Reflections On Kant'S Critique of Pure Reason Penelhum Macintosh

the first
critique
REFLECTIONS PENELHUM &
ON KANT'S MACINTOSH
CRITIQUE OF PURE REASON
'\
THE FIRST CRITIQUE:
ReRections on Kant's
Critique of Pure Reason

WADSWORTH STUDIES IN
PHILOSOPHICAL CRITICISM
Alexander Sesonke, General Editor
HUMAN UNDERSTANDING:
Studies in the Philosophy of David Hume
META-MEDITATIONS:
Studies in Descartes
PLATO'S MENO:
Text and Criticism
PLATO'S REPUBLIC:
Interpretation and Criticism
LIMITS OF LIBERTY:
Studies of Mill's On Liberty
ARISTOTLE'S ETHICS:
Issues and Interpretations
BERKELEY'S PRINCIPLES OF HUMAN

KNOWLEDGE:
Critical Studies
HOBBES'S LEVIATHAN:
Interpretation and Criticism
MILL'S UTILITARIANISM:
Text and Criticism
THE FIRST CRITIQUE:

Reflections on Kant's Critique of Pure Reason
THE FIRST CRITIQUE:
Reflections on Kant's
Critique of Pure Reason
edited by
Terence Penelbum and J. J. MacIntosh
UNIVERSITY OF CALGARY
WADSWORTH PUBLISHING COMPANY, INC.

BELMONT, CALIFORNIA
© 1969 by Wadsworth Publishing Company, Inc., Belmont, California. All
rights reserved. No part of this book may be reproduced in any form, by
mimeograph or any other means, without permission in writing from the
publisher.
L. C. Cat. Card No.: 73-81083
Printed in the United States of America
WADSWORTH STUDIES IN
PHILOSOPHICAL CRITICISM
The idea of a series of Studies in Philosophical Criticism devel-

oped in response to a growing problem in American universities. Phi-
losophy can be taught most successfully in small classes; philosophical
understanding grows in the course of a dialogue where problems are
discussed from diverse points of view bymen who differ in experience
and temperament. But with the increase in college enrollments, the
size of introductory classes has grown larger and the possibility of a
dialogue between professor and students more remote. Our hope is
that the Studies in Philosophical Criticism will make a dialogue of
sorts possible in a class of a hundred, or a thousand, as well as in smaller
classes and seminars. Each volume in the series contains a collection of
critical writings related to a single classical philosophical text, such as
Descartes' Meditations or Plato's Republic. These critical writings are
not substitutes for the classical work, but supplements to it. They should
be read in conjunction with the classical text. So used, they will bring
to bear on the problems raised by Descartes, Hume, or Plato that diver-
sity of voices and viewpoints which is the heart of the dialogue-and
also, we hope, will prompt the student to add his voice to the discus-
slOn.
In selecting material for the volumes in the series, the editors have
not searched primarily for writings which provide a definitive analysis
of the classical text, but have rather selected those papers they thought
might be most useful in undergraduate courses in philosophy, both to
provoke students into serious engagement with the text and the prob-
lems found there, and to present them with a variety of philosophical
styles and idioms. Most of the writings reprinted are quite contempo-
rary; they were selected not only for their excellence but also as an
indication that many of the classical problems of philosophy persist as
centers of current controversy. We believe this format also achieves
one prime desid~ratum: it acquaints the student with both the great
works of the philosophical tradition and the most contemporary con-
cepts, techniques, and modes of thought. _ ~ "~~ n r-
CONTENTS
INTRODUCTION 1
Lewis White Beck

KANT'S STRATEGY 4
Richard Robinson
NECESSARY PROPOSITIONS 18
T. D. Weldon
KANT'S PERCEPTUAL VOCABULARY 34
J aakko Hintikka
ON KANT'S NOTION OF INTUITION
(ANSCHAUUNG) 38
Barry Stroud
TRANSCENDENTAL ARGUMENTS 54
W. H. Walsh
KANT ON THE PERCEPTION OF TIME 70
Lewis White Beck

THE SECOND ANALOGY AND THE PRINCIPLE
OF INDETERMINACY 89
S. Korner
ON THE KANTIAN FOUNDATION OF SCIENCE
AND MATHEMATICS 97
Jonathan Bennett
THE SIMPLICITY OF THE SOUL 109
Vlll Contents
Jerome Shaffer
EXISTENCE, PREDICATION, AND THE ONTO·
LOGICAL ARGUMENT 123
Peter Remnant
KANT AND THE COSMOLOGICAL ARGUMENT 143
SUGGESTED FURTHER READINGS 147

INTRODUCTION
Kant's Critique of Pure Reason, though one of the greatest philosophi-

cal writings, is notoriously a source of despair even to the most
assiduous student. Its difficulty is due not only to the fact that it breaks
so mu~l.! n~.:v groun9 but also to its obvious cief~cts of style_and
presenta.tjon::-most particularly to its t~<.:.h!lic;:.~lity_and to .Kanes unique
c~Pllsity for combining prolixity with extreme compression at impor-
tant place~ in_._his arguments. One result of this is an understandable
tendency for the student to feel satisfied with attaining a grasp of its
main doctrines and to be reluctant to debate them. But among con-
temporary philosophers an eagerness to debate Kant's doctrines is
clearly on the rise; and to encourage this interest, some of the essays
that are part of this debate are collected in this volume. Our collec-
tion is not intended to compete with the standard expository com-
mentaries or to add to the exegetical controversies which abound
within them; however, there is more discussion here about Kant's
idiom than the aim of this series has required in the case of other
philosophers.
The first two essays deal with the way in which Kant saw his
task. In "Kant's Strategy," Lewis White Beck shows how Kant ranges
himself against the two major philosophical traditions whose skeptical
and speculative excesses he wished to avoid. Beck emphasizes that to
defeat them Kant does not eclectically accept details which please him
on each side, but rather attacks the principle accepted by both tradi-
tions concerning the source of human knowledge. It is this strategy
that makes Kant the philosophical revolutionary he thought himself to
be. But in his own formulation of the problem, in the Introduction to
the Critique, Kant writes in terms of the unglamorous technicalities of
analytic and synthetic judgments. These terms have since found their
way into philosophical jargon, and their widespread use has inspired
negative reactions in recent years. Richard Robinson examines Kant's
use of the notion of necessity in propositions, and finding it a muddle,
makes stimulating suggestions for clarifying it.
In the Aesthetic and Analytic, Kant attempts to justify his claims
that the human understanding possesses knowledge of certain synthetic
n~~ssarytruths. The justification is presented largely in terI?s of a
complex theory about the mechanism of our perceptual expenence, a
1
2 The First Critique
theory which has often embarrassed his late~ adr:ure.rs. W ~ may wis~
to rid him of it, but there is no doubt that In hIS VIew this theory IS
essential to the argument. Our extract from T. D. Weldon's study of
Kant shows the risks run by the English reader in interpreting it. The
paper by Jaakko Hintikka concerns the theory of mathematics de-
veloped in the Aesthetic, with its puzzling doctrine of Space and Time
as pure intuitions which provide the mathematician with his subject-
matter and the sensory faculty with its form. In expounding the
changes which the key concept of intuition undergoes as Kant's theory
of mathematics develops, Hintikka is able to offer both an indication of
the difficulties in Kant's developed theory and an account of how a
description of mathematical knowledge in terms of the form of our
sensory faculty could have seemed plausible.
The heart of the Analytic is, of course, the Transcendental Deduc-
tion of the Categories, in which Kant attempts to demonstrate the
objective validity of his table of pure concepts of the understanding;
that is, that there must be a world of objects to which they apply.
Similar attempts in recent years to refute philosophical skepticism
concerning our knowledge of a world of objects have made it seem
that the essential form of Kant's proof could be presented without the
perceptual theory that accompanies it in the Critique. Arguments
embodying this essential form of proof are known as transcendental
arguments. They attempt to demonstrate that the validity of the con-
cepts being justified is a necessary condition of all intelligible experience
and, hence, even of a coherent statement of the skeptic's doubts. Barry
Stroud examines contemporary versions of such reasoning and suggests
that at most they establish that certain beliefs, rather than certain facts,
are needed for our experience to be intelligible. This paper raises dif-
ficulties for "the much-heralded 'Kantian' turn in recent philosophy"
and prompts reexamination of the possibility of separating readily palat-
able elements in Kantian thought from those we would like to disregard.
Critical examination of the Analytic of Principles has tended to
focus on the Analogies of Experience, where Kant attempts to demon-
strat:.. the _obj~ctive application of _~~_e.'"three categories, !lr pure con-
cepts, of relati.!l!!.' The transcendental argument here dra~vLheavily on
t!t~ .f"eci"iijreinenrs of our perception of objects and events in time.
W. H. Walsh's essay is a lucid and sympathetic discussion of Kant's
att~mpts to show the concepts of substance, cause, and reciprocity to
be indispensable. It has frequent! y been said that in trying to show this
Kant represents the particular requirements of Newtonian physics as
necessary conditions of all human thought. Lewis White Beck argues
Introduction 3
that the famous Indeterminacy Principle does not refute the Second
Analogy, which in fact has to be drawn upon in arguments designed to
establish that principle.
S. Korner draws a different conclusion about the implications of
changes in mathematics and physics since Kant's time. He notes that
Kant's justification of the objective validity of the categories rests on
the universal application of the mathematical and physical principles
that embody them and argues that the existence of alternative geom-
etries or mechanics forces us to change our picture of those very facts
about human thinking that Kant was attempting to justify. However,
mathematics and physics must make crucial use of some categories and
synthetic a priori principles, even though these need not be the ones
Kant thought.
The skeptic is the main philosophical enemy in the Analytic, but
in the Dialectic the main enemy is the rationalist metaphysician who
tries to erect a scientific system using a priori concepts that Kant
thinks have no application beyond experience. Literature available on
the Dialectic is less extensive than that on the Analytic-perhaps
because philosophers feel more necessity to refute the skeptic now
than the metaphysician; perhaps because of the common but question.
able belief that the Dialectic is easier; perhaps because its many anti·
speculative arguments do not depend strictly on the epistemology that
comes before them. This last fact, however, is itself a source of the
destructive power of Kant's criticisms: He is not content merely with
general reasons for expecting speculative metaphysics to be confused
and empty, but he exposes and analyzes the very confusions that his
epistemology enables him to predict. We have included, first, an essay
by Jonathan Bennett on Kant's treatment of the second Paralogism.
The Paralogisms are relatively neglected compared to the Antinomies,
and yet Kant's exposure of the pitfalls of rational psychology is of
topical interest in the philosophy of mind. Bennett's essay is also a
welcome foretaste of his expected work on the Dialectic. Our last two
selections deal with the refutations of natural theology. Jerome
Shaffer's paper is one of the best of many recent searching examina·
dons of Kant's famous adage that existence is not a real predicate. By
contrast with Kant's attempt to refute the Ontological Proof by his
dictum about existence, his argument that the Cosmological Proof
reduces to the Ontological has been too easily dismissed-perhaps
because St. Thomas, while using the one, officially rejects the other.
Peter Remnant's essay defends Kant against recent attempts to brush
this argument aside.
KANT'S STRATEGY*
Lewis White Beck
Great strategists are soldiers who win wars, or at least delay defeat, by
their genius in efficiently marshalling the forces at hand. Wars are not
won just by strategy, but they may be lost by it. For victory, good
strategy must be carried out by effective tactics; but good tactics may
be wasted by being used in the wrong place, at the wrong time, and
against the wrong enemy. While the great strategist will keep in mind
the tactical capabilities and weaknesses of his men, so as not to plan
battles that cannot be won in the field, the history of warfare recog-
nizes great strategists who did not win, not because their strategy was
wrong but because this or that tactical move was not successfully
made. Often we discern strategy best in a general who loses, for often
we can attribute to his victorious opponent such virtues as stubborn-
ness, single-mindedness, and courage, or ascribe his victory to such
gifts of fortune as superior numbers, better weapons, and good luck.
I wish to use these metaphors of strategy and tactics in an exposi-
tion of Kant's philosophizing. We know of Kant's keen and learned
interest in military matters. Though most of his metaphors which refer
to his work are drawn from law and the natural sciences and there are
few military metaphors, l is it too far-fetched to consider his work as
intellectual warfare, and thus to draw on a different set of metaphors?
In particular, I see Kant as engaged in a two-front war. Germany has
• Reprinted by permission of the author and publisher from Journal of the

History of Ideas (1967), pp. 224-236.
1 The ric~est. source of Kant's military me~aphors is his late polemic against
Schlosser, VerkundJgung des nahen Abschlusses emes Traktats z1Im ewigen Frieden
in der !,hilosopbie. (1796). In th~s he propo.sed his theory of freedom as a ground
on WhICh dogmatists and skeptIcs could sIgn a treaty of peace. Nothing in the
present essay is incompatible with that polemic, but I hope that I have come closer
to the historical and epistemological citadel of his philosophy than he did in this
Tendenzschritt. (Titl~s are given ~~ German when there is no English translation
of the work In questIOn.) A sensltlve and comprehensive study of Kant's meta-
p~ors, "The. Fabric of M~taphor in the Critique of Pure Reason," by David Tarbet,
WIll be published shortly In the Journal of the History of Philosophy.
4
Kant's Strategy 5
found, twice in this century, and Prussia found, in Kant's lifetime, that
to fight a two-front war it must use one army on two fronts and be
able to shift this army rapidly from one to the other. In Kant's two-
front war, his talent as a strategist was shown by his finding a
philosophical argument which could be used effectively against two
very differel1t opponents at the same time. I shall try to show the
pattern of this strategy by showing how Kant waged one battle before
he began writing the Critique of Pure Reason and then repeated its
plan later and on a larger scale. But before doing that, we must first
carefully identify the opponents against whom this strategy was to be
used.
At the very end of the Critique of Pure Reason, in the chapter
entitled "The History of Pure Reason," Kant sets forth three great
perennial divisions in metaphysics as the theory of the scope and
function of reason. He gives three dichotomies: between intellectual-
ists and sensualists in regard to the object of knowledge, between
empiricists and noologists (rationalists) in regard to the origin of
knowledge, and between naturalists and scientists (users of the scien-
tifische Methode, i.e., systematic, "scholastic" philosophers) in regard
to the methods of knowledge. Among the latter, Kant distinguishes
two types: those who proceed dogmatically, like Wolff, and those who
proceed skeptically, like Hume.
These three ways of dividing possible philosophies are logically
independent of each other, but in fact we find certain family affilia-
tions among some of them. The great intellectualists have been noolo-
gists and dogmatists; the great sensualists have been empiricists and
either naturalists or skeptics. There are, then, two great coalitions
opposed to each other, not only in Kant's time but again and again
since Socrates and the Sophists met in battle.
Let us see what were the elements in the two coalitions which
were opposed to each other. We may think of Kant as standing
isolated between them, looking for future allies in both camps (though
of course this is not the whole historical truth because Kant himself
was at one time a member in good standing of one of the alliances and,
in the later sixties, seems to many historians of philosophy simply to
have changed sides for a time). But the mature Kant was never entirely
uncritic;l of the rationalists, and never fully committed to a skepticism
based on empiricism. So let us imagine him as standing between,
unwilling to commit himself to either; this was, . perhaps, his own
conception of his position as when he stirs up trouble between the
opposing forces, entices them into battle with each other so that they
will destroy each other. 2
We shall examine first the composition of the entente founded in
modern times by Locke. There were three principal allied powers
here, all agreeing on one point: that all we know we learn from
experience. These allied powers were: skepticism in metaphysics,
naturalism in ethics, and something I shall call for lack of a better name
skepticism tempered with naturalism in the theory of knowledge.
Hume is the best example of all three. Kant believed that, though
Hume was not a consistent skeptic, he was saved from Pyrrhonism
only by his good sense and by a fortunate error he made in the estima-
tion of mathematical knowledge. But Kant believed that a consistently
developed empiricism could lead to skepticism not only in meta-
physics, which is as far as he thought Hume pushed it, but also in our
natural and mathematical knowledge. Since, as Hume3 put it and Kant
believed, "Nature is always too strong for principle," such a skepticism
in metaphysics was quarantined by a recognition that what little
knowledge we have suffices for the ordinary affairs of life. As Locke
himself saw, "The candle that is set up in us shines bright enough for
all our purposes," but does not cast a light into the regions into which
we can never venture. Naturalism in ethics, a kind of empiricism in
regard to our knowledge of right and wrong and a eudaemonism or
hedonism in the definition of good and bad, has regularly been
associated with metaphysically skeptical but this-worldly naturalistic
empiricism. Kant did not equally oppose all the members of this
coalition of ideas. He came to share the empiricists' skepticism of
metaphysics but always rejected the naturalism of healthy common
sense, which he called "mere misology reduced to principle."4 And he
always opposed the naturalistic ethical theory even when praising
Shaftesbury, Hutcheson, and Hume for their "beautiful discoveries" in
the method of ethical inquiry. 5
Kant's strategic question in dealing with this coalition must have
1
been: how could he maintain skepticism in metaphysics-to which he
was pushed by his study of Burne and his own discovery of the
2 Critique of Pure Reason, A 422-3 = B 450-1. The same thought with a legal
rather than a military metaphor occurs at A 530 = B 558. Hume uses the same
imagery in The Natural History of Religion, last paragraph.
3 Hume, Enquiry Concerning Human Understanding, Sect. XII Part ii end
4 Critique of Pure Reason, A 855 = B 883. " .
• I) Nachricht vo"! .der Hinrichtung seiner Vorlesungen, Gesammelte Schriften,
Prusslan Academy edItIon, II, 311-2. (Henceforth all page references in parentheses
are to this edition.)
Kant's Strategy 7
antinomies which he called "the most fortunate perplexity into which
pure reason has ever fallen"6-without falling victim to eudaemonism
in ethics and to a jejune appeal to common sense in the conduct of life
and the development of science? How could he oppose Burne without
falling in with Reid, Beattie, and Oswald, who he thought were very
uninspiring company? How could he give up a supernatural meta-
physics without making a metaphysics out of naturalism?
Let us now look at the opposing coalition of the rationalists. Here
we find dogmatism in metaphysics, to which Kant adhered in the
fifties. Opposed to empiricism was the Leibnizian epistemology which
proceeded to solve even the simplest problems by an argument ob-
scurum per obscurius. It was an epistemology which explained the
simplest facts learned through sense experience by an appeal to pre-
established harmony, and the theorems of physics by appeal to the-
odicy. In ethical theory, there was a Scotism making the good depend
upon God-another instance of getting the cart before the horse-and
a theory of freedom which Kant called a "wretched subterfuge"
grounding nothing more than the freedom of a turnspit or marion-
ette. 7 Again we find that Kant was not equally inimical to all these
opponents; as I have said, early in life he espoused its dogmatism, and
in the sixties he accepted the "wretched subterfuge" in ethics he was
later to condemn. Hence his strategic position must have been some-
thing like this. Having turned his back upon dogmatism in metaphysics
(after 1770), how could he save an ethical theory which required a
metaphysical rather than a quasi-physical or hyperphysical founda-
tion? How could he give up metaphysics because it was not empirical?
How could he defend our knowledge of nature by giving a theory of
synthetic judgments known a priori, without defending metaphysics
which consisted of nothing but synthetic judgments known a priori?)
Had Kant's strategy been that of divide et imp era, he would have
been an eclectic philosopher drawing a bit from here and a bit from
there. His philosophy would itself have been a coalition system of the
kind he explicitly condemned,S and he would have been as forgotten
now as other eclectics and compromisers.
I have been speaking in very broad terms of Kant's attitudes
6 Critique of Practical Reason, Beck translation (New York, 1956), 111 (V,
107). (All page numbers to the second Critique are to this translation, followed by
Academy pagination in parentheses.)
7 lbtd., 99, 101 (IV, 96, 97).
S Critique of Practical Reason, § 3, p. 23 (V, 24).
towards a coalition of views into two different Weltanschauungen.
Kant seldom talked on this synoptic level; he was generally concerned
with specific philosophical doctrines, whatever their source. But we
have a foretaste of his grand strategy in the tactics of a particular battle
he waged in the late sixties and early seventies, so we shall examine this
before considering the later and larger campaign. I refer to the disputes
concerning the nature of space and geometry.
It might be thought that empiricism could have no tenable theory
of mathematics. If all the material of knowledge comes from expen-
ence, all universal statements are inductive and only probable. But
since universal mathematical statements are not merely probable, the
thesis of empiricism must be false. The empiricist had two ways of
dealing with this syllogism. First, he could, with Hume in the Treatise,9
deny the minor premise and make geometry an empirical science
which only approximately fits the observed facts of nature. Or, second,
he might deny, with Hume in the InquirylO that mathematics is
knowledge of nature and assert that it is only the logical manipulation
of symbols. Kant suggests that Hume took the latter point of view in
order to keep from giving up mathematical certainty,l1 apparently not
knowing that Hume had once committed himself to the former
explicitly skeptical conclusion. But given a Newtonian physics and
theory of space, the second of Hume's views is in practice as skeptical
as the first, because there is no way to justify the application of a
merely logically necessary geometry to the space of physics, and, with
very few exceptions, the XVIIIth century had not yet distinguished
9 Treatise of Human Nature, I, Part III Sect. i: "The reason why I impute
any defect to geometry is, because its original and fundamental principles are
derived merely from appearances; and it may perhaps be imagin'd, that this defect
must always attend it, and keep it from ever reaching a greater exactness in the
comparison of objects or ideas, than what our eye or imagination alone is able to
attend" (Selby-Bigge, 71).
.. lO./nquiry. C~~cerning Human Und~rstanding, the first paragraph in Sect.
Vll, and III Sect. IV: If any term be defined III geometry, the mind readily of itself
s?bstitutes <?n a~l occasions, ~he definiti~n of t?e term defined ..." and ~'Proposi~
nons <?f thIS kind [geo~etrlcal and arithmetIcal] are discoverable by the mere
operatIon of thought, wIthout dependence on what is anywhere existent in the uni-
verse. Though there never were a circle or triangle in nature, the truths demon-
strated by Euclid would for ever retain their certainty and evidence" (Yalden-
Tho~son, 61, 24) ...But when H';lme returns to the subject matter of the Treatise in
~ectIon XII Par.t 11 of the Inquzry (":alden-Thomson, 162-64), i.e., to the applica-
tIon of ge.ometrlcal concepts to experl~nce, he does so for the purpose of pointing
out the dIscrepancy between geometrical concepts, as commonly (and he believes
erroneously) understood, and our spatial perceptions, and thus repeats his first
answer.
11 Critique of Practical Reason, 13 (V, 13).
Kant's Strategy 9
between interpreted and uninterpreted axiom systems and decided that
mathematics belonged among the latter. 12
Leibniz, on the other hand, never played with an empiricistic
theory of mathematics, but consistently developed the formalistic
theory like that of Hume's later work, and his theory therefore
suffered the same infirmity as that in Hume's Inquiry-it was hard to
see why a system of analytic judgments should have any objective
reference to the real disposition of things in space. But whereas Hume
could have no theory of how this is possible and, in all probability,
thought (as he did in the Treatise) that it was not possible, Leibniz did
try to show that the propositions of mathematics are necessarily true
of space. He did so through two hypotheses. (1) Our perceptions of
the physical world are only confused conceptions: hence clear and
distinct sense perceptions, of the kind we get through counting and
mensuration, must conform to, if they are indeed not identical with,
clear and distinct conceptions of numbers and magnitudes. (2) But
what we perceive as spatial is not really outside us in an absolute
Newtonian space which we could know, if at all, only a posteriori.
Rather, space is a pbenomenon bene fundatum, a phenomenon well-
founded in the logical relations holding between substances (monads)
or states of substances. Geometry is founded on logic; it is a logic
which can be mapped spatially as a representation of simultaneously
existing incompatibles; space is simply the order in which we perceive
compresent possibilities. There is, therefore, no problem as to why the
logical relations of concepts fit the intellectual representations we have
of space; in a sense, the former generate the latter. 13
If we compare the outcome of Hume's Inquiry with Leibniz's
theory of mathematics, we find that these two philosophers agreed on
one point and disagreed on another. (a) They agreed, to use Kant's
12 See E. W. Beth, The Foundations of Mathematics (Amsterdam, 1959) on

efforts to draw this distinction in the XVlIIth century. We may say that Hume's
inconsistency, documented in notes 9 and 10 supra, arises because in the Treatise
he is concerned with applied geometry and in the Inquiry with pure geometry. But
inasmuch a:; the distinction did not clearly exist, each is mixed up with the other
in both the books. Kant recognized/the distinction (Critique of Pure Reason, B 145;
d. also Thought on the True Estimation of Living Forces, § 115) but did not see
how important-perhaps how fatal-it was for his entire position.
13 See Kant's very clear statement of Leibniz's problem and solution, Critique
of Pure Reason, A 275-6 = B 331-2. The flaw in Leibniz's argument is that Leibniz
does not distinguish between the sensible appearances and things in themselves, and
fails to do so because he makes sensible appearances only confused intellectual
representations of things in themselves (substances). But, according to Kant, they
are not; hence the problem of the relation of mathematics to the perceptual content
is left untouched.
terminology, that mathematical judgments are analytic and logically
necessary. (b) They disagreed as to whether mathematics had any
necessary objective reference-Leibniz affirming it and Hume deny-
ing it.
In 1768 Kant . . . formulated the hypothesis that the thesis (a)
was false: 14 Mathematical knowledge is not analytic and logically
necessary. But it is necessary, in the sense that it necessarily applies to
our experience of objects in space (here Kant agrees with Leibniz),
but it does not apply, even approximately, to anything but objects of
experience (and here Kant agrees with the earlier Humean view).
When Kant formulated this view by saying that mathematics
contains synthetic judgments, he is disagreeing with both Hume and
Leibniz. 15 His answer to it leads directly into the great campaign I
have spoken of. Our preliminary study of Kant's treatment of mathe-
matical knowledge from 1768 to 1770 points the way to the larger issue
in two respects. First, the answer to the question, "How are synthetic
judgments a priori possible?" or "How is pure mathematics possible?"
gives the substantive thesis of Critique of Pure Reason, to which I shall
tum shortly. But more to my purpose here is the fact that Kant's
strategy in this criticism of Hume and Leibniz is in form and pattern
exactly the same as his larger strategy in the Critique of Pure Reason as
a whole.
Kant did not tell us his strategic secrets, and perhaps he was not
fully aware of his stratagem. But in our own century, this stratagem
has been formulated in what is sometimes called "Ramsey's Maxim." In
cases where two opposed arguments seem internally sound but where
their conclusions are incompatible and hence a stalemate is created,
14 On tbe First Grounds of tbe Distinction of Regions of Space. The argu-

ment is that logically indiscernible entities should be identical, but they are not, as
shown by incongruent counterparts in geometry; therefore geometrical figures are
not adequately defined by non-intuitive predicates or relations.
15 It is often said that Kant's initial discovery (or claim) from which every-
thing else followed was that there are synthetic judgments known a priori. This is
not true. Both Locke and Descartes had admitted judgments which, by the later
Kantian criteria, were synthetic and known a priori. It was to explain their apriority
that Leibniz and Hume had denied their syntheticity and replaced them with
verites de raison and relations of ideas, respectively. But what forced Leibniz and
Hume to do so (Hume much more consistently than Leibniz) was the assumption
by both Descartes and Locke that we are endowed with an intellectual intuition
and that all sense experience is a posteriori. Kant rejected both these assumptions,
while Hume denied only the first. Only Kant's denial of the latter permitted him to
go beyond Hume ~nd to support the thesis that we have synthetic a priori sensible
knowledge. (In this paper, however, I am more concerned with the contribution
of sense to the syntheticity of knowledge than to its apriority: but I believe Kant's
underlying strategy could be illustrated also in his theory of the latter.)
Kant's Strategy 11
Frank P. Ramsey16 wrote: "It is a heuristic maxim that the truth lies
not. in one of the two disputed views but in some third possibility
W?IC~ has not y.et been thought of, which we can only discover by
reJec~mg somethmg assumed as obvious by both the disputants." Two
theones, X and non-X, may be reconciled or both refuted by finding
that they have a common false element. Upon analysis, X may be
found to be A + Y and non-X may be found to be A + non-V; and
A may be found to be false. When the falsity of A is seen, a new
theory can be developed without it. Though Y and non-Yare still
contrad~ct?ry, just as X and non-X are still contradictory, these
contradIctIOns no longer matter; they are left behind in philosophical
debate, because what made them seem frustratingly important was that
they seemed to be the only possible corollaries of A, and now A itself
has been given up.
We have seen how Kant applied this maxim in the disputes
concerning mathematics. H ume and Leibniz disputed whether mathe-
matics applied to experience; and each could give excellent reasons for
denying and asserting it, respectively. But they both agreed that
mathematical judgments were logically necessary. This agreed-upon
principle was, according to Kant, false.
We can illustrate this pattern again by referring to Kant's later
attempt to resolve the space antimonyP Rationalists said space was
finite; empiricists said it was infinite. Equally good proofs existed on
both sides. The statements are contradictory, but Kant said they were
both false. What he should have said is that both were statements in
compound judgments, "Space is the real form of objects existing
intrinsically, and it is finite," and "Space is the real form of objects
existing intrinsically, and it is infinite," and both these compound
judgments are false because the first conjunct is false. If we deny the
first conjunct, then the dispute about the second no longer matters
since the second conjuncts are about something which does not exist at
all.
Other illustrations of this strategy could be given. But let us now
look at the over-all strategic situation. Can Kant find some principle
accepted by both coalitions but nevertheless false? If so, rejecting that
common principle will defeat both parties simultaneously, or at least
break up by their internal affiliations and establish a new center of
power in Kant's own counter-thesis which will attract allies from both
16 Frank P. Ramsey, The Foundations of Mathematics (London, 1931), 115-6.

17 See the very relevant paper of J. E. Llewelyn, "Dialectical and Analytical
Opposites," Kant-Studien, LV (1964), 171-4.
12 T he First Critique
coalitions. Kant's attempt to discover this common, but false, funda-

mental principle is his strategy in fighting his two-front war; but
unlike most two-front wars, here a victory on either front will be a
victory on both. Perhaps we should even change our metaphor, and
say that Kant's strategy is to show that a single position which is
essential to both sets of opponents is untenable.
The common principle Kant thought he found in both and
thought he could show to be false is: There is but one ultimate source
or faculty of knowledge. The point at issue between the Lockeans and
the Leibnizians was: What is this single source of knowledge? That
there was such a source was the unexamined dogma of both, and if
Kant could show it to be false he would have broken up both co-
alitions.
Leibniz, he tells us, intellectualized appearances while Locke
sensualized all the concepts of intellect. 18 For Leibniz and Wolff,
increasing the distinctness of a representation raised it from the level of
feeling and sensation to that of thought; sensation and feeling are
confused thought. The de facto synthetic connections discovered
empirically between representations are to be replaced by logical
connections between well-defined concepts resulting from their analy-
sis. Synthetic empirical knowledge is not knowledge but only a pre-
stage to rational knowledge of things as they must be. We human
beings have to be satisfied with experience, but it is a poor substitute
for rational knowledge of necessary connections. Empirical episte-
mology as a theory of scientific methodology is at best only an Inter-
imserkenntnistheorie.
Locke was too much influenced by Descartes to deny in prin-
ciple, and too much of a man of good sense to deny in practice, that
there is a difference between the experience of seeing the compatibility
of two ideas and that of seeing a sequence of logically unrelated ideas.
He never formulated one theory that would account for both; but
both are there. The differences between him and Leibniz are not as
pointed as the rubrics of empiricism versus rationalism make them
appear. Even Hume recognized diverse sources for mathematical
knowledge and for knowledge of matter of fact; he still sees the
difference between thinking and perceiving and, in fact, he insists that
there is a gulf between them. How, then, can Kant say Locke sensual-
ized intellectual concepts?
The clue is again found in Hume's theory of mathematics in the
18 Critique of Pure Reason A 273 = B 327.

Kant's Strategy 13
Inquiry. Precisely because mathematical knowledge is knowledge of
the ~e!a:ion of ideas, and thus does not fall under the general rule of
empIrICISm that our knowledge comes from experience, Hume saw
that mathematics has no existential import. By completely intellectual-
izing our mathematical knowledge, Hume cut it off from reaching the
full fruition of knowledge, which is to be knowledge about existing
things. The connection with objects, which would be necessary if
mathematics were to be full knowledge,19 was broken by making
mathematics logically analytic. Therefore the intellect that had to do
with real existence, that is, the intellect that gives knowledge, must be
an intellect which has been sensualized. It was this sensualized intellect
which Kant thought produced the skepticism inherent in empiricism,
just as Leibniz's intellectualized senses could produce only dogmatism.
As early as 1762 Kant saw the differences between the methods
of mathematics and those of metaphysics. By 1768 he saw the differ-
ence between mathematics and logic. Certainly in 1770 he had a theory
of two cognitive faculties, the sensibility and the intellect, and of two
different worlds, the sensible and the intelligible. No matter how much
clarity we might achieve in our analyses of the content of the empiri-
calor sensible world, we could never raise it to the level of the logi-
cally necessary and metaphysically evident; indeed, we have, as a
matter of fact, far more clear and distinct ideas of the sensible world
than we have of the intelligible. In metaphysics, we have only a few
clear and distinct ideas; but our obscure metaphysical notions are not
thereby rendered sensible to us. 20
The Inaugural Dissertation of 1770 represents Kant's first stra-
tegic coup: he can save geometry from skepticism, by showing how it
can apply to objects a priori, and he can go on with the dogmatic
metaphysics he derived from his rationalist teachers. He thinks that the
troubles of the Leibnizians in their metaphysics arose from their
mixing concepts which applied only to the senses with those which
were produced by pure reason. By a clear separation of the spheres of
the two cognitive faculties, therefore, he could save metaphysical
dogmatism and, almost as a by-product, avoid skepticism in mathe-
matics. Now this may appear to have been no victory for him at all on
a strategic level. For Kant seems to have attacked both Leibniz and
19 Ibid., B 147.
20On the Form and Principles of the Sensible and Intelligible Worlds ("In-
augural Dissertation"), §§ 7, 8: On the Distinctness of the Principles of Natural
Theology and Morals ("Prize Essay") I, § 3 (II, 280).
Hume on their mathematical theories, but to have drawn a set of
conclusions entirely acceptable to the rationalists, viz., the apriority of
mathematics and the possibility of an a priori metaphysics. The letter
to Marcus Herz of February 1772 is the last ma~ifesto of his rational-
ism, however, because his "recollection of Hume" soon thereafter
awoke him from his dogmatic slumber.
We often speak of what followed as "Kant's reply to Hume," but
Kant is ~eplying to both Leibniz and Hume. His question was: How
not to be a dogmatist in metaphysics without being a skeptic in our
knowledge of nature. Hume's skepticism was all of one piece: no
objective necessary knowledge of matter or fact either in or beyond
experience. Leibniz's dogmatism was all of one piece: a priori knowl-
edge of both what is in and what is beyond experience. Kant wanted to
break these two continuities; and he saw that each was based on a
theory of one source and one kind of knowledge.
We know little of the history of what went on in Kant's mind
between February 1772 and April 1781. But it is clear that two things
did happen. One was that Kant, having settled to his own satisfaction
the problem of how mathematics is possible, turned his attention to the
conditions of our knowledge of existing objects and found that we can
know only those given both to thought and sense. Second, Kant con-
tinued to think about the book on the "metaphysics of morals" which
he had been planning to write for twenty years. 21 We shall see now
how his strategy evolved so as to solve both questions at once: how to
save the rational features of science from Hume's attack and the
irreducibly empirical features from Leibniz's and Wolff's; and how to
save the conception of a metaphysics of morals from Hume's natural-
ism and empiricism and from Leibniz's and Wolff's dogmatism which
did, as he said, "war against it. "22 To show this, I shall quote two
passages from the Critique of Pure Reason which are absolutely
essential to his total philosophy, and I shall show how they are direct
consequences of his challenge to the principle agreed upon by all his
opponents.
The first is: "Thoughts without content are empty, intuitions
without concepts are blind. It is just as necessary to make our concepts
sensible, that is, to add the object to them in intuition, as to make our
intuitions intelligible, that is to bring them under concepts. "23 The
21 See my Commentary on Kant's Critique of Practical Reason (Chicago,

1960),5-10.
22 Critique of Pure Reason :a xxx.
23 Ibid., A 51 = B 75.
Kant's Strategy 15
reason for this statement is found in Kant's distinction between
analytic and synthetic judgments. Analytic judgments relate concepts
to each other by finding one contained in the intension of the other;
synthetic judgments are syntheses of concepts which are held together
by their common reference to something given, which Kant calls X.24
In a mathematical judgment, this X is a pure intuition or construction
of space; in perceptual knowledge, it is a phenomenal object which is
given to me in a set of intuitions which are related by a rule I follow in
determining the order in which I entertain them, a rule so formed that
they will conform to a rule of judgment in logic. But when I try to
make a synthetic judgment about something not given in intuition, I
find that I can only relate my concepts analytically and not bring them
into any. relation to an object; there is no intuitive X which is, as it
were, the glue to hold the concepts together either a priori or a
posteriori. If, however, there were only one source of knowledge,
which Kant calls an intellectual intuition or intuitive intellect, then the
act of thinking an object would lead to its representation in me-and
dogmatic metaphysics would be possible.
Notice Kant's strategy here. There are two factors involved in
knowing: sensibility and understanding. Neither alone can give us
knowledge; either alone is blind or empty. (Empiricism is blind;25
rationalism is empty.) Knowledge comes from the application of one
to the other. Dogmatism is the policy of claiming rational knowledge
beyond what can be perceived; rationalism is inherently dogmatic; it
can best be dogmatic precisely where there is in principle no perceptual
source or test for its claims. So that what makes an answer to Bume
possible-the rules of relating representations to each other introduce a
synthetic element a priori into our empirical knowledge-also makes
an answer to Leibniz possible: there is a perceptual or intuitional
element in all a priori knowledge that is not merely and emptily
logical.
The second sentence is: "I have, therefore, denied knowledge in
order to make room for faith."26 This well-known sentence is the
foundation of Kant's ethical theory, because it makes it possible for
him to accept the rationalistic thesis of the Third Antinomy without
taking also the dogmatism which universally attended it. The thesis of
the antinomy states that there is a "causality of freedom" and the
24 Ibid., A 9 = B 13.
25 "Empiricism is based on touch, but rationalism on a necessity which can
be seen"-Critique of Practical Reason, 14 (V, 14).
26 Critique of Pure Reason, B xxx.
antithesis is that there is only a "causality of nature." Both of these

statements may be true if the common presupposition is that the
sensible world in space and time is the only world to which our
concepts apply. But there is no evidence whatsoever that the thesis is
true; all that Kant says he has shown is that it is thinkable, i.e., not self-
contradictory, and that it is not even contradictory to the principle of
the mechanism of nature if the common presupposition is false. Now
this common presupposition is precisely what was challenged by
Kant's solution to the Hume-Leibniz controversy over knowledge; it is
the presupposition that there is one world of actual objects to be
known in only one way (other ways being conveniences or customs of
the limited human knower). Had that controversy not been solved as
it was, Kant tells US,27 nothing could have saved the causality of
freedom, nothing could have prevented the export into metaphysics of
the principle of sufficient reason found essential to physics.
Unless a theory that sharply divides physics from metaphysics
can be established, metaphysics can only be an extension of physics and
the only metaphysical causality will be mechanical. The theory of the
sharp division is the theory that while metaphysics uses only one
cognitive faculty and hence is unable t9 give theoretical knowledge,
physics, which does give us knowledge, requires two independent
sources, viz., sensibility and reason. Here again we have the negative
test on what Kant says about an intellect that would be unitary in its
sources of knowledge. If we had an intuitive intellect, the antinomy
would not arise,28 but the antinomy is the most fortunate perplexity
into which pure reason could ever have fallen.
I have said that Kant's strategy is to break up the coalitions led by
empiricism and rationalism by finding and denying the axiom they
agreed upon. Bergson wrote that every great philosopher has said only
one thing,29 and James remarked that any worthwhile system of
philosophy can be written on a post card. It is an amusing jeu d'esprit
to take a philosopher's ten or twenty volumes and try to compress
27 Critiq1!e of Pract!cal Reason, 98 (V, 95); Critique of Pure Reason, B xxix.

!~e Kemp SmIth transl~tIon of the ldtter sentence is unclear. The sense is: "Since
It IS only on. th.e assumptIon of fr~edom that its negation [i.e., mechanism] contains
a~y contradICtIOn, wh!le ~he dental of mechanism contains an obvious contradic-
tIon, freedom, and wI~h It morality, ,,:"ould have to yield to the mechanism of
~1ature [on. t~e assumptIOn .that speculatIve reason does not permit freedom except
In contradIctIOn to mechamsm]."
28 Critique, of Judgment § 77; Critique of Practical Reason, 103 (V, 100).

29 La pensee et Ie mouvant (Paris, 1934), 141.
Kant's Strategy 17
them to post card length. My proposal for doing this to Kant's will be
disappointing, since hardly anyone nowadays will deny the sentence
but many will deny that it is the seminal thought in Kant. But it was a
highly disputable proposition in his day, and I think that some of tl:e
lasting importance of Kant is shown by the fact that it is no longer
disputed. The sentence would be: In order to know and to act, it is
necessary both to see and to think.
I hesitate, of course, to say that this insipid statement is the sum
and substance of Kant's philosophy. But when I see how much of his
philosophy depends on it, how much is an elaboration and defense of
it, and how many of his polemics are against those whose philosophy
was an implicit denial of it, I think there is merit in this as a summary
if one insists upon post card brevity in the history of philosophy.
I have expressed the opinion elsewhere 30 that Kant is too com-
plex a philosopher to be pressed into any single mold, whether that
mold be made in Marburg, Heidelberg, Gottingen, or elsewhere. The
logistic, metaphysical, positivistic, existential, psychologistic, and axio-
logical interpretations of Kant all have the merit, to be sure, of
bringing to the fore what is at most background in the others. It is
natural to use any great philosopher of the past as either an arsenal or
as a target in later philosophical conflicts, and thus Kant has been
regarded at one time as a founder of positivism, at another as its
opponent; as the destroyer of metaphysics, and as the philosopher who
made it "scientific"; as the chief critic of psychologism, and as the man
who ruined epistemology by making it psychologistic. 31 It is not
likely, therefore, that Kant will appear as a living force in contempo-
rary philosophy if we see him primarily as the arbiter between an
empiricism and a rationalism that are themselves no longer living
movements in philosophy. Yet if we try to see Kant in his own time
and try to give full weight to his explicit statements about Hume,
Leibniz, and Wolff, we return to the classica}32 picture of Kant which
I have been outlining.
30 Studies in the Philosophy of Kant (New York, 1965),52 n.

31 See Max Wundt, "Wandlungen des Kantbildes in der deutschen Geistes-
geschichte," Universitas (Stuttgart), 14 (1959), 51-58; Wolfgang Ritzel, Studien
zum Wandel der Kantauffassung (Meisenheim/Glan, 1952).
32 I call it "classical" because Hegel, the true founder of the history of philos-
ophy, so regarded Kant (Lectures on the History of Philosophy, iii, 429); but it
was, until very recently, so naively reproduced that it could equally well be called
the "textbook picture."
NECESSARY PROPOSITIONS*
Richard Robinson
~
1. NECESSARY
What is a 'necessary truth', and what is a 'necessary proposition'?

There was a symposium on necessary truth in 1947. The first and
second symposiasts assumed without comment that every philosopher
knows what is meant by the phrase 'necessary truth'. The third
symposiast, Mr. Kneale, commented as follows: 'Admittedly the phrase
"necessary truth" belongs to philosophers' jargon rather than to com-
mon speech, but its origin is clear enough. Philosophers say that a
proposition is a necessary truth if it is impossible that it should not be
true, i.e. if there is no possible alternative.' His discussion consisted
largely in arguing that the other symposiasts had lent themselves to a
doctrine about necessary truth which contradicts the known and
agreed meaning of the phrase 'necessary "truth'.
I maintain, on the contrary, that neither the origin nor the
meaning of the phrase 'necessary truth' is clear enough; and that
different philosophers have used the phrase in different senses and have
given different accounts of what it means.
The phrases 'necessary truth' and 'necessary proposition' do not
have one single inevitable meaning by the meanings of their com-
ponent words and the way in which these components are here put
together. On the contrary, the rules for 'necessary' and for 'proposi-
tion', and the way in which the two words are here combined, leave
open an indefinite number of possible meanings for the phrase, an
indefinite number of meanings which the intelligent layman might
assign to it, so far as he was not prevented from doing so by the
context in which he found it.
(1) In the first place, he might quite reasonably guess that a
• Reprinted by permission of the author and publisher from Mind (1958)

pp.289-304. '
18
Necessary Pro positions 19
necessary proposition was a proposition which it is necessary for us
men to believe, or for some of us to believe. A proposition could be
necessary for us to believe because it had not occurred to us to doubt
it, or because it seemed to us obviously true, or because something
compelled us to believe it, or for other causes. Or it could be necessary
for us to believe if a certain purpose was to be fulfilled. This meaning
is suggested by the delightful example which Kneale quoted: 'Who-
soever will be saved: before all things it is necessary that he hold the
Catholick Faith'. The statement of the Catholick Faith is a 'necessary
proposition' in the sense that it is necessary for you to believe this
proposition if you are to be saved. I will call this the 'compulsory-
belief' sense of the phrase 'necessary proposition'. .
(2) The origin of the expression 'necessary proposition' appears
to be Aristotle's expression 'avaYKa{a 7rpcham ..' in his Prior Analytics, i,
8-12. Aristotle makes unambiguously clear what he means by the ex-
pression there. He means by an 'avaYKa{a 7rpOTam..' a premiss or proposi-
tion which contains the expression 'i~ avaYK,.,..', ie. which says that A
'necessarily' belongs (or does not belong) to B. In English we may say
that, roughly, a proposition is necessary in this sense if it is of the form
'S must be P', or'S cannot be P', or'S is necessarily P'. A proposition
cannot be necessary in this sense unless it contains 'must' or 'cannot' or
'necessarily' or some equivalent word. The proposition that 'A father
is a parent' is not necessary in this sense. To use the technical language
of later logicians, Aristotle's 'necessary proposition' is a modal proposi-
tion of the apodeictic kind. I will call this the 'Aristotelian' or the
'apodeictic' sense of the phrase 'necessary proposition'.
These first two kinds of necessary proposition are independent of
each other. An Aristotelian necessary proposition mayor may not be a
compulsory-belief proposition; and a compulsory-belief proposition
mayor may not be an Aristotelian necessary proposition.
(3) Leibniz wrote: 'A truth is necessary when the opposite
implies contradiction, and when it is not necessary it is called con-
tingent' (Letter to Coste, 1707; p. 480 of Wiener's Selections). 'Oppo-
site' is a vague word; but it is clear here that Leibniz meant that a truth
is necessary when the denial of it implies contradiction. There are
some truths to deny which is to contradict oneself, either explicitly or
implicitly. That is, there are some truths whose contradictory is self-
contradictory. Out of this definition of 'necessa~~ t~uth' we ca~ .mak.e
the following definition of a 'necessary proposltlon: a proposttton 1S
necessary if either itself or its contradictory is selfcontradictory . I will
call this the 'analytic' kind of necessary proposition, because Leibniz
remarked that 'when a truth is necessary, its reason can be found by
analy sis, resolving it into more simple ideas and truths, until we come
to those which are primary' (Monadology, 33). I will also call it the
'Leibnizian' kind of necessary proposition.
This third kind of necessary proposition is independent of each
of the previous kinds. That is, an analytic proposition mayor may not
be a compulsory-belief proposition, and it mayor may not be an
apodeictic proposition. To assert a necessity, e.g. 'Men must die', is not
the same thing as to have a selfcontradictory contradictory. Nor is the
fact that I must believe a statement the same fact as the statement's
having a selfcontradictory contradictory.
This third kind of necessary proposition differs from each of the
others in that its contradictory is also necessarily a necessary propo-
sition.
(4) I have a reason, which will appear later, for saying that some
people mean by a 'necessary proposition' a proposition which asserts a
universal connection with unrestricted generality. In this sense propo-
sitions of the following forms are necessary: 'Absolutely any A is B';
'No A whatever is B'; 'If anything were A it would also be B'; and just
'All A is B' when this is understood to be unrestrictedly general. In this
sense it is a necessary proposition that 'All bodies are heavy'. In this
sense the contradictory of a necessary proposition is not a necessary
proposition. I will call this the 'universal' sense of 'necessary propo-
sition'.
This fourth sense is also independent of all previous senses. A
universal-necessary proposition mayor may not also be a compulsory-
belief necessary proposition. It mayor may not also be an Aristotelian
necessary p~~position. It mayor may not also be a Leibnizian neces-
sary proposItlon.
How did Kant understand the phrase 'necessary proposition'?
For example, what did he mean by 'ein notwendiger Satz' in Kritik der
reinen Vernunft (B 3)? He offers no explicit definition, so far as I have
observed.
There is some reason for saying that he understood it in sense 1,
as a proposition which we are compelled to believe, owing to the
nature of the human mind. Yet on reflection it seems that to say this
would be to substitute Kant's conclusion for his definition. Kant
concludes, more or less, that we are compelled to believe certain
propositions; but he concludes this precisely from the premiss that
those propositions are necessary. He is here not making the uninterest-
ing inference that we are compelled to believe these propositions
Necessary Propositions 21
be~au.se they are pr_opositions which we are compelled to believe. He is
thmkmg that we are compelled to believe them because they are
necessary in some other sense. It seems, therefore, that Kant's 'neces-
s~~y proposition' is not our first sense, the compulsory-belief propo-
sItIOn.
He certainly did not understand it in the Aristotelian sense. A
?ecessary proposition for him, at least in the Critique of Pure Reason,
IS not a modal proposition of the apodeictic kind; for he gives examples
of necessary propositions in which there is neither a 'must' nor a
'cannot' nor a 'necessarily'. (He also gives examples in which a 'must'
does occur.)
He certainly did not understand it in the Leibnizian sense either,
as the analytic necessity of a proposition such that either itself or its
contradictory is selfcontradictory. For it is one of his most important
premisses that there exist propositions which are necessary although
they are not analytic.
Did he have our fourth sense in mind? Did he mean by a 'neces-
sary proposition' a universal proposition of unrestricted generality?
No again, for he explicitly distinguishes necessity from universality,
and declares them to be two distinct attributes of a priori propositions
(B4).
(5) Kant must have intended, therefore, some fifth sense of the
phrase 'necessary proposition', distinct from all those I have indicated.
But he did not tell us what it is. In this respect his Critique of Pure
Reason reminds me of his use of the word 'freedom' in his Foundation
of the Metaphysic of Ethics. He there makes great and astounding use
of the word 'freedom'; and, although this word cries out for definition
when used in philosophy, he does not define it_
Can we collect this fifth or Kantian sense of 'necessary proposi-
tion' from the way he uses the phrase? Necessity enters the Critique of
Pure Reason with the words: 'First, then, if there is a statement which
is thought along with its necessity, it is an a priori judgment' (B3). Of
all the obscure sentences! Next he tells us that 'necessity and strict
universality are sure signs of a priori knowledge, and also belong
together inseparably' (B4). Next he says 'it is easy to show' that we
actually know some necessary judgements; for such are all the proposi-
tions of mathematics, and the proposition that every change must have
a cause. In this last the concept of a cause obviously contains the
concept of a necessity of the connection with an effect (B5). We also
collect without difficulty from what Kant says that according to him
the necessary is equivalent to the a priori. All necessary propositions
are a priori (Pure Reason, B4); and all a priori statements are necessary
(Werke, viii, 228-9, Akad. Ausg.). Later (Pure Reason, B17) we are
told that it is clear that the following are necessary: 'In all changes of
the physical world the quantity of matter remains unchanged', and 'In
every communication of motion, action and reaction must always
equal each other'.
I regret to say that I think that the concept of a necessary
proposition is now a muddle, and that this muddle began with Kant. I
think that Kant's concept of a necessary proposition is nothing defi-
nite, but just a confusion of the four clear concepts of a necessary
proposition which I have indicated. Kant's necessary proposition is
none of those four necessary propositions precisely, but it is all of
them confusedly.
Kant's necessary proposition is partly and confusedly the Aristo-
telian necessary proposition, that is, the proposition which declares
that something must be so. This is suggested by those of his examples
which contain the word 'must'. It is more than suggested by his talk of
statements which 'think necessity' (B4), or 'contain' necessity, or
'whose assertion brings necessity along with itself.' When Kant said
metaphorically that a statement 'contained' necessity, he meant in part
that it asserted necessity, i.e. asserted that all S must be P.
Kant's necessary proposition is partly and confusedly the uni-
versal proposition. He speaks of necessity and universality as two
marks of the a priori. He says that they 'belong together inseparably'.
He does not realize that they belong together inseparably because they
are one and the same thing. The necessity which he has in mind there
is just the universality. The unrestricted universal 'All S is P' has
necessity in that it entails that, if anything were S, it would necessarily
be P also. It is a necessary proposition in that it necessitates each S to
be P. If absolutely all Ss trre P, then any particular S must be P. There
is no other necessity about many of Kant's examples than just their
universality, which necessitates something about every particular fall-
ing under the subject-term. Take for example his 'In all changes of the
physical world the quantity of matter remains unchanged'. There is
nothing necessary about it except that if it were true it would
necessitate each particular change to preserve the total quantity of
matter.
Kant thought he had found a necessary proposition whenever he
felt compelled to believe (sense 1) a proposition which either asserted
that so~ething must be so (sense 2), or had a selfcontradictory
contradIctory (sense 3), or asserted something with unrestricted uni-
versality (sense 4).
Kant did not sufficiently consider the relations between necessity
and truth. Is a necessary proposition by definition a true proposition?
If it is, we had better apply the appropriate tests for truth in each case
before declaring any proposition necessary. If a necessary proposition
need not by definition be a true proposition, some necessary proposi-
tions will be false. Some of the propositions, of course, which some of
us are compelled to believe (sense 1) are false; and some of the
propositions which say that something must be so (sense 2) are false;
and some of the propositions which say that something universally is so
(sense 4) are false.
A Leibnizian 'necessary truth' is necessarily true, both by its
name and by the definition which Leibniz gave it. But Kant, when he
proceeded from Leibniz's 'necessary truths' to his own 'necessary
propositions', and declined to accept Leibniz's criteron as a criterion
for all necessary propositions, failed to observe that his 'necessary
propositions' were no longer guaranteed to be true. He was far too
confident that he knew when he had a necessary truth. More precisely,
while he rightly doubted the confidence of the dogmatic metaphysi-
cians, he remained himself far too much of a dogmatist in physics.
While he points out that Newton's propositions clearly make universal
and synthetic assertions, he passes over in silence his belief that they
make true assertions (BI7-I8). I think this is because, or partly because,
he assumes that universality entails necessity and necessity entails
truth. But his kind. of necessity does not entail truth; only Leibniz's
does so.
How do we tell, according to Kant, whether a statement is
necessary, in the case where it is a synthetic necessity, i.e. its denial is
selfconsistent? His idea seems to be that the criterion of necessity is
that the statement both asserts something to be necessary and 'forces
itself upon us' as true. 'You must, therefore, overruled by the necessity
with which this concept forces itself upon you, admit that it has its
seat in your faculty of a priori knowledge' (B6). If there is a statement
to the effect that any S must be P, and I feel myself obliged to believe
it, that shows that it is a necessary statement. For Kant, it goes without
saying, if it is a necessary statement it is a true statement...
Why was Kant confident t~at necessary proposI.t10nS wer.e t~ue
propositions? It was probably mamly a confidence whIch he unJustIfi-
ably copied from Leibniz. Leibniz had spoken only of necessary truths,
and had given them a definition which made them indeed necessarily
true. Kant spoke of necessary propositions, and tacitly repudiated
Leibniz's definition without giving another; but he went on assuming
that they were all necessarily true.
But I think there was another cause also. Kant knew that no
description of what we have experienced can entail an unrestrictedly
universal statement, a knowledge which he expressed by statements
like 'Experience does not give strict universality'. Therefore a man
would be a fool or a knave who would make an unrestricted generali-
zation merely on the basis of experience. Therefore the great Newton,
neither fool nor knave, must have made his unrestricted generalizations
from some other source. There must be some inner light which gives
us these synthetic unrestricted generalizations about experience; and
the inner light is infallible. That is his idea in this passage: 'If a judge-
ment is thought in strict universality, that is, so thought that absolutely
no exception is allowed to be possible, then it is not derived from
experience, but is valid absolutely a priori' (B4).
He takes it for evident that we sometimes assert absolute necessi-
ties whose contradictories are selfconsistent, and that when we do so
we are sometimes right to do so. So do I. But he further seems to take
it for evident that when we do so we are always right to do so. He
seems to assume that the mere fact that we do make these outrageously
sweeping assertions, going beyond all possible experience, by itself
shows that they are true, 'schlechterdings apriori gtiltig'. In meta-
physics it does not show that they are true; but in physics and
mathematics it apparently does. I believe that that is because he thinks
that we are not so foolish as to induce anything. No one would be so
foolish as to assert an unrestrictedly universal statement about the
physical world on the ground of man's restricted experience. Yet we
do make unresuictedly universal statements about the physical world.
Therefore we must have a nonempirical source for these statements.
And he assumes that a nonempirical source would never deceive us, as
many philosophers have assumed.
2. ANALYTIC
Kant held, at least in effect, that Leibniz's definition of 'necessary

~ruth~'. had been ;00 n~rrow. Leibniz's definition was in reality a
aefimnon only. of . analync' necessary truths, and 'synthetic' necessary
truths fell outsIde It. Kant then had to explain what he meant by these
words 'analytic' and 'synthetic'.
!"Ie did not take what seems the obvious way of explaining them.
He dId not say: 'An analytic truth is any truth whose denial is self-
contradictory; but a synthetic truth is any truth whose denial is
selfconsistent; thus Leibniz was really defining analytic truth; he was
not defining necessary truth as he supposed himself to be.'
This was the obvious way for I\ant to explain his distinction
between analytic and synthetic statements, because it would have
shown where he stood with Leil;miz, it would have given a clear and
sharp distinction, and it would have been true. It would have been
true, because Kant's distinction between analytic and synthetic truths
is equivalent to Leibniz's distinction between truths whose denials are
selfcontradictory, and other truths. At any rate Kant himself believed
the two distinctions to be equivalent; for on a few occasions he does
substitute Leibniz's distinction for his own. (Pure Reason, B12, 14, 15,
190; Werke, viii, 248.) The last of these passages speaks of the Law of
Contradiction as the principle of analytic judgements; and the last but
one says: 'If a judgement is analytic, ... its truth must always be
capable of being adequately recognized according to the Law of
Contradiction'.
But this obvious way of explaining the distinction is rare and
secondary in Kant. His primary and his usual explanation is to say that
analytic statements are those in which the predicate is contained in the
subject. In the Critique of Pure Reason he introduces the distinction
thus: 'In all judgements, in which the relation of a subject to the
predicate is thought (to mention only affirmative judgements; after-
wards it will be easy to apply it to negative judgements), this relation
is possible in two ways. Either the predicate B belongs to the subject A
as something implicitly contained in this concept A; or B lies wholly
outside the concept A, though connected therewith. In the first case I
call the judgement analytic, in the other case synthetic' (BlO).
There is some doubt about the force of the relative clause in the
first sentence of this passage: 'In all judgements, in which the relation
of a subject to the predicate is thought'. Does it give a description of all
judgements as such, or does it mark off a species of judgements from
judgements in general? If the former, Kant thought that all judgements
think the relation of a subject to a predicate, and intended his division
to divide all judgements. I think, but not confidently, that Kant here
intended to mark off a species of judgements from judgements in
general, and to divide only this species. My grounds for this interpreta-
tion are (1) that two persons, each very familiar with the German
language, have assured me that the text must mean this; and (2) that, as
Mr. Warnock has pointed out to me, Kant's famous doctrine that
existence is not a predicate implies rather obviously that the judgement
that 'God exists' is not in subject-predicate form, and therefore that
some judgements do not 'think the relation of a subject to a predicate'.
There were disadvantages in this manner of making the distinc-
tion. (1) In the first place, it seems a pity to conceal the continuity
with Leibniz. (2) Secondly, Kant's words are much less clear and
unambiguous than those of Leibniz. 'Contained in' is a metaphor, and
not susceptible of clear interpretation without going back to the words
of Leibniz. (3) Thirdly, Kant's division is a division of a species of
judgements only, whereas Leibniz had given a criterion which cor-
rectly and usefully divides the whole class of judgements.
Such being the disadvantages of Kant's new criterion, it seems
that he would have done better to stick to Leibniz's criterion, merely
adding his new version of what it is that Leibniz's criterion divides. He
would then have said: 'Leibniz's distinction between those statements
whose denial is selfcontradictory, and the others, is clear and good; but
Leibniz was mistaken in supposing it to be the distinction between
necessary and contingent statements; it is really a distinction of its
own, which I will baptize as the distinction between analytic and
synthetic statements; and it is not the case that all necessary statements
are analytic; on the contrary, some of them are synthetic'. So far as I
see, if Kant had said this he would have left himself free to make all the
important points of the Critique of Pure Reason, while at the same
time avoiding the disadvantages of the containment-criterion.
Why then did Kant adopt his new criterion? What overcompen-
sating advantages did he see in it, which caused him to relegate Leib-
niz's criterion to the status of an occasional accessory? Or was it
merely that the criterion occurred to Kant and the disadvantages of it
did not occur to him?
I have sometimes thought that perhaps Kant changed from the
contradiction-criterion to the containment-criterion because he felt
unequal to the task of showing that mathematical statements are
synthetic by the contradiction-criterion, but able to show that they are
synthetic by the other criterion. He believed that mathematical axioms
and theorems are synthetic, though the proofs by which the theorems
are derived from the axioms are analytic. He therefore believed that,
for example, the statement '7 + 5 = 12' cannot be shown to be true by
the contradiction-criterion. From this it follows, though he does not
say so, that the contradictory statement, '7 + 5 #- 12', is not
selfcontradictory and cannot be shown to lead to contradiction. But
could he actually demonstrate to us the fact that '7 + 5 #- 12' leads to
no contradiction? At any rate he did not try to do so in the Critique of
Pure Reason. After implying vaguely that it does not lead to contra-
diction, he turned fo the containment-criterion and argued that '7 + 5
= 12' is synthetic by this criterion. Here is a translation of this
remarkable slide: 'At first one would think that the statement "7 + 5
= 12" was a merely analytic statement, following from the concept of
a s~m of seven and five in accordance with the Law of Contradiction.
But when one looks more closely one finds that the concept of the sum
of 7 and 5 contains nothing more than the union of the two numbers
into one, without its being in any way thought which this number is
that includes the two' (B1S). This· passage has suggested to me the
guess that Kant adopted the containment-criterion because he could
not shO\v in detail by Leibniz's criterion that mathematics is synthetic.
Whether he could show it or not, it appears that he was right. It
appears that mathematics, or a very large part of it, is synthetic. It is
perfectly clear now, at any rate, that Euclidean geometry is synthetic,
because it has contraries which are selfconsistent. The ordinary propo-
sitions of Euclid which we learn in school, such as his postulate of
parallels and his theorem of Pythagoras, are synthetic. Professor Ayer
has unfortunately misled many undergratuates by writing that the
'propositions > [of a geometry] < are purely analytic propositions'
(Language, Truth, and Logic, 2nd edn., p. 83).
Here is another guess as to why Kant adopted the containment-
criterion. Perhaps he thought that the dogmatic metaphysicians, whom
he wished to undermine, derived their confidence in their statements
from the assumption that all these statements were mere unfoldings of
concepts. If so, he might well judge that he would shake the assump-
tion by bringing it into full consciousness, and explicitly contrasting
statements that clearly are mere unfoldings of concepts with other
statements.
These are mere guesses. I cannot estimate their probabilities, and
I am unable to answer the question why Kant changed from the
contradiction-criterion to the containment-criterion.
Whatever his hopes and considerations were, it seems now that
more was lost than was gained by the introduction of his new
criterion. He confused the notion of a necessary statement, and he
introduced a new notion, analytic statement, which was unsatisfactory
as he introduced it, and has given rise to much bewilderment since.
Although Kant implied that the contradiction-criterion is equiva-
lent to his containment-criterion, he also implied that the distinction
between analytic and synthetic was his own discovery. 'The distinc-
tion between analytic and synthetic judgements has perhaps not previ-
ously been entertained' (Pure Reason, B19; more strongly in the Streit-
schrift, Werke, viii, 232). In his Streitschrift (Werke, viii, 248 f.) he
represented Leibniz's two principles, that of Contradiction and that of
Sufficient Reason, as in their day 'a new and remarkable indication of
inquiries which were yet to be made in metaphysics (and which in fact
have but recently been set on foot),. He meant apparently that
Leibniz's two principles foreshadowed his own distinction. between
analytic and synthetic necessities and the need for special explanations
of the latter. He suggested that Leibniz's Principle of Sufficient Reason
was not so much a principle as an indication of the need for a principle.
3. PROPOSALS
What we need in this area is in the first place a division of

statements or propositions. We want a division of propositions rather
than of 'judgements'; and Kant makes his division into analytic and
synthetic a division of propositions in some of his other works, though
he makes it a division of judgements in the Critique of Pure Reason.
In the next place, it seems that we want a division of all state-
ments, not a division of a subclass of statements, though of course we
want many of those too, for other purposes. Many thinkers from
Leibniz down have offered us here a division of true statements only.
Leibniz defined the division so that it had to be a division of true
statements only. Later thinkers have talked about it as if it were a
division of true statements only, but have not always defined it so that
it had to be. Ayer at first defines 'analytic' (L. T. & L., 2nd edn., p. 16)
so that an analytic proposition must be true; but later he defines it (p.
78) so that 'A father is not a parent' is an analytic proposition; for his
phrase 'its validity depends' is equivalent to 'whether it is true or not
depends'. There has been an undesirable silence about the relation of
false propositions to this distinction. Are they or are they not m-
cluded?
I think we ought to include them, for the following reasons.
(1) Much confusion has arisen, especially in Kant, from assum-
ing that all 'necessary propositions' are true propositions while putting
nothing into their definition to make them so.
(2) Leibniz's distinction, between truths whose denials are self-
~ontradictory, and other truths, points to a division of all propositions
mto two classes, namely those whose truth-value is, and those whose
truth-value is not, determined by themselves alone, because either the
assertic;>n or the denial of them is selfcontradictory. This is a character-
istic which the proposition that 'A father is not a parent' shares with
the proposition that 'A father is a parent', and which makes them more
like each other than either of them is like the proposition that 'Apples
were scarce in England in 1943'.
(3) If analytic-synthetic is understood to be a division of true
propositions only, we shall have to wait until A.D. 2000 before we know
whether it is a synthetic proposition that 'There will be a king of
Scotland in A.D. 2000'. Yet we can obviously say now that it is a
synthetic proposition if it is true. Similarly, that 'There is life on Mars'
is obviously synthetic if there is life on Mars. But, as I do not know
whether there is life on Mars, I do not know whether 'There is life on
Mars' is a synthetic proposition. This is rather absurd.
Against these reasons there is the argument that we have got into
the habit of saying 'That is analytic', and perhaps also 'That is neces-
sary', in the sense of 'That is analytically true'. But there is not much
force in this. We can get out of habits we have got into. It is a bad
habit, a muddled habit, because we often define 'analytic' and 'neces-
sary' in such ways that an analytic proposition and a necessary proposi-
tion need not be a true proposition. On reflection it seems clear that we
want to be able to tell that a synthetic proposition is synthetic before
we know whether it is true or false and independently of that. It is not,
however, possible to tell of an analytic proposition that it is analytic
before we know whether it is true or false, because the criterion of
analyticalness is such that the seeing that a proposition is analytic is
necessarily also the seeing what its truth-value is.
In the next place, if we are to have a division of all propositions,
we must not merely include the false propositions but also make our
division exhaustive. The way to make a division exhaustive is to make
it on a single principle. Ayer pointed out that Kant made the mistake
of having two principles. It is a curious fact that, in the very next
paragraph after criticizing Kant for having two principles in his divi-
sion, Ayer himself proposes a division according to two principles! He
proposes that we 'say that a proposition is analytic when its validity
depends solely on the definitions of the symbols it contains, and
synthetic when its validity is determined by the facts of experience'
(L. T. & L., 2nd edn., p. 78). Here we have two different principles of
division:
(1) A proposition is analytic when its validity depends solely on
the definitions of the symbols it contains; otherwise it is synthetic. Let
us call this his 'definition-principle'.
(2) A proposition is synthetic when its validity is determined by
the facts of experience; otherwise it is analytic. Let us call this his
'experience-princi pIe'.
Some of the recent doubts about the analytic-synthetic division
arise from the undetected duplicity of Ayer's principle of division.
There is in fact a huge subclass of propositions which are synthetic by
one of his principles and analytic by the other, namely the ethical and
in general the practical propositions. The statement that 'An apple
should be juicy and sweet' is synthetic by his definition-principle, but
analytic by his experience-principle. Professor Urmson in his article
'On Grading' (MIND, 1950) reached the conclusion that the analytic-
synthetic division does not apply to ethical propositions; inspection of
his article will reveal that this was because he was assuming Ayer's
double principle for the division. He felt that he could not call the
above statement analytic because its validity is not determined solely
by the definitions of the symbols it contains, and could not call it
synthetic because its validity is not determined by the facts of experi-
ence. He concluded that 'the question whether the connexion between
two sets of characteristics is analytic or synthetic is a question which is
designed to be asked where the related characters are descriptive' (p.
157, ct. p. 161).
But there is no need to abandon the hope of an exhaustive
division of all definite propositions. We need only to avoid having two
principles of division at the same time. Which of Ayer's principles is it
better to keep? In my opinion, it is better to keep the one that is nearer
to Leibniz, and better still to use a principle still nearer to that of
Leibniz. If we say that analytic propositions are those of which either
the assertion or the denial is selfcontradictory, it follows easily that
'An apple should be juicy and sweet' is a synthetic proposition, and
ordinary practical propositions are usually synthetic. Many of them,
however, are too indefinite to fall into either class.
Practical propositions resemble logical propositions in not having
their truth-value determined by the course of events, but differ in that
they can be both asserted and denied without selfcontradiction. Practi-
cal propositions resemble historical propositions in that they can be
both asserted and denied without selfcontradiction, and in that the
course of events is relevant to their truth-value. But the way in which
the course of events is relevant to the truth-value of practical proposi-
tions is different from that in which it is relevant to the truth-value of
historical propositions. Historical propositions propose to describe the
course of events, and therefore the course of events determines di-
recdy whether they are true or false. Practical propositions, on the
other hand, propose rather to alter or to evaluate the course of events.
This makes the course of events relevant to them, but relevant in a
different and indirect way, which I cannot characterize precisely. This
peculiar kind of relevance is hinted at by the following observation:
the. practical proposition, 'Justice must be done though heaven falls',
seems to be false or rejectable because it proposes a policy and denies
that the course of events has any relevance at all to the choice of this
policy. Experience is relevant to practical statements in that a man who
asserts a practical statement in defiance of all possible experience is
acting improperly.
Mr. Hare gave the following account of the distinction: 'A
sentence is analytic if, and· only if, either (1) the fact that a person
dissents from it is a sufficient criterion for saying that he has misunder-
stood the speaker's meaning or (2) it is entailed by some sentence
which is analytic in sense (1). A sentence which is not analytic or
selfcontradictory is called synthetic' (Language of Morals, 41). This
has one and only one criterion, and that is a great improvement on
Ayer's definition. It has, however, two serious defects. First, it makes
the division a division of sentences. Questions and commands are
sentences; but it is not clear how one 'dissents from' a question; and it
is not clear that we wish to make this a division not merely of all
propositions but still more widely of all sentences. The other defect is
that the division is made by reference to a subject, who understands or
misunderstands. It is thus a subjective or epistemological division, a
division of propositions with reference to the attitude of subjects
towards them. I think that probably Hare's division is only apparently
or superficially subjective; but there is no need to be even apparently
subjective; and the division we have in mind is a division of proposi-
tions according to their intrinsic natures, not according to their
relations to subjects. The analyticalness of a proposition is distinct
from the certainty with which men may adhere to it; and there is
hardly any limit to what some men will dissent from.
We have the two sets of words, 'neces~ary-contingent' and
'analytic-synthetic'. But have we two distinctions to make? It seems
not, for it seems undesirable to follow Kant and speak of 'necessary
synthetic propositions'. A synthetic proposition can be necessary only
in the sense of either (1) being as a matter of fact true, or (2) being
firmly believed, or (3) being required to be believed by some cause or
for some purpose, or (4) being the assertion that something is neces-
sarily so, or (5) being the assertion of an unrestricted generalization. It
seems clear that to call a proposition necessary on anyone of these
grounds, or on any combination of them, is to invite confusion and
gain nothing. We have, therefore, only one distinction to make here.
Should we retain both ways of referring to this one distinction?
If not, which is the better way? Each of them is bad, because each of
them is very liable to misinterpretation and fails to carry its meaning
on its face. It would be far more generally intelligible to talk of 'self-
determining and unselfdetermining propositions', those which do and
those which do not determine their own truth-value. I prefer Leibniz's
words to Kant's words, partly because Leibniz gave a clear distinction
in the first place and Kant did not, and partly because Kant's words
have given rise in recent literature to a horrible noun which I do not
use in this paper, and which I cannot bring myself even to mention in
quotation-marks! It is, however, easier to talk of 'analytic falsehoods'
than of 'necessary falsehoods' in the same sense.
Necessary propositions are sometimes thought not to be really
propositions, or to be always very unimportant propositions. Both
views appear to be mistaken. Necessary propositions are not questions,
or exclamations, or commands, or incomplete parts of a sentence. Since
they are not commands, they are not rules. They are really proposi-
tions, in that propositions are those sentences which have a truth-value,
those sentences which call for intellectual assent or dissent by all
rational beings. Every language, consisting of a certain number of
words and a certain number of rules for combining them, yields a
certain number of sentences which invite, not answers, nor obedience,
but judgments of assent or dissent. These are the total possible
propositions in that language. Half of them are false; and the other
half, being the contradictories of the false ones, are true. Some of them
are contingent and the rest are necessary.
Professor Britton wrote that 'necessary propositions are not
propositions and are not true or false in the same sense as empirical
propositions' (Proc. Arist. Soc. Suppl. (1947), p. 82). It seems to me
that the criterion is different but the sense is the same. The criterion of
the falsehood of a necessary proposition is that it is selfcontradictory,
whereas the criterion of the falsehood of an empirical proposition is
that the world is other than the proposition asserts it to be. But the
sense of calling it false is the same in each case. To call it false is to say
that it is a proposition which ought to be rejected by all users of
propositions, and its contradictory ought to be accepted by them.
The view that necessary propositions are not really propositions
is argued by Mr. Hamlyn in MIND for 1956. He points out that we
often use tautologies with practical effect. 'Enough is enough', we say;
and this has the effect of making somebody cease from pouring sugar
into his cup, much as if we had said 'Stop!'. 'If the voltage gradient
between the electrodes is high enough, the primary ions produce many
secondary ions by collision with the gas molecules.' This looks like a
necessary truth, because 'high enough' must mean 'high enough to
make the primary ions produce many secondary ions', etc. But it is
used as equivalent to the contingent statement: 'There is a height at
which the voltage gradient causes the primary ions to produce', etc.
This use of a necessary truth to represent a contingent assertion is
common.
Necessary truths, however, are not always uttered as idiomatic
ways of formulating commands or contingent truths. They are some-
times uttered for their own sakes and intended to be taken literally.
Nor are necessary statements always unimportant or trifling, at
least if anything is important beyond barely keeping alive. It is often
important and difficult to know whether a given statement is necessary
or contingent. Having ascertained that it is necessary, we do not
always lose all interest in it. If we had unlimited mental capacity, we
should be glad to know the truth-value of every proposition contained
in the language.
Nor are necessary statements artificial constructions, except in
the strained sense in which all words are artificial. Language is a part of
culture, and culture is 'natural' to all natures that are teachable.
Language cannot develop far without inevitably coming to include the
possibility of necessary statements. Necessary statements are no more
'man-made' than contingent statements.
KANT'S PERCEPTUAL
VOCABULARY*
T.D. Weldon
It is fair to say that a great deal of Kant's admitted failure to make

clear what he meant by 'knowledge of objects', 'objects of perception',
and 'objects of scientific inquiry' resulted from his tendency to use
perceptioll and sensation words which occur in both Leibnizian and
Lockean ways of talking, without noticing the very different employ-
ment of them which the two systems require. In addition, his vocabu-
lary of epistemological words was quite inadequate to the distinctions
he needed to draw, and therefore his use of them is generally both
ambiguous and vague. No clear analysis of his uses is practicable except
at enormous and unprofitable length, and no English equivalents to his
terms can be consistently employed in translation which are not seri-
ously misleading. Roughly, however, his most important German
words with the Latin equivalents of the Leibnizians and the English
renderings of Kemp Smith are these: 1
Kant Leibniz Kemp Smith

V orstellung Repraesentatio Representation
Erscheinung Phaenomenon Appearance
Anschauung Intuitio Intuition
Empfindung Sensatio Sensation
These need some further comment.

1. V orstellung is an umbrella word corresponding to Locke's
'idea', 'whatsoever is present to the mind when it thinks'. Hence it
refers indifferently to concepts, thoughts, images, and any kind of
apprehension. Feelings are generally excluded .
. . • Reprinted from Kant's Critique of Pure Reason, second edition, by per-

mISSIon of The Clarendon Press, Oxford. Pp. 124-127. Title supplied by editors
1 Kant's own table is at B 376. It is not very helpful. .
34
KantJs Perceptual 1/ ocabulary 35
The least misleading English equivalent is probably 'idea'. But
Kemp Smith not unnaturally felt he must avoid this because Kant uses
Idee technically to stand for the 'Ideas of Pure Reason' " God Freedom ,
and Immortality. It might be better to use 'notion' for these, following
Berkeley, though Kant himself would not have approved.
2. Erscheinung is simply not translatable by anyone word or
phrase. Kemp Smith decides in favour of 'appearance', but this is mis·
leading almost always. Frequently 'phenomenon', which Kant some-
times, though not often, uses as Leibniz had done, is preferable. The
point is that for Kant it makes sense to say that Erscheinungen have
backs and insides; that they are solid and obey the laws of mechanics.
But these are not permissible ways of using 'appearance' in English. On
the other hand, it is impossible to stick to 'phenomenon' or just 'thing'
since Kant also wants to say that Erscheinungen are Inbegriffe der
V orstellungen, and this produces the same feeling of cross-category
discomfort as Berkeley's 'We are fed and clothed with ideas.' We do
not use 'idea' and 'thing' like that in English.
Among professional philosophers, it is true, resistance to this kind
of talk has been reduced by Russell's recommendation to make 'thing'
equivalent to 'class of appearances' and to the phenomenalist conten-
tion of the 1930's that statements about material objects might be
replaced without change of meaning by statements about sense-data.
But this, too, is against ordinary use and I do not think Kant would
have accepted it. He did not think he was using a technical language
here.
My objection to translating Erscheinung by any of the 'appear-
ance', 'sensum', 'sense-content' group of words is not merely on
grounds of scholarship. It is rather that such translations tend to
commit Kant to a two-world view with its egocentric implications.
They push him into a terminology like that of Descartes and Locke in
which it is linguistically proper to talk as if the furniture of the world
contained both plates and looks of plates. 2
3. Anschauung. 'Intuition' is not a good English equivalent since
it at least suggests a special kind of faculty or achievement. We need
rather a standard ordinary-language word to cover 'seeing' and 'hear-
ing'. 'Perception' is probably the best available. In fact Kant seems
nearly always to have seeing in mind when he uses Anschauung. When
2 Kant realized the unclearness of Erscheinung and tried to do something

about it in the Opus Postumum by distinguishing between Erscheinungen and
Erscheinungen von den Erscheinungen. Ak. xxii. 339 and 363-5. Too late!
he wants to use a wider term he prefers Wahrnehmung, which Kemp

Smith translates as 'perception'.
4. Empfindung. 3 Kant's terminology is neither precise nor stable,
but generally speaking he uses 'sensation' (Empfindung) to refer to
neat or non-spatial data as contrasted with 'intuition' (Anschauung)
which refers to normal (visual) perception. He then takes for granted
a psycho-physical hypothesis as to what happens when I see some-
thing. 4 Objects, or rather light rays reflected from objects, operate on
(affect) the optic nerve and generate impressions which in themselves
are neither spatially nor temporally ordered (Empfindungen). In addi-
tion to being receptive of these impressions, our minds have a form-
imposing structure of their own. This is presumably an ordinary
empirical scientific hypothesis, and it suggests the question 'Might the
form of outer sense be atrophied or destroyed and impressions still be
received?' Kant seems generally to have favoured an affirmative an-
swer,5 and modern experiments confirm this.6 He knew, too, of
Locke's discussion with 'the thinking Mr. Molyneux' and later contro-
versy on the same point. 7 But he certainly did not think he was
propounding a scientific hypothesis.
The whole of this story is probably best thought of by the

analogy of coloured spectacles, as Paton does, though perhaps the
camera or even the sausage machine are even more suitable. Kant seems
rather surprisingly to have found no difficulty in holding (a) that there
were both spectacles and data observable in theory without spectacles,
(b) that he could examine the spectacles (in the Aesthetic) and find
out how they worked and what they did, (c) that he could not in
practice examine the pre-spectacular data, the raw materials, and give a
description of them. He could inspect the sausages and the sausage
machine, but not the sausage meat before the machine had formed it.
All this he seems to regard as just obvious (though it ceased to be so
when he began asking questions about synthesis). It is dismissed in a
paragraphS of the Critique. Probably it seemed to him the concern of
3 This and the succeeding paragraph are from pp. 110-111 of the original
text and replace Weldon's reference to them on p. 126 [eds.).
• • 4 Immanuel Kant's Menschenkunde oder philosophische Anthropologie,
LelpZ1§, 1831, pp. 64 ff. Compare Kant's Letter to Sommering, Ak. xii. 31.
See his discussions of unsynthesized manifolds, A 115 ff., and references in
Eisler, Kant Lexicon, ·'Mannigfaltig'.
6 J. Z. Young, Doubt and Certainty in Science, pp. 61 ff.
7 Locke, Essay, II. ix. 8 and footnote (Fraser); Kant, Menscbenkunde p. 63.
sB3~ ,
Kant's Perceptual Vocabulary 37
physiology and he would have referred a questioner to Haller's text-

book9 for further information.
The upshot is, though it must be repeated that Kant's handling of
his own terms would by no means satisfy a legal authority, that to have
an empirical intuition is to have an idea which is at least fairly clear and
distinct. Further,10 'The undetermined object of an empirical intuition
is called appearance', that is, the object of an idea is called 'thing' or
'phenomenon'. This is certainly vague and uninformative, but, as a
preliminary statement, it is not at all mysterious. 'Undetermined'
merely emphasizes the deliberate lack of precision in Erscheinung.
Kant does not want to tie himself down to 'body' or 'material thing' in
a restricted sense. A rainbow is an Erscheinung, so is a thunderstorm,
and so is a writing-desk.
In fact Kant's perceptual vocabulary does not in German en-
courage or discourage the adoption of a two-world view. It is too
vague to commit him to anything. The Kemp Smith renderings, which
are too firmly established in this country to be altered without generat-
ing more confusion among students who do not use the German text,
are much more committal. They all tend to edge Kant into the
phenomenalist camp of the 1930's. The real criticism of Kant's epis-
temological vocabulary is not that it is technical or complicated but
that it is far too simple to do the job he requires of it. He juSt has not
enough words with which to draw the distinctions within 'perceiving'
which he must draw in order to expound his concept of the 'object of
perception'. Hence he uses Erscheinung to stand indifferently for:
1. Things as they look (sound, etc.) to a normal observer under normal
conditions.
2. Things as they look to a specially equipped observer under more or
less controlled conditions (microscopes, telescopes, X-rays).
3. Ideas, perceptions occurring to normal observers under abnormal
conditions (Gestalt tricks, mirages).
4. Ideas, perceptions to abnormal observers under normal conditions
(pink rats).
(3) and ,(4) are supposed to be called Schein rather than Erscheinung,
but Kant does not always remember this. The footnote on B 69-70
shows where this inadequate vocabulary can lead him.
9 Haller, Elementa physiologiae corporis humani, 1762. See Ak. xii. 34.
10 B 34.
ON KANT'S NOTION O,F
INTUITION (ANSCHAUUNG)*
laakko Hintikka
In his brilliant essay on the Foundations of Arithmetic, Frege observes

that on different occasions Kant uses the word Anschauung (usually
translated by 'intuition') in seemingly different senses. In particular,
the relation of this notion to sensibility (Sinnlichkeit) varies. As the
term Anschauung is defined in Kant's logic, "there is absolutely no
mention of any connection with sensibility, which is, however, in-
cluded in the notion of intuition in the Transcendental Aesthetic, and
without which intuition cannot serve as the principle of our knowl-
edge of synthetic judgements a priori."1
This aperfu of Frege's poses several interesting problems for a
student of Kant. What exactly is meant by Anschauung in Kant's
writings? What is the relation of this notion to sensibility? If there
sometimes is no connection between the two, how did Kant at other
times come to relate them to each other? Of all the occasions on which
Kant employs the notion of intuition, when are we entitled to assume a
connection with Sinnlichkeit and when not? The purpose of the
present essay is to offer a few comments on these questions.
Prima facie, the relation of intuition to sensibility in Kant is
determined by the ordinary meaning and by the etymology of the
word Anschauung. There is an obvious connection in German be-
tween the term Anschauung and the verb schau en (to view). This
connection suggests an intimate relation between the notion of intui-
tion and the senses, especially the sense of sight. It suggests that the
•. This is the first publication of this article. All rights reserved. Permission
to reprmt must be obtamed from the author and publisher. English translations
which .occur in square brackets after German quotations have been inserted by
the editors and are by Kemp Smith, in the case of the Critique and Petra von
Morstein, in other cases. '
1 Gottlob Frege, Die Grundlagen der Aritlmtetik, eine logisch-mathematische
Untersuchung uber den Begriff der Zahl, Breslau, 1884. Reprinted with an English
translation by J. L. Austin, Blackwell's, Oxford, 1950. (See p. 19.)
38
On Kant's Notion of Intuition (Anschauung) 39
~eaning of Anschauung should be something like an image or a mental
pIcture, and furthermore also one's capacity of conjuring up such
"views" or images. This is in fact roughly the present-day ordinary
meaning of the word, especially insofar as the noun Anschauung is
connected with the cognate adjective anschaulich which means "intui-
tive" in the sense of "graphic" or "vivid". It is tempting to think that
this meaning is what guided Kant in his use of the term Anschauung.
From this point of view, it might even seem less than happy to
translate Anschauung as "intuition". In fact, this has been claimed by
Bertrand Russell in his History of Western Philosophy.2 Before Rus-
sell, Edward Caird had gone as far as to translate Anschauung as "per-
ception" and not as "intuition".3
This same interpretation underlies the most common view of the
role of the concept of Anschauung in Kant's philosophy of mathe-
matics. Kant's statements that the mathematical method turns on the
constant use of Anschauungen are usually taken to mean that in
mathematics we have to appeal all the time to our geometrical or
temporal imagination. A blunt statement of this view is again due to
Bertrand Russell: "Kant, having observed that the geometers of his day
could not prove their theorems by unaided argument, invented a
theory of mathematical reasoning according to which the inference is
never strictly logical, but always requires the support of what is called
'intuition' ."4
Although all these opinions are undoubtedly inspired by per-
fectly genuine and important features of Kant's views on mathematics,
they are nevertheless gravely misleading as far as the meaning of the
term Anschauung and its role in Kant's philosophy of mathematics are
concerned. The sense of Anschauung on which the opinions of Russell,
Caird and others are based is a beautiful example of what C. S. Lewis
has called a "dangerous sense"; it is a sense which we find tempting to
read back to the texts because it fits the context, although it is in reality
different from the actual sense the earlier writer was relying on.a
2 Bertrand Russell, History of Western Philosophy and Its Connection with

Political and Social Circumstances from the Earliest Times to the Present Day,
London, 1946, p. 734.
3 Edward Caird, The Critical Philosophy of Immanuel Kant, Glasgow, 1889,
vol. 1, p. xi. The line of thought of Russell and Caird is echoed by several recent
writers, e.g. by S. Korner in Kant, Penguin Books, Harmondsworth, Middlesex,
1955, pp. 27, 59, 80,95 etc.
4 Bertrand Russell, Introduction to Mathematical Philosophy, London, 1919,
p.145.
a C. S. Lewis, Studies in Words, Cambridge, 1960 (second edition, 1966).
It is easily shown that all attempts to dispense with the translation
of Anschauung as "intuition" are bound to fail. The term was not
taken by Kant from the ordinary usage of his time, although his choice
may have been influenced by the "dangerous" (= modern) sense of the
word which was already gaining currency in his time. Kant took it
over from the philosophical jargon current in Germany in the eight-
eenth century. In this jargon, Anschauung enjoyed a well-established
status as a technical or semi-technical term. It had been introduced into
the German philosophical terminology (probably by Christian Wolff)
as a translation of the (mediaeval) Latin term intuitus. 6 Wolff's suc-
cessors, including Kant, were aware of its intended synonymy with the
Latin term. For instance, A. G. Baumgarten's Metaphysica, which was
used by Kant as a textbook, in so many words identifies "ein an-
schauendes Erkenntniss" with cognitio intuitiva. 7 In the Critique of
Pure Reason Kant resorts to the term intuitus in explaining his sense of
Anschauung,8 and in his Dislfertation of the year 1770, which was
composed in Latin, the tenn intuitus is used precisely in the same way
as Anschauung is used in the comparable parts of his critical writings.
For these reasons, it is scarcely possible to deny that the term
"intuition" is the best possible translation of Kant's term into English.
Indeed, our observations suggest that the meaning Kant associated with
the term Anschauung can be elucidated as well and even better by
considering the use of the term "intuition" and its cognates in the
philosophical usage of the eighteenth and seventeenth centuries than
by considering the meaning of Anschauung in Kant's non-philosophi-
cal contemporaries. This is made even more important by the con-
siderable fluctuations that there seem to have been in the ordinary
senses of Anschauung in German. Before it was connected by Wolff
with intuition, this word (and its cognates) could be used with an
almost opposite purpose. When Jacob Bohme wanted to tell that he is
following what we would probably call intuition and not appearances,
h~ could express this by saying that he is writing "nach Geist und Sinn,
rucht nach dem Anschauen".
How, th~n, was the te~m ~'intuition" used by Kant's predeces-
sors? One pertment observatIon IS that there was no inseparable con-
6 Se~ Hannann Paul, Deutsches W orterbuch, Vierte Auflage Halle Saale

1935; Monz Heyne, Deutsches Worterbuch, Leipzig, 1890. ' ,
~ A. G. Baumgarten, Metaphysica, Halle, 1757, § 620.
.. A 320. =:= B 377. (Letters A and B refer, as usual, to the first and second
edItIons of Cnttque of Pure Reason, respectively.)
ne~t~on between t,?e notion of intuition and the sensibility in the
wrItmgs of most seventeenth- and eighteenth-century philosophers.
Descartes could even contrast intuition to the 'fluctuating testimony of
the senses' as well as to 'the blundering constructions of imagination'. 9
For him, intuition was a form of activity of the intellect, and it is in
fact sometimes called by commentators intellectual intuition. As Alex-
andre Koyre observed, in Descartes "I 'intuition est cette fois-ci une
intuition intellectuelle et non intuition mystique".lo Similarly, Spinoza
reserves the attribute imuitiva for, a mode of scientia entirely different
from sensible observation. l l When writers of this kind employ the
term intuitus, they are not referring to any real connection with
sensibility. What they are stressing is a particular analogy between an
act of intuiting and an act of seeing, viz. the immediacy of both the
acts.
This is particularly clear in Leibniz. He does not contrast intui-
tive knowledge to non-sensible knowledge but to symbolic knowledge.
For him, the kno\vledge of a concept is intuitive if we have in mind all
the marks that go into the definition of the concept; knowledge is
symbolic if it is not immediate in the sense that we make use of words
or other symbols the meaning of which we remember but do not have
explicitly in mind for the moment. 12 This distinction does not seem to
have anything to do with a distinction between truths derived from
perception and truths independent of perception. Intuitive knowledge
could for Leibniz be of truths of reason as well as of factual truths. IS
Conversely, Baumgarten, who also contrasted intuitive knowledge to
symbolic knowledge, and for whom neither of them seemed to have
more intimate a relation to sensibility than the other, indicated that in
both cases we may have a relation of signs to their meanings which is
established "percipiendo" (Lac. cit.).
This was also the meaning of Anschauung which was uppermost
in the minds of Kant's philosophical contemporaries: " ... in der
Philosophie versteht man durch die anschauende Erkenntniss, eine jede
9 Regulae ad directionem ingenii, rule 3. Cf. L. J. Beck, The Method of

Descartes, Oxford, 1952, ch. iv.
10 A. Koyre, Essai sur l'ldee de Dieu et les Preuves de son Existence chez
Descartes, Paris, 1922, p. 183.
11 Cf. Harry Wolfson, The Philosophy of Spinoza, Cambridge, Mass., 1934,
vol. 2, pp. 153-155.
12 Leibniz, Nouveaux Essais sur l'Entendement Humain, IV, 2, § 1; "Medita-
tiones de cognitione, veritate et ideis," Acta Eruditorum, November 1684.
13 Nouveaux Essais, loco cit.
Erkenntniss, die wir durch die Empfindung erlangen, oder da wir uns
die Sache selbst oder doch ihr Bild vorstellen; die sinnliche, bildliche
Erkenntniss im Gegensatze der symbolischen, da man eine Sache unter
Worten oder and ern Zeichen denkt."14. [... in philosophy one takes
as intuitive each piece of knowledge which is acquired by means of the
senses, or by representing to ourselves the thing or its picture; that is,
representative knowledge, or sensible knowledge, as opposed to sym-
bolic knowledge in which we think of things in terms of words or
other symbols.]
One of the most important ideas the tenn Anschauung was to
convey was for Kant, too, the immediacy of its relation to an object.
For him, 'knowledge ... is either intuition or concept (intuitus vel
conceptus). The former relates immediately to the object and is single,
the latter refers to it mediately by a feature which several things
may have in common' (A 320 = B 376-377): 'Intuition is a representa-
tion such as would depend directly on the presence of the object'
(Prolegomena §8) . . . .1 5
These quotations also show what for Kant was the alternative to
an immediate relation to objects. It was a reference to objects by
means of certain marks or characteristics which may be shared by
several objects, i.e., a reference to objects by means of general con-
cepts. Hence, another way of saying that Anschauungen have an im-
mediate relation to their objects is to say that they are particular ideas
or 'representations' (V orstellungen) in contradistinction to general
representations or concepts. This, it seems to me, is the basic meaning
of Kant's term "Anschauung." It is in this way that he defines the term
in Logic (as observed by Frege): 'Die Anschauung ist eine einzelne
Vorstellung (representatio singularis), der Begriff eine allgemeine (rep-
resentatio per notes communes) oder reflectirte V orstellung (represent-
atio discursiva).'16 [An intuition is a single representation (representatio
singularis), a concept is a general representation (representatio per
notes conununes) or a representation of reflection (representatio dis-
14 J. Chr. Adelung, Auszug aus dem grammatisch-kritischen Worterbuch

der bochdeutschen Mundart, Leipzig, 1793. Cf. also E. Schmid's Worterbuch zum
leichtaren Gebrauch der Kantischen Scbriften (jena, 1798) on Anschauung
(quoted in my paper, "Kant's 'New Method of Thought' and His Theory of
Mathematics", Ajatus, vol. 27 (1965), pp. 37-47).
15 Further evidence is provided by H. Vaihinger Commentar zu Kants
Kritik der rein en Vernunft, Stuttgart, 1881-1892, vol. 2, pp. 3,24.
16 Logic (Immanuel Kants Logik: Ein Handbuch zu Vorlesungen ed by
G. B. Jiische, Konigsberg, 1800), § 1. ' .
On Kant's Notion of Intuition (A nschauun9) 43
cursiva).J Kant abided by the definition in his writings, as witnessed
?y the many passages in which he contrasts intuitions or particular
Ideas to concepts or general ideas. The following examples are fair
samp,les: 'Der Raum ist kein diskursive oder, wie man sagt, allgemeiner
Begnff von Verhaltnissen der Dinge liberhaupt, sondern eine reine
Anscha~ung' [SP.ace i,s not a discursive or as we say, general concept
of relatIons of thmgs m general, but a pure intuition. J (A 24 = B 39);
'Zur Konstruktion eines Begriffes wird also eine nicht empirische An-
schauung erfordert, die folglich, also Anschauung, ein einzelnes Objekt
ist . . . ' [For the construction of a concept we therefore need a non-
empirical intuition. The latter must, as intuition, be a single object
. , .J (A 713 = B 741; Kant's italics); 'Begriffe .. , bedlirfen doch,
urn ihnen Bedeutung und Sinn zu verschaffen, einen gewissen Gebrauch
in concreto, d.i. Anwendung auf irgendeine Anschauung .. .' [Con-
cepts . , . still require, to provide them with meaning and sense, a
certain use in concreto, i.e. application to some intuition ... J (Prole-
gomena §8).
In contrast, a concept (Begriff) was defined by Kant as always
being a general concept-a definition to which he did not always con-
form himself.
Our quotations also show that Kant does not distinguish between
an intuition representing an individual object and an intuition itself
being an individual object. This is largely due to the psychologistic
terminology he uses. Abstracting from this terminology, we may say
that Kant's notion of intuition is not very far from what we would call
a singular term, An intuition is for Kant a 'representation'-we would
perhaps rather say a symbol-which refers to an individual object or
which is used as if it would refer to one.
We may observe that Kant did not usually speak of intuition of
particular sensible objects, as one probably would expect him to do if
the word meant something like a 'mental picture'. Instead, he speaks of
intuitions as corresponding to general concepts and as being themselves
particular ideas.
One likely factor in this transformation of the concept of intui-
tion from expressing immediacy to indicating individuality was the
influence of Johann Heinrich Lambert. Descartes had tried to present a
general method for human knowledge which turned on analyzing our
ideas into their simplest componentsP In this method, intuition had a
17 See Regulae, especially rules 5 and 6,

well-defined place: it pertained to those simplest elements and to their
combinations. Lambert attempted a somewhat similar analysis of
human knowledge. 18 One feature of his theory which distinguishes it
from Descartes' is that he tried to specify the character of the simplest
ideas we have to postulate. According to Lambert, the simplest con-
cepts are all individual concepts. Since these were what intuition per-
tained to, it was not far-fetched to connect individuality and intuitiv-
ity in this way.
From the historical point of view, the close relation of intuitivity
to individuality makes the connection between intuitions and sensi-
bility very natural. Aristotle had already held that 'it is sense-percep-
tion alone which is adequate for grasping the particulars ...' (An.
Post. 81 b7). In connecting intuitions with sensibility, Kant is thus to
some extent following old Aristotelian precedents. The immediate
source of this sort of a view may nevertheless have been one of his
German predecessors several of whom put forward similar ideas. 19 But
what becomes surprising on this view is that mathematics should have
anything to do with intuitions according to Kant. In Kant's primary
sense of the word, the intuivity of mathematics which he sets so great a
store on means that mathematics turns essentially upon the use of
particular ideas, whereas a modern reader would certainly expect
mathematics to deal with ideas of a considerable generality. I want to
argue, however, that this is precisely what Kant's theory of the
mathematical method amounts to, although I cannot do so fully in this
paper. 20
In spite of the Aristotelian precedent, even the relation of intui-
18 J. H. Lambert, Anlage zur Architectonic, oder Theorie des Ersten und

Einfachen in der philosopbischen und matbematiscben Erkenntnis, Riga, 1771.
Cf. also Lambert's letters to Kant, especially the one dated February 3, 1766, where
all simple concepts are asserted to be individual concepts.
19 Thus]' A. Eberhard writes in the journal Pbilosopbiscbes Magazin which
h~ himself edited (vol. 3, 1789, pp. 292-293): "So unterscheidet Wolf Sinne, Ein-
bIldungskraft und Verstand dadurch, dass die erstern das gegenwartige, die andern
das abwesende Einzelne, der Verstand hingegen das Allgemeine vorstellt (Wolf:
Psych. rat. § 86). Eben das geschieht in den deutschen Uebersetzungen von A. G.
Baumgartens Metaphysik und in Meyers Untersuchung fiber die Natur des Ver-
standes." [!hu~ Wolf distin~uishes between the senses, the imagination and the
understandmg In the followmg way: the first presents the particular which is
present, the second presents the particular which is absent, while the understanding
f>resents the general. (Wolf: Psych. rat. § 86). The same distinction is made in the
German translations of A. G. Baumgarten's Metapbysics and in Meyer's investiga-
tions into the nature of the understanding.]
20 I have tried to do so a little more fully in my paper "Kant on the Mathe-
matical Method", Tbe Monist, vol. 51 (1967), pp. 352-375. '
tions to sensibility becomes something of a problem. For Kant insists
that the connection between the two is not a mere logical consequence
of the definition of intuition. On the contrary, it is said to be due to
human nature, to the peculiar nature of uns Menschen. Kant says
repeatedly that it is not impossible that there should be other beings
which can have intuitions by means other than senses. According to
him, " . . . man kann vor der Sinnlichkeit doch nicht behaupten, dass
sie die einzige mogliche Art der Anschauung sei" [. . . we cannot
assert of sensibility that it is the sole possible kind of intuition.] (A 254
= B 310: cf. A 27, 34, 35,42, and 51).
How, then, did he come to hold that there is in fact an intimate
connection between intuitions and sensibility? What were his reasons
for assuming such a connection? In order to approach these questions,
it is advisable first to distinguish between a priori and a posteriori
intuitions. Intuitions, in general, were for Kant V orstellungen or ideas
which are due to the immediate presence of their objects or which are
used as if they were due to the presence of objects. 21 In the first case,
we have to do with an empirical (a posteriori) intuition; in the second,
with an a priori one. Here saying that a posteriori intuitions are due to
objects is according to Kant, the same as to say that they are results of
an affection or influence on our mind by objects. Kant defined sensi-
bility as the very capacity of the mind of being affected in this way.
Hence, the a posteriori intuitions are connected with sensibility virtu-
ally per definition em. The problem of the relation of intuitions to
sensibility is therefore essentially a problem of the relation of a priori
intuitions to sensibility.
According to Kant, the most important a priori intuitions are
those of space and of time. Showing that all our intuitions are bound
up with sensibility thus amounts to proving, first and foremost, that
the ideas of space and time are inseparably tied up to human sensibility.
This Kant tries to prove by showing that space and time are the
forms of our sensibility. His arguments to this effect are given in the
Transcendental Aesthetic. The reasons, therefore, which Kant may
have offered for holding that the a priori intuitions are related to
sensibility can only be given by the whole of Transcendental Aes-
21 Cf. Prolegomena §§ 7-8; also A 19. As pointed out by H. J. Paton (Kant's

Metaphysic of Experience, London, 1936, vol. I, pp. 93-94) the latter passage is
not only an explanation of Kant's concepts, but also a statement of what he wants
to prove concerning them.
thetic. This is indicated by Kant's own remarks at B xxv-xxvi and B
146.22
This view of the role of the Transcendental Aesthetic in estab-
lishing a relation between the two is verified by the way Kant actually
argues there for the subjectivity ('ideality') of space and time. Let us
take the so-called Metaphysical Exposition of space as an example. This
argument may be said to consist in three steps. First Kant tries to show
that space is a necessary idea a priori; then he argues that space is an
intuition; and from these facts he concludes that space is subjective in
origin. What interests us here is the question: What does he mean by
saying that space is an intuition? The answer is given by the way he
argues for his thesis that space is an intuition. It is an intuition because,
he says, 'we can represent only one space', because speaking of differ-
ent spaces only means speaking of different parts of one and the same
space, because, in a word, 'space is essentially one'. If possible, this is
clearer still in the inaugural Dissertation: "III, § 14, 2. The idea of time
is singular, not general . . . 3. The idea of time is therefore an intui-
tion." (Cf. § 15c: "The concept of space is thus a pure intuition, since
it is a singular concept.")
This shows quite clearly that in his actual arguments for the
subjectivity of space and time Kant by intuitivity predominantly
meant individuality. It shows that the connection between intuitions
and sensibility was, for Kant's purposes, first established by the argu-
ments presented in the Transcendental Aesthetic. In fact, the first time
sensibility is mentioned by Kant in his discussion of space is when he
draws conclusions from the facts he thought he had established, viz.
from the necessity, apriority, and intuitivity of space. The possibility
of a necessary intuition a priori, he claims, can only be understood by
assuming it to be subjective in origin, to be due to the structure of our
sensibility. Far from being one of the premises of Kant's arguments,
the connection between intuitions and sensibility is thus one of their
conclusions. 23
The view we have thus reached has a number of consequences
for any attempt to interp~et Kant's philosophy. It implies that one
must not presuppose any connection between intuitions and sensibility
. 22 Witness also ~ant's emphasis on the phenomenal character of space and

tIme as one of the matn results of the Transcendental Aesthetic (A 42 = B 59).
=
According to Kant's own explanations (A 19-20 B 33-34) this phenomenal char-
acter of intuitions amounts to their being due to our sensibility.
23 This fact is perhaps clearer still in those arguments of Prolegomena which
correspond to the Transcendental Aesthetic.
in interpreting those parts of the first Critique which precede the
Transcendental Aesthetic or which are presupposed in the arguments
given in the Aesthetic. I am not saying that Kant did not have the
connection in mind when he wrote these parts of the first Critique.
Most likely he did. But I am saying that in so doing he did not con-
tribute to the clarity of his argument and that any consistent attempt
to make sense of what he in fact did write must not assume the con-
nection before the arguments by means of which Kant thought he
established the connection.
Among the parts of the Critique of Pure Reason which precede
the discussion of space and time there is the section of the Einleitung in
which Kant explains his celebrated distinction between synthetic and
analytic judgements. In so doing he makes use of the notion of intui-
tion. If what was said above is correct, the notion of intuition must
here be taken in a purely 'logical' sense as referring to particular ideas
in contradistinction to general concepts and not in the 'psychological'
sense in which the term implies a connection with sensibility.
There is another important part of the first Critique of which the
same is true. This part does not precede the Transcendental Aesthetic
in the order of exposition followed by Kant in the Critique of Pure
Reason. There are ample reasons, however, for believing that it should
precede the discussion of space and time in the Transcendental Aes-
thetic if a logically correct order of presentation were followed.
I am referring. to Kant's theory of mathematics as it is presented
in the first part of the Transcendental Doctrine of Method toward the
end of the first Critique (A 712-738 = B 740-766). The priority of this
theory to the other form of Kant's theory of mathematics developed in
the Transcendental Aesthetic is shown by the following facts.
(1) The argument for the subjectivity of space which is called
by Kant the Transcendental Exposition presupposes the doctrine of
mathematics explained in the Methodenlehre. The argument proceeds
from the premises that geometry is based on the use of a priori intui-
tions and that the theorems of geometry have an apodictic character to
the conclusion that the a priori intuitions of geometry must be bound
up with the nature of our sensibility. The premises of this argument
are discussed and elaborated in the Methodenlehre. If the results of the
Transcendental Aesthetic were presupposed in the Doctrine of
Method, the Transcendental Exposition would be a petitio principii.
(2) This dependence of Kant's full theory of mathematics on the
preliminary doctrine set forth in the Methodenlehre is made explicit in
the Prolegomena-in the work whose professed purpose was 'das
Ganze zu iibersehen', to make clear the structure of Kant's argument
as a whole. In the Prolegomena, Kant refers to the discussion of
mathematics in the Doctrine of Method both in explaining the syn-
thetic nature of mathematical judgements and in the beginning of the
argument which corresponds to the Transcendental Exposition. 24
( 3) There are statements and whole passages in the Doctrine of
Method which are hardly intelligible if one assumes that by 'intuitions'
Kant means 'mental pictures'. For instance, Kant there explains the
nature of algebra in a way that makes it turn upon the use of intuitions.
This is very awkward to those who interpret Anschauung as a "mental
picture" of some sort, for whatever is the rationale of algebraic
symbols, it is certainly not to furnish us with more vivid images of the
things under consideration, or to make the discussion more 'intuitive'
in the ordinary sense of the word. The impossibility of understanding
Kant's views on algebra from the point of view of an inseparable
connection between intuitions and sensibility is perhaps best illustrated
by Professor C. D. Broad's interesting essay on 'Kant's theory of
mathematical and philosophical reasoning', which is based on the
assumption of a connection of this kind. 25 Broad points out, quite
correctly, that what Kant says of algebra could equally well have been
said by Leibniz and could with the same right be said of symbolic
logic. Kant does not establish any connection between algebra and his
theory of mathematics as being concerned with the forms of our sensi-
bility. In this sense, Broad can say that 'Kant has provided no theory
whatever of algebraic reasoning'.
This is a solid objection against Kant's full theory of mathematics
developed in the Transcendental Aesthetic. It leaves entirely incom-
prehensible, however, how Kant could have made statements that so
completely differ from his professed theories. It seems to me that
Kant's views on algebra can only be understood if the term 'intuition'
is in them taken in the purely logical sense without any reference to
sensibility. They belong, in other words, to the preliminary theory of
mathematics into which the subjectivity of space and time does not yet
enter and which forms the starting-point and the premises of Kant's
full theory. (Of course, this does not mean that they can be reconciled
with the 'full' theory.)
From the point of view of the preliminary theory, Kant's pro-
24 See the Academy Edition of Kant's works, vol. 4, pp. 272 and 281, and
d. p. 266.
25 Proceedings of the Aristotelian Society, vol. 42 (1941-42), pp. 1-24.
noun~ements on algebra become perfectly intelligible, whether they
are :Ig~t .or. wron?": ,they me~ely amount to saying that algebra is
nommahstIc m Qume s sense: Its symbols (variables) have individuals
(individual numbers) as their values, not classes or attributes of
numbers.
(4) In a group of excellent papers to which the present author is
indebted for important insights, Professor E. W. Beth showed that the
theory of mathematics presented in the Transcendental Doctrine of
Method is in many of its essential features already to be found in
Kant's pre-critical works. 26 It is hence historically prior to the 'critical'
doctrines of the Transcendental Aesthetic. The latter, therefore,
should already for this reason be read in the light of the former, and
not vice versa.
F or these reasons, we have to distinguish two different, though
not unrelated, levels of Kant's philosophy of mathematics. There is the
broader and vaguer doctrine which we have called the preliminary
theory. It constitutes the background and the starting-point of the
more typically Kantian theory which we have called his full theory of
mathematics. The fuller theory is mainly explained in the Trans-
cendental Aesthetic, in the Dissertation, in the § 9-13 of the Pro-
legomena, as well as in some parts of the Transcendental Analytic. The
most important sources of the preliminary theory are Kant's discussion
of the "pure reason in its dogmatic employment" in the Methoden-
lehre, the remarks on mathematics in the introduction to the first
Critique, the § 6-8 of the Prolegomena, and the pre-critical prize essay
of the year 1764. In the later writings, the different aspects of Kant's
philosophy of mathematics understandably tend to merge into each
other. The 'full' theory is the better known of the two. In fact, the
'preliminary' doctrines seem to be almost completely neglected. Beth's
papers, in which their significance is recognized, have not received the
attention they deserve.
The main difference which we have found between the two
levels of Kant's mathematical philosophy is that in the preliminary
theory no connection is assumed between intuitions and sensibility
26 See E. W. Beth, "Kants Einteilung der Urteile in analytische und syn-

thetische," Algemeen Nederlands Tijdschrift voor Wijsbegeerte en Psychologie,
vol. 46 (1953-54), pp. 253-264; "Critical Epochs in the Development of the Theory
of Science", British Journal for the Philosophy of Science, vol. 1 (1950); La crise
de La raison et La logique, Paris, 1957; The Foundations of Mathematiq, Amster-
dam, 1959, pp. 41-47; Mathematical Tbought, Dordrecht, 1965; and other papers
referred to in these works of Beth's.
while in the full theory Kant tries to show that all intuitions are
sensible. The distinction can be put in' other ways, too. One of the
central notions of Kant's philosophy of mathematics is that of con-
struction. 27 Following Broad, we may distinguish between symbolic
and instantial constructions. By a symbolic construction, I mean here
the production of a particular object to stand for a number of objects
with some common characteristic. Under certain conditions, we can
deal with this 'illustrative' construct as if it were one of the objects it
signifies even if it is not really one of them. For instance, we can
represent the body temperatures of a person by means of lines as in a
fever chart. There is no non-conventional connection between the
two; yet for certain purposes it is more facile to consult the fever chart
than to measure actual temperatures. The comparison of the lengths of
different lines is easier than the direct comparison of temperatures and
brings out the long-range tendencies better than the latter.
By instantial construction, on the other hand, I mean the produc-
tion of an actual instance of the objects under consideration. The
difference between the two levels of Kant's philosophy of mathematics
may be characterized by saying that in the preliminary theory con-
structions may be purely symbolic, while in the full theory Kant
makes an effort to narrow them down to instantial ones. For, according
to the full theory, "each of us carries about with him a mind-depen-
dent intuitum in which he can imaginatively construct without help
from sensible experience any figure that he may choose to define."28
The imaginative constructions of this kind exhaust all the geometrical
relationships, for these very relationships are created by the mind who
therefore has the power to instantiate all of them. 29
By distinguishing the two levels of Kant's philosophy of mathe-
matics, I am placing myself under the obligation of actually giving an
account of the two theories. This involves a reconstruction of Kant's
distinction between analytic and synthetic judgements (in mathe-
matics) in terms not including sensibility.30 It also includes the recon-
27 This term is defined in A 713 = B 741 and it is also used there to char-
acterize the kind of knowledge Kant thought of as being peculiar to mathematics
HB roa d ,op. CIt.
. .
29 This connects Kant's theory of space and time with his basic idea that
~uman ~eason :.'~a~ insight only in that which it produces itself after a plan of
Its own (B Xlll-XlV). Cf. my comments on the tradition which this basic idea
represents in "Kant's 'New Method of Thought' and His Theory of Mathematics"
Ajatus, vol. 27 (1965), pp. 37-47. '
30 I have sk~tched .the foun?at.ions of ~uch a reconstruction in a sequence of
four papers publIshed 10 Deskrtptton, ExtStenz und Analytizitiit, ed. by Paul
stru~ti~n of the basic features of Kant's preliminary theory of mathe-
matIcs 10 such a way that no connection between intuitions and sensi-
bilit,Y is assumed. I have tried to deal with some of these problems in
earlIer papers, and I hope to return to them in future. 31 In so far as
these later attempts of mine are successful, they will reinforce the
thesis of. the present essay concerning the lack of any direct conceptual
connectIOn between Kant's notions of intuition and his concept of
sensibility.
The difference between the two levels of Kant's mathematical
philosophy calls for a few further historical comments. The main
difference between them is, we have found, that in the preliminary
theory one is not allowed to assume any connection between intuitions
and sensibility, while in the full theory Kant tries to argue that all our
intuitions are sensible. The preliminary theory forms, we said, the
premises of the full theory. This may also be put by saying that for
Kant the preliminary theory of mathematics poses certain problems or
difficulties. And Kant thinks that the only way of solving these prob-
lems is to assume that all our intuitions are subjective in origin and,
more specifically, due to the structure of our sensibility.
From this it follows that we cannot at all understand Kant's
actual arguments without starting from a clear-cut notion of intuition
as a particular idea in contradistinction to general concepts and having
no necessary relation to the senses. For one of the main things he was
out to explain was how intuitions in this sense can be successfully
employed a priori in mathematics. The explanation consisted in relat-
ing intuitions to sensibility. Thus the explanation Kant gave for the
possibility of mathematics in his full theory served to erase the very
distinction from which he started. It largely consisted in assimilating
the logical sense of the term intuition in Kant to the naive psycho-
logical meaning of 'intuition' as an image or a mental picture or a
particular idea in Frege's sense of 'idea'. The choice of the term
'Anschauung' for a representative of a particular object was no doubt
to anticipate this 'solution'.32
Weingartner, Munich and Salzburg, 1966. See especially the paper entided "Kant
Vindicated", and cf. also my paper, "Are Logical Truths Analytic?", Philosophical
Review, vol. 74 (1965), pp. 178-203.
31 In addition to the papers already mentioned there is also a long paper in
Finnish, "Kantin oppi matematiikasta", Ajatus, vol. 22 (1959), pp. 5-85.
32 The "psychologistic" misinterpretation of Kant's term Anschauung was
already ripe at the time he was still writing, as witnessed e.g. by Herder's Meta-
kritik (first edition, pp. 53, 56, 60, 72 etc.).
This helps us to understand the differences between Kant and a
logicist like Frege. This psychologization of the concept of intuition
was one of the main historical outcomes of Kant's theory of mathe-
matics, although it was partly based on a misinterpretation of Kant. In
general, I think it fair to say that his theory of space and time opened
the doors for the wildest and vaguest psychologism in general logic
and, in particular, in the philosophy of mathematics that we can find at
any period of the history of logic. 33 It is this psychologism that Frege
was reacting against more than against Kant himself. This psycholo-
gism was, I think, one of the reasons why Frege was so bitingly critical
of his contemporaries and why he had to go to the very first funda-
mentals of logic to argue for his view. He found the prevailing ideal-
istic tradition completely useless. What I have been arguing here is that
Kant's starting-points were far less psychological than his final views.
The effects of the full Kantian theory were, I think, most unfortunate
for the history of logic and philosophy of mathematics. And among
the unfortunate and ironic effects of this part of Kant's philosophy was
that it served to confuse the very arguments and concepts by means of
which Kant tried to establish his theory.
APPENDIX
Our interpretation of the role of the notion of intuition in Kant is

further supported by an analysis of Kant's discussion of the so-called
'paradox of incongruent counterparts'. The alleged paradox lies in the
fact that there are objects in space-say my right and left hand-
which have exactly the same 'internal' properties but which neverthe-
less cannot be brought to coincide (to occupy the same region in
space) by any possible movement. Objects of this kind are, Kant
thinks, conceptually indistinguishable but nevertheless distinguished by
their incongruence.
We shall not ask here whether Kant is right in thinking that this
state of affairs constitutes a genuine puzzle or even whether his de-
scription of the 'paradox' is correct. What is interesting about the
alleged paradox from our point of view is the fact that at different
times Kant drew from it conclusions which are apparently contrary to
33 There were of course a few exceptions, for instance Bolzano who was
both .an anti-psycholog~st in his ?wn logical theory and also in many respects quite
well mformed concermng the hIstory of the concept of Anschauung. C. S. Peirce,
too, was remarkably accurate concerning the implications of the term in Kant, as
show? inter alia by his discussion of "Questions Concerning Certain Faculties",
question 1.
each other. In his paper on "The Ground of the Distinction between
different Regions in Space" of the year 1768, he thinks that the
paradox shows the existence of absolute space, while in the Pro-
legomena (1783) and in Die metaphysische Anfangsgrunde der Natur-
wissenschaft (1786) he claims that it shows the intuitive and subjective
nature of space. It has been pointed out often that especially the latter
use of the 'paradox' is, as it stands, a non sequitur. (See, e.g. C. F.
Gauss, Wissenschaftliche Werke, vol. 2, p. 177.)
The gist of the 'paradox' is shown by the use Kant makes of it in
the Dissertation. There the existence of "incongruent counterparts" is
given as one of the arguments for the claim that "conceptus spatii est
intuitus purus, cum sit conceptus singularis". Kant seems to have
thought that the fact that the "incongruent counterparts" have exactly
the same (internal) properties implies that they cannot at all be distin-
guished by means of general concepts. From this he concludes that
their difference is due to their relation to some third individual thing.
This individual thing is identified by Kant with space. Hence, in the
Dissertation he mentions the incongruent counterparts (without speak-
ing of a 'paradox') as a proof for the fact that space is an intuition in
his primary sense of the word, i.e. an individual thing in contradistinc-
tion to general concepts;
In 1768, Kant was mainly concerned with the dichotomy be-
tween Newton's theory of absolute space and Leibniz's theory of space
as the order of coexisting things. By showing as he thought he had
succeeded in doing by means of the 'paradox', that space is something
singular and yet different from all the individual bodies in space, he
thought he had shown the existence of absolute space, i.e. shown that
Newton was right.
In 1783, however, he had already hit upon a compromise between
the two positions. He had become aware of other possibilities than the
theories of Newton and Leibniz. Hence, he could no longer infer the
objective existence of absolute space from the individuality of space.
Meanwhile, he had also succeeded (to his own satisfaction, at least) in
connecting all our intuitions (singular con~epts) with s~~sibility. ~e
cause of this, he felt free to use the term Anschauung III the naIve
sense or, which amounts to the same, to infer from the "intuitivity" of
space in the sense of singularity that space is subjective or "ideal".
Unless the peculiar implications of the term "Anschauung" in the first
Critique and in the Prolegomena are recognized, we cannot at all
understand the connection there is between the apparently discrepant
consequences he drew from exactly the same premises in 1768 and in
1783.
TRANSCENDENTAL ARGUMENTS*l
Barry Stroud
In recent years there has been widespread use of arguments described

as Kantian or "transcendental" which have been thought to be special,
and perhaps unique, in various ways. What exactly is a transcendental
argument? Before looking closely at some specific candidates it will be
useful to see some of the general conditions which such arguments
must fulfill.
Kant recognized two distinct questions which can be asked about
·concepts. 2 The first-the "question of fact"-amounts to "How do we
come to have this concept, and what is involved in our having it?".
This is the task of the "physiology of the human understanding" as
practised by Locke. But even if we knew what experiences or mental
operations had been required in order for us to have the concepts we
do, Kant's second question-the "question of right"-would still not
have been answered, since we would not yet have established our right
to, or our justification for, the possession and employment of those
concepts. Although concepts can be derived from experience by
various means, they might still lack "objective validity", and to show
that this is not so is the task of the transcendental deduction.
For example, Kant considered it:
a scandal to philosophy and to human reason in general that the existence

of things outside us ... must be accepted merely on faith, and that if any-
one thinks good to doubt their existence, we are unable to counter his doubts
by any satisfactory proof. 3
The transcendental deduction (along with the Refutation of Idealism)

is supposed to provide just such a proof, and thereby to give a com-
• Reprinted by pennission of the author and publisher from the Journal of

Philosophy (196,8), pp. 241-256.
1 I am indebted to many friends and colleagues for their criticisms of an
earlier version of this paper. I would like to thank in particular Martin Hollis and
Thomas Nagel.
2 Kant, Critique of Pure Reason, tr. N. Kemp Smith, London 1953 A 84 ff.
. B xl.
aK ant, op. CIt., ' ,
54
Transcendental Arguments 55
plete answer to the sceptic about the existence of things outside us. We
c.an therefore. get some understanding of Kant's question of justifica-
tIon ~y lookmg at the challenge presented by the epistemological
sceptlc. 4
Since the. traditional epistemologist asks how it is possible to
know anything at all about the world around us, he is not interested
only in the specific question of whether there really is a tomato on the
table. Consequently, he will not be answered if we simply appeal to
one alleged matter of fact in order to support our claim to know
another. You cannot show the sceptic that you're not hallucinating,
and hence that you know there is a tomato on the table, simply by
asking your wife if she sees it too-hallucinations of your wife's re-
assuring .words are epistemologically no better off' than hallucinations
of tomatoes. At every point in the attempted justification of a knowl-
edge-claim the sceptic will always have another question yet to be
answered, another relevant possibility yet to be dismissed, and so he
can't be answered directly.
Doubts about whether some particular hypothesis is true can
often be settled by following the ordinary, well-known ways of estab-
lishing matters of so-called empirical fact. But the sceptic maintains
that the whole structure of practices and beliefs on the basis of which
empirical hypotheses are ordinarily "supported" has not itself been
shown to be reliable. As long as we have a public objective world of
material objects in space and time to rely on, particular questions about
how we know that such-and-such is the case can eventually be settled.
But that there is such a world of material objects at all is a matter of
contingent fact, and the sceptic challenges us to show how we know it.
According to him, any justification for our belief will have to come
from within experience, and so no adequate justification can ever be
given. Transcendental arguments are supposed to demonstrate the im-
possibility or illegitimacy of this sceptical challenge by proving that
certain concepts are necessary for thought or experience, but before
trying to see exactly how they are thought to do this it will be instruc-
tive to consider a possible objection to what has been said so far.
If transcendental arguments are meant to answer the sceptic's
question, and if, as many believe, that question makes no sense, then
there will be little point in considering the exact nature of these alleged
4 When I speak of "the sceptic" I do not mean to be referring to any person,

living or dead, or even to the hypothetical upholder of a fully articulated philo-
sophical position. I use the expression only as a convenient way of talking about
those familiar philosophical doubts which it has been the aim of the theory of
knowledge at least since the time of Descartes to settle.
arguments. This is reminiscent of the line taken by Carnap.5 He, like
Kant, distinguishes between two types of questions-ordinary em-
pirical questions on the one hand, which are raised and answered from
"within" a framework of concepts, beliefs, and recognized procedures
of confirmation, and, on the other hand, questions raised by the sceptic
or metaphysician about this framework, raised, so to speak, "from
outside". To ask whether there are any objects more than ten billion
miles from the earth is to ask an "internal" question to which there is
an objectively right answer. It is a genuine "theoretical" issue which
can be settled by discovering the truth of certain empirical statements.
But to ask simply whether there are any objects at all is to ask an
"external" question about the existence of the system of spatio-
temporal material objects as a whole, and this is not a "theoretical"
question with an objectively right answer at all. It is a "practical"
question, a request for a decision as to whether or not we should think
and talk in terms of material objects. Since there is no set of true
propositions which would answer an "external" question, the issue
cannot be settled by gathering evidence.
The belief that "external" questions must be answered in the
same way as ordinary empirical questions is what leads the epistemolo-
gist to the sceptical impasse. Carnap avoids scepticism by denying this
and claiming that statements like 'There are material objects' assert
nothing about the world at all, and hence that we couldn't conceivably
lack knowledge of their truth value. They have no truth value-they
merely serve to express a policy we have adopted, or a convention
with which we comply.
If this conventionalist line is to be successful there must be no
need for us to conceive of the world in terms of material objects in
space and time; it must be perfectly possible for us to find the world
and our experience intelligible in other terms. But transcendental
arguments are supposed to prove that certain particular concepts are
necessary for experience or thought; they establish the necessity or
indispensability of certain concepts. Therefore conventionalism of this
sort will be refuted if a sound transcendental argument can be pro-
duced. If there are particular concepts which are necessary for thought
or experience then it is false that, for everyone of our present con-
cepts, we could dispense with it and still find the world or our experi-
ence intelligible. A sound transcendental argument therefore would
5 R. Carnap, "Empiricism, Semantics and Ontology", Appendix A in M ean-

ing and Necessity, Chicago, 1956.
show that it is wrong to think (with the conventionalist) that the only
possible justification of our ways of thinking is "pragmatic" or practi-
cal, and equally wrong to think (with the sceptic) that they can be
justified only by collecting direct empirical evidence of their reliabil-
ity. Although these look like difficult demands to meet, they represent
the minimum conditions which Kant set for the success of a trans-
cendental argument.
Recent attempts to demonstrate the "absurd" or "paradoxical"
nature of sceptical questions have taken various forms. It has been
argued that seeing a tomato in the clear light of day, when other
people say they see it too, when I can reach out and feel it, is simply
what we call "finding out that there is a tomato there". This is the best
possibl~ case of knowing of the existence of a tomato, and since situa-
tions like this certainly do occur, it follows that we do know that there
are tomatoes, and hence that there are material objects. But from the
fact that this is the best possible case of knowing of the existence of a
tomato the most that follows is that "If this isn't a case of knowledge
of the external world then nothing is", or, in the more familiar ex-
ample, "If this isn't a case of acting of one's own free will then nothing
is". But the truth of such conditionals does not threaten the sceptic; it
is precisely because they are true that he is able to challenge all of
knowledge by considering only one or two examples. In addition to
establishing conditionals of this sort, then, one would also have to show
that it's false that there is no knowledge of the external world. But any
attempt to show that by an appeal to other empirical facts would lead
back onto the sceptic's treadmill. .
Defenders of the paradigm-case argument failed to see that the
sceptic need not deny that we can make all the empirical distinctions
that we do make (e.g., between what we call "hallucinatory" and what
we call "non-hallucinatory" perceptions), or that we all apply certain
concepts (e.g., "of his own free will") in certain circumstances and
withhold them in others. In Kant's terms, these are answers to "ques-
tions of fact" and so are not sufficient to answer the "question of
justification". It is not a sufficient refutation of the sceptic who doubts
that p to present him only with a conditional to the effect that if not-p
we couldn't possibly do A. What is in question is whether we ever
"validly" or "justifiably" do A. This is shown, in the extreme case, by
the obvious weakness of the argument which runs: If no-one ever
acted freely, then the ascription of praise and blame would be impos-
sible. But we do ascribe praise and blame. Therefore it is false that no-
one ever acts freely.
In order to demonstrate the absurdity of scepticism, the paradigm-
case argument had to rely on a theory of meaning to the effect that, at
least for some words, if those words are to have the meaning they do
have in our language, there must actually be things or situations to
which they have been, and perhaps still are, truly applied. If this were
true of the word 'X', for example, then from the fact that the question
"Are there really any X's?" makes sense it would follow that the
answer to it is "Yes". This has been thought sufficient to demonstrate
the "absurdity" of the sceptic's question. 6 But this theory of meaning
is highly doubtful, for reasons that will be given later. In the meantime
I shall examine some subtler and more persuasive recent anti-sceptical
arguments.
The first half of Strawson's Individuals, which is certainly Kant-
ian in ton.e, gives the impression of relying on transcendental argu-
ments to establish the absurdity or illegitimacy of various kinds of
scepticism. Strawson starts by saying:
(1) We think of the world as containing objective particulars in a single

spatio-temporal system.
He emphasizes that this is a remark "about the way we think of the

world, about our conceptual scheme",7 and he wants to discover some
of the necessary conditions of our thinking in this way. In discovering
these conditions Strawson claims to have demonstrated that the scep-
tic's doubts are illegitimate since they amount to a rejection of some of
the necessary conditions of the existence of the conceptual scheme
within which alone such doubts make sense. s This can be understood
in two ways, depending on what the sceptic is thought to doubt.
Strawson sometimes takes the sceptic to doubt or deny:
( 6) Obj ects continue to exist unperceived.
Only on this understanding of the sceptic is there any plausibility in

the claim that he is merely a "revisionary" metaphysician who rejects
6 See J. O. Urmson, "Some Questions Concerning Validity" is Essays in

Con~eptual 4nalysis, ed. A. Flew.. London, 1956, p. 120.. According to the still
fas~IOnable. VIew that all m~thematIca~ truths are true by VIrtue of the meanings of
theIr constIt~;nt words, ~hIS assumptIOn would also render ",absurd" all questions
of the fo~m Does 3695 tImes 1583 re~lly equal 5748785?". GIven the meanings of
the constItuent words and numerals, It follows that the answer is "Yes". Has the
question therefore been "exposed"as "absurd"?
7 P. F. Strawson, Individuals, London, 1959, p. 15.
sIbid., p. 35.
our conceptual scheme and offers a new one in its place. 9 But if the
sceptic doubts or denies (6), and if the truth of what the sceptic
doubts or denies is to be a necessary condition of those doubts' making
sense then Strawson would have to show that (6), a statement about
the way things are, follows from (1), a statement about how we think
of the world, or what makes sense to us. How could such an inference
ever be justified?
Strawson's argument is this. The sceptic's doubts about the con-
tinued existence of objects make sense only if (1) is true. But it is a
necessary truth that:
(2) If we think of the world as containing objective particulars in a single

spatio-temporal system then we are able to identify and reidentify particulars.
And again, necessarily:
(3) If we can reidentify particulars then we have satisfiable criteria on the

basis of which we can make reidentifications.
Strawson's argument actually stops here, thus showing that he regards

what has been established as sufficient to imply his diagnosis of
scepticism, but it is clear that it does not follow from (1)-(3) alone
that objects continue to exist unperceived. The most that has been
explicitly established is that if the sceptic's statement makes sense then
we must have satisfiable criteria on the basis of which we can re-
identify a presently observed object as numerically the same as one
observed earlier, before a discontinuity in our perception of it. And
this does not imply that objects continue to exist unperceived if it is
possible for all reidentification statements to be false even though they
are asserted on the basis of the best criteria we ever have for reidentifi-
cation. Only if this is not possible will Strawson's argument be
successful.
A principle that would explicitly rule out this alleged possibility
would be:
(4) If we know that the best criteria we have for the reidentification of
particulars have been satisfied then we know that objects continue to exist
unperceived.
Either this is a suppressed premiss of Strawson's argument or it is what

he means by "criteria for reidentification of particulars"-in either
9 Ibid., pp. 35-36.

case it is required for the success of his attack on scepticism. But the
argument now comes down to the claim that if we think of the world
as containing objective particulars then it must be possible for us to
know whether objects continue to exist unperceived. We couldn't
make sense of the notion of unperceived continued existence without
having criteria of reidentification, and if we have such criteria then we
can sometimes know whether objects continue to exist unperceived. I
shall call this result, which is the conclusion of the argument from (1)
to (4), the verification principle. If this principle is not true Strawson's
argument is unsound.
It does not follow from (1)-(4) that we actually do know that
objects continue to exist unperceived, and hence that (6) is true, but
that conclusion will follow if we add to the verification principle one
more premiss to the effect that:
(5) We sometimes know that the best criteria we have for the reidentifica-
tion of particulars have been satisfied.
The fact that (5) is needed shows that it was wrong to interpret
Straws on as making a purely deductive step from how we think, or
what makes sense to us, to the way things are. (6) is not a consequence
of (1) alone, but only of the conjunction of (1) and (5), and so there
is an additional factual premiss which enables Strawson to make the
otherwise questionable transition. And this in turn shows that Straw-
son was wrong to take the sceptic to be denying (6). If the truth of
what the sceptic denies is a necessary condition of that denial's making
sense, and if, as we've seen, it is not the case that the truth of (6) is a
necessary condition of the sceptic's making sense, then the sceptic
cannot be denying (6). On his grounds, to deny this would be just as
unjustified as our asserting it-he argues only that our belief that ob-
jects Gontinue to exist unperceived can never be justified.
If this is so, then the factual premiss which warrants the infer-
ence to (6) is obviously superfluous. The verification principle which
the argument rests on is: if the notion of objective particulars makes
sense to us then we can sometimes know certain conditions to be
fulfilled, the fulfillment of which logically implies either that objects
continue to exist unperceived, or that they do not. The sceptic says
that we can never justify our acceptance of the proposition that ob-
jects continue to exist unperceived, but· now there is a direct and
conclusive answer to him. If the sceptic's claim makes sense it must be
false, since if that proposition could not be known to be true or known
to be false it would make no sense. This follows from the truth of the
verification principle. Without this principle Straws on's argument
would have no force; but with this principle the sceptic is directly and
conclusively refuted, and there is no further need to go through an
indirect or transcendental argument to expose his mistakes.
Strawson's apparently more complicated account of scepticism
about other minds is essentially the same as this. In order for me to
understand, or make sense of, talk of my experiences, I must at least
understand the ascription of experiences to others. But it is a necessary
condition of my understanding this that I be able to identify different
individuals as the subjects of such ascriptions. And this in turn is
possible only if the individuals in question are such that both states of
consciousness and corporeal characteristics are ascribable to them. But
talk of identifiable individuals of this special or unique type makes
sense only if we have "logically adequate kinds of criteria" for ascrip-
tions of such predicates to them. Hence' "the sceptical problem does
not arise"-its very statement involves the pretended acceptance of a
conceptual scheme and at the same time the silent repudiation of one
of the conditions of its existence".l0 But what the sceptic "repudiates"
is the possibility of my knowing that there are any states of conscious-
ness other than mine, and so Strawson's characterization of the sceptic
is correct only if my possession of "logically adequate criteria" for the
other-ascription of a particular psychological state implies that it is
possible for me to know certain conditions to be fulfilled, the fulfill-
ment of which logically implies either that some particular person
other than myself is in that state or that he is not. This must be either a
suppressed premiss of Strawson's argument or an explanation of "logi-
cally adequate criteria".
As before, then, the sceptic is seen as maintaining both that (i) a
particular class of propositions makes sense and that (ii) we can never
know whether or not any of them are true. For Strawson the falsity of
(ii) is a necessary condition for the truth of (i), and the truth of (i) is
in turn required for the sceptic'S claim itself to make sense. Therefore
the success of Strawson's attack on both forms of scepticism depends
on the truth of some version of what I have called the verification
principle.
In Self-Knowledge and Self-Identify Shoemaker argues against
the other-minds sceptic as follows.!1 A person who understands 'I am
10 Ibid., p. 106.
11 S. Shoemaker, Self-Knowledge and Self-Identity, Ithaca, 1963, pp. 168-9.
in pain' cannot utter those words sincerely and without a slip of the
tongue unless he is in pain. Therefore, if it is possible to know whether
another person understands the word 'pain' it must be possible to know
whether another person is in pain. But the word 'pain' could not have
an established meaning if it were not possible for people to be taught
its meaning and possible for us to determine whether a person is using
it correctly. Therefore to assert, as the sceptic does, that it is logically
impossible for one person to know of another that he is in pain is to
imply that the word 'pain' has no established meaning. But if the word
'pain' has no established meaning, then the putative statement that it is
logically impossible for one person to know of another that he is in
pain has no established meaning either. Therefore, either what the
sceptic says has no established meaning, or it is false.
This _conclusion is the same as Strawson's, but in summarizing the
argument Shoemaker makes a further claim for it which appears to be
mistaken. He says:
Of any sentence that appears to say that it is logically impossible to know

that another person is in pain we must say either that it actually expresses no
statement at all or that it expresses a statement that is necessarily false.1 2
But it does not follow from the necessity of the conditional "if the
sceptic's statement makes sense, then it is false" that the sceptic's state-
ment is a necessary falsehood. Although Shoemaker does not go on to
draw any conclusions from this summary of the argument that do not
follow from the argument itself, later on he does claim that:
It is a necessary (logical, or conceptual) truth, not a contingent one, that

when perceptual and memory statements are sincerely and confidently
asserted, i.e., express confident beliefs, they are generally true. 13
One argument he gives for this starts out as follows:
(I) A primary criterion for determining whether a person understands such

terms as "see" and "remember" is whether under optimum conditions the
confident claims that he makes by the use of these words are generally true. 14
It is essential for anyone's using the words 'see' and 'remember' cor-
rectly-and hence for their having the established meanings they
have-that statements made by the use of those words be generally
true. Therefore, if perceptual and memory statements were not gen-
12 Shoemaker, op. cit., p. 170.

l3/bid., p. 229.
I4/bid., p. 231.
erally true then 'see' and 'remember' would not have the meanings
they appear to have, and there would be no perceptual or memory
statements.
To say that the words 'see' and 'remember' would not have the
meanings they do have unless the statements people made by the use of
those words were generally true is explicitly to rule out the possibility
of our understanding those statements when they are, unknown to us,
always false, or false most of the time, although they appear to be true
and hence we believe them. Therefore this argument too depends on
the truth of the verification principle. But more is needed in order to
prove that it is a necessary truth that perceptual and memory state-
ments are generally true. The most that has been established is that the
putative statement that it is not the case that perceptual and memory
statements are generally true is either false or meaningless. But this
alone does not imply that it is a necessary falsehood, and so does not
imply that it is a necessary truth that perceptual and memory state-
ments are generally true.
The rest of the argument is:
(II) So to suppose that (a) it is a contingent fact, which could be otherwise,

that confident perceptual and memory statements are generally true is to
suppose that (b) we have no way of telling whether a person understands
the use of words like "see" and "remember", or means by them what others
mean by them, that (c) we can never have any good reason for regarding
any utterance made by another person as a perceptual or memory statement,
and that (d) we could therefore never discover the supposedly contingent
fact that perceptual and memory statements are generally true. And this is
a logically absurd supposition.1 5
But the conclusion that it is a necessary truth that perceptual and

memory statements are generally true does not follow from this alone
because (b), (c), and (d) do not follow from (I) and (a). All that
follows is that it is a contingent fact that any person understands 'see'
and 'remember'. And that this is a contingent fact does not itself imply
that (b) we can have no way of telling whether it obtains or that (c)
we can never have any good reason for regarding any utterance as a
perceptual or memory statement, since the contingency of 'p' does not
in general imply that we can never find out that p. Without some
independent support for this last step the argument would fail. Given
(I), (c) and (d) do follow from the assumption that perceptual and
memory statements are not generally true, but they do not follow
15 Ibid., pp. 231-2. In this and the previous quotation I have inserted numerals
and letters into Shoemaker's text.
from the quite different assumption that it is a contingent fact that
perceptual and memory statements are generally true.
Shoemaker's independent argument is that in trying to discover
by inductive means the allegedly contingent fact that perceptual and
memory statements are generally true I could not rely on anything
that I believe on the basis of observation or memory. But there is no
other way in which I could come to know it, therefore I could never
know it. From the assumption (shared by the sceptic) that if it is a
contingent fact that p then our acceptance of 'p' can be supported only
by experience or by inductive means, and the fact that we could not
rely on perception or memory in order to establish that our perceptual
and memory beliefs are generally true, Shoemaker concludes that it is a
necessary truth that those beliefs are generally true. But this does not
follow, and the most that he has shown, as he himself sometimes points
out,16 is that a conditional statement to the effect that if . . . then
perceptual and memory beliefs' are generally true is a necessary truth.
What should the antecedent of such a conditional be? Shoemaker
says that "it follows from the logical possibility of anyone's knowing
anything about the world that perceptual and memory beliefs are gen-
erally true",17 but this alone raises no difficulties for the sceptic who
denies that we can know anything about the world. He too insists on
the truth of that conditional. It is no accident that those concerned
with all of our knowledge of the world have concentrated on percep-
tion and, to a lesser degree, on memory.
Rather than dealing with the conditions of knowledge, then,
those conditionals must assert that the truth of what the sceptic doubts
or denies is a necessary condition of the meaningfulness of that doubt
or denial. But even this could fail to be a conclusive refutation of the
sceptic. If only a restricted class of propositions is in question it is
always open to the sceptic to accept the argument and conclude that
talk about, say, the continued existence of unperceived objects really
doesn't make sense to us. Although he wouldn't, and needn't, say this
at the outset, he would be forced into it by an argument which relied
on the truth of the verification principle. Far from refuting scepticism,
this would make it stronger. Not only would we be unable to know
whether the proposition allegedly expressed by a certain form of
words is true, we would not even understand those words. 18 A suc-
16 Ibid., p. 238.
17 Ibid., p. 235.
18 That this result follows from. an applicati.on o! the ,:eri!ication principle
seems to me more an argument agamst the venficatton pnncIple than against
cessful anti-sceptica~ argument will therefore have to be completely
general, a~d deal wIth the ~ecessary conditions of anything's making
sense, not Just wIth the meanmgfulness of this or that restricted class of
propositions.
Furthermore, it won't be enough to deal simply with all of
language as it now is. David Pears described the conclusions of
Straws on's arguments as "conditional necessities" to the effect that
such-and-such is necessary if we are to think and speak as we now
do. 19 But even if such conditionals are true, it is still open to the
conventionalist to claim that no "theoretical" justification has been
given for our acceptance of the propositions the sceptic doubts or
denies, since we could simply give up our present ways of thinking and
speaking (of which they are the necessary conditions) and adopt
others (of which they are not). Transcendental arguments must yield
more than "conditional necessities" in this sense-they must make
these sceptical and conventionalist replies impossible.
Kant thought that his transcendental proofs counted in a unique
way against both scepticism and conventionalism because their con-
clusions were synthetic and could be known a priori. They are shown
to have this status by a transcendental argument which proves that the
truth of its conclusion is a necessary condition of there being any
experience or thought at all. If the conclusion were not true, there
could be no experience to falsify it. For Kant, proofs that such-and-
such is a necessary condition of thought or experience in general there-
fore have a special feature which is not shared by other proofs that one
thing is a necessary condition of another,20 and because they have this
feature they can answer the "question of justification".
scepticism. Ayer expresses a somewhat similar belief in discussing Strawson. See

The Concept of a Person and Other Essays, London, 1963, p. 110.
19 Philosophical Quarterly, 1961, p. 172.
20 "Through concepts of understanding pure reason does, indeed, establish
secure principles, not however directly from concepts alone, but always only in-
directly through relation of these concepts to something altogether contingent,
namely, possible experience. When such experience (that is, something as object
of possible experiences) is presupposed, these principles are indeed apodeictically
certain; but in themselves, directly, they can never be known a priori. Thus no
one can acquire insight into the proposition that everything which happens has its
cause, merely from the concepts involved. It is not, therefore, a dogma, although
from another point of view, namely from that of the sole field of its possible em-
ployment, that is, experience, it can be proved with complete apodeictic certainty.
But though it needs proof, it should be entitled a principle, not a theorem, because
it has the peculiar character that it makes possible the very experience which is its
own ground of proof, and that in this experience it must always itself be pre-
supposed." Kant, op. cit., A 737.
Suppose we have a proof that the truth of a particular proposi-
tion S is a necessary condition of there being any meaningful language,
or of anything's making sense to anyone. For brevity, 1 will say that
the truth of S is a necessary condition of there being some language. If
we had such a proof we would know that S cannot be denied truly,
because it cannot be denied truly that there is some language. The
existence of a language is a necessary condition of anyone's ever as-
serting or denying anything at all, and so if anyone denies in particular
the proposition that there is some language it follows that it is true.
Similarly, it is impossible to assert truly that there is no language. This
suggests that there is a genuine class of propositions each member of
which must be true in order for there to be any language, and which
consequently cannot be denied truly by anyone, and whose negations
cannot be asserted truly by anyone. Let us call this the "privileged
class".
There are some propositions which it is impossible for one par-
ticular person ever to assert truly. For example, Descartes cannot assert
truly that Descartes does not exist-his asserting it guarantees that it is
false. Also, there are some propositions which it is impossible for a
particular person to assert truly in a certain way, or in a particular
language. 1 can never truly say (aloud) '1 am not now speaking', but
everyone else can sometimes say this of me without falsity, and I
myself can write it or think it without thereby demonstrating that it is
false. Similarly, DeGaulle cannot truly say 'DeGaulle cannot construct
an English sentence', but anyone else can truly say this of DeGaulle,
and he himself can truly say in French that he cannot construct an
English sentence. Furthermore, there are some propositions which it is
impossible, not just for one person, but for any member of a particular
class of people to assert truly. A Cretan cannot assert truly that every
statement made by a Cretan is false-if he does assert this it must be
false-but of course any non-Cretan can assert this ,without thereby
guaranteeing its falsity. But the "self-guaranteeing" character of the
members of the privileged class is more general than that of any of
these. There is no one, whoever he might be, whatever language he
might speak, or whatever class of people he might belong to, who
c~~ld truly deny any of the members of the privileged class of propo-
SItIons.
Now no true proposition could be denied truly by anyone. But
for any proposition S which is a member of the privileged class, the
truth of S follows from the fact that somebody asserted it, or denied it,
or said anythin.g at .all, and this does not hold for all true propositions
generally. It mIght also be argued that since a necessary truth couldn't
be false under any circumstances, it couldn't be denied truly under any
circumstances either, and hence that all necessary truths belong to this
class. This might be so, but from the fact that a proposition is a
member of the privileged class it does not follow that it is a necessary
truth, and so it seems that there are some propositions, such as 'There
is some language', the truth of which is necessary for anyone's ever
asserting or denying anything, but which are not themselves necessary
truths. 21 It could have been, and undoubtedly was, the case at one time
that there was no language, and it probably will be :tgain. Although it
could not be truly denied,. still it might have been, and might yet
become false.
The existence of the privileged class is obviously important, since
if it could be proved that those propositions which the sceptic claims
can never be adequately justified on the basis of experience are them-
selves members, then from the fact that what the sceptic says makes
sense it would follow that those propositions are true. This would be a
way of replying to the sceptic while still acknowledging the con-
tingency of the things he questions. If those propositions could be
shown to belong to the privileged class there would appear to be no
more sceptical questions left open, as there are at every point when we
try to answer his questions directly. In general, giving an answer to the
question 'What are the necessary conditions of X?' does not tell one
way or the other about the answer to the question 'Do those conditions
obtain?' But in the special case of asking for the necessary conditions
of there being some language, giving an answer to the first implies an
affirmative answer to the second. One's asserting truly that the truth of
S is a necessary condition for there being some language implies that S
is true. Therefore there isn't another question about the truth-value of
S yet to be answered, and anyone who denied that we know it and still
demanded empirical evidence for its truth would have failed either to
have understood or to have been convinced by the argument. In either
case the proper reply would be to go through the argument again.
21 The tendency to confuse these two different kinds of necessity has seemed
an almost inevitable occupational hazard in transcendental philosophy, with its
claims to establish necessary or "conceptual" truths (cf. Shoemaker). If to say that
a proposition is "necessary" or "conceptual" is only to say that it must be true in
order for us to have certain concepts or for certain parts of our language to have
the meanings they have, then it does not follow that "necessary" or "conceptual"
truths are not contingent. Perhaps my privileged class will provide a way of keep-
ing these different kinds of necessity distinct.
The question now arises whether there is anything special, and
perhaps unique, about transcendental arguments even when they deal
with the necessary conditions of language in general, or of anything's
making sense. Is it only because Strawson's and Shoemaker's arguments
are limited in scope that they depend on an appeal to the verification
principle? There are some general reasons for being pessimistic on this
question. Although it seems to me unlikely that there should be no
members of the privileged class, we have yet to find a way of proving,
of any particular member, that it is a member. More specifically, we
have yet to show that those very propositions which the epistemologi-
cal sceptic questions are themselves members of this class. It is ob-
viously extremely difficult to prove this, and not just because talk
about "language in general" or "the possibility of anything's making
sense" is so vague that there seems to be no convincing way of decid-
ing what it covers and what it excludes. That is certainly a difficulty,
but there are others. In particular, for any candidate S, proposed as a
member of the privileged class, the sceptic can always very plausibly
insist that it is enough to make language possible if we believe that S is
true, or if it looks for all the world as if it is, but that S needn't actually
be true. Our having this belief would enable us to give sense to what
we say, but some additional justification would still have to be given
for our claim to know that S is true. The sceptic distinguishes between
the conditions necessary for a paradigmatic or warranted (and there-
fore meaningful) use of an expression or statement and the conditions
under which it is true.
Any opposition to scepticism on this point would have to rely on
the principle that it is not possible for anything to make sense unless it
is possible for us to establish whether S is true, or, alternatively, that it
isn't possible for us to understand anything at all if we know only what
conditions make it look for all the world as if S is true, but which are
still compatible with S's falsity. The conditions for anything'S making
sense would have to be strong enough to include not only our beliefs
about what is the case, but also the possibility of our knowing whether
those beliefs are true; hence the meaning of a statement would have to
be determined by what we can know. But to prove this would be to
prove some version of the verification principle, and then the sceptic
will have been directly and conclusively refuted. Therefore, even
when we deal in general with the necessary conditions of there being
any language at all it looks as if the use of a so-called transcendental
argument to demonstrate the self-defeating character of scepticism
would amount to nothing more and nothing less than an application of
some version of the verification principle,22 and if this is what a
trans7endental. argument is then there is nothing special or unique, and
certamly nothmg new, about this way of attacking scepticism.
What we need to know at this .point is whether or not some
version of the verification principle is true. It is not my intention to
discuss that issue now, but I do want to insist that it is precisely what
must be discussed by many of those who look with favour on the
much-heralded "Kantian" turn in recent philosophy. It could be that
we are not as far as we might think from Vienna in the 1920's.
For Kant a transcendental argument is supposed to answer the
question of "justification", and in so doing it demonstrates the "objec-
tive validity" of certain concepts. I have taken this to mean that the
concept 'X' has objective validity only if there are X's, and so demon-
strating the objective validity of the concept is tantamount to demon-
strating that X's actually exist. Kant thought that he could argue from
the necessary conditions of thought and experience to the falsity of
"problematic idealism" and so to the actual existence of the external
world of material objects, and not merely to the fact that we believe
there is such a world, or that as far as we can tell there is.
An examination of some recent attempts to argue in analogous
fashion suggests that, without invoking a verification principle which
automatically renders superfluous any indirect argument, the most that
could be proved by a consideration of the necessary conditions of
language is that, for example, we must believe that there are material
objects and other minds in order for us to be able to speak meaning-
fully at all. Those propositions about what we believe or about how
things seem would thereby have been shown to belong to the privi-
leged class. Although demonstrating their membership in this class
would not prove that scepticism is self-defeating, it would refute a
radical conventionalism of the kind outlined earlier. It would then be
demonstrably false that, for everyone of our present concepts, we
could dispense with it and still find our experience intelligible. But
until this much has been shown, not even part of the justification Kant
sought for our ways of thinking will have been given.
22 This suspicion is strongly confirmed by Judith Jarvis Thomson's excellent
account of the verificationism in Malcolm's argument against the possibility of a
private language (American Philosophical Quarterly, 1964). Stuart Hampshire's
discussion of the necessary conditions for any language in which a distinction can
be made between truth and falsity, while of the required generality, will have
force against scepticism only if it is interpreted as resting on a verification prin-
cipal (i.e., if in, ord~r for us to "succ~ssfully identify" an X, X's must act~~lly
exist). HampshIre hImself does not dtrectly apply the argument to SCeptICISm
(Thought and Action, London, 1959, chapter 1).
KANT ON THE PERCEPTION
OF TIME*
W.H. Walsh
This essay amounts to a commentary on some of the leading doctrines

of the Analogies of Experience, whose main contention I take to be
that(we should not be in possession of a unitary time-system unless
certain things were true, and indeed necessarily true, of the world of
experienced factJ A unitary time-system is one in which all temporal
ascriptions-all dates and durations-are directly relatable; it makes
sense inside such a system to ask of every supposed happening whether
it preceded, followed or was simultaneous with anything else which is
taken to happen. Kant assumes, obviously correctly as it seems to me,
that the temporal system we have at least purports to be unitary in this
way. He also assumes, again as I see it uncontrovertibly, that state-
ments assigning dates to events or durations to processes are (intended
to say something about the o!>j~~ivelworld, instead of to record what
particular persons happen to feel. We do contrast real with apparent
duration ("the struggle lasted for ten minutes, though it felt like an
age"), but it is the former which necessarily occupies our primary
attention, for {only if we first fix the real position of some things in the
temporal process can we speak effectively of the apparent position of
other thingsJ The real, here as elsewhere, is the normal, the apparent
the deviant, and you cannot understand the deviant until you grasp
that from which it deviates. pur chief aim in operating a system of
temporal concepts must accordingly be to say what objectively is the
case.)
It turns out, however, that the achievement of this aim is less easy
here than in some other instances. Direct perception, at least if we
correlate the data of different senses, or even of one sense at different
times, enables us to say that physical objects possess certain properties.
But time, as Kant himself is constantly saying, "cannot be perceived":
• Reprinted by pennission of the author and publisher from The Monist

(1967), pp. 376-396.
70
Kant on the Perception of Time 71
events do not come to us with their dates stamped on them, and the
fact that a precedes b in my experience does nothing to show that a
precedes b in reality. The special difficulties of the establishment of
objective time-determinations are such, Kant believes, that we can
make genuine temporal judgments only if the experienced world has a
certain necessary form.( It must be a world in which nothing is abso-
lutely created or absolutely annihilated, one where all change is
transformation) It must be a world in which events are not loose and
separate in the way Hume took them to be, but rather where the very
fact that something occurs means· that something else must have
occurred or be about to occur, i.e. one where there are necessary
connections between events. Finally, it must be a world in which
different physical things do not operate in causal independence of one
another, but form (part of a system all of whose members are in
thoroughgoing causal reciprocity) That these things are the case in the
world we have to deal with Kant says is not just a fact but a necessary
fact: it is bound up with our having a certain consciousness of time
whose characteristics we all recognise. 1
In what follows I shall be trying to sketch and evaluate Kant's
arguments for these striking conclusions. As he says himself, many
persons before him had accepted them as true, but few, if any, had
subjected them to serious examination and none had offered a satisfac-
tory proof of their validity. The 'philosopher' who, asked how much
1 The general principle of the Analogies is best stated in the formula Kant
uses in the first edition (A 176-7): "All appearances are, as regards their existence,
subject a priori to rules determining their relation to one another in one time."
I take it that the last three words here are the crucial ones. The fonnula in the
second edition ("Experience is possible only through the representation of a
necessary connection of perception~"-B 218) would apply to the Tra':lscendental
Deduction as well as to the AnalogIes, and does not brIng out the specIal concern
of the latter with dates and duration. Kant's reference in the first edition passage
to "existence" is in contrast to what he had tried to establish in the Axioms of
Intuition and Anticipations of Sense-Perception, where it was the internal struc-
ture of appearances which occul?ie~ his attention. In the Analogies he seeks ~o
show that, quite apart from theIr mternal structu~e, the very f~ct that. certam
items occur in our experience commits us to the belIef that other Items WIll occur
or have occurred, and so permits us to move necessa~ily from th~ exi~tence of .o.ne
thing to that of another. Naturally, he finds somethlllg paradoxlca.l III our abIlIty
to make demands on fact in this way; his solution to the paradox IS to argue that
~e are dealing not with an independently existi~g wo:ld! but .with one which. is
merely phenomenal) I s?all not be c~ncerned WIth th~s ISSU~ III the pr~sen~, dIS-
cussion but I try to brIng out Kant s caveat that he IS dealmg only WIth phe-
nomen~" or "appearances" by speaking of "the experienced world" or "the world
of experienced fact."
smoke weighs, replied that you could get the answer by subtracting
the weight of the ashes left from the weight of the wood burnt (B
228/ A 185) was in fact assuming the principle of the first Analogy; so
were the ancients when they produced their formula Gigni de nihilo
nihil, in nihilum nil posse reverti. Similarly the principle that nothing
happens without a cause is accepted as axiomatic by philosophers and
nonphilosophers alike in daily life. One can call attention to tl,te special
necessity we attach to such principles by pointing out that, if someone
says that something happened for no reason at all (without any
connection with anything that preceded) or that something may have
gone clean out of existence, this is taken as a joke: the implication is
that we are not prepared to subject principles of this kind to serious
doubt. But it is one thing to establish this as a matter of fact, and quite
another to find a justification for treating them in this way. What
makes Kant's position intriguing is just that he thinks he can provide
such a justification. Still more impressively, the justification he offers
of one principle is, as we shall see, closely bound up with those he gives
of the others: the three Analogies, although formally separate, belong
intimately together, with the result that the argument in anyone case
derives support from the argument in the others. This is not to say that
it has to be accepted, either as a whole or in part; it is merely to call
attention to an immediate point of strength in Kant's case and to
suggest that its rejection may involve more far-reaching consequences
than may at first appear.
Kant makes remarks both at the beginning and at the end of the
Analogies about the form of the proofs he offers( He maintains that
neither a conceptual nor an empirical proof is in point here, what is
wanted being rather a "transcendental" proof, "from the possibility of
expe~ience. "We need to start from the fact that we are able to apply a
certam system of concepts, and then ask what must be true if this
situation is to obtain. I have drawn attention elsewhere2 to some of the
difficulties in this notion, and shall merely say now that, for all Kant's
disclaimers{the suspicion must remain that he offers a series of analytic
arguments after all, built largely round his idea of what is involved in
being an event) I do not mean this to imply that his contentions are, in
my opinion, essentially arbitrary: he could and would assert that his
concept of an event was framed to fit the facts of our actual temporal
2 See my article "Philosophy and Psychology in Kant's Critique," Kant-

Studien,57 (1966), 186-198.
experience. But even if this is true, as no doubt it is, the form of the
argument would not be unique in the way Kant says it is .
. With this by way of preliminary let us now proceed to the

det~Ils of Kant'~ .cas:. Th.e reason why the Analogies, unlike the
AxIOms and AntIC~pa~IOns, Involve three special principles in addition
to the general pnnciple already referred to is that(time has three
modes, duration, succession' and coexistence) (B . 219/ A 177). The
obvious inference from this would be that the first Analogy is con-
cerned with the perception of duration. In fact, its purpose is wider
than this: Cit seeks to lay down a general condition which must be
satisfied if we are to have a single, continuing time-system, as opposed
to a set of particular temporal judgments which can be brought into no
relationship with one another) If all times are to belong to a single series
(and we all behave as if they must) we must believe,(kant maintains,
that there is something in the experienced world which endures
through all time, something which persists as the underlying substance
of things, though its manifestations or modifications constantly change.
At bottom, in this scheme of things, nothing is created or annihilated;
what is fundamentally there continues to exist, unchanged in quantity)
The idea is perhaps most easily made intelligible if we take it, as Kant
himself was inclined to do, in terms of the classical doctrine of matter,
the configurations of which were supposedly constantly changing
though it was itself indestructible and though its quantum in nature
was neither increased nor diminished whatever changes occurred. But
it is important to notice that all Kant needs to make good his point is
that there is something in the world of experience which endures
through all time; he has no call, nor indeed any alj'thority, to say what
that something is. To put it in his own terms: l that the concept of
substance has application in the experienced world is a truth we can
know a priori; what its application is we can find out only by empirical
mean4Kant's position about substance is in fact exactly parallel to his
position in the rest of the Analytic of Principles. In the Anticipations
of Sense-Perception, for example, he tries to show that every sensation
must have a determinate degree; only on this presupposition, he argues,
are we justified in asking quantitative questions about, e.g. the intensity
of an illumination or the depth of a colour. But he never pretends that
we can anticipate experience here in more than formal terms:(to find
out in what degree a sensation is present we need to have recourse to
experience) Similarly in this passage: to discover what form the
permanent takes we must go to the scientist, not the critical phi-
losopher. 3
In the summary proof he added in the second edition at the
beginning of the first Analogy (B 224-S{Kant first remarked that time
itself "remains and does not change,')since it is in time that all changes
must be thought to take place. But tIme itself cannot be perceived, and
hence it follows that "there must be found in the objects of percep-
tion ... the substratum which represents time in general." To say
the least, this is not very lucid. The case is put rather more convinc-
ingly in the opening paragraph of the first edition version, where we
read that, in order to determine whether "the manifold of appearance"
is, as "object of experience," coexistent or successive, "we require an
underlying ground which exists at all times, that is, something abiding
and permanent, of which change and coexistence are only so many
ways (modes of time) in which the permanent exists" (A 182/B
225-6). But for anything like an effective argument in support of
"Kant's conclusion we have to turn to the last pages of the first
Analogy. If we were willing to allow that "new things, that is, new
substances, could come into existence," we read in one passage (B
229/A 186), we should "lose that which alone can represent the unity
of time, namely the identity of the substratum." The reason for this, as
given in a later paragraph (B 231/A 188), is that "this permanent is
what alone makes possible the representation of the transition from one
state to another." "A coming to be or ceasing to be that is not simply a
determination of the permanent but is absolute, can never be a possible
perception" (ibid.). "If we assume that something absolutely begins to
be, we must have a point of time in which it was not. But to what are
we to attach this point, if not to what already exists? For a preceding
empty time is not an object of perception" (same paragraph). If sub-
stances could come into being or cease to exist, "appearances would
relate to two different times, and existence would flow in two parallel
streams-which is absurd" (B 231-2/ A 188).
I shall now attempt a free reconstruction of this somewhat
elusive line of thought, designed to bring out what I take to be its main
31t was in his Metaphysical First Principles of Natural Science (1786) that
Kant laid down ~s the "first law of mechanics" t.hat "in all changes of corporeal
natur~ t?~ qua?,tlty o~ mat.t~r on the whole remams unchanged, neither increased
nor dlmtnished .(~erhn ~dlt1on, IV, 54I). In.this w?rk Kant professes to apply the
~esults .of the CrItIcal phl!osophy, but not wIthout Introducing empirical concepts,
I? pa.rt1~ular th~t. of motIon: The concept of matter involves the concept of mo-
tlon tn Its defimnon, accordmg to Metaphysical First Principles, IV, 480.
points. Let me remark first that inspection of the whole passage reveals
that Kant is concerned not just with the unity but also with the
continuity of time, which he says can be assured only if we suppose
that the underlying substance or stuff of the experienced world-
whatever it is that undergoes change-persists unaltered in quantity.
Without continuity of substance in this sense we could not have
continuity of time. Why not? First, for the general reason that "only
the permanent can change":(we can take cognizance of alterations only
if we see them against a background that persists} If there were
nothing stable in our experience-if we lived in a world more Hera-
cleitean than that of Heracleitus, with everything 'flowing' at the same
rate-we could not even appreciate its instability.{Kant makes use of
this arg~ment in the second edition Refutation of Idealism when he
claims that knowledge of our inner states is possible only if we also
have outer experience.}The mental world, as Hume put it (Treatise, p.
252, ed. Selby-Bigge), is one where perceptions "succeed each other
with an inconceivable rapidity, and are in a perpetual flux and move-
ment." In these circumstances, to say no more than that different
perceptions are related as earlier and later, we require to be conscious
of something permanent, and this must be something outside the series
of perceptions (B 275-6). However, this argument alone will certainly
not give Kant all he wants in the first Analogy, for the persisting
things it demands need only be relatively persistent. We all know that
as a matter of fact a dating system is possible only because the physical
world contains relatively stable and long-lasting objects like the earth
and the sun; that there are such objects is, one supposes, an empirical
matter. But the first Analogy puts forward what is claimed to be an a
priori as opposed to an empirical requirement; the permanence or
persistence it speaks of is absolute, persistence through all time rather
than for a long time. Kant must therefore have, or suppose himself to
have, further arguments in support of his case.
In fact, these arguments are all indirect: they take the form of
asking what the situation would be if the principle of the first Analogy
did not hold. Suppose in the first place that the principle were to be
breached by the creation of a new substance, which would then
presumably manifest itself in what, in our present language, we should
describe as a series of happenings. (How should we integrate these
happenings with the rest of our experience?)How indeed should we be
justified in speaking of them as happenings at all? Since by definition
they would be accidents of a substance which was totally unrelated to
anything else in existence, their history would be separate from that of
the rest of the experienced world. We could not accordingly say
whether what happened to this new substance was happening before,
after or simultaneously with any other events; we should find ourselves
in the paradoxical position of having two wholly unrelatable time-
series, one which we had already, the other which we had newly
acquired. But would even this language be legitimate in the circum-
stances sketched? It is surely plain that it would not. We could not
date the emergence of the new substance, and in consequence would
not be justified in saying that the "new" time-series had "just been
acquired." We can say)hings like "At such-and-such a time Vesuvius
first erupted" because we can see this occurrence as part of the history
of the earth, which in turn is part of the history of the supposedly
persistent physical universe. But we could not say "At such-and-such a
time a new substance began to manifest itself," since ex hypothesi
other happenings would have no connection with this supposed event.
The first appearance of a new substance would accordingly be pre-
ceded by nothing but empty time which, as Kant says, is not a possible
object of perception.
Consider now the opposite contingency in which substance is
thought of not as being created but as annihilated. It might be sup-
posed that the annihilation of substance-its going clean out of exist-
ence-could be a datable occurrence, since it would follow on a
regular series of happenings in which something taken to be permanent
changed in an orderly way. Provided that events of this sort were
relatively rare-if transformation were the general rule and annihila-
tion only exceptional-we could at least recognise them as events. But
there are difficulties about this too. Unlike all other events, an occur-
rence of this sort would have predecessors but no successors; as the
'absolute termination of a series it could not be said to belong to the
history of anything. More seriously, it is hard to see how what we
should describe as subsequent to it could be said to be really subse-
quent to it; as an event with no outcome it could not be seen as a
regular part of the time-series. Once more, it would be followed by
.nothing but empty time, and as such could not be perceived.
Substance then must be taken as permanent because neither the
creation nor the annihilation of substance can be experienced. It fol-
lows that we must suppose that everything that happens must belong
to a single history, the history of eternal (phenomenal) substance.
Were this not so we should be without a framework inside which to
elaborate a unitary system of temporal relations. The position would
be that we should be presented with a series of distinct histories which
could be brought into no relationship to one another; the temporal
questions we now think it appropriate to raise on all occasions simply
could not be posed. It is important to observe that the same predica-
ment would threaten us if we followed the lead of some commentators
and argued that relatively persistent objects of the kind previously
referred to would be enough to meet Kant's requirements. That they
would not we can see by reflecting that there might be, and doubtless
are, relatively persistent physical objects in different galaxies. It would
be possible in these conditions for different groups of intelligent beings
to make temporal judgments which were wholly unrelatable; with
nothing more to go on than is here presumed we could not ask
whether events in the one galaxy were or were not simultaneous with
those in another. Kant is right in holding this to be a paradox, and in
arguing that we in fact make more extensive demands about the unity
and continuity of time than would be possible in the conditions
sketched. We need to provide for overall continuity, and that is why
we need not relative but absolute permanence in the substance of
things.
It scarcely needs to be emphasised that the substance for which
Kant argues in the first Analogy is not metaphysical substance. He is
not talking about things ·in general, but about the world of experience;
the characteristics of substantia phaenomenon, as he calls it (B 186/A
146), can accordingly be quite different from those of substance in the
metaphysical sense. Substantia phnenomenon, to mention one point
only, is something which essentially manifests itself in time, which
would certainly not be true of, for example, Leibniz's monads. Sub-
stance in the first Analogