucidate the ways in
which the diagonalization argument and the 'proof mutually articulate
each other and the ways in which that mutual articulation is relied on
to construct the diagonalization/'proof.' This analysis of the simultane-
ous development of a pair - (a mathematical construction/a proof) -
clarifies features of some types of mathematical theorem proving but,
as the reader will see, it is not intended as a prototypic account of
mathematicians' work.
'A Technical Lemma' discusses a second component of the proof of
Godel's theorem - the proof of the lemma that primitive recursive
relations are numeralwise expressible in first-order arithmetic. This
lemma has the character of being a 'technical residue' of the proof of
Godel's theorem as a whole. In this chapter, I attempt to explain how
its character as a 'technical residue' arises from within the work of
proving Godel's theorem itself.
The main body of the proof of Godel's theorem consists of a sequen-
tially ordered list of functions and relations and the interwoven proofs
that those functions and relations are primitive recursive. This list of
propositions and proofs will be referred to in the book as the 'schedule
of proofs.' The chapter 'Primitive Recursive Functions and Relations'
addresses the practical techniques of showing that a function or relation
is primitive recursive. The idea of this chapter is to indicate how those
practical techniques provide for their own demonstrable adequacy,
thereby radicalizing the notion that those techniques are 'grounded.'
The material presents a curious backdrop to the work of constructing
a naturally accountable schedule of proofs.
'A Schedule of Proofs' begins by outlining a particular schedule of
proofs and ends by describing four of that schedule's features in terms
of the lived-work of its production. First, a Godel numbering is shown
to be a technique of proving rather than just a correspondence defmed
between the language of a formal system and the natural numbers.
Next, the ordering of the propositions of the schedule is shown to have
a 'directed' character. Third, the propositions of the schedule of proofs
are shown to be selected and arranged so as to provide for and exhibit
the accountable orderliness of the work of producing that schedule.
Fourth, I provide some initial materials on the way in which the work
18
 INTRODUCTION

of producing a consistent notation for a schedule of proofs articulates

that schedule as one coherent object.
The detailed analysis of the work of proving Godel's theorem is
brought to a conclusion in 'A Structure of Proving.' Here, the problem
is raised of specifying what identifies a proof of Godel's theorem, over
the course of its production and review, as a naturally accountable
proof of just that theorem. The analysis of the book up to this point
allows some immediate constraints to be placed on prospective solu-
tions: primarily, that such a solution cannot be found by reference
to an idealized proof or to an idealized collection of mathematical
practices. What I propose is that a proof of Godel's theorem has a
'structure of proving' - that is, that a schedule of proofs is arranged
in such a way so as to exhibit the accountable orderliness of the work
of producing that schedule, and in that this is so, that schedule simul-
taneously provides directions for its own recovery as a naturally
accountable mathematical object, both in its material detail and as
pointing to the diagonalization/'proof and, therein, to the necessary
bridging step of the technical lemma. More generally, the idea that a
proof has a structure of proving is that the material proof of a theorem
is available in practice as a template of a course of action, that course
of action being the work of recovering the proof for which the material
presentation stands as its projected, natural and practically adequate
account.
The chapter 'A Structure of Proving' brings to an end Part II of the
book. Part III, a short concluding chapter, ties together the book's
overall argument. Its point is this: even though Part II was restricted to
the analysis of the work of proving one particular theorem, the results
of that analysis provide for further, similarly detailed investigations
of mathematicians' work. What was discovered about the natural
accountability of a proof of Godel's theorem seems to resonate
throughout mathematical praxis. In that this is so, the argument turns
back on itself: By discovering the origins of mathematical rigor in and
as the work of producing naturally accountable, ordinary mathematics,
conventional foundational studies, in so far as they are a mathematical
discipline, must essentially hide those origins in and as the self-same
practices of doing recognizably adequate, rigorous lffiathematics.
The book concludes with an appendix that points to the consequen-
tiality of my studies of mathematicians' work for the investigation
of the relationship between mathematics and theoretical physics.
This, then, outlines the argument of the book. Let me offer some
advice about reading it. First, given the claims of the book, it was
impossible for me to turn away from the examination of the material
detail of mathematical praxis. I have tried to write the book so that
its arguments are intelligible over and above the mathematical analysis.
The mathematically uninitiated might do well to treat the symbols
19
INTRODUCTION

as untranslated hieroglyphics. The hieroglyphics should, however, be

inspected and, perhaps, by the end of the book, they will begin to take
on a fuller life.
A more serious drawback for the mathematically unsophisticated is
that I do not think that the descriptiveness and generalizability of the
book's findings and the faithfulness (or lack thereof) of my analyses
to actual mathematical praxis can be decided without reference to
and knowledge of the details of that praxis. Whether or not, in the end,
my research brings clarity to the problem of the foundations of math-
ematical practice, I hope that my book invites the reader into the
serious investigation of it.
One last bit of advice. My terminology and definitions (like those of
the opening pages) may strike the reader (particularly one unfamiliar
with ethnomethodological investigations) as being strange, difficult,
obfuscating, or the like. I have tried, on all occasions, to make my
terminology descriptively precise. Nevertheless, that precision, or lack
thereof, cannot be decided by ruminating over the definitions them-
selves. The reader is urged to read the definitions but not to dwell on
them, to read through the book, but not to dwell on the mathematics,
to use the book to find the phenomenon that I am attempting to
describe, and then to return to the book to see if I have described that
phenomenon accurately or not.

20
Part I
Introduction. The Phenomenon:
The Existence of Classical Studies of
Mathematicians' Work

The aim of this book is to examine the consequentiality of conventional

studies of the foundations of mathematics for investigations of the
natural accountability of mathematicians' work. In order to get the
point of that examination, some preliminary definitions must be given.
The book's motivating concern is to make the natural accountability
of mathematics' ordinary organizational objects capable of being
investigated as a real-world phenomenon. In particular, mathematical
proofs are going to be discussed as being organizational objects. The
reference to proofs will always be to the particular here and now
proofs that fill mathematical practice. It is to proofs as 'lived work,'
to proofs as they are tied to temporally developing and, then,
historicized and reviewed displays of written material. The reference
will never be to some conception of proofs in general.
The expression 'the natural accountability of a proof or 'the natural
accountability of a proof to its production cohort' will be used to
point to the availability of a proofs practical objectivity, of its anony-
mity as to authorship, of its transcendental character, of its rigor 1
as a witnessed and witnessible feature of the work of proving the
theorem, of the work of a proofs situated accomplishment. The notion
of natural accountability refers not only to the fact that proofs are
done for and done among mathematicians, but that they are done
recognizably as proofs; they are produced as the objects that they
recognizably are.
Reference to a proofs natural accountability as an essentially local
phenomenon is to the assertion that a proofs accountability consists
of the witnessible work of its production.
Last, the term 'classical foundational studies' will be used to refer
to studies of the natural accountability of mathematical proofs for
which the essentially local character of that accountability is irremedi-
ably hidden in and as the practices that make up the accountability
23
INTRODUCTION. THE PHENOMENON

of those studies themselves.

The point of this book is this: conventional studies of the founda-
tions of mathematics are, in fact, classical studies.

24
1 A Review of the Classical
Representation of Mathematicians'
Work as Formal Logistic Systems

The first task in coming to some understanding of the proposition that

conventional studies of the foundations of mathematics are classical
studies of mathematicians' work is to review what a formal logistic
system looks like and to examine whether or not, or in what sense,
such a system is descriptive of the material details of mathematical
practiCe. The 'theory of groups' will serve this comparison well.
Group theory should be particularly accessible to formal representation
and analysis because it is defmed syntactically as the consequences of a
given set of axioms. In contrast, a theory like the theory of functions
of a . real variable is defined in a partially semantic way, partly by
axioms and partly as just all things true about functions of a real
variable.
A group may be defined in the following way: Let G be a nonempty
set and let * be a mapping of G X G into G; that is, * associates a
unique element of G to each pair of elements in G. x * y will be
referred to as the product of x and y in G. G is a group if and only if
the following conditions hold:

(Gl) VX,y,z E G x * (y * z) = (x * y) * z (associative law)

(Read: 'for all x, y, and zinG, the product ofx with the product
of y and z is equal to the product of the product of x andy with z.')
Now, let .3 mean 'there exists.' The next axiom posits the existence
of an identity element:
(G2) 3e E G such that 'Vx E G e * x = x * e ,;:, x.
In words, G2 says that there is an element e of G such that the product
of e with any other element of G is always equal to that other element.
Before stating the third and last axiom in the definition of a group,
25
MATHEMATICIANS' WORK AS FORMAL LOGISTIC SYSTEMS

it is possible to prove the theorem - to be identified as Theorem T -

that the identity element of a group posited in axiom G2 is unique.
This is an example of a, perhaps, ultimately simple proof, and I will
use this proof as a means of developing and illustrating the contrast
between actual mathematical practice and the representation of that
practice within a logistic system. To prove the theorem, consider first
that to say that the identity element of a group is unique is to say that
if there were two identity elements, they would be equal. The proof
of the theorem is given in one line and goes like this: if there were
two identity elements e and e', then
e = e' * e = e'.
The last axiom in the definition of a group posits the existence of
'inverses':

(G3) Let e be the identity element posited in G2. Then Vx E G 3y E G

such that x * y = e = y * x.

In order to develop our comparison between actual mathematical

practice and the representation of that practice within a logistic system,
the rudiments of mathematical logic will first have to be developed.
The real aim of the underlying logistic system of a formal theory is
to provide an exact, syntactically specified notion of a 'formal proof.'
Reference to a formal proof (a proof within a formal system) will be
distinguished from references to the 'informal' proof of ordinary
mathematics by speaking of the former, as is sometimes done, as
'deductions.' Theorems of a formal theory, as opposed to theorems
of ordinary mathematics (including those about such a formal theory),
will be denoted as *theorems*. All the details cannot be given here;
the following discussion is only intended to give a flavor of what
is involved.
The immediate object of this presentation is to represent the proof
of Theorem T in a formalization G* of group theory. The first step
in developing G* is to specify its 'language' by providing a collection
of signs to serve as the primitive symbols of the language and by giving
exact formation rules for building up its syntactic objects - like
'variables,' 'terms,' and 'well-formed formulas (wffs)' - from these
primitive symbols through the operation of concatenation. In general,
the parameters of a language will change from theory to theory, as in
the case of G* where there is only one predicate letter .:1:., to be inter-
preted as* in G, in additioh to the two-place predicate for equality.
The underlying logistic system of a formal theory should contain
as *theorems* all wffs (well-formed formulas) that are 'tautologically
true.' This can be accomplished by the following three axioms:
26
 MATHEMATICIANS' WORK AS FORMAL LOGISTIC SYSTEMS

A::>.B::>A
A :::> [B :::> C] . :::> . [A :::> B] :::> [A :::> C]
~~A::>A

where A, B, and Care any wffs.

Additional axioms for first-order predicate calculus are
V'xA :::> S~A I
where S~A I indicates the process of 'substitution,' x is an individual
variable, and tis a term 'free for x in A,' and
V'x[A :>B):::> .A:::> V'xB
where x is an individual variable having no 'free' occurrences in A.
A two-place predicate, represented by the standard equality symbol,
and the axioms
V'x[x = x]
x = y :::>. A(x.x) =A(x,y)
where A is any wff obeying a certain restriction, introduces the notion
of equality into the logistic system.
Finally, the (non-logical) axioms for the formal theory of interest
(in this case, elementary group theory G*) are added to the logical
axioms. The reason for speaking of G* as being 'elementary' group
theory will be explained below. With ! as a three-place predicate
symbol defining the group operation in G*, the axioms for G* are the
following: 1
(G* 1) V'xV'y[3z(.:!:(x,y,z)) A
V'uV'v[:!:(x,y,u) A :!:(x,y,v) :::> u = v]}
(G*2) V'xV'yV'zV'sV'tV'u[~(x,y,r) A !(r,z,s) ·A
!(y,z,t) A !(x,t,u) :::> s = u]
(G*3) 3e[V'x[!(e,x,x) A !(x,e,x)] A
V'x3y[!(y,x,e) A !(x,y,e)].
G* 1 formalizes the notion that * is a function on G X G; G*2 corres-
ponds to the associative law G 1, and G*3 to both the existence of an
identity G2 and the existence of inverses G3. 2
The final component of our logistic system is the rules of inference.
For our system, we take these to be modus ponens,
from A andA ::>B, to infer B,
and generalization,

27
MATHEMATICIANS' WORK AS FORMAL LOGISTIC SYSTEMS

if x is an individual variable, from A to infer "xA.

Modus ponens is used in the following fashion: let the fact that A is a
*theorem* be written 1- A. Similarly, if A :J B (read: A 'implies' B)
is a *theorem*, write 1- A :J B. An application of modus ponens
allows the line 1- B to be written. In the vernacular, if A implies B
and A, then B. An application of generalization would allow the line
1- "xA to be written after the line 1- A.
At long last, the notion of a deduction (in G*) - the counterpart
of the notion of a proof (in group theory) - can be defined. A deduc-
tion of An is a finite sequence A 1 , ••• , An of wffs of the language
of a particular theory (G*) such that for each i, 1 ~ i ~ n, one of the
following three conditions holds:
(1) Ai is an axiom,
(2) for some r, s < i, Ar is of the form As :J Ai,
(i.e., Ai results from Ar and As by an application of modus
ponens), or
(3) Ai is of the form "xAs for some As, s < i
(i.e., Ai results from As by an application of generalization).
A wff A is a *theorem* (of G*) when A is the last wff of some deduc-
tion (in G*). As is customary, a turnstile r (rG*) will be used to
indicate that the formula following the turnstile is a *theorem* (a
*theorem* of G*).
The preceding presentation has glossed over many features of the
definition of a formal logistic system and formal theory. The material
in hand does allow the reader to get an idea of how the theorem T of
group theory that was proved earlier can be deduced as a *theorem*
Tin G*. The deduction that follows is, in fact, only an abbreviation
of a full derivation of T. The steps of a full deduction can be recovered
through the use of the justificatory comments that follow each line of
the argument. An actual deduction of T* in G* would be written
without the use of abbreviations, comments, or metatheorems. In
practice, as will become apparent, this is never done except in the
most elementary of cases.
The theorem T that a group has unique identity element may be
'translated' 3 into G* as the sentence T,
3e"e'["x[!(e',x,x) A !(x,e',x)] :J e = e'].
A 'deduction' of this sentence in G* follows:
LetA be the wffVx(.!(s,x,x)J\ .!(x,s,x))J\
"x3y(!(y,x,s) J\ !(x,y,s)), and let B be the wff
"x(!(e',x,x) 1\ !(x,e',x)). Then

28
 MATHEMATICIANS' WORK AS FORMAL LOGISTIC SYSTEMS

G*l; G*2;A;B I- !(e',s,s) (universal instantiation from

B; detachment)
G*l; G*2;A;B 1- !(e',s,e') (universal instantiation from
A; detachment)
G*l; G*2;A;B f-!!(e', s, s)A ~ (e', s, e') :::> s =
(universal instantiation twice in
G*l; detachment; universal
instantiation two more times)
G* 1 ; G*2; A; B 1- s = e' (modus ponens from last line
and conjunction of first two
lines)
G* 1; G*2 ;A 1- B :::> s = e' (deduction theorem)
G*l; G*2;A I- Ve'(B :::> s = e') (generalization on a deduction
from hypotheses)
G*l; G*2;A I- 3eVe'(B :::> e = e')
(existential generalization)
G*l; G*2; 3eA 1- 3eVe'(B :::> e = e!)
(existential instantiation, since s
does not occur in either G* 1,
G*2, or 3eVe'(B :::> e = e'))
Since 3eA is the same as G*3 and 3eVe'(B :::> e = e') is the same as T,
we have shown
rG* r.
In mathematics instruction, after the definition of a group is given
and some of a group's elementary4 properties are proved - as, for
example, the uniqueness of a group's identity element - the topic
that is ordinarily introduced next is that of subgroups. A 'subgroup'
is a subset of a group G that is itself a group with respect to the group
operation (*) already present in G. The attempt to represent the
study of subgroups in a formal system increases the complexity of
that formal system tremendously. Whereas, previously, 'quantification'
(Vx; 'for all x') was used only with respect to individual variables -
variables which range over the elements of a group in their intended
interpretation - the formal study of subgroups necessitates quantifi-
cation over predicate symbols (VA) as well as over individual variables.
It becomes necessary to express things like 'all subsets of a group with
a certain structure .. .' and 'if A is a subgroup of G, then ... ,' and a
first-order theory is not adequate to this task. Parenthetically, this is
the reason for referring to G* as elementary (that is, first-order) group
theory; it is a theory that is restricted to studying relationships between
29
MATHEMATICIANS' WORK AS FORMAL LOGISTIC SYSTEMS

the elements of a given group and not between subsets of a group.

The material presented so far offers a direct contrast between the
proof of a theorem in ordinary mathematics and the representation of a
proof (if feasible) in a formal system. Hopefully that contrast will lend
cogency to the following series of claims.
1 The representation of in-progress research within a formal system
is treated by mathematicians engaged in that research as being com-
pletely irrelevant to their research.
2 The translation of current work into the deductive schema of a
formal theory is never contemplated nor seen as being feasible.
3 The representation of work in a formal logistic system is never
carried out as a check on either one's own or on one's colleagues'
work.
4 Although the ability to represent mathematical reasoning in a
formal system (as, for example, axiomatic set theory) is spoken of as
the ultimate source of the legitimacy of the reasoning of a mathematical
argument, such a check is always proposed as an 'in principle' possibility
and is, in practice, completely infeasible.
5 The study of formal logistic systems are not found by mathemati-
cians to be illuminating of the natural technology of their situated
inquiries into mathematical structure. Such studies are completely un-
enlightening as to the origins of mathematical creativity and discovery.
6 The representation of an argument in a formal system is not, in
any ordinary sense, descriptive of the work of the original proof. On
the contrary, in an immediate sense, it is obfuscating of that work.
This last point can be elaborated through the example of the
proof e = e' * e = e'. In the first place, the attempt to formalize that
proof - to deduce the translation of the theorem T in G* - makes
the proof of that theorem problematic. Secondly, it makes problematic
the obvious thing being proposed by the single line e = e' * e = e',
tliat it is proposed in a single line, and that such economy is essential
to the obviousness of that proposal. The claim that needs to be advanced
is this: the line e = e' * e = e' is, in fact, read as a proposal; the proposal
can be articulated as the claim that 'e' * e' can be read in two specific
ways; when that line is read. as a proposal, that reading initiates a
'small-time' inquiry that is almost immediately satisfied in the way
that it asks that the axioms for a group be reviewed as to their 'mean-
in~'; that 'meaning' comes to reside in the seeable thing proposed about
'e * e'.
7 A ubiquitously recognized feature of mathematics instruction is
that the occasioned need to make such 'small-time' inquiries a theor-
etical enterprise turns the teaching of mathematics into a hopeless
enterprise.

30
2 An Introduction to Godel's
Incompleteness Theorems:
Their Metamathematical Interpretation
Contrasted with the Proposal to Study Their
Natural Accountability in and as the Lived-Work
of Their Proofs

The overall concern of this book is to examine the bearing of conven-

tional foundational studies on the investigation of the natural account-
ability of mathematicians' work. In the previous chapter, the formal
representation of mathematical practice was compared to that practice
itself, and it was argued that that representation was not, in any straight-
forward way, descriptive or illuminating of mathematicians' work.
However, the guiding inspiration of foundational studies is not that the
formal systems that are studied are literally descriptive of mathematical
practice, but that they represent the essential features of that practice,
and that by proving things about formal systems, things are actually
being proved about the essential, innate, or inherent structure of
mathematics itself. According to this account, the natural accountability
of mathematicians' work is but an intimation of the transcendental
character of mathematical reasoning.
In Part II of this book, the feasibility of this program of conventional
foundational studies will be addressed through an examination of a
proof of one of Godel's incompleteness theorems. The argument that
will be made is not that a proof of Godel's theorem does not prove
what others have claimed it to prove; instead, the origins of the rigor of
a proof of Godel's theorem will itself be examined and the claim
advanced that that rigor consists of its local work. Thus, this argument
points to the primordial character of the activity of doing mathematics
over some conception of mathematics-in-itself.
After completing this argument, we will be in a position to return to
our initial proposal that conventional foundational studies are, in fact,
classical studies.
The present chapter is intended mainly as an introduction to Godel's
incompleteness theorems, to their metamathematical interpretation,
and to the contrast between the interpreted content of Godel's theorems
and the natural accountability of proofs of Godel's theorems as rigorous
31
INTRODUCTION TO GODEL'S INCOMPLETENESS THEOREMS

proofs of just those theorems.

The theorem of Godel that I am going to discuss concerns any
'reasonable' 1 first-order axiomatization of number theory. My presen-
tation begins, however, with a particular axiomization of that theory
which will be referred to as elementary (Peano) arithmetic and denoted
by the letter P. Pis obtained from the logistic system of the last section
by removing the axioms of group theory, adding axioms for elementary
arithmetic, and adjusting the parameters of the language by eliminating
the three-place predicate symbol .! and adding a constant symbol 0,
a one-place function symbol S (for the successor operation), and the
two-place function symbols ± and ..!... For the parts of the proof of the
theorem that will be of interest here, an actual set of axioms need not
be given, and I refer the reader to conventional sources. A schema of
mathematical induction must be included in the axioms of P if the
theorem is to be 'embedded' in P itself.
As our terminology indicated, the *theorem* of G*, that was
deduced earlier was an example of a *theorem* 'proved' within a
formal system. In contrast to this, the theorem of Godel to be discussed
is a metatheorem; it is a theorem that concerns the structure of P as a
formal system. The theorem is not deduced in P, but is proved as a
theorem of ordinary mathematics about P.
In order· to state the theorem, the notion of the consistency of a
formal system must be introduced. P is said to be consistent if there is
no wff A of P such that both A and -A ('not'-A) are *theorems*
of P. The import of this notion is that if, for some wff A, both f-pA
and 1-p"'A then by the tautology -A :::> • A :::> B, all the wffs of P
would be *theorems*. If this were so, the claim that mathematics
concerns 'platonic objects' (in this case, number theory) would (in
the case of number theory) be shattered; number theory would have
no definite set of properties. Furthermore, the elemental character
of number theory would multiply this result throughout mathematical
practiCe. On the other hand, the proof of the consistency of P could
be interpreted as saying that ordinary mathematical reasoning does, in
fact, prove things about mathematical objects.
Although this line of reasoning may well have been at the basis of
their research, the formalists (referring principally to Hilbert and his
co-workers) at least claimed to understand mathematics as itself being
investigations within formal systems that, in practice, were treated and
studied in an informal manner. To the formalists, P made explicit what
was already implicit in ordinary mathematical investigations of number
theory. The proof of the consistency of P was itself taken as supplying
the absolute grounding of that theory and, therein, the adequacy of
ordinary mathematical reasoning about it.
The discussion of the philosophical consequencc:s of GOdel's
theorem will be facilitated by first speaking about the theorem that
32
 INTRODUCTION TO GODEL'S INCOMPLETENESS THEOREMS

Go del almost proved: if P is consistent, then there is a sentence J of P

such that neither J nor -J are deducible in P. By extending the pre-
viously noted generality of Godel's theorem to this one, the theorem
says that in any 'reasonable'· first-order axiomatization of number
theory (including, for example, P with J added as an additional axiom),
if that system is consistent, there will be some sentence S of that
system such that neither S nor -s are deducible in it. This is often
expressed by saying that 'number theory' is essentially incomplete.
Finally, if it were possible to supplement this theorem with the fact
that one of the sentences S or "'Sis intuitively true of number theory,
the theorem could be interpreted as saying that there is no adequate
formalization of number theory.
Let us contrast the theorem 2 that Godel actually did prove to the
one just mentioned. Godel's theorem says that there is a sentence J of
P such that (1) if Pis consistent, J is not deducible in P, and (2) if P
is w-consistent, then "'J is not deducible in P. This is the theorem
that will occupy us, and it will be referred to either as 'GOdel's first
theorem' or simply as 'Godel's theorem.'
w-consistency (the definition of which will not be needed for now)
is a stronger condition than consistency. Correspondingly, by weaken-
ing GOdel's theorem, one obtains the theorem that says that if P is
w-consistent, there is a sentence J such that neither J nor "'J is a
*theorem* of P. Since, in practice, number theory is assumed to be
both consistent and w-consistent, the philosophical impact of GOdel's
theorem is left unmarred. The thing that was seen to be so remarkable
about the. theorem was that it was proved by frrst indicating how such
a sentence J could be constructed and then by showing that the assump-
tion of either J or "'J led to a contradiction. Because Go del's 'formally
undecidable' sentence J is not deducible in P, it can also be seen to be
intuitively true. 3
The tremendous significance of this result for traditional studies
of the foundations of mathematics was eclipsed (at least philoso,P..hically)
by the significance of a consequence of just the first part ( 1) of Go del's
theorem, a consequence indicated by GOdel himself in the same paper
in which his first result appeared.
To state this result and briefly indicate its significance, the technique
of 'Godel numbering' will have to be introduced. As will be discussed
later, it is possible to assign 'Godel numbers' to each symbol, formula
and deduction in P. Letting g denote this assignment, m = g(M) would
be the GOdel number of the formula M. Now, let k(m) = Sffi(O) be the
numeral in P corresponding to the number m. Then k(g(M)) = k(m) =
Sffi(O), and k(g(M)) would be the numeral in P .corresponding to the
Go del number of the formula M.
Using a GOdel numbering, it is possible to 'arithmetize' Godel's
first theorem and prove this *theorem* as a deduction in P. To do this,

33
INTRODUCTION TO GODEL'S INCOMPLETENESS THEOREMS

a formula Consp is constructed in P which can be interpreted as saying

that Pis consistent, and, similarly, a formula -Th(k(j)) is constructed,
where j = g(J), and which can be interpreted as saying 'k(j) is not a
*theorem* of P' - kG) being the numeral representing the Godel
number of the sentence J which can be interpreted as saying 'I am not
a *theorem* ofP.'
Godel argued that the 'arithmetization' of the first part of his
theorem - Cons :::> -Th(k(j)) - is deducible in P. The implication
of this result is tfiis: If Cons~ is a *theorem* of P, then so is -Th(kG))
by modus ponens. Next, -Th(k(j)) can be shown to imply J. But
if J is a theorem of P, by the actual proof of GOdel's first theorem (and
as suggested by the interpretation of J itself); P can be shown to be
inconsistent. Thus, for P to be consistent, Consp cannot be deducible
in P. This is GOdel's second incompleteness theorem.
Generalizing the statement of this theorem, as with Godel's first
theorem, to any 'reasonable' axiomatization of number theory, Godel's
second theorem showed that if number theory was consistent, the
arithmetical statement of its consistency could not be deduced as a
*theorem* of number theory, or, in the vernacular, that one could not
prove the consistency of number theory in number theory itself. This
result can be generalized even further. In any formal system, like
axiomatic set theory, strong enough to defme the natural numbers
and have the axioms of P as *theorems*, a similar statement concerning
that system's consistency can be deduced. Since set theory is understood
to be strong enough to define all of mathematics, the consistency of
mathematics, via Godel's theorem, was interpreted as being unprovable
within mathematics itself. Thus, the transcendental character of math-
ematical reasoning - the disengagability of that reasoning from the
actual, temporally-situated, circumstantial, at-the-board or with-pencil-
and-paper work of doing mathematics - was found to be incapable
of mathematical demonstration.
I have developed the statement and interpretation of Godel's
theorems to the extent that I have in order to give the reader a sense
of the way in which those theorems (and the theorems of conventional
foundational studies in general) can be seen to be descriptive of math- -
ematical practice. An oddity of this interpretive descriptiveness is also
illustrated by Godel's work. On one hand, his work was tremendously
consequential for the development and practice of foundational studies;
on the other, it had no effect whatsoever on the way in which ordinary
mathematical investigations were actually conducted. Contrarily, the
mathematicalization of foundational studies was greatly accelerated.
This curious situation points to a paradox at the heart of the attempt
to interpret the results of conventional foundational studies in terms
of mathematical. practice. The paradox is this: the daily and identifying
work of professional mathematics consists, for mathematicians and
34
 INTRODUCTION TO GODEL'S INCOMPLETENESS THEOREMS

mathematical logicians alike, of the work of discovering and establishing

demonstrably rigorous mathematical arguments.
Part II of this book initiates the investigation of how, in fact, this is
so, through a descriptive review of a proof of Godel's first theorem
itself. The argument to be made is that the rigorous character of the
reasoning of that proof resides in the local work of the proof itself.
As the reader will recall, P is being used to designate a particular
axiomatization of elementary Peano arithmetic, and the theorem that
says 'there is a sentence J such that if P is consistent, J is not deducible
in P, and if P is w-consistent, then -J is not deducible in P' is being
referred to as Godel's first theorem.
I want to now outline the argument that I will make. In the proof
of his theorem, Godel introduced two major techniques of meta-
mathematical research: 'Godel numbering' and 'the double-
diagonalization procedure.' Together, these techniques give the work of
proving his theorem a directional character which is actualized as a
programmatically accomplished, intrinsically sequentialized series of
tasks. This way of proving will be called a 'structure of proving.' I begin
my analysis by introcfucing the notiQns of Godel numbering and double-
diagonalization and by indicating that they do provide for the proof
of GOdel's theorem as a programmatic series of tasks. A review of
some of the work of that program will then elucidate three major
claims: (1) the proof-specific structure of proving Godel's theorem is
made available in and as the local work of actually proving the theorem;
the structure of the work and the work itself are mutually identifying
as a course of action; (2) the naturally accountable objectivity, rigor
or truthfulness of that work -whether the proofs demonstrable clarity,
its flawed and inadequate character, or its evident erroneousness -
consists of that identification; and (3) that (1) and (2) are so points
to the organizational or 'social' character of mathematicians' work.

35
Part II A Descriptive Analysis of
the Work of Proving Godel's First
Incompleteness Theorem
3 Godel Numbering and Related
Topics:
Background Materials for a Proof of
Godel's Theorem

In this chapter, I give the mathematical definitions that are needed to

discuss Godel's theorem and indicate, in a preliminary fashion, how
these definitions are coordinated over the course of proving Godel's
theorem. Primitive recursive functions and relations, numeralwise
expressibility, and representability will be given exact definitions/
and a specific Godel numbering will be partially described and illustrated.
The meaning of these definitions in and as the detailed, in-situ work
of proving Godel's theorem will be developed in the chapters that
follow.
Although primitive recursion functions and relations are of great
importance to the proof of Godel's theorem, only the definitions of
them will be needed for now.
A numerical function is, for some m > 0, a function f : Nm -+ N,
where N stands for the natural numbers (i.e., 0, 1, 2, ... ; the numbers
of number theory). A numerical relation R is, for some m > 0, a subset
of Nm. The numerical functions Z (the zero function), S (the successor
function), and If (the projection functions)
Z(x)= 0
S(x) = x + 1

are referred to collectively as the initial functions. If h is an n-place

numerical function and g 1, ... , gn are m-place numerical functions,
then the m-place function f defined by
f(x1, ... , Xm) = h(gt (Xt, ... , Xm), ... , gn(Xt, ... , Xm))
is said to be obtained from g1, ... , gn and h by substitution. If g is
an m-place numerical function and h is an (m + 2)-place numerical
function, then the (m + I)-place function defined2 by
39
GODEL NUMBERING AND RELATED TOPICS

f(x 1, ... , Xm, 0) = g(x 1, ... , Xm)

f(x1, ... , Xm, S(y)) = h(x1, ... , Xm, y, f(x1, ... ,Xm, y))
is said to be obtained from g and h by primitive recursion.
The primitive recursive functions are inductively defined as only
those functions obtained from the initial functions by a finite number
of substitutions and primitive recursions. From the fact that this is an
inductive definition, it follows that if the initial functions are shown
to possess some property/ and if that property is shown to be preserved
by substitution and primitive recursion,4 then every primitive recursion
function possesses that property.
Let the characteristic function of an m-place numerical relation be
denoted by KR,
1 if(x 1, ... ,xm)ER
KR(XJ, ... ,Xm)= (
0 if (x1, ... , Xm) E Nm - R.
An m-place numerical relation is a primitive recursive relation if its
characteristic function is a primitive recursive function.
The next definitions that need to be introduced are those of
'numeralwise expressibility' and 'representability.' The idea of numeral-
wise expressibility is that if a relation W is numeralwise expressible,
then there is a formula W of P that 'defines' that relation in P.
Technically, an m-place numerical relation W is numeralwise expressible
in P if and only if there is a wff W(x 1 , . .. , Xm) of P with m free
variables such that the following conditions hold:
(1) if (a 1, ... , am) E W, then l-p W(k(aJ), ... , k(am))
(2) if (a 1, ... , am) ti. W, then l-p ~ W(k(a 1), ... , k(am))
where, as before, k(a) is the numeral ofP corresponding to the number a.
For a numerical function, the notion corresponding to the numeral-
wise expressibility is that of representability. An m-place numerical
function f is representable in P if and only if there is a wff F(x 1 , •.• ,
Xm, xm+ 1 ) of P with m+ 1 free variables which has the properties
that, for any natural numbers a 1, ... , am+l,
(1) if f(a 1, ... , am)= am+l, then l-p F(k(a 1 ), •.. , k(am), k(am+d),
and
(2) 1-p 3!xm+l F(k(ai), ... , k(am),Xm+ 1).
('3!x .. .'is read 'there exists a unique x such that .. .')
The idea of both numeralwise expressibility and representability
is that some formula of P 'performs a "role"' in P similar to the 'role'
performed by the original relation or function in number theory.
For the discussion of Godel's theorem the relationship between
representability and numeralwise expressibility that is needed is this:

40
 GODEL NUMBERING AND RELATED TOPICS

if R is an m-place numerical relation and if its characteristic function

K R is representable in P by KR (x 1 , . . • , Xm+l }, then R is numeralwise
expressible in P by KR(x 1 , ••• , Xm, S(0)). 5
Later, I will discuss the fact that all primitive recursive functions are
representable in P. From this fact, from the fact that primitive recursive
relations have, by definition, primitive recursive characteristic functions,
and from the results just mentioned it follows that primitive recursive
relations are numeralwise expressible in P.
In the next section, the numeralwise expressibility of primitive
recursive relations will be given greater emphasis than the representa-
bility of primitive recursive functions. Godel numbering permits
syntactic properties of P to be associated with numerical relations.
These relations can be shown to be primitive recursive; hence, there are
formulas of P corresponding to those relations that numeralwise express
them. In this manner, certain formulas of P can be seen to formulate
properties ofP's own syntax.
Later, in the discussion of 'the technical lemma,' the importance of
representability will come to the fore.
The last major definition to be introduced is that of Go del number-
ing itself. The predominant way of introducing this topic is simply to
give a specific assignment of numbers to the primitive symbols, expres-
sions,6 and finite sequences of expressions of P (that assignment being
defined as a Godel numbering) with the anticipation/foreknowledge
that the work of proving Godel's theorem will come to show that that
particular assignment fulfills the requisite properties of its intended use.
Typically, some of the 'computability' properties of that numbering are
explicitly mentioned. I begin my presentation in a similar manner by
illustrating a conventional type of Godel numbering. Afterward, I will
indicate the practical requirements on a Godel numbering for the proof
of Godel's theorem.
A Godel numbering is specified by first assigning numbers to the
primitive symbols, then, using these, to the expressions, and, then,
using these, to the sequences of expressions of P. Consider, for example,
the string of symbols (x"') which will be abbreviated x 3 and used
syntactically as an individual variable. Under the following assignment
of numbers
( 3
) 5
7
X 9
(x"') can be written as an ordered 6-tuple of numbers (3, 9, 7, 7, 7, 5).
Now consider the prime numbers arranged as an increasing series:
41
GODEL NUMBERING AND RELATED TOPICS

2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, ... '

or symbolically, p 1 , p2 , p 3, .... (p 0 will be defined as 1.) To the 6-
tuple (3, 9, 7, 7, 7, 5) associate the number p~ • p~ • p~ • pJ • p~ • p~
or 23 • 3 9 • 5 7 • 7 7 • 11 7 • 13 5 • This is the GOdel number g(x3) of
X3.
Next, for finite sequences of expressions, Godel numbers can be
assigned as follows: Consider a = x 4 , x 3 as a sequence of expressions
(and not the expression obtained by concatenating the symbols of x 4
with those of x 3). Then the GOdel number of a is 2g(x4) • 3g(x3).
As there is some ambiguity is using g to denote the Godel numbering
function, I will use g only for the Godel numbers of expressions, as
withg(x3).
Finally, .with this numbering, it is theoretically possible to work
backwards and determine if a given number is a Godel number and, if
so, what symbol, expression or sequence of expressions corresponds to
it. Given any number, one can, in principle, uniquely factor it as a
product of primes and, from this decomposition, reconstruct the
number's corresponding 'element' of P if it has one. Thus, a prime
factorization of g(x 3 ) would yield g(x 3 ) = 23 • 39 • 57 • 7 7 • 11 7 •
13 5 , from which (3, 9, 7, 7, 7, 5) and then (x"') could be recovered.
The major definitions that have been introduced so far - primitive
recursive functions and relations, numeralwise expressibility, represen-
tability, and Godel numbering - can be brought together in a list of
three related practical requirements on a Godel numbering. By 'practical
requirements,' I mean that a Godel numbering needs to fulfill these
requirements in order that a proof of Godel's theorem be carried out
with sufficient detail to exhibit that proof's rigorous character. 7
Let Lg(P) denote the set of all primitive symbols, expressions, and
finite sequences of expressions of P. A Godel numbering is an assign-
ment of numbers to the elements of Lg(P) that (minimally) obeys the
following restrictions:
( 1) That assignment is one-to-one - each element of Lg(P) is
assigned only one number, and each number has at most one corres-
ponding element in Lg(P). Technically such an assignment is an
'injection' from Lg(P) into N; informally, it is a 'renaming' of the
elements of Lg(P) as natural numbers. Under the association of
elements of Lg(P) with natural numbers established by a particular
Godel numbering, to speak about a specific number can be intuitively
interpreted as speaking about the element of Lg(P) that corresponds
to it, and to speak about an element of Lg(P) can be interpreted,
vice versa, as speaking about the number corresponding to it.
In the presence of such an assignment, let G(x,y) 8 denote the
numerical relation 'x is the number of a deduction of the wff which
results when k(y) is properly substituted for a specified individual
variable z in the wff with number y .'
42
 GODEL NUMBERING AND RELATED TOPICS

(2) The numerical relation G(x,y) defmed under the correspond-

ence established by a particular assignment of numbers to the elements ·
of Lg(P) can be shown to be primitive recursive.
(3) That assignment allows the 'proof' that G(x,y) is primitive
recursive to be accomplished by 'demonstrating' that G(x,y) is the
result of the construction of a chain of functions and relations, built
by substitution and primitive recursion, from the initial functions
and from previously established primitive recursive functions and
relations of that chain. These demonstrations are understood to repre-
sent the explicit display of a sequence of primitive recursive functions,
built · stepwise according to the definition of primitive recursive
functions, for either the functions themselves or for the character-
istic functions of the relations involved. Later, this notion will be made
precise as a 'formal proof and the explicit construction as a 'formal
construction sequence.'
The w.ork-related point of this restriction is that the practically
available rigor of the main body of work of proving Godel's theorem
'consists' of that sequentialized series of 'demonstrations.' 9
In order to give the philosophical significance of this restriction,
the following result must be stated: 10 the representability of primitive
recursive functions in P is proven by giving explicit wffs representing
the initial functions and by showing how explicit wffs representing a
function obtained by substitution or primitive recursion can be con-
structed from the explicit wffs representing the functions used in that
substitution or primitive recursion. The philosophical point of restric-
tion (3) is that, following from this method of proof, there is a pro-
jectable construction in P11 of a wffG(xh x~ corresponding to the
projecta:ble formal construction in number theory and, thus, in principle
the formally undecidable sentence of Godel's theorem is capable of
being constructed in P and materially exhibited as a formally un-
decidable sentence.
Despite their theoretical multiplicity, there is essentially only one
commonly used G6del numbering and only several explicitly described
types of such numberings in the literature. The reason for this is that a
Godel numbering must demonstrably fulfill the practical requirements
listed over the course of a proof of Godel's theorem. Theoretically,
that the construction of restriction (3) can be made is the heart of the
proof of Godel's theorem; in principle, most of the work of that
construction is trivial; practically, the construction is a matter of
extended labor and care. The presence of an explicitly described, proof-
specific Godel numbering in a proof of Go del's theorem is essential for
the practical objectivity of the temporally situated and developing,
detailed work that makes up the demonstrations of restriction (3).
The point of listing these three restrictions is this: the explicit
statement of these practical requirements - as opposed to concealing
43
GODEL NUMBERING AND RELATED TOPICS

them in an abstract definition of Godel numbering - begins to locate

the local work and the situated, discovered-in-course, material detail
of proving Godel's theorem.
A severe misconception will result if the detailed co-ordination
of definitions pointed to here is understood as a transcendental property
of the (platonic) objects or properties defined by them - it turns
Godel's discoveries into an heroic achievement, not a human one.
Rather than the various definitions fitting together as pieces of a
jigsaw puzzle, the definitions are, in a natural way, tied to the existence
of a Godel numbering. As we shall see, Godel numbering is not just a
numbering; it is a-numbering-as-a-technique-of-proving. In conjunction
with the diagonalization procedure/'proof to be discussed next, Godel-
numbering-as-a-technique-of-proving provides an endogenous organiza-
tion for the work of proving Godel's theorem. The proof-specific
Godel numbering (as a technique of proving) exhibits -over the course
of the work of proving Godel's theorem, as the accomplishment of that
work of proving - a technical compatibility with the practical techni-
ques of working with primitive recursive functions and relations. That
'compatibility' - as a name for the thing that the work of proving
Godel's theorem makes practically available - makes up the practical
objectivity of that work itself.

44
4 The Double-Diagonalization/
'Proof':
Features of the Closing Argument of a Proof of
Godel's Theorem as Lived-Work

The sequence of defmitions through which the formally undecidable

sentence of Godel's theorem is constructed will be referred to here 1 as
the 'double-diagonalization procedure' or as one or more of the possible
variants of this name.
In the previous chapter it was noted that the numerical relations
corresponding under a specific Godel numbering to certain syntactic
properties of Pare primitive recursive and, therefore, are numeralwise
expressible in P. In this way,2 formulas of P can be interpreted as
'speaking about' P itself. Heuristically, the double-diagonalization
procedure constructs a wff which 'says' 'I am not a *theorem* of
P.' The fundamental point of this construction is that the formula so
obtained has the syntactic property that neither it nor its negation are
deducible in P if P is consistent or w-consistent, respectively.
By the '"proof"' of Godel's theorem, I refer to the statement of
Godel's theorem and the argument following that statement (placed
after the diagonalization procedure) that the sentence constructed
through the diagonalization procedure has the syntactic properties
just mentioned. The 'proof speaks of one aspect of the practical
organization of the proof of Godel's theorem through which the
accountable structure of the work of that proof is exhibited. The
proof of Godel's theorem and the 'proof are two distinct, but over-
lapping sets of practices.
This chapter initiates the task of elucidating the endogenously
organized work of the naturally accountable, practically objective
proof of Godel's theorem. First, I recall for the reader the material
presentation of the diagonalization procedure and the 'proof. Some
summarizing comments are made about the work of constructing
the undecidable sentence in terms of the jointly articulating character
of the diagonalization procedure and the 'proof. The pairing of the
diagonalization procedure/'proof points to· the character of that pair
as a 'closing argument.'
45
THE DOUBLE-DIAGONALIZA TION/'PROOF'

For the 'proof that follows, we will need the definition of w-

consistency, and I begin by stating it. 3 P is said to be w-consistent
if for no wff F(x), both f-p 3xF(x) and f-p -F(k(O)), ... ,
f-p -F(k(n)), ... for all n EN. Intuitively, w-consistency means that
when any wff F(x) is interpreted in number theory, it cannot be
simultaneously false for all natural numbers and true for some natural
number. If P is w-consistent, then it is consistent or, equivalently, if
P is not consistent, it is also not w-consistent. This is so because if P
is Iiot consistent, all formulas ofP are deducible in P.
For the double-diagonalization procedure, we may proceed as
follows: Let subg~ )(X, y)4 be the Godel number of the wff which
results from 'properly' substituting the numeral k(x) for the individual
variable x 2 in the wff with GOdel number y. I.e., subg~ 2 ).(x, y) =
g(S~fxy41) where g(A) = y. Let «P(u) = subg~ )(u, u). Next, let
ded(x, y) denote the relation 'x is the Go del num~er of a deduction
in P of the wff with Godel number y.' 1/l(u) and ded(x, y) can be
shown to be a primitive recursive function and relation, respectively,
from which it follows 5 that the relation defmed by G(x, u) <>
ded(x, 1/l(u)) is also a primitive recursive relation. Let G(x 1, x 2 ) be a wff
numeralwise expressing Gin P.
Defme I as '"'"'3x 1 G(x 1, x 2 ) and let g(I) = i.
Defme J as -3x 1 G(x 1, k(i)) and let g(J) = j.
Finally, note that 1/l(i) = j, that is, that the Godel number of the
wff resulting from the substitution of k(i) for x 2 in the wff with
Godel number i (i.e.,/) is the GOdel number of J.
Godel's First Incompleteness Theorem (1) If P is consistent, J is
not deducible in P. (2) If Pis w-consistent, ":' J is not deducible in P.
Proof of (1): If J is deducible in P, that is, if l-p J or
(*) l-p- 3x1 G(x1, k(i)),
then, for some r, ded(r,j). Since
ded(r,j)~ ded(r, 1/l(i)) ~ G(r, i),
the numeralwise expressibility ofG gives
f-p G(k(r), k(i)).
Thus,
(**) f-p 3x1 G(x1, k(i)).
But (*) and (**) are contradictory, so if Pis consistent, J cannot be
deducible in P.
Proof of (2): Suppose - J is deducible in P. Then l-p "'J and
(***) f-p 3x 1 G(x 1 , k(i)),
46
 THE DOUBLE-DIAGONALIZA TION/'PROOF'

But if t-p "'], then ded(r, j) cannot hold for any r E N, since, other-
wise, we would have t-p J (by definition of ded(r, j) above), contra-
dicting regular consistency. But if ded(r, j) does not hold for any r E
N, G(r, i) ~ ded(r, ¢(i)) ~ ded(r, j) also does not hold for any r.
Using the numeralwise expressibility of G, one obtains
t-p "'G(k(l), k(i))

(****) j-p "-'G(k(n), k(i))

for all n E N. (***) and (****) contradict the assumption of w-

consistency. Thus, if Pis w-consistent, "'J is not deducible in P.
The following list of features of the diagonalization and 'proof
gives some indication of the objectifying work of their production.
However, the intention of this chapter is not to provide a detailed
analysis of that work, but to point to the fact that the diagonalization
and 'proof form an intrinsically related pair of objects and that together,
in terms of the larger structure of proving Godel's theorem, they
constitute an 'ending' for the proof of that theorem.
(I) The diagonalization procedure and 'proof do not form a pair
of 'a discovery'/'the proof but are the practically accountable, proof-
specific reconstruction of something previously 'known' to be capable
of such a reconstruction. 6
(2) A produced feature of the diagonalization procedure is its
detailed notational consistency. Consider the selection of the variable
for which the substitution of k(i) is made, indicated in the function
subg( )· In the diagonalization above, x 2 was used, making that selection
consistent with G(x 1 , x 2 ), the wff numeralwise expressing G(x, u)
in P. Thus, G(x 1 , x 2 ) could be quantified over x 1 , and x 2 was left as
the variable free for the substitution of k(i). Had x 1 been chosen as
the variable for substitution, the formula G(xn, xt) would have
appeared, calling forth an explanation of the ordering of the variables
in G and, perhaps, an explanation of the availability of x 1 as a free
variable in a wff numeralwise expressing G.
There are other ways of bringing about this notational consistency,
and I mention two of them. G and ded could be redefined by reversing
the order of the variables in their predicative notation, thus obtaining
G(u, x) ~ ded(¢(u), x). G(x 1 , x 2 ) could then be used to numeralwise
express G, x 2 would be the variable for quantification, and x 1 would
47
THE DOUBLE-DIAGONALIZA TION/'PROOF'

be used as the variable for the substitution. The benefit of this pro-
cedure is that the intention of the choice of the first individual variable
goes unnoticed, whereas the choice of the second individual variable
raises the question of why the second was selected and not the first.
Another technique of bringing about a consistent notation is to camou-
flage the problem through the use of meta-variables like x, y, and z.
(3) Associated with the problem of developing a consistent notation
is the problem of developing a notational system adequate to the cir-
cumstantial, proof-specific details of the diagonalization and 'proof.
Different authors develop different notations for the diagonalization,
and the two problems - that of developing a notational system and
that of developing a consistent notational system - arise together,
are a thematic concern of working out the material presentation of
the diagonalization and 'proof, and are solved simultaneously. The
notion of a 'consistent' notation is, in practice, equivalent to a notation
that, in its orderly presentation, provides sufficient analytic detail
for the diagonalization and 'proof.
( 4) The existence of a proof-specific notation for the diagonalization
procedure is essential to the diagonalization's adequacy as the sequence
of defmitions leading to the observation ¢(i) = j and the 'proof'.
(5) That the mathematician is able to disengage the practically
objective, material presentation of J from the real-time, real-world
work of producing an orderly, proof-specific, analytically adequate
notation is the accomplishment of that· work. In the presence of an
adverse arrangement of notation - discoverable in the course of devel-
oping the diagonalization and 'proof - the mathematician, as a matter
of familiar practice, immediately begins to reconstruct that notation
so as to exhibit its orderly, proof-specific adequacy. Only when the
work of such a reconstruction is repeatedly frustrated does the claimed
adequacy of the potential demonstration begin to be called into question.
( 6) The reconstruction of the diagonalization procedure and 'proof
mutually articulate each other, not just as fmished objects, but as
temporally developing, mutually elaborating constructions. A person
engaged in proving G6del's theorem discovers, over the course of
writing out the diagonalization and the first part of the 'proof, the
necessary technical details that permit their joint development. The
projected first part of the 'proof is consulted to elicit the necessary
structure of the diagonalization; the second half of the 'proof' is
recovered with the undecidable sentence in hand from the very way
that w-consistency is formulated to permit that 'proof.
Typically, the work of this mutual, temporally developing articulation
is summarized by the mathematician by speaking of the material
display of the diagonalization as the 'remembered thing' that it is then,
retrospectively, seen to be.
(7) The diagonalization procedure and 'proof are recovered together,
48
 THE DOUBLE-DIAGONALIZATION/'PROOF'

but as a pair of distinct objects. The work of the production of that

pair maintains, as an integral feature of that work, the distinction
between the diagonalization and the 'proof. The two are constructed
as the pair 'the diagonalization' /'the proof as the practically account-
able structure of the work of their construction.
(8) The diagonalization and 'proof are organized in such a way
as to evince 'a proper way of doing.'
Consider, for example, the placement of the crucial observation
that </l(i) = j. </l(i) = j is not proved, or, rather, is proved like the proof
e = e' * e = e' discussed in Chapter 1. </l(i) = j is a proposal that points
to the thing that if> can be understood as saying. The factual character
of </l(i) = j resides in and as its placement in the sequenqe of defmitions
preceding it, that placement making transparent the temporally extend-
ing reasoning behind the recovery of the meaning of </l(i). If the equation
</l(i) = j were placed in the 'proof' just before the point where it was
ftrst needed, that assertion would exhibit, instead of its factual character,
its status as requiring the work of its justification. 'One' would have
to, and would, 'go back' and review the sequence of defmitions of
the diagonalization to recover the reasoning that gives </l(i) = j its
factual character. In this way, the diagonalization/'proof would be
(mentally) reorganized into a proper ordering of topics.
(9) The 'proper' order of the items making up the diagonalization
and 'proof - and of a mathematical proof in general - are made
available through the work of their material presentation. The need
for certain results, notational corrections, abbreviations, restrictions,
and the like is discovered as a circumstantial feature of the thing that
needs to be done immediately.
At the blackboard, 't the time of the apparent disarray, the re-
organization of the work is immediately indicated, commonly by
boxing off the needed result with chalk lines from the rest of the work
on the board and by offering elaborating comments or by reviewing
the board so as to indicate via the notational apparatus available there,
how the appropriate emendations could be made. The produced objec-
tivity of both the need and the adequacy of the emendations and
corrections allows those emendations and corrections to be interpreted
as having come before the work that indicated their need.
The continual reorganization of the material displays of mathemati-
cians' work is an omnipresent feature of mathematical practice. The
immediate point is that (i} the organization of the work of proving
is an abiding concern in that the practical objectivity of that work
recognizably depends on its orderliness and (ii) the practically account-
able, analytically adequate orderliness of the work is a produced
feature of that work itself.
(1 0) The discoverable orderliness of the 'proof' as a natural course
of reasoning is tied to the way in which the diagonalization procedure
49
THE DOUBLE DIAGONALIZATION/'PROOF'

produces the sentence J. The diagonalization provides a pattern that

recurs in the 'proof, and, therein, immediately presages the 'proof
as an ordinary course of reasoning. J is constructed from G(Xt' x2),
which is itself obtained from ded arid cp through the use of the numeral-
wise expressibility of primitive recursive relations. In the 'proof, the
same type of construction sequence is used to prove the contradictory
character of asserting either J or - J. The assertion of either J or - J
is first translated into an assertion of assertions concerning the Godel
numbers i and j, ded, and cp. A theorem or theorems are then obtained
by numeralwise expressing those assertions in terms of G(x 1 , x 2 ).
( 11) The orderliness of the diagonalization procedure is not inherent
in that procedure, but is tied to the things that must be done with it
in the 'proof.
(12) The discovered orderliness of the 'proof makes up its practical
objectivity as a 'proof.' Each proposition of the 'proof appears from
within the work of recovering its reasoning as the natural, sequentially
next proposition, 'seeably' 7 factual in and as its placement in the
sequence, and, therein, making available the purported claim that the
chain of propositions proves the theorem.
( 13) What is being recovered in the mutual articulation of the diagon-
alization and 'proof are the material details through which the orderly
way of working through the diagonalization and 'proof is made
apparent. The work of recovering the diagonalization/'proof is the
work of formating 8 the diagonalization and proof so as to provide for
that orderly course of reasoning.
( 14) The existence of a formally undecidable sentence of P is funda-
mentally tied to the existence of a sentence of P adequate to the
methods of the 'proof .9 What the diagonalization procedure does is
to embed the proposition that there exists a sentence with certain
undecidability properties into the techniques of proving that have been
developed over the course of the work leading to the diagorialization
and 'proof.
(1 5) The structure of the diagonalization/'proof as a formated pair
is oriented to as part of the natural organization of the work of proving
Godel's theorem. Once enough apparatus is developed to permit the
construction of J, the 'hard-labour' of the proof is over. From within
the work of proving Godel's theorem, the diagonalization procedure/
'proof, as a forma ted pair, are oriented to and worked for as a 'closing
argument.'

so
5 A Technical Lemma:
A Lemma Used in the Proof of Godel's Theorem;
Its Origins as a Technical Residue of the Work of
Proving Godel's Theorem within that Self-Same
Work

The numeralwise expressibility of the relation G defined above is of

crucial importance for the diagonalization/'proof of Godel's theorem,
both for the construction of the sentence J and for the proof of that
sentence's peculiar syntactic properties. However, the proof that G is
numeralwise expressible in P emerges not as part of the diagonalization/
'proof, nor as an aspect of G particularly, but as a part of the larger
structure of proving Godel's theorem. As we shall see later, the proof
of Godel's theorem develops through the sequentialized demonstrations
of the primitive recursiveness of various numerical functions and
relations corresponding to syntactic features of P under a specific
Godel numbering. These demonstrations lead to the construction of G
as a primitive recursive relation. The numeralwise expressibility of G is
obtained by proving that all primitive recursive relations (and, hence,
G) are numeralwise expressible in P.
Although the fact that primitive recursive relations are numeralwise
expressible in P is, thus, of critical importance, the proof of this fact
has the character of being a 'technical residue' of the work of proving
Godel's theorem. The aim of this chapter is to indicate why this is so.
That it is so will allow us, in the chapters that follow, to give greater
attention to the places where the practically available rigor of the proof
of Go del's theorem actually lies.
To develop the argument of this chapter some background informa-
tion must be given. In his original paper, 1 Godel did not use the formal
system P, 2 but worked within a logistic system similar to that of
Whitehead and Russell's Principia Mathematica (PM). 3 There are two
immediate differences between PM and P: (1) PM has only one non-
logical 'constant,' indicated simply through predicative notation and
interpreted as saying, in the case of F(x ), that x has the property F,
whereas P has four constants 0, S, +, and·; and (2) PM has an infmite
number of predicate variables of all 'types,' whereas P has only individual
51
 A TECHNICAL LEMMA

variables, that is, variables of type 1. Furthennore, in PM, it is possible

to quantify over a variable of any type, pennitting the expression
'VF(F(x)=JV))to be a wffofPM. The strength of a logistic system like
PM is illustrated by the fact that equality, introduced in P through
explicit axioms, can be defined in PM; in PM, x = y is a notational
abbreviation for the wff VF(F(x) = F(y)). In that PM has axioms
of comprehension,4 all the properties of equality can be deduced from
this definition as *theorems* of PM.
For the discussion that follows, the Peano axioms must also be
introduced. 5 Briefly, the Peano axioms assert the existence of a set N
·and a function S, S: N ~ N, such that S is injective but not surjective
(an axiom of infinity) and such that if any subset L of N has the
property that L (/. S(N) and S(L) C L, then L = N (the induction
axiom). The collection of all true propositions following from these
axioms is referred to as Peano arithmetic. It can be shown that there is
a unique isomorphism between any two mathematical theories that
satisfy the Peano axioms, and together with a certain amount of infor-
mal set theory, these axioms are sufficient for the development of the
theory of natural numbers.
A formal statement of the induction axiom given above minimally
require~ a second-order logic (that is, a logistic system allowing quanti-
fication over predicate variables) and cannot be given in a first-order
system like P. Basically, P is an approximation of Peano arithmetic
in a first-order logistic system. In PM, on the· other hand, the Peano
axioms can be fonnulated and a set N and a relation Scan be defined
such that within iJ6 the Peano axioms hold. The pair (N, S) so defined
is referred to as an 'inner model' of Peano arithmetic in the logistic
system of Principia Mathematica. That PM was constructed so as to
have such an inner model, and the elaborateness of PM, was, in good
measure, determined by the attempt to produce such a model without
incorporating any potential contradictions into its construction.
Godel himself did not introduce Peano arithmetic as an inner model
in PM but, instead, superimposed a set of axioms for Peano arithmetic
on the basic logistic system of PM. This procedure insures the existence
of an infinity of 'individuals' within a specified type, an issue of elabor-
ate concern in the original fonnulation of PM. Godel also did not use
the ramified type theory of PM, but a theory of simple types alone.
The simplification thus introduced into PM made GOdel's formal
system - for the purposes of his incompleteness theorems - essentially
an axiomatic formulation of second-order Peano arithmetic that included
an axiom of comprehension. In the following, I will refer to Godel's
system - with some terminological inaccuracy - as (formal) Peano
arithmetic or PA.
For the proof that primitive recursive relations are numeralwise
expressible in PA, it is first necessary to show how to define wffs in PA
52
 A TECHNICAL LEMMA

that 'correspond' to the initial functions and that 'correspond' to

functions obtained by substitution or primitive recursion from numeri-
cal functions with previously established 'corresponding' wffs. The
notion of 'correspondence' is elaborated - and the proof of the
numeralwise expressibility essentially completed - by showing (1)
that the 'associated' wffs so defined act as functions in PA 7 and (2)
that if f(a 1 , . • • , am) = b, then nA F(k(at), ... , k(am), k(b)),
where F is the wff of PA 'corresponding' to f. 8 These conditions are,
of course, those of the (strong) 9 representabilityof a function in PA.
Let us suppose that 'corresponding' wffs of PA could be found for
the primitive recursive functions in the sense that the construction
of these wffs in PA mimic the construction of their associated functions
in number theory. Then (2) and the restriction of(l) to numerals (that
is, representability) would intuitively follow in that for any (m+ I)-
tuple of numerals (k(at), . . . , k(am). k(b)), the substitution of
these numerals in F would yield a variable-free formula 10 of PA and,
hence, a formula either deducible in PA or one whose negation is
deducible in PA. If rpA F(k(a 1), • . . , k(am), k(b)), then for any
other (m+l)-tuple (k(a 1)', . . . , k(am), k(c)}, c =F b, l-PA -F(k(a 1),
... , k(am), k(c)).
Continuing in this intuitive fashion, consider the construction of
the wffs to be associated with primitive recursive functions. The wff
S corresponding to S, either by construction in PM or by explicit
postulation in PA, acts as a function in PM or PA, respectively. The
wffs representing the projection functions and the constant function
Z are fairly obvious (xi = y and 'o'y-S(y, x), respectively), and
that they act as functions to be espected, 'o'y -S(y, x) by definition
and Xi = y by the properties of equality. Also easily defined is the wff
corresponding to a function obtained by substitution from numerical
functions previously shown to have corresponding wffs. The only
potential difficulty arises in the case of a function obtained by primitive
recursion. Yet, in informal set theory, the Peano axioms assure that a
function obtained by primitive recursion is well-defined. In that PA
and PM contain a formalization of the Peano axioms, the construction
of a wff 'corresponding' to a function obtained by primitive recursion
can mimic the definition of such a function in number theory, and the
proof that such a function is well-defined can similarly be translated
into PA or PM, implying that that wff acts as a function in PA or PM.
The preceding remarks are only heuristic, and what remains is the
work of materially proving that such 'corresponding' wffs can be
defined and that (1) and (2) hold for them. That material presentation
does, however, follow the heuristic sketch just given. Once the corres-
ponding wffs are found, they are seen to be 'natural' definitions; the
proofs of (1) and (2) follow by induction on the length11 of the con-
struction of a primitive recursive function in number theory; in both
53
A TECHNICAL LEMMA

cases, the proofs for the wffs corresponding to the initial functions
are elementary, and the proofs for the wff corresponding to a function
obtained by substitution require only slightly more ingenuity. In the
case of (1 ), the proof for the wff associated with a function obtained
by primitive recursion is technically difficult, but mimics the proof of
informal set theory that a function so obtained is well-defined, and,
finally, the proof of (2) primarily consists of checking to see that the
right formula holds for the right numerals. This synchrony between a
natural way of proving and the orderliness of the eventual proof as
the material realization of that way of proving makes up the straight-
forward character of the proof of the numeralwise expressibility of
primitive recursive relations in PA. G6del himself noted that the actual
proof was direct, if somewhat laborious, and gave only a brief indication
of it in his paper. 12
One last aspect of the proof of the numeralwise expressibility of
primitive recursive relations in PA needs to be introduced - that of its
'role' as a 'structure of accountable inference' in the larger proof of
Godel's theorem. As we shall see later, a proof-specific Godel number-
ing is essential to the proof of G6del's theorem in that the work of
demonstrating the primitive recursiveness of the needed numerical
functions and relations involves that numbering's explicit presence.
In contrast to this use of Godel numbering, the lemma of the numeral-
wise expressibility of primitive recursive relations and its proof are
detached from the detailed work of the material presentation of the
larger proof. Instead - as with the passage from the statement G(r, i)
to the *theorem* 1-pA G(k(r), k(i)) - the lemma provides only an
accountable structure for the writing of various lines of the diagonaliza-
tion and 'proof.' 13
The material presented so far can be summarized in three points.
First, the proof of Godel's theorem develops by demonstrating that the
numerical functions and relations corresponding under a specific Godel
numbering to syntactic features of PA have the structure of primitive
recursive functions and relations. That G is a primitive recursive relation
is the aim of those demonstrations, and that G is numeralwise expres-
sible in PAis a result of the fact that it is primitive recursive. Thus, the
question of the numeralwise expressibility of G emerges as part of the
larger structure of proof of Godel's theorem. Second, the proof of the
numeralwise expressibility of primitive recursive relations is straight-
forward in the sense that the intuitive, 'natural' way of proving that
proposition is realized as the orderliness of the proof. In that this is so,
the proof of that proposition becomes a technical exercise. And third,
the role of the proposition and its proof in the proof ofG6del'stheorem
is only to provide an accountable structure for writing various lines of
that proofand is not involved with the detailed, material presentation
of the proof of Godel's theorem itself.
54
 A TECHNICAL LEMMA

Once the proof of GOdel's theorem was given for PA (interpreted as

PM), the importance of that theorem was immediately established.
The proof of the theorem for the formal system P is essentially the
same as that for PA, except for the question of the numeralwise expres-
sibility of primitive recursive relations in P. Although the initial
functions and functions obtained by substitution should be (and are)
representable in P, P does not have the formal apparatus that PA has
to mimic definitions by primitive recursion. Thus, the problem of
proving the numer.alwise expressibility of primitive recursive relations
in P devolves on finding some way of translating the definition of a
function by primitive recursion in number theory into a formal con-
struction of a wff representing that function in P from the wffs
representing the functions used in that definition.
This problem can ·be made more explicit. Suppose that the (m + I)-
place function f is obtained by primitive recursion from the m-place
numerical function g and the (m + 2)-place numerical function h, and
furthermore, suppose that f(a 1 , ... , am, b) = c for some numbers
a1, ... , am, b and c. Then, from the defmition of primitive recursion,
there exists a sequence of numbers n 0 , ••• , Db = c such that
no = g(a1 , ... , am, 0)
n 1 = h(a 1 , ... , am, 0, n 0 )
n2 =h(a1, ... ,am, l,n 1)

Db-1 = h(a 1, ... , am, b -l,nb-2)

c =Db= h(a 1 , ... , am, b, Db-1).
If a function obtained by primitive recursion from functions represent-
able in P is itself to be representable in P, then it is necessary to show
(1) that there is a numerical function {j such that, for every finite
sequence of natural numbers n 0 , ••• , Di , ... , nb, there exists natural
numbers r:and s for which {j (r, s, i) = ni for 0 ~ i ~ b, and (2) that
there is a wff representing this function in P. Once this wff is obtained,
it can be used to define the values that the wffs representing g and h
must assume in order to construct a wff representing a function ob-
tained by primitive recursion. The work that remains is the technically
involved proof that the wff so obtained actually represents the original
function in P .14
That such a function {j exists is a known result of (ordinary) number
theory. However, suppose for a moment, that the definition of {j could
not be given in such a way so as to show that it is representable in P.
In the first place, that function certainly would be representable in
55
A TECHNICAL LEMMA

the system P with an additional function symbol and axioms for

exponentiation. 15 Secondly, if {3 were not representable in P - and,
by implication, if functions obtained from representable functions
by primitive recursion were not representable in P - then P might not,
retrospectively, be considered an adequate formalization of number
theory.
Let us now return to the theme of this chapter: that the proof of
the numeralwise expressibility of primitive recursive relations - and,
more particularly, that the proof of the numeralwise expressibility
of the relation G - is a 'technical residue' of the proof of Godel's
theorem. To speak of that proof as a 'technical residue' is to ~make
reference to the recognized coherence and structure of the work of
proving Godel's theorem from within that work. In the chapters that
follow, this endogenously produced orderliness of the proof of
Godel's theorem will be identified as a 'structure of proving.' Antici-
pating the description of this 'structure of proving,' the material in this
chapter indicates why, as part of that 'structure of proving,' the proof
of the numeralwise expressibility of primitive recursive relations has the
character of being a 'technical residue.' In the way that the problem
arises and is formulated within the proof of Godel's theorem, in the
way that it consists of working out the technical details of its argument,
in that its proof is anticipated in PM, in that its proof is an indication
of the adequacy of P as a formalization of number theory, and in that
its 'role' in the proofofG6del's theorem is not intrinsic to the material
presentation of the proof, but only supplies an accountable structure
for certain parts of that presentation all point to the intuitive accuracy
of the formulation of that proof as a 'technical residue.'

56
6 Primitive Recursive Functions
and Relations:
An Initial Discussion of the Irremediable
Connection between a Prover's Use of the
Abbreviatory Practices/Practical Techniques of
Working with Primitive Recursive Functions and
Relations and the Natural Accountability of a
Proof of Godel's Theorem

So far, two organizational components of the work of proving Go del's

theorem have been identified: an ending argument consisting of the
construction of the diagonalization/'proof and the proof of a 'technical
lemma' asserting the numeralwise expressibility of primitive recursive
relations in P. The major remaining task of the following chapters is
that of disclosing the orderliness and rigor of the proof of Godel's
theorem as the local work of their joint accomplishment. In order to
do this, one more aspect of that proof needs to be examined - namely
the abbreviatory practices used in working with primitive recursive
functions and relations. The present chapter reviews these practices
and gives a brief indication of their role in the proof of Go del's theorem.
The reader will recall that primitive recursive functions are defined
inductively as only those functions obtained by a finite number of
substitutions
f(xl, · · · , Xrn) = h(g1 (xl, · · · , Xrn), · .. , gn(Xl, · · · , Xrn))
and primitive recursions
f(x1, ... , Xrn, 0) = g(x1, ... , Xrn)
f(x 1 , ... , Xrn, S(y)) = h(x 1 , ... , Xrn, y, f(x 1 , ... , Xrn, y))
from the initial functions: the zero function, the successor function,
and the projection functions -
Z(x)= 0,
S(x) = x + 1, and
If(x 1, ... , Xj, ... , Xrn) = Xj,
respectively. The reader will also recall that a primitive recursive relation
is a numerical relation R whose characteristic function KR is a primitive
57
PRIMITIVE RECURSIVE FUNCTIONS AND RELATIONS

recursive function. KR serves to 'identify' the members of R; KR can

assume only the values 0 and 1, and if KR(x1 , ••• , Xm) = 0, then
(x~o ... ' Xm) e R, and if KR(Xl' ... ' Xm) = 1, then (xl' ... ' Xm)
ER.
For the discussion that follows, several more definitions will be
needed. First, a formal construction sequence for a primitive recursive
function f is defined as a finite sequence of functions f 0 , ••• , fn with
the properties that (1) for every 0 ~ i ~ n, fi is either an initial function
or is obtained from the functions of the collection {f0 , ••• , fi-ll by
substitution or primitive recursion and (2) f = fn. The definition of
primitive recursive functions guarantees that every primitive recursive
function has at least one (and hence, many) associated formal construc-
tion sequences. A formal proof that a given numerical function f is a
primitive recursive function consists of the explicit display of a formal
construction sequence and the demonstration that f satisfies the defin-
ing equation(s) of the last function of that sequence. Finally, the
expression 'seen/shown' will be used to refer to the fact that what is
seeably true about a mathematical object is seeably true in that that
'seeing' is embedded in the practical techniques of making the thing
seeably true, accountably so.
As has already been pointed out, the body of the proof of GOdel's
theorem consists of a series of propositions and proofs that certain
functions and relations are primitive recursive. The proofs of that
series are never formal proofs in the sense just defined; instead, they
are understood by mathematicians as representing such formal proofs
through the use of various abbreviatory practices. In what follows,
I will briefly examine some of the features of these informal proofs
and the abbreviatory practices associated with them.
In the first place, explicit construction sequences for functions are
never actually displayed. The informal proof that a function or relation
is primitive recursive is placed within a sequentialized series of similarly
informal demonstrations, and the intervening steps between the previous
construction and the demonstration in question are assumed to be
easily recovered from the formula(s) defining the new function or
relation. For example, once addition and an operation like 'limited
subtraction,' symbolized by .....,
x~y={x-yify~x
0 ifx<y
have been shown to be primitive recursive, the material proof that the
equality relation on N2 is primitive recursive is simply the equation
K= (x, y) = 1 ~ ((x ~ y) + (y ~ x)). The adequacy of this proof
consists of the work of finding the seeable/showable structure of K=
as the result of a series of substitutions and of computing the value of
K= when x = y and when x < y. 1
58
 PRIMITIVE RECURSIVE FUNCTIONS AND RELATIONS

The example of K= illustrates a second abbreviatory practice used

in working with primitive recursive functions and relations: the check
or argument that the equation(s) demonstrating the primitive recursive-
ness of the function or relation in question does define that function or
relation is not given, but is partially left to the reader and partially
understood to be fulfilled by the construction of the equation(s)
itself. Thus, the need for the calculation indicated above - that the
equation K=(x, y) = 1 when x = y and= 0 when x =F y- is left to
the reader, and in the .case of the proof that addition is a primitive
recursive function, the equations
x+O=x
x + S(y) = S(x + y)
are either assumed to be recognized as the recursive definition of
addition or assumed to be understood in the sense that since ordinary
addition satisfies these equations, and since the equations define a
unique function, that function must be ordinary addition. As in the
example of K=, the calculations that are needed are generally obvious,
but they are of interest here in that once 'one' writes the formula
for L (or for term(x) below), 'one' immediately checks the formula
against the criteria used in obtaining it, and the exclusion of that work
from the material presentation of the informal proof is part of the work
of trivializing the omnipresence of such local, situationally-specific
work as constituents of the practically adequate formula itself.
A third abbreviatory practice concerns the conventions for naming
functions and relations. In the case of K=, rather than specifically
naming the numerical function (x, y) i-+ (x ...._ y) + ( y _._ x) and the
constant function SZ, the function K= is identified through its value
at (x, y) - K= (x, y) = 1 ...._ ((x ...._ y) + (y ..._ x)). The numerical function
of addition is identified simply as x + y; multiplication, as xy; 'limited
subtraction,' as x ..... y; and so forth. The initial functions Z, S, and Ifl
are written as 0, x + 1, and Xi, respectively.
Fourth, the material" demonstration of the primitive recursiveness
of a function is never an explicit display of a definition by substitution
or primitive recursion. Instead, unspecified, informal ways of writing
epitomizing demonstrations are employed. In the elementary case of
addition, the recursive equations
x+O=x
(*)
x + S(y) = S(x + y)
are 'understood' to stand proxy for the primitive recursive equations
that can be constructed from them as
1/(x, 0) = I~ (x)
59
PRIMITIVE RECURSIVE FUNCTIONS AND RELATIONS

11(x, S(y)) = (s(ln) (x, y, 77(x, y))

which, in turn, seeably/showably compute the same values as (*)and,
hence, define addition of natural numbers - 7] = +. In the case of
primitive recursion relations, equations for their characteristic functions
- except in a few elementary cases like equality - are not given.
Instead, various ways of constructing primitive recursive relations from
previously established primitive recursive functions and relations are
used to establish the primitive recursiveness of the relation in question.
Finally, the example just given points to a fifth ab brevia tory practice:
the use of theorems showing that if certain functions and relations
are primitive recursive, then a particular way of defining a function or
relation from them will also be primitive recursive. Constructing a
formula in such a way that it seeably/showably fits the template
provided by such a theorem constitutes an adequate demonstration
that that function or relation is itself primitive recursive. An elementary
example of such a theorem is the proposition that if S and T are m-
place primitive recursive relations, then 'S and T' will be one also.
(Proof: 2 Ks and T(x 1 , ... , Xm) = Ks(x 1 , ... , Xm) • KT(xl, ... ,
Xm ).) The logical operations of 'or,' 'not,' and 'implies,' applied to
primitive recursive relations, also provide primitive recursive relations.
The following three propositions illustrate how these abbreviatory
practices fit together, supporting and elaborating their joint use. 3
(1) If T is an (m + 1)-place primitive recursive relation, then so is
'3y ~ z T(x 1 , ... , Xm, y)' where (a 1 , ... , am, b) E '3y ~ z
T(a 1 , ... , am, y)' if and only if 3y ~ b T(a 1 , ... , am, y). (Proof
omitted.)
(2) If g1, ... , gy and h are m-place primitive recursive functions
and T 1, ... , Tv are disjoint m-place primitive recursive relations, then
the function defined by
g1(x 1 , ... , Xm) ifT 1(x 1 , ... , Xm)

gv(x 1 , ... ,xm)ifTv(xl, ... ,xm)

h(x 1 , ... , Xm) otherwise
is also primitive recursive. Proof: f(x 1 , ... , Xm) =
KT, (xl, ... , Xm) • gl (xi, ... , Xm) + ... + KTv(x 1, ... , Xm) •
gv(xl, ·. ·, Xm) + (1 ..._(KT, (xl, ... , Xm)+ ... + KTv(x 1 , ... , Xm))) •
h(xl' ... ' Xm).
(3) Define JJY ~ z T(x 1 , ... , Xm, y) as the least y ~ z for which
T(x1, ... , Xm, y) holds if there exists such a y and if not, defme
60
 PRIMITIVE RECURSIVE FUNCTIONS AND RELATIONS

p.y ~ z T(x 1 , ... , Xm, y) = 0. If T is an (m +I)-place primitive

recursive relation, then the function defined by (x 1, ... , Xm, z)
~-+ p.y ~ z T(x 1 , ... , Xm, y) is primitive recursive. Proof:

(*) JlY ~ 0 T(x1, ... , Xm, y) = 0

JlY ~ S(z) T(x 1, ... , Xm, y) p.y ~ z T(x1, ... , Xm, y) if

3y ~ z T(x 1, ... , Xm, y)

z + 1 ifT(x 1, ... , Xm, S(z)) and
~(3y~zT(x 1 , ... ,xm,y))
0 otherwise.
(Partial exegesis: First note that (x 1, ... , Xm, z) E '3y ~ w T(x 1,
... , Xm, y)' has been replaced by 3y ~ z T(x 1, ... , Xm, y) and
~(3y ~ z T(x 1 , ... , Xm, y)) has replaced (x 1 , ... ,Xm, z) E not-
'3y ~ w T(x 1, ... , Xm, y).' The proof is by recursion; (*) evaluates
the function at (x 1, ... , Xm, 0), and the right side of the remaining
equation defines a primitive recursive function by cases, following from
Proposition (2) above.)
The last example that I will give is the proof that the relation term(x)
is primitive recursive, where term(x) denotes the set of numbers corres-
ponding under the Godel numbering to the terms of P. As with the
examples just given, this example illustrates how the various abbrevia-
tory practices are used in conjunction with each other. However, there
is an important difference between the previous examples and that of
term(x). The demonstrations that numerical functions like addition,
multiplication, and exponentiation are primitive recursive functions
and that numerical relations like 'x divides y>4 and 'x is a prime number'
are primitive recursive relations can be5 understood as definitions of
addition, multiplication, etc. In contrast, the syntax of P is specified
independently of any consideration of its eventual translation into
numerical functions and relations via a Godel numbering. Consequently,
term(x) is, a priori, just a collection of numbers obtained through the
correspondence of Lg(P) 6 to N under a specific Godel numbering, and
the demonstration that term(x) is primitive recursive is the demonstra-
tion that the characteristic function of this (a priori) 'unstructured'•
set of numbers is a primitive recursive function.
In order to present such a demonstration, a precise definition of
the terms of P must be given. The terms of Pare defined inductively as
only those expressions of P determined by the following conditions:
(1) 0 is a term; (2) individual variables (i.e., x 0 , x 1, x 2 , •••) are terms;
(3) if v is a term, then so is S(v); and (4) if v1 and v2 are terms, then
+(v 1 v2 ) and ·(v 1 v2 ) are terms, also. 7 A term can be thought of as
potentially naming a natural number in P; S(O) 'names' 1 in P, S(l')
61
PRIMITIVE RECURSIVE FUNCTIONS AND RELATIONS

'names' the successor of v. The other definitions that will be needed

are those of L(x), (x)i, x * y, and var(x). L(x) counts the number of
primes in the prime factorization of x; (x)i gives the exponent of the
i-th prime in the prime factorization of x; and x * y is the function that
'concatenates' the prime factorizations of x and y. All three of these
functions are primitive recursive. var(x) is the primitive recursive
relation that holds if and only if x is the Godel number of an individual
variable of P.
A materially adequate demonstration that the set of Go del numbers
corresponding to the terms of P is a primitive recursive relation can now
be given, and it consists of the formula 8

term(x) ~ x * 0 and 3y :s;;; i~ pf

L(x) {
[(y)L(y) = x] and

Vi :s;;; L(y) ([i = 0] or [(Y)i = g(O)] or

[var((y)i)] or 3j < i [(y)i = g(S() * (y)j * g())]
or 3j < i 3k < i ([(y)i = g(+() * (y)j * (y)k * g( ))]
or [(y)i = g(() * (y)j * (Y)k * g( ))]))) ·
Given its placement in the course of work that makes up the proving
of Godel's theorem, this formula seeably /showably defines a primitive
recursive relation. In that the formula makes available a methodic
procedure for assigning to each and only the terms of P a number
that satisfies it, the adequacy of the formula as exactly defining the
Godel numbers of the terms can be checked by 'reasoned' calculations
comparing the formula to that numbering and to various examples of
terms constructed as 'typical' cases. Finally, that the bound on y (i.e.,
L(x)
n pf is large enough is established through considerations of the
i::;:Q
structure of the Godel numbering as that structure becomes available
over the course of developing both the formula and the bound. Thus,
together, these observations point to the in-situ determination of the
adequacy of the formula defining term(x) as a primitive recursive
relation as being tied to the local practices that make up the work of
writing and inspecting that formula.
With these examples before us, it is possible to clarify certain aspects
of the abbreviatory practices used in working with primitive recursive
functions and relations. By doing so, we will be led to the pointed
relevance of these practices for the proof of Go del's theorem.
1 The use and omnipresence of these practices are almost trans-
parent. The abbreviatory practices are introduced gradually, coincident
with the increasingly extended constructions of primitive recursive
functions and relations, and when they are introduced, they have the
62
 PRIMITIVE RECURSIVE FUNCTIONS AND RELATIONS

initial character of being either notational conventions or theorems of

ordinary mathematics. The cumulative effect of these practices as a
body of learned, practical techniques for working with primitive
recursive functions and relations makes its appearance only when the
presentation of a proof of primitive recursiveness - like that of term(x)
above - is disengaged from the sequence of such propositions and
proofs that was organized to include that proof at its particular place
in that sequence. For the reader of· the proof of Godel's theorem,
the proofs of the series of propositions and proofs concerning the
construction of the necessary primitive recursive functions and relations
are understood practically not as self-sufficent demonstrations, but as
guides and summaries of the work that, in the end, finds the adequacy
of them. The work· of finding the practical accountability of that
sequence of propositions and proofs serves to instruct the reader in
the abbreviatory practices that are being used. As we shall see, for
the person engaged in proving Godel's theorem, the abbreviatory
practices are essential to constructing the practically objective proof
that a function or relation is primitive recursive. In either case, the
transparency of those practices turns out to consist of the work of
producing the practically objective proofs of that primitive recursiveness.
2 From within the work of their proofs, the informal demonstra-
tions that certain functions and relations are primitive recursive are
not viewed as standing proxy for a formal rendering of them; they are
the practically adequate demonstrations of the propositions for which
they stand as proofs. That they can, later, be considered as representing
formal proofs is essentially tied to the ways in which the abbreviatory
practices are developed over the course of increasingly extended con-
structions of primitive recursive functions and rela,tions and to the ways
in which the informal proofs can be given increasing detail as approxi-
mations of formal ones.
3 Although introduced and reflectively considered as 'abbreviatory
practices,' these practices, in fact, make up the learned, practical
techniques of working with primitive recursive functions and relations.
The point is not that formal proofs cannot be given nor that these
practices are not adequate to the construction of such formal proofs.
The point is that the practical objectivity and accountability of the
informal demonstrations does not depend on giving such formal proofs,
but is available in and as the practical techniques of working with
primitive recursive functions and relations.
The existence of these abbreviatory practices and their connection
with the practical techniques of working with primitive recursive
functions and relations leads to the major point of the immediate
discussion. The real problem in working with primitive recursive jUnc-
tions and relations, at least for the proof ofGodel's theorem, is that of
accountably showing that each of a sequentialized series of jUnctions
63
PRIMITIVE RECURSIVE FUNCTIONS AND RELATIONS

and relations is, in fact, primitive recursive. The way that this is done is
(1) by writing - where 'writing' is a trivialization of the work of writing
at the proper place in a sequentialized series of such writings - a
practically objective formula or formulas for a function or relation in
terms of known primitive recursive functions and relations and in terms
of known ways of building primitive recursive functions and relations
that seeably/showably insures the primitive recursiveness of it, and (2)
by checking in a locally j:letermined, practically adequate manner that
the formula or formulas do compute the correct values, where the
calculations themselves are 'reasoned' procedures, practically adequate
to the task at hand.
Consider for a moment a formal construction sequence for ordinary
multiplication of natural numbers- n,I~, S, s(ID.
+(x, 0) = IHx) }
{
+(x, S(y)) = (S(I~)) (x, y, +(x, y)) , I~, +(I~, ID, Z,
·(x, 0) = Z(x) }
{
•(x, S(y)) =(+(I~, ID) (x, y, ·(x, y)) -and compare it with
the 'abbreviated proof
X • 0= 0
x • S(y) = xy + x.
The formal primitive recursive equations do not make apparent what
ordinary numerical function they define, and it would be difficult
to find the formal primitive recursive equations without having recourse
to the abbreviated proof. The difficulty that is illustrated here in the
case of multiplication becomes severe as the functions and relations
involved become more complicated. In that the adequacy of an equation
defining a primitive recursive function or relation as a description of a
pre-given function or relation is made available through the work of
writing and inspecting that formula as seeably/showably adequate
to that claim, the point to be made is this: theabbreviatory practices/
practical techniques make that writing and inspection possible. It is
this aspect of the abbreviatory practices that is crucial to the practical
objectivity - to the 'rigor' - of the work of proving Godel's theorem.

64
7 A Schedule of Proofs:
An Extended Analysis of the Lived-Work of
Producing the Body of a Proof of Godel's Theorem

A
A Schedule of Proofs

The body of the proof of GOdel's theorem consists of a schedule of

proofs that certain functions and relations are primitive recursive and
that certain ways of building functions and relations from other func-
tions and relations preserve primitive recursiveness, that schedule
leading to, and organized so as to permit, the construction ofG in the
diagonalization/'proof as a primitive recursive relation~ The reader will
recall that the wff G(x 1 , x 2 ) numeralwise expressing Gin P was used
to construct the formally undecidable sentence J and that the numeral-
wise expressibility ofG was used in both parts ofthe 'proof to transfer
facts about membefship in G (e.g., G(n, i)) to *theorems* of P (e.g.,
1-p G(k(n), k(i)). Once the schedule of proofs has been completed
and the theorem asserting the numeralwise expressibility of primitive
recursive relations in P has been established, the diagonalization/
'proof can be constructed, bringing to an end the proof of Godel's
theorem.
In this section, I recall for the reader the general pattern of ordering
of such a schedule of proofs 1 and then describe some of the general
features of that schedule of proofs of the lived work of its production.
By a 'schedule of proofs,' I refer both to the ordered sequence of
proportions of that schedule and to the proofs of those propositions,
the latter generally consisting of a single formula. I have omitted these
proofs from the outline that follows. A few marginal comments have
been added to indicate the schedule's topical organization.
1 The constant functions Zn(x) = n, n = 0, 1, 2, ... are primitive
recursive.
2 The functions obtained by permuting variables, identifying variables,
adding dummy variables, and substituting constants for variables in

65
A SCHEDULE OF PROOFS

primitive recursive functions are primitive recursive.

3 Addition, multiplication, exponentiation, and 'limited subtraction'
are primitive recursive functions.
4 Let f be an (m + I )-place numerical function and (informally) let
y y
~ f(xl, ... , Xm, k) and ll f(xl, ... , Xm, k) denote the
k=O k~O
functions (x 1 , ... , Xm) >-+ f(x 1 , ... , Xm, 0) + ... + f(x 1 , ..• ,
Xm, y) and (x 1 , ... , Xm, y) ~ f(x 1 , ... , Xm, 0) • ... • f(x1,
•.. , Xm, y), respectively. Then iff is a primitive recursive function,
y y
SOare ~ f(xl, ... , Xm, k) and ll f(xl, ... , Xm, k).
k=O k=O
5 =, -=1=, <,and~ are primitive recursive relations.
6 The logical operations of'not,' 'and,' 'or,' 'implies,' and 'if and only
if,' applied to primitive recursive relations, produce primitive
recursive relations.
7 The relation obtained by substituting a primitive recursive function
for a variable in a primitive recursive relation is primitive recursive.
8 Let R be an (m + I)-place numerical relation and let 'Vz ~ y R(xt.
... , Xm, z) and 3z ~ y R(xh ... , Xm, z) denote the (m +I)-
place relations that hold for (x 1 , ... , Xm, y) ifR(xh ... , Xm, z)
holds for all z ~ y or for some z ~ y, respectively. If R is a primitive
recursive relation, then so are 'Vz ~ y R(xh ... , Xm, z) and 3z ~
y R(xh ... , Xm, z). .
9 If T 1, ... , Tv are pairwise disjoint, m-place primitive recursive
relations, and if g 1 , ... , gy and hare m-place primitive recursive
functions, then the function f defined by the equation

f(x 1 , . . . , Xm) =

for all i, I ~ i ~ v.

is primitive recursive.
IO If R is an (m + I)-place numerical relation, let /J.Z ~ y R(x 1 , ... ,
Xm, z)(informally) denote the function of(x 1 , ... , Xm, y) defined
by the equation
66
 A SCHEDULE OF PROOFS

the least z ~ y such that R(x 1 , ••• ,

xm, z) if there is such a z
{
JJ.Z ~ y R(x 1 , .•• , xm, z) =
0 otherwise.
If R is an (m + 1)-place primitive recursive relation, then
JJ.Z ~ y R(x 1 , •.• , Xm, z) is a primitive recursive function.
The preceding propositions provide the background facts and techniques
that are needed for the proofs of the propositions that follow. The
next group of propositions supply the apparatus for working with the
G6del numbering that was introduced earlier.
11 The 2-place numerical relation xly ('x divides y with no remainder')
is primitive recursive.
12 Let prime(x) hold if and only if xis a prime number. Then prime(x)
is a primitive recursive relation.
13 The function Pn (i.e., n ~ Pn) giving, for each n, the n-th prime
number is primitive recursive (p 0 := 1).
14 The 'decoding function' (x)n (i.e.,(x,n) ~-+(x)n) giving the exponent
of Pn in the prime factorization of xis a primitive recursive function. 2
((x)0 := 0; (O)n := 0.)
15 Let the function L(x) give the number (in the serial ordering of the
prime numbers) of the largest prime with a non-zero exponent in
the prime factorization of x or give 0 if xis 0 or 1.3 L(x) is a
primitive recursive function. (y)
16 For every natural number y, y = 0 or y = p 0 (y)• p?'), · ... · PL(y)(y)
Define (informally) x * y as the function mapping (x, y) to the
value
L(y)
X* y =X • fi p(Y)k
L(x)+k ·
k=O
x * .Y is a primitive recursive function.
The remaining propositions assert the primitive recursiveness of the
functions and relations that correspond under the previously specified
Godel numbering to certain syntactic categories and operations of the
formal system P.
17 The mapping of a natural number n to the G6del number of the
individual variable Xn (i.e., n ~--+ g(xn)) is a primitive recursive
function.
18 Let var(x) be the relation that holds if and only if xis the Godel
number of a variable of P. var(x) is primitive recursive.
19 formterm(x), holding if and only if xis the G6del number of a
67
A SCHEDULE OF PROOFS

'formation sequence of terms,' is a primitive recursive relation.

20 Define the relation term(x) as holding if and only if x is the Go del
number of a term of P. term(x) is primitive recursive,
21 The relation formwff(x), holds exactly when xis the Godel number
of a 'formation sequence of wffs,' is primitive recursive.
22 wff(x) is a primitive recursive relation, where wff(x) holds if and
only if xis the Godel number of a wff of P.
23 Let occur(w, x, y, z) hold exactly when z is the Godel number of a
wff, x is the Godel number of a variable, and z can be written as z =
w * x * y. 4 occur(w, x, y, z) is a primitive recursive relation.
24 bound(w, x, y, z), indicating roughly that an occurrence of the
variable with Codel number xis bound in the wffwith Godel
number z, is a primitive recursive relation.
25The relations free(w, x, y, z), which holds if an only if occur(w, x,
y, z) and not•bound(w, x, y, z), is primitive recursive.
26 Let S(x, t, a) give the Godel number of the wff that results when
one free occurrence of the variabie with Godel number x in the wff
with Godel number a is replaced by the term with Godel number t
if xis the Godel number of a variable x, tis the Godel number of a
term, a is the Godel number of a wff A, and x does occur free in A,
and let S(x, t, a) = a otherwise. S(x, t, a) is a primitive recursive
function.
27 sm(x, t, a), iterating the operation of S(x, t, a) m times, is a
primitive recursive function of m, x, t, and a.
28 The replacement, conveyed through Godel numbers, of the variable
x throughout a wff A by a fonnulaically determined variable occur-
ring neither in A nor in a specified term t is a primitive recursive
function of x, t, and a where x, t, and a are the Godel numbers
corresponding to x, t, and A, respectively.
29 Defme Sub(x, t, a) as giving the Godel number of the wff that
results when the term with Godel number t is substituted for the
free occurrences of the individual variable with Godel number x in
the wff with Go del number a when, in fact, x is the Godel number
of an individual variable, t is the Godel number of a term, and a is
the Godel number of a wff, and as giving a otherwise. 5 Sub(x, t, a)
is a primitive recursive function.
30 Let sub(x, n, a) equal the Godel number of the wff that results when
k(n) is substituted for the free occurrences of the individual variable
with G6del number x in the wff with Godel number a when, in fact,
xis the G6del number of an individual variable and a is the Go del
number of a wff, and let it equal a otherwise.6 sub(x, t, a) is a
primitive recursive function.
31 freefor(x, t, a) is a primitive recursive relation, where freefor(x, t;a)
holds if and only if x, t, and a are the Godel numbers of a variable x,
a term t, and a wff A, respectively, and if 'tis free for x in A.'
68
 A SCHEDULE OF PROOFS

32 The relation notfree(x, a), which holds if and only if x is the Godel
number of a variable with no free occurrences in a wff with G6del
number a, is a primitive recursive relation.
33 axiom 1 (x), ... , axiomi(x), ... , axiomk(x) are primitive recursive
relations, where i enumerates the axioms of P and axiomi(x) holds
if and only if x is the Godel number of an instance of axiom i.
34 Define the relation deduct(x) as holding if and only if x is the Godel
number of a deduction. deduct(x) is primitive recursive.
35 The relation ded(x, y), which holds only when xis the Godel
number of a deduction of the wff with Godel number y, is a
primitive recursive relation.
Proposition 35 and its proof complete a schedule of proofs for the
proof of Godel's theorem. Once sub(x, n, a) and ded(x, y) have been
shown to be primitive recursive, the practical techniques of working
with primitive recursive functions and relations make available (and
are available as) the seeable/showable adequacy of the demonstration
that G(x, u) is a primitive recursive relation:
G(x, u) <> ded(x, cfl(u))
<> ded(x, subg{x 2 )(u, u))
<> ded(x, sub(g(xt), u, u)).

Thus, once the lemma asserting the numeralwise expressibility of primi-

tive recursive relations in P has been established, the diagonalization/
'proof can be constructed and the proof ofGodel's theorem completed.

B
A Schedule of Proofs as Lived-Work
In the remainder of this chapter, I describe four general features of the
schedule of proofs in terms of the lived-work of producing that schedule.
The material that is introduced here prepares the reader for the analysis
in the next chapter of what identifies a schedule of proofs as a naturally
accountable schedule of proofs for a proof of Go del's theorem.

1 As it is used in developing the schedule ofproofs, a Godel numbering

is not an abstractly defined co"espondence between the symbols,
expressions, and sequences of expressions ofP and the natural numbers;
it is a technique of proving
Disengaged from the work of proving Godel's theorem, a Godel num-
bering merely establishes a correspondence between the symbols,
expressions and sequences of expressions ofP and the natural numbers.
69
A SCHEDULE OF PROOFS

Under this correspondence, the syntax of P can be translated into

numerical functions and relations, but, speaking again in a manner
disengaged from the work of proving Godel's theorem, the numerical
functions and relations so defined are in no way assured analyzable
mathematical structures. In practice, however, a specific Godel num-
bering does provide for the mathematical analyzability of these
functions and relations.
As a hypothetical example, suppose that the numerical relation
R corresponding to some syntactic feature (R of P under the Godel
numbering with which we are working consists of all the natural num-
bers for which tlte exponent of the largest prime in each of these
numbers' prime factorizations is 2 - that is, R is the set of numbers
22 , 2·3 2 , 22 ·3 2 , 2·3·5 2 , 22 ·3 2 ·5 2 , 23 ·3·5 2 , etc. Given the apparatus
for working with primitive recursive functions and relations developed
in Propositions 1 through 16, the defmition ofR can be written as
R(x) ~ (x)L(x) = 2.
The immediate point is that this formula, besides being the seeable/
showable counterpart of the verbal definition of R, seeably/showably
exhibits R as a primitive recursive relation.
The artificiality of this example comes from the fact that (R is
presented in terms of the structure of the numerical relation R corres-
ponding to it. 7 In contrast, in that the syntax of P is defmed prior to,
and independently of, the specification of a Godel numbering, the
functions and relations appearing in Propositions 19 through 35 do not
have a predetermined mathematical structure. 'Godel numbering as a
technique of proving' refers to the fact that a proof-specific Godel
numbering not only allows the prover to define these numerical func-
tions and relations, but allows him to demonstrate, through the work
of uncovering the analyzable mathematical structure that that number-
ing provides, that these functions and relations are primitive recursive.
The relation term(x), for example, is defined only as the image under
the Godel numbering of the terms of P. It is the technique of working
with that Godel numbering that makes this set of numbers accessible
to the proof that it is, in fact, a primitive recursive relation.
As a means of recalling for the reader this technique of proving, I
conclude the discussion of it by outlining the work of producing a
formula that defines term(x) as a primitive recursive relation.
The reader will recall that the terms ofP are defined inductively and
include only those expressions of P determined by the following
COJ1ditions: (1) 0 is a term, (2) the individual variables are terms, (3) if
vis a term, then soisS(v),(4) ifv 1 and v 2 are terms, then so is +(v 1 v 2 ),
and (5) if v1 and v2 are terms, then so is -(v 1 v2 ). term(x) is then
defined as the image of the terms of P under the Godel numbering, or,
in other words, term(x) holds if and only if x is the Godel number of
70
 A SCHEDULE OF PROOFS

an expression of P satisfying one of the five conditions just listed. The

proof that term(x) is primitive recursive consists of translating this
definition into a mathematical formula that (i) given its placement in
the schedule of proofs, seeably/showably exhibits the primitive recur-
siveness of the relation that it defines, and, simultaneously, (ii) can be
read to find that the relation defined by the formula exactly identifies
the Godel numbers of the terms of P.
Let us begin the work of proving that term(x) is primitive recursive
by examining the case that x is the Godel number of 0. A numerical
relation consisting of a single number a is primitive recursive, as can be
justified by the fact that, since = is a primitive recursive relation,
Propositions 2 and 7 imply that {x I x = a} is primitive recursive. In
that we are only interested in the defming condition x = g(O), we
may write
term(x) ~ x = g(O) or ... ,
and, by Proposition 6, if the condition following 'or' defines a primitive
recursive relation, term(x) will be primitive recursive, also.
Next, consider the case, denoted var(x), that xis the Godel number
of an individual variable of P. As before, let 3, 5, 7, and 9 be the Godel
numbers assigned to the primitive symbols(,),', and x, respectively. If
the individual variables of P are enumerated as follows
xo := (x)
Xt := (x')
Xz := (x")

n
. ...----...
x 0 := (x"' · · . ')

x will be the Godel number of an individual variable if and only if x E

{g(xo), g(xt), g(xz), ... , g(x0 ), ••• } . Computing the values for
several of the g(x1), beginning with i = 0, one obtains the collection
{23·39·55, 23·39·57·75, 23·39·57·77·115, 23·39·57·77·117·135,
... } . The orderly way in which these numbers are arrived at points to
the formula for g(x 0 ) as 23 • 3 9 ·5 7 • .•. • p~+2 • p~+ 3 . With the apparatus
of Propositions 13 through 16 as background, this may be written as
the seeably/showably primitive recursive function
n n
n ~g(x 0 )=2 3 ·3 9 * ll Pf *2 5 =g((x)* ll Pf *g()).
i=O i=O

71
A SCHEDULE OF PROOFS

Let us return to the problem of demonstrating that the relation 'x

is the Codel number of an individual variable' is primitive recursive.
The condition that x is the Godel number of an individual variable
translates into 'x = g(xn) for some n,' and, by writing this more
formally, the formula
var(x) ~ 3n (x = g(xn)
is obtained. Although this formula does not exhibit vai(x) as a primitive
recursive relation, the practical techniques of working with primitive
re:cursive functions and relations insure the primitive recursiveness of
var(x) if (f) the relation x = g(xn) can be shown to be primitive recur-
sive and (ii) an upper bound for n as a function of x alone can be
found. But g(xn) was just shown to be a primitive recursive function,
implying that x = g(xn) delmes a primitive recursive relation. Since n
will always be less than g(xn),8 we have
var(x) ~ 3n..;;; x (x = g(xn))
seeably/showably defining var(x) as a primitive recursive relation, 9 and
teirn(x) ~ x = g(O) or. var(x)or ...
will define a primitive recursive relation if the condition following the
second 'or' defines one.
Consider, now, the remaining three cases - that x is the Godel
number of an expression of P either of the form (3) S(v), ( 4) +(v 1 v2 ),
or (5) ·(v 1 v2 ), where v, v1 and v2 are themselves terms. By temporar-
ily using 'comers' to indicate the Godel number of a primitive symbol,
these conditions can be written as
(3a) term(v) and x = fSI • 3rn* v * 2rn
(4a) term(vd and term(v2) and x = 2r+l • 3rn * v1 * v2 * f n
(Sa) term(v 1) and term(v2) and x = 2r·l • 3rn * v1 * v2 * 2rn
where term(v), of course, indicates that v is the Godel number of a
term. Without using 'comers,' these conditions become
(3b) term(v) and x = g(S() * v * g( ))
(4b) term(vd and term(v2) and x = g(+() * v1 * v2 * g( ))
(Sb) term(v 1 ) and term(v2) and x = g(·() * "1 * v2 * g( ))
g being used exclusively for the Godel numbers of expressions. In that
the right side of each of the equations for x see ably /showably defines a
primitive recursive function of either v or of both v1 and v 2, the
equations for x define primitive recursive relations of x and, appropri-
ately, of either v or ofv 1 and v2 .
Let us examine the case that x is the Godel number of a term of the
72
 A SCHEDULE OF PROOFS

form S(v). Someone engaged in proving Godel's theorem might begin

developing a formula that exhibits the primitive recursiveness of this
relation by first translating it into the formalism
3v [term(v) andx = g(S( )* v * g( ))]
and then proceed by fmding a bound on v depending on x alone. In
that consideration of the structure of the Godel numbering shows that
the GOdel number of an expression will always be greater than the
Godel number of a 'part' of it, 10 x itself will be a bound on v. Thus,
the prover arrives at the formula
3v =E;; x [term(v) and x = g(S() * v * g( ))].
However, the real problem remains: the apparent circularity of this
formula in a definition of term(x) must be removed and it must be
removed in such a way that the formula that results defmes term(x)
as a primitive recursive relation.
By proceeding in a similar manner for the cases of +(v 1 v2) and
·(v 1v2), one obtains the formulas
(4c) 3v1=E;;x 3v2=E;;x [term(vd and tenn(v2) and
x = g(+() * v1 * v2 * g())]
and
(Sc) 3v 1=E;;x 3v2=E;;x [tenn(v 1) and tenn(v 2) and
x = g(·() * v1 * v2 * g())],
again presenting tbe problem of their impredicativity but with the
further complication that two possibly distinct variables are being
specified as the Godel numbers of terms.
Before I give the construction of a formula seeably/showably ex-
hibiting tenn(x) as a primitive recursive relation, an important point
must be made. In using a given solution to find how the Godel num-
bering provides for a numbering of the terms of P adequate to that
solution and, at the same time, in using that solution to find the
method of constructing the terms of P that makes that numbering
possible, the reader works toward the recovery of pre-existent, objec-
tive structures of the syntax of P and of the GOdel numbering. In
contrast, for someone engaged in developing the schedule of proofs,
the problem of removing the impredicativity in the definition of
term(x) and cotemporaneously showng that term(x) is primitive
recursive is the problem of finding a method, using the Godelnumber-
ing, of associating numbers with the terms of P with the property that
the structure of that method, as it is made available through the use
of the apparatus for working with the Godel numbering, identifies
exactly the Godel numbers of the terms. In other words, the prover
73
A SCHEDULE OF PROOFS

looks to discover that aspect of the structure ofthe techniques of work-

ing with the GOdel numbering that will provide for what a solution
could consist of. For the prover, the solution is not molded to the
known structure of the GOdel numbering and the definition of the
terms of P; in the presence of the problem, the prover uses and develops
the techniques of working with the Godel numbering to discover a
mathematically identifiable method of numbering the terms adequate
to what a solution then becomes.
As a way of working toward the discovery of a formula for term(x),
consider a particular term t of P, for example, 11
t= +(x 1S(·(xs0))).
t can be viewed as being built up in a succession of steps, each in
accord with the inductive definition of the terms ofP:
Xs, 0, ·(xsO), SHxsO)), x 1 , +(x 1S(·(x 5 0))).
Denoting this sequence of terms as T, T will be referred to as a forma-
tion sequence fort. The point of constructing such a formation sequence
is that·a Godel number y can be assigned to it, namely
y = ~(x 1 ). p§(O). p§(·(x,O)}. pfS<·(x,O))). ~(x~).
pf+(x 1 S( ·(x 1 0)))),
with the following properties: (i) the apparatus of Propositions 13
through 16 allows each of the exponents of the prime factors ofy to be
checked one at a time to see if it is the Godel number of a term -
provided there is some way to make that determination, and (ii) given
that y represents the formation sequence of a particular term t, it is
not necessary to examine whether or not g(t) = g(S(v)) or g(t) =
g(+(v1v2) or g(t) = g(·(v 1v2)) for all possible terms v or v 1 and
v2 , but only whether or not one of these equations holds for terms
v or v1 and v2 with Godel numbers occurring as the exponents of
primes smaller than PL(y) in the prime factorization of y. In fact,
(ii) holds for the exponent of any prime factor ofy when one replaces
PL(y) with the prime number of which it is the exponent.
The generalization of the idea of this construction and numbering
proceeds by mathematically identifying that construction and num-
bering as a methodic procedure: first, a formation sequence of terms
is defined as a finite sequence of expressions of P - t 1, ... , ti, ... ,
tm - such that, for each i, i = 1, ... , m, one of the following con-
ditions holds:
(i) ti is 0
(ii) !i is an individual variable
(iii) tj is S(tj) for some j < i
(iv) ti is +{tjfk) for some j < i, k < i
(v) ti is ·(tjtk) for some j < i, k < i.
74
 A SCHEDULE OF PROOFS

The terms of P can then be seen to be only those expressions of P

that are the final expressions in a formation sequence of terms.
Lett be a term and let r be a formation sequence for it:
(6) r= t 1 , •.• , tj, ... , tk, ... , ti> ... , tm,
where tm = t. The G6del number of r will be
( 7 ) y= IJ¥<rl) •...• pf<rj) •...• p~<fk) •...• Prtv ..... P~tm).
Given the availability of this numbering, a number x will be the
Godel number of a term if and only if there exists some number y
such that the last exponent in the prime factorization of y is x (i.e.,
(y)L(y) = x) and, for each exponent in the prime factorization of v
(i.e., V1 ~ L(y), i =I= 0), that exponent is either the Godel number of
0 (i.e., (y)i = g(O)) or the Godel number of a variable (i.e., var((y)j))
or the Godel number of an expression of the form S(tj) for some
expression tj where the Godel number of tj is an exponent of some Pj
in the prime factorization of y with Pj < Pi (i.e., 3j < i [(u)i = g(S( )
* (v)j * g( ))]) or ...
The rendering into standard English of these conditions is both
unnecessary and obfuscating. The prover's familiarity with the practices
of working with primitive recursive functions and relations allows him
to work directly from the method of numbering indicated in (6) and
(7) and formulate the conditions on term(x) by rendering them into
the apparatus for constructing primitive recursive relations. In this
way, the prover directly constructs the formula
3y {[(y)L(y) = xJ and Vi~ L(y) ([i =OJ or [(y)i = g(O)]
or [var((y)i)J or 3j < i [(y)i = g(S() * (y)j *g())]
or 3j < i 3k < i ([(y)i = g( +() * (y)j * (y)k * g( )) ]
or [(y)i = g(·() * (y)j * (y)k * g( ))]))} •
seeably/showably defining a primitive recursive relation in x if a bound
for y depending on x alone can be found. 12
In order to find this bound, let us return 13 to the juxtaposition of
t, r, andy:
t = +(x 1S(·(xs0)))
T = x 5 , 0, -(x 5 0), S( ·(x 5 0)), x 1 , +(x 1 S(·(xs0)))

y = pytx5 ). p~(O). p§({x 50)). p§f.S'(·{x 50))). ~(xl)

• pg(+(x 1S({x 5 0)))).

Here, one notices that at least one formation sequence for t (i.e., r)
will have six constituents, one for each primitive symbol oft that is not
a parenthesis. Thus, there will be at least one number y {i.e., the y
75
A SCHEDULE OF PROOFS

computed above) such that y is the Godel number of a formation

sequence of t and y has the first six prime numbers as its only factors.
Let x · = g (t). In that the function L (x) counts the parentheses as
well as the other symbols in t, L (y) will be less than or equal to L (x).
Further, in that the Godel number of a 'part' of an expression is less
than the Godel number of the expression itself, 14
y = ~(,xs) • M(O) • p§( ·(,xs 0)) • p§4S'( ·(,x50))) • pfX1)
• pg{+(X 1 (·(,x 5 0))))
L(x)
.,;;:: Px • • Px .,;;:: Px • • Px _ II Px
""" 1 • • • 6 """ 1 • • • L(x) - i '
i=O
L(x)
and II pf will be a bound on y.
i=O
The methodic character of this computational reasoning in the case
of the immediate example provides the practical accountability of its
L(x)
generalization. Thus, if there is no y ..;;; II pf such that y is the Godel
i=O
number of a formation sequence of a term with G6del number x, x
will not be the Godel number of a term.
Finally, a review of the original formula to examine its adequacy
shows that it excludes the case that x = 1, but not that ofx = 0, both
0 and 1 not being the Godel numbers of terms. 15 Adding this condition,
one obtains the formula
L(x)
term(x) ~ x * 0 and 3 y..;;; II pf { [(Yk(y) = x] and
i=O
'Vi..;;; L (y) ([i = 0] or [(Y)i = g(O)] or
[var ((Y)i)] or 3 j <i [(y)i = g(S() * (y)j * g(_))]
or 3j < i 3k < i ([(y)i = g(+() * (y)j * (y)k * g( ))]
or [(y)i = g(-() * (y)j * (y)k * g( ))]))}
seeably/showably defining term(x) as a primitive recursive relation.

2 The schedule of proofs has a 'directed' character: it leads to and is

organized .so as to permit, the construction ofG as a primitive recursive
relation

An initial sense of the 'directed' character of the schedule of proofs can

be obtained by working back from the definition of G and establishing
'lines of dependence' between the various propositions of that schedule.
76
 A SCHEDULE OF PROOFS

I will briefly indicate how this can be done.

Let us begin with the construction of G. That G is primitive recur-
sive can be seen to depend on the demonstrations that sub and ded
are primitive recursive and, further, on the use of Propositions 2 and
7. 16 If we leave aside the matter of the placement of Propositions 2
and 7, the dependence of the primitive recursiveness of G on that of
sub and ded can be depicted graphically as follows:

/G(x,u)

ded(x, y) I \~ub(x, n, a)

Working backwards, consider the demonstration that ded(x, y) is

primitive recursive. Let A 1 , . . . , Am = A be a deduction of the wff A,
a1 , .•. , am = a be the Godel numbers corresponding to A 1 , .•. , Am,
respectively, and let b = p~' • ... • piWt. Then b is the Godel number
of a deduction of the wff with Godel number a - that is, ded(b, a).
But if A~, ... , A~ =A is another deduction of A and if a~, ... , a~=;= a
are the yodel numbers corresponding to A~, ... , A~, then b' = p~' •
... • p~s is also the Godel number of a - ded(b', a). From these
examples it can be seen that ded(x, y) holds if and only if x is the
Godel number of a deduction (i.e., deduct(x)) and if the exponent of
the largest prime number in the prime factorization of x is the Godel
number ofy. In symbols,
ded(x, y) ¢> deduct(x) and (xk(x) = y.
Given its placement in the schedule of proofs, this equation constitutes
the materially adequate proof that ded(x, y) is a primitive recursive
relation, 17 but for now, its importance for us lies in the fact that it
exhibits the dependence of the primitive recursiveness of ded(x, y)
on that of earlier propositions in the schedule of proofs. Graphically,
we have the following situation:

/G(x,u)

ded(x, y) / \ ~b(x, n, a)
doduc(~ I \ I I\
77
A SCHEDULE OF PROOFS

As one last example, let us examine the proof of the primitive

recursiveness of the relation deduct(x) which holds exactly when x
is the Godel number of a deduction. A deduction in P is defmed as a
finite sequence of wffs A 1 , •.• , Ai> ... , Am subject to the conditions
that each wff Ai of that sequence is either an instance of one of the
axioms of P, Ai results from some As, s < i, by generalization -that is,
Ai is of the form VxAs for some individual variable x and some As,
s < i - or Ai results from some A 8 and At, s < i, t < i, by modus
ponens - that is, As =At :::> Ai for some s < i and t < i. If a= A 1 ,
... ,1j, ... , Am is a deduction of P, then the Go del number of a will
be p¥VI 1 ) • ••• • Pf(Aj) • ... • Plh<Am)_ Now let a be any natural number
other than 0 or 1, 0 and 1 not being the Godel number of deductions
in P. The prime factorization of a yields a = p~a)t • ... • pfa)j • ... •
p~~~(a). Hence, a will be the Godel number of a deduction in the
formal system P if and only if a does not equal 0 or 1 and, for all i less
than or equal to the 'length' of a, L(a), the exponent of Pi in the prime
factorization of a, (a)i. is either the Godel number of an instance of an
axiom of P, the Godel number of the wffthat results from a wffwhose
Godel number is among the (a)j, j < i, by generalization, or the Godel
number of the wff that results from two wffs whose Godel numbers are
among the (a)j, j < i, by modus ponens. The materially adequate
demonstration that deduct is primitive recursive consists of the trans-
lation of the above analysis into symbols:
* *
deduct(x) ~ x 0 and x 1 and
Vi :E;; L(x) {i = 0 or axiom 1 ((x)i) or ... or axiomk((x)i) or
3j < i 3y < x (var(y) and (x)i = g(V) * y * (x)j)
or 3j < i 3s < i ((x) 8 = g(() * (x)j * gp) * (x)i * g( )))}
where axiom 1 , ••• , axiomk enumerate the axioms of P. As with
the examples above, the primitive recursiveness of deduct is seen to
depend on the prior demonstrations that the various functions and
relations occurring on the right-hand side of the equation are primitive
recursive and that the ways of constructing functions and relations used
in the right-hand side of the equation preserve primitive recursiveness.
Working back from the construction of G in this manner illuminates
the fact that the schedule of proofs provides for, and builds to, the
eventual construction of G in the diagonalization/'proof'. However, in
that this procedure of working backwards depends on the availability
of the schedule of proofs as a completed object, it gives only an initial
and retrospective sense of the 'directed' character of the lived-work
of producing that schedule. I want to now indicate what the directness
of that work actually consists of.
The discussion that follows will be facilitated by the availability of
a referentially-specific defmition of the syntax of a formal system. In
78
 A SCHEDULE OF PROOFS

order to give such a definition, some preliminary terminology must be

introduced. First, by a syntactic definition, I will refer to a definition
identifying some feature of a formal system in which the definiendum,
a syntactic object, is specified entirely in terms of ordered arrange-
ments of the primitive symbols of that system. As is customary, the
terms 'formation rules' and 'axiomatics' 18 will be used to refer to the
hierarchies of syntactic definitions that -are needed for - and that
culminate, respectively, in - the definitions of the wffs and the
*theorems* of that system. For the system P, the formation rules are
given by specifying the concatenated sequences of primitive symbols
that constitute the variables, the terms, and the wffs. The axiomatics of
P include 19 the constellation of definitions surrounding the notions of
substitution - namely, those concerning bound and free occurrences
of individual variables, that of a term being free for a variable in a wff,
and that of the operation of substitution itself - as well as the specifi-
cation of the axiom schemata and rules of inference, the defmition of
a deduction, and the culminating defmition, that of the *theorems*
e
of P. Finally, by the syntax20 of a formal system, 1 will refer to the
collection of syntactic objects that are defined by the formation rules
and axiomatics of that system.
The reason for giving this defmition of the syntax of a formal
system is that it makes precise the notion that a Godel numbering,
temporarily considered as merely establishing a correspondence between
the language 22 of a formal system and the natural numbers, renders the
syntax of a formal system as a collection of numerical functions and
relations. The syntax of a system so rendered will be referred to as the
arithmetized syntax of that system, therein providing a descriptive
definiteness to the vernacular expression that a Godel numbering
'arithmetizes' the syntax of a formal system.
With these definitions in hand, let us return to the discussion of the
'directed' character of the schedule of proofs.
In the schedule of proofs outlined above, Propositions 1 through 12
represent a selection of the rudimentary facts concerning primitive
recursive functions and relations that were well-known prior to Godel's
undecidability paper of 1931. Against the background of practices
represented by these results, Propositions 13, 14, 15 and 16 are rela-
tively easy to prove. However, the fact that they are relatively easy to
prove offers no motivation internal to the work of proving Godel's
theorem for their introduction. That motivation comes from the fact
that with the apparatus supplied by these propositions, the additional
mathematical structure that the Godel numbering gives to the arithme-
tized syntax of P is such that the prover can begin to write formulas
seeably /showably defining the functions and relations of the arith-
metized syntax as primitive recursive ones. More accurately, the need
for the apparatus ofPropositions 13 through 16 arises co-temporaneously
79
A SCHEDULE OF PROOFS

with the envisioned use of a numbering like a Godel numbering as a

technique of proving. With, for example, Godel numbers at and a 2
assigned to formal expressions A 1 and A 2 , the attempt to write the
formula for the Go del number of A 1:JA 2 indicates the need for
introducing an operation like that of *, thereby allowing the prover
towriteg(A 1 :JA2)=a1 *g(:J)*a2.
From within the work of attempting to write, and of developing
an apparatus for writing, the Godel numbers of syntactic objects like
that of A 1 :J A 2 and, for example, like the individual variables, the
increasing orderliness of that work exhibits the possibility that the
entire syntax of P - rendered as numerical functions and relations
under the Godel numbering - can be shown to consist of primitive
recursive functions and relations. Although, in the end, the project
that is thus initiated cannot be completed, only the culminating
relation, that of theorem(x), 23 cannot be shown to be primitive recur-
sive. The part of the project that can be realized is sufficient for the
construction of the diagonalization/'proof'.
The point is this: given the arithmetization of the syntax of P under
a Godel numbering, the presence and availability of the Godel number-
ing as a technique of proving sets up the program of demonstrating that
that arithmetized syntax consists of primitive recursive functions and
relations. The notion of the 'directed' character of the schedule of
proofs is a reference to this prospective enterprise as it unfolds in and
as the work of producing the schedule of proofs.

3 The selection and a"angement of these-particular propositions as

composing this-particular, intrinsically sequentialized order of proving
is the situated achievement of the work of producing the schedule of
proofs

The aim of the following four topics, (a)-(d), is to descriptively enrich

and elaborate this proposal by recalling for the reader the lived-work
of enunciating and organizing the schedule of proofs. In that the
accountably-ordered schedule of proofs is the achievement of this
lived-work, the fact that the finished schedule is an accountable order
of proving - and can, therefore, as an example, be rendered as a struc-
ture of logical dependencies - does not explicate the schedule's
temporally-situated and temporally-developing construction. Thus, by
making the development of the schedule available, in its technical
detail, as a production problem, the following material sets in relief
the radical problem of specifying the coherence of the work of produc-
ing the schedule as that schedule is actually being constructed. This
same material also provides a curious solution to that problem - that
the coherence of the work (or of all the things that make up and are
80
 A SCHEDULE OF PROOFS

spoken of as that 'coherence') inhabits and is the work itself.

(a) Six themes concerning the lived-work of producing a schedule of

proofs
The major part of topic (a) will be devoted to the review of the con-
struction of the 'concatenation' function * as the primitive recursive
function,
L(y)
X *y =X • II P~fx)+i .
i=O
The idea will be to place that construction within the encompassing
work of producing the schedule so as to find, in *'s construction, what
it means to speak of that construction as being 'within' in the first
place. In the course of adequately defining *, a significant portion of
the schedule of proofs will be developed, and topic (a) will be brought
to a close by extracting six summarizing themes from that material.
As a means of opening the discussion, let us begin by reviewing
the proof that the set of Godel numbers of terms of P is a primitive
recursive relation. The reader will recall that I began by considering
the particular term
t = +(x 1 S(-(x 5 0))).
A formation sequence T,
T = x 5 , 0, ·(x 5 0), S(·(x 5 0)), x 1 , +{x 1S({xs0)))
was then associated with t. Corresponding, under the Godel numbering,
to this sequence of terms is a sequence of numbers, namely
g(x 5 ), g(O), g({xs 0)), g(S(·(x 5 0))), g(xd, g(+(x 1 S(-(x 5 0)))),
from which the 'sequence number'
y = pytx"5 ) . p§(O). p§(·(x 5 0)). p~(iS'(·(x 5 0))). p~(x,)
• pg(+(x,S(·(x 5 0))))

was finally constructed. The idea of this numbering was that it made
available a methodic procedure, embedded in the techniques of con-
structing primitive recursive functions and relations, for determining
whether or not a given number x was the Godel number of a term,
that procedure being.articulated in the formula
L(x)
term(x) ~ x * 0 and 3y,;;;;; II
i=O
p~ {[(Yk(y) = x] and

'Vi,;;;;; L(y) ([i = 0) or [(y)i = g(O)] or

[var((y)i)] or 3j < i [(Y)i = g(S() * (y)j * g( ))]

81
A SCHEDULE OF PROOFS

or 3j < i 3k < i ([(y)i = g(+() * (y)j * (y)k * g( ))] or

[(Y)i = g(·() * (y)j * (y)k * g( ))}))}.
Now the prover, in beginning to construct a schedule of proofs, may
not foresee the exact construction of such a formula nor may he
anticipate the problematic details of such a construction for the partic-
ular, syntactically-specified theory in which he is working. On the
other hand, in that the construction of such formulas is the projected
thing to which his proving must come to, he can and does inspect the
projected schedule as a way of informing the production of it. In this
way, the prover may come to envision the following functions as
comprising the apparatus for working with the Godel numbering while,
at the same time, recognizing that the adequacy of that apparatus
will be determined over, and is contingent on, the course of working
out the schedule itself:

(i) a set of 'encoding' functions, one for each natural number m;;;;.. 1,
such that
0: Nm ~N

O:(a 1, ... ,am) '"+(a 1, ... ,am)=p~' • ... • p~m.

One of these functions maps, for example, the sequence of numbers
g(x 5 ), g(O), g( ·(x 5 0))), g(S(-(xsO))), g(x1 ),
g(+(x1S(-(x 5 0))))
corresponding to the formation sequence T to the sequence number
y = py<xs) • p~(O) • p§< · tx 5 0)) • p§(S( · tx 5 0))) • p~tx 1 )
• pt(+(x 1 S( ·(x 5 0))))

thereby 'encoding' the formation sequence T in the number y.

(ii) a 'decoding' function ( ) , mapping a given number x into the

exponent of the serially-numbered •-th prime in the prime factor-
ization of x:
O.:NXN~N

( )i : x = P~' • · · · • pfi • · .. • P~ ~'+ ai.

This function provides the device for inspecting each of the ex-
ponents in the prime factorization of a number to determine, for
example, whether or not those exponents correspond to the
Go del number of terms comprising a formation sequence for t.

(iii) a 'concatenation' function. If A and B are sequences of primitive

82
 A SCHEDULE OF PROOFS

symbols of P, then so is the expression AB formed by concaten-

ating the symbols of A with those of B. Let g(A) = p~1 • • • • •
pllm and g(B) = p~t • ... • p~n. Then g(AB) = p~t • ... • p~.
P~+l • ... • P~+n· The 'concatenation' function * maps (g(A),
g(B)) to g(AB), mirroring the process of concatenating the sym-
bols, expressions, and sequences of expressions of P with an
operation defined on pairs of numbers. Provisionally,24 let us
define* as
* :NXN-+N
* .· (x • y) ~ x * y = Palt • · · · • pam
m • pbt
m+ 1 • · · · • phn
m+n
where x = p~t • ... • p~ andy = p~t • ... • p~n.
Then, as an example, the Godel number of A :J B can be written
g(A :J B) = g(A) * gp) * g(B) once, of course, the associativity 25
of* has been established.

In addition to the projected construction of these functions, the prover

also has, in beginning to construct the schedule of proofs, in and as the
increasing articulation of the accountable work of that construction,
that way of working as a familiar and remembered course-of-action.
Thus, the prover may start the construction of the schedule by showing
that first the divisibility relation, then the set of prime numbers, and
then the function enumerating the primes are primitive recursive: 26
() The 2-place numerical relation x I y is primitive recursive.
Proof· xly ~ 3n :s;;; y (y = n • x). 27
( ) Let prime(x) hold if and only if xis a prime number. Then prime(x)
is a primitive recursive relation.
Proof· prime(x) ~ 1 < x and
Vy :s;;; x (ylx implies y = 1 or y = x)). 28
( ) The function Pn giving, for each n, the n-th prime number is primitive
recursive. (p 0 := 1.)
Proof· p 0 = 1
Pn+l = f.J.X :s;;; 0 (prime(x) and Pn < x). 29
This is not to say that the writing of these propositions is automatic or
mechanical. The prover, in a manner to be illustrated shortly, may have
to find this or a similar order as an appropriate order for the propo-
sitions, or he may more or less remember the propositions and their
order but, in the course of working out their proofs, come to establish
and, therein, to see again that particular order as an accountable order
83
A SCHEDULE OF PROOFS

of work. Even if the prover is using one or several previous proofs of

Godel's theorem as guides, he will still have to find for himself, over
the course of writing their proofs, in the material details of that writing,
that the selection and arrangement of the propositions given by those
proofs is, in fact, an accountable order of proving. However, the aim of
this discussion is neither to examine all the details of the schedule's
construction - leaving aside the question of how the relevant and
identifying details of the schedule emerge from within and are tied to
the work of that construction; a question that I will return to in the
next chapter - nor is it to assert that the construction of the schedule
is done in some specific manner. The material that has just been outlined
simulates the circumstances and material detail that might surround
the prover as he undertakes the construction of the 'concatenation'
function * as a primitive recursive function in the midst of constructing
the schedule itself. It is with this material as background that I now
want to turn to the examination of the work of constructing that
function.
Let us begin that examination by reconsidering the definition of *
that was given earlier:
X * y = pal • • pam
1 • · • m • pb1
m+l • · · · • pbn
m+n
where x = p~1 • ... • p~m andy= p~ 1 • ... • p~. An immediately
recognizable problem with this definition, independently of whether
or not it defines a primitive recursive function, is that the numbers 0
and 1 do not have prime factorizations and, hence, that * is not well-
defined. As a means of working toward this problem's solution, let us
first consider the case when x = 1. By using the fact that 1 was defined
as the 0-th prime p 0 - or, alternatively, by introducing such a modi-
fication into the definition of Pn at this point in the development of
the schedule30 - the prime factorization of any number x = p~1 •
. . . • p~ can be written as x .= p~ • p~1 · ... • p~m where s is an
arbitrary natural number. 31 Both for definiteness and in order to
see what it will come to in the work that follows, let us temporarily
adopt the convention that s = 0. The aim of introducing this device of
adding pg to the prime factorization of a number is that, by defining
the prime factorization of 1 as pg , the formula
x=poo • pal.
1 • • •
• pam
m
will hold, for some sequence of primes p(), ... , Pm and some sequence
of exponents 0, a 1 , •.. , am, for all x > 0, and, further, the formula for
* can then be written as
x * y = Poo • pal
1
• • · · • pam
m • Por • pbl
m+1 • · · · • pbn
m+n
where r, in this case, is arbitrary. 32 For definiteness, let r = 0 although
we will need to recall, at an appropriate time in our later work, that this
84
 A SCHEDULE OF PROOFS

choice was arbitrary and, therefore, that it can be changed. Provisional

on the adequate definitional explication of our conventions in and as
part of the developing schedule, we can define* as
X * y = Poo , pa11 , • · • , pam
m • Poo • pbm1 +1 , · · · , pbn
m+n
for all x and ~ greater than 0, x = pg • p~ 1 • • • • • p~m, andy = pg •
p~i • ... · Pnn, thus extending the initial definition of* to the cases
when x = I andy = I.
As a next step in developing the formula for *, let us consider the
case when x = 0. In this case, x simply does not have a prime factoriza-
tion - that is, 0 =I= pg • p~ • ... • p~m for any possible sequence of
prime numbers p0 , •.. , Pm and exponents 0, a 1 , .•. , am. In the
presence of this circumstance, we are forced to look for a different kind
of modification of the formula for *. One possibility, which on its
introduction appears as a 'natural' solution,33 is to define* as
x * y = x • Poo . pbl
m+l • · · · • pbn
m+n
which yields x * y = 0 when x = 0 and which can be further clarified as
n
o II phi
x*y=x· Po. m+l·
i=l
That x * y = 0 when x = 0 can be interpreted as saying that the null,
or empty, sequence of symbols of P concatenated with a non-null
sequence B is a null sequence. Although, in this interpretation, the
formula for * seems to lead to a false conclusion, the Godel numbers
of any member of Lg(P) 34 will always be greater than 0 or, for that
matter, I. Thus, in its projected use - as in writing g(A :::>B) = g(A)
* g(::>) * g(B) where A and B are already identified as non-null
sequences of symbols - the formula appears, at least for the moment,
as being unproblematic.
Finally, let us address the case when y = 0. Here, even the last
formula does not work. In view of the difficulties presented by this
case, one alternative is to provisionally define * 'by cases,'35
0 ify = 0
n
x*y= x • pg • II P~+l otherwise, where x = 0 or
i=l
x = pg • p~I • ... ·p~m and
Y = Poo • pbl
1
• • • •
• pbn
n
and, thereby, to also further provide for the primitive recursiveness of*.
The sketch of the development of a formula for * that has been given
to this point may strike the reader as the depiction of a fully 'rational'
and 'orderly' course of inquiry and discovery that a prover of Godel's
85
A SCHEDULE OF PROOFS

theorem might undertake and whose work would then come to fulfill.
By this I mean that the reader may have understood the work of
developing the last equation as being intrinsically connected to the
sequential placement of the various approximating equations in the
reasoned argumentation that I have used to motivate their introduction
and modification. If this is so, a reinterpretation of the preceding
discussion must be given. An actual prover will work his way to the last
equation by writing formulas similar to the ones displayed and by
modifying formulas already written - by crossing out symbols, adding
symbols, inserting symbols with arrows, and the like. In doing so, he
will provide for the recovery of the devices that are occasioned by his
work (and that are being locally employed as prospective solutions of
the problems that the written formulas manifest) in and as the found
order and in-course rearrangement of his writings. In other words,
the prover may articulate the course of reasoning that will lead again
to the last equation 'only' in and as the accountable organization of
the notes of his page of writings. To say this, however, is not to say that
the prover is not always in the presence of the practically accountable,
real thing that he is doing. Instead, it is intended to point to the local
character of the 'rationality' of what the prover is doing as that 'ration-
ality' inhabits the course of the prover's work and, therein, to point to
the fact that that 'rationality' is betrayed by the devices of a 'descrip-
tive narrative' like the one that I have given. Thus, for example, the
prover will see in his introduction of pg in the exgression x * y = p~1
• 0 •• b• pam
m • pblm+t •. . . . • pbn e , X * y -- P0. • P~
m+n (t" .. 1 • • • • • pam
m •
Po • Pnl+l • ... • P~+n) the purposeful thing that he is doing, and he
will also fmd that device as a reflectively uninteresting, natural conse-
quence of those doing - as something not needing to be specifically
elaborated as an appropriate, adequate, or efficacious procedure dis-
engaged from the further material development of the definition of*.
Similarly, the occasioned need for introducing pg need Qot develop in the
fashion that I have narrated, but could be motivated by considerations
of the manipulatable things that can be done with formulas already on
the page of working notes.
These matters will be addressed again later. For the moment, let
us return to the construction of *.
In that the prover of Godel's theorem knows 36 that the problem-
atic character37 of the initial definition of * when x or y equals 0 or 1
can be circumvented by defining* by cases, the prover will also antici-
pate that the problems arising from these cases are not serious for the
function *-proper that is intended to mirror the concatenation operation
on Lg(P) with an operator on N X N. In fact, immediately following
his writing the preliminary definition and seeing that definition's
inadequacies, the prover may, as part of a natural course of reasoning,
come to inspect the projected use of * to discover whether or not
86
 A SCHEDULE OF PROOFS

those cases are consequential to its eventual, adequate definition.

A potentially more serious problem in defining * comes from the
fact that the prover must be able to determine 38 a natural number's
prime factorization as a primitive recursive function of that number
in order to modify the formula
X * y = Pal, •
· · · • pam
m • Pb'
m+l • · · · • pbn
m+n
or n
X * y = X ' pg 'll P!a +i
i=l
so as to seeably/showably define * as itself a primitive recursive func-
tion. In particular, one of the functions that will be needed is one that
gives the 'len~th' of a natural number or, more precisely, that gives the
serial-number 9 of the largest prime in a given number's prime factoriz-
ation. Let thisfunction be denoted by L. Then, ifx = p~ • ... • p~m
and y = p~ 1 • ••• • p~, the prover wants L(x) = m and L(y) = n.
With this function in hand, he will be able to write
L(y)
X
* y =X. Po0 b·
n PL\x)+i•
•
i=l
bringing the definition of* closer to exhibiting its primitive recursiveness.
Once again, there is something slightly deceptive in the way of
speaking of the discovered need for the function L. It is on the occasion
of the need for such a function - as, for example, when the prover is
surveying a prellininary definition of* like
x * y -- pa,
1
• • · ·
• pllm
m • Pb'
m+1 • · · · • pbn
m+n
to see what can be appropriately modified about that definition - that
the prover will recall the presence of just such a function/device in the
proofs of Godel's theorem with which he is familiar. 40 However, after
seeing the materially purposeful thing that the introduction of L could
do, the prover may not set about constructing a formula defining it.
Instead, the prover may defer that construction and first use L to
develop the formula for * as
L(y)
X
*y= X • Po0 • n Pl.\x)+i
h·

i=l
and, using the 'decoding' function ( ) , further or co-temporaneously41
develop it even as •
L(y)
x*y= x• n P~&)+i•
i=O
therein producing a seeably/showably primitive recursive definition of
* subject to the adequate definition of L and ( ) •. If, in fact, he works
87
A SCHEDULE OF PROOFS

in this manner, he will also note, over the course of the work of writing
such formulas, the proper reorganization of that work as the account-
able work of the locally-obtained, projectively final version of that part
of the schedule of proofs on which he is currently working.
What is needed to define Las a seeably/showably primitive recursive
function? Well, what is L(x)? We want L(x) to be the number of the
largest prime that divides x or, in terms compatible with the ways of
constructing primitive recursive functions that are presently available,
we want it to be the number m such that Pm lx but Pkfx for all k >
m. As a means of obtaining that number m, we can use the least number
operator:
L(x) = Jlm ~ 0 (Pm lx and Vk ~ 0 (m < k implies Pkfx)). 42
By supplying proper upper bounds for m and k, this formula becomes
L(x) = J1ffi ~ x (pm lx and Vk ~ x (m < k implies Pkfx)),
defining a primitive recursive function, the 'length' of x. Furthermore,
by checking this definition for the potentially problematic cases x = 0
andy= 0, one finds that L(O) = 0 and L(l) = 0 as they 'should'. 43
In that L is needed for later propositions of the schedule (as, for
example, in the formula defining term(x)) and in that the serial charac-
ter of the schedule is to be maintained (as part of what will later come
to be called the schedule's 'structure of proving'), the proposition that
L is primitive recursive is enunciated as a separate proposition of that
schedule. In that that proposition has been found to be needed for the
proof that x * y is primitive recursive, it is also found to properly
precede the proposition concerning x * yin the 'fmished' schedule of
proofs.
'Next,' let us consider an appropriate modification of our new
formula
L(y)
X
* Y = X • Po0 •
II lv
Pl.\x)+i
i=l
so as to obtain, in a primitive recursive manner, the exponents b 1 , •• ~ ,
bn of the prime factorization of y. By using the 'decoding' function ( )
with b 1 = (y)i, b 2 = (y) 2 , ••• , bn = (y)L(y)• this formula can be re'-
written as
L(y)
X *y = X • pg . II P~&)+i
i=O
and the need for ( ) will, therein, either be·recalled or, if its use was
anticipated, be recalled again in and as the first instance demanding
its articulation and use.
A definition of (x)i as a primitive recursive function of x and i can
be found as
88
 A SCHEDULE OF PROOFS

{x)i = Jlk ~ x (pf I x and pf+l l x). 44

Checking this formula for the potentially problematic cases x = 0, x =
1, and i = 0, one finds that (l)i = 0 for all i > 0, but for the cases when
i = 0 or when x = 0, the formula is 'apparently' not well-defmed. 45
The solution of this problem may come about as the 'discovered
gestalt'46 of the following circumstances: By reviewing the work of
coming to the formula
L(y)
X * y= X • n "L(x)+t
Po0 •
D.(y)j .
i=l
as a feature of inspecting that formula to find a next thing to be done,
the prover will find again (as something already known) the arbitrari-
ness of the selection of pg in that formula. Seeing, 'therefore,' that he
can write PL(x) in the place of p 0 , the prover can obtain
L(y)
X
* Y-
- X • PL(x)
o • II1 PL(x)+l
(y)i
i=l
which is appropriate, by definition of L(x), even when x = 0 and
x = 1. If (y ) 0 is defined as equaling 0 for all y, then P1(x) = p~~~) and
L(y)
X * y= X • Poo • n p(y)i
L(x)+t.
i=l
L(y)
- X • p(Y)o • fi p{y)j
- L(x) L(x)+i
i=l
L(y)
=x • ll P~~h+i '
i=O
the last line defining * for all x, y E N with x * y = 0 if x = 0 and with
x * y = x if y = 0. This being the case, the prover will defme (x) 0 := 0
for all x, implying (0) 0 := 0 as well. The appropriate defmition of().
therein emerges as
0 if x = 0 or i = 0
(x). = {
1 Jlk ~ x (pflx and pf+ifx) otherwise
and with it, *'s definition as well:
L(y)
X *y =X • fi p(y)j
L(x)+t'·
i=O
a seeably/showably primitive recursive function of x andy. In particu-
lar, x * y is defined when y = 0, and the formula is applicable without
introducing special conventions for the prime factorization of 1.
Furthermore, given this way of working, the proofs of the primitive
89
A SCHEDULE OF PROOFS

recursiveness of ( ) and L naturally precede that of *, and the proofs

of() andL are found to be independent of one another and, therefore,
arbitrarily ordered in relation to each other. In that the definition of*
depends ·on that of L, the definition of L 'should' directly precede that
of*.
One last problem with the formula for * remains, although that
problem may make its appearance only later in the development of
the schedule. The problem is this: For almost all of the proofs of the
schedule concerning the arithmetized syntax of P, a prover (heuristi-
cally)47 uses the *-operator to exhibit the 'structure' of a natural
number x by envisioning that number as the Go del number of a member
of a particular subset of Lg(P) and then 'decomposing' it into its
constituent parts. For· example, a prover will want to write formulas
like x = g(A) * g(:>) * g(B) exhibiting the fact that if g(A) and
g(B) are the Godel numbers of wffs A and B of P, then xis the Godel
number of the wff A :::> B. Now the ordered sequence of primitive
symbols of P, a 1 ••• ~· that is being meta-syntactically abbreviated
as A:::> B-as, for example, if A= a 1 ••• ~-1• :::> = ak, andB =
~+1 ... ap - remains the same whether that ordered sequence is
considered as the sequence of symbols making up A :::> concatenated
with the ordered sequence of symbols identified as B, or it is considered
as the symbols of A concatenated with those of:::> B, or the sequence
of primitive symbols is considered as being partitioned in some other
way. The prover wants a similar property to hold for the *-operator;
for example, he will want to be able to unambiguously write
X= g(A :::>B)= g(A) * g(:>) * g(B)
independently of whether g(A :::>B) is decomposed as
g(A :::>B)= {g(A) * g(:::>)} * g(B)
or as
g(A :::>B)= g(A) * {g(:::>) * g(B)}.
Formulated more generally, the property of* that needs to be estab-
lished is its associativity - that is, that
(x * y) * z = x * (y * z)
for all x, y, zEN.
In a sense, the associative property of* on thesetofG6de/ numbers48
is already ensured by the fact that the definition of * articulates the
methodic character of the Godel numbering and the Godel numbering
was constructed so as to assign unique numbers to different elements
of Lg(P) independently of the various ways in which those elements
can be constructed by concatenating the primitive symbols that com-
pose them. The prover, however, on the occasion of questioning the
90
 A SCHEDULE OF PROOFS

associativity of *, will not necessarily reflect on the reasonableness of

* being associative, but, instead, will 'simply' examine49 * 's definition
L(y)
X *y =X ' II p(y)i
L(x)+l
i=l
to see if, in fact, that property holds. Such an examination locates the
one problematic case as being y = 0. Using *'s definition, the prover
will come to calculate
(x * 0) * z = x * z
x * (O * z) = z.
In that the equality
x * z=x
does not generally hold, the prover will, therefore, not be able to assert,
without qualification, that * is associative.
Two solutions to this problem seem to be available. As I suggested
earlier, the prover could redefine * by cases, as in
Oify=O
X* y= [ ·L(y)
x • II p~{~)+i otherwise
i=O
thereby ensuring the associativity of* by defining x * y to always equal
0 when either x or y equals 0. Such a device is somewhat artificial: the
L(y)
formula x • II p~{~)+i already adequately defmes * as a primitive
i=O
recursive function on N X N, and, in that this is made available to the
reader by offering y = 0 as a separate case, the distinction between
*
y = 0 and y 0 in a definition of* by cases instigates an examination
and discussion of how * will actually be used in the proofs of the
schedule that follow its introduction. However, in contrast to this
artificiality but tied to the examination of which it speaks, the defmition
of * by cases has a potentially serious consequence for the development
of the schedule: in that the values of* are being chosen for some of its
arguments the prover must look to the projected use of* to see if such
a choice will affect the proofs that follow it.
A second solution to the problem raised by the desired associativity
of * is afforded by the fact that, since all Godel numbers are greater
than 0 and since * will be applied only to Godel numbers, the associa-
tivity of * need only be established when x, y, and z are all greater
than 0. The prover - for example, in examining the justification for
L(y)
modifying the defmition x * y =x • II p~~)+i when y = 0, at the
i=O
91
A SCHEDULE OF PROOFS

time of that examination - may come to see, as a pointed relevance for

his immediate work, that all GOdel numbers are greater thaJ;J. 0 and,
therein, that the case when y = 0 is actually of no consequence for this
later work. Rather than modifying the definition of * and motivating
that modification for the reader, the prover will simply define * by the
formula
L(y)
X *y=X •
.
II p(y)j_ .
L(x)+t
t=O
and then note, in one way or another, that while the associativity of*
will be needed later in the schedule, that associativity need only be
established when x, y, and z are all non-zero. 50 • 51
Let us now - i.e., at this point in the construction of the schedule
of proofs - look back at the apparatus that was originally envisioned
as being needed to show that the arithmetized syntax of P is made up
of primitive recursive functions and relations. Of that apparatus, only
the 'encoding' functions
() : (at, ... , am) H <at, ... , am>= P~1 • ••• • p~
have yet to be shown to be, or reconstructed as, or shown not to be
primitive recursive. But here, the displayed equation already seeably/
showably defines a primitive recursive function: the formula for (a 1 ,
... , am> can be rewritten as 52
<at , ... , am> = P~1 • ••• • p'tfll
_ J.Vl(a1 , ••• , am) • • IF(a1 , ••• , am) • • I~(a 1 , ••• ,llffi.)
-p m ··· P m ··· P m ·
Z 1 (a10 ••• , am) Zi (a10 ••• , am) Zm (a10 ••• , am)
where If is, as the reader will recall, the projection function mapping
(al> .. - , ai, ... •l!rn) onto its i-th COOrdinate ai and zr is a Situationally-
occasioned notational 'innovation'53 for the function that extends the
domain of the primitive recursive function Zi (a) = i from N to Nm;
in that the relevantly exhibited and exhibitable 'component' functions
If, Zf, exponentiation, the mappings
(at' ... 'am) \-+ Zf (at' ... 'am) 1-+ Pzm(a
i 1, ••• , am ) = Pi
i = 1, ... , m, and the fmite product
(at, ... , am)-+- at • ... • am
are primitive recursive, 54 then so is <at, ... , am). 55 The immediate
point is that the prover, in seeing how to go about showing that <at,
... , am> .= p~ • . . . • p:7f defines a primitive recursive function,
also sees what is (evidently) needed for such a demonstration and,
therein, as a pointed relevance of that inspection, fmds that the
proof of the primitive recursiveness of ( ) is not dependent on the

92
 A SCHEDULE OF PROOFS

primitive recursiveness of L, ( ) , or *; that the proposition asserting

the primitive recursiveness of ( 5 may be placed irrespectively of the
relative positioning of the assertions concerning L, ( ) , or *; and, in
that L, ( ) , and * do exhibit dependencies among· their proofs56
and, therein: in that the produced arrangement of those propositions
and their proofs compose an orderly course of proving among them-
selves, that the proposition concerning ( ) 'should not' be inserted
within the arrangement of L, ( ) , and *.
Let me give the schedule of proofs as it has been developed to this
point: 57
( ) The 2-place relation xly is primitive recursive.
Proof" xly~ 3n ~ y (y = n • x)
( ) Let prime(x) hold if and only if xis a prime number. Then prime(x)
is a primitive recursive relation.
Proof· prime(x) ~ x > 1 and
Vy ~ x (ylx implies (y = 1 or y = x)).
()The function Pn giving, for each n, the n-th prime number is
primitive recursive. (p 0 := 1.)
Proof· Po= 1
Pn+l = JJ.X ~ (pn)0 + 1 {prime(x) and Pn < x}.
To see that (p 0 ) 0 + 1 is an upper bound on x, it is enough to note
that (p 1 • ••. • Pn) + 1 either is a prime number or is divisible by
some prime number greater than p 0 , for it then follows that Pn+l
~ (pl • · · · • Pn) + 1 < (pn) 0 + 1.
()Define (x)0 asthe exponent ofp 0 in the prime factorization ofx if
x > 1 and n > 0 and as 0 otherwise.* Then (x)0 is a primitive
recursive function of x and n.
Proof· { 0 if x = 0 or i =0
(x)n =
J.Lk ~ x (pklx
n
and pk+llx)
n
otherwise
() Let the function L(x) give the number n of the largest prime Pn in
the prime factorization of x or give 0 if xis 0 or 1.** L(x) is a
primitive recursive function.
Proof" L(x) = J.Ln ~ x {p 0 lx and Vk ~ x (n < k implies Pkfx))
()For every natural number y, y = 0 or y = pfr)o • p~Y)t • ... • p~&1(y).
* E.g., (294 )4 = (2• 3• 72) 4 = (pl·P~ ·p~ ·p~ )4 = 2.
** E.g., L(294) = L(pl·p~ ·p~ ·p~) = 4. L(x) defines the 'length' of x.
93
A SCHEDULE OF PROOFS

Define x * y as the function mapping (x, y) to the value

L(y)
X * y =X • fl p~~)+i"t
i=l
Then x * y is a primitive recursive function of x and y.
It follows from the definition that x * (y * z) = (x * y) * z
for all x, y, and z greater than 0. Thus, the finite 'product' of
numbers a 1 >-0, i = 1, ... , n, can be written unambiguously
as a 1 * ... *an.
58 ( ) For each n > 0, define (x 1 , ••• , Xn) by the equation
<x1, ... , Xn) = P~ 1 p~n·*
• ••• •

Then (x 1 , ••• , xnhs a primitive recursive function of(x 1 , •

. . . ,Xn).
I now want to bring topic (a) to a close by developing six themes
that summarize and enrich the preceding discussion.
(i) There are two related ways in which the part of the schedule
that was presented above need not have been constructed in the manner
I have described. First, both the preceding text and the accompanying
footnotes have already emphasized the variations of materially-specific
and materially-motivated reasoning that could lead to the same, or to a
recognizably similar, schedule of proofs. Second, a naturally accountable
schedule of proofs need not, in the end, be identical to the schedule
that was partially constructed. Thus, were a prover to have envisioned
the need for 59 the primitive recursive functions 60
lifx=O
sg(x) = {
Oifx=FO
Oifx=O
sg(x) = {
lifx=FO

1x-y l=(x..!..y)..!..(y..!..x)
and had he 'similarly'61 introduced/exhibited the function rm(x, y)
giving the remainder upon division of y by x as a primitive recursive
function,

t E.g., 294 * 6 = (2·3·7 2 ) * (2·3) = (p~ ·p~ ·p~ •p:) * (p~ ·p~)
= P~ ·p~ ·p~ ~P: .+pl ~N
= 2•3•7 2 ·11~13= 42,042.
:j:E.g., u,t,o, 2>= p~ ·p~·p~·pr= 2·3·7 2 = 294.

94
 A SCHEDULE OF PROOFS

rm(x, 0)= 0
rm(x, S(y)) = [S(rm(x, y)] • [sg( lx- S(rm(x, y))J )],
•then,'62 writing KR for the characteristic function of a numerical
relation R, the proofs of the primitive recursiveness of xly and
piime(x) could have been given as 63
()The 2-place numerical relation xly is primitive recursive.
Proof" K 1(x, y) = sg(rm(x, y))
( ) Let D(x) be the number of divisors of x if x > 0 and be 1 if x = 0.
Then D(x) is a primitive recursive function.

Proof: D(x) = ~ sg(rm(i, x))

i=O
( ) Let prime(x) hold if and only if x is a prime number. Then prime(x)
is a primitive recursive relation.
Proof· Kprime(x) = sg[(D(x)..!.. 2) + sg(Jx- 1 J) + sg(x..!.. 0)]
It is clear, then, that the materially-specific detail and the materially-
exhibited reasoning bf any one such schedule may differ from other
practically objective, comparably adequate schedules of proofs. But the
ethnography that I gave earlier did not attempt to document either
invariant material specificities of all schedules of proofs or invariant
'cognitive processes,' nor did that ethnography have such invariant
descriptions as its projected goal. Instead, that ethnography - and the
preceding example as well - make. available, as an inspectable and
researchable phenomenon, the finding that the work of producing
a schedule of proofs is, in every particular case, from within that
work itself, constrained by the material character of its own developing
argument and that it is always from within just-this materially specific,
just-this increasingly articulated, endogenously and developmentally
organized way-of-working/mathematical object that a prover will
further articulate and organize, as the endogenous work of his proving,
both that object and, simultaneously, that way of working.
(ii) Once again, let me begin by making several observations. First,
even though the lived-work of producing a schedule of proofs is, in
each particular case, constrained by the material character of its devel-
oping argument, there is nothing about a particular way of working,
from within the developing course of the work itself, that is self-
exhibiting of its uniqueness in providing a solution for the problem-at-
hand. As a feature of proving's work, the proof of a theorem can,
potentially, always be given in a different and hitherto unrealized
manner. Second, the solution of an unsolved problem - as the very
thing that such a projected solution has come to be in that the problem
is recognizably unsolved and, therein, not yet adequately understood -
cannot be specified before it is found, nor can the existence of a
95
A SCHEDULE OF PROOFS

solution be guaranteed to be made accessible through any particular

way of working. And third, as a prover develops a 'known' proof of a
theorem (as, for example, on the request of a student), the familiar
efficacy of 'the way of the theorem's proof can be betrayed by the
'adversity' of the self-same, notationally-specific, temporally develop-
ing, endogenously organized manner of proving - that 'adversity'
consisting, for example, of the further work that the prover's current
methods recognizably project or of the curious 'loss' of the orderline~s
of the argument and, therein, simultaneously, the 'loss' of the vision of
the thing that needs to be proved. In the face of such circumstances,
the prover will search for alternative ways of proving that will circum-
vent (what are available as particularly-his) present troubles.
These observations lead to and set in relief a further observation, one
made available by the ethnography as well: when a mathematician is
working on a problem, he is actually searching for and cultivating 'some-
thing,' in and as the developing and projected writings on the working-
page or blackboard, which, when found, makes up the thing that can
then be spoken of as a mathematical discovery. Speaking particularly of
the work of producing a schedule of proofs, what I propose is this: the
lived-work of producing a schedule of proofs constructs that schedule,
as its accomplishment, as an accountable course of mathematical
proving, and it is that accountable course of proving, as it has come to
be embodied in the schedule over the course of the schedule's construc-
tion, that is the thing that is being looked for as, and that then makes
up, the mathematical discovery64 (of that schedule). Furthermore, and
in consequence, in that that discovery consists of the just-this endo-
genous organization of work practices, that discovery is irremediably
and exclusively available in and as a local enterprise.
(iii) Consider now the 'finished' portion of the schedule of proofs
that was given above. In the earlier discussion of the 'directed' character
of a schedule of proofs, I began that discussion by indicating how the
ordering of the propositions of a schedule could be graphically repre-
sented as 'lines of dependence,' either as

/~~
(xln T)~<x,, . .,x.>
£!!.

I
96
 A SCHEDULE OF PROOFS

or, replacing the functions and relations with the associated numbers
of the propositions of the schedule concerning them, as

/16~
14 15 <x 1, ... ,Xn>

1~~
I
~12
11
On being presented with such a graph, a prover will fmd in its depiction
the availability of a method of representing the orderliness of the entire
schedule of proofs,65 and this, partly because that method does not
seem to be peculiar to the schedule itself. The graph, in its disclosure
as a method or representation, provides for essential and inessential
features of the-work-of-the-schedule's-production/the-schedule's
organization, and, therein, provides as well for a transcendental or
platonic schedule of proofs that is disengaged from the local work of
its production. The graph, together with its associated method,
renders66 the schedule of proofs as what could be called a 'structure of
logical dependencies.'
In miniature, let us suppose that A, B and Care propositions of the
schedule and that proposition A is used to prove proposition B which,
in tum, is used to prove proposition C. Then the ordering of these
propositions could be represented as A ~ B ~ C, where the arrows
provide a similar function as the serially ordered numbers and 'reason-
ably' oriented line segments in the graph above. The immediate point is
that to make the observations that A, Band C are needed in the schedule;
that A, Band C adequately articulate a portion of the schedule and that
the three make up a practically distinguishable and practically separable
sequence of proofs; that A is used to prove B, that B is used to prove C,
and that A is not directly used to prove C; that B is a necessary inter-
mediate proposition between A and C; that the propositions are appro-
priately arranged as first A, then B, and then C: that A is, in fact,
actually needed to prove B, and that B is actually needed to prove C -
to make these observations and, thereby, to come to fmd the adequacy
and cogency of A ~ B ~ C as a representation of the orderliness of a
part of the schedule - already places the prover within the 'perspective
of the proof's work.' In this way, then, the 'analysis' of the orderliness
of the propositions of the schedule that is provided by A ~ B ~ C is
already tied to the availability of the lived-work of proving just-those
propositions; in that the propositions of the schedule are already
97
A SCHEDULE OF PROOFS

chained to their associated proofs and in that together, those propo-

sitions and proofs make up a naturally accountable course of proving,
the enunciation and arrangement of the propositions thereby comes
to exhibit, via this rendering, a transcendental order of work. The
prover does not fmd the orderliness of the 'completed' schedule by
first, if ever, consulting the logical requirements for such an order, but,
instead, finds that orderliness by returning those propositions to the
work of their proofs from within which they then take on their order
properties. Thus, to draw 'lines of dependence', a prover must first
fmd the associated work of the schedule's production; to see that
proposition B is needed to prove proposition C is already to fmd
oneself within the work of proving that the schedule both provides
and is a part.
The larger point is this: in that a schedule of proofs is the endo-
genously organized, temporally and materially developing construction
that it has come to be over the course of its production, that schedule
takes on - not just its order properties as they become available
through a schematic rendering of its 'logical structure' -but all of its
properties as the naturally accountable object that it demonstrably is.
(iv) Today, if a mathematician were to go about proving-again
Godel's theorem, and were he to do so by explicitly constructing a
schedule of proofs, it is unlikely that he would actually go through the
detailed work of constructing such a schedule that I indicated and
outlined above. Instead, he would probably construct a schedule of
proofs in immediate consultation with established texts giving such
schedules. This being the case, what then is the relevance of the pre-
ceding ethnography as a description of mathematicians' work? The
relevance is this: although a mathematician might not undertake the
construction of a schedule of proofs without consulting, over the
course of his work, other, already 'completed' schedules, and although
he might not, thereby, work out for himself the material detail and
organization of the schedule, a mathematician is completely familiar,
in and as his own work as a practising mathematician, of what 'finding'
and, therein, articulating and organizing, mathematical proofs consists
of, in its material detail, as lived-work. In viewing another prover's
schedule of proofs or in juxtaposing a number of such schedules so as
to fmdwhat needs to be done for the particular schedule that he himself
is. building, the mathematician reads through those schedules to fmd-
and in reading them, finds - what they make available as the account-
able work of producing just-their schedules of proofs, respectively.
A mathematician need not go through the entire reconstruction of a
given schedule to have access to that schedule as an adequate summary
of a naturally accountable way of proving, and this is so not because
of what the mathematician 'knows,' but because of what he is utterly
familiar with as the endogenously organized workings of his own praxis.
98
 A SCHEDULE OF PROOFS

The ethnography that I gave earlier descriptively explicates this familiar,

yet, for the mathematician, unremarkable work in the particular case
of the construction of a part of a schedule of proofs.
(v) Point (iv) lends itself to the discussion of another and, as the
reader will see, related issue. In order to introduce it, let me begin by
offering two different ways of speaking abut the 'received' history of
a mathematical theorem. 67 The first and a 'weak' use of this notion is
that a theorem and its proof are available to a mathematician (working
in the particular field in which the theorem occurs) not just as a theorem
and its •proof, but as a theorem and proof that arise from within and
that are responsive to the proper history of their discovery and develop-
ment. That history is understood to include the accountable historical
development of the methods that are used in the proof, it provides
the accountable· context from within which the theorem took, and
now takes on, its importance, it provides an accountable history for the
development of the current and alternative statements of the theorem,
of its current and alternative proofs, and of its current uses in the field,
it is available as the history of the work of particular people, etc.
The second and 'strong' use of the notion of the ·~ceived' history
of a: theorem refers, in contrast to the preceding, to the 'state of the
art' ofthe theorem's proof. Under this usage, the 'received' history of
a theorem is not seen to consist of a proper history, but of 'received'
ways of proving. The reason for this usage being 'strong' and the other
usage being 'weak' is that it is by reason of the availability to a mathe-
matician of that 'state of the art' that allows him to examine and
construct an appropriate history of the theorem in question. It is in
the presence of naturally accountable proofs of the theorem that a
mathematician is able to fmd and argue what the proper history of that
theorem and its proof is.
Having distinguished these two different ways of speaking about the
'received' history of a theorem, let us return to the discussion of the
work of producing a schedule of proofs. The connection between the
two is this: in that, over the course of constructing a schedule of
proofs, a prover consults other 'completed' schedules as a means of
constructing his own, a prover finds as an incidental feature of that
consultation, what the 'received' history of the proof of Godel's
theorem - in the strong sense - at least in part, consists of. And, at
the same time, a prover fmds the place of his own work within that
'received' history as well. The point, then, is this: it is in this way~ in
that a· schedule of proofs, over the course of its production, is con-
structed in consultation with already 'completed' schedules of proofs
- that the material presentation and organization of such schedules
are preserved, repeated and modified as appropriate, efficacious and
accountable ways of proving Go del's theorem.
(vi) Finally, it is now possible to begin to make descriptively precise
99
A SCHEDULE OF PROOFS

what it means to speak of the intrinsic orderliness and the intrinsic

sequentialization of the schedule of proofs. The idea is this: on every
occasion when a prover comes to examine a schedule of proofs to see
if the definitions of that schedule 'fit' together or to review the arrange-
ment of the propositions or, in general, to find the accountable order-
liness of a given schedule of proofs, the prover does so by returning
those 'questions' to the work of producing the schedule that the given
schedule itself makes available. And, in fact, in the very way that they
arise, those 'questions' are already tied to, occur from within, and are
not disengageable from the work of producing the schedule of proofs.
To see, for example, that L(O) should equal 0 or to see that certain
propositions should be arranged in such and such an order is not to
find, by reflection on the properties of natural numbers, that 0 has no
prime factorization and, 'therefore,' that L(O) = 0, nor is it to fmd
that the nature of logical inference 'requires' a specific organization of
propositions. Instead, to 'understand' the appropriateness of the de-
finitions and the organization of a schedule is to find oneself, as a prover,
inextricably engaged in the work of proving Go del's theorem. To speak
of the intrinsic orderliness of a schedule of proofs is to refer to the fact
that the orderliness of a schedule of proofs is exhibited in and as a
course of mathematical proving.
(b) The constrnction of a schedule of proofs so as to provide an
apparatus within itself for the analysis of the work of its own
constrnction
As part of the work of 'working through' an established schedule of
proofs, a prover will use and interrogate the previous propositions
of that schedule so as to provide an exegesis (and, therein, to establish
the adequacy) of a given proof as a defmition of a primitive recursive
function or relation. Consider, for example, the following proof of
the primitive recursiveness of the divisibility relation (Proposition 11):
xly <* 3n.,.; y {y = n • x).
On 'coming to' this formula while working through another prover's
schedule, a prover might68 thereupon compose - as a temporally
realized, materially exhibited course of reasoning - a sketch derivation
similar to the one below:
n·x (Proposition 3)
y=z (Proposition 5)
y= n•x (Proposition 7)
3n .,.; w (y = n • x) (Proposition 8)
3n.,.; y (y = n • x) (Proposition 2)
the reasoning of which can be elaborated as follows: in that multiplication,
100
 A SCHEDULE OF PROOFS

n • x, and the equality relation, y = z, are, respectively, a primitive

recursive function (by Proposition 3) and a primitive recursive relation
(by Proposition 5), the equation y = n • x defines a primitive recursive
relation in that it results from the 'substitution'69 of a primitive recur-
sive function for a variable in a primitive recursive relation (Proposition
7); from the primitive recursiveness of y = n • x, the technique of
bounded quantification (Proposition 8) then insures that 3n :E;;; w
(y = n • x) defines a primitive recursive relation of x, y and w; from
which it follows, by 'identifying' the variables y and w (Proposition
2),70 that 3n :E;;; y (y = n • x) defmes a primitive recursive relation of
x and y alone.
A reviewer of a schedule of proofs may not actually construct such
a derivation, and even if he does, if he does so with no intention of
preserving that derivation, he need not rewrite it (as the derivation
above) so as to exhibit, in its material presentation, the achieved
orderliness of its reasoning. Independently, however, of whether or not
a particular reviewer actually constructs such a derivation of the formula
3n :E;;; y (y = n • x), the derivation given above does point to the
following phenomena: first, a reviewer, in and as the 'working through'
of another prover's schedule of proofs, engages in the situationally
relevant work that makes up the 'checking' of the proofs of that
schedule, and second, it is through that work that a reviewer comes to
find, as his - though not idiosyncratic - achievement, both the
accountable adequacy of the proofs of that schedule and, as well,
simultaneously and inseparably, the accountable adequacy of the
initial propositions of that schedule in supplying just the apparatus
necessary for making the adequacy of the later proofs demonstrably
and analytically available. In this way, the derivation
n·x (Proposition 3)
y=z (Proposition 5)
y= n • x (Proposition 7)
3n :E;;; w (y = n • x) (Proposition 8)
3n :E;;; y (y = n • x) (Proposition 2)
not only comes to exhibit the accountable analyzability of the assertion
that 3n :E;;; y (y = n • x) defmes a primitive recursive relation, but it
exhibits the adequacy of the initial propositions of the schedule (Propo-
sitions 1-10 of the schedule given earlier) in providing an apparatus for
such a demonstration as well.
In the discussion that follows, I review the work of producing the
formula
xly ~ 3n :E;;; y (y = n • x)
101
A SCHEDULE OF PROOFS

as a proof of the primitive recursiveness of the divisibility relation. The

aim of this review is twofold: first, it will illuminate how a prover,
in constructing a schedule of proofs, selects and articulates the initial
propositions of the schedule - and, therein, cultivates and refines
locally developing practices of proving - so that the-initial-propositions/
the-way-of-working-to-which-they-are-inseparably-tied come to com-
pose, over the course of their development, an accountably adequate
apparatus for the analysis of the later proofs. At the same time, this
review will permit the work practices of a prover constructing a schedule
of proofs to be contrasted to those of a prover reviewing another
prover's schedule. Although I speak here, and in the following, of an
'original prover' and a 'reviewer,' these terms are actually being used to
refer to contrasting work circumstances and styles. In consequence,
the material that follows can be seen to simply open for inspection
the ways in which, for prover and reviewer alike, the adequacy of the
initial propositions as an analytic apparatus is itself the achievement
of the local work of a schedule's production and inspection and, therein,
more generally, to open for further inspection the ways in which a
schedule of proofs, as a temporally developing object/organization-of-
practices, provides the background for that schedule's further enuncia-
tion and organization.
To begin the review of the work of developing a proof of the primitive
recursiveness of the divisibility relation, let us start, as an actual
prover might, by writing the conventional definition of the divisibility
relation as
xly <> x =I= 0 and 3n (y = n • x)
where, as the way in which that formula is written and thereby recalled,
a prover will verbalize that formula - cotemporaneously with his
writing it - as saying that x divides y if and only if x does not equal
zero and there exists a number n such that y equals n times x. 71 In
considering this formula as the initial part of a prospective and poten-
tially realizable course of writing that, at the same time, that formula
itself entails, a prover will modify or rewrite it so that y will appear
as an upper bound on the existentially quantified variable n, thus:
xly <> x =I= 0 and 3n EO; y (y = n • x).
That a prover will do this is integrally tied to the recognized utility of
the technique of bounded quantification as a means of constructing
primitive recursive relations. However, that the writing of this modified
formula is tied to the recognized utility of that technique is not the
phemenon of immediate interest; instead, the phenomenon of interest
is the fact that the analyzability of that connection is explicitly worked
out, post facto, as a thematic element of a prover's work in developing
a schedule of proofs. Specifically, in coming to write 3n EO; y (y = n • x)
102
 A SCHEDULE OF PROOFS

as replacing 3n (y = n • x), a prover will identify, over the course of

that writing, those features of the technique of bounded (existential)
quantification that are of pointed relevance to establishing the applica-
bility of that technique to the developing problem-at-hand, therein
simultaneously providing for the prospective work of making analytic
the assertion that 3n .;;;;; y (y = n • x) does, in fact, defme a primitive
recursive relation. In consequence, if a prover, at this point in his work,
were to provisionally72 formulate the technique of bounded quantifi-
cation- as, for example,

if R(x 1 , •.• , Xm, z), is an (m + 1)-place primitive recursive relation of

(x 1 , •.• , Xm, z), then the relation 'defmed' 73 by 3z.;;;;; y (R(x 1 , ••. ,
Xm, z)) is a primitive recursive relation of(x 1 , .•. , Xm, y)

- that formulation would not be tailored 74 so as to immediately

apply to the transformation of the formula 3n (y = n • x) to 3n .;;;;; y
(y = n • x), but instead, it would be tailored so that, in and as the
contrast it provides between its own generality and the particular,
material-specific transformation of 3n (y = n • x) to 3n.;;;;; y (y = n • x),
it provides as well for exactly those details of proving that need to be
checked to insure that 3no;;;;y (y = n · x) does, in and of itself,
accountably define a primitive recursive relation. In this way, whether
or not the technique of bounded quantification is then and there
explicitly formulated, the adequacy of 3n .;;;;; y (y = n • x) as defming
xly as a primitive recursive relation comes to seeably/showably depend
on the satisfaction of two conditions: (1) that y = n • x, in itself,
defmes a primitive recursive relation and (2) that if xly holds between
x andy, then y is, in fact, an upper bound on the quantified variable n.
Concerning the later condition, a prover will differentiate and examine
various 'cases' provisionally adequate to showing that y is such an upper
bound and, as that examination's achievement, will establish the
demonstrable adequacy of those cases for that assertion. 75
Our review of the lived-work of producing the proof
xly ~ 3n.;;;;; y (y = n • x)
has taken us to the point at which a prover will have before himself the
definition
xly ~ x ::#= 0 and 3n.;;;;; y (y = n- x)
as a specifically-textured, developing, provisional proof of the primitive
recursiveness of the divisibility relation, its provisional character being
partially formulated, as the prover's achievement, by the assertion that
if y = n • x defmes a primitive recursive relation, then so will 3n .;;;;; y
(y = n • x). I now want to introduce - both as a means of continuing
103
A SCHEDULE OF PROOFS

the descriptive exegesis of the work of producing the 'final' proof and
as a means of contrasting a prover's and a reviewer's work in relation to
that proof- an example of what, in practice, would be an unexplicated,
situationally-occasioned, locally-efficacious method 76 that a prover
working through another prover's schedule of proofs might employ as a
means of establishing the practical adequacy of that schedule's proofs.
This method, already used in introducing this topic, is this: let a dis-
played formula stand for the assertion that that formula defines either
a primitive recursive function or a primitive· recursive relation (as the
case may be), and let a number enclosed in parentheses stand for the
number of the proposition of that schedule of proofs that either
'justifies' that assertion or that 'justifies' it as following from the
formulas that are displayed above it. Thus, a directed sequence of
such formulas and numbers

A (a)
B (b)

G (g)

is constructed as, and composes, as a practically accountable course

of reasoning and writing, a 'derivation' establishing the last formula G
of the sequence as defming (again, as the case may be) either a primitive
recursive function or relation.
As one last preliminary, let us recall a statement of the technique
of bounded quantification as it might appear in a schedule of proofs
(as, for example, Proposition 8 of the schedule outlined earlier):

Let R be an ( m + I )-place numerical relation and let "iz EO; y (R(x 1 ,

... , Xm, z)) and 3z EO; y (R(xt. ... , Xm, z)) denote the (m + I)-place
relations that hold for (xt. ... , Xm, y) if R(xt. ... Xm,z) holds for all
z EO; y or for some z EO; y, respectively. Then if R is a primitive recursive
relation, so are "iz EO; y (R(x 1, ... , Xm, z)) and 3z EO; y (R(x 1, ... , Xm,
z)).

Now, perhaps as the beginning of a marginal comment through

which the prover intends to preserve for himself a course of reasoning
adequate to the text's assertion, perhaps as the scratch paper work
through which he is searching for such a course of reasoning and
writing, a prover, on coming to the proposition and proof
I04
 A SCHEDULE OF PROOFS

11 The 2-place relation xly is primitive recursive.

Proof: xly = 3n <; y (y = n • x).
while working through another prover's schedule of proofs, might write
y= n • x
3n <; y (y = n • x)
which can be translated as follows: if y = n • x defines a primitive
recursive relation, then 3n <; y (y = n • x) will defme one by con-
struction ·from y = n • x as justified by the technique of bounded
quantification." One of the problems that this partial and anticipatory
derivation exhibits is the problem of justifying that, on the basis of
that particular schedule of proofs, y = n • x defines a primitive recur-
sive relation. But another problem arises as well, whose unanticipated
character resides in the fact that it is by virtue of the material specificity
of the displayed equations that that problem comes to be discovered
and exhibited. The problem is this: although the statement of Propo-
sition 8 that was given above does not explicitly prohibit the use of a
variable (like y) that already occurs in a formula defming a relation R
(like that defined by y = n • x) as the upper bound on the quantified
variable (like n in 3n <; y (y = n • x)), neither does that statement
explicitly permit such a formal procedure. 78
In seeing the difficulty that the notationally-specific derivation
given above makes available - in seeing the problem of justifying on
the basis of the statement of Proposition 8 the writing of 3n <; y
(y = n • x) as a formula properly following the writing of y = n • x
- a prover will (as one possibility) modify or rewrite that derivation
so that it appears as
y= n • x
3n <;w (y = n • x) (8)
using w rather than y as the variable serving as an upper bound for n
since this new transformation, in contrast to the one using y, is now in
accord with the developed relevancies of Proposition 8 an4 where,
parenthetically, the notational device of citing the proposition number
might start to be developed as a situationally and materially motivated
accounting procedure. At the same time, however, that a prover will
come to write 3n <; w (y = n • x) as following y = n • x, and as the
directed character of that writing, a prover will have already seen
that 3n <; y (y = n • x) results from 3n <; w (y = n • x) by symboli·
cally replacing w with y, therein 'anticipating' that transformation as
a prospectively accountable course of writing and reasoning. In this
way, the new problem that comes to be posed, as an integral feature
of the work of producing an accountable derivation, is that of justifying
105
A SCHEDULE OF PROOFS

the writing of 3n :e;; y (y = n • x) as a forrimla properly following

3n :E;; w (y = n • x), a problem that we can represent as
y=n·x ()
3n :E;; w (y = n • x) (8)
3n :E;; y (y = n • x) ( )
Given these circumstances, a prover might thereupon interrogate the
propositions of the schedule preceding Proposition 11 and come to
write/reason as follows: if 3n :E;; w (y = n • x) defmes a primitive
recursive relation R in three variables -that is, if
R(w,y,x)~ 3n:E;;w(y= n • x)
-then
R(y, y,x)~ 3n:E;;y(y = n • x)
is obtained from R(w, y, x) by 'substituting' the variable y for the
variable w. But Proposition 2 of the schedule (asserting that the variables
in a primitive recursive function can be 'identified') seeably/showably
justifies such a substitution,79 and in consequence, the derivation that
was started above can now be accountably written as 80
y=n·x ()
3n :E;; w (y = n • x) (8)
3n :E;; y (y = n • x) (2)
As I indicated earlier, the necessity for developing (as in the deriva-
tion above) the argument that the primitive recursiveness of the relation
defmed by 3n :E;; y (y = n • x) follows by the technique of bounded
quantification from the primitive recursiveness of the relation defined
by y = n • x potentially arises from within a reviewer's materially-
specific attempt to recover the original prover's work as an accountable
course of writing and reasoning. Contrastingly, given the different
material circumstances of the original prover's work in constructing a
schedule of proofs - summarized by pointing to his attempt to produce
a primitive recursive definition, by pointing to his facility in writing
certain formulas and, cotemporaneously, noting to himself the justifica-
tion of such writing as applications of the technique of bounded
quantification, and by pointing to his attempt to produce a primitive
recursive defmition of the divisibility relation specifically and, therein,
to his immediate concern for, and his immediate work in, establishing
the adequacy of y as an upper bound for n in the defmition of xly -
the original prover 'may not give' 81 cfoser scrutiny "to the course of
combined writing and reasoning that assures that the primitive recursive-
ness of 3n :E;; y (y = n • x) accountably follows from that of y = n • x.
106
 A SCHEDULE OF PROOFS

Thus, without specifically reviewing the applicability of the technique

of bounded quantification to that transformation, a prover constructing
a schedule of proofs may come to establish the practical adequacy of
the formula
xly ~xi= 0 and 3n .;;;;; y (y = n • x)
as a primitive recursive defmition of xly contingent only on the
primitive recursiveness ofy = n • x. 82
Let us turn then to the examination of how a prover, having con-
structed a schedule of proofs 'so far,' comes to establish y = n • x
as defining a primitive recursive relation. 83 The prover's problem is this:
he must develop the 'earlier' parts of thi-s-particular-schedule-of-proofs/
the-work-practices-in-which-that-schedule-is-irremediably-embedded so
as to provide an apparatus/the-techniques-internal-to-that-schedule
for adequately analyzing y :;::: n • x as such a definition. To do so, a
prover might begin by seeing, as a consequential manipulation of
signs, that the equation
y= n • x
is formed by joining
y=
with

where, at the same time and intrinsic to that manipulation, the prover
would recognize the equality relation and multiplication as a primitive
recursive relation and function, respectively. However, in addition to
the formal character of this construction, a prover will have already
temporalized that construction as well - that is, n • x is seen as being
placed 'within' the relation 'already' defined by the equality sign.
A prover will then develop this temporal ordering in and as an account-
able course of proving: for example, a prover might begin by writing
K=(x, y)
(where K=(x, y) is intended 'to stand for' 84 the characteristic function
of the equality relation) and by writing after K==(x, y)
x=y,
thereby associating x = y with K=(x, y) and, therein, providing for
the projected analysis of it. 85 What a prover then wants to do is to
'substitute' 86 n • x for y in K=(x, y). Such a substitution, however,
will result in K=(x, n • x) and x = n • x, neither of which corresponds
to y = n • x. An efficacious choice of variables 87 will remedy this
difficulty: by rewriting the equation as 88
107
A SCHEDULE OF PROOFS

y=z
the prover can then 'substitute' n • x for z so as to obtain y = n • x
and K=(y, n • x) and, therein, realize as well the 'temporal' character
of the construction ofy = n • x. Finally, evoking the functional notation
f(n, x) = n • x to clarify the construction, the prover can write

y= n • x K=(y, f(n, x)),

the latter as, prospectively, the characteristic function of the relation
defmed by y = n • x. In that K=(y, f(n, x)) can be seen as being
constructed from the primitive recursive functions K= and f by sub-
stitution, K=(y, f(n, x)) already seeably/showably defines a primitive
recursive function, and all that remains in order to show that y = n • x
defines a primitive recursive relation - as a recognized, practical require-
ment arising from within the way in which K=(y, f(n, x)) was con-
structed -is to check that K=(y, f(n, x)) is, in fact, the characteristic
function of y = n • x. By working backward through the formulas,
a prover will perceptually organize and 'rehearse' those formulas,
as an accountable course of reasoning, so as to be able to read
K=(y, f(n, x), as 'saying that if y = n • x, then K=(y, f(n, x)) = 1
and if y =F n • x, then K=(y, f(n, x)) = 0. 89 Having in this way insured
that K=(y, f(n, x)) is the characteristic function of y = n • x, the
prover has demonstrated the primitive recursiveness of y = n • x
by making that equation analytically available as defining such a
relation from within a 'completed' course of work.
As the reader will recall, the question of the primitive recursiveness
of y = n • x was addressed as the projected last step in demonstrating
the primitive recursiveness of the relation defined by 3n ~ y (y =
x
n • x). With the primitive recursiveness of y = n • now established,
a prover (or a reviewer) will see that he could construct (or, respectively,
has found and could preserve by writing) a 'unidirectional' or 'linear'
derivation of 3n~y (y = n • x) ad defining a primitive recursive
relation.
In a short while I will give such a derivation, and I will thereby
complete the derivation that I had started earlier in the discussion.
However, before I do this, I want to review a slightly different way of
showing that y = n • x defines a primitive recursive relation. The reason
for giving this alternative construction can be developed as follows:
first, in that the characteristic function of the equality relation is
primitive recursive by a previous construction, 90 the construction of
y = n • x given above can be formulated as, simply, the substitution
of a primitive recursive function, n • x, for a variable, z, in a primitive
recursive function, K=(y, z), and this construction preserves primitive
recursiveness by the definition of the class of primitive recursive
functions. Second, on coming to a problem similar to the one for
108
 A SCHEDULE OF PROOFS

y = n • x as, for example, the problem of showing that x =I= 0 defines a

primitive recursive relation, a prover will find, in the case ofx =I= 0, that
the solution to this problem consists 'first' - as an accountable temporal
organization of work -of realizing that 0 can be written as Z(y), the
primitive recursive zero function, and by so writing 0, 'then' of seeing
- as the result of the seeable/showable availability of a course of
argumentation appropriately mimicking the one developed for y =
n • x - that K*(x, Z(y)) can be similarly 'derived' as the character-
istic function of x =I= 0. Thus, even though the prover would fmd an
immediate need for a construction similar to that used for y = n • x,
the increasing familiarity and naturalness of the work involved in such
constructions, the (now) obviousness of those constructions as an
accountable line of argumentation, and the fact that, as an extractable
line of argumentation, it is already enunciated in the definition of the
class of primitive recursive functions, all offer no motivation to a prover
for articulating the construction of y = n • x as a separate technique
of constructing primitive recursive relations. 91 In contrast, the method
=
of construction y n • x that I will give now does exactly the opposite;
it suggests - in and as the problematic details of its own accountable
production - the need for formulating just such a technique as a
separate proposition of the developing schedule of proofs.
Let us suppose that a prover (or a reviewer), after seeing that the
insertion of n • x in y = _ provides a potentially accountable con-
struction of y = n • x, began the work of proving that y = n • ~ defines
a primitive recursive relation by naming that relation T,
T(y, n, x) ~ y = n • x.
Given this as the material origins of his work -and, in particular, the
fact that he has associated a 'name' T with the relation defined by y =
n • x - a prover might write next
E(y,x)~y= z
and
f(n,x)= n • x,
thereby 'naming' the equality relation and multiplication E and f,
respectively. Then, as the now recognized and realized directed course
of his writing, the prover can construct y = n • x as
T(y, n, x) ~ E(y, f(n, x)).
With the formulas
T(y, n, x) ~ y = n • x
E(y, z)~ y= z
109
A SCHEDULE OF PROOFS

f(n, x) = n • x
T(y, n, x) ~ E(y, f(n, x))
before the prover - at least as a perceptually organized course of work
-the 'question' that arises is that of what to do next. More specifically,
in coming to write
T(y, n, x) ~ E(y, f(n, x)),
the prover has not only 'named' the component parts of that formula,
but he has given the equation y = n • x an incipient analytic structure.
Part of the prover's search for the next thing to do consists of reviewing
the construction of T to see if, in elucidating that incipient structure,
that structure can be made explicit as part of a proof of the primitive
recursiveness of y = n • x.
What a prover will do next is to write
KT(y, n, x) = KE(y, f(n, x)),
and in writing this equation, and in seeing it as an appropriate thing
to write at this point, a prover will come to recognize, as that equation's
developingly 'self-explicated gestalt, that what he must do now is to
show, in that KE(y, f(n, x)) apparently92 equals KT(y, n, x), that
KE(y, f(n, x)) defmes a primitive recursive function. (As part of that
'gestalt,' a prover will justify the fact that this is what he must do by
reconstructing for himself -through the use of the material detail of
his course of writing - the fact that a relation R is, by definition,
primitive recursive if its characteristic function KR is and, therefore,
that Twill be primitive recursive if KT is.) Thus, in writing
KT(y, n, x) = KE(y, f(n, x)),
a prover has already envisioned that formula as part of this-yet-to-be-
completed and, therein, conjectured course of accountable work from
within which it will (prospectively) take on its character, and be estab-
lished, as a naturally accountable part of a demonstration of the
primitive recursiveness of y = n • x.
Among the things that the prover has done, one of the things that
he will recognize that he has done is to have formulated the original
problem of showing that the relation defined by y = n • xis primitive
recursive as the problem of showing that the construction of the
relation T as
T(y, n, x) ~ E(y, f(n, x))
exhibits T as such a relation. Moreover, in that this realization is tied
to the prover's having written (or to his having envisioned the writing of)
KT(Y, n, x) = KE(y, f(n, x)),
110
 A SCHEDULE OF PROOFS

the prover will not only undertake his ensuing work as that of finding
this problem's solution, but he will have already found the material
origins of that work in the formula
KT(y, n, x) = KE(y, f(n, x))
itself. That is, in the way in which the prover has come to formulate
the problem of showing the primitive recursiveness of y = n • x as the
problem of showing the primitive recursiveness of the relation T
defined by
T(y, n, x) * E(y, f(n, x)),
the prover already has a prospective way of solving that problem
through the examination of the equation
KT(y, n, x) = KE(y, f(n, x)).
Now, in his previously corning to write the formula
T{y, n, x) * E(y, f(n, x)),
a prover will have already identified E and f as, respectively, a primitive
recursive relation and function. In consequence, the primitive recursive-
ness of E and f is now available to the prover as a witnessible and
remarkable feature of
KT(y, n, x) = KE(y, f(n, x)).
In that this is so, a prover will find, in and as the organizing work of
'seeing,' that
KT(y, n, x) = KE(y, f(n, x))
exhibits 93 KT as being constructed by the substitution of a primitive
recursive function, f(n, x) = n • x, for a variable, z, in the primitive
recursive function KE(y, z), the ·characteristic function of the equality
relation. Thus, in that the construction of KT is now available as a
practically accountable instance of the general process of defining a
primitive recursive function by substitution, the formula
KT(y, n, xJ = KE(y, :f(n, x))
has come to seeably/showably exhibit the primitive recursiveness of
KT itself.
As the reader will recall, a prover, working to establish the primitive
recursiveness of y = n • x in the fashion that I described first, as an
endogenously articulated feature of that work, needed to check that
the function K=(y, f(n, x)) was the characteristic function of y =
n • x. From within the present way of working a different problem
arises: Earlier in writing
KT(y, n, x) = KE(y, f(n, x))
111
A SCHEDULE OF PROOFS

beneath
T(y, n, x) ~ E(y, f(n, x)),
a prover would have seen in that arrangement, as the proper temporality
of that arrangement, that that arrangement can be seen as one indicat-
ing a relation of proving - that is, that
KT(y, n, x) = KE(y, f(n, x))
'follows from'
T(y, n, x) ~ E(y, f(n, x)).
And in seeing that this was so, the prover would have also realized that
he needed, as part of his current work, to establish that the one did,
in fact, 'follow' from the other. So far, the prover has only shown
that if
KT(y, n, x) = KE(y, f(n, x)),
then KT is a primitive recursive function. Thus, given the material
development of the problem itself, the problem that remains is that of
showing that
T(y, n, x) ~ E(y, f(n, x))
and
KT(y, n, x) = KE(y, f(n, x))
can be embedded in a course of work such that they stand in the
relationship of the latter being provable from the former.
Returning to this problem after establishing KT as a primitive
recursive function, a prover will see, as the incipient part of a solution
to it, that the notation T(y, n, x) can be read as (and, in fact, can now
be retrospectively recalled as an abbreviation for)
(y,n,x)ET,
i.e., that the 3-tuple (y, n, x) is a member of the relation T. In that
the prover sees this, he is able to read the formula
T(y, n, x) ~ E(y, f(n, x))
as saying that
(y, n, x) E T ~ (y, f(n, x)) E E, 94
therein providing for the following line of argumentation: since the
characteristic function of an m-place relation R will be 1 if (x 1 , •.• ,
xm) E R and will be 0 if (x 1 , ••• , Xm) f1. R, KT(y, n, x) will equal
1 or 0 as (y, n, x) is or is not a member ofT, as (y, f(n, x)) is or is not
a member of E, as KE(y, f(n, x)) equals 1 or 0. 'Thus' -that is, as
112
 A SCHEDULE OF PROOFS

the naturally accountable consequence of this reasoning -

KT(y, n, x) = KE(y, f(n, x)),
and in that 'KT(y, n, x) equals KE(y, f(n, x))', as the-proofs/the-
work-of-the-proofs own accomplishment in developing what needed to
be shown about y = n • x to insure that it defines a primitive recursive
relation, the demonstration that y = n • x defines a primitive recursive
relation is thereby brought to a close.
We have now developed a second WC\Y of establishing the primitive
recursiveness of y = n • x. As I suggested earlier, this alternative way of
working makes available the following contrast: from within the
argument-specific work of the earlier analysis of y = n • x, a prover will
find no apparent, material motive for articulating the demonstration of
the primitive recursiveness of y = n • x as a separate technique of
constructing primitive recursive relations. In contrast, a prover working
in the second manner, in coming upon the problem of showing that the
construction of
T(y, n, x) <* y =n • x
as
T(y, n, x) <* E(y, f(n, x))
preserves the primitive recursiveness of E and f, comes on that problem
as a problem in its own right for which he then (in the sense of temporal
succession) finds its solution as a discovered course of writing and
reasoning. In consequence (but possibly a consequence realized in
coming to see and work out, liS a recurring argument, the accountably-
same construction in subsequent parts of the schedule), a prover might
formulate 95 this construction of y = n • x as itself an extractable
course of reasoning (Proposition 7):
( ) The relation S obtained by substituting a primitive recursive
function f for a variable in a primitive recursive relation R is
primitive recursive.
And, by extrapolating from the particular proof for y = n • x, a prover
can obtain that proposition's proof:
Proof: Let S be the relation defined by
S(x1, · · · ,:x;__1, Yt. · · · • Yn,X;_+1• · · · • xm)<*
R(xl' ... 'xi-1' f(yl' ... 'Yn), xi+1' ... 'xm).
Then 96
Ks(xl • · · · • X;.-1 • Y1 • · · · • Yn• X;.+1 • · · · • xm) =
KR(xl • · · · • Xi-1 • f(yl • · · · • Yn), X;.+1• · · · 'Xm).
Hence, if R and fare primitive recursive so is S.
Finally, we should note that if the 'original' schedule of proofs did not

113
A SCHEDULE OF PROOFS

include this proposition, then a reviewer working in a manner similar

to that just described, in coming to find the need to account for the
construction of T as a primitive recursive relation, would find the
faulted character of the schedule and, therein, would find as well the
need to develop and add such a proposition to it.
Let us now return to the developing material proof of the primitive
recursiveness of the divisibility relation,
xly * x * 0 and 3n ~ y (y = n • x).
In writing this equation, in and as the way in which the writing of it
is embedded in the work of developing the schedule of proofs, a prover
will have already noted that
x =F. 0 and 3n ~ y (y = n • x)
*
consists of 'x 0,' 'and,' and '3n ~ y (y = n • x)' and that if 'x 0' *
and '3n ~ y (y = n • x)' define primitive recursive relations separately,
*
then so will the conjunction 'x 0 and 3n ~ y (y = n • x).' 97 Further-
more, as we have seen above, in developing a sketch-proof that 3n ~ y
(y = n • x) defines a primitive recursive relation, a prover must 'first'\IIS
show that y = n • x defines a primitive recursive relation as well. The
point to be made now is that once a construction of y = n • x as a
primitive recursive relation has been given, an analogous construction

*
*
can be given for x 0 and, conversely, once a construction exhibiting
x 0 as defining a primitive recursive relation has been given, a similar
construction for y = n • x follows from it: thus, by seeing that 0 can
be interpreted as the zero function Z(w) = 0, a primitive recursive
function by definition, the prover, in a fashion analogous to the con-
struction of y = n • x, will cotemporaneously see (as a seeably writable
*
course of argumentation) that x 0 can be constructed by substituting
Z(w) for y in x = y. 99 This done - and, moreover, this establishing
the practical accountability of the last relevant detail of
x ly * x * 0 and 3n ~ y (y = n • x)
as the relevance of the details of that fonnula has itself been produced
and identified over the course of the formula's construction and inspec-
tion - the prover can finally assert, as part of the developing schedule
of proofs,
( ) The 2-place relation xly is primitive recursive.
Proof: xly * x * 0 and 3n ~ y (y = n • x)
as the practically adequate statement and proof that the divisibility
relation is primitive recursive.
Similarly, let us now return to the reviewer's examination of a proof
of the primitive recursiveness of the divisibility relation as he comes to
114
 A SCHEDULE OF PROOFS

that proof while working through another prover's schedule of proofs.

For reasons that will be given shortly, the proof that will appear in a
'finished' schedule will not be the one given above, but, instead, it will
most likely exclude the condition that x =I= 0:
xly~ 3n~y(y = n • x).
Now, as I indicated at the beginning of this topic, a reviewer, working
in the manner that I have described, in and as that work, may compose
the following course of writing and reasoning as a derivation of the
primitive recursiveness of 3n ~ y (y = n • x): 100
y=z (5)
n·x (3)
y= n • x (7)
3n ~ w (y = n • x) (8)
3n ~ y (y = n • x) (2)
thereby not only finding and exhibiting 3n ~ y (y = n • x) as defining
a primitive recursive relation but finding and exhibiting, as something
found and exhibited in course, the adequacy of the initial propositions
of the schedule in providing such an analysis. In completing such an
analysis (after having already examined the displayed formula
xly ~ 3n ~ y (y = n • x)
as a definition of the divisibility relation), the reviewer finds, therein,
as the closure of his work in reviewing that proof, the practical adequacy
of the material proof in-itself as that proof appears in the schedule.
As the reader will recall, I began the discussion of this topic by
using the very same analysis of the construction of 3n ~ y (y = n • x)
to point out that the initial propositions of a schedule of proofs (Propo-
sitions 1-10 of the schedule in which we are working) provide an
apparatus for analyzing the proofs of the subsequent propositions of
that schedule as defining primitive recursive functions and relations.
The review of the lived-work of actually constructing and inspecting
a proof of the primitive recursiveness of xly then gave us access, not
to the adequacy. of that apparatus in and of itself, but, instead, as
real-worldly researchable phenomena, to how the practical adequacy
of that apparatus is tied to the local work of constructing or reviewing
such a schedule and to how that schedule-specific apparatus is itself
developed from within that local work. Furthermore, the preceding
discussion has indicated both how and the way in which a prover
comes to articulate and, thereby, formulate as separate propositions,
just the techniques that he needs to construct the later proofs of
the schedule as seeably/showably defining primitive recursive functions
115
A SCHEDULE OF PROOFS

and relations. In summary, one aspect of the local work of developing

the schedule should, perhaps, be emphasized: rather than the initial
propositions being constructed in a step-like, even if retrospective/
prospective manner - for example, the need for a particular proposition
occasioning that proposition's formulation and retrospective placement
in the schedule - what the preceding material makes available is that a
schedule is always and unavoidably further articulated and organized
from within its development as a schedule-of-proofs/a-way-of-organizing-
proving's-practices constructed 'so far.' In this way, the work of devel-
oping the schedule can be seen to be much more of a continual shaping
which finds, over its course, the adequacy of, the recurring adequacy
of, and the need to modify, both that schedule's material presentation
and, inseparably, that schedule's developingly and endogenously
articulated, proof-specific practices of proving than a construction
of various elemental pieces that are arranged so as to make up the whole.
Before leaving this discussion, one last point needs to be made.
A contrast has already been provided between the proof of the primitive
recursiveness of the divisibility relation as that proof was obtained in
the preceding discussion,
xly <* x * 0 and 3n..;;; y (y = n • x),
and the proof of that fact as it will more likely appear in a 'finished'
schedule of proofs,
xly <* 3n...: y (y = n • x).
*
The exclusion of the condition x 0 in the latter may strike the reader
as curious in that, as we have seen, the definition
xly <* x * 0 and 3n..;;; y (y = n • x)
more closely conforms with the conventional definition of the divisib-
ility relation and, second, as we have also seen, the inclusion of the
condition x * 0 does not affect the primitive recursiveness of the
definition. The point is this: a prover, in consulting other, previously
constructed schedules of proofs while constructing his own (or a
reviewer, in working through another prover's schedule) will similarly
question the absence of this condition. The inquiry that is thereby
initiated finds that the exclusion of the condition x * 0 from the
definition of xly really only adds one element 010 to that relation.
On then inspecting the subsequent propositions and proofs of the
schedule that are practically available as being dependent on the defini-
tion of xly - like that of prime(x), for instance - a prover (or reviewer)
will find that the inclusion of 010 in the definition of division is, in
fact, practically irrelevant to the. schedule of proofs. In this way, then,
as the provers before him, a prover will drop the restriction that x
not equal 0 from his proof of the primitive recursiveness of x ly.
116
 A SCHEDULE OF PROOFS

(c) The correspondence between the propositions of a schedule of

proofs and the syntax of formal number theory as an achievement
of the schedule ofproofs itself
A schedule of proofs is typically organized in the following manner:
an initial group of propositions ( 1-10 in the schedule of proofs outlined
earlier) introduces those elementary primitive recursive functions and
relations and those techniques of constructing primitive recursive
functions and relations that are needed for the proofs of the later
propositions of the schedule; a second group (I 1-16) then supplies
an apparatus for working with the proof-specific G6del numbering,
and a·final group of propositions (17-35) constructs the syntax of the
particular logistic system under examination - that syntax being
'arithmetized' by the G6del numbering - as consisting of primitive
recursive functions and relations. Now the analyses of topics (a) and
(b), although they pointed to, and were suggestive of, pervasive features
of the lived-work of selecting and arranging the propositions of a
schedule, they did this by attending particularistically to the work
of producing the propositions and proofs of the first two of these
three groups. The aim of the present topic is first to formulate, on
behalf of the reader, a criticism concerning the applicability of those
analyses to the work of developing the arithmetized syntax and, then,
to reply to that criticism.
Earlier in the book I indicated that the proofs of the primitive
recursiveness of the functions and relations of the arithmetized syntax
were ambiguous in the following respect: one of a prover's problems
in developing or reviewing a schedule of proofs is to show (or to verify)
that the numerical functions and relations of the arithmetized syntax
are, in fact, primitive recursive, and this amounts to showing (or verify-
ing) that, in defining such functions and relations, those definitions
provide a seeable/showable one-to-one correspondence between, for
example, the elements of term(x) and the terms of P. Thus, although
it is possible to provide abstractly for the collection of elements of such
a relation independently of the exhibition of that collection's primitive
recursiveness, in practice, at the same time that a prover is working
out the demonstration of the primitive recursiveness of that collection,
he is doing this by articulating the definition of an arbitrary member of
it (whose arbitrariness is itself tied to the self-same demonstration).
In this way, then, as lived-work, the proof of the primitive recursiveness
of a component of the arithmetized syntax is the same as the primitive
recursive definition of it. As a means of referring to the ambiguity of
the proofs of a schedule as both proofs and definitions (an ambiguity
that is formally avoided in the written presentation of the schedule),
I now want to introduce and use the term 'definition/proof.'
As the reader will recall, I argued earlier that a prover's ability to
write and inspect the definitions/proofs of a schedule was tied to the
117
A SCHEDULE OF PROOFS

prover's familiarity with the abbreviatory practices/practical techniques

that are used in working with primitive recursive functions and relations.
I later argued, in point 1 of the present chapter, that a Godel number-
ing not only 'arithmetizes' the syntax of a formal language, but that,
in doing so, it initiates - and, as a technique of proving, is itself an
integral part of - the project of demonstrating that that arithmetized
syntax can be defined as, and, thus, that it consists of, primitive recur-
sive functions and relations. The notion of the 'directed' character
of a schedule of proofs was then introduced to refer to this project
as it unfolds in and as the work of producing a schedule.
The interrelatedness of these three themes - that the surrounding
techniques of working with primitive recursive functions and relations
permit the practically accountable writing and inspection of the proofs
of a schedule, that a Godel numbering is itself a technique of proving,
and that a schedule of proofs has a directed character - lies in the
fact that the directedness of the work of producing a schedule, as that
directedness is realized over the course of developing the schedule
itself, is made possible by, and is realized from within, the increasingly
articulated techniques of working with primitive recursive functions,
relations, and the Godel numbering. Moreover, once the propositions
and proofs concerning the arithmetized syntax have been developed
and arranged, as part of their development and arrangement as part of a
schedule of proofs (or once a reviewer has worked through a schedule
for a first time), a prover's (or a reviewer's) cultivated familiarity with
those techniques then insures that the prover (or the reviewer) can
work through the schedule, once again, in just such an accountably
sequentialized manner, at just such a pace, with just those proof-
specific material details, so as to exhibit, and, therein, to realize, as a
procedural way of working, the propositions concerning the arithmetized
syntax as having been constructed in correspondence with, and by
mimicking, the constructive and hierarchical definitions of the syntax
of the logistic system as those definitions were specified prior to and
independently of the construction of the schedule itself. In this way,
then, the correspondence between the definitions of the schedule and
the definitions of the 'original' syntax - as just this accountable corres-
pondence, as a property of both the syntax and the schedule - is
made available by, and is the achievement of, a schedule of proofs as
a schedule-of-proofs/the-naturally-accountable-work-of-its-production.
Now the reader, by turning away from the natural accountability
of a schedule as a local achievement, may have reasoned as follows:
in that the definitions of the syntax of a formal language are 101 specified
prior to the construction of a schedule, and in that the constructive
and hierarchical character of those definitions demands that a prover
would have already worked out a proper, sequentialized ordering of
them, it should then follow that the order of the propositions of a
118
 A SCHEDULE OF PROOFS

schedule concerning the arithmetized syntax is a necessary order

available to a prover prior to the development of that arithmetized
syntax; that while the proofs of those propositions still need to be
constructed as materially-specific proofs, a prover need only work
through a prearranged list of propositions corresponding to the pre-
viously available definitions of the syntax of the language in question,
and, therefore, that the surrounding techniques of working with primi-
tive recursive functions and relations allow the prover to work through
the proofs of those propositions as a pre-formulated sequence of work.
Through this reasoning, then, the reader may have seen a way of
arguing that the 'directedness' of an-always-particular schedule of
proofs - now interpreted as the programmatic fulfillment of a pre-
established correspondence between a schedule of proofs and the
syntax of a logistic system - is only dependent on the lived-work
of producing that schedule as the means through which such a schedule
is constructed, that that work relies on the correspondence between a
schedule of proofs and the syntax of a logistic system as an objective
relationship between natural, mathematical objects, and, therein,
and more generally, that a schedule's discovered and demonstrable
orderlinesses are, in fact, properties of a transcendental, objective
proof of Godel's theorem that the particular proof represents and to
which the particular proof provides access. Thus, a reader's criticism
of topics (a) and (b) may be this: that, at least for the part of a schedule
of proofs concerning the syntax of the formal system in question, the
orderliness of a schedule is not irremediably tied to the local practices
of their discovery/exhibition, but, instead, speaks on behalf of a trans-
cendentally existent and objective proof of Godel's theorem.
As a means of addressing this criticism, I want to begin by bringing
together several preliminary observations. These observations attend
to the actual details of a schedule and the work of its production as
those details provide a contrast between the syntax of a formal system
and its 'representation' (via a Godel numbering) in a schedule of proofs.
By doing so, these observations offer, as a radical problem, not the
problem of demonstrating that such a correspondence between syntax
and schedule does not exist, but, instead, the problem of showing what
such a correspondence consists of as praxis.
The first observation is this: given the practical exigencies of con-
structing a schedule of proofs, a prover may come to introduce certain
functions and relations into the schedule that do not correspond to
defmitions used in specifying the original syntax of the formal language.
One example from the schedule of proofs in which we are working
is that of the function n .i-+ g(x 0 ). As we shall see below, this function
is introduced into the schedule partly to clarify the defmition/
proof of var(x) that immediately follows it and partly, in and as the
way in which it clarifies that definition by maintaining the sequential
119
A SCHEDULE OF PROOFS

character and pace of the work of working through (once again) the
propositions and proofs of the schedule, to provide easier access to and
to exhibit the accountable orderliness of the schedule. Another
example is provided by the introduction of the function sub(x, n, a).
sub(x, n, a), which specializes and modifies the arithmetized counter-
part Sub(x, t, a) of the syntactically-defmed operation of substitution,
is introduced into the schedule specifically for its later use in the
diagonalization /'proof.
Second, in developing a schedule of proofs, a prover may find the
need - again, in and as the practical exigencies of constructing that
schedule - to articulate further the formal system itself by introducing
syntactic definitions that were not seen to be needed for that system's
original specification. An example of such an addition from within the
schedule in which we are working is the definition and the proof of the
primitive recursiveness of the (arithmetized) class offormation sequences
of terms. Similarly, in topic (d) below, we will see that the attempt to
modify the schedule of proofs in such a way that it will be applicable
to extensions of our original system P necessitates 102 the definition
(and the proof of the primitive recursiveness) of the function arg(x)
giving the number n of terms that follow a function symbol f with
Godel number x for the concatenation of that function symbol, a left
parenthesis, n terms, and a right parenthesis, in that order, to itself be
a term of the language. The reader may well argue, regarding these
examples, first, that the notion of a formation sequence of terms
merely articulates the inductive definition 103 of the class of terms
in a manner compatible with the Godel numbering/the-techniques-of-
working-with-that-numbering, and, second, that the notion of a function
mapping a function symbol f to the number of argurnents 104 that it
takes is implicit in the definition of the function symbols as a class
of primitive symbols. 105 The immediate point, however, is that the
need to define the class of formation sequences of terms and the
function arg(x), and the need to include those definitions in the
schedule of proofs, arise from within the work of that schedule's
construction.
Third, the placement and fitting of the 'new' components of the
arithmetized syntax in a schedule of proofs are bound to the developing
orderliness of a prover's work in constructing that schedule as the work
of producing a sequentialized series of hierarchically building definitions/
proofs whose construction (i.e., whose accountable work of proving)
programmatically corresponds, as its achievement, to the exhibitedly-
prior ·specification of the 'original' syntax; that is, that placement
and fitting are bound, not to the transcendentalized schedule whose
idealized, formal properties are merely adequate to a similarly trans-
cendentalized proof of Godel's theorem, but to the naturally account-
able mathematical object - now understood as the pair the-material-
120
 A SCHEDULE OF PROOFS

proof/the-practices-of-proving-to-which-that-proof-is-inseparably-tied -
that the developing schedule of proofs is coming to be. Thus, although
the need for a separate proposition concerning the function n ~ g(xn)
could have been circumvented by including the proof of that function's
primitive recursiveness in the definition/proof of var(x) as follows:
( ) Let var(x) be the relation that holds if and only if x is the Godel
number of a variable. var(x) is a primitive recursive relation.
Proof.· 106
var(x) ~ 3n..:;; x (x = g(xn))
where g(xn) is primitive recursive by theformula
n
g(xn) = g((x) IT Pf * g( ))
*
i=O
thereby giving the proof of the primitive recursiveness of var(x) a
nested structure, or, alternatively, the introduction of g (xn) could
have been entirely avoided by simply defining/proving-the-primitive-
recursiveness-of var{x) as

var(x) ~ 3n..:;; x (x = g((x) * .~ p( * g()))

1=0
thereby including the formula for g(xn) in the defmition/proof of
var(x) itself, these arrangements set in relief the fact that the following
organization of the schedule maintains and exhibits the sequential
and sequentially-paced character of the work of its proving: 107
17 The function g(xn) giving, for each n, the Godel number of the
n-th variable, is primitive recursive.
Proof.·tos
n
g(xn) = g((x) * rr p( * g( )).
i=O
18 Let var(x) be the relation that holds if and only if xis the Godel
number of a variable. var(x) is a primitive recursive relation.
Proof."
var(x) ~ 3n ..:;; x {x = g{xn)).
The inclusion and positioning in the schedule of the definition of a
formation sequence of terms and the definition and proof of the
primitive recursiveness of its arithmetized counterpart, formterm(x),
provide a second illustration of the ways in which a prover will arrange
the propositions and proofs of a schedule so as to evince the program-
matic character of the correspondence between the original definitions
of the syntax and the primitive recursive arithmetization of them under
the proof-specific Godel numbering. As the reader will recall from the
121
A SCHEDULE OF PROOFS

discussion of Godel numbering as a technique of proving, the introduc-

tion of the relation formterm(x) into the schedule -and, simultaneously,
the need for explicitly defining the notion of a formation sequence of
terms - was avoided by 'including' the definition of formterm(x)
in the definition/proof of term(x) thus:
L(y)
term(x) <>X-=/= 0 and 3y ~ ll pf [(y)L(y) = x) and
i=O
Vi~ L(y) ([i = 0] or [(y)i = g(O)] or
[var((y)i)] or 3j ~i ((y)i = g(S() * (y)j * g( )))
or 3j ~ i 3k ~ i ([(y)i = g(+() * (y)j * (y)k * g( ))]
or [(y)j = g(·() * (y)j * (y)k * g( m)))
In that previous discussion, in the way in which the definition/proof of
term(x) was constructed- and, specifically, in that both the notion of
a formation sequence of terms and the definition of sequence numbers
encoding such sequences were developed so as to permit that construc-
tion - the practical adequacy of that formula for term(x) as (1) defining
term(x) and (2) defining it as a primitive recursive relation was already
part of the intelligibility of the formula. In the present context,
however, in that the formula for term(x) has been isolated from the
work of its construction, that formula loses its immediate availability
as, and naturalness in, providing for and demonstrating the corres-
pondence between the elements of term(x) and the sequences of
primitive symbols defined as the terms of the formal system P. 109
Given this circumstance - or, rather, as the practical and naturally
available requirements of composing a recognizably adequate schedule
of proofs that this formula helps exhibit - a prover, 'instead' of giving
the proof of the primitive recursiveness of term(x) as above, will more
likely work up to that proof in a manner similar to the following, 110
Define a formation sequence of terms as a finite sequence of
terms t 1 , ••• , ti, ... , tm having the property that, for each i,
i = 1, ... , m, one of the following conditions holds:
(i) ti is 0
(ii) ti is an individual variable
(iii) ti is S(tj) for some j < i,
(iv) ti is +(t/k) for some j < i, k < i
(v) ti is {tlk) for some j < i, k < i.
() formterm(x), holding if and only if x is the Godel number of
a formation sequence of terms, is a primitive recursive relation.
Proof·
formterm(x) <:>xi= 0 and xi= 1 and
122
 A SCHEDULE OF PROOFS

Vi~ L(x) {(i = 0) or ((x)i = g(O)) or

(var((x)i)) or
3j < i[(x)i = g(S() * (x)j * g( ))] or
3j < i ;lk < i [(x)i = g(+() * (x)j * (x)k * g( ))] or
3j < i 3k < i [(x)i = g(·() * (x)j * (x)k * g( ))]
() Let term(x) hold if and only if xis the GOdel number of a term
of P. term(x) is a primitive recursive relation.
Proof: L(y)
term(x) ~ 3y ~ ll pf {formterm(y) and [(y)L(y) = x]} .
i=O
And, finally, as one last example of this aspect of the structuring
of a schedule, let us consider the placement in a schedule of the propo-
sition concerning sub(x, n, a). Even though sub(x, n, a) is not used in
the proofs of the later propositions of a schedule, even though it is
needed only for the diagonalization/'proof, and even though it does
not directly correspond to a previously defined syntactic function, the
definition/proof of sub(x, n, a) immediately follows that of Sub(x, t, a),
the arithmetical counterpart of the syntactically defined operation of
substitution, that being the recognizably proper place, as part of the
produced and identifying structure of a schedule, for the work of
sub(x, n, a)'s introduction.
A fourth observation is this: neither the ambient techniques of
working with primitive recursive functions, relations, and a proof-
specific G6del numbering, nor the abstract correspondence that a
G6del numbering provides between the language of a formal system
and the natural numbers, make obvious or automatic the ways in
which the various definitions of the syntax of the formal system can,
under that particular G6del numbering, be formulated and exhibited as
primitive recursive functions and relations; on the contrary, for a
prover engaged in constructing the seeably/showably primitive recursive
arithmetizations of the necessary syntactic objects for a schedule of
proofs, the arithmetization of each such object is its own technical
problem. 111
Fifth, over the course of constructing a schedule of proofs a prover
will not maintain a strict reliance on what becomes retrospectively
available as the pre-existent, pre-established specification of the syntax
of the logistic system in which he is working; rather, as a means of
making that construction, a prover will re-examine, modify and possibly
restructure the syntax itself: a prover, in working out the definitions/
proofs of the arithmetized syntax, may discover various aspects of the
original syntax that were not, but now need to be made explicit, like
that of the notion of a variable occurring in a formula or of it having
123
A SCHEDULE OF PROOFS

no free occurrences in a formula; a prover may come to consider

whether or not commas should be included as primitive symbols or to
review the way in which the functions symbols were defined and the
precision required for that definition; a prover may even come to
examine, in the presence of the adverse material detail of his current
way of working, the efficaciousness and perspicacity of using a different
notational system or a different set of axioms and rules of inference.
In summary then, the specification of a formal system can be, and in
fact is, continually evaluated over the course of constructing a schedule
of proofs and can be and is changed if and when the need arises. 112
Moreover, that this is so points to the fact that the accountable detail
of the syntax of a logistic system, for a prover, is irremediably tied
to the embeddedness of those details in a structure of proving's practices
and it is by recourse to that embeddedness that a prover can modify
and restructure the 'original' syntax. As we shall see in the chapter 'A
Structure of Proving,' a prover's ability to determine the class of formal
systems for which Godel's theorem holds becomes available to him
through the inspection and analysis of a particular, materially-present
proof of that theorem to find what can be modified or changed in the
specification of the syntax for that particular proof to remain a proof
of Godel's theorem for that modified system. 113
At this point, given the preceding observations, the reader may now
see himself as being invited to suppose that the 'directed' character of
a schedule of proofs - rendered as the: sequentialized, programmatic
construction of the arithmetized syntax in correspondence with a pre-
established hierarchical construction of the pre-existent syntax of a
logistic system - is an 'entirely' produced feature of the work of a
schedule's construction and, thus, that the notion of the directedness
of the lived-work of producing a schedule is only a retrospectively
available attempt to provide an interior orderliness for that lived-work.
One last observation must, therefore, be made: although the sequential
and programmatic character of a schedule's construction, as an account-
able feature of a finished schedule of proofs, is a thematic concern of
a prover in constructing a schedule, although that sequential and
programmatic character is specifically worked out over the course of
a schedule's construction, the lived-work of that construction maintains
and is informed by the directedness and the developing realization of
the programmatic character of a schedule over the entire course of a
schedule's construction.
Given this last observation, it would seem that we have been led to
a paradox - on one hand, we have seen that the properties of a schedule
of proofs are essentially tied to the local work of a schedule's produc-
tion and review; on the other, it is nevertheless the case that over the
course of that local work, that work retains its sense as the working
out of an objectively and transcendentally ordered course of work
124
 A SCHEDULE OF PROOFS

that that self-same work exhibits and to which that work provides
increasingly technical access. In this way, then, these observations
set in relief, and can now be reviewed in light of, the following
phenomenon: 114 it is over the course of and through the work of
working out the definitions/proofs of the arithmetized syntax, in
and as the mutually discovered and produced compatibility of those
defmitions/proofs with the techniques of working with primitive
recursive functions and relations, that a prover finds, as a process
of rediscovery/construction, the accountable orderliness of the original
specification of the formal syntax; that temporally developing re-
discovery/construction, in the ways in which it is tied to the lived-work
of organizing a schedule as an accountable order of proving, uncovers
the orderliness of the original specification of the syntax in greater
detail and as a technical, mathematical (and, therein, a mathematically
analyzable) object; in consequence, in that the orderliness of the
original definitions of the syntax of a formal system are re-achieved
through the work of developing the schedule, the work of a schedule's
construction is given its continuing sense and direction as the discovery
and the working out of a prior, objective, and transcendentally-ordered
course ofwork.

(d) A review of the work ofgeneralizing a schedule of proofs so as to

elucidate the character of the development and organization of a
schedule as a radical problem, for the prover, in the production of
social order
On various occasions in the analysis of topics (a), (b) and (c), we have
been reminded that the lived-work of selecting and arranging the
propositions of a schedule of proofs is, in fact, the work of re-proving
a theorem that has already been proved. This circumstance can be
elaborated as follows: a mathematician undertaking the construction
of a proof of Godel's theorem has himself been a prior witness to the
work of that proof as an achievable organization of proving's practices;
he has renewable access to the structure and detail of that proof by
consulting established texts, and he uses the increasingly exhibited
structure of his proof, as the exhibited structure of a proof of Godel's
theorem, to project the work that he still needs to do and to locate his
immediate work within that developing, yet familiar structure. Thus,
even though the innovative character of his work in bringing to material
exhibition a proof of Godel's theorem brings him into the presence of
his own ability as a prover and, simultaneously, into the presence,
structure, and originality of the proof itself, a prover of Godel's theorem
- in that his work is a re-proving - interprets, trivializes, projects,
reviews and interrogates (as a means of further developing) his work
as the work of the recovery of the familiar, remembered, inessentially
modified, and proper articulation and organization of that proofs
125
A SCHEDULE OF PROOFS

structure and necessary detail.

In the present topic, in contrast to examining the work of proving
Godel's theorem as the work of its re-proving, I want to indicate how
the orderliness of a schedule of proofs for Godel's theorem may arise
as a vital and critical production problem for a contemporary prover
from within the work of proving the theorem. To do this, I begin
by outlining one of the ways in which such a problem can arise and,
at the same time, inseparably, by indicating what the character of
that problem, as practice, is. I then proceed to work out a solution
for the problem and to indicate what is required of such a solution
for it to be one. The solution that is given is not original, and I have
relied throughout my treatment of it on the proof of Godel's theorem
found in Mendelson's book, Introduction to Mathematical Logic.
The point of this topic is not to give an original formulation and
proof of Godel's theorem, but to show, given the existence of already
established proofs of that theorem, how the enunciation and organiza-
tion of a schedule of proofs can still become a matter of critical
attention and interest.
The treatment of this topic will have a technical character; it will
be summarizing, and a close analysis will not be given.
In part, the material that follows has been included in order to
give completeness to the discussion of the work of producing a schedule
of proofs. However, in that this material illustrates how the adequacy
of an established schedule can seriously be called into question from
within the work of that schedule's production, and in that it illustrates
how a prover will further develop the practices of proving Godel's
theorem (and, therein, further articulate the accountable structure of
that proof) so as to organize and extract from those practices (as their
discovered orderliness) an accountable structure of proving adequate
to the solution of the problem that has been posed, the reader may
gain a sense of the radical character of a prover's problem, as a problem
in the local production of social order, of organizing and extracting
an accountable structure of proving from within the lived-work of
proving to which that structure is irremediably tied. The material in
this topic will bring to a close the treatment of the lived-work of
selecting and arranging the propositions of a schedule of proofs.
In order to indicate how such a problem can arise, let us begin by
recalling, once again, the proof of the primitive recursiveness of term(x)
as that proof was given in the discussion of Godel numbering as a
technique of proving:

term(x)-¢> x * 0 and 3y L(x)

EO; i~O pf
{
[(y)L(y) = x] and

Vi EO; L(y) ( [i = 0] or [(y)i = g(O)] or [var((y)J]

126
 A SCHEDULE OF PROOFS

or 3j < i ((y)i = g(S() * (y)j * g( ))) or

3j < i 3k < i ([(y)i = g(+() * (y)j * (y)k * g( ))) or
[(y)i = g(·() * (y)j * (y)k * g( ))))))
The reader will also recall that our development of this formula was
tied to our interrelated construction of what became explicitly formu-
lated as a 'formation sequence of terms' and the discovered compatibility
of the Godel numbers of such formation sequences with the methods
of constructing primitive recursive functions and relations: to say that
an expression of P is a term is to say that there exists a finite sequence
of expressions of P, t 1 , •.. , ti, ... , tm -i.e., a formation sequence of
terms - such that, for each i, i = 1, ... , m, one of the following
conditions holds:
(i) ti is 0
(ii) ti is an individual variable
(iii) ti is S(tj) for some j < i
(iv) ti is +(t/k) for some j < i, k < i
(v) ti is -(t/k) for some j < i, k < i.
Then, to each term t such as
t = +(x 1 S(·(x 5 0))),
we can associate a formation sequencer, as, for example,
T = x 5 , 0, ·(xsO), S(·(xsO)), x 1 , +(x 1 S(·(x 5 0))),
and, in turn, a Godel numbering assigns a Godel number y to such a
formation sequence, as with
y= P¥(x 5 ) . p~(O). p§(·(x 5 0)). p§($(·(x 5 0))). p~(x 1 )
• p&(+(x 1 S( ·(X 5 0)))).

It is the structure of the Godel numbering that is illustrated here that

was used and articulated in the proof of the primitive recursiveness of
term(x) above.
Let us now ·modify the proof of the primitive recursiveness of
term(x) in the following way: rather than including the notion of a
formation sequence of terms within the proof of the primitive recur-
siveness of term(x), we can, as we did in topic (c), first define the
relation formterm(x)- holding if and only if xis the Godel number of
a formation sequence of terms - as a separate relation of the arithme-
tized syntax of P. The proof that term(x) is a primitive recursive
relation then consists of two definitions/proofs:
() formterm(x), holding if and only if x is the Godel number of a
formation sequence of terms, is a primitive recursive relation.

127
A SCHEDULE OF PROOFS

Proof·
formterm(x) ~ x =F 0 and x =F 1 and
Vi.;;;; L(x) [i =F 0] or [(x)i = g(O)] or
[var((x)i)] or
3j < i [(x)i = g(S() * (x)j * g( ))] or
3j < i 3k < i [(x)i = g(+() * (x)j * (x)k * g( ))] or
3j < i 3k < i [(x)i = g(-() * (x)j * (x)k * g( ))] } .
() Let term(x) hold if and only if x is the Godel number of a term of
P. term(x) is a primitive recursive relation.
Proof· L(x)
term(x) ~ 3y.;;;; n pf { [formterm(y)] and
i=O
[(Yk(y) = x]} .
The reader should note that, except for the technically necessary
provision that i =F 0, the conditions following the first bounded quanti-
fier Vi .;;;; L(x) in the definition/proof of fonnterm(x) have been
arranged so as to be in serial correspondence with the five conditions
used to define a formation sequence of terms and that the definition/
proof of term(x) 'says' that there exists a number y such that y is the
Go del number of a formation sequence or terms t 1 , . . . , tm,
y = P¥(c,) • ... • pgrm),
that tm is a term since it has such a formation sequence, and that x is
the Godel number of the last exponent of y, that is, that x is the
Godel number of the term tm.
Although this material restructuring the proof of the primitive
recursiveness of term(x) is not critical to the argument that follows,
the structure of the definition/proof of term(x) that that material
restructuring exhibits is. In order to see this, let us suppose that we
'wished' 115 to generalize 116 the proof of Godel's theorem to 'extensions'
P', P", P"', ... of our original logistic system P. An example of such an
extension would be that of adding the 2-place function symbol E to
the primitive symbols and of adding two nonlogical axioms,
E(x 1 0) = S(O)
and
E(x 1S(x 2)) = {E(x1x2)xt),
to the axioms of P. The intended interpretation of E in N is exponentia-
tion, E(x y) being interpreted as xY. In the section 'A Structure of
Proving,' I will consider the question of whether or not the proof of
128
 A SCHEDULE OF PROOFS

Godel's theorem, as it has been developed in this book, can be modified

so as to provide a proof of Gociel's theorem for such an extension P' of
P - that is, whether or not, given the consistency ( w-consistency) of
P', there exists a constructable sentence J' of P' such that neither J'
(nor ~ J') are deducible in P'. This is the problem of determining;·
given a proof of G6del's theorem for P, the class of formal systems to
which that theorem applies. In the present discussion, however,
I only want to consider whether or not it is possible to modify the
proof of the primitive recursiveness of term(x) so that it still holds
given the addition of new function symbols to P like that of E and,
hence, of new terms such as E(x 1 x 2 ). This is obviously related to
the more general problem.
If we consider just the case of the addition of the 2-place function
symbol E, a solution to our problem is not hard to fmd and can be
reconstructed as the following orderly procedure: first, an 'appro-
priate'117 Godel number needs to be assigned to the symbol E; a new
clause (vi) added to the definition of a formation sequence of terms,
namely
(vi) ti isE(x_rk) for some j < i, k < i;
and, corresponding to this clause, a new condition added to the proof
of the primitive recursiveness of formterm(x), thus,
3j < i 3k < i [(x)i = g(E() * (x)j * (x)k *g( ))].
The proof that term(x) is primitive recursive can then remain exactly
as before:
L(x)
term(x) ~ 3y E;; n pf { [formterm(y)] and [(y)L(y) = x]} .
i=O
Now, what this example makes apparent - and what, in fact, is
made available not by first examining the case of the addition of one
particular function symbol like E, but as the accountable structure of
the reasoning of the definition/proof of term(x) that is illustrated by
the case of the addition of E - is that the same procedure articulated
for the function symbol E can be used for the addition of any finite
number of function symbols rr
to the primitive symbols of P, where
m indicates the number of arguments that the function symbol takes
and r enumerates, and thereby distinguishes between, the m-place
function symbols. 118 For this more general case, appropriate G6del
numbers again need to be assigned to each of the ;trn - assuredly
possible since only a finite number of such assignments are needed -
an additional clause needs to be added to the definition of a formation
sequence of terms for each of these new function symbols, and, corres-
ponding to each such clause, a new condition needs to be added to the
disjunction in the defmition/proof of formterm(x),
129
A SCHEDULE OF PROOFS

3n 1 < i ... ~nm < i [(x)i = g(f~() * (x)n, * ... * (x)nm * g( ))].
The proof of the primitive recursiveness of term(x), again, remains
as before,
L(x)
term(x) ~ 3y ~ IT pf { [formterm(y)] and [(y)L{y) = x] } .
i=O
The availability of this straightforward procedure for generalizing
the proof of the primitive recursiveness of term(x) to the case where a
finite number of new function symbols have been added to P provides a
background for asking whether or not it is possible to prove the primitive
recursiveness of term(x) for extensions of P that contain an infinite,
rather than just a finite, number of function symbols. However, this
question cannot be answered by further elaborating the method that we
used in our previous proof: in the case of an infinite number of func-
tion symbols, our previous definition of a formation sequence of
terms would have to include a new clause for each new symbol, and our
previous definition/proof of formterm(x) would have to include a
corresponding infinity of conditions. There is no provision in our
definition or elaborations of the notion of primitive recursiveness for
such a construction, and, in fact, a major theme in the historical devel-
opment of the notion of primitive recursiveness was that primitive
recursive functions should have finite, effective procedures for calcu-
lating their value from any given argument. Minimally, the proof of the
primitive recursiveness of term(x) for such an extension of P depends
on a prover being able to make the definition/proof either of form-
term(x) or, in a schedule not using formterm(x), of term(x) self-
referential in the sense that, for any given term t, there would be a
primitive recursive procedure for deciding, through the use of the
Godel numbering and in such a way so as not to refer to an infinite
list of possible constructions, whether or not the subsequences of
symbols occurring in t were themselves terms. The problem is not
solved by realizing that for the Godel number of any one particular
term, the number offunction symbols occurring in a formation sequence
for that term is finite: in that the function symbols that would need
to be examined would depend on the particular term that was given,
an infinite list of possible constructions from the function symbols
would still be necessary in order for a procedure developed around
this idea to be applicable to any possible given term.
In the material that follows, a solution to this problem is given. 119
Unlike the discussions of topics (a) and (b) however, I will not try to
exhibit the lived-work of discovering and articulating the-organization-
of-practices/the-materially-specific-schedule-of-proofs that makes up
such a solution, although I will try to present the material in a
manner that reflects the character of that solution as an (endogenous)
130
 A SCHEDULE OF PROOFS

organization of proving's work. The aims of the discussion will be

these: first, the solution makes available, as its accomplishment, what a
prover needs to discover, as an accountable course of reasoning, from
within the lived-work of developing a schedule of proofs; thus, in
comparison to the schedule of proofs outlined earlier, the solution
makes available the features of that previous schedule that need to be
modified or changed; and third, in these two ways, in that the solution
points to the lived-work of producing a practically adequate schedule
of proofs, the solution also indicates how the accountable orderliness
of a schedule of proofs can become a critical production problem for a
contemporary prover in and as the course of proving-again G6del's
theorem.
We wish to show that when an infinite number of new function
symbols have been added to the formal system P, the terms of this new
system - or, more exactly, the Go del numbers of the terms of this new
system - still constitute a primitive recursive relation. Let us begin,
however, by first introducing a device that is known as course-of-
values recursion: 120 As the reader will recall, a numerical function
f (of one variable) is defined by primitive recursion when the value
of that function at 0 is given and when the value of that function at
the successor of each number x, S(x), is defined in terms of the value
of the function at x - that is, if a is a natural number and his a numer-
ical function of two variables, then the function f defined by the
equations
f(O) =a
f(S(x)) = h(x, f(x))
is said to be defined by primitive recursion. In contrast to the definition
of a function by primitive recursion, a function f is said to be defined
by course-of-values recursion when the value of that function at S (x)
depends not only on the value of f at x, but on at least some of the
values f(i), i < x.
Now, let f be a numerical function of one variable and define a
function t:#, called the course-of-values function for f, by the equation
1ifx=O
F(x) = ( x-1 .
n nf(i)
n+t ifx> 0
i=O
The utilityofthisfunction comes from the fact that F(x + 1) 'encodes'
the values of f(i) for 0 ~ i ~ x since
X
r..#(x + 1) = ll pf(i) = pf(O) •
1 ..
• pf(x)
i+l 1 · · · x+l
i=O
and from the fact that these encoded values can then be recovered
131
A SCHEDULE OF PROOFS

from f4t by using the 'decoding' function ( ). , thus

f(i) = (f#(x + l))i+l.
Since the course-of-values function f4t at S(x) encodes the values
of a numerical function f for all i < S(x), this function's usefulness
for explicit definitions by course-of-values recursion is clear. But more
to the point, in order to generalize the definition/proof of term(x)
to our extension of P, we will need to use two theorems concerning
this function and course-of-values recursion:
Theorem 1. lfh(x, z) is a primitive recursive function and f(x) =
h(x, f#(x)), then fis a primitive recursive function.
Proof· By definition,
f#(O) =1
f#(S(x)) = f#(x) • p~~if#(x)).
Thus, the primitive recursiveness off# follows from that of h by
primitive recursion. Since
f(x) = (f#(S(x)))x+l,
where ( ). is the 'decoding' function, f is primitive recursiveness by
substitution.
Theorem 2. If H(x, z) is a primitive recursive relation and a relation
R(x) holds exactly when H(x, KR#(x)) holds, KR being the
characteristic function of R, then R is a primitive recursive relation.
In that it is not at all obvious, as of yet, how a prover might come to
need and articulate the second theorem, nor, for that matter, is it
obvious just what the second theorem actually proposes, I will defer
its proof until later.
Let us return to the problem of proving the primitive recursiveness
of term(x) when the language of P has been supplemented by the
addition of an infinite number of function symbols where m and r rr'
are now any natural numbers greater than 0. For defmiteness, let/~ :=
S, fi :.= +, and /~ := · in our new system, and let us call this new
system p+_ In previously defining/proving-the-primitive-recursiveness-of
term(x) for P, we associated with each term of P a (non-uniquely
determined) formation sequence for it. For example, the term + (·
(x 1 O)x 2 ), by our previous method, could be associated with the
formation sequence
r = x1, 0, ·(x1 0), x2, +(·(x1 0) x2)
having the Godel number
M(x 1 ) • p~(O) • ~(·(x 1 0)) • p{{x 2 ) • p~{+( ·(x 1 O)x 2 )) •

132
 A SCHEDULE OF PROOFS

However, we saw that our method of proving the primitive recursive-

ness of term(x) for P, as that method was tied to this way of associating
formation sequences and Godel numbers with the terms of P, was not
conducive to the definition/proof of term(x) for p+_ In contrast, then,
let us now associate a different type of formation sequence with a
term f~(t 1 ••• tm) of p+, where t 1 , . . . , tm are also terms of p+,
namely the sequence
m m m m m m
fr, fr (, fr (ti> fr (tl t2, .. ·, fr (tl t2 · · · tmJr (tl t2 ... tm).
Then the term +(-(x 10)x 2), or fi(fhx 10)x 2), is now to be associated
with a unique 'formation sequence'
+, +(, +(·(x 10), +(:(Xt 0) x 2 , +(·(xt 0) x2)
which, in turn, has the Godel number
p'l(+) • ~(+() • p~(+(·(x 1 0)) • p~(+(·(x 1 O)x 2 ) • M(+(·(x 1 O)x 2 )).

Next, let us define a relation f-symbol(x) which holds if and only if x

is the Go del number of a function symbol of p+ and a function arg(x),
m ifx = g(f~)
arg(x) ={ ,
0 otherwise
which gives the number m of terms that properly follow a given func-
tion symbol ;:n for the concatenation of that function symbol, a left
parenthesis, m terms, and a right parenthesis, in that order, to itself be
a term of p+_ I will leave for later the problem of arranging the Godel
numbering in such a way so as to permit the demonstration that f-
symbol(x) and arg(x) are primitive recursive. With these definitions in
hand, then, we can articulate the methodic character of our new way
of associating formation sequences, and thereby Godel numbers, to
the terms of p+ as the following, provisional definition/proof of term(x)
as a primitive recursive relation, where the contingencies of that
definition/proof are themselves locally identified and cultivated features
of that formula's construction:
term(x) ~ x = g(O) or var(x) or
3y ~ 0 {[y =f. 0] and [x = (y)L(y)] and [f-symbol((y)t )]
and [L(y) = arg((y)t) + 3] and [(y) 2 = (y) 1 * g(( )] and
Vu ~ L(y) ([(2 < u) and (u ~ arg((y)t) + 2)] imply
3v ~ x (term(v) and (y)u = (y)u-"-1 * v)) and
[(y)L(y) = (y)L(y)..d * g( ))] } ·
L(x)+3
The check that IT pf is an upper bound for y is left to the reader.
i=O
133
A SCHEDULE OF PROOFS

An immediate problem with this definition/proof of term(x) is that

it is 'impredicative' - the presence of term(v) in the defmition of
term(x) introduces a circularity into the definition. But by considering
our new method of associating with a term ti of p+- a unique formation
sequence and, thereby, the unique Godel number n of that formation
sequence, a prover can see/show that if a function symbol rr
is to
appear in ti, then the Godel number of Jfl must be less than n. Thus,
only a finite number of, in principle, determinable function symbols
need to be examined as being used in forming ti. Even more importantly,
however, we can see from the new formation sequence
+, +(, +(·(x 1 0), +(·(x 1 O)xz, +(·(xlO)xz
for +( ·(x 1 O)x 2 ) that each of the successive elements of thE;l sequence
consists of an addition of symbols to the immediately preceding element
of that sequence and, therefore, that the Godel number of each succeed-
ing element of that sequence is greater than that of all those elements
that have preceded it and that the GOdel number of a term - i.e., of
the last element of such a construction sequence - has the largest
Godel number of all the elements in that sequence. Since, then, all the
Godel numbers needed to show that some number x is the GOdel
number of a term are less than x, Ktenn (x), the characteristic function
of the terms of JY, depends only on the values of Ktenn(Y) for y < x,
and the apparent impredicativity of the preceding definition seems to
be avoidable through the use of the course-of-values recursion. More
precisely, we want to show that term(x) is a primitive recursive relation
or, equivalently, that Ktenn is a primitive recursive function or that
{ x I Ktenn(x) = 1 } is a primitive recursive relation. Let us replace
term(v) in our provisional defmition of term(x)
term(x) = x = g(O) or var(x) or
L(x)+3
3y EO; II pf {[y =F 0] and [x = (y)L(y)] and
i=O
[f-symbol((y)l)l and [L(y) = arg((y)I) + 3] and
[(y)2 = (y)l * g(( )] and
Vu EO; L(y) ([(2 < u) and u EO; arg((y)I) + 2)] imply
3v EO; x (term(v) and (y)u = (y)u.!.l * v)) and
[(y)L(y) = (y)L(y).!.l * g( ))]}
with (z)v+t = 1 and defme a new relation H(x, z),
L(x)+3
H(x, z) ~ x = g(O) or var(x) or 3y EO; II pf {(y =F 0] and
i=O
[x = (Yk(y)] and [f-symbol((y) 1 )] and [L(y) = arg((y)I) + 3]
134
 A SCHEDULE OF PROOFS

and Vu ~ L(y) ([(2 < u) and u ~ arg((y)I) + 2)] imply

3v ~ x ((z)v+l = 1 and (Y)u = (y)u.!.l * v))
and [(y)L(y) = (y)L(y).!.l * g( ))] } ·
Since (Kterm # (x))v+t = 1 implies, by the definition of the course-of-
values function, that Kterm(v) = 1, it follows, by defmition of a charac-
teristic function, that v is the Godel number of a term and that x is the
Godel number of a term if and only if H(x, Kterm # (x)). Thus, term(x)
is defined by
term(x) ~ H(x, Kterm# (x)),
and we only need to show that Kterm(x) = KH(x, Kterm# (x)) is a
primitive recursive function. But as the reader will recall, Theorem 1
stated earlier says that if KH(x, z) is a primitive recursive function and
Kterm (x) = KH(x, Kterm~ (x)), then Kterm (x) is a primitive recursive
function as well. Since, subject to the demonstration of the primitive
recursiveness of f-symbol(x) and arg(x), H(x, z) is a primitive recursive
relation, KH is a primitive recursive function, and we are done. More-
over, if we rewrite this last argument, we see that that argument is a
proof of a particular case of our second theorem concerning course-of-
values recursive (with R(x) = Kterm(x)):
Theorem 1. Ifh(x, z) is a primitive recursive function and f(x) =
h(x, f#(x)), then f is a primitive recursive function.
Theorem 2. If H(x, z) is a primitive recursive relation and a relation
R(x) holds exactly when H(x, KR# (x)) holds, KR being the
characteristic function of R, then Ris a primitive recursive relation.
Proof" SinceR(x) holds if and only if H(x, KR# (x)),
KR (x) = KH(x, KR #(x)),
and it follows from Theorem 1 that KR is a primitive recursive
function, and, hence, that R is a primitive recursive relation.
To finally complete the proof that term(x) is a primitive recursive
relation for JY, we need to show that f-symbol(x) and arg(x) are,
respectively, a primitive recursive relation and ·a primitive recursive
function. I will not belabor the technical detajls involved in working
out these proofs. 121 'First,' the Godel numbering must be re-specified
so as to accommodate the infinite number of function symbols. This
may be done by defining
g(f~) = 9 + 8 (2m·3r)
and arranging the other Godel numbers so that this is possible. The
'appropriateness' of this numbering can 'then' be seen to consist of the
following: first, the introduction of the primitive recursive function
135
A SCHEDULE OF PROOFS

qt{x, y), the quotient of y divided by x, allows arg(x) to be defined as

the primitive recursive function
arg(x) = {qt{8, x ...!..9)) 1 ;
second, the numbering g(f~) = 9 + 8 {2m•3r) affords an easy proof
that f-symbol{x) is a primitive recursive relation, namely
f-symbol(x) ~ 3m .;;;; x 3r.;;;; x (x = 9 + 8 {2m· 3r));
and, third, this numbering provides a method for generalizing the
proof of Godel's theorem to logistic systems that extend P through the
introduction of an infinite number of constant symbols ar (witha 1 :=
0) and an infinite number of relation symbols AT
(withAi :=)as well
as an infinite number of function symbols: for example, we can further
modify our Godel numbering so as to permit the definitions
g(ar) = 7 + 8r 1 .;;;; r
g(f~) = 9 + 8 (2m·3r) 1 .;;;;m, 1.;;;; r
g(A:U) = 11 + 8 (2m ·3r) 1 .;;;; m, 1 .;;;; r
and by doing so, thereby insure that the Godel numbers of the primitive
symbols of the extended system are all odd and distinct, that the
relations constant(x), f-symbol(x), and R-symbol{x) of the arithmetized
syntax of that system are demonstrably primitive recursive, and that
g(f~) and g(A~) can be 'decoded' so as to determine, in a primitive
recursive manner, the number of arguments that IF and Arm take,
and all these as relevant features of the Godel numbering that are
discovered and developed over the course of, and in and as, the
materially-specific work of producing a schedule of proofs for a proof
of Godel's theorem.
With these last details concerning the assignment of Godel numbers,
we have, in a sense, completed the definition/proof of term(x) for an
extension P'" of the system P. In a finished schedule of proofs, however,
the reasoning indicated in the discussion above would be arranged,
not in the fashion of the preceding discussion, but as an orderly course
of proving's practices providing, in and as the accountably-exhibited
work of proving-again Godel's theorem, a schedule of proofs for that
proof. Moreover, that ordering of the schedule would not simply be
retrospectively imposed on the lived-work of the discovery and articu-
lation of the definition/proof of term(x), but, instead, that arrangement
of practices would itself arise from within, and, in doing so, would
already be tied to, the components of that definition/proof as the
components of a temporally discovered, temporally developing, and
temporally articulated gestalt of practices needed for, and accountably
making up, the proof of the primitive recursiveness of term{x): thus,
the proposition that qt(x, y) is a primitive recursive function and the
136
 A SCHEDULE OF PROOFS

propositions needed to give such a proof would be placed in the earlier

part of a schedule of proofs; the introduction of the techniques of
course-of-values recursion would probably follow the t._echniques for
working with primitive recursive functions and relations; the Godel
numbering might be specified just prior to the propositions concerning
the apparatus for working with that numbering, and the positioning
of the propositions concerning arg(x), f-symbol(x), term(x) and the
deletion or modification of formterm(x) would all be placed in
accountably appropriate and practically efficacious places in the latter
part of the schedule. And through this arrangement as well, as the
reader will see in the next chapter 'A Structure of Proving,' the account-
able structure of the schedule of proofs as a schedule of proofs for a
proof ofGodel's theorem is thereby both, and at the same time, further
articulated and maintained throughout the lived-work of producing
a proof of the primitive recursiveness of term(x) for JY.
Let me bring this material to a conclusion.
In this topic, I have tried to indicate the reworking of our original
schedule of proofs that is necessary to prove the primitive recursive-
ness of term(x) for an extension JY of our original language P. As the
reader will recall, the discussion of topic (d) began with the observation
that the work of proving Godel's theorem now has, for a contemporary
prover, the character of being the re-proving of an established theorem.
What the review of the modification of the defmition/proof of term(x)
for JY has shown is that the accountable orderliness of a schedule of
proofs for a proof of Godel's theorem can still arise, for a prover
proving-again Godel's theorem, as an open problem whose solution
is not guaranteed in the sense that it is not available as an extractable
course of materially-exhibited reasoning from a previously established
schedule of proofs and that, therefore, the identifying orderliness of a
schedule of proofs can still arise for a prover as a critical and vital
production problem. Thus, the preceding discussion has opened up
another problem: in searching for the extension of the definition/
proof of term(x), in the way in which the accountable orderliness
of our schedule of proofs was thereby re-opened as a production
problem, the problem of descriptively analyzing what constitutes
the identifying orderliness of the lived-work of producing a schedule
of proofs for a proof of Godel's theorem - as that identifying orderli-
ness is made available from within that work itself- becomes available
as a radical problem in the study of the production of social order.
I will return to this matter in the next chapter 'A Structure of Proving.'

4 The work of providing a consistent notation for a schedule of proofs

articulates that schedule as one coherent object122
For brevity, I will refer in this section to the work of a prover in
137
A SCHEDULE OF PROOFS

developing and using a schedule of proofs' always proof-specific notation

as 'notation's work.' A review of the material in the preceding section
shows, in addition to that work's omnipresence and diversity, its
occasioned and practical character, its unremittingly circumstantial
specificity, and, through the exhibition of some of its inspectable
detail, that work's utter familiarity to and unremarkableness for the
experienced prover. In this section I bring to a close the discussion of
the work of producing a schedule of proofs by attempting to fmd,
in notation's work, a distinctive order-productive phenomenon. The
section begins by recalling the definition of the function Sub(x, t, a).
Although the naturally accountable achievement of that definition is
the provision for a function, designated as Sub(x, t, a), whose properties
are independent of its particular notational representation, I argue
that not only is such an achievement made available through a prover's
locally developing notational apparatus, but it is the embeddedness
of the objects of a schedule of proofs in notation's work that provides
both for the analyzability of those objects and for their utility for
developing that schedule. As a first step in my argument, I describe,
in a naturalistic manner, some of the accountable detail ofthe notational
designation Sub(x, t, a). This material provides (what will later become
available as) a background of notational practices which, in turn, will
enable us to examine a particular instance of notation's work. A pro-
visional analysis of that work will then bring the discussion of this
section to a close.
To begin, then, let us recall the definition of Sub(x, t, a):
g(s; A I) if x = g(x), t.= g(t), a= g(A),
x an individual variable, t a
{
Sub(x, t, a)= term, and A a wff

a otherwise
where Sf AI denotes the wff that results when a term tis substituted
for the free occurrences of an individual variable x in a wff A.
Now, although the symbols x, t and a make this definition nota-
tionally specific, the definition of Sub(x, t, a) is constructed (and
necessarily so for it to be a naturally accountable mathematical
defmition) so as to provide for x, t, and a only as a scheme of reference
- no properties either of the symbols x, t, or a or of any 'meanings'
associated with them are exhibited as necessary features of that defini-
tion. In this way, the definition of the numerical function denoted as
Sub(x, t, a) provides for that function as an 'ideal' (or 'real' or 'trans-
cendentalized') object disengaged from the local work circumstances
of its development and use. If, however, we modify our definition of
Sub(x, t, a) by replacing x, t and a with a, y, and x, respectively,
we obtain
138
 A SCHEDULE OF PROOFS

g(s; XI) if a= g(a), y = g(y), x = g(X),

a an individual variable, y a
(
Sub(a, y, x) = x term, and X a wff

otherwise
which, though not 'wrong,' begins to indicate an (as-of-yet-unidentified-
as-practice) confusion. We can, of course, go even further: by removing
the alphabetic coordination between the numerical and syntactic
variables, we can write the definition of Sub(w, y, x) as
. g(S~ WI) ifw = g(a), y = g(x), x = g(W),
a an individual variable, x a
[
Sub(w,y,x)= term, and Wawff

x otherwise
thereby obtaining a definition that seriously obscures the object that
that definition offers for consideration.
In the presence of such possibilities, a mathematician might well
argue that while a prover does, in fact, develop and use particular
notational devices in constructing such a definition, he does so as a
means (though, perhaps, a necessary one) for exhibiting the indepen-
dence of that definition from what that definition makes available,
as its achievement, as simply a convenient choice of notational desig-
nations for it. One of the aims of this section is to begin to find what
the lived-work of such an achievement actually consists of.
Before attempting this, and as a means of providing background
material for that attempt, I want to first review some of the account-
able details of the notation Sub(x, t, a) as those details provide for
the practical accountability and efficacious use of Sub(x, t, a) as part
of a schedule of proofs. One such detail, for example, is tied to the
fact that the definition of the syntactic operation of substitution -
defined as the. operation of replacing the free occurrences of an
individual variable x in a wff A by a term t and denoted as s; A I -
although it is itself defined in such a way that its variables serve as a
scheme of replacement, uses its syntactic variables x, t, and A in a
manner that is coordinated throughout the development and discussion
of the syntax of P - namely, that xis understood as representing an
arbitrary individual variable, that t represents an arbitrary term, and
that A represents an arbitrary wff. The point, as it regards Sub(x, t, a),
is this: the designation of the arithmetized counterpart of the substi-
tution operation as Sub(x, t, a) used the (never made explicit) alpha-
betic pairing of the variable x with the Godel number of an individual
variable x, the variable t with the Godel number of a term t, and the
variable a with the Godel number of a wff A
139
A SCHEDULE OF PROOFS

x~x

t~t

a~A

as a local 123 device for recalling the 'meaning' or 'interpretation' of

Sub(x, t, a) in terms of Sf A I .
A second accountable feature of the notational designation Sub(x, t,
a) that I wish to call to the reader's attention is that the arguments
of the function so designated are arranged in a particular order - that
x is the first variable, that t is the second, and that a is the third - and,
thereby, as a developing aspect of the reliance of a prover on the
association of x with x, t with t, and a with A; that if the numerical
value m 1 of the first variable is the Godel number of an individual
variable, if the numerical value m2 of the second variable is the Godel
number of a term, and if the numerical value m 3 of the third variable
is the Godel number of a wff, then the value of Sub(m 1 , m 2 , m 3 ) 124
is the Godel number of the wff that results when the term with Godel
number m 2 is substituted for the free occurrences of the variable with
Godel number m 1 in the formula with Godel number m 3 . Although
the ordering of these variables is theoretically arbitrary- for example,
the variables could be arranged as in Sub(a, t, x) but still correspond
to S? A I - that ordering is tied to a way of remembering the association
of the particular variables in Sub(x, t, a) (independently of, in addition
to, and in the absence of, the mnemonic use of the alphabet) with the
particular variables in Sf A I, thus,

~
Sub(x,t~AI

Specifically, this arrangement of variables is tied both to the temporal

organization of the writing of Sf A I as
S ~ sx ~ Sf ~Sf A ~ Sf A I
and, thereby, to a way of seeing the symbols S? AI as a temporally
organized configuration/:.: 5 and, as well, to the facility of constructing
the function
IP{u) = sub(g(x 2 ), u, u) = g(S~(u) Ul)
140
 A SCHEDULE OF PROOFS

in the diagonalization/'proof by fixing the first variable in sub(x, n, a)

as opposed to fixing the second variable in a function sub(n, x, a)
and writing
I,O(u) = sub(u, g(x 2 ), u),
a formula that would result from writing Sub(t,x, a), instead of Sub(x,
t, a), as the arithemetized counterpart to S~ A I, instead of SJ A I,
where, reversing the placement of t and X, sr
A I now stands for the
operation of substituting a term t for the free occurrences of x in A. 126
Finally, as a means of pointing to a third accountable detail of
'Sub(x, t, a)', I want to contrast the notation for Sub(x, t, a) with that
for the function sub(x, n, a) defined by the equation
g(S~(n) A I) if x = g(x ), a = g(A ),
xis an individual variable,
(
sub(x, n, a)= and A is a wff

a otherwise.
These equations define sub(x, n, a) as Sub(x, g(k(n)), a) where k(n) =

--· n
S( ... S(O) ... ), the numeral representing the number n in P. As with
all the components of the arithmetized syntax, Sub(x, t, a) and sub(x,
n, a) use an abbreviation of the syntactic operations or relations that
correspond to them as a mnemonic device for associating the arithme-
tized syntax with the original syntax of the formal language. In the
case of Sub(x, t, a) and sub(x, n, a), however, both of these functions
correspond to the operation of substitution, and since they cannot be
accountably distinguished through the alphabetic designation of their
variables (both functions simply being defined on N3 ), then, again as a
local and locally transparent device, the two functions are distinguished
by capitalizing the first letter of one of them.
Later in this section, I will return to, and reconsider, what I have
spoken of as the accountable 'details' of a notational designation - in
this case, of Sub(x, t, a) -as a residue of a prover's notational practice.
In fact, the immediate utility of the preceding material is that it provides
a background of such practices against which we can examine a prover's
work in developing a schedule of proofs' notation.
In outlining a schedule of proofs at the opening of this chapter, and
throughout this book as a whole, I have been following closely the
proof of Godel's theorem that is found in Robbin's Mathematical
Logic: A First Course. 127 In the schedule of proofs given earlier,
however, instead of adopting Robbin's notation for the relations
occur(a 1 , x, a2 , a), bound(a 1 , x, a2 , a), and free(a 1 , x, a 2 , a), I used
w, x, y, and z as the variables for these relations, writing them as
occur{w, x, y, z), bound{w, x, y, a), and free(w, x, y, a) in Propositions
141
A SCHEDULE OF PROOFS

23, 24 and 25, respectively. The point of my doing so at that time, as a

revision of Robbin's schedule considered in detachment from the work
of its proving, was to emphasize that those relations were defined on
N4 , specifically disengaging their definitions from the (informal,
unexplicated) alphabetic correspondence that was being used in
Robbin's proofs. 128 Now, there is nothing critically wrong with the
notation that I adopted; in fact, the difference between Robbin's and
my notation is accountably inessential to the definition/proofs of those
relations. Nevertheless, from within the lived-work of developing or
reviewing a schedule of proofs, Robbin's notation exhibits its preferred
character. We will begin to have descriptive access to notation's work
by addressing the problem of determining the local motives - in and
as the work of producing/recovering the schedule of proofs as a naturally
accountable course of proving - for so preferring Robbin's notation.
First, consider the definition/proof of the relation occur using my
notation:
occur(w, x, y, z) <:> wff(z) and var(x) and z = w * x * y.
What this definition 'says' is that occur(w, x, y, a) holds if and only if z
corresponds, under the Godel numbering, to a wff Z, w corresponds
to an initial string W of primitive symbols of Z, x corresponds to an
individual variable x, y corresponds to a terminal string Y of primitive
symbols of Z, and Z can be written as the concatenation of the symbols
of W, x~ andY in that order, Z = WxY.
At first, the defmition/proof of occur(al> x, a2, a) in Robbin's
notation may seem only a little different:
occur(a 1 , x, a2 , a)<:> wff(a) and var(x) and a= a 1 * x * a2 .
In contrast to the earlier one, this last definition maintains the use of a
as the Godel number of a wff to be designated as A and maintains, 129
as well, the use of x as the Godel number assigned to an individual
variable to be designated as x. However, the more significant difference
is that, rather than using an arrangedly arbitrary assignment of the
letters w and y to represent the GOdel numbers of initial and terminal
sequences W and Y of primitive symbols in the wff Z corresponding
to the number z, Z = WxY, the definition/proof of occur(a 1 , x, a2 , a)
uses the designations a 1 and a2 as a means of providing a further
correspondence between A and its decomposition as A = A 1 xA 2 •
This association of a1 , x, a2 and a with A =A 1xA 2 is the achievement
of the-proofs-notation/that-notation's-associated-work and is nowhere,
except in and as that work, made explicit.
The differences between these two definitions/proofs becomes
clearer in the subsequent definition/proof of bound(a 1 , x, a2 , a).
Roughly, bound(a 1 , x, a2 , a) asserts that the occurrence of x in A
indicated as A 1xA 2 is a bound occurrence of that variable - that is,
142
 A SCHEDULE OF PROOFS

that that x occurs in a well-formed part 'tlxC where the exhibited

occurrence of x in 'tlxC may or may not be the bound occurrence
of x indicated in bound(a 1 , x, a2, a). Robbin's definition/proof of
this relation is
bound{a 1 , x, a2 , a)~ occur(a 1 , x, a 2 , a) and
3c 1 ~a 3c2 ~a 3c ~a (wff(c) and L{c!) ~I..( a!)
and L(c2 ) ~ L(a2) and a= c 1 * g(V') * x * c * c 2 ).
If we compare this definition/proof to one using the notation bound(w,
x, y, z),
bound{w, x,y, z) ~ occur(w, x, y, a) and
3u ~ z 3v ~ z 3p ~ z (wff(p) and L{u) ~[(ad
and L(v) ~ L(y) and z = u * g('t/) * x * p * v),
we see, first, that in Robbin's proof the numerals 1 and 2 are used to
develop and further articulate the decomposition of a formula A
associated with the number a into distinguished sequences of symbols
- Ct indicating an initial sequence cl of A' c2 a terminal sequence
C2 of A, and c, with no subscript, a wff C 'between' C1 and C2 -
with the formula A being decomposed into possible sequences of
primitive symbols A = C1V'xCC2 • Although the relationship between
c 1 , c2, and c is slightly different from that between a 1 , a 2 , and a, the
similarity between the use of these letters and subscripts is recognized
by a prover in and as its practical consequentiality for finding and
exhibiting the naturally accountable definition/proof of bound(a 1 , x,
a2 , a); those similarities, as well as the difference in their use, are
witnessed in and as the work that that notation 'performs' in the
proof. But while a prover depends on such notational practices, those
practices need not be specified nor is it immediate, in each particular
case, how that use would be specified.
In contrast to Robbin's definition/proof of bound(a 1 , x, a2 , a),
the adequacy of the second proof is not as immediate as Robbin's;
a prover is required to do more work to find it. Once the achievement
of Robbin's definition/proof is available to a prover, however, the
prover is aided in finding the adequacy of the definition/proof of
occur(w, x, y, z) by the analogy between that definition/proof and the
earlier one. In both cases, as a prover's achievement, the work that the
notation of those proofs does in making the adequacy of those proofs
available is hidden in and as the natural accountability of the proofs
themselves.
Given this circumstance, an additional, more complex example
may help illustrate the work that Robbin's notation performs in finding
and exhibiting the natural accountable definitions/proofs of his schedule.
143
A SCHEDULE OF PROOFS

To this end, consider the problem of proving the primitive recursiveness

of the relation freefor(t, x, a) (Proposition 31) which is defined as
holding if and only if there are no free occurrences of an individual
variable x (corresponding, under the Godel numbering, to the number
x) in a well-formed part Vyc of the wff A (corresponding to the
number a) where y is a variable occurring in the term t (corresponding
to the number t). The idea of this definition is that if freefor(t, x, a)
holds, then, in substituting t for x in A, a variable y in the term twill
not occur within the scope of a universal quantifier in y which would
make that occurrence of y a bound occurrence.
For the immediate purposes of proving the primitive recursiveness
of freefor(t, x, a) - specifically, by providing the necessary level of
detail to provide for that proofs associated work - let us first intro-
duce the relation free(a 1, x, a 2 , a):
free(a 1, x, a2 , a)~ occur(a 1, x, a 2 , a) and not-bound (a 1, x, a 2 , a).
free(a 1, x, a2 , a) holds if and only if the occurrence of x inA indicated
by A 1xA 2 is not a bound occurrence and, as a consequence, that a
term t can be substituted for x at that place.
With this background, Robbin then gives the now somewhat trans-
parent definition/proof of freefor(t, x, a), further exhibiting and
developing his notational practices in doing so: 130
freefor(t, x, a)~ term(t) and var(x) and wff(a) and
3a1 ~a 3a2 ~a {free(a 1, x, a2 , a) implies
not-3c 1~ a 3c2 ~ a 3c ~ a 3y ~ a 3t 1 ~ t
3t 2 ~ t [wff(c) and var(y) and L(ct) ~ L(at)
and L(c 2 ) ~ L(a 2 ) and t = t 1 * y * t 2 and
a=c1 *-g(V)*y*c*c2]}.
With such an organization of notational practices, the associated work
of the material display through which the natural accountability of
that proof is achieved is readily made available and exhibited; without
such systematic use of notation, the proof of the primitive recursiveness
of freefor(t, x, a) becomes extremely complex, as the following
definition/proof of it, using a 'more arbitrary' 131 notation, illustrates:
freefor(x, y, z) ~ term(x) and var(y) and wff(z) and

3u ~ z 3y ~ z {free(u, y, v, z) implies
not-3r ~ z 3s ~ z 3p ~ z 3q ~ z 3b ~ t 3c ~ t
[wff(p) and var(q) and L(r) E;;i L(u) and L(s) ~ L(v) and x =
b • q * c and z = r * g(V) * q * p * s] } .

144
 A SCHEDULE OF PROOFS

Given the need for such a number of numerical variables in this proof,
along with the obscurity of the work required to define/prove freefor(x,
y, z), a prover would find the need to begin to systematize his notation
and to introduce that orderliness into his schedule of proofs.
With this material in hand, it is now possible to begin to find in a
schedule of proofs' notation, as that schedule's notation is tied to its
lived-work, a distinctive order-productive phenomenon in the study
of mathematicians' work.
Already in this chapter we have seen that a schedule of proofs is a
curiously paired. object the-material-schedule/the-practices-of-proving-
to-which-that-schedule-is-irremediably-tied. To speak of the naturally
accountable schedule of proofs is to speak of that paired object. And,
similarly, to speak of the naturally accountable definition/proof of
the relation freefor(t, x, a) is to speak of the intrinsically tied pair
the -material- proof /the-practices-of-proving- to-which-that-proof-is-
irremediably-tied. In this way then, the 'material proof
freefor(t, x, a)~ term(t) and var(x) and wff(a) and
3a 1 ~a 3a2 ~ a {free(a 1, x, a2 , a) implies
not-3c 1 ~ a 3c 2 ~ a3c ~a 3y ~a 3t 1 ~ t
3t2 ~ t [wff(c) and var(y) and L(c 1) ~ L(a!)
and L(c2 ) ~ L(a2 ) and t = t 1 * y * t2
and a = c 1 * g('V) * y * c * c2]} ,
in coming to be, as the achievement of a prover's lived-work, the
naturally accountable account of the work of its proving, is, as that
achieved pairing of material proof and practice, the naturally account-
able definition/proof of freefor(t, x, a).
Consider the following drawing taken from a book on Gestalt
psychology:

From Principles of Gestalt Psychology by K. Koffk:a, copyright© 1935

by Harcourt Brace Jovanovich, Inc., renewed 1963 by Elizabeth Koffka.
Reproduced by permission of the publisher.

The point of this drawing is that although, in some sense, it is a 'chaotic

jumble of lines,' 132 the lines of that drawing provide enough detail so
145
A SCHEDULE OF PROOFS

that they are seen as composing, and are organized as, the face of
a plumpish, spectacled man. The analogy to notation's work is this:
in the definition/proof of freefor(t, x, a), the notation that is. used
serves to articulate the work to which that definition/proof, as the
naturally accountable definition/proof of freefor(t, x, a), is irremed-
iably tied. The point is that, from within the lived-work of proving,
the achievement of notation's work is not that the notation provides
a correspondence between symbol and referent nor, for that matter,
that it provides a correspondence between symbol and practice, but
that - in the way in which the orderliness of the material detail of a
proof and that orderliness's associated work are mutually developed
and intrinsically complementary, and in the way in which that comple-
mentarity is itself a developing and produced feature of the work of
proving - a proofs notation comes to articulate just those practices
to which the naturally accountable proof is itself tied.
The material in this section can now be drawn together. First, if
we understand by a schedule of proofs' notation the accountable order-
liness of that proofs material detail, then that orderliness -as a prover's
achievement in constructing or working through a schedule of proofs -
articulates (in the sense of an intrinsic complementation) the work of
producing the naturally accountable schedule of proofs. This being
the case, the fact that the designation Sub(x, t, a) offers itself (again,
as its achievement) as providing a correspondence between its 'account-
able detail' and that detail's intended referent is itself a disengaged
residue of notation's work - that work only being available at the
mathematical work-site and hidden even there in and as that work's
accountable accomplishment. And, last, if we now understand that a
reference to the consistency of a schedule of proofs' notation is, in
fact, a reference to that notation's exhibited and accountable orderli-
ness, then a prover's work in providing such a consistent notation is the
work of locally articulating just those practices which, paired with the
schedule's material detail, make up the coherence of that schedule as
a naturally accountable course of provings' work.
Before leaving the discussion of a schedule of proofs' notation, it is
now possible to address one of the notational details of the schedule
of proofs (as it is presented in this book and in Robbin's Mathematical
Logic as well) that may have troubled the reader. As the reader will
recall, the proof of the primitive recursiveness of the function n ~ g(xn)
enumerating the Godel numbers of the individual variables of P was
previously given as
n
g(xn) = g((x) * ll Pf * g( )),
i=O ~
where xn is a11 abbreviation for (x" ... ') and 7 is the Godel number
that has been assigned to the primitive symbol '. The troublesome
146
 A SCHEDULE OF PROOFS

aspect of this proof is only alluded to by pointing out that this proof
is the only place in our schedule where the actual Godel number of
a symbol appears; throughout the rest of the entire schedule the intro-
duction. of the particular numerical assignments of the proof-specific
Godel numbering has been avoided through the use of the function,
designated by g, that maps the expressions (but not the primitive
symbols or sequences of expressions) of our formal system P onto
their Godel numbers. What is troublesome about the use of the number
7 is this: in the way in which the rest of our schedule of proofs does
not explicitly use the actual assignment of specific Godel numbers,
that schedule works out and exhibits the relevant strncture of that
numbering for a proof of Godel's theorem. 133 Against this background,
the use of the specific Godel number of ' in the demonstration of the
primitive recursiveness of the function n f.+ g(xn) offers the possibility
that a proof of Godel's theorem may be dependent on just such a
particular assignment of GOdel numbers.
Now, given that the presence of the 7 in the displayed proof, and
given as well that it is 'apparent' 134 to a prover that the schedule of
proofs is not so dependent on this particular assignment of Godel
numbers, the question is thereby raised as to why the 7 still appears
in the proof.
The answer to this question is found through an examination of the
notational work that would be involved, and the examination of the
notational apparatus that would have to be built, so that that 7 could
be removed.
As the reader will recall, the symbol g was defined as, and has
been used throughout our work in developing the schedule of proofs
to denote, the mapping of the expressions of Ponto their Godel num-
bers. However, in the proof of the primitive recursiveness of n ~ g(xn)
n
g(xn) = g((x) * n p{ * g( )),
i=O
the Godel number of the primitive symbol 1 (i.e., 7) and not the Godel
number of the expression consisting of 1 alone (i.e., 2 7 ) is needed. Now,
the reason that the function g cannot be used to give this number - the
reason that g cannot be used to denote the Godel numbering as it is
defined on the. entire collection of primitive symbols, expressions and
sequences of expressions of P - is that a Godel numbering must assign
different numbers to 1 when 1 is considered as a primitive symbol, as
an expression, and as a sequence of expressions of P. The detail of the
symbolism g(') fails to distinguish between these different ways of
interpreting1
•

Given this circumstance, there seem to be two ways to avoid the

presence of 7 in the proof of nf.+ g(xn): One way would be to write
the exponent of Pi as (g('))i since g(') = 2 7 (the Godel number of the
147
A SCHEDULE OF PROOFS

expression ') and, then, (g(')) 1 evaluates g(') at the exponent of p 1 in

g('), (g(')) 1 = (2 7 ) 1 = 7. A second method would be to use some
notation, like 'corners' 135 I I, to indicate the Godel numbers of the
primitive symbols of P. However, in that the proper introduction of
that notation would have to occur before its immediate use in the
schedule 136 - thereby providing for the independence of the distinc-
tion between I I and g from the locally contingent details of the work
of proving Godel's theorem - the introduction of a special designation
for the Godel numbers of primitive symbols in addition to such a
designation for the Go del numbers of expressions would necessitate,
as well, an additional notational designation for the Godel numbers
of sequences of expressions. A prover might, for example, 137 use g, g+,
and g++, respectively, for these three functions. ·
These two devices suffer the following difficulties: In the first
case, the decipherment of (g(')) 1 exhibits the artificiality of first
raising the Godel number of the symbol ' to an exponent and then
determining what that exponent is. The use of (g(')) 1 not only obfus-
cates the structure of the proof by making a prover decipher it, but
it then exhibits its artificiality in that self-same decipherment. In that a
specific Godel number has already been introduced in defining that
numbering, it is easier and less contrived to simply use that assigned
number. In the second case, if the reader is offered the distinction
between g, g+, and g++ -and, thereby, the relevance of that distinction
to the work of proving Godel's theorem - that notation then requires
a prover to do more work on reading and working through the schedule
of proofs. Every occurrence of g in the present schedule would have to
be changed to g+, yet the only material motivation for introducing that
distinction is the single occurrence of the 7 in the proof of the primitive
recursiveness of nf.+g(x 0 ). 138 In this way, the distinction between g,
g+, and g++ would only distract from the clarity of the material presenta-
tion in providing for more work than is required to find the material
presentation of the schedule as a naturally accountable schedule of
proofs.
In summary, then, we have found not only the material motives for
using the 7 in the proof of the primitive recursiveness of n ~ g(x 0 ),
but, at the same time, the advantage, in that it reduces the burden of
a schedule's notation, of simply defining one function g defined exclu-
sively on the expressions of P.

148
8 A Structure of Proving

A
The Characterization Problem: The Problem of Specifying What
Identifies a Proof of Godel 's Theorem as a Naturally Accountable Proof
of Just That Theorem; The Texture of the Characterization Problem
and the Constraints on Its Solution; The Characterization Problem as
the Foundational Problem

Over the course of the preceding chapters we have come to rediscover,

in increasing, technical detail, the natural accountability of a proof of
Godel's theorem as being irremediably tied to the local, lived-work of
that proofs production. In doing so, we have therein come to redis-
cover a proof of Godel's theorem as a naturally accountable proof.
One major problem remains, however, for the analysis of a proof of
Godel's theorem as lived-work: what is it that identifies a proof of
Godel's theorem, over the course of its production, from within the
developing course of its own lived-work, as a naturally accountable
proof of just that theorem? What is it that is identifying of the local
work of its proof, not as the work of proving, but as the work of
proving Godel's theorem?
In this chapter I will refer to the problem of specifying what identi-
fies a particular proof of Godel's theorem as a naturally accountable
proof of just that theorem as the 'characterization problem.' By a
'solution' to that problem, I will refer here to a descriptive analysis of
the work of producing a proof ofGodel's theorem that provides, in and
as the inspectable details of the lived-work of that theorem's proof,
technical access to that proofs characterization problem as a problem,
for provers at the mathematical work-site, in the local production of
social order. The aim of this chapter is to begin to develop such a
solution, and, as a first step in that development - as well as a means
of iridicating the texture and richness of the characterization problem
149
A STRUCTURE OF PROVING

itself - the present section indicates some of the immediate constraints

that can be placed on what a solution to the characterization problem
could be. The analyses of the preceding chapters allow this to be done
in a brief and summarizing fashion.
To begin, then, a first constraint on a solution to the characterization
problem is this: the characterization problem cannot be solved by
construing the natural accountability of a proof of Godel's theorem
as a property inherited from, or as a reflection ofthe apodictic character
of, an ideal or idealized proof of Godel's theorem. On one hand, there
is nowhere available such ·an ideal, transcendental, self-evidential and
self-evidentially complete proof to consult; on the other, it is the
materially present, practically objective, naturally accountable proof
of Godel's theorem that provides the grounds for speaking of a proof of
Godel's theorem as being disengaged from the work of its production
and review in the first place. Similarly, the characterization problem
cannot be solved by positing the transcendental existence of the mathe-
matical objects about which that proof is concerned and then using that
claimed existence to provide for the self-evidential character of those
object's properties. In that it is the naturally accountable proof that
makes available and exhibits both itself and those mathematical objects
as 'real,' propertied objects, a solution to the characterization problem
begins not by taking that natural accountability for granted, but by
examining the work that makes up that natural accountability as an
achievement.
A second and closely related constraint to be placed on a solution
to the characterization problem is that such a solution cannot make
use of an idealized version of mathematical reasoning to, in that way,
insure the transcendental properties of a naturally accountable proof of
Godel's theorem. The reduction of the reasoning of such a proof to,
for example, instances of the use of modus ponens 1 not only trivializes
the reasoning of the proof, but irremediably hides the distinctiveness of
the discovered structure of reasoning that that proof makes available.
The rendering of a naturally accountable proof as a derivation within a
formal system - and, therein, the purported demonstration of the dis-
engaged character of that proofs reasoning from its lived-work- cannot
fmd the reasoning of that proof particularly; in all but the most elemen-
tary cases, the work of fmding such a rendering is absurdly complicated;2
and in the case of a proof of Godel's theorem, when an (abbreviated)
rendering is even attempted, that rendering is produced, and becomes
intelligible, only by using a naturally accountable proof of Godel's
theorem as a guide to that rendering's interpretation, that rendering
itself being tied to the local work practices of its own production to
insure its accountability as a rendering of the proof of Go del's theorem. 3
By disallowing the characterization problem to be solved by refer-
ence to a disengaged version either of a proof of Godel's theorem, of

150
 A STRUCTURE OF PROVING

such a proofs reasoning, or of the objects that are considered in such a

proof, the first two constraints serve to relativize the characterization
problem as a problem arising from within, and whose solution is main-
tained through, a prover's local organization of a materially-specific
course of work as this-particular proof of Godel's theorem. In this way,
then, the first two constraints point to a third one as well: that the
practical objectivity and practical accountability of a proof of Godel's
theorem, as the proof of just that theorem, are temporally-developing
features of the work of producing and reviewing such a proof.
One of the ways in which a solution to the characterization problem
can be provided for is by using a comparison between different proofs
of Godel's theorem to argue that the variations between those proofs
are inessential features ofthem and, therefore, that a particular proof
of Godel's theorem is, in fact, the material-specific realization of an
objective, work-transcendent proof that has been fitted to the circum-
stantial particulars of its immediate production. The constraints that
have already been introduced indicate the irrelevance of (the posited
existence of) such a transcendental proof to the solution of the charac-
terization problem - a solution must provide for the recognizability
and identifiability of a proof of Godel's theorem in and as the local,
temporally-situated, temporally-developing, endogenously organized
work of a material-specific proofs production and review. Nevertheless,
the existence of variations between proofs of Go del's theorem point to
a phenomenon internal to the production of a particular proof, namely
that provers will recognize the features of such a proof, over the course
of that proofs development, as being peculiar to that particular proof;
provers will see a developing proof as a particular way of proving
Godel's theorem. A mathematician watching a student work through a
proof of the theorem will, at times, recognize the student's work as
being faulted, plodding or disorganized, or will recognize that the
material detail of the student's work is unmotivated by the circum-
stances of the proof given so far, but, at the same time, he will see how
the student's work can be reorganized so as to provide an adequate
proof of Godel's theorem. That a solution to the characterization
problem provides for the transcendental character of the work of
proving Godel's theorem from within the local work of that proofs
production is a fourth constraint on such a solution.
A fifth constraint is that a solution must provide for the intelligi-
bility of three somewhat mysterious features of the work of producing
such a proof: the first is that certain details of a proof, later available
to a reviewer as being obviously necessary for that proof, will be
curiously missed by the prover when he was constructing the proof;
second, and conversely, from within the work of producing such a
proof a prover will find it necessary to introduce certain details into
that proof that, in a later review, he will see as unnecessary and even
151
A STRUCTURE OF PROVING

unmotivated by the encompassing proof; and third, the reasoning of

part of the argument of the proof - while apparent from within the
work of its development - may, on later review, seem unsuggestive
and even unintelligible, and the prover will have to rework the reason-
ing of his argument to find the cogency, or the faulted character, of
the argument that he had previously given. In summary then, this fifth
constraint, in requiring that a solution to the characterization problem
provide for the intelligibility of these three features of mathematical
work, can be seen to raise the question of what constitutes the relevant
and identifying detail of a proof of G6del's theorem from within the
work of proving, or the work of reviewing, that proof.
A sixth constraint is that the natural accountability and coherence
of a course of proving as, identifiably, a proving of Godel's theorem
cannot be made to depend on a prover's ability to articulate the work
of that proof in any other way than the way in which that work is done
as recognizably adequate mathematical practice. The analyzability of a
proof of Godel's theorem arises from within and is essentially tied to
professional mathematical praxis; it is not tied to the reflective avail-
ability of that practice in terms other than those provided by the
practice itself. Despite the fact that, on reflection, a given proof never
explicates every feature of its reasoning, despite the fact that such
reflection offers the possibility of an endless regress of further explica-
tions, nevertheless, a given proof, as its achievement, will be recognizably
the practically adequate, practically objective, practically complete
proof of Godel's theorem; in a proofs achievement as an exhibited,
naturally accountable organization of mathematical practice, that
proof will therein, simultaneously, exhibit the pointlessness of disput-
ing its adequacy as such a proof. 4
Finally, a seventh constraint to be placed on a solution to the
characterization problem is that, despite - or against the background -
of the other constraints of this list, a given proof of Godel's theorem is,
nevertheless, an 'organizational' or 'social object';5 it is recognizably a
practically objective, practically adequate proof of that theorem; it is
an analyzable object with demonstrable properties; it has a transcen-
dental presence as such a proof of Godel's theorem; while apparently
dependent on the material details of its presentation, it exhibits, as
its achievement, its independence from the particular details of that
presentation; at a given stage in the construction of a proof, the work
that has been done in proving the theorem can be contrasted with, and
constrains, the work that still remains to be done; a proof of Godel's
theorem is seen to have an atemporal character over the course of its
temporal development and construction; for a mathematician proving
Godel's theorem, and for mathematicians party to such a proving, the
work of proving that theorem is the witnessible and recognizably
real thing that is being done in and as the material details of that work;
152
 A STRUCTURE OF PROVING

that this is so - that Godel's theorem is the real thing that is being
proved - is not available to a mathematician as an idiosyncratic per-
ception of his own or another prover's work, but as something that is
objectively so about that work, as something that anyone 6 can see;
the work of proving Godel's theorem is available to the mathematician,
from within the course of that work, as a matter of practical assess-
ment, comment and review; and, finally, the way in which the proving
is done is, in fact, a moral way of proving, it is recognizably a proper
way to prove, it is recognizably a proper way to prove Godel's theorem,
and the proving is done in such a way so as to be, recognizably, just
such a proper way of proving.
This, then, is a list of constraints that can initially be placed on a
solution to the characterization problem. That they are actual con-
straints is not a matter of principle or of proper argumentation; they
are constraints in the way in which they serve to summarize some
of the features of a-proof-of-Godel's-theorem/the-lived-work-of-its-
production that a proof of Godel's theorem, over the course of this
book, has been disclosed to be. It is this object to which the term
'a proof of Godel's theorem' refers, and it is the problem of specifying
what identifies such a proof, to its local production cohort, from
within and as the lived-work of its production, as a naturally account-
able proof of Godel's theorem, that makes up the characterization
problem for a proof of Godel's theorem. In this way, then, the charac-
terization problem is, in fact, a radical problem in the production of
social order.
In the next section I will begin to work toward a 'solution' to the
characterization problem. It is clear, however, that while I have used
the term 'characterization problem' to refer exclusively to the charac-
terization problem for a proof ofGodel's theorem, the characterization
problem and the constraints placed on its solution have a much greater
relevance to the study of mathematicians' work. That this is so can be
gained, in the first instance, from some of the problems involved in
teaching beginning mathematics students 'what a proof consists of'
and, therein, 'how to prove': each mathematical problem, no matter
how similar to others, is nevertheless a distinctive problem, and a
student must learn, in and as the fmding of a given problem's relevant
detail, how to form up a course of work that stands as a proof of that
problem's solution; for each problem, it is impossible to specify before-
hand what would make up a naturally accountable proof of it; even
when the similarity between the proofs of two propositions is almost
identical, this does not insure that a student will fmd what, as practice,
that similarity consists of for proving the second proposition after
having seen a proof of the first one;7 proofs are not and (generally)
cannot be found by consulting specifications of the rules of proper
logical inference to fmd an appropriate way of proceeding; and while
153
A STRUCTURE OF PROVING

the 'vernacular' use of rules of logic are relevant to the adequacy of

a proofs reasoning, the insistence on characterizing a proof as a scheme
of deductive inference is a pedagogic device for teaching what, in the
end, is known and available to a prover as mathematical praxis. If we
add to these brief observations just two more - that a discovered
proof is a discovered naturally accountable proof and that professional
mathematicians, in proving theorems, are engaged in teaching colleagues
the-particular-way -of- proving/ the- particular- way -of-reasoning that
makes up a proof of the theorem in question - then we come to see
that the characterization problem - as the problem of specifying what
it is about a particular proof that identifies that proof as the naturally
accountable proof of just the theorem that that proof proves - can,
in fact, now be understood, not as a foundational problem, but as the
problem of the foundations of mathematics.

B
Generalizing the Proof of Gooel's Theorem (As a Means of Gaining
Technical Access to the Characterization Problem)

In the remainder of this chapter I want to develop the notion that

for provers a-proof-of-Godel's-theorem/the-lived-work-of-its-proving is,
and is recognizable and identifiable as, an organization of proving's
practices. I will speak of this organization of practices as the 'structure
of proving'· of a proof of Godel's theorem, and later in this chapter I
am going to propose that this 'structure of proving,' available only in
and as the lived-work of that proofs production, provides a provisional8
solution to the characterization problem. Before doing so, this section
will first address a technical problem in the proof of Godel's theorem
as that proof has been presented here - namely, the fact that the
adequacy of that proof (for proving the thing that it accountably
proves) depends on the witnessible generalizability of that proofs
reasoning. By elaborating on this problem and its solution, this section
provides background material for the chapter's larger argument.
Let us begin, then, by considering the work that remains for our
proof of Godel's theorem to be a practically adequate proof of that
theorem. As it was stated earlier, Godel's first incompleteness theorem
says (roughly) 9 that if the formal theory designated by Pis consistent,
then neither the particular, seeably/showably exhibitable sentence J
nor that sentence's negation "' J are deducible in P. Since a formal
theory is defined to be complete if and only if, for every sentence 10
S of that theory, either S or"' Sis deducible in it, then, by definition,
the proof that neither J nor "' J is deducible in P -or, in other words,
that J is a 'formally undecidable' sentence of P- shows that Pis not a
complete theory. 11
154
 A STRUCTURE OF PROVING

By putting together the various parts of the proof ofGodel's theorem

as they have been developed to this point, we have thus far shown that
P is not a complete theory. But this statement of Godel's theorem is
certainly problematic: as the reader will recall, P was originally con-
structed as a 'model' of the theory of the natural numbers (that is,
as a model of elementary arithmetic), and 'Godel's theorem,' as it has
just been stated, does not assert that there exists an undecidable propo-
sition of number theory, but only that there exists an undecidable
proposition of P; the possibility remains that only P, and not number
theory, is incomplete or, stated in a fashion more compatible with the
Hilbertian program for foundational studies, that there exist different
formalizations of number theory that are both consistent and complete.
In fact; an immediate candidate for such a formal theory is simply
our original theory P with the undecidable sentence J now added to P
as an additional axiom. 12 •13 By construction, J would be deducible in
this new theory, and the problem would thereby arise of showing
either that there exists or that there does not exist an undecidable
proposition of it. Thus, our proof of Godel's theorem for P 'could be
seen' 14 to leave open the question of the applicability of that theorem
to other logistic systems that contain 'models' of number theory;
thus; to leave open the consequentiality of that theorem for the pro-
gram of the formal investigation of number theory; thus, to leave open
the consequentiality of it for the (classical) investigation of the foun-
dations of mathematics in general.
This, of course, is not the case: the proof ofGodel's theorem demon-
strates what is. sometimes referred to as the essential incompleteness
of P -namely, that the incompleteness of P cannot be remedied by
adding to P, for example, any class of axioms the Godel numbers of
which make up a primitive recursive set - and, in fact, that proof
makes available the essential incompleteness of a much larger class
of formal systems modelling elementary arithmetic than those built
from the particular system P.
Our problem, then, in descriptively analyzing the natural account-
ability of a proof of Godel's theorem, is that of finding how the
generaJizability of Godel's theorem is tied to the lived-work of that
theorem's proof. As a means of doing this, I am going to review the
entire proof of Godel's theorem: axioms for the system P will be
introduced, ·a Godel numbering will be specified, a schedule of proofs
will be given, and the diagonalization/'proor will be presented once
again. The aim of so forming up the various 'parts' of a proof of Godel's
theorem into a coherent proof is to exhibit the essential connection
between the generalizability of that proof and the availability of that
proof, to a prover, as a 'structure of proving.'
To begin then, let me introduce, although in an abbreviated manner, 15
a specific set of axioms and rules of inference for the formal system
155
A STRUCTURE OF PROVING

now to be denoted as the system P. For the primitive symbols ofP let
us take (, ), ', x, ~. -, V, =, 0, S, f, and·, and let us adopt the follow-
ing as axioms and axiom schemata where A,B, and Care wffs ofP: 16
1 A:::>(B~A)
2 ((A ~ (B ~ C))~ ((A ~B) ~ (A ;>C))
3 (-(-A))~A
4 VxA ~ S:!j A I, where x is an individual variable and t is a term which
is free for x in A
5 Vx(A ~B) ~ (A ~ VxB), where x is an individual variable
having no free occurrences in A
1? ' ' '
6 (x1 = x2) ~ (x1 = X3 ~ x2 = X3)
7 (x 1 = x 2) ~ (S(xt) = S(x 2))
8 -(S(xt) = 0)
9 (S(xt)=S(x2))~(x 1 =x2)
10 x 1 +O=x 1 =
11 x1 +S(x2) =S(x1 +x2)
12 x 1 ·0=0
13 X1 · S(x2) = (x1 · x2) + x1
14. S~ AI~ (Vx(A ~ S~exlAI) ~ VxA).
The rules of inference of P will be modus ponens
fromA andA ~B. to infer B
and generalization,
if x is an individual variable, from A to infer VxA.
Now, let us introduce a specific Godel numbering for the language of
P. The Godel numbers of the primitive symbols of P are defined as
follows:
( ... 3
) ... 5
, ... 7
X ... 9
~ ... 11
13
v 15
= 17
0 19
s ... 21
+ ... 23
',' ... 25
If ~ = s 1 ••• Sn is a sequence of primitive symbols of P (i.e., ~ is an
'expression' of P) and Si is the Godel number assigned to the symbol
si, then the GOdel number assigned to ~ is g(~) = p~l • ... • p:P,

156
 A STRUCTURE OF PROVING

where Pi is the i-th prime number, p 0 := 1, and the symbol g is used

exclusively for the Godel numbers of the expressions of P. If A = 0:1>
... , O:n is a sequence of expressions of P, then the Godel number
assigned to A is Pi(a,) • ... • p~(an).
Before giving a 'schedule of proofs' for P, let us recall the definition
of the class of primitive recursive functions and that of the class of
primitive recursive relations. First, if his ann-place numerical function
and g1, ... , gn are m-place numerical functions, then the m-place
function f defined by
f(x 1, ... , xm) = h(g 1{x 1, ... , xm), ... , gn(x1, ... , Xm))
is said to be obtained from g 1, ... , gn and h by substitution. If g is
an m-place numerical function and h is an (m + 2)-numerical function,
then the (m + I)-place function defined by
f(x1; ... , Xm, 0) = g(x1, ... , Xm)
f(x 1 , •.. , Xm, S(y)) = h(x 1, ... , Xm, y, f(x1, ... , Xm, y))
is said to be obtained from g and h by primitive recursion. And if a
numerical function is either the zero function (denoted by) Z, the
successor functionS, or one of the projection fuctionsiF, 0 <i, O<m,
=0
Z(x)
S(x) = x + 1

that function is said to be an initial function. Then the primitive recur·

sive functions are defined inductively as only those functions obtained
from the initial functions by a fmite number of substitutions or primi-
tive recursions; less precisely, they are the functions generated from the
initial functions by the 'operations' of substitution and primitive
recursion.
By definition, an m-place numerical relation R is a primitive recursive
relation if and only if its characteristic function K R,
1 if(x 1, ... ,xm)ER
KR(x1, ... ,xm)= { '
0 if(x 1, ... , Xm)ENm- R
is a primitive recursive function.
A number of functions and relations will now be shown to be
primitive recursive, among them functions and relations corresponding
under the Godel numbering to the syntax of P. The list begins with
some propositions indicating some elementary primitive recursive
functions and relations and some procedures for constructing primitive
recursive functions and relations (Propositions 1-10); the proofs of
157
A STRUCTURE OF PROVING

these propositions will not be given. 18

I The constant functions Zn(x) = n, n = 0; I, 2, ... are primitive
recursive.
2 The functions obtained by permuting variables, identifying
variables, adding dummy variables, and substituting constants for
variables in primitive recursive functions are primitive recursive.
3 Addition, multiplication, exponentiation, and 'limited subtrac-
tion,' .£.., are primitive recursive functions. y
Let f be an (m + I)-place numerical function and let l; ·f(x 1 , •.• ,
y k=O
xm, k) and II f(x 1 , ••• , Xm, k) denote the functions
k=O
(x 1 , ... , Xm, y) ~f(x1, ... , Xm, 0) + ... + f(x1, ... Xm• y)
or
(x 1, ... , Xm, y) ~f(xl> ... , Xm,O) • ... • f(x 1 , ... , Xm, y),
respectively. y
4 Iff is a primitive recursive function, then so are l; f(x 1 , ... ,
y k=O
Xm, k) and fl f{Xt. ... Xm, k).
k=O
5 =, =/=, <, and .so; are primitive recursive relations.
6 The logical operations of 'not,' 'and,' 'or,' 'implies,' and 'if and
only if,' applied to primitive recursive relations, produce primitive
recursive relations.
7 The relation obtained by substituting a primitive recursive function
for a variable in a primitive recursive relation is primitive recursive.
Let R be ail (m + I)-place numerical relation and let Vz .so; y R(x 1,
... , Xm• z) and 3z .so; y R(x1, ... , Xm• z) denote the (m + I)-place
relations that hold for (x 1 , ... , Xm, y) if and only if R(x 1 , ... , Xm•
z) holds for all z .so; y or for some z .so; y, respectively.
8 If R is a primitive recursive relation, then so are Vz .so; y R(x 1,
... , xm, z) and 3z .so; y R(x 1, ... , Xm, z).
9 If T 1, ... , Tv are pairwise disjoint m-place primitive recursive
relations, and if g1, ... , gv and h are m-place primitive recursive
functions, then the function defined by the equation
gt(Xt, ... , Xm) ifT t(Xt, ... , Xm)
g2(x1, ... , xm) if T2(x1, ... , XnJ)

gv(xl> ... , xm) ifTy(xl> ... , Xm)

h(x1, ... , Xm)if(xt, ... , Xm)EtTi foralli, I.s;;;i.s;;;v
I 58
 A STRUCTURE OF PROVING

is primitive recursive.
If R is an (m + I)-place numerical relation, let p.z..;;;; y R(x 1 , •.• ,
Xm, z) denote the function of (x 1 , •.. , Xm, y) defined by the equation
the least z ..;;;; y such that
JJ.Z..;;;; y R(x 1 , •.• , xm, z) = { R(x 1 , •.. , Xm, z) if there is
such a z
0 otherwise.
IO If R is an (m + I)-place primitive recursive relation, then p.z..;;;; y
R(x 1 , •.. , Xm, z) is a primitive recursive function.
II The 2-place relation xly is primitive recursive.
Proof· xly ~ 3n..;;;; y (y = n • x).
I2 Let prime(x) hold if and only ifx is a prime number. Thenprime(x)
is a primitive recursive relation.
Proof" prime(x) ~ x > I and 'fly..;;;; y (y!x implies (y = I or y = x)).
I3 The function Pn giving, for each n, the n-th prime number is
primitive recursive. (p 0 := 1.)
Proof· Po= I
Pn+ 1 = JJ.X..;;;; (pn)n + I {prime(x) and Pn < x } .
To see that (pn)n + 1 is an upper bound on x, it is enough to note that
(p 1 • ••• • Pn) + 1 either is a prime number or is divisible by some
prime greater than Pn, for it then follows that Pn+ 1 ..;;;; (p 1 • ••. • Pn)
+ 1 < (Pn)n + 1.
14 Define (x)n as the exponent of Pn in the prime factorization
of x if x > 1 and n > 0 and as 0 otherwise. Then (x)n is a primitive
recursive function of x and n.

Proof" (x)n = { 0 if x = 0 or i = 0
p.k..;;;; x (p~lx and p~+ 1 lx) otherwise.
15 Let the function L(x) give the number n of the largest prime Pn
in the prime factorization of x or give 0 if x is 0 or 1, i.e., L(x) defines
the 'length' of x. L(x) is a primitive recursive function.
Proof· L(x) = p.n..;;;; x (pnlx and V'k'..;;;; x (n < k implies Pklx)).
For every natural number y, y = 0 or y = p{r)o • p?')1 •
p(y)(L()Y). Define x * y as the function mapping (x, y) to the value
L y L(y)
x * y= x• n p~fx}+l.
i=O
16 x * y is a primitive recursive function of x and y.
159
A STRUCTURE OF PROVING

It follows from the definition that x * (y * z) = (x * y) * z for all

x, y, and z greater than 0. Thus, the fmite 'product' of natural numbers
ai > 0, i = 1, ... , n, can be written unambiguously as a1 * ... *an.
17 The function nj-+ g(xn) giving, for each n, the GOdel number
of the n-th individual variable, is primitive recursive.
n
Proof: g(xn) = g((x) * n p{ * g( )).
i=O
18 Let var(x) be the relation that holds if and only ifx is the Godel
number of a variable. var(x) is a primitive recursive relation.
Proof· var(x) ~ 3n.;;;;; x (x = g(xn)).
Define a formation sequence of terms as a finite sequence of terms
'tm such that, for each i, i = 1, ... 'm, one of the
tl, ... 'tj, ...
following conditions holds:
(i) ti is 0
(ii) ti is an individual variable
(iii) ti is S(tj) for some j < i
(iv) ti is +(tlk) for some j < i, k < i
(v) ti is ·(tjtk) for some j < i, k < i.
19 formterm(r), holding if and only if Tis the Godel number of a
formation sequence of terms, is a primitive recursive relation.
Proof"
formterm(r) ~ T =I= 0 and T =I= 1 and
'Vi.;;;;; L(r) {[i = 0] or [(r)i = g(O)] or
[var((r)i)] or 3j < i [(r)i = g(S() * (r)j * g())]
or 3j < i 3k < i [(r)i = g(+() * (r)j * (r>Ic * g( ))]
or 3j < i 3k < i [(r)i = g(-() * (r)j * (r)k * g( ))]} .
20 Let term(t) hold if and only if tis the G6del number of a term
of P. term(t) is a primitive recursive relation.
L(t)
Proof·term(t)~ 3T< fl pf {formterm(T)and[(Tk(T)=t]}
i=O
To see that the bound on Tis adequate, let t be the G6del number of
a term t = s 1 ••• Si ... sm. where the Si are primitive symbols of P.
By reviewing the definition of a formation sequence of terms, the
reader will see that there exists at least one formation sequence for t
of length less than or equal to m. If T is the Godel number of such a
L(t)
formation sequence, then T.;;;;; p~ • ... • p:n = fl pf.
i=O

160
 A STRUCTURE OF PROVING

In analogy with the definition of a formation sequence of terms,

let us call a finite sequence A 1 , •.• , Aj, ... , Am of wffs a formation
sequence of wffs if and only if, for each A j, i = 1, ... , m, one of the
following conditions holds:
(i) Ai is of the form (r = s) where rands are terms
(ii) Ai is(~ Aj) for some j < i
(iii) Ai is (Aj :::> Ak) for some j < i, k < i
(iv) Ai is (VxAj) for some individual variable x and for some j < i.
21 formwff(a), which holds exactly when a is the Godel number of
a formation sequence ofwffs, is a primitive recursive relation.
hoof'
formwff( a)~ a =I= 0 and a =I= 1 and Vi,;;;; I.{a) {[i = 0] or
3t <a 3s <a [term(t) and term(s) and
(a)i = g(() * t * g(=) * s * g())] or
3j < i [(a)i = g((~) * (a)j * g( ))] or
3j < i 3k < i [(a)i = g(() * (a)j * gp) * (a)k * g( ))] or
3j_ < i 3x <a [var(x) and (a)i = g((V) * x * (a)j * g( ))]
22 Let wff(a) hold if and only if a is the Godel number of a wff
of P. wff(a) is a primitive recursive relation.
L(a)
Proof' wff(a) ~ 3a,;;;; II Pi [formterm(a) and (ak(a) =a].
i=O
Define the relation occur(a 1 , x, a2 , a) by the condition occur(a 1 ,
x, a2 , a)~ wff(a) and var(x) and a = a 1 * x * a 2 . occur(a 1 , x, a2 , a)
holds if and only if a is the Godel number of a wff A, x is the Godel
number of a variable x that occurs in A, a 1 is the Godel number of a
part of the formula A to the left of an occurrence of x, and a 2 is the
Go del number of the part of the formula A to the right of that occur-
rence ofx. Then
23 occur(a 1 , x, a2 , a) is a primitive recursive relation.
24 Let bound(a 1 , x, a2 , a) hold if and only if a is the Godel number
of a wff A, x is the Godel number of a variable x which occurs in A,
and the occurrence of the variable x in A indicated by occur(a 1 , x, a 2 ,
a) is a bound occurrence of that variable. Then bound(a 1 , x, a2 , a)
is a primitive recursive relation.
Proof·
bound(a 1 , x, a2 , a)~ occur(a 1 , x, a2 , a) and
3c 1 ,;;;; a 3c 2 ,;;;; a 3c,;;;; a {wff(c) and
161
A STRUCTURE OF PROVING

L(ct) ~ L(a!) and L(c2 ) ~ L(a 2 ) and

*
a= C 1 g('o') * X * C * c2 } •

Let the relation free(a 1, x, a2 , a) hold if and only if occur(a1, x,

a2 , a) and not-bound(a 1, x, a2 , a), i.e.
free(a 1, x, a2 , a)~ occur(a 1, x, a2 , a) and not-bound(a 1, x, az, a).
free(a~o x, a2 , a) indicates a free occurrence of a variable x ( corres-
ponding to the number x) in a wff A (corresponding to the number a).
25 free(a 1, x, a2 , a) is a primitive recursive relation.
Define a functionS 1(x, t, a) as follows:
1-1b ~ t *a* t { 3a1 ~a 3a2 ~a [free(a1, x, az, a) and
b = a1 * t * a2 ]}
if 3a 1 ~a 3a2 ~a [free(a 1, x, a2 , a)]
a otherwise.
Then if x is the Godel number of a variable x, t is the Godel number of
a term t, and a is the Godel number of a wff A, S1(x, t, a) gives the
G6del number of the wff resulting from the replacement of one free
occurrence of x in A by t.
26 S1(x, t, a) is a primitive recursive function.
27 sn(x, t, a), iterating the operation of S 1(x, t, a) n times, is a
primitive recursive function of n, x, t, and a.
Proof: S0 (x, t, a)= a
sn+l(x, t, a)= S1 (x, t, sn(x, t, a)).
Defme M(x, t, a) = SL(a) (x, g(xL(a)+L(t)), a). M(x, t, a) replaces all
the free occurrences of x in A by a variable xL(a)+L(t) that does not
occur either in t or in A.
28 M(x, t, a) is a primitive recursive function.
Defme the function Sub(x, t, a) by the equation

f g(S7 Ai) ifx = g(x), t = g(t), a= g(A),

x an individual variable, t a
Sub(x, t, a) = \ term, and A a wff.

a otherwise.
29 Sub(x, t, a) is a primitive recursive function.
Proof: Sub(x, t, a)= sL(a) (g(xL(x)+L(a)), t, M(x, t, a)).
30 Define a function sub(x, n, a) by the equation
162
 A STRUCTURE OF PROVING

g(Sf(n) A I) if x = g(x), a= g(A), x an

individual variable and A a wff,
sub(x, n, a)= n
..---..
where k(n) = S(... S(O) ...)

a otherwise
sub(x, n, a) is a primitive recursive function.
Proof: First, g(k(n)) is a primitive recursive function of n by
primitive recursion, since
g(k(O)) = g(O)
g(k(S(n))) = g(S() * g(k(n)) * g( )).
Then, sub(x, n, a)= Sub(x, g(k(n)), a).
31 freefor(t, x, a) is a primitive recursive relation where freefor(t, x,
a) holds if and only if t, x, and a are the G6del numbers of a term t,
a variable x, and a wff A, respectively, and if t is free for x in A.
Proof"
freefor (t, x, a)~ term(t) and var(x) and wff(a) and
3a1 ,.;,:; a 3a2 ,.;,:; a {free(a 1 , x, a2 , a) implies
not-3c 1 ,.;,:; a 3c2 ,.;,:; a 3c,.;,:; a 3y,.;,:; a 3t 1 .;;;;; t
3t2 ,.;,:; t [wff(c) and var(y) and L(c 1),.;,:; L(a 1)
and L(c 2 ),.;,:; L(a 2 ) and t = t 1 * y * t 2 and
a1 = c1 * g(V') * Y * c * Cz]} .
32 notfree(x, a), which holds if and only if xis the G6del number of
a variable x, a is the G6del number of a wff A, and x has no free occur-
rence in A, is a primitive relation.
Proof· notfree(x, a)~ not-3a 1,.;,:; a 3a2 ,.;;; a [free(a 1 , x, a2 , a)].
33.1 axiom 1(w), which holds if and only ifw is the Godel number of
an instance of the axiom scheme (A :) (B:) A)), is a primitive recursive
relation.
Proof" axiom 1(w) ~ 3a,.;,:; w 3b,.;,:; w {wff(a) and wff(b)
and w = g(() *a* g(J() * b * gp) *a* g()))} .

33.4 axiom 4 (w), which holds if and only if w is the G6del number of
163
A STRUCTURE OF PROVING

an instance of the axiom scheme (('V'xA) :::> S1 AI), is a primitive recur-

sive relation.
Proof: axiom 4 (w) ~ 3a ~ w 3x ~ w 3t ~ w {wff{a) and
var(x) and term(t) and freefor(t, x, a) and
w = g(((V) * x * a * g( P) * Sub(x, i, a) * g( ))} .

33.6 axiom 6 (w), which holds if and only if w is the Godel number
of axiom 6, is a primitive recursive relation.
Proof· axiom 6 (w) ~ w = g(((x 1 = x 2 ) :::> ((x 1 = x3) :::> (x2 = x3)))).
33.7 axiom 7 (w), which holds if and only if w is the Godel number
of axiom 7, is a primitive recursive relation.
Proof· axiom 7 (w) ~ w = g(((x 1 = x 2 ) :::> (S(xd = S(x 2 )))).

33.14 axiom 14 (w), which holds if and only ifw is the GOdel number
of an instance of the axiom scheme (S~ AI :::> (('V'x(A :::> S~(x) AI)):::>
(VxA))), is a primitive recursive relation.
Proof: axiom 14 (w) ~
3a ~ w 3b ~ w 3c ~ w 3x ~ w {wff{a) and var(x) and
b = Sub(x, g(O), a) and c = Sub(x, g(S(x)), a) and
w =g(()* b * g((('V') * x * g(()* a* gp) * c* g()):::>('V') *
x * a * g( ))))} .
33 axiom(w), which holds if and only ifw is the Godel number of an
axiom or of an instance of an axiom scheme, is a primitive recursive
relation.
Proof· axiom(w) ~ axiom 1 (w) or . .. or axiom 14 (w).
34 Define deduct(A) as holding if and only if A is the Godel number
of a deduction. ded).lct(A) is a primitive recursive relation.
Proof· deduct(A) ~ A=I= 0 and A=I= 1 and
Vi~ L(A) {i = 0 or axiom(A)i) or
3j ~ i 3k ~ i [(A)k = g(() * (A)j * gp) * (A)i * g( ))]
or 3j < i 3x <A [var(x) and (A)i = g(('V') * x * (A)j * g( ))] } .
164
 A STRUCTURE OF PROVING

The last two conditions correspond to the use of modus ponens and
generalization.
3519 The relation ded(A, a), which holds if and only if A is the Godel
number of a deduction of the wff with Godel number a, is a primitive
recursive relation.
Proof" ded(A, a) ~ deduct(A) and (A)L(A.) = a.
The reader will recall that an m-place numerical relation W is
numeralwise expressible in P if and only if there exists a wff W(x 1 ,
... , x m) of P with m free variables such that
(i) if (a 1 , •.. , ~) E W, then l-p W(k(at), ... , k(am))
(ii) if (a 1 , . . . , ~) e W, then l-p ~ W(k(at), ... , k(am))
a
----...
where k(a) = S( ... S(O) . .. ) is the numeral of P corresponding to
the number a. We then have the following
Theorem. 20 Every primitive recursive relation is numeralwise
expressible in P.
Finally, let us defme P to be w-consistent if there is no wff F(x)
of P such that both l-p 3xF(x) and l-p ~F(k(O)), ... , l-p ~F(k(n)),
... for all n.
The proof of Godel's first incompleteness theorem for P is brought
to a close with the following construction of the sentence J and the
proof that that sentence is formally undecidable in P.
First, let I,O(u) = sub(g(x 2 ), u, u). I,O(u) is the primitive recursive
function such that, if u is the Godel number of a wff U, then I,C(u)
gives the Godel number of the wff that results when the numeral k(u)
representing u in Pis substituted for x 2 in U. Next, define the primitive
recursive relation G(A, u) by the condition that
G(A, u) ~ ded(A, ip(u))
~ ded(A, sub(g(x 2 ), u, u)).
Roughly, G(A, u) holds if A is the Go del number of a deduction of the
wff that results when k(u) is substituted for x 2 in the wff with Godel
number u. Since G(A, u) is primitive recursive, there is a wff that
numeralwise expresses Gin P. Let G(x 1 , x 2 ) be such a wff.
Defme I as 3x 1 G(x 1 , x 2 ) and let g(/) = i.
Define] as ~3x 1 G(x 1 , k(i)) and let g(J) = j.
Now, observe that j = I,C(i), that is, tht j is the Godel number of the
wff that results when k(i) is substituted for x 2 in the wff with Godel
number i.
J can be interpreted as 'saying' that there does not exist a number n

165
A STRUCTURE OF PROVING

such that n is the Godel number of a deduction of the formula that

results when the numeral k{l) is substituted for x 2 in the wff with
Godel number i, or, in other words, J 'says' that J is not deducible in,
and hence is not a theorem of, P.
The 'proor of Godel's theorem can now be given.
Godel's First Incompleteness Theorem (for P). (1) If Pis consistent, J
is not deducible in P. (2} If Pis w-consistent,- J is not deducible in P.
Proof· (1) If J is deducible in P- that is, if
(*) l-p- 3x 1G(xi, k(i)),
-then, for some r, ded(r,j}. But since
ded(r, j} ~ ded(r, IP(i)) ~ G(r, i),
the numeralwise expressibility of G gives
l-p G(k(r), k(i}},
from which it then follows that
(**)
(*) and (**) cannot both hold if Pis consistent. Thus, if P is consistent,
J is riot deducible in P.
(2) Suppose that - J is deducible in P, that is
(***) l-p 3x 1G(x1, k(i)}.
Since P is assumed to be w-consistent, it follows that Pis consistent as
well; hence J cannot be a theorem ofP. But if J is not a theorem, then,
by definition of ded(r, j},
G(n_, i) ~ ded(n, IP(i)) ~ ded(n, j}
cannot hold for any n EN. Using the numeralwise expressibility ofG,
one obtains
l-p- G(k(l}, k(i))

(****) 1-p "'G(k(n), k(i))

for all n E N. Together,(***) and(****) contradict the assumption of

w-consistency. Thus, ifP is w-consistent,- J is not deducible in P.
This, then, brings to a close the proof of Godel's theorem for the
system P. Once again, it says, roughly, that if Pis a consistent system,
then there exists a sentence J of P such that J is formally undecidable
166
 A STRUCTURE OF PROVING

in P - i.e., that there exists a sentence J of P such that neither J nor

-J are deducible in P. In other words, it says that if P is consistent,
then it is incomplete. Moreover, in that the proof of the numeralwise
expressibility of primitive recursive relations in P gives an inductive
procedure for constructing formulas that numeralwise express those
relations,21 it is possible, in principle, to construct and exhibit the
actual sentence J, although in practice such a construction is infeasible
if not impossible.
Now, let us return to our original question concerning the generality
of this proof.
As I reminded the reader earlier, P was initially constructed as a
formal 'model'22 of number theory. The central problem that Godel's
theorem is intended to address is not whether or not P is complete,
but whether or not ordinary number theory itself is complete, or, more
precisely, whether or not there is any formal 'model' 'of number
theory such that that 'model' can be shown to be complete.
A first step in addressing this problem is taken by realizing that,
since the axioms and rules of inference ofP 'represent' true propositions
of number theory and general, or epitomizing characterizations of,
methods of ordinary mathematical inference, then a first-order formal
system constructed as a 'model' of number theory would certainly
seem to have to include, among its theorems, all the theorems of P.
By this reason~g. then, a complete 'model' of number theory would
have to contain - or, in more conventional terminology, would have
. to be an 'extension' of- our original system P. 23
Let us consider one such extension of P.
First, note that since the sentence J is not deducible in P, and since
J can be intuitively interpreted as 'saying' about itself that it itself
is not deducible in P, J is, intuitively, a true sentence of number theory.
Let us add J as a new axiom (axiom 15) to the axioms of P there by
obtaining an extension P' ofP. In that! is an axiom of this new system,
J is certainly deducible in it, and, thus, if P is consistent, P' will properly
contain the theorems of P. In other words, we have 'enlarged' our
'model' of number theory, and the ~uestion that now confronts us is
whether or not this enlarged system P is a complete one.
In order to show that P' like P, is an incomplete system, we can
reason as follows: In that J is a constructable formula of P, we can,
in principle, calculate the GOdel number. g(J) of J and insert the
proposition
33.15 axiom 15 (w), which holds if and only ifw is the GOdel
number of J, is a primitive recursive relation.
into our already constructed schedule of proofs. Modifying Proposition
33 accordingly by adding a new condition to the definition of axiom(w)
corresponding to axiom 15 (w), we next obtain, in a new Proposition 34,
167
A STRUCTURE OF PROVING

the primitive recursive relation ded'(A., a) which holds exactly when A

is the Godel number of a deduction of the wff of P' with GOdel number
a. Following the procedure in the proof of Godel's theorem for P, we
can then construct a primitive recursive relation G'(A, a) and use the
numeralwise expressibility of primitive recursive relations to obtain
x
a new relation G'(x 1 , 2 ), corresponding to the previously constructed
G(x 1 , x 2 ) of P, numeralwise expressing G' in P'. A new sentence J'
can then be constructed in the diagonalization argument, and, with
only the notational replacement ofP' for P,J' forJ, ded' forded, G' forG,
andG' forG, the 'proof ofGodel's theorem given above forPthen applies
to the new system P'. Hence, we conclude that ifP' is consistent, then
it is incomplete.
The pattern of the preceding argument should be clear. If, for
example, we next add.!' toP' as an axiom, we obtain a new system P".
Following the same procedure as above, we can construct a new primi-
tive recursive relation ded"(A., a) and can use this relation to construct
a sentenCe J" which is formally undecidable in P", More generally,
consider a new set of axioms for a first-order 'model' p+ where p+
is an extension of P - that is, where all the theorems of P are theorems
of p+ as well. 24 Then as long as it is possible to 'effectively determine'
the deductions of p+ through the use of the Go del numbering - that is,
as long as ded+(A., a) is a primitive recursive relation - we can, follow-
ing the same procedure given for P, construct an undecidable sentence
.r of P'". In this way, then, we come 'to see' 25 that the system Pis
not only incomplete, but essentially incomplete in the sense that the
incompleteness of P cannot be remedied by the addition of a 'reasonable'
(i.e., primitive recursive) set of axioms.
We can, in fact, go further -we can consider a system!l'that 'extends'
P not only by adding additional axioms, but by extending the language
of P through the -addition of new constant, function, and relation sym-
bols. For example, we might consider the system obtained from P by
adding to it the 2-place relation symbol<, the 2-place function symbol
E, and the axioms 26
(i) 'fiX 'fly [x<S(y) =(x <y or x =y)]
(ii) 'fix (x < 0)
(iii) 'fiX 'fly (x <y or x = y or y <x)
(iv) 'fix [E(x, 0) = S(O)]
(v) "'x Vy [E(x, S(y)) = E(x, y) ·xj.
(The intended interpretation of< and£ inN is, of course, the inequality
relation and the exponentiation function, respectively.) In a manner
similar to the modification of the schedule of proofs that we made in
adding J to P as an axiom, we can modify the schedule so as to accom-
modate the symbols < and E and the axioms for them. In fact, in
section 3(d) of the last chapter27 the modifications in our schedule
168
 A STRUCTURE OF PROVING

that are necessary for adding a countably infinite number of function

symbols .ff', 1 ~ m, 1 ~ i, were already considered. In this way then,
we have seen some of the variations that can be made in the logical
system P that do not affect the 'structure' of the proof of Godel's
theorem, and we can further appreciate the essential incompleteness
of P in the following manner: if S' is 'any' consistent, formal system
that extends P - where the precise meaning of 'any' is now to be
understood as itself a discoverable and accountable feature of the
proof of Godel's theorem- thenS'is incomplete.

c
A Structure of Proving: The Availability to a Prover of the Proof of
GOdel's Theorem as a Structure of Practices; The Proof as the Pair
The-Proof/The-Practices-of-Proving-to-Which-That-Proof-is-
Irremediably-Tied
The aim of this section is to briefly summarize the discussion of the
lived-work of proving Godel's theorem and of the proof of that
theorem as lived work.
In the preceding section, I reviewed the proof of Godel's theorem
for the formal system P, and then indicated how that proof can be
generalized to any 'reasonable' extension of P. Following the publication
of Godel's original paper in 1931, Godel's work in proving his incom-
pleteness theorem for a simple theory of types was (in a certain sense)
sharpened by Rosser28 so as to avoid the necessity of assuming w-
consistency; Godel's notion of 'rekursiv' functions, as found in the
1931 paper, was refonnulated (in part, by Godel himselt) 29 as defining
the subclass of primitive recursive functions of a larger class of functions,
this larger class then being designated as the recursive functions(proper);
the notions of 'nurileralwise expressibility,' 'representability,' 'defma-
bility ,' and 'decidability' were developed and applied; Church30 showed
that the set of theorems of formal number theory - i.e., the set of
Godel numbers of the theorems of formal number theory - is un-
decidable - i.e., not recursive; Tarski 31 demonstrated that the syntax
of a formal system like P is not adequate to express its own semantics
- i.e., that the set of Godel numbers of the wffs of formal number
theory that are 'true' in N is not 'arithmetical'; the question was raised
as to the generalizability of Godel's theorem - i.e., the question of
which formal systems have results comparable to the incompleteness
theorem; new topics were developed such as the notion of degrees of
unsolvability; Gentzen32 was able to prove that a different formulation
of formal number theory was consistent, opening the possibility of the
reinterpretation of Hilbert's notion of 'fmitary' methods for meta-
mathematical investigations, etc., etc.
This brief indication of the tremendous consequentiality of Godel's
169
A STRUCTURE OF PROVING

work for, and of the pervasive influence of his work on, classical studies
of mathematical practice might invite us into the reflection that the
•meaning' of Godel's theorem and of its proof have an historical and
historically unfolding character, that Godel's work was motivated by,
and set within, a specific historical context, that the ways in which we
look at and interpet that theorem today is influenced by the various
ways in which that theorem is embedded in its received histories, and
the like. But it is against this envisioned historical flux that the proof of
Godel's theorem, then as now, stands as a recognized and recognizable
achievement as the naturally accountable proof of the theorem that it is
seen, from within the lived-work of its proof, to witnessibly and
demonstrably prove. As I have tried to show over the course of this
book, that achievement cannot be attributed to a work-transcendent
state of mathematical objects, but, instead, the origins and substance
of it must be found within the lived-work of proving GOdel's theorem
itself. This problem was formulated in Section A of this chapter as the
problem of specifying, as lived-work, what identifies a proof of
Godel's theorem as the proof of just that theorem -what, as practice,
is identifying of that proof as its own demonstrably exhibited achieve-
ment. If we understand by a •solution' to this problem a formulation of
it that provides further descriptive and technical access to the problem
itself as a problem, for provers, in the local production of social order,
then, in the remainder of this section, I will try to suggest such a
possible solution.
Let us return to the fact that the proof of Godel's theorem for the
system P can be •generalized' to any •reasonable' extension a' of P.
The first point that needs to be made is that this sense of generalization
is not the same as the generalizations, sharpenings, modifications
and reformulations of that theorem and its proof that followed the
publication of Godel's original paper. As I tried to indicate in the
preceding section, the importance of Godel's theorem (for P) is not just
that the system P is incomplete, but that Pis essentially incomplete,
and the generalizability of the proof of Godel's theorem for P to
extensions of that system is itself an integral and necessary part of the
proof ofGOdel's theorem.
Now the reader will recall from Section B that the generalization
of Godel's theorem to extensions of P was obtained by first proving
Godel's theorem for P and then indicating how that proof provided
for both the changes that were needed, and the modifications in the
proof that were needed to incorporate those changes, for such a general-
ization. In this way, then, in that the generalizability of the proof is
made available to the prover by his finding/being-directed-to-find the
extractable, accountable structure of the proofs original argument, the
proof of Godel's theorem has an explicitly self-referential character:
the proof of Godel's theorem closes by pointing to the accountable
170
 A STRUCTURE OF PROVING

structure of that proof such that that proof is applicable to 'any'

fonnalization of number theory having a 'reasonable' set of axioms.
The important point that needs to be made is not that our proof
of Godel's theorem has such an explicitly self-referential character, 33
but that that proof-specific self-reference points to the availability,
to a prover, of the proofof Godel's theorem as a structure of proving
- that is, as an organization of the practices of proving for proving-
again Godel's theorem. Summarily, that structure of proving can be
construed in tenns of the introduction of, and the intentionality of the
introduction of, a Godel numbering; the fact that that numbering
allows the syntax of a fonnal system to be 'arithmetized'; the intro-
duction of primitive recursive functions and relations and the construc-
tion of the arithmetized syntax as consisting of such functions and
relations; the fact that, by showing that primitive recursive relations
are numeralwise expressible in the fonnal system, the arithmetized
syntax can be expressed in the syntax itself and that, in consequence,
the diagonalization argument/'proof can be given. The important
point, however, is not this list of topics, but that this list is, in fact, a
reference to an exhibited, orderly course of proving; it is a reference
to the actual lived-work of proving-again Godel's theorem. These
topics and their arrangement provide an indication of how to prove
Godel's theorem - of a way of working that proves G6del's theorem
- and it is the availability to a prover of the proof ofGodel'stheorem
as such a structuring of the practices of proving that make it pointless
to dispute that proofs witnessed achievement. 34
As an initial summary, then, we can say that a proof of Godel's
theorem, as a structure of proving's practices, provides a course of
practical reasoning and action - namely, the course of reasoning
and action that is the identifying, naturally accountable, lived-work
of proving-again Godel's theorem. In this way, the proof of Godel's
theorem is a pedagogic object- it teaches provers how to prove Godel's
theorem, and it does this by providing, metaphorically, in the material
proof, a template of that course of action. More descriptively, however,
we must ask how that organization of practices is achieved in and as the
detailed, lived-work of proving Godel's theorem. What can now be
proposed is this: a proof of Godel's theorem is itself an intrinsically
paired objectl5 the-material-proof/the-practices-of-proving~to-which­
that-proof-is-irremediably-tied; a proof of Godel's theorem is produced
as such a paired object; it is recovered as such a paired object; and it is
in the realization of a proof ofGodel's theorem as such a paired object
- and, therein, in the way in which the material proof is constructed
as, and comes to stand as, an account of the work of its production,
thereby articulating the proof-pair as such a pairing of material detail
and practice -that a proof ofG6del's theorem is simultaneously realized
as a structure of practices for proving, identifiably, just that theorem.
171
Part III Conclusion
9 Summary and Directions for
Further Study

A
Classical Studies of Mathematical Practice: A Review of the Book's
Argument 1

In the introduction of this book, I introduced the notion of classical

studies of mathematical practice - that is, of studies of the rigor, or
natural accountability, of mathematicians' work for which the essentially
local character of that accountability is irremediably hidden in and as
the local work practices that make up the natural accountability of
those studies themselves. I then proposed that 'conventional' studies
of the foundations of mathematics are, in fact, such classical studies.
In order to develop this proposal, it was necessary to gain real-world
access to the rigor of a mathematical argument as a local phenomenon,
and the body of this book was an attempt to do this through an
extended descriptive analysis of the lived-work of proving one of
Godel's incompleteness theorems.
The book's argument was developed, roughly, as follows: among
studies of the foundations of mathematics, Kurt Godel's two incom-
pleteness theorems are among the most celebrated. The first of these
theorems 'says' that any consistent formal system strong enough to
'represent' elementary arithmetic is incomplete - that is, roughly, that
there are true propositions of arithmetic that cannot be proved in such
a formal system. Godel's second incompleteness theorem 'says' that
the consistency of such a formal representation of arithmetic cannot
be demonstrated within that system itself or, even more loosely, that
the consistency of mathematics cannot be proved mathematically.
In attempting to give a precise description of these results in an
early chapter, we found that the accuracy of such a description
is tied to the material detail and lived-work of actually proving
those theorems, and, thus, even more importantly, we found that the
175
SUMMARY AND DIRECTIONS

interpretability of those theorems rested on the prior availability of

their proofs as naturally accountable mathematical arguments. The
specification of what, as praxis, makes up the naturally accountable
rigor of the proofs of Godel's theorems is necessarily prior, as a founda-
tional· question, to the interpretation of those theorems in terms of
mathematical practice.
The body of this book was devoted to an extended analysis of the
lived-work of proving Godel's first incompleteness theorem. Through
that analysis I argued in an evidential manner that the rigor of a proof of
Godel's theorem consists of its local, lived-work - that a proof of
Go del's theorem was produced, in such a way so as to be the recognizably
and analyzably adequate proof of just that theorem. Although this
argument was made through the particular analysis of the work of
proving G6de1's theorem, what was uncovered as the work of that proof
resonates throughout mathematical practice: by attending to the work
of proving Godel's theorem particularly, we gained material and techni-
cal access to the work of proving. The generality that was uncovered
was not one of fmdings, but of the discovered existence of a phenome-
non - that of the naturally accountable proof. In this way, then,
the analysis of the proofofGodel's theorem pointed to the discoverable
origins of mathematical rigor in and as the lived-work of doing naturally
accountable ordinary mathematics and to mathematical rigor as itself a
problem, for provers, in the local production of social order.
To this point in our analysis we have treated the proof of Godel's
theorem as a proof of ordinary mathematics independently of its
'metamathematical' interpretation. Godel's theorem is, of course, one
of the great studies of the foundations of mathematics, and it is at this
point that the book's argument turns back on itself: in so far as con-
ventional foundational studies - like Godel's theorem and its proof -
are mathematical, they themselves are irremediably tied to the practices
of doing recognizably adequate, rigorous mathematics; as does ordinary
mathematical practice, conventional studies of mathematical founda-
tions hide the origins of mathematic rigor in and as the rigor of their
own, local, lived-work. Thus, the book returns to its opening proposal,
that 'conventional' studies of the foundations of mathematics are
properly characterized and investigated, as praxis, as classical studies
of mathematicians' work.
In addition to the proposal of the existence of classical studies of
mathematicians' work, the argument of this book has a further conse-
quence for the study of the relationship between mathematical practice
and the classical study of that practice. We would like to know if there
is something 'natural' in the pairing of classical studies and mathematical
practice; we would like to know if the origins of classical studies lie
within mathematical practice itself. We have seen in this book that the
rigor of a mathematical proof not only lies within, but is hidden within,
176
 SUMMARY AND DIRECTIONS

that proofs lived-work, that a proof consists of the pair the-material-

proof I the-practices-of-proving-to-which-that-proof-is-irremediably-tied,
and that a proof is cultivated so as to realize the material proof as a
disengaged version, or account, of that proofs lived-work. In this way,
a mathematical proof is itself a classical study of its own practices.
Conjecturally, then, a mathematical proof is itself a classical study of
practical action; it is a classical study of the work of mathematical
theory proving, and mathematics, as a discipline, is a classical science
of practical action and practical reasoning. 2

B
Prospectus: Mathematicians' Work as Structure BuDding

In this book I have addressed the problem of the nature and consti-
tution of mathematical rigor - often spoken of as the problem of the
foundations of mathematics - as a problem, for provers, in the local
production of social order, and I have attempted to show that so
formulated the foundations of mathematics can be investigated in and
as the inspectable details of mathematical practice. This book is the
beginning and not the end of this project. What I hope to have demon-
strated, however, besides the feasibility of this project, is that one of
the tasks that does not remain is that of building philosophical, his-
torical or sociological elaborations of this way of posing foundational
questions. The material in this book provides, I hope, a starting point
for the investigation, as a real-wordly researchable matter, of what
makes up the natural accountability of mathematicians' work; it
provides a starting point for the discovery of what mathematical proofs
identifiably consist of as lived-work, and it provides a starting point
for the further examination of the proposal that the rigor of a proof is
irremediably tied to the work of its local production. Moreover, in that
the discovery of a mathematical proof is the discovery of a naturally
accountable way of proving, this book provides an initial means of
animating the question of the nature of mathematical discovery, not
by romanticizing or psychologizing the mathematician, but by provid-
ing access to a mathematical discovery as lived-work.
In bringing this book to a close, I want to suggest a topic for further
exploration which, because of the singular concentration in this book
on the descriptive analysis of the lived-work of proving Go del's theorem,
could not lre developed in the book itself. The idea is this: for mathe-
maticians, the heart of their profession is not theorem proving per se;
that it is is itself a classical version of their work. What I wish to propose
is that the sustaining life of professional mathematics lies in what might
be called 'mathematical structure building' - that is, the envisionment
and construction of structures of theorems and proofs that have their
177
SUMMARY AND DIRECTIONS

motivating origins in, and are directed to the development and reformu-
lation of, a current state of mathematical practice. What at this time I
can only suggest is that the naturally accountable work of theorem
proving provides essential and unique access to mathematical structure
building; without knowing, as praxis, the work of theorem proving,
the discovery, construction, recognition and consequentiality for
mathematical practice of such mathematical structures is impossible.
In closing, then, the questions that I wish to raise are whether or not it
is possible to find mathematical structure building in and as the lived-
work of doing professional mathematics; whether or not, by using the
now available material on the lived-work of naturally accountable
proving, it is possible to discover a natural technology in mathematicians'
situated inquiries into mathematical structures; and whether or not, if
such a natural technology exists, that technology is integrally tied to
creativity in the work of mathematical discovery.

178
Appendix
The Use of Ethnomethodological
Investigations of Mathematicians'
Work for Reformulating the Problem
of the Relationship between Mathe-
matics and Theoretical Physics as a
Real-World Researchable Problem in
the Production of Social Order 1

Over the course of the past several years, I have been engaged in studies
of mathematicians' work and, particularly, the study of the relationship
between that work and the foundations of mathematics. The central
theme of those investigations has been that the problem of the founda-
tions can be formulated as a problem in the production of social order
and that so formulated, it can be investigated in and as the work of
mathematicians in locally producing, for and among mathematicians,
accountably ordinary mathematics. This appendix serves as another
introduction to those studies by addressing their possible consequential-
ity for another problem in the foundations of science: the problem of
characterizing the relationship between mathematics and theoretical
physics.
The problem of characterizing the relationship between mathematics
and physics has its origins in the omnipresence, utility and effectiveness
of mathematical formalisms and derivations in the work of theoretical
physicists. The question that is usually asked - what is the nature of
mathematics and the nature of physics or physical reality that provides
for such a mysterious compatibility? - has the effect of turning the
real-worldly, material investigation of the mathematical practices of
physicists into a problem concerning the naturally and constructively
theorized content of the two disciplines.
The aim of this appendix is to indicate that the problem of charac-
terizing the relationship between mathematics and physics, as it is
traditionally conceived, is a pseudo-problem; that that problem can be
reformulated as a problem in the study of the production of social
order} and that that reformulation makes the mathematical methods
of physicists capable of being empirically researched as an autonomous
praxis.
The idea of my presentation is not to argue each of these claims
separately, but to present an argument that provides, at once, for their

181
APPENDIX

joint intelligibility. Mathematical studies of the so-called foundations

of physics have, as a major goal, the development of properly mathemati-
cal methods 'substantiating' the mathematical methods actually used by
theoretical physicists. I will argue first, that the (theoretically presumed,
or programmatically established) existence of such correspondences
needs to be separated from the idea that such a correspondence confers
a platonic or transcendental character on the mathematical reasoning
of physicists, and second, that once this is done, the descriptive ade-
quacy of physicists' mathematical reasoning becomes a curious and
interesting phenomenon in its own right.
To this end, I am going to begin by offering a contrast between the
mathematical practices of physicists and those of professional mathe-
maticians. A mathematical theorem - the divergence theorem - will
be briefly discussed first from the perspective of physicists and then
from the perspective of mathematicians. That contrast will set in relief
the presupposition of traditional attempts to characterize the relation-
ship between mathematics and physics that the mathematical practices
of physicists can be spoken of as, and therein identified with, those of
mathematicians. Later in the appendix, I will return to the physicists'
proof of the divergence theorem and examine it as a form of instruction
in a method of physical reasoning.
A typical statement and proof of the divergence theorem from the
perspective of physics instruction is the following: let S be a 'closed
surface' and D, the surface S and its interior. Let g = (g 1 , g2 , g3 ) be
a vector field defined on D and It= (n 1 , n 2 , n 3 ), the outward unit
normal vector field onS. Then the divergence theorem says
fs <l,it>dA = J0 (~· t) dV.
A picture will convey the physical significance of the theorem. In the
figure below, let S be the 'closed surface,' let g r~resent the flow of
some physical process, like heat, through S, and let n be the unit normal
vector field on S.

The left-hand side of the divergence theorem is called the flux of the
vector field through Sand measures, for example, the total heat current
directed out from the surface. The divergence theorem says that the
182
 APPENDIX

flux of the vector field through the surface S equals the integral of the
divergence of the field taken over the interior of S.
A proof of the theorem can be given in the following fashion.
The first step of the proof is to reduce the problem of proving the
theorem to that of proving it for an infinitesimal cube. This is done by
showing that the total flux of a vector field through a volume is equal
to the sum of the fluxes out of each part of the volume when that
volume is dissected into smaller pieces.
The second step of the proof begins by considering an infinitesimal
cube.

Coordinate axes are arranged so that they line up with the edges of the
cube.

The infinitesimal lengths of the sides of the cube will be denoted dx,
dy and dz. In the figure, we also depict the unit normal on one of the
faces along with the vector gat
that point.

I
I
I
dy
}---
;'
;'
;'

"' dx
The flux of the vector field gthrough the cube,
fc <&it>dA=
will equal the sum of the fluxes through each face of the cube. Using a
183
APPENDIX

first-order linear approximation to calculate the flux through two

opposing faces, we get

f c (~
g, it> dA --~
ax dx dy dz + ...
and adding similar results obtained from the other sides
a!Jl . a!J2 a!J3
f C <l rt) dA = ~ dx dy dz + 3y- dx dy dz + ~ dx dy dz.
The sum on the right equals the divergence of g times the volume of
the infinitesimal cube,
('il • g) /), v.
Using the first step of the proof, we sum these infinitesimal volumes
throughout D and arrive at the divergence theorem.
From a standpoint of rigorous mathematics, there are a number of
difficulties with this proof, even and especially when that proofis given
in greater detail. Although the existence of such difficulties lies at the
heart of the study of the mathematical foundations of physics, the
particular troubles in the case-at-hand are incidental to the present
discussion. What is important is to offer, at least on the basis of immed-
iate visual perception, a direct contrast between the physical version of
the theorem and the theorem as seen by professional mathematicians.
An actual proof of this theorem would get too involved, but I hope to
convey some sense of that proofs technical character.
For mathematicians, the setting of the theorem is immediately
generalized to an object of arbitrary finite dimension. 3 Let M be a k-
dimensional, compact, oriented C2 manifold-with-boundary and let w
be a k- 1 differential form on M. Then the generalized Stokes' theorem
says that
fM dw=faM w
where am has the induced orientation. This theorem can be proved to
the satisfaction of professional mathematicians and, once proved, it can
be specialized to the divergence theorem in the case of 3 dimensions.
To do this, one defines a 2-dimensional volume element on the tangent
space ofM

dA(vx, wx) = det ( : ) .

n(x)
It can be shown that the following relationships hold:
n 1 dA = dy 1\ dz
n2 dA = dz 1\ dx
n3 dA = dx 1\ dy.
184
 APPENDIX

Then, by letting
w = g1 dy Adz + g2 dz Adx + g3 dxA dy,
a 2-form on D, and substituting these definitions and relationships into
Stokes' theorem, a precise though appearientially similar version of the
divergence theorem will result.
This. presentation of the divergence theorem illustrates the discrep-
ancies between physicists' mathematical reasoning and that of profes-
sional mathematicians. The initial point that I want to make is this:
in the reflective discussion of the relationship between mathematics and
physics, physicists' mathematical practices are predominantly spoken
of as, and therein identified with, those of professional mathematicians.
In the case-at-hand, the physicists' proof of the divergence theorem
is understood to stand proxy for the more rigorous methods of
mathematics. This presumed identifiability of methods underlies the
traditionally-conceived problem of characterizing the relationship
between the two disciplines.
The obvious descriptive inadequacies of such an identification -
illustrated by the example of the divergence theorem and uniformly
recognized and acknowledged by mathematicians, physicists, and
formal methodologists of science - makes the pervasiveness of its
presupposition extremely curious. There are, I believe, two comple-
mentary reasons4 for that pervasiveness, the first of which will be
conveyed through another example. At the end of the appendix, I
will try to bring these examples together.
In that physicists derive the heat equation from Fourier's law of
heat conduction and the law of the conservation of energy (we will
write this as follows),
Ji =-KV'U
,j.
ess
ll.u = ut
it is generally supposed that the adequacy of Fourier's law as a descrip-
tion of a physical situation will insure the similar descriptive adequacy
of the heat equation. 5 Another way of putting this is that the mathe-
matical derivation (represented by 'the arrow' in the figure above) is
presumed to be analytic, or that it is understood to preserve the truth-
value of the mathematical descriptions.
In practice, when they are actually engaged in their work, physicists
have a much more circumspect regard for their derivations and mathe-
matical descriptions than this account provides. Nevertheless, the account
does reflect the way in which physicists will sometimes speak about
their mathematical practices when, no longer actually at work, they
come to speculate reflectively about them. In that their mathematical
185
APPENDIX

derivations are constructively developed over the course of collegial

work sessions and blackboard discussions, those derivations have
their retrospectively considered production independence naturally
embedded in and as their own communally recognized and accountable
efficaciousness.
At this point, I am going to introduce a distinction between the
practical analyzability and accountability of physical mathematics
as that mathematics is encountered in the work arrangements of theor-
etical physicists and the problematic character of that mathematical
reasoning when it is disengaged from that situated, collegial work. I
will refer to the work of that disengagement as the work of an objecti-
fying science. It is from the perspective of an objectifying science that
the retrospective efficaciousness of the mathematical reasoning of
physicists poses the problem that that efficaciousness could not (at least
seemingly) be dependent on the idiosyncracies, circumstantialities, and
even dubiously legitimate mathematical methods that pervade the
mathematical work of physicists. That such methods actually compose
such efficacious derivations is attributed to the existence of rigorous
mathematical methods to which physicists' mathematical reasoning
·alludes and for which it stands as a degraded counterpart.
A second and closely related basis for the thesis of the identifiability
of the mathematical methods consists of the unavailability of non-
formal methods for technically analyzing the actual use of mathematics
by physicists, The known methods by which the reasoning of physicists
can be techilically and critically analyzed ~ again from the standpoint
of an- objectifying science ~ are those of mathematics proper, and,
consequently, to be engaged in the mathematical review of physicists'
writings is to already be involved in a reconstructive enterprise that has
the identifiability of methods as its programmatic ideal.6
In order to summarize all this, let us depict the assumed existence
of a platonic mathematical counterpart to physical reasoning with a
double arrow so that the initial situation would be mirrored in another
situation .written as follows:
h = -KV'U
+
ess ess
6-u = ut 6-u = ut
The demonstrated existence of such properly mathematical methods
are understood to supply the foundations for the mathematical methods
of physicists in that they ·are seen as providing an account of those
methods .that is; at least apparently, independent of the actual methods'
situated production and use.
Having laid out this material, I am now in a position to come to the
main point of the argument. As I indicated earlier and have presented in
186
 APPENDIX

the previous chapters, my investigations of mathematicians' work

allow me to evidentially argue that the rigor of mathematicians' reason-
ing consists of the essentially local work of mathematicians in producing
for and among mathematicians the accountably ordinary objects of
mathematical discourse. Thus, that the mathematical practices of
physicists can be shown, in some sense, to correspond to those of
mathematicians (as depicted in the diagram above) can no longer be
seen as providing an analytic foundation for physical reasoning. Instead,
the situated achievement of some mathematicians in showing that such
partial identifications can be made becomes available for what, in fact,
it is: the demonstration of the partial translation of one set of practices
into another, demonstrated to the satisfaction of one of the groups of
practitioners, through which suggestively mathematical methods
become rendered as properly mathematical ones.
That the foundations of mathematics, properly conceived, reside in
and as the practices that make up mathematicians' work free the
question of the relationship between mathematics and physics from the
goal of detaching that relationship from its dependence on the situated
features of physical theorizing, and return the question to its origins
and substantiating conditions in the practice of theoretical physics.
Thus, we are taken to the heart of the matter - whether or not it is
possible to recover the practical analyzability and accountability of
physicists' mathematical methods as, at once, a deeply technical enter-
prise (as the mathematicians have it) and, also, as an enterprise whose
analyzability is essentially available as local, situated accomplishments.
I do not have an answer to that question, but I would like to offer
one last example as a way of clarifying it and drawing this material
together. Returning to the derivation of the heat equation from
Fourier's law, one way of speaking about that derivation (the 'arrow'
in the illustration) is to say that it depends centrally on the divergence
theorem. A more or less formal rendering of that derivation would
make that appear to be the case. Instead, let us look at a more physical
derivation.
The flow of heat through a solid is, again, reduced to the problem of
the flow of heat through an infmitesimal cube.

Letting gin our previous derivation be replaced with It, the heat current
density, the same reasoning as before gives the total heat current out
187
APPENDIX

of the cube as the divergence of Ittimes the volume of the cube,

(V'. It) t:N.
As the heat in the object will be proportional to its temperature, the
negative of the change of temperature will, by the law of the conser-
vation of energy, equal, up to the constant of proportionality, the loss
of heat through the surface of the cube; that is, it will equal the flux.
Thus,
(V' • K) D.V = fc- ( ~ it) dA = - p Ut D.V,
where p is the constant of proportionality and the subscript indicates
partial differentiation with respect to time. Substituting Fourier's law
for the heat current density It, one obtains
-K V' • V'u =-put
or, essentially, ~-

ess
D.u = Ut

the heat equation.

What this example seems to do is to make the physicists' proof of
the divergence theorem given earlier available as the embodiment of a
method of reckoning. The physicist's proof of the divergence theorem
serves as instruction in a method of reasoning that, as a method, has the
interesting feature that it preserves the physical interpretability of the
equations throughout the derivation. The efficacy of the derivation as it
concerns physical phenomena does not originate in, nor is it established
and maintained by, a correspondence between that derivation and
mathematicians' conceptions of proper deductive inference. Instead, it
is tied to the constructively and developmentally provided physical
interpretability of the equations as that interpretability is constructed
and maintained throughout the derivation.
The presence and use of mathematical methods that preserve the
interpretability of physical descriptions offers a promising way of
initially examining theoretical physicists' actual mathematical reasoning.
However, I think it would be wrong to begin looking for instances of
structures throughout mathematical physics as straightforward as the
example of the divergence theorem, particularly as a means of construc-
ting a documented argument. The relationship between the mathematical
formalisms and derivations of physicists and their descriptive content is,
inevitably, __going to tum out to be extremely subtle in that it must
depend on the ways in which a production cohort can make that
descriptiveness available through and as the details of their work.
What the example does suggest is that the mathematical practices of
physicists can be examined independently of mathematicians' practices

188
 APPENDIX

and that the foundations of mathematical physics might well consist of

the concrete, exhibited/exhibitable ways in which that descriptiveness
is produced and maintained as a situated accomplishment.
In the philosophy, history and sociology of science, the problem of
characterising the relationship between mathematics and physics has
a docile presence - that problem is disengaged from considerations of
physicists' actual work practices; it is amenable to the endless discourse,
historical reviews, and argumentation that make up the practice of
those disciplines; it is ignored, commented on, addressed or put aside
as a matter of choice. In the work of physicists, where the physical
adequacy of mathematical methods is a continual and crucially impor-
tant concern, that problem takes on a different life entirely. For
physicists, the problematic features of the relationship between mathe-
matics and physics are both practically formulated and practically
solved as situated, daily features of their work. By examining that
relationship as the work of locally producing practically objective
mathematical descriptions, the liveliness of that relationship is re-
discovered, and the problem of characterizing that relationship is
reopened as a problem in the study of the production of social order.

189
Notes

Introduction

The distinction between classical and ethnomethodological

studies of work, due to Harold Garfinkel, is a technical one, and I
will only elaborate on it here in so far as it applies to the study of
mathematicians' work and only in so far as a few brief comments
may initially help the reader. The increasingly technical access,
as real-worldly researchable matters, that current ethnomethodo-
logical studies of mathematicians' work provide to the
investigation of the rigor of a mathematical argument as that rigor
is essentially tied to the lived-work of that argument's production
and exhibition provides as well technical access to the witnessed,
yet ignored character of that connection in previous studies of
mathematicians' work. In that these previous studies rely on the
lived-work of doing rigorous mathematics to find and identify
naturally accountable mathematical argumentation - and,
thereby, to furnish the adequacy and cogency of their own
studies - the essential connection between that lived-work and
mathematical rigor is irremediably hidden within those studies in
and as the efficaciousness and adequacy of their own practices.
Rather than speaking in a non-technical manner of 'traditional' or
'conventional' studies of mathematicians' work, we speak of
these studies as 'classical' studies and, thereby, point to and seek
to elucidate those studies as themselves practical enterprises tied,
for their practical accountability, to the various professional
disciplines from within which they arise and have their proper
and accountable origins. In brief, then, the distinction between
classical and ethnomethodological studies of mathematicians'
work recommends the existence of classical studies as its own
phenomenon in the study of practical action and practical
reasoning.
2 As far as possible, given the constraints of the meetings, the
intention of my talk was to bring the reader into the presence of
190
 NOTES TO PAGES 2-8

the lived-work of doing mathematics and, by doing so, to offer

that work for analysis as practical action. In presenting this
material in a lecture, a prover uses his embodied presence to the
blackboard and to the audience to achieve the exhibited precision
of his work and talk. Thus the fhere:l- noted in the text is
replaced or accompanied by a pointing out - by indicating the
tracing of - the intercepted arc

where the drawing itself is a temporally developing, temporally

organized achievement that can be indicated as follows:

OOG
e0®®
In presenting the talk as a written text, I have retained the use of
multiple pictures, arrows and bracketed indexical expressions that
initially served the purposes of a prepared lecture to provide
instruction for the reader for finding, in the definiteness of the
mathematical thing that is then found through their use, the
lived-work to which that definiteness is tied. Thus, the reader
will not be offered a key to the literary devices being used in the
text, but, instead, is advised to seek out in the text's surrounding
materials, just what those devices come to, as practical action,
in and as the definite thing that they come to exhibit, and to
attend to that work as the lived-work of mathematical theorem
proving.
3 As an illustration of a 'theory' of such objects, consider, for
example, the following definitions of an angle and of its measure
(taken from Gustave Choquet, Geometry in a Modern Setting
(Boston: Houghton Mifflin Company, 1969), p. 79): Let,f+ be
the group of even isometries of the Euclidean plane, andTthe
normal subgroup of translations. Then an angle can be defined as
an element of the factor group
.!11= .J+I r,
the angle between a pair of half-lines (A,B) with common vertex
as the canonical image in the factor groupJ+/5"of the isometry
transforming A into B, and a measure of angles as a continuous

191
NOTES TO PAGES 8-11

homomorphism ..p,
..p: R~d
from the additive group of reals with the natural topology to the
angles topologized by first identifying the angles with the multi-
plicative group of complex numbers of modulus unity. The
measure of a given angle a is then defined as the inverse image of
a under ..p, m(a) = ..p- 1 (a). A less exotic example would be the
following definition of an inscribed angle: 'Let A and B be
distinct points of a circle and let AB be one of the arcs of the
chord AB. We say that an angle is inscribed in this arc if its vertex
X is a point of the arc and if its sides are the rays [XA and [XB.'
(Howard Levi, Foundations of Geometry and Trigonometry
(Englewood Cliffs: Prentice-Hall, Inc., 1960), p. 280.)
The reader should note that the notation in the preceding
definition speaks on behalf of the work of developing that
notation so as to be able to disengage the mathematical objects
under consideration from the remarkable relevancies of the work
of proving for which that notation was developed as just such a
device.
4 The reader may here wish to argue that the need for 'rigorous'
definitions of these objects is exhibited from within the local
work of the proof. This is certainly the case: as in note 6 below,
on entertaining the possibility that

8
depicts an angle and, thereby, on seeing the possibility of a fourth
case involved in proving our theorem, a prover, having in just this
way raised the question of what an angle actually is, would find
for himself, as the witness to the natural accountability of his
own work, the necessity of formulating a definition of an angle
which would exhibit that definition's independence from the
problematic detail of the proof for which that definition provides
a· solution. The point to be made is that the need for such a
definition and the adequacy of that definition are both, and
irremediably, local accomplishments tied to the analyzable detail
of an exhibited angle that such a definition is seen to provide
and to recover. In this way, then, such definitions not only arise
from within, but are answerable to, and must exhibit their
necessary properties as part of, the local, lived-work of the proof
itself·.
5 Let me briefly sketch a different proof of our theorem. (The
reader is referred to Levi, Foundations of Geometry and Trigon-
ometry for details.) The idea of the proof is to show 'first that the
measure of an inscribed angle depends only on the measure of
the intercepted arc. Given this result, the inscribed angles are

192
 NOTES TO PAGES 11-14

partitioned into the class where the intercepted arc is less than
180 °, the class where the intercepted arc equals 180 °, and the
class where the intercepted arc is greater than 180 °. For the first
case, we need only prove our theorem for the particular case
where one of the ~dges of the angle is a diameter of the circle.
The proof woul~ be the same as that of the first case in our
original proof, For the second case, we need to prove the theorem
only for the situation depicted below:

And in the third case, we can proceed by proving that angles

inscribed in opposite arcs of a circle are supplementary

m(o:) + m(p) = 180°

from which case three then follows.

To make this argument work, however, a prover would 'first'
('first' reflecting the locally produced and exhibited orderliness
of the work of proving the theorem) have to show that all the
angles inscribed in the same arc of a circle are, in fact, congruent
to each other, and to do so, a number of other theorems need to
be articulated and proved - e.g., one that states that given four
distinct points~· . , A4 on a circle, exactly two of the chords
A1A2, A1A3, A 1A4, A2A3, A2A4, A3A4, have a point in
common and one that states that when the length of two sides
of one triangle are proportional to the lengths of two sides of
another triangle and, in addition, the corresponding included
angles of the triangle are congruent, the triangles are similar.
The existence of this other proof of our proposition allows me
to advance another claim: what is involved in proving such a
theorem is the articulation of a locally discovered and locally
constructed course of proving as a chain of dependent
propositions and that, for the mathematician, the heart of his
work is not the proving of theorems in themselves, but the
mathematical structure building to which that proving provides
essential and unique access and, thus, that those structures are
themselves local accomplishments or, said differently, that the
structures consist of their demonstrable and exhibited properties
as built structures and that those structures are 'nothing' other
than the organized practices of their construction.
6 As an aside, let me note that a prover, on seeing the possibility
of including

193
NOTES TO PAGES 11-28

8
as an inscribed angle, might not only admit e as the repre-
sentative of an additional equivalence class of inscribed angles,

defining the measure of e

but in proferring the proof of our theorem for this case by
so that the theorem holds, a

explicating the definition of an angle such that e.

prover might review his proof and come to see the need for
would be
seen to be the one as well. Moreover, the definition would have to
be given so as to preserve the properties of angles and measures
of angles exhibited over the course of, and needed for, the proof
of our theorem. The import of this aside is not to suggest the
transcendental necessity of such a definition; I have already argued
that the need for such a definition is itself an achievement of a
prover's local work and that the adequacy of such a definition is
essentially tied to that work for that adequacy. Instead, the
import of this example is that it points to the radically self-
organizing (metaphorically, 'hermetic') character of a
mathematician's work in producing and exhibiting accountable
mathematical structures.

Part I Introduction

My use of the word 'rigor' is adopted from conversational usage

among mathematicians where it is used to refer to the witnessible
(and, thereby, to the apparent, visible, recognizable) adequacy of
a proof in demonstrating its own truthfulness, its own objectivity,
its own accountability.

Chapter 1 A Review of the Classical Representation of Mathematicians'

Work as Formal Logistic Systems

This formalization of the axioms of group theory is a slight

modification of the one found in Donald W. Barnes and John M.
Mack, An Algebraic Introduction to Mathematical Logic (New
York: Springer-Verlag, Inc., 1975), p. 41.
2 Both the informal axioms for G and the formal ones for G* could
be 'weakened,' but such considerations would have greatly
complicated our presentation.
3 To the novice, the notion of 'translation' will be seen to
obviously trivialize the work involved in being able to 'see' a
formal sentence as the translation of a statement of ordinary
mathematics.
194
 NOTES TO PAGES 29-42

4 'Elementary' is used here in its usual sense as 'simplest' or 'most

basic.'

Chapter 2 An Introduction to Gooel's Incompleteness Theorems

1 'Reasonable' refers to the condition that the set of Godel

numbers for the axioms of the system is a primitive recursive
relation.
2 As it is standardly rendered in P.
3 By slightly modifying GOdel's argument, Rosser (1936) was able
to show that the flrst result concerning regular consistency also
holds. However, that modification introduces changes in the
intuitive interpretation of the undecidable sentence. See F.
Feferman, 'Arithmetization of Metamathematics in a General
Setting,' Fundamenta Mathematicae 49 (1960), pp. 35-92.

Chapter 3 Gooel Numbering and Related Topics

1 Provisional on the complete specification of P.

2 That f is ''well-deflned" requires proof. I take this occasion to
remind the reader explicitly that I do not intend my treatment of
Godel's theorem to itself be 'rigorous,' but to examine, at
perspicuous places in a proof of Godel's theorem, what constitutes
the 'rigor' of such a proof. The material reviewed is intended to
recall for the reader various features of a proof of Godel's
theorem and to serve as a guide to the work of that proof.
3 E.g., that they are representable in P.
4 E.g., that _the representability of the functions g 1 , . . . , gn and h
in the definition of substitution and the representability of g and h
in primitive recursion insure the representability of the functions
obtained from them by substitution and primitive recursion,
respectively.
5 That this result is an immediate consequence of the deflnitions
is exhibited by the following orderly way of proceeding: (i) If
(a 1 , ••• , llro)E R, then KR(a 1 , ••• , am)= 1, and from the
defmition of representability, this implies 1--p. KR (k(ad, ... ,
k(am ), S(O)). (ii) If (at, ... , am) f1. R, then Kp(at, ... , am)
= 0; hence, 1--p KR(k(ad, ... , k(am), 0). If 1--p KR(k(ad,
... , k(am), S(O)), then 1--p 0 = S(O) by the uniqueness property
(2) of the deflnition of representability. The axioms of P insure
that 1--p 0 =I= S(O); hence; by a tautology, 1--p '""KR (k(a 1 ), ••• ,
k(am). S(O)). Thus,R(x 1 , ••• , Xm) = KR(x 1 , ... , Xm, S(O))
numeralwise expresses R in P. ~
6 The expressions of P consist of all flnite, concatenated sequences
of primitive symbols of P.
7 The actual practical character of these restrictions will not be
available until the material details of the 'structure of proving'

195
NOTES TO PAGES 42-7

of Go del's theorem are finally made analytically available.

8 Here I adopt the notational convention that (x, y) E G can be
written as G(x, y). The theoretical interchangeability of these
notations- available first in the elementary situations where that
convention is introduced and, later, through systematic methods
of replacing the predicative notation with the set-theoretic one -
will be contrasted with the essential character of the collective
use of notational conventions for the practical observability of
the rigor of the work of proving Godel's theorem.
9 'Consists' is contained in quotation marks because the identity
between the rigor of the proof and these 'demonstrations' is
available in and as the 'structure of proving' the theorem.
'Demonstration' is contained in quotation marks because these
demonstrations have yet to be explicated as the local work of
proving Godel's theorem.
10 To be discussed later.
11 As opposed to an actual, explicitly displayed formal construction
sequence.

Chapter 4 The Double-Diagonalization/'Proof'

There is some difference among authors as to what is designated

as the double-diagonalization. The sequence of definitions used in
constructing the formally undecidable sentence does contain
the double-diagonalization of the function subg(x )·The first
diagonal.ization is cp( •) =subg(x ) (·,·);the 'doubl~­
diagonalization' is cp (i). The apparent fact that cp (i) = j -
apparent in its specific placement within the work for which the
sequence of definitions stands as the material display - is
sometimes referred to as the 'double-diagonalization trick.'
(See below for the definitions of the functions and the numbers i
and j that are introduced here.)
2 The stress on this phrase points to the circumstance that the
interpretability of these formulas is tied to a way of mathe-
matically proceeding.
3 In writing up the diagonalization, I made use of the treatment of
it in Raymond Wilder, The Foundations of Mathematics, 2nd
edn, (New York: John Wiley and Sons, 1965), pp. 272-3.
4 The selection of x 2 is incidental to the 'proof, but must be
consistently maintained throughout the diagonalization and
'proof. That that selection is incidental is available from
reflective consideration of the work of the 'proof.
5 The practical techniques of working with primitive recursive
functions and relations will be addressed in a separate section
devoted to them.
6 Historically, knowledge of the sentence J as the-sentence-J-with-
its-undeducibility-properties was instrumental in formulating
Godel's theorem. The 'natural' theorem concerning undeducibility

196
 NOTES TO PAGES 47-53

is 'if Pis consistent, then there is a sentenceS of P such that

neither S nor ~ S are deducible in P .' Go del's discovery of the
sentence J as the sentence with its undeducibility properties, and
Godel's ability to formulate that sentence's inadequacy for
obtaining the proof of the 'natural' theorem as w-consistency,
allowed Godel to arrive at the statement of his first incomplete-
ness theorem.
7 As a gloss for the reasoner's work of its facticity.
8 I am borrowing the term 'formating' from Harold Garfinkel.
Garfinkel speaks of a formated queue as a queue which exhibits,
in and as the positioning of its production cohort, its queue-
specific order of service.
9 More generally, to speak of a mathematical discovery is to
simultaneously speak of a discovered way of proving.

Chapter 5 A Technical Lemma

'Uber formal unentscheidbare Satze der Principia Mathematica

und verwandter Systeme I,' Monatschefte fur Mathematik und
Physik, 38: 173-98,1931. An English translation by van
Heijenoort is available in Jean van Heijenoort, From Frege to
Godel: A Source Book in Mathematical Logic 1879-1931
(Cambridge: Harvard University Press, 1967), pp. 596-616.
2 Godel, in fact, did use the symbol P to designate an (unramified)
logistic system similar to that of Principia Mathematica together
with specific axioms for Peano arithmetic. Godel's system P
corresponds (roughly) to the system PA given below.
3 Alfred North Whitehead and Bertrand Russell, Principia
Mathematica (Cambridge: Cambridge University Press), Vol.
I (191 0; second edn, 1925), Vol. II (1912; second edn, 1927),
Vol. III (1913;second edn, 1927).
4 Positing the existence of a predicate for every wff.
5 The formulation of the Peano axioms that follows is taken from
Chih-Han Sah,Abstract Algebra, (New York: Academic Press,
1967), pp. 16-17.
6 I.e., the universal quantifier is relativized by the J2_redicate serving
to identify N. The existence of such a predicate N requires the
explicit assumption of an axiom of infinity or an equivalent of it.
7 I.e., has the property that ifF 'corresponds' in PA to an m-p1ace
numercial function f, f-FA 3!yF(x 1 , •.. , Xm, y).
8 Technically, the expression F(k(ad, ... , k(am), k(b)) would
be written in PA as Za1 (xd :::> ••••• :::> Zam(xm) :::>. Zb(Xm+l)
:::> F(xl,. -·, Xm, Xm+l ), where Zn(x) is the wff of PA serving
to name the number n. The notation used in the text is
consistent with Godel's notation and with the earlier discussion
of numeralwise expressibility and representability in P.
9 For the definition of strong representability, the condition of
ordinary representability that, for any numbers a 1 , . • . , am,

197
NOTES TO PAGES 53-60

1-pA 3!xF(k(ad, ... , k(am), x) is replaced by the condition

1-pA 3!yF(x 1 , .•• , Xm, y). Strong representability implies
representability. The idea of strong representability is that the
new condition insures that F 'acts' like a function (in PA), and
condition (2), the same for representability and strong representa-
bility, insures that, when this 'function' is restricted to its
appropriate domain, it takes the values that it should.
10 E.g., one whose interpretation in ordinary number theory would
be an equation like 5 + 2 = 7.
11 The notion of the 'length' of a construction of a primitive
recursive function will be defined in the next section in terms of a
'formal construction sequence.' Each primitive recursive function
can be associated with a 'genealogy,' and the induction
mentioned above is on the length of that function's 'lineage.'
12 See p. 607 and the accompanying footnotes of the van Heijenoort
translation cited above. For the mathematical details of the proof
of numeral wise expressibility of primitive recursive relations in
PA, the reader is directed to the book by Joel W. Robbin,
Mathematical Logic: A First Course, (New York: Benjamin/
Cummings Publishing Company, 1969). I take this occasion to
express my deep debt to this book and to Herbert Enderton and
Louis Narens for helping me to undezstand the material presented
in this section. Responsibility for the text is entirely my own.
13 In principle, the proof of the numeralwise expressibility of
primitive recuzsive relations in PA shows how a wff G numeral-
wise expressing G can be constructed in PA. The discussion above
does not seek to minimize the philosophical import of this fact,
but to indicate that the actual construction of G in PAis not
essential to the practical objectivity of the proof itself.
14 For the actual construction and proof, see Elliot Mendelson,
Introduction to Mathematical Logic, (Princeton: D. Van
Nostrand, 1964), pp. 131-4.
15 For such a presentation, see Herbert B. Enderton,A Mathematical
Introduction to Logic, (New York: Academic Press, 1972).

Chapter 6 Primitive Recursive Functions and Relations

The case x < y is 'seeably/showably' adequate to the actual

condition x =I= y. At the end of this chapter, the checks that are
actually performed will be elaborated as 'reasoned computations.'
2 As is customary, the designation 'Proof', will be used as the
indication that what follows that designation is a conventional,
informal proof of the preceding proposition. I.e., 'Proof:'
indicates an exhibit of the material detail of a practically
accountable proof and not an exegesis or interpretation of the
work in which that proof occuzs and to which that proof makes
reference.
3 The statement of these propositions and their proofs are similar

198
 NOTES TO PAGES 60-8

to those found in Robbin, Mathematical Logic: A First Course.

4 I.e., the set of pairs of numbers (x, y) such that x divides y with
no remainder.
5 In practice, the primitive recursive equations for a given
numerical function or relation are thought of as providing a
definition consistent with the function or relation as it is
informally and practically known. Thoralf Skolem first indicated
how the functions and relations of elementary arithmetic could
be defined 'recursively.' ( 'Begriindung der elementaren
Arithmatik durch die rekurrierende Denkweise ohne Anwendung
scheinbarer Veriinderlichen mit unehdlichem Ausdehmungs-
bereich' ('The Foundations of Elementary Arithmetic Established
by means of the Recursive Mode of Thought, without the use
of Apparent Variables Ranging over Infinite Domains'), Videns-
kapsselskapets skrifter, L Matematisk-naturvidenskabelig klasse,
no. 6, 1923; translated by Stefan Bauer-Mengelberg and appearing
in Van Heijenoort, From Frege to Godel, pp. 302-33). Gi:idel,
in the same paper in which he presented his incompleteness
theorem ( 1931 ), gave the first precise definition of the notion
of (primitive) recursiveness.
6 The set of all primitive symbols, expressions and sequences of
expressions of P.
7 The comma usually placed between v1 and v2 in +(v 1 v2 ) and
·(v 1 v2 ) is omitted here in that it is used only to increase the
legibility of the expression. It is not a primitive symbol of P.
8 I thank Herbert Enderton for helping me work out this
formula.

Chapter 7 A Schedule of Proofs

On the whole, I have followed Robbin's presentation in

Mathematical Logic: A First Course, and I have used the notation
found there. The reader should note that Robbin proves Godel's
theorem for first-order recursive arithmetic and that changes
need to be made to make his proofs ade~uate for the system P.
2 E.g.,ifx=25=5 2 =2°•3°·5 2 =p?·p 2 ·p~,then(x)2 =Oand
(xh = 2.
3 E.g., if x = 25 = PY •pg • p~, then L(25) = 3. L defines the 'length'
of a natural number.
4 Roughly, occur(w, x, y, z) indicates that the individual variable
with Godel number x occurs in the wff with Godel number z.
5 I.e., Sub(x, n, a) equals g(S"J A I) or a, where x, t, and a are the
Go del num hers corresponding to x, t, and A, respectively. The
reader should note that Sub(x, t, a) does not guarantee that 't
is free for x in A.'
6 I.e., sub(x, n, a)= g(S~(n) A I) or a, respectively. Since k(n)
is a constant, k(n) 'is free for x in A,' and sub(x, n, a) represents
a proper substitution without needing the apparatus for specifying

199
NOTES TO PAGES 68-73

through Godel numbers the requirement that 't be free for x

inA.'
7 This artificiality is heightened by the fact that the numerical
relation R need not be the image under a Godel numbering of any
syntactic feature IR.of P at all. As an illustration of how this
may happen, suppose that the Godel numbers of the primitive
symbols of Pare all odd numbers greater than 1. Then 4 satisfies
the definition of R, but it is not the Godel number of a symbol,
expression, or sequence of expressions.
8 Com~aring n with g(xn) one obtains 0 < 2 3 •3 9 ·5 5 , 1 < 2 3 •
3 9 •5 ·7 5 ,2<2 3 ·3 9 ·5'1·7 7 •11 5 ,etc.
9 Mo.re elaborately, if x = g(xn) defines a primitive recursive
relation, then 3n :e;;; k(x = g(xn)) defines a primitive recursive
relation of x and k by Proposition 8, and 3n :e;;; x (x = g(xn))
defines a primitive recursive relation of x alone by Proposition 2.
However, two points need to be made. First, although this
elaboration accounts for the method indicated in the text, the
procedure followed in the text is informed by the familiar
techniques of working with primitive recursive functions and
relations and does not look to such an elaboration for the in-
course justification of the construction of 3n :e;;; x (x = g(xn))
as defining a primitive recursive relation. The second point is
that speaking of the formula 3n :e;;; x (x = g(xn)) as seeably/
showably defining a primitive recursive relation refers to the fact
that, given the placement of that formula in the course of work
of which the schedule of proofs is its natural account, that
formula is itself the materially adequate demonstration that the
relation defined by it is primitive recursive and, furthermore, it
refers to the fact that the adequacy of that formula is tied to the
ability to provide elaborations such as the one just given.
10 That this is so is easy to understand but notationally awkward to
prove. If a 1 , a 2 , .•• , O:n is an indexed sequence of not necessarily
distinct primitive symbols and a 1 , ••• , an are the Godel numbers
corresponding to each ak the expression a = a 1 a 2 ••• an has
Godel number 23 1 ••• Pnn· Consider the Godel number of any
proper subset of the O:j, i = 1, ... , n, concatenated in the order
of the increasing index. Each symbol present in a but missing
in the new expression represents the deletion of an exponent in
the prime factorization of g(a) and the downward shift of the
exponents of the larger_pri~ factors. For examnle , if a= (S
((O)), then g(a) = 21 CT.31 1 ·s1 Cl.fCl·llrv1 ·131 n·l7rn,
with 'corners' indicating the Godel number of the symbol that
they enclose. The Godel number of the expression (3 = S ((0, a
'part' of a, ·is 2rs1. 31Cl· 51(1· ~01. Thus, to obtain the
Godel number of {3 from the Godel number of a, the exponent of
2 is deleted in g(a), the other exponents shifted downward, and
the two greatest prime factors dropped completely:
21Cl·31SI.srn-7rCl·l!IOI·l3rn·l7rn ~

200
 NOTES TO PAGES 73-9
2m• 31SI· Sl(l• ']1(1·11101·131)1·171)1 ~
21 s1. 31 n· 51 (1.71 01·11 f");::J.13hQ ~ 21s1. 31 n· 51 (1- 7101.

Each deletion and each shift produces a number smaller than the
one before that operation.
The formalization of this reasoning is uninstructive and is
omitted here. In fact, even the elaboration that has been given is
never presented in a proof of Go del's theorem: the assertion that
x is a bound on v is understood as a sufficient basis for the
reconstruction of this reasoning and, therein, the uncovering of
one more aspect of the structure of the Go del numbering and its
compatibility with the techniques of working with primitive
recursive functions and relations.
11 The choice of tis not completely arbitrary in that it has been
constructed so as to include all the term building operations of P.
In that this construction is intentional, results concerning t
particularly are seen to be capable of generalization to the entire
class of terms of P.
12 In that x = g(O) and var(x) have been shown to define primitive
recursive relations independently of the construction of the
formula for term(x), the presence of them in the above formula
is unproblematic.
13 I.e., as part of the arranging of practices that makes up the search
for a bound on y.
14 Cf. note 10.
15 The circumstance that L(x) = 0 for x = 0 and x = 1 suggests
the need to test the formula for these cases.
16 By definition, G(x, u) is equivalent to ded(x, sub(g(x 2 ), u, li)).
If sub(x, n, a) defines a primitive recursive function, then sub
(g(x 2 ), n, a) defines a primitive recursive function by Proposition
2 (the substitution of a constant for a variable in a primitive
recursive function), and by Proposition 2 again (identification of
variables), sub(g(x 1 , u, u) defines one, also. Thus, if ded(x, y)
defines a primitive recursive relation, G is primitive recursive
by Proposition 7 (the substitution of primitive recursive function
for a variable in a primitive recursive relation).
17 This proof and several of those that follow have been taken from
Joel W. Robbin, Mathematical Logic: A First Course (New York:
Benjamin/Cummings Publishing Company, 1969).
18 Or transformation rules.
19 The placement of the definitions concerning the operation of
substitution - that is, whether they are part of the formation
rules as 'grammatical' entities or are part of the axiomatics - is
somewhat problematic for the abstract definition of a formal
system. The important point here is that in practice, for a
particular formal system, the terms 'formation rules' and
'axiomatics' are understood to refer to a definite and semi-ordered
collection of syntactic definitions. Definiteness of reference has

201
NOTESTOPAGES79~4

been supplied above by fiat in order to avoid a potentially

interesting but, also, a potentially distracting discussion of the
vernacular character of the recognizable definiteness of such
definitions (including that of the syntax of a formal system). I
thank Harold Garfinkel for bringing this matter to my attention
and for offering his reflections on it.
20 Or 'logical syntax'.
21 The first person singular pronoun is assuredly not being used to
indicate originality of conception, but only to emphasize (1)
the attribution of a referential definiteness to the notion of the
syntax of a formal system and (2) the fact that the definition
of syntax that is given refers to syntactic objects and not to their
definitions or, alternatively, that this defmition emphasizes the
strictly extensional character of the syntactic definitions that are
involved.
22 That is, the primitive symbols, the expressions, and the sequence
of expressions.
23 Theorem(x) is defined as holding if and only if x is the GOdel
number of a *theorem* of P.
24 The faulted character of this definition will be discussed later in
the text, and an adequate defmition will be given.
25 'Associativity' refers to the property that, for all x, y, and z,
(x * y) * z = x * (y * z). Because of this property, the 'product'
x * y * z is unambiguous. The importance of the associativity of
* will be indicated later in the text.
26 Empty parentheses are used in the propositions shown later in
the text to indicate that although those propositions are serially
ordered on the page of working notes or are indexed on that
page as an ordered sequence, they are not yet specifically
placed in the increasingly articulated, fmished schedule.
27 Read: 'x divides y if and only if there exists a natural number
n less than y such that y.= n • x.'
28 Read: 'xis~ prime number if and only if xis greater than 1 (i.e.,
x =I= 0 and x =I= 1) and for all numbers y less than x, if y divides
x then y is either 1 or x.'
29 Read: 'The (n + 1)-th prime is the least number x (less than or
equal to D) such that x is prime and x is greater than the n-th
prime.' In order to use the IJ" or least number operator, an upper
bound on x must be found. The 'empty' square is a
circumstantially-occasioned, situationally-transparent device
that a prover could use to signal the need for such a bound
without being distracted from the work of finding the main body
of the formula.
30- For instance, Po could have been defined earlier as 2 as is done in
Stephan Cole Kleene, Introduction to Metamathematics (New
York: D. Van Nostrand Inc., 1952).
31 The arbitrariness of s comes from the fact that p 0 = 1s = 1 for
all s.
32 The arbitrariness of r comes from the fact that p~ = 1 for all r.

202
 NOTES TO PAGES 85-8

33 The notion that this solution appears as a 'natural' one can be

elaborated by pointing to the way in which that solution is
embedded in the surrounding work of developing that formula;
the 'naturalness' of the solution is tied to the fact that x no
longer needs to be decomposed into its prime factorization,
that the formula applies when xis either 0 or 1, that x * y will
be primitive recursive if the remainder of the formula can be
constructed so as to exhibit its primitive recursiveness, that the
formula portends the use of the finite product construction
n
X. pg • n P~+i as indicated immediately below in the text,
i=O
and, finally, in that all this is so, that this solution, when it is
found, already fits within and further develops the immediate
course of work that seeks an adequate definition of * and that
provides the background for what an adequate definition of it
could be ..
34 Lg(P) was defined previously as the 'language' of P - i.e., as the
set of all primitive symbols, expressions, and sequences of expres-
sions of P.
35 I.e., by using the technique of constructing primitive recursive
functions that appears as Proposition 9 in the schedule of proofs
outlined earlier.
36 Or knows as a possibility.
37 Given an initial way of working similar to the one depicted above.
38 In the sense in which such a 'determination' is given later in the
text.
39 I.e., the inverse image under the mapping nf+ Pn·
40 For that matter, the prover may examine a familiar text or texts
to find, through that consultation, either initially or exactly how
to proceed. An elaboration of this issue will be given in points
(iv) and (v) of the summary of the topic (a).
41 The idea of speaking of the 'co-temporaneous' use of both Land
( ), in writing the formula for x * y is that the use of both of
these functions may arise while working on writing the 'final'
equation (i.e., the equation that is found to need no further
modification throughout the rest of the work of developing the
schedule), but that Land ( ), are neither simultaneously invoked
nor 'remembered' in a necessarily temporally-ordered fashion.
'Co-temporaneous' refers to the fact that the need and use of
these functions may arise in the presence of one another and
as the 'working-at' and 'working out' of the same immediate
problem.
42 Read: 'The "length" of x, L(x), is the number m of the least
prime Pro such that Pro divides x but Pk does not divide x for all
k > m.' The upper bounds on m and kneed to be large enough
to·insure the adequacy of the formula, but since x < Px for all x,
x itself will serve as these bounds.
43 That L(O) and L(l) 'should' have these values is a reference not

203
NOTES TO PAGES 88-9

only to the fact that 0 and 1 have no conventionally understood

prime factorizations, but also to the way in which the values of
L at 0 and 1 'fit' into the schedule of proofs as a developing
system of local relevancies. In the case of L, this 'fit' is illustrated
by the compatibility of L( 1) = 0 with writing 1 as pg , and it will
be illustrated further later in the text; the more general point
will be elaborated when I come to summarize some of the
material in topic (a) by speaking of the 'intrinsic' orderliness of
the schedule of proofs (point (vi)).
44 Read: 'The exponent of the i-th prime in the prime factorization
ofxis the least number k (less than x) such that pf divides x but
Pi +l does not.'
45 This is technically incorrect. In the cases mentioned, the J.I.-
operator is defined (by the convention indicated in the schedule
of proofs outlined on pages 65-9) to equal 0, and this is
exactly the value that will be explicitly stipulated for ( ), in the
text below. In the list that follows, l give possible 'reasons' for
the prover, as he is developing the schedule, to find the 'inade-
quacy' of the definition of ( ), as that 'inadequacy' is noted
above. In speaking of them as possible 'reasons,' I do not intend
them as an exhaustive list of distinct cases, but as an attempt to
provide for the indefinite number of ways a prover may come to
develop an always this-particular schedule of proofs. Of course,
it remains that a particular prover may not find this specific
matter problematic, and this, too, as being tied to his way of
developing, in its material specificity, a schedule of proofs. 'His
way,' however, would be capable of similar elaborations. The list
is as follows: (1) Although the prover is using the J.I.-operator
when he is writing the formula for ( ). , he may not have as-of-
yet explicitly worked out its definition. (2) In working with a
half 'verbal' translation of the J.I.-operator, the prover may find
through that translation -in that that 'translation' does not
make available the precise definition of the J.l.-operator - the
inadequacy of the definition of ( ), . (3) The fact that ( ). is well-
defined by the equation given in the text does not mean that the
definition given by the J.l.-operator is appropriate to the case at
hand, and the prover may find the need to explicitly work this
out. (4) The prover may explicitly define ( ). when x = 0 and
i = 0 as a means of making explicit just what is needed for the
local contingencies of the work of producing the schedule of
proofs. (5) In developing his schedule, the prover may be
consulting other texts giving such schedules of proofs, and he
may see in those texts (a) the lack of such a specification and,
although by working back through the definitions, he may find
( ). to be adequately defined, he may nevertheless feel that the
accountable work of the schedule's production has not been
made sufficiently available; (b) the lack of such a specification
and, by incorrectly finding ( ). as not being well-defined, thereby
finding as well his own contribution to the previous provers'

204
 NOTESTOPAGES89~2

work by correcting that definition; or (c) in that the writers

of those texts may have used other conventions in developing
their schedules of proofs, the prover may find himself with the
continual task of appropriately translating the other schedules
into his own current way of working, and the apparent need to
define ( ) for x = 0 and i = 0 may arise in and as the circum-
stances of such translational work.
46 In his seminars, in speaking about the work of a scientific
discovery, Harold Garfinkel suggested and elaborated on the
reinterpretation of varous themes from Gestalt psychology -
like that of the figure-ground relation and that of the gestalt-
switch - in terms of the situated and endogenously developing
practices of finding and exhibiting a scientific discovery. Rather
than understanding, for example, the 'ground' as the current state
of (physical, astronomical, chemical, etc.) knowledge and
experimentation concerning the domain of phenomena against
which the discovery is then referred, and the 'figure' as (what
amounts to a statement asserting the existence of) the discovered
object itself, Garfinkel suggested that the 'ground' be interpreted
as the situated, known, and recognized practices making up
adequate and efficacious laboratory procedure and that the
'figure' be interpreted as the cultivated and discovered organization
of those practices to find again the thereby naturally accountable
and propertied object. By this note I do not mean to attempt to
summarize Garfinkel's work on this matter, but to point out that
my reference to the 'discovered gestalt' of the material indicated
in the text is a reference to the discovered coherence of the
definitions that are given against the background of the surround-
ing practices of finding those definitions, and that that coherence
consists of the ways in which, when once found, the propositions,
their proofs, and the associated definitions seeablyfshowably
elaborate one another as making up an accountable course of
proving. The reader is referred to Harold Garfinkel, Michael
Lynch, and Eric Livingston, 'The Work of the Discovering
Sciences Construed with Materials from the Optically Discovered
Pulsar,' Philosophy of the Social Sciences 11 (1981 ), pp. 131-58.
47 This qualification is being used to signal (1) that this formulation
of the prover's use of the *-operator stands in relation to the
actual work practices that make up that use and, therefore, as
such a formulation, can be examined for its faulted character,
and (2) that this is so is not crucial for the discussion that follows.
48 In distinction from N, g(Lg(P)) being a proper subset of N.
49 Where 'examine' is being used as a gloss for the work of using and
interrogating the formula with the intention of determining
whether or not, or the conditions under which, * is associative.
50 Two alternatives to the course of reasoning that I have sketched
should be mentioned explicitly. In that the prover may only
come to see the need to assert the associative property of * when
he finds that he needs that property for proving later propositions

205
NOTES TO PAGE 92

of the schedule, that assertion and an appropriate qualification

on the domain of its validity may occur as retrospective
modifications on the previously established formula x * y = x •
L(y)
n P~tx}+i· On the other hand, the prover may be constructing
i=O
his schedule of proofs in close consultation with texts providing
such schedules, and the prover may adopt the device used above
from those texts in that he comes to see, through that consul-
tation, the utility of it.
51 Even though only the case y = 0 creates problems for the
associativity of *, the stated restriction on that associativity is
that x, y and z are all non-zero. That a prover will write the later
statement as the restriction and not the more specific one of
y =I= 0 (or the even more precise condition that y =I= 0 or that
y = 0 and either x = 0 or z = 0, also) points to the following
phenomenon: a prover will know, as part of the background
of work against which he is writing and organizing the account-
able schedule of proofs, the details of the schedule's reasoning
that he is not making explicit; he will recognize that certain
statements can be made more precise and that certain arguments
can be elaborated, but he will not attend to such precision or
provide such elaborations. It is not, however that the prover is
leaving out some of the details of the proof, for the associated
quantifier - all of the details of the proof- does not have a clear
referent. Instead, a mathematician reviewing a proof will point
to the omission of details that he considers as necessary in provid-
ing for the accountable work of which the written proof is a
residue and for which the written proof serves as a guide. The
prover, in turn, is responsive to the problem of setting out those
details of a proof that are identifying of that proof as a naturally
accountable proof of the theorem in question. In the case at
hand, in writing the restriction on x, y and z as the restriction
that they all be greater than 0, the prover is speaking faithfully
of the witnessed details of the work of proving Godel's theorem
while, at the same time, not holding himself to the endless
elaboration of them. In that a proof is written and read as a
pedagogical object -in that it is intended as and is held
accountable for making available or 'teaching' a way of proving -
a proof invites the reader into reworking the proof to find how
the written proof does, in fact, stand as the accountable work of
proving the theorem in question. Once again, in the case at hand,
the prover leaves it to the reader to see that the condition on the
associatj.vity of * is both adequate and too restrictive, and, in
seeing that it is too restrictive, that a concern over this precision
is of no consequence to the accountable production of the
schedule of proofs.
52 Parenthetically, in developing the presentation given here, I
initially wrote the equation for (a 1 , ••. , am> without exhibiting

206
 NOTES TO PAGE 92

m
the i-th term p 1im . Having written that equation and having
zi
then come to see what could be done - as a prospective course
of action within the course of the ongoing analysis - to provide
further a structure of construction for (a 1 , ••• , am), I then
m
introduced the term p1im which took on its relevance in develop-
Zi
ing the analysis of the next note but one by making notationally
available what the i-th term of the product would be. The point
that I wish to make is that this is just one of an endless number of
examples of the way in which the specific, material detail of a
written course of accountable mathematical proving is cultivated
over the course of that proofs production. (I am indebted to
Herbert Enderton for pointing out that, in that the functions
(a 1 , •.. , am> are defined for each m separately, there is no need
to indicate a dependence of any one of those functions on the
function n ~ Pn enumerating the prime numbers. Thus, for
example (al, a:2) = 2a, • 3a:Z, = (Z~ (a 1 , a:2 ))~ • (Z~ (a 1 , a 2 ))~ .
Moreover, given this analysis, the arrows from Pnto (x 1 , •.. ,
xn> on page 96 and from 13 to (x 1 , •.• , xn> on page 97 can be
eliminated. The analysis that is given in the text was based on the
fact that to write a general fo.rmula for (a 1 , ... , am>, a prover
'will' write (a 1 , .•• , am>= P~' • ... • p~in, and given this
circumstance, a prover 'might' - as I did - proceed to obtain
the constants Pi by composing the function Pn with that of
zr' thereby arriving at (what is now available as) the somewhat
'erroneous' construction of (a 1 , .•• , am> that appears in the text
and these accompanying notes. The major point of the text and
notes should, however, be remembered: for a prover engaged in
producing or reviewing a schedule of proofs, the primitive
recursiveness of (a 1 , ..• , am> does not depend on that prover
actually working out an explicit construction of (a 1 , . • . , am>
from the definition of a primitive recursive function, but on his
seeing a way of showing that it can be so constructed. Later, in
the chapter 'A Structure of Proving,' I point out that the relevant
detail of a proof is tied to the structure of that proofs local
work and that, in this way, what is available at the mathematical
work-site as the relevant detail of a proof can be different from
within different work circumstances. This does not mean that
'anything goes,' but that 'what goes' and 'what is correct' are
themselves, in each particular case, tied to, and discovered as,
local work practices. This technical aside does not affect the
argument in the text. I am grateful to Herbert Enderton for
his assistance; responsibility for the text, however, is mine alone.)
53 Whose unremarkable character (and in consequence, the some-
what exaggerated use of the word 'innovation') is connected not
only to that notation's intended/discoverable meaning, but, as

207
NOTES TO PAGES 92-3

well, to the presence of the similar use of the superscript in the

notation for the projection functions Ii.
54 The finite product (a 1 , ... , am) >-+ a 1 • ... • am is primitive
recursive in that it itself is seeably/showably constructed by
the substitution of a previously constructed primitive recursive
function (a 1 , ... , am -1)-+ a 1 • ... • am -1 for a variable in
the 2-place primitive recursive function of multiplication and
such a construction requires only a finite number (in this case,
m - 1) of steps. Exhibiting that substitution through the formula
m-1· )
(a 1 , ... , am) ~( .n ai • am
1=0
simultaneously indicates how an inductive proof could be given.
55 In even greater detail, by using the notation ex(a, b) for ab,
(a 1 , , .. , am) can be analyzed as being constructed as
(a 1, ... , am) ~ (ex(pz~(a, , ... ,am)• 1r (a1, ... , am)), ... ,
ex(pz~(a, , ... ,am)• I~ (a,, ... , am)))

~ { mn 1 Ii(ex(p 1 , Ilfl(al, ... ,am)), .. ·, ex(pm,

~~(a1, ... , am)))} • I~(ex(pl, Ilfl(al, ... , am)), ... ,
ex(pm, I~(a 1 , ... , am)))= p~' • ... • p~m.
56 Given the way of working from within which those propositions
and their proofs have been constructed. For further amplification
of this point, see the summarizing material (i)-( vi) on pages 94-100.
57 The propositions and proofs that follow in the text are presented
as they might appear in a finished schedule of proofs. In
constrast, the prover who has developed the schedule to this
point will (probably) only have those propositions and proofs
enunciated and arranged as, for example:
xly is p.r.
Pf: xly ~ x';l(o A 3n.;;;;; x (y = n • x)
prime(x) is p.r.
Pf: prime(x) ~ (x > 1) A 'v'y,;;;;; x { ylx :::> (y = 1 V y = x)}
- nf+ Pn is p.r.

- <~L · · · J~> =Pi' • · · · • P~m

Pf: by Eiefffiffien construction
In fact, the physical ordering of the propositions on the page
might be different; (possibly) arrows or sequential numberings
superimposed on one another (as provers' situationally-occasioned

208
 NOTES TO PAGES 93-4

devices) might be used, or propositions and proofs might be

sandwiched between other propositions and proofs on the
working page, so as to mark the accountable reorganization of
those propositions and proofs and, therein, the accountable
orderliness of the prover's work as well. The surrounding way of
speaking about those propositions and proofs - that is, whether
or not, or to what degree, the distinction between the syntax
language, the meta-language, and the language of ordinary
number theory should be applied (e.g., whether to use:::> or
'implies' in the proofs of propositions given in ordinary number
theory), whether or not such distinctions should be specifically
mentioned and used in the text, whether the 'proofs' of the
propositions should be separated from the propositions and
annotated as 'proofs' or, instead, be given as definitions with
accompanying 'verbal' interpretations (as GOdel did in his original
paper), whether functions such as n~-+ Pn should be written in
this fashion or abbreviated as, in this case, Pn, whether
motivational remarks and/or suggestive names should accompany
the introduction of the 'apparatus' ( ). , L, *,and (),or, instead,
the reader should be left to come into that 'motivation' in and as
the further work of proving the propositions of the schedule,
etc., etc. -may not have been 'decided' or 'finalized' as of
yet, and the prover might postpone, or may not have yet come
into the relevance of, working out such details for the 'final'
writing up of the proof. ('Decided' is placed within quotation
marks in that, once again, the term is being used to refer to the
practices that make up the writing of the schedule as those
practices are tied to the material detail of the schedule itself,
and not, as might be imagined, to refer to a psychological process
that can be disengaged from those practices and that detail.)
58 Although the initial presentation of the 'encoding' functions ()
may have suggested the centrality of those functions for the
construction of the schedule of proofs, their utility is hindered
by the fact that a separate function is needed for each number
of arguments. In that the Godel numbering already 'encodes'
sequences of numbers in single numbers, much of the work
of producing the schedule can be formulated as the work of
'decoding' a given number so as to exhibit the 'structure' of its
construction. For the formal theory P, the prover does not need
to use any of the functions () as a prover would discover over the
course of his working through the entire schedule. In the 'final'
version of the schedule, the proposition concerning ( ) would be
omitted. The reader will find an example of the use of ()in a
schedule of proofs in Joel W. Robbin, Mathematical Logic:
A First Course (New York: Benjamin/Cummings Publishing
Company, 1969), pp. 103-4.
59 Or to have retrospectively introduced on the occasion of their
apparent need, or to have begun the construction of the schedule
by first laying out the following functions as being prospectively

209
NOTES TO PAGES 94-100

useful for the construction of other primitive recursive functions,

etc., etc., each of the alternatives providing a formulation of the
work-site practices that it simultaneously glosses.
60 Respectively, the characteristic function of the set { 0} , the
characteristic function of the set N - { 0} . and the absolute
value of the difference of two numbers. sg(x) is defined by
primitive recursion - sg(O) = 0, sg(S(x)) = 1 -and sg(x) is
then defined as 1 -'- sg(x). The reader will recall that x -'- y,
equaling x- y if x ~ y and 0 otherwise, was introduced earlier
as a primitive recursive function.
61 Where 'similarly' refers to the ways in which the introduction
of that function is already embedded in the surrounding work
of proving that provides for that function's (prospectively)
efficacious use.
62 Where the appropriateness of using 'then' is now to be under-
stood as having its origins within the achieved availability of the
accountable work of the schedule's production.
63 A sequence of proofs similar to the one that follows can be found
in Mendelson, Introduction to Mathematical Logic. The reader
should note that Mendelson defines the characteristic function on
a subset A of Nm as being 0 on A and 1 on Nm - A, and that I
have followed the more standard mathematical practice (if not
the standard practice of mathematical logicians) in defining the
characteristic function as 1 on A and as 0 on Nm - A.
64 To speak of the produced schedule as a mathematical discovery
may seem to the reader a usage more properly reserved for more
significant 'results' as, for example, the proof of Godel's theorem
as a whole. The point, however, is that this use of 'mathematical
discovery' does not concern the importance of a result, but
speaks instead of the lived-work of proving that inhabits and is
identifying of mathematical practice.
65 If, however, the graph included the first ten propositions of the
schedule, the number of lines would obscure the immediate
availability of the thing that such a representation proposes.
Similarly, for the prover constructing such a graph, a central
theme and a recurring problem of that construction is the
arrangement of the graph so as to exhibit the organizational
features of the schedule that that graph is intended to represent.
66 I am indebted to Harold Garfinkel for the notion that such a
graph 'renders' (the schedule of proofs/the lived-work of the
schedule's production that is identifying of the schedule as such)
as a 'structure of logical dependencies.'
67 Again, I am indebted to Harold Garfinkel for making the notions
of a 'received view,' 'received history,' and 'received topic'
available as a matter of ethnomethodological interest.
68 The word 'might' is used here to indicate that the inquiry that is
embodied in the work of composing such a derivation is itself
part of, and arises from within, a developing, material-specific
way of working through a schedule of proofs; the usage is not

210
 NOTES TO PAGES 100-2

intended to imply that a prover 'chooses' between 'subjectively'

available, alternative courses of reasoning and action.
69 Although the use of the word 'substitution' can be justified in
terms of the relation defined by the composition of the character-
istic function of the equality relation, K_ Oi , I~) with (I~, I~ •
I~), the definiteness of the reference to 'the relation that results
from "substitution"' (as with the definiteness of reference
provided by the use of 'identification' later in the text) does not
reside in its exegesis as an accountable description, but is first
provided for as a definiteness of action in writing the derivation
itself- specifically, the equation y = n • xis written below (and
'thereby' follows from) the equation y = z with n • x (as a purely
symbolic or 'formal' entity) being seen as replacing, or as being
substituted for, z. That it is through such definiteness of action
that the statements of the initial propositions (in this case,
Proposition 7) take on their practical utility serves as a reminder
that the reasoning given in this paragraph stands only in relation
to and renders the combined course of writing and reasoning
that makes up and is the derivation itself.
70 In greater detail, Proposition 2 permits the identification of
variables y and w in the characteristic function of the relation
defined by 3n ~ w (y = n • x). Alternatively, in that y (viewed
as a projection function) is itself primitive recursive, the primitive
recursiveness of 3n ~ y (y = n • x) follows from that of 3n ~ w
(y = n • x) by Proposition 7, the substitution of a primitive
recursive function for a variable in a primitive recursive relation.
The relationship between Proposition 2 and Proposition 7 will be
developed later in the text.
71 It might be argued that a prover is aware, over the course of
writing/verbalizing the displayed formula, that that formula and
verbalization are practical expedients whose approximate
character can be made available, for example, by (interpretatively)
amplifying the given defmition so as to read: if x andy are
natural numbers (repaired in writingfverbalizing-to-<>neself to
read 'for all x andy natural numbers'), then xis said to divide y
if and only if x =/:= 0 and there exists a natural number n such that
y equals n timex:
'Vx, yEN, xly ~ x =/:= 0 and 3n (y = n • x).
In this way, the earlier definition comes to be seen as modeling,
in known and analyzable ways, the 'real' properties and objects
that it represents and to which it refers. A first point to be made
about such an argument is that a prover neither uses nor relies
upon such exactitude of expression in order to write the formula
xly ~ x =/:= 0 and 3n (y = n • x).
as, at that time, the practically adequate defmition of the
divisibility relation. Secondly, if incorporated into a schedule of
proofs, such exactitude would only obscure the developing

211
NOTES TO PAGES 102-3

orderliness of a prover's work. But the larger point is this:

whether or not
xly ~ x 4= 0 and 3n (y = n • x)
is recognized by the prover, as he is developing the schedule, as
being potentially remediably-vague, the precision that may come
to be required of the written formula, as such a requirement
may become motivated by the detailed circumstances of the
prover's immediate work, is itself required to be available in
and as the accountably-historicized, exhibitably-relevant material
detail of that work itself. In this way, the accountable
corrigibility or incorrigibility of the definition is essentially tied
to that definition's placement within an accountable course of
proving's work. A prover, on later coming to see the need to
assert that the definition refers to natural numbers x andy, will
extract from his course of work - as an historicized, material-
specific object- the reasoning and detail needed to make that
fact apparent (and, in this particular case, utterly 'trivial').
72 The use of 'provisionally' needs some amplification. First, the
term is somewhat misleading: although it appears to be
descriptive, it glosses the ways in which such a formula is, for a
prover, part of, and remediably available as part of, an ongoing
course of work. On the other hand, in that this term signals the
utterly practical character of a prover's work in constructing
a schedule of proofs, it reminds us, once again, that the
accountability and analyzability of the schedule is a practical
achievement and, thus, recommends, once again, the work of
that achievement as the phenomenon of interest.
73 The_quotation marks around 'defined' are used here to indicate
that at this point in developing a schedule of proofs, a prover
'might' not (in the sense of note 68) thereupon explicitly
formulate a definition of the relation in question.
74 Both in terms of the particular relations involved and in terms of
the notational specifics in which those formulas are written.
75 Specifically,.a prover 'may' distinguish the cases when y = 0
and when y > 0. If y = 0, then y = 0 • x for all x and xly for
all x with n = 0. Thus, if y = 0, xly ~ 3n ~ y (y = n • x).
On the other hand, if y > 0 andy= n • x (i.e., xly), then
neither n nor x can be equal to 0. An inductive proof shows that
for all x, y > 0, y ~ y • x. Hence, y is again an upper bou.nd on n.
The reader should note that it is unlikely that an actual prover
would articulate his reasoning to this extent. Instead, a prover
would probably see - perhaps through a review of a few
particular cases examined as exemplars - that the distinction
between the cases when y = 0 and when y > 0 seeably/showably
provides for a line of argumentation similar to the one just given,
and he would anticipate both the theoretical simplicity and
picayune detail of materially elaborating that argument and,
thereby, of further articulating and exhibiting its reasoning.

212
 NOTES TO PAGES 104-6

76 To speak of the following way of working as a 'method' is not to

thereby claim that it consists of a definite set of procedures and
abbreviatory practices. Instead, the methodic character of this
way of working is itself increasingly developed and articulated
over the course of the prover's work, thereby taking on expressive
capabilities and analytic properties not provided by its summar-
izing description. In fact, the adequacy of the description of that
method is available to a prover only after having come to see how
its methodic character is tied to and developed as both a way of
working and, specifically and crucially, as a way of working that
has, as its accomplishment, the adequate demonstration of the
adequacy of the proofs in question.
77 The reader should note that although the availability of the
technique of bounded quantification was used in writing the
displayed equations and, for the prover, is part of that display as
a derivation, a prover may not have seen as of yet the pointedness
of incorporating that use in the material display itself.
78 One way of determining whether or not Proposition 8 could be
used directly.- and, thus, an alternative procedure to that
developed later in the text - would be to review the various
constructions that are used in proving Proposition 8 to see if
the particular aspect of its notation that is in question reflects
a substantive matter or is, instead, merely an inadvertent result of
the particular formalism in which that proposition was written.
By working backward through a proof of Proposition 8, a prover
might find (as his practical achievement) that, at some stage in
the construction of that proof, a technique like the one
formulated in Proposition 2 was 'already' used; therein, as a
seeable/showable consequence of that finding, the prover will
have found the accountable irrelevance of a new variable for the
upper bound on n.
79 At this point in constructing the derivation, a prover might
realize only that what is involved is an 'identification' of variables
and that Proposition 2 permits such identifications. If this were
so, it would remain for the prover's retrospective review of his
own work to clarify the fact that the substitution in question
is actually being made in KR(w, y, x), the characteristic function
of the relation R(w, y, x), and not in R(w, y, x) itself, thereby
formulating the immediate problem as a proper instance of
Proposition 2.
80 If a prover were to have adopted a different method of working
out such a derivation - for example, by associating a relation
symbol with the relation that an equation defines:
T(y, n, x) ~ y = n • x ()
S(y, w, x) ~ 3n :E;; w (T(y, n, x)) ~ 3n :E;; w (y = n • x) (8)
R(y,x)~S(y,y,x)~3n:E;;y(y=n· x) (2)
-then this new method would have raised problems for the

213
NOTES TO PAGES 106-7

derivation that were not exhibited by the method described in

the text. An example of such a problem is the fact that the
relation xly 'transposes' the variables in R(y, x), a course of
writing again provided for by Proposition 2 of the schedule
outlined earlier. In practice, of course, a prover will not hold
himself to any one 'method,' but will use those devices that
become relevant to the developing material detail of the
particular problem-at-hand as, in fact, was done in the text
above. Over the course of constructing such a derivation, a prover
will, as part of that work itself, do that work in such a way that it
could be articulated as an accountable definite procedure;
fu.rther, he will look to that work's prospective achievement
in finding and exhibiting such a derivation in order to retro-
spectively insure - as a rehearsed course of writing and reasoning
- the accountability of that work itself. At the same time it
should be noted that a prover will find no immediate, material
motivation for his actually articulating as a definite procedure the
situated way of working that he is both using and developing.
81 'May not give' is used here in preference to 'may fail to give,' the
point being that the latter expression preserves the transcendental
necessity of such a review whereas the former is consonant with
our examination of the relevant and problematic details of such
formulas as those details are tied to and embedded in the course
of writing/reasoning of which they themselves are a part.
82 The condition 'x =I= 0', not considered in the discussion
immediately following in the text, will be briefly examined at the
end of topic (b). However, for completeness, let us note the
following: first, if both 'x =I= 0' and '3n ~ y (y = n • x)' define
primitive recursive relations, then so does 'x =I= 0 and 3n ~ y
(y = n • x)' in that definitions formed through the use of
sentential connectives (i.e., 'not,' 'and,' 'or,' 'implies,' and 'if and
only if') preserve primitive recursiveness (Proposition 6). In
writing out and then reviewing the defmition/proof
xly ~ x =I= 0 and 3n ~ y (y = n • x),
a prover constructing a schedule of proofs will justify, and
thereby come to see the need for explicitly articulating as part
of that schedule, the technique of constructing primitive recursive
relations by conjunction ('and') and negation ('not')- from
which it seeably/showably follows, as a prospective course of
proving, that relations constructed through the use of any of the
sentential connectives preserve primitive recursiveness in that all
the connectives can be defined in terms of 'and' and 'not' alone
- and to establish the primitive recursiveness of the equality
relation. (x =I= 0 can be seen to be defined as not-(x = 0).) In
practice, a contemporary prover would most likely not come to
retrospectively construct the initial propositions of the schedule
in quite this fashion, but instead would begin constructing a
schedule by preparing a preliminary list of primitive recursive

214
 NOTES TO PAGE 107

functions, relations and techniques of constructing such functions

and relations and/or by bringing together some of the lists of such
functions, relations, and techniques that are available in the
literature, selecting from such lists those functions, relations, and
techniques found to be needed over the course of constructing
the remainder of the schedule and those of practical necessity for
that list's orderly development as a coherent course of work.
As a second point, let us note that the justification that x =I= 0
defines a primitive recursive relation is the same as that for the
relation defined by y = n • x; once the justification for either of
them is developed as an extractable course of reasoning, the proof
of the other follows as well. Specifically, if y = z defines a
primitive recursive relation and n • x is a primitive recursive
function, then y = n • x defines a primitive recursive relation in
that it results from the substitution of a primitive recursive
function for a variable in a primitive recursive relation
(Proposition 7). But then, since x =I= y is a primitive recursive
relation and Z(w) = 0 is a primitive recursive function (by
definition), the relation defined by 'substituting' Z(w) for yin
x =I= y - that is, x =I= 0 - is primitive recursive as well.
Finally, the reader should note that ( 1) the efficacious choice
of variables, and the seeable/showable irrelevance of such a
choice, is itself a part of a prover's work in producing a
notationally consistent derivation of the formula
xly ~ x =I= 0 and 3n ~ x (y = n • x)
as, for example, the work of such an arranging is available in the
'choosing' of x =I= y and Z(w) = 0 above so as to 'deduce' the
notationally-specific formula x =I= 0, and (2) the fact that a
prover can read symbols like 0 as the function Z(w) = 0 in order
to do what he needs to do with them is tied to the fact that such
formulas are already embedded in an ongoing course of work
and, further, to a prover's ability, on the occasion of such a need,
to historicize that work so as to find, therein, the materially
justifiable motives for just such an interpretation and manipula-
tion of symbols.
83 One final digression may help to set in relief the type of
constraints that are placed on such work.
Roughly speaking, a function is said to be effectively
computable if there is a finite, 'mechanical' procedure for
calculating the value of that function for any of its possible
arguments. A relation is said to be effectively computable if its
characteristic function is. Now, in that the initial functions
Z(x) = 0
S(x) = x + 1
Ifl (x 1 , ..• , xi, ... , xm) = xi
can be seen to be computable, and in that the definition of a

215
NOTES TO PAGE 107

function by substitution or primitive recursion can be seen to

give an effective procedure for calculating the values of that
function from the values of the functions used in its definition,
the class of functions generated from the initial functions by
substitution and primitive recursion -i.e., the primitive recursive
functions- would seem therefore to all be computable as well.
And, in fact, the definition of the class of primitive recursive
functions can be thought of as an (early or preliminary) attempt
to formalize the notion of effective computability itself.
Let us consider"the relation R consisting of all3-tuples (y, n,
x) such that y = n • x. In that, given any 3-tuple of natural
numbers, one need only calculate n • x and compare that number
toy to see if that 3-tuple is a member of R, it is certainly to be
hoped, however the notion of computability is formalized, that
R will be able to be shown to be such a relation. In a similar
fashion, the divisibility relation can be seen to be 'computable':
it is defined by the formula
xly <> x :f= 0 and 3n (y = n • x)
and, as I indicated earlier, y is itself an upper bound on the
number n. Thus 316 since 3 :f= 0 and
6=1=0·3=0
6=1=1·3=3
6 = 2·3
but 3 f 5 in that, if 3 I 5, there would be some number n less than
5 such that 5 = n • 3 and
5=1=0·3=0
5=1=1·3=3
5=1=2·3=6
5=1=3·3=9
5=1=4·3=12
5=1=5·3=15.
These considerations can be seen to provide heuristic evidence
that the divisibility relation and the relation defined by y = n • x
are primitive recursive, and, in consequence, they occasion the
question of the praxeological origins of their character as being
evidential and accountably heuristic, rather than their being
corrig~bly demonstrative. To answer this question, consider the
follo~ing: first, various attempts have been made to formalize
the notion of (what is referred to as) effective computability;
the notion itself has a 'received' history, and from within that
history, that notion (and the term 'effective computability')
is preserved and maintained as an informal one used in contrast
with its various formalizations. Secondly, as part of the received

216
 NOTES TO PAGE 107

history of 'effective computability,' various attempts to formalize

it - including the use of Turing computability, Markov
algorithms, A.-definability, and general (as opposed to the more
restrictive 'primitive') recursion -can all be shown to define the
same class of functions; this class of functions properly includes
the primitive recursive ones, and, therefore, that a function is
'computable' does not necessarily imply, even heuristically,
that it is primitive recursive. (Parenthetically, Church's thesis
is the proposal that this larger class of functions is the appropriate
formulation and formalization of 'effective computability.')
Most importantly, however, the circumstance that occasions
the preceding remarks and that leads us back into the text is this:
the heuristic character of considering the computability of the
relation defined by y = n • x consists of the recognition that
there is no immediately accessible way, given the local work
practices from within which the problem arises (as described
in the text) of turning the reasoning concerning the computability
of y = n • x into an accountable course of proving.
84 Previously, KR was referred to 'as' the characteristic function of
a relation R, whereas, in the present situation, I have spoken of
K_ as 'standing for' the characteristic function of the equality
relation. This change in usage points to a phenomenon in its
own right. In order to work through the present problem, a
prover comes to rely on properties of the function that K_
names. In doing so, the distinction between the symbol K-=:_ and
the function for which it stands becomes a feature of the -
problem's solution. The first point that needs to be made is that
the omnipresent 'naming' found in mathematical argumentation
- as, for example, in using K_ to 'name' the characteristic
function of the equality relat1on - is available to the mathema-
tician (1) as a practice and (2) in that it is a practice, as a practical
resource for the work of proving. Let us make two further
observations: First, the formulas K_ (x, y) and x = y, as they
have come to be written here, are not related as one being
'derived' from the other, but, as the reader will see, they
constitute a pair of mutually elaborating expressions. Second, in
*
showing that the equation x 0 defines a primitive recursive
relation (as will be done later), a prover re-reads 0 as the constant
function Z(y) = 0. The point of these observations is that they
offer the diversity and richness of mathematicians' notational
practice as a real-worldly researchable phenomenon and, therein,
provide a contrast between that practice and the rendering of it
as, for example, in the specification of a logistic system and, as
well, in conventional theories of signs and reference.
85 Even though an equation for K=(x, y) is available,
K=(x,y)= l...!...((x...!...y)+ (y...!...x)),
a prover does not use such an equation, nor is such an equation
enlightening, in showing that y = n • x defines a primitive

217
NOTES TO PAGES 107-13
recursive relation. Instead, as the reader will see below, a prover
will 'simply' use the association of x = y with K_ (x, y) or, more
exactly, the association of y = n • x with K_ (y ,n • x) as it is
developed from the association of y = z with K_ ( y, z).
86 'Substitute' is placed in quotation marks as a way of signaling
the fact that the term refers, at this point, to a manipulation of
symbols. Shortly, however, and then retrospectively as well, that
term will take on its character as referring to the construction of
a function by the process of 'substitution.'
87 And, therein, a prover's recognition of the practically accountable
irrelevance of such a choice; therein, once again, a prover's
recourse to the material detail of his work as being embedded in
the developing course of the work itself.
88 In the display on page 108, y = z is written- or, at least,
recognized as the appropriate equation - before the prover writes
K_ (y, z); it is from the recognition that the substitution of n • x
for z in y = z yields y = n • x that an appropriate choice of
variables forK_ is obtained.
89 It might be argued that such a check and, similarly, that some of
the 'steps' used in developing the formula K_(y, f(n, x)) are
unnecessary. The point here is not that such-work is or is not
necessary, but that, first, a prover does such work; second, it is
through such work that a prover comes to establish K_ (y, f(n,
x)) as the practically adequate formula for the problem-at-hand;
and third, it is against the background of the achieved adequacy
of K_ (y, f(n, x)) in solving that problem that the essential or
incidental character of the work of its production is then assessed.
90 I.e., K_ (x, y) = 1 -'- ((x-'- y) + (y-'- x)).
91 Parenthetically, however, a prover might review and/or articulate
as a separate proposition some of the techniques of manipulating
variables in primitive recursive functions as, for example, the
ability to hold all but one of the variables constant when defining
a function by substitution.
92 In writing
KT(Y, n, x) = KE(Y, f(n, x))
after and beneath
T(y, n, x) ~ E(y, f(n, x)),
a prover will recognize that the former does not necessarily
'follow' from the latter. I will return to this matter later.
93 As the function defined by KT(Y, n, x) = KE(Y, f(n, x))
and only incidentally as the characteristic function of the relation
defined by y = n • x.
94 I.e., '(y, n, x) is a member ofT if and only if (y, f(n, x)) is a
member of E.'
9 5 For completeness, the following formulation of this proposition
and its proof are themselves locally worked out both in relation
to each other and as a recurring and circumstantially renewed

218
 NOTES TO PAGE 113

topic over the course of developing a schedule ofproofs as a

completed whole. Thus, the proposition and proof that follow in
the text have their own histories in coming to be the practically
adequate statement of that proposition and that statement's
practically adequate proof. It should also be noted that the
generalizability of both the problem and its solution is already
provided for, not because a prllctically accountable solution has
been achieved, but in that the way in which that solution was
obtained already incorporated,.over the course of its development,
the character of its accountable detail as instances of more
general and previously recognized constructions.
96 Here, the prover leaves it to the reviewer to find how to establish
the fact that the one line follows from the other. For convenience
- and as an accountable procedure that promises the generality
of its prospective accomplishment - the reviewer might let R be
a 2-place relation and fa 1-place function. Then, as in the text
above, a method of showing that S(x, y) ~ R(x, f(y)) 'implies'
Ks(x, y) = KR(x, f(y)) is tied to seeing that S(x, y) can be
rewritten in set-theoretic notation as (x, y) E S. As in the text,
S(x, y) ~ R(x, f(y))
can be written as
(x, y) E S ~ (x, f(y)) E R
and, in that the characteristic function of a subset U of a set Vis
1 on U and 0 on V- U, Ks(x, y) can be seen to equall if and
only if KR (x, f(y)) equals 1 and Ks(x, y) can be seen to equal 0
if and only if KR (x, f(y)) equals 0.
Two things should be noted: first, that the prover does not
elaborate this argument is not because he does not see that such
an argument needs to be given, but that in order to make such
an argument precise, a larger analytic apparatus would have to be
developed, and such work would distract from the larger
structure of the accountable work of producing the schedule of
proofs, and that, at this point in the prover's work, it is available
to the prover that a reviewer should be able to give such an
exegesis and to see, in doing so, that that exegesis is incidental to,
and does not provide critical detail for, the development of the
schedule itself. Second, the reader should also note that while
words like 'then,' 'it follows,' 'thus,' and 'implies' are rendered
in classical studies of mathematical practice as the notion of
'material implication' -i.e., approximately, from the purely
formal expressions A :::> B and A, the prover can then write B -
or, possibly, that of 'logical consequence' -i.e., approximately,
for every possible assignment of truth values true and false to
some set of propositions S, if whenever all if> E Shave the truth
value true, a proposition () has the truth value true, then () is a
'logical consequence' of S -in actual mathematical practice
these words are used both as indicators and as summaries of

219
NOTES TO PAGES 113-15

situationally-specific, endogenously organized, naturally account-

able courses of argumentation. Words like 'then' and 'it follows'
serve to articulate the argument being given- for example, by
pointing to work that actually needs to be done while, at the
same time, maintaining the proof-specific level of appropriate
detail, or by calling the readers' attention to what has just been
argued in such a way that he thereby comes to see that he needs
to reorganize that argument so as to provide for and find the
self-evident character of the thing being offered as that
argument's consequence - and, in this way, these words take
on an endless diversity of what could be called argument-
specific meanings.
97 Proposition 6 of the schedule outlined earlier reads 'the logical
operations of "not," "and," "or," "implies," and ''if and only
if," applied to primitive recursive relations, produce primitive
recursive relations.' In working through the schedule, this
proposition is 'understood' and applied as formulating the fact
that if P(x 1 , ... , Xm) and Q(x 1 , ... , Xm) are conditions that
define primitive recursive relations P and Q, then the expression
P(x 1 , ... ; Xm) 'TI Q(x 1 , ... , Xm) defines a primitive recursive
relation P 'TI Q as well, where 'TI is a sentenial connective and P 'TI Q
is the relation defined by the appropriate set-theoretic
construction on the sets P and Q corresponding to the connective
'T/. The point here is that the cogency and 'meaning' of the
statement of this proposition is integrally tied to, and is available
to and recognized by the prover as being tied to, the familiar
practices of constructing primitive recursive relations that that
statement is seen to formulate in an adequate and efficacious
manner.
98 I.e., as a retrospectively reconstructed proper order of work.
99 In the fashion of what we have come to speak of as a 'derivation,'
this argu.111ent can be elaborated as
x=y (5)
X -:f= y (6)
Z(w) = 0 (by definition)
X -:f= 0 (7)
100 The reasoning of which can be elaborated, as I did earlier, as
follows: in that the equality relation, y = z, and multiplication,
n • x, are, respectively, a primitive recursive relation (by
Proposition 5) and a primitive recursive function (by Proposition
3 ), the equation y = n • x defines a primitive recursive relation
in that it results from the 'substitution' of a primitive recursive
function for a variable in a primitive recursive relation
(Proposition 7); from the primitive recursiveness of y = n • x,
the technique of bounded quantification (Proposition 8) then
insures that 3n ~ w (y = n • x) defines a primitive recursive
relation of X, y and w; from which it follows, by 'identifying'
220
 NOTES TO PAGES 115-21

the variables y and w (Proposition 2), that 3n ~ y (y = n • x)

defines a primitive recursive relation of x and y alone.
101 At least in principle, and in that they are retrospectively available
as having been, and in that they are repaired over the course of
developing the schedule as having properly been.
102 From within the way of working that is described in topic (d)
and as an accountable aspect of both the schedule and the syntax
that that way of working makes apparent.
103 In the formation rules of a logistic system, the class of terms are
typically defmed by the following rules: 1) the constant symbols
(e.g., 0) are terms, 2) the individual variables are terms, 3) iff is
ann-place function symbol and t 1 , •.. , tn are terms, then f(t 1 ,
... , tn) is a term, and 4) a formula is a term if and only if it can
be shown to be one on the basis of 1), 2) and 3). This deftnition
is to be contrasted to the reformulation of it in terms of a forma-
tion sequence of terms as that reformulation is given later in
the text.
104 I.e., the (unique) number of terms which, when that number of
terms is concatenated with that function symbol and appropriate
parentheses, makes up a term itself.
105 Or, depending on the specification of the logisitic system, as a
class of expressions.
106 The reader will recall that the individual variables in our system
Pare x, x', x", etc. and that the Godel number that has been
assigned to the symbol ' is 7.
107 It should be emphasized that the point being made here does not
refer to the particular arrangement of these propositions in
isolation from the rest of a schedule of proofs, but, instead,
concerns the structure of a schedule of proofs as a totality
whose natural accountability as a course of proving just that
schedule is maintained over that self-same course of proving.
The point is that enough of the sequentialized, hierarchical
structure of the arithmetized syntax must be present in a
schedule of proofs so as to provide access to and to exhibit, in
and as that schedule's own material detail, that schedule/the-
accountable-work-of-its-production as just such a naturally
accountable schedule of proofs. Thus, for the proof of the
primitive recursiveness of var(x), any of the three alternative
proofs that are given in the text could, in fact, be included in
a schedule of proofs, the selection of any one of them not being
critical to the exhibited structure of that schedule. This holds
similarly for the example of term(x) given below, and in the case
of sub(x, n, a), the misplacement of that function's deftnition/
proof in the diagonalization argument would, most likely, appear
as an awkwardness to a prover reviewing that proof and, in once
again working through the other propositions and proofs of the
schedule to find the interpretable 'meaning' of sub(x, n, a) as that
'meaning' is required for the diagonalization/'proof, that prover
would, in that way, find as well the proper place for sub(x, n, a)
in that schedule.

221
NOTES TO PAGES 121-2

108 One feature of this proof should be noted. The reader will recall
from the discussion of the formula
L(y)
X *y= X • fl pl{l~)+i
i=O
in topic (a) that the term (i.e., the component) p~l~)+O (when
L(y)
i = 0) was included in the product ll p~6b+i specifically to
i=O
make that formula applicable to the case when y = 0. Having
discovered this solution to the problem of defining x * y, a
prover, on confronting the need to make the equation for g (xn)
applicable to the case n = 0 as well as that of n > 0, can now use
that previous solution as, and now has it available as, a device
of proving. On coming to and in writing the proof given here, a
prover will see the efficaciousness and accountable adequacy
of letting the index i include the case i = 0 in the product
n
g((x) * ll P{ *g())
i=O
and will, on seeing the utility of that device, know it as just such
a previously considered, familiar, accountable procedure.
109 In anticipation of topic (d), we should note that it is the
exhibited correspondence between the terms of P and the Godel
numbers of the terms of P that allows a prover to extend the
proof of Godel's theorem for P to logistic systems that properly
contain it. (Parenthetically, this is not to say that a prover could
not (with additional effort) recover from the preceding formula
for term(x) the necessary structure of the Godel numbering and,
therein, the construction of those certain sequences of terms to
which the adequacy of that formula is tied. The point, however,
which contrasts to this circumstance and which recalls the gestalt
theme of the figure/ground is this: a prover, in the course of
writing up a proof of a theorem, is engaged in the selection and
articulation of those details of the work of that proof that
provide for the recovery of that proof as an accountable course of
work, and that extractable course of work, in and as a materially-
embodied course of argumentation, comes to exhibit those
properties of the proof-relevant mathematical objects that are
accountably adequate to and, therein, revelatory of the descrip-
tive character of the assertion of the theorem itself. (I am
indebted to Harold Garfinkel for this formulation.))
110 The construction that follows bears a recognizable similarity to
that of var(x), and, later, to those of wff(x) and ded(x): first,
a general property is shown to be primitive recursive - like that
of a number x being the Godel number of a formation sequence
of terms - and, then, the desired relation - e.g., term(x) -is
shown to consist of those numbers that possess that property and
obey some further restriction -in this case,

222
 NOTES TO PAGES 122-5

L(x)
term(x) ~ !ly.;;;;; ll pf (formterm(y) and (y)L(y) = x).
i=O
Although this type of construction, as a repeating pattern within
a schedule of proofs, can be retrospectively disengaged from the
schedule as an incidental and nonessential feature of that
schedule in providing for the proof of Godel's theorem, over the
course of working through a schedule this repeating construction
is part of the exhibited and developing naturalness of the
arithmetized syntax as being constructed in programmatic and
sequential correspondence to the specification of the original
syntax of the system; it is the achievement and maintenance of
just this correspondence that makes up the accountable adequacy
of the schedule of proofs, and it is this that a prover relies on
later when he uses the functions and relations of the arithmetized
syntax in the diagonalizationf'proof'.
111 This. observation has already been illustrated in the discussion of
Godel numbering as a technique of proving. The fact that the
notions of a bound and of a free occurrence of a variable in a
formula, and that the notion of the substitution of a term for a
variable in a wff, are developed in the schedule of proofs in which
we are working as the extended series of Propositions 23 through
29 -introducing, in the following order, the functions/relations
occur(w, x, y, z), bound(w, x, y, z), free(w, x, y, z), S 1 (x, t, a),
sm(x, t, a), M(x, t, a), and, finally, Sub(x, t, a)- is a reminder
of the existence of such work and provides a perspicuous place
in a schedule for that work's further elucidation.
112 It should be noted that Godel did not use the logistic system of
Whitehead and Russell's Principia Mathematica in his original
paper but, instead, defined a formal system that, on one hand,
was more conducive to his methods of proving his theorem and,
therein, to the exhibitable and exhibited adequacy of those
methods for such proving and, on the other hand, a system that
retained what the new system simultaneously offered as the
essentially relevant features of Whitehead and Russell's system
for the problem at hand.
113 Parenthetically, the mathematical procedure of using a given
materially-specific proof and the structure of proving that that
proof both makes available and comes to exhibit to find, through
modifications (or variations) of its arguments, the essential
or essentially necessary (or ideal) structure of that proof bears
a similarity with one interpretation of Husserl's notion of the
phenomenological reduction as the free, imagined variation of
a perceived object so as to elucidate the essence of that object
as an intentional structure of consciousness. Thus, we are here
reminded of the possible origins of Husserl's phenomenology in
his early training and apprenticeship as a mathematician.
114 The summary characterization that follows stands here in place of
a detailed, descriptive review, in the fashion of topics (a) and (b),

223
NOTES TO PAGES 125-9

of the lived-work of developing this part of a schedule of proofs.

115 Here, 'wished' is being used as a gloss for the local and material
motives for a prover's inquiry into the accountable structure of
a proof of Godel's theorem as that inquiry is formulated as the
need to generalize that proof. Later, in 'A Structure of Proving,'
I will indicate how the need to so extend the proof of Godel's
theorem is 'natural' in the sense that it is tied to the account-
ability of the work of proving Godel's theorem and, therein,
to the structure of the proof of Godel's theorem itself.
116 I.e., to show that Godel's theorem holds for such extensions
of P as well as for P itself.
117 The appropriateness of such a number is tied to the necessary
things that need to be done with the Godel numbering as those
requirements become available over the course of constructing
the original proof of Godel's theorem. Thus, for example, the
Godel number of E could be defined as the first positive odd
number greater than those numbers assigned to the primitive
symbols of P. The constraints on such a definition, however, do
not have to do with the fact that the assigned number is, in and
of itself, distinct and/or odd, but, for example, with the fact that
if the primitive symbols of the language are all assigned odd
numbers, then the Godel numbers of the expressions and
sequences of expressions of the language will all be even numbers
(given our proof-specific Godel numbering), thereby permitting,
given any natural number, a strictly mechanical determination
of whether or not that number is the Godel number of a primitive
symbol of the language and, if it is, then which one. The notion
of the appropriateness of the assignment of Godel numbers
becomes more critical when the language of the formal system
is extended to having an infinite number of primitive symbols.
This matter will be addressed in the text below.
118 In .that the natural num hers are used in the defintiion of the t:U,
and in that the system Pis itself being used to articulate and
define (what is then understood as at least part of) the essential
structure of the natural numbers, it is not technically correct
to introduce the t:U as new primitive symbols of P. Although
this problem can be circumvented, as in the case of the individual
variables, by using the symbol' as inti := "t', the use of such
an austere notational system makes the definitions/proofs of the
schedule of proofs more difficult and, thereby, obscures the
structure of their exhibited reasoning. In that the replacement
of the symbols frm with technically correct notation is seen to
present no difficulties in principle -i.e., to present no problems
other than the technicalities of manipulating the notion for
reworking the definitions/proofs of the schedule -the use of
frm does not exhibit a damaging fault of a schedule of proofs.
As is customary, I have simply used the notational abbreviations
t:U later in the text as if they were primitive symbols; in
particular, I have followed Mendelson's presentation in

224
 NOTES TO PAGES 129-40

Introduction to Mathematical Logic.

119 As I have already indicated, the solution that will be given is
found in Mendelson's Introduction to Mathematical Logic. Of
necessity, some of the details of Mendelson's presentation have
been changed to fit the proof of Go del's theorem as that proof
has been developed here.
120 In addition to Mendelson's treatment (pp. 129-30) of course-of-
values recursion, the reader is referred to the exposition in
Stephan Cole Kleene, Introduction to Metamathematics (New
York: ·D. Van Nostrand Company, 1952), pp. 231-3.
121 Again, the solution that is offered here has been taken from
Mendelson, Introduction to Mathematical Logic.
122 The material in this section does not provide a finished argument,
but only some working notes directed to the recovery of a
proof's notation as that notation is tied to a prover's work at the
mathematical work site. I hope to address this relationship, with
materials other than those from a proof of Godel's theorem, in a
separate and later paper.
123 Atother places in the schedule, xis used simply to denote a
numerical variable as in the definition of the 'decoding' function
(x)j, where, as a further example, i is understood as indexing the
natural numbers in a serial fashion where as, in the diagonalization/
'proof, the letter i is used to stand for the Godel number of a
particular formula.
124 Parenthetically, the reader should note that the serial indexing of
m 1 , m 2 , and m 3 is here used as a local, occasioned and
unexplicated device for associating the variables with their
appropriate places in the symbolism Sub(x, t, a).
125 I take this occasion to point out a curious feature of the temporal
writing of s: A I and, therein, to a more general concern of
provers in collaborative work and in classroom and public
presentations: the writing of
s: s: sx s: s:
A I asS~ ~ ~ A~
A I does not correspond to the cotemporaneous descriptio rtf
elaboration of that writing as denoting the substitution oft for
x in A. In order to coordinate that writing and description a
prover may prefer to read s: A !"as 'the replacement of x by t
in A.' Alternatively, a prover may preserve the description and
write t prior to x, losing, however, the emphasis on the fact that
xis the variable that is being substituted for, or a prover could
produce s: A I as a gestalt, maintaining both the temporality of
the writing and the description of it as the substitution oft for x
in A but announcing, perhaps, 'oft for x' only after having
written s: and, then, pointing to t and x as they are named.
Three points need to be made: first, in writing a proof, a prover
will rely on the seeable things that are provided for by the
material notation and not on the verbal names or descriptions
of that notation; second, while the temporal coordination of
blackboard writing and the description of the notation being used
is an omnipresent feature of mathematical presentations, the

225
NOTES TO PAGES 140-7

exact manner in which blackboard displays are produced has not,

to my knowledge, been subjected to examination; and third,
the nature and consequentiality of that coordinated work, for
mathematical practice, cannot be decided prior to an actual
examination of those practices.
126 Quine, for example, adopts the convention of indicating the
variable x being substituted for as the subscript and the term t
being §Ubstituted as the superscript, writing, in his notation,
x' SF~ x to mean that the formula x' results from the
substitution of y' for the free occurrences of y in x. (Willard
Van Orman Quine, Mathematical Logic, revised edn (Cambridge:
Harvard University Press, 1951), p. 301.) By modifying the
notation used in the text, this would result in writing S~ A I
and, there by, of bringing its description as 'the-substitution of the
term t for x in A' into accord with the temporality of its writing.
127 Joel W. Robbin, (New York: Benjamin/Cummings Publishing
Company, 1969).
128 The reader will note, however, that I did not make such changes
uniformly throughout the schedule of proofs for reasons tied,
for example, to the interpretability of the notation for Sub(x,
t, a).
129 The variables in occur(w, x, y, z) are arranged in alphabetical
order, exhibiting, in that way, the arbitrariness of their selection
and, therein, the fortuitousness of x corresponding (if occur
(w, x, y, z) holds) to the GOdel number of an individual variable.
130 An initial candidate definition for freefor(x, t, a) might be the
relation T(t, x, a) defined by
T(t, x, a)~ term(t) and var(x) and wff(a) and
3a 1 ~ a 3a2 ~ a {free(a 1 , x, a 2 , a) implies not-3c 1 ~ a
3c2 ~a 3y ~a 3t1 ~ t 3t2 ~ t [var(y) and L(c 1 ) ~ L(a 1 )
and L(c 2 ) ~ L(a 2 ) and bound(c 1 , y, c 2 , a)]}.
T(t, x, a), however, is too restrictive a definition in that the
bound occurrence of y in A indicated by bound(c 1 , y, c 2 , a) may
occur in a well-formed part of A not containing the free occur-
rence of x. The more extended condition given later in the
text remedies this problem.
131 Where that arbitrariness, and the opacity of that notation, is now
to be understood in terms of the work associated with a schedule
of proofs' notational development.
132 K. Koffka, Principles of Gestalt Psychology (New York:
Harcourt, Brace and Co., 1935), p. 172.
133 Although, as we have seen in point 1 of this chapter, a prover may
use the specific assignment of Godel numbers to develop some of
the proofs of a schedule, the fact that that work has no residue
in the finished schedule does not mean that the prover is conceal-
ing that work from his reader. A prover is held to exhibit, in and
as the material detail of a proof, the work of proving that is

226
 NOTES TO PAGES 147-50

adequate to, and exhibiting of, that material detail as composing

a naturally accountable proof. No residue of the use of specific
Godel numbers is found in the schedule in that that work is not
work that is tied to the naturally accountable schedule of proofs.
It is the lack of that connection which gives to a prover's illus-
tration of and commentary on such work its character as only
heuristic and pedagogically directed remarks.
134 Generally, in the way in which the Godel numbering is otherwise
being used; specifically, as the reader will see shortly, in a manner
similar to that discussed later in the text.
135 As I did in some of the notes of point 1 of this chapter.
136 Such notation probably being introduced at the same time that
the Godel numbering is defined.
13 7 J. Donald Monk adopts this notation in his Mathematical Logic
(New York: Springer-Verlag Inc., 1976).
138 It should be pointed out that, given the need for different uses
of a Godel numbering, the designation of functions g, g+, and g++,
and designating them in such a way so as to suggest the relation-
ship between them, may be critical to the development of some
mathematical arguments. The reader is again referred to J. Donald
Monk, Mathematical Logic (New York: Springer-Verlag Inc.,
1976).

Chapter 8 A Structure of Proving

I.e., the rule of inference that states from 'A' and 'A implies B',
one can infer B.
2 Cf. the proof in Chapter 2 of the uniqueness of the identity
element of a group.
3 L. Chwistek gives an abbreviated proof, in a particular formal
system, of a theorem that he describes as 'correspond[ing] to
the first theorem of Godel,' but where 'the contradiction does
not appear in f-0 0 but in f-2 2.' Chwistek states this theorem as
I= [ 10] ::::> V f-o G(2) h ~ G(2) f-o h = .0(5).1(5)[5] is a
theorem
where GL(E) is an abbreviation of
3 [.O(L)L] XLYLZLfiLA =Ell [.2(L).l(L)]y'L.ZL[.O(L)]
1\Im.O(L) EX:L A (.ZLhX:LuL) [.O(L)] ~ f-L uL
and G(L) is an abbreviation of
GL(ll [.2(L).l(L)] ii.l(L} G.l(L) (ii.l(L))).
This theorem allows him to prove 'Godel's [second] theorem'
I= [101 ::::> h (32) ~ h (54)= .0(5).1(5)[51 h (32) =
.0(3).1(3)[3]
in a manner that 'is quite formal and does not differ in any way

227
NOTES TO PAGES 150-4

from ordinary arithmetical proofs.' The relevant articles here

are L. Chwistek and W. Hetper, 'New Foundations of Formal
Metamathematics,' Journal of Symbolic Logic, 3, 1, March 1938,
pp. 1-36 and L. Chwistek, 'A Formal Proof of Godel's Theorem,'
Journal of Symbolic Logic, 4, 2, June 1939, pp. 61-9.
4 I am indebted to Harold Garfinkel for this last observation.
5 The notion of an organizational object that I am using here is due
to Harold Garfinkel, and I have relied on his writings and lectures
in developing the list that follows. Responsibility for the material
in the text is my own. ·
6 'Anyone' refers to the fact that the proof is available to provers
as a practically objective thing, as an 'organizational object.' For
provers, if someone cannot see that such a proof has been given,
if a person cannot see that the work done to prove the theorem
was the work that was required of that proof, then reasons are
given for that person's inability to 'understand' that proof- e.g.,
that that person does not have a 'mathematical mind,' that
he has not been 'trained' as a mathematical logician, etc.
(Parenthetically, in that it is the inability of someone to do the
work that makes up the recognizably adequate proofs of
mathematics that exhibits that person's mathematical incompet-
ence (and, conversely, in that it is the ability of someone to do
that work that identifies him, in the company of other provers,
as a prover himself), what the formulations of a person's mathe-
matical incompetence can be seen to always lack is the
specification of the features of the lived-work of proving, from
within that lived-work, that are identifying of that work as such.)
7 E.g., on seeing the proof that a homomorphism ifJ of a group A
to a group B maps the identity 1A of A to the identity 1B of B -
the main 'idea' of which is that, by writing lA as lA lA, one can write
ifJ(iA) = ifJ(lAlA) = ifJ(lA) ifJ(lA)
- a beginning prover may not see the 'relevance' of this same
'idea' to proving that ifJ(x- 1 ) = [ ifJ(x)r 1 thus,
lB = ifJ(lA) = ifJ(xx- 1 ) = ifJ(x) ifJ(x- 1 ).
8 As the reader will see, the expression 'a structure of proving' is
intended as an initial, descriptive formulation of the
phenomenon of which it speaks. Thus, the problems that remain
are not those of analyzing and constructively interpreting the
notion of a 'structure of proving,' but those of using that notion,
if possible, to gain deeper and more technical access to mathe-
maticians' work practices. By pointing out that the notion of a
'structure of proving' provides only a provisional solution to the
characterization problem, I wish to emphasize that it remains for
later investigations to come to a more competent understanding
of the characterization problem.
9 As I have alieady indicated, a more precise statement of the
theorem is that, for the particular sentence J constructed in the

228
 NOTES TO PAGES 154-5

diagonalization argument, (1) if Pis consistent, then J is not

deducible in P, and (2) if Pis w-consistent, then~ J is not
deducible in P.
10 I.e., for every wff of P with no free variables.
11 Parenthetically, the reason Godel's 'second' theorem is also
referred to as an incompleteness result is that that theorem 'says'
that one of the 'undecidable' sentences of formal number theory
is the (a) sentence that 'arithmetically' asserts that that theory
is consistent; thus, the more common formulation- if the theory
of arithmetic is consistent, then its consistency cannot be
demonstrated within the theory itself.
12 Moreover, in that J can be interpreted as 'saying' that it itself i$
not deducible in P, and in that J is not deducible in P by the
proof ·of Godel's theorem given earlier, J represents an 'intuitively'
true proposition of elementary arithmetic (where 'intuitively'
refers to the informal character of the procedure whereby J·is
given its interpreted meaning).
13 In that the non-logical axioms of P are true propositions of
number theory under their standard interpretation in N, and in
that the logical axioms and rule of inference of P correspond as
abstractions or epitomizing characterizations of familiar and
regularly recurring techniques of naturally accountable mathe-
matical reasoning, the difficulty in not being able to show that
Pis complete -assuming that ordinary number theory is itself
consistent and complete - does not seem to reside in the fact
that P is too strong a theory in the sense that restrictions in the
specification of P might result in a system p- that still
'represented' number theory but in which a comparable
statement of Godel's theorem did not hold. This reasoning
underlies the discussion of the possible extension of P to a
complete theory. Two comments, however, are appropriate.
First, one of the attempts to circumvent the apparent achieve-
ment of Godel's second incompleteness theorem in regard to
the formal investigation of elementary arithmetic was based on
the diagnosis that the proofs of Godel's theorems depended on an
unrestricted use of universal quantification and that, in this way,
Pis too strong a theory. {See Alonzo Church, Mathematical
Logic, Lectures at Princeton University, October 1935-January
1936, notes by F .A. Ficken, eta/.) Second, Gerhard Gent zen was
able to prove the consistency of a system representing elementary
arithmetic, but used the inferential procedure of transfinite
induction up to the ordinal

Eo =wWW.

in order to do so. Transfinite induction up to Eo is not itself

deductible in our system P, and, because of this, Gentzen's
proof of the consistency of 'arithmetic' may be said to rely on
methods stronger than those used in 'arithmetic' itself.

229
NOTES TO PAGES 155-6

Gentzen's original papers appear in translation in M.E. Szabo

(ed.), The Collected Papers of Gerhard Gentzen (Amsterdam:
North-Holland Publishing Company, 1969). The reader is also
referred to discussions of Gentzen's work in Elliot Mendelson,
Introduction to Mathematical Logic (Princeton: D. Van
Nostrand Company, Inc., 1964), Howard Delong, A Profile of
Mathematical Logic (Reading: Addison-Wesley, 1970), Stephan
Cole Kleene, Introduction to Metamathematics (New York:
D. Van Nostrand Company, 1952), and Gaisi Takeuti,Proof
Theory (Amsterdam: North-Holland Publishing Company,
1975).
14 The reader may wish to argue that, properly understood, the
proof of Godel's theorem, as it has been presented for P,
already answers this question. What is at stake, however, is not
the fact of whether this is or is not so, but the specification of
the work required on the part of a prover to see that this is so.
15 The syntax of this system will not be precisely defined, and the
reader is referred to conventional sources for such definitions.
Particularly careful treatment of such matters is to be found in
Alonzo Church,lntroduction to Mathematical Logic (Princeton:
Princeton University Press, 1956), Vol. 1. For the material that
follows, I have used, and wish to acknowledge my indebtedness
to, Elliot Mendelson, Introduction to Mathematical Logic
(Princeton: D. Van Nostrand, 1964) and Joel W. Robbin,
Mathematical Logic: A First Course (New York: Benjamin/
Cummings Publishing Company, 1969).
16 I will not make explicit all the parentheses in the formulae that
follow, nor will I place the function symbols+ and ·to the left
of their arguments as would be required by an explicit statement
of the formation rules of the system. In stating the axioms, such
conformity to the formal specification of the syntax of P would
only obfuscate the meaning of the axioms as interpreted in N.
On the other hand, it is important to note the presence of such
abbreviations in that, in later constructing the schedule of proofs,
we will use the unabbreviated form of some of these axioms.
17 In this note, I indicate a technical revision of the specification of
the axioms of P.
A first-order theory Tis said to be a first-order theory with
equality if x 1 = x 1 and x = y :::> (A :::> A 1) are theorems of T,
where x andy are individual variables, A is a wff, A' is obtained
from A by replacing some, but not necessarily all, occurrences of
x in A by y' andy is free for x at those places in A where it
replaces x. (Eliott Mendelson, Introduction to Mathematical
Logic (Princeton: D. Van Nostrand, 1964 ), p. 75). Define the
atomic formulas of a theory T to be the formulas of the form
Aj (t 1 , . . . , t n) where A j is an n-place predicate symbol of T
and t 1 , . . . , tn are terms ofT. (In P, the only predicate symbol
is=, and the atomic formulas are written in the form t 1 = t 2 .)
Then, it can be shown that if x 1 = x 1 and if x = y :::>(A :::>A 1)

230
 NOTES TO PAGES 156-8

hold for the atomic formulas A of a first-order theory T, then Tis

a theory with equality. (Mendelson, p. 76.) Let us number these
last two formulas 6' and 7', respectively. Now, while axioms 6
and 7 in the text are themselves sufficient to show that P is a
first-order theory with equality (Mendelson, p. 107), it is
customary and convenient to distinguish between a basic under-
lying logistic system (of which the quality relation, if present,
is considered a part) and the specific axioms of a theory under
construction. The axioms of P can be brought into accord with
this policy by replacing axioms 6 and 7 with new axioms 6' and
7', and the new axiom 7' can be introduced into the schedule
of proofs as follows: First, define atomic( a) as holding if and
only if a is the Godel number of an atomic formula of P.
atomic(a) is primitive recursive since
atomic(a) ~ 3t 1 ~a 3t 2 ~a (term(t 1 ) and term(t 2 ) and
a=t 1 *g(=)*t 2 ).
Then
axiom 7 •(w) ~ 3x ~ w 3y ~ w 3a ~ w 3b ~ w {var(x) and
var(y) and atomic(a) and L(a) = L(b) and
Vi ~ L(a) ((a)i = (b )j or [ (a)j = x and (b )j = y])
and w = g((() * x * g(=) * y * g() ~ ()*a* gp) *
b * g()))} .
Since a is the Godel number of an atomic formula A -from
which it follows that there are no quantifiers in A - we need not
be concerned over whether or not y is free for x in A. (The idea
of this definition/proof was taken from Herbert B. Enderton,
A Mathematical Introduction to Logic (New York: Academic
Press, 1972), p. 224.)
18 I have made no attempt at originality in developing the 'schedule
of proofs' that follows; most of the statements and proofs of it
follow closely those found in Joel W. Robbin, Mathematical
Logic: A First Course (New York: Benjamin/Cummings
Publishing Company, 1969). I have made some changes -in
wording and detail -and, as always, responsibility for the text is
my own. In addition to these minor changes, I have modified
Robbin's presentation in two other ways: First, some
modifications were necessitated by the fact that Robbin proves
Godel's theorem for first-order primitive recursive arithmetic,
whereas I provide a proof of it for the system P. Second, Robbin,
as is now customary as well as pedagogically efficient (both for
reasons internal to the proof of Godel's theorem and in order
to introduce topics - e.g., primitive recursive functions and
relations - independently of their use in a proof of that theorem),
divides the schedule of proofs into distinct, topically identified
sections. In this, I have instead followed Godel's original

231
NOTES TO PAGES 158-67

formating of the propositions as a displayed 'schedule.' On the

other hand, however, Godel introduces the necessary elementary
primitive recursive functions·and relations and techniques for
constructing primitive recursive functions and relations prior to
beginning the schedule itself; his schedule begins with the
assertion of Proposition 11 below. (See Kurt GOdel, 'pn
Formally Undecidable Propositions of Principia Math~ematica and
Related Systems I,' in translation from German in Jean van
Heijenoort, From Frege to Godel: A Source Book in Mathe-
matical Logic 1879-1931 (Cambridge: Harvard University Press,
1967), pp. 596-616. Finally, in constructing a 'diagonalization/
"proof"' for the proof of Godel's theorem, I made use of the
'diagonalizationj"praof'" found in Raymond Wilder, The
Foundations of Mathematics, 2nd edn (New York: John Wiley
and Sons, 1965), pp. 272-3.
19 In that the construction of a schedule of proofs is informed by
the project of arithmetizing the syntax of P, one last proposition
offers itself for inclusion in the schedule, namely
36 theorem(a), which holds if and only if a is the Godel
number of a theorem, is a primitive recursive relation,
as wt:ll as a potential proof of it,
theorem(a) ~ 3A.,;:;; f(a) {ded(A., a)} ,
where f(a) is a primitive recursive function of a. However, the
construction of such a function f(a) is not forthcoming; in fact,
if such a primitive recursive function existed - and, hence, if
theorem(a) was a primitive recursive relation - then, given any
formula M of P, one could compute g(M) and, in principle,
calculate whether or not M was deducible in P. This, however,
contradicts a consequence of Church's theorem that if Pis w-
consistent, then there can be no 'effective procedure' for
determining, for an arbitary wff M, if M is or is not a theorem of
P. Explicit mention of the relation theorem(a) is made, for
example, in the schedules of proofs found in Godel's original
paper, in Robbin, Mathematical Logic: A First Course, and in
Joseph R. Shoenfield, Mathematical Logic (Reading: Addison-
Wesley, 1967). ·
20 A proof of this theorem can be found in Elliot Mendelson,
Introduction to Mathematical Logic (Princeton: D. Van
Nostrand, 1964), pp. 131-4.
21 For example, see the proof of Elliot Mendelson, Introduction
to Mathematical Logic (Princeton: D. Van Nostrand, 1964 ),
pp. 131-4. .
22 Throughout this book, I have used the notion of P being a
'model' of N in the ordinary sense of P being a representation of
it. In mathematical logic, however, a modelrolof a set of
sentences is defined by first defining the notion of logical impli-
cation (indicated notationally as F=~a) as distinguished from

232
 NO fES TO PAGES 167-8

the notion of deducibility (with its notational designation l-p o).

Given this definition, a sentence o is true in!n(andiDlis said to
be a model of o) if and only if 1= 111 o. is said t-o be a model of
a set I: of sentences if and only if it is a model of every member
of I:. (See Herbert B. Enderton, A Mathematical Introduction to
Logic (New York: Academic Press, 1972), p. 84 and the pages
immediately preceding p. 84 for prepamtory material.) In this
sense, then, N = {INI, 0, S, +, · } is said to be a model of the
axioms of P, thus reversing our way of informally speaking of P
as a 'model' of N.
23 For a little more detail, however, see note 17 on p. 230. Besides
the criticisms found there, other criticisms of the reasoning in
the text above are possible as, for example, the fact that our
system Pis not finitely axiomizable. What I wish to stress here,
however, is not the endless refinability of Godel's theorem and
its proof, but the fact that the proof of that theorem is witness~
ibly exhibiting of its naturally accountable accomplishment. I
will briefly return to this matter at the beginning of section C
of this chapter.
24 As noted later in the text, such an extension need not be limited
to the language of P, but may be an extension of that language
as well.
25 One of the aims of the next section is to begin to elucidate
what it means 'to see' that this is so. Robbin, for example,
writes: 'It is, of course, not particularly surprising that a
particular set of axioms is not complete. The full impact of the
Godel theorem comes on realizing that the proof of the Godel
theorem works for any reasonable set of axioms for arithmetic.'
(Joel W. Robbin, Mathematical Logic: A First Course (New
York: Benjamin/Cummings-Publishing Company, 1969), p. 115,
italics mine.) The problem that I want to point to is that of
determining what, as pmctice, such a 'realization' consists of.
I now quote Robbin's 'concluding' remarks in full as an
illustration of how an author might 'tempomrily close' the
discussion of the proof of Godel's f'trst theorem. (For my
qualification concerning the tempomry chamcter of the closure
of the discussion of the proof the reader is again referred to the
next section of this chapter.)
It is, of course, not particularly surprising that a particular set
of axioms is not complete. The full impact of the GOdel
theorem comes on realizing that the proof of the GOdel
theorem works for any reasonable set of axioms for arithmetic.
So long as the set of Godel numbers of axioms is primitive
recursive, the Godel technique [the diagonalization procedure]
can be applied to yield a true sentence which is not provable.
Thus, if we form a new formal language RA~ by adding J as an
axiom to the axioms of RA [first-order primitive recursive
arithmetic], we have a new two-place predicate ded'(o, a),
which says that o is the Godel number of a deduction in RA'

233
NOTES TO PAGES 168-71

of the wff with GOdel number a, and by exactly the same

construction as above, we get a sentence J' which is true but
is not a theorem of RA'. Thus, as long as we demand of a
formal language that there be an effective procedure for
recognizing which sequences of wffs are deductions (i.e.,
ded(a, a) is primitive recursive), there is no formal language
whose theorems consist of precisely the true sentences of
arithmetic.
26 These axioms have been taken from Herbert B. Enderton, A
Mathematical Introduction to Logic (New York: Academic Press,
1972), p. 194.
27 See pp. 125-37.
28 Barkley Rosser, 'Extensions of Some Theorems of Godel and
Church,' Journal of Symbolic Logic, 1, 3 (1936), pp. 87-91.
29 In his paper 'An Unsolvable Problem of Elementary Number
Theory,' Alonzo Church writes in a footnote: 'This definition
is closely related to, and was suggested by, a definition of recur-
sive functions which was proposed by Kurt Godel, in lectures
at Princeton, N.J., 1934, and credited by him in part to an
unpublished suggestion of Jacques Herbrand. The principal
features in which the present definition of recursiveness differs
from Godel's are due to S.C. Kleene,' American Journal of
Mathematics, Vol. LVIII, no. 2 (1936), p. 351, n. 9.
30 Alonzo Church, 'An Unsolvable Problem of Elementary Number
Theory,' American Journal of Mathematics, 58,2 (1936), pp.
345-63.
31 Alfred Tarski, 'The Concept of Truth in Formalized Languages,'
trans. from the German translation of the original Polish text
(1933) by J .H. Woodger in Alfred Tarski, Logic, Semantics, aRd
Metamathematics (Oxford: Clarendon Press, 1956), pp. 152-278.
See as well Robbin's treatment in Joel W. Robbin, Mathematical
Logic: A First Course (New York: Benjamin/Cummings Publish-
ing Company, 1969), pp. 119-21.
32 Gerhard Gentzen, 'Die Widerspruchsfreiheit der reinen Zahlen-
theorie,' Mathematische Annalen, 112 (1936), pp. 493-565,
translated as 'The Consistency of Elementary Number Theory'
in M.E. Szabo (ed.), The Collected Papers of Gerhard Gentzen;
'Neue Fassung des Widerspruchsfreiheitsbeweises fiir die reine
Zahlentheop.e,' Forschungen zur Logik and zur Grundlegung der
exakten Wissenschaften, New Series, no. 4, Leipzig (Hirzel),
(1938), pp. 19-44, translated as 'New Version of the Consistency
Proof for Elementary Number Theory' in M.E. Szabo (ed.),
op. cit., pp. 252-86.
33 It should be noted that it is not necessary for a proof of Godel's
theoremto incorporate this self-referential feature explicitly;
a more 'contemporary' way of proving Godel's theorem is to first
prove a 'diagonal lemma,' to use that lemma to show that the set
of Godel numbers of the theorems of any consistent extension
of P is not recursive (Church's theorem), to show that if a formal

234
 NOTES TO PAGES 171-81

system is axiomatized and complete, the set of Godel numbers

of the theorems of the system is recursive, and, in consequence,
to conclude that any consistent extension of Pis incomplete.
(See, for example, Joseph R. Shoenfield,Mathematical Logic
(Reading: Addison-Wesley Publishing Company, 1967), pp. 131-
2). The relevance of the self-referential character of Godel's
(and our) proof to the discussion of what identifies a proof of
a particular theorem as a proof of just that theorem is some-
what indirect and is indicated in the text that immediately
follows.
34 I am indebted to Harold Garfinkel for the observation that a
proof's achievement is the pointlessness of disputing that
proof's reasoning and the properties of the mathematical objects
that that proof makes descriptively available. Responsibility for
the use of this expression in the text is, of course, my own.
35 The formulation of a proof as consisting of such a pair is due to
Harold Garfinkel. He speaks of the first part of the pair not as
the 'material proof' but as a 'proof account.' The pointedness
of this observation will be briefly indicated later and in the next
chapter.

Chapter 9 Summary and Directions for Further Studies

The discovery and formulation of 'classical studies of work,'

the availability of 'classical studies' as a phenomenon from within
ethnomethodological studies of work, and the discovery and
formulation of the discovering sciences as classical sciences of
practical action are all due to Harold Garfinkel. I am indebted
to him for making this material freely available. I am indebted
to him as well for the overall perspective from within which this
section was written. Responsibility for the text, however, is
mine alone.
2 The discovery and formulation of a mathematical proof as a
classical study of practical action, and the discovery and formu-
lation of mathematics in particular, and of the discovering
sciences in general, as classical sciences of practical action and
practical reasoning are due to Harold Garfmkel. I am indebted
to him for offering, and thus for bringing my studies to, these
both epitomizing and deeply suggestive formulations of
mathematicians' work.

Appendix

A reference for the material from physics used in this appendix

is Richard P. Feynman, Robert B. Leighton and Matthew Sands,
The Feynman Lectures on Physics (Reading: Addison-Wesley,
1964), Vol. II, Chapters 2 and 3. Michael Spivak's Calculus on

235
NOTES TO PAGES 181-6

Manifolds (Menlo Park: W.A. Benjamin, 1965) is a source for the

material from mathematics.
2 This appendixwas written before the book. At this point
in the appendix, an attempt to abstractly formulate what it
means to speak of the relationship between mathematics and
physics as a problem in the production of social order would
introduce a severe distraction into its presentation. A sense of
what this means should become clearer by the end of the
appendix. The reader is asked to wait until he has read the body
of the book to begin to see how the phrase 'a problem in the
production of social order' takes on its descriptive sense within
a technical argument, its use being animated by and clarifying of
the material details of the mathematicians' praxis.
3 The statement of the theorem and the material concerning it is
taken primarily from Spivak, Calculus on Manifolds.
4 I speak here of two 'reasons' for the pervasiveness of this pre-
supposition. What is intended, however, is an initial way of
locating the embeddedness of conventional analyses of mathe-
matical physics within the methods of doing those analyses.
5 Very briefly, Fourier's law is an imprecise law that says that the
flow of heat in an isotropic medium is, infinitesimally, in the
direction of the maximum decrease of temperature and in magni-
tude, proportional to that decrease. Fourier's law can be
empirically checked, whereas the adequacy of the heat equation
is found by comparing its solution, subject to the specification
of initial and boundary conditions, to the empirical object. The
heat equation is a linear partial differential equation that gives the
local law of the change of temperature. In the discussion that
follows, I am not concerned with proper constants (asK) and
write Llu ~ Ut for an essential equality. Ll is the Laplacian; the
subscript t stands for partial differentiation with respect to time.
6 This appendix, having been written prior to the body of the book,
has postponed the analysis of mathematicians' praxis. Neverthe-
less, I point out here that the methods of reading theoretical-
physics-as-matpematics simultaneously interrogate and organize
a text so as to elucidate, and therein build, a properly mathe-
matical argument. From within these practices, the mathematics
of theoretical physics stands immediately juxtaposed to its
potentially mathematical rendering. These practices stand in
contrast to the diverse and as of yet unanalyzed ways in which
physicists examine such writing from within their praxis as
theoretical physicists.

236
Bibliography

Baccus, Melinda. 'Sociological Indication and the Visibility Criterion of

Real World Social Theorizing.' Unpublished paper, Department of
Sociology, UCLA, 1976.
Baccus, Melinda. 'Multipiece Truck Wheel Accidents and Their Regu-
lation.' Unpublished paper, Department of Sociology, UCLA, 1981.
Barnes, Donald W. and Mack, John M. An Algebraic Introduction to
Mathematical Logic. New York: Springer-Verlag Inc., 1975.
Bellman, Beryl L. The Language of Secrecy. Unpublished monograph,
Department of Sociology, University of California, San Diego, 1981.
Boolos, G. and Jeffrey, R. Computability and Logic. Cambridge:
Cambridge University Press, 1974.
Burns, Stacy. 'A Comparison of Geertzian and Ethnomethodological
Methods for The Analysis of the Production of Member-Relevant
Objects.' Unpublished paper, Department of Sociology, UCLA,
1978. .
Burns, Stacy. 'A Compari~on of March-Olson and Ethnomethodology
on "Circumstantial" Analysis of Decision-Making Work.' Unpub-
lished paper, Department of Sociology, UCLA, 1978.
Burns, Stacy. 'Lecturing's Work.' Unpublished paper, Department of
Sociology, UCLA, 1978.
Burns, Stacy. 'Becoming a Lawyer at Yale Law School.' Unpublished
paper, New Haven: Yale Law School, 1981.
Choquet, Gustave. Geometry in a Modern Setting. Boston: Houghton
Mifflin Company, 1969.
Church, Alonzo. 'An Unsolvable Problem of Elementary Number
Theory.' American Journal of Mathematics 58,2 (1936): 345-63.
Church, Alonzo. Mathematical Logic. Lectures at Princeton University,
1935-6. Notes by F.A. Ficken, H.G. Landau, H. Ruja, R.R. Single-
ton, N.E. Steenrod, H.H. Sweer and F.J. Weyl.
Church, Alonzo. Introduction to Mathematical Logic, Vol. I. Princeton:
Princeton University Press, 1956.
Chwistek, L. 'A Formal Proof of Gi:idel's Theorem,' Journal of Sym-
bolic Logic 4, 2 (1939): 61-9.

237
BIBLIOGRAPHY

Chwistek, L. and Hetper, W. 'New Foundations of Formal Mathe-

matics.' Journal of Symbolic Logic 3, 1 (1938): 1-36.
Delong, Howard. A Profile of Mathematical Logic. Reading: Addison-
Wesley Publishing Company, 1970.
Eglin, Trent. 'Introduction to a Hermeneutics of the Occult: Alchemy.'
In E. Tiryakian (ed.), On the Margin of the Visible: Sociology, the
Esoteric, and the Occult. New York: John WHey and Sons.
Enderton, Herbert B. A Mathematical Introduction to Logic. New
York: Academic Press, 1972.
Fauman, Richard. 'Filmmakers' Work: On the Production and Analysis
of Audio-Visual Documents for Social Sciences.' Unpublished paper,
Department of Sociology, UCLA, 1980.
Feferman, F. 'Arithmetization of Metamathematics in a General
Setting.' Fundamenta Mathematicae 49 (1960): 35-92.
Feynman, Richard P., Leighton, Robert B. and Sands, Matthew. The
Feynman Lectures on Physics. Vol. II. Reading: Addison-Wesley
Publishing Company, 1964.
Garfi~kel, Harold. 'A Conception of and Experiments with "Trust"
as a Condition of Concerted Stable Actions.' In O.J. Harvey (ed.),
Mo.tivation and Social Interaction. New York: The Ronald Press
Company, 1963.
Garfinkel, Harold. Studies in Ethnomethodology. Englewood Cliffs:
Prentice-Hall, Inc., 196 7.
Garfinkel, Harold and Burns, Stacy. 'Lecturing's Work of Talking
Introductory Sociology.' Unpublished paper, Department of Sociol-
ogy, UCLA, 1979.
Garfinkel, Harold, Lynch, Michael and Livingston, Eric. 'The Work of
the Discovering Sciences Construed with Materials from the Optic-
ally Discovered Pulsar.' Philosophy of the Social Sciences 11 (1981):
131-58.
Garfinkel, Harold and Sacks, Harvey. 'On Formal Structures of Practical
Actions.' In John C. McKinney and Edward Tiryakian (eds), Theor-
etical Sociology: Perspectives and Developments. New York:
Appleton-Century-Crofts, 1970.
Gentzen, Gerhard. 'Die Widerspruchsfreiheit der reinen Zahlentheorie.'
('The Consistency of Elementary Number Theory.') Mathematische
Annalen 112 (1936): 493-565. English translation appearing in
M.E. Szabo (ed.), The Collected Papers of Gerhard Gentzen, pp.
132-213.
Gentzen, Gerhard. 'Neue Fassung des Widerspruchsfreiheitsbeweises
fur die reine Zahlentheorie.' ('New Version of the Consistency Proof
for Elementary Number Theory.') Forschungen zur Logik und zur
Grundlegung der exakten Wissenschaften, New Series, 4, Leipzig
(Hirzel), (1938): 19-44. English translation appearing in M.E.
Szabo (ed.), The Collected Papers of Gerhard Gentzen, pp. 252-86.
Godel, Kurt. 'lJber formal unentscheidbare Sitze der Principia Mathe-
matica und verwandter Systeme 1.' ('On Formally Undecidable
Propositions of Principia Mathematica and Related Systems 1.')
Monatschefte jiir Mathematik und Physik 38 (1931): 173-98.

238
 BIBLIOGRAPHY

English translation appearing in Jean van Heijenoort, From Frege to

Godel: A Source Book in Mathematical Logic, 1879-1931, pp.
596-616.
Gurwitsch, Aron. The Field of Consciousness. Pittsburgh: Duquesne
University Press, 1964.
Kleene, Stephan Cole. Introduction to Metamathematics. New York:
D. Van Nostrand Company, Inc., 1952.
Koffka, K. Principles of Gestalt Psychology. New York: Harcourt,
Brace and Co., 1935.
Leiberman, Kenneth. Understanding Interaction in Central Australia:
An Ethnomethodological Study of Australian Aboriginal People.
Doctoral Dissertation, Department of Sociology, University of
California, San Diego, 1980.
Leiberman, Kenneth. 'The Economy of Central Australian Aboriginal
Expression: An Inspection from the Vantage of Merleau-Ponty and
Derrida.' Semiotica, in press.
Levi, Howard. Foundations of Geometry and Trigonometry. Engle-
wood Cliffs: Prentice-Hall, Inc., 1960.
Livingston, Eric. 'An Ethnomethodological Approach to the Arts.'
Unpublished paper, Department of Sociology, UCLA, 1976.
Livingston, Eric. 'Mathematicians' Work.' Paper presented in the
session 'Ethnomethodology: Studies of Work,' Ninth World Congress
of Sociology, Uppsala, Sweden, 1978.
Lynch, Michael. Art and Artifact in Laboratory Science: A Study of
Shop Work and Shop Talk in a Research Laboratory. Doctoral
Dissertation, School of Social Sciences, University of California,
Irvine, 1979.
Lynch, Michael. Art and Artifact in Laboratory Science. London:
Routledge & Kegan Paul, 1985.
Lynch, Michael, Livingston, Eric and Garfinkel, Harold. 'Temporal
Order in Laboratory Work.' In Karen Knorr and Michael Mulky
(eds), Science Observed: Perspectives on the Social Study of Science.
London; Sage Publications Ltd., 1983.
Lyndon, Roger C. Notes on Logic. Princeton: D. Van Nostrand Com-
pany, Inc., 1966.
Mendelson, Elliott. Introduction to Mathematical Logic. Princeton: D.
Van Nostrand Company Inc., 1964.
Merleau-Ponty, Maurice. Phenomenology of Perception. Translated by
Colin Smith. London: Routledge & Kegan Paul, 1962.
Monk, Donald J. Mathematical Logic. New York: Springer-Verlag Inc.,
1976.
Morrison, Kenneth. Reader's Work: Devices for Achieving Pedagogic
Events in Textual Materials for Readers as Novices to Sociology.
Doctoral Dissertation, Department of Sociology, York University,
1976.
Morrison, Kenneth. 'Some Researchable Recurrences in Sciences and
Social Science Inquiry.' Unpublished paper. Toronto: York Univer-
sity, 1980.
Pack, Christopher. 'Features of Signs Encountered in Designing a

239
BIBLIOGRAPHY
Notational System for Transcribing Lectures.' Unpublished paper,
Department of Sociology, UCLA, 1975.
Pack, Christopher. 'Towards a Phenomenology of Transcription.'
Unpublished paper, Department of Sociology, UCLA, 1975.
Pack, Christopher. Interactional Synchrony Reassessed -An Attempt
to Apply the Discovery of Interactional Synchrony in Neonates as
an Index of Attachment in Non-organic Failure to Thrive Syndrome.
Doctoral Dissertation, Department of Sociology, UCLA, 1982.
Pollner, Melvin. 'Notes on Self-Explicating Settings.' Unpublished paper,
Department of Sociology, UCLA, 1970.
Quine, Willard Van Orman. Mathematical Logic, revised edn. Cambridge:
Harvard University Press, 1951.
Robbin, Joel W. Mathematical Logic: A First Course. New York:
Benjamin/Cummings Publishing Company, 1969.
Robillard, Albert R. 'Applied Behavioral Analysis.' Unpublished paper,
Department of Human Development, Michigan State University,
1980.
Robillard, Albert R. and Pack, Christopher. 'The Clinical Encounter:
The Organization of Doctor-Patient Interaction.' Unpublished
paper, Department of Human Development, Michigan State Univer-
sity, 1980.
Robillard, Albert R., White, Geoffrey M. and Maretzki, Thomas W.
'Between Doctor and Patient: Informed Consent in Conversational
Interaction.' Unpublished paper, University of Hawaii, Department
of Psychiatry, 1981.
Rosser, Barkley. 'Extensions of Some Theorems of Godel and Church.'
Journal of Symbolic Logic 1, 3 (1936): 87-91.
Rosser, Barkley. 'An Informal Exposition of Proofs ofGodel's Theorems
and Church's Theorem.' Journal of Symbolic Logic 4 (1939):
53-60.
Sacks, Harvey. 'On the Analyzability of Stories by Children.' In John
J. Gumperz and Dell Hymes (eds), Directions in Sociolinguistics:
The Ethnography of Communication. New York: Holt, Rinehart and
Winston, 1972.
Sacks, Harvey. 'Unpublished Lectures.' Department of Sociology,
UCLA and School of Social Sciences, University of California,
Irvine, 1964-75.
Sacks, Harvey, Schegloff, Emanuel and Jefferson, Gail. 'A Simplest
Systematics for the Organization of Tum-Taking on Conversation.'
Language 50 (1974): 696-735.
Sah, Chih-Han.Abstract Algebra. New York: Academic Press, 1967.
Schegloff, Emanuel. 'Sequencing in Conversational Openings.' American
Anthropologist 70 (1968): 1075-95.
Schegloff, Emanuel. 'Notes on a Conversational Practice: Formulating
Place.' In David Sudnow (ed.), Studies in Social Interaction.
Schegloff, Emanuel and Sacks, Harvey. 'Opening Up Closings.' Semiotica
8 (1973): 289-327.
Schenkein, Jim (ed.), Studies in the Organization of Conversational
Interaction. New York: Academic Press, 1978.

240
 BIBLIOGRAPHY

Schrecker, Friedrich. 'Doing a Chemical Experiment: The Practices of

Chemistry Students in a Student Laboratory in Quantitative Analy-
sis.' Unpublished paper, Department of Socioloay, UCLA, 1980.
To appear in Harold Garfmkel (ed.), Studies of Work in the Dis·
covering Sciences.
Shoenfield, Joseph R. Mathematical Logic. Readins: Addison-Wesley
Publishing Company, 1967.
Skolem, Thoralf. 'Begriindung der elemetaren Arithmetik durch die
rekurrierende Denkweise ohne Anwendung scheinbarer Verinder-
lichen mit unendlichem Ausdehnungsbereich.' (The Foundations of
Elementary Arithmetic Established by Means of the Recursive Mode
of Thought, Without the Use of Apparent Variables Ranging over
Inimite Domains.') Videnskapsselskapets skrifter, I. Matematisk·
naturvidenskabelig klasse, no. 6 (1923). English translation appear-
ing in Jean van Heijenoort. From Frege to Godel: A Source Book
on Mathematical Logic, 1879-1931, pp. 303-33.
Spivak, Michael. Calculus on Manifolds. Menlo Park: W.A. Benjamin,
Inc., 1965.
Sudnow, David (ed.), Studies in Social Interaction. New York: The
Free Press, 1972.
Sudnow, David. Ways of the Hand. Cambridge: Harvard University
Press, 1978.
Sudnow, David. Between Two Keyboards. New York: Alfred Knopf
and Company, 1980.
Szabo, M.E. (ed.}, The Collected Papers of Gerhard Gentz en. Amsterdam:
North-Holland Publishing Company, 1969.
Takeuti, Gaisi, Proof Theory. Amsterdam: North-Holland Publishing
Company, 1975.
Tarski, Alfred. 'The Concept of Truth in Formalized Languages.'
Translated from the German translation of the original Polish text
(1933) by J.H. Woodger in A. Tarski, Logic, Semantics, and Meta-
mathematics: Papers from 1923 to 1938, pp. 152-278.
Tarski, Alfred. Logic, Semantics, and Metamathematics: Papers from
1923 to 1938. Trans. J.H. Woodger. Oxford: Clarendon Press, 1956.
Turner, Roy (ed.), Ethnomethodology, Selected Readings. Baltimore:
Penguin Books Inc., 1974.
van Heijenoort, Jean. From Frege to Godel: A Source Book in Mathe-
matical Logic, 1879-1931. Cambridge: Harvard University Press,
1967.
Whitehead, Alfred North and Russell, Bertrand. Principia Mathematica.
Cambridge:' Cambridge University Press. Vol. I, 1910 (second edn,
1925). Vol. II, 1912 (second edn, 1927). Vol. III, 1913 (second edn,
1927).
Wieder, D. Lawrence. 'Behavioristic Operationalism and the Life-World:
Chimpanzees and Chimpanzee Researches in Face-to-Face Inter-
action.' Sociological Inquiry 50 (1980): 75-103.
Wilder, Raymond. The Foundations of Mathematics. Second edn,
New York: John Wiley and Sons, 1965.

241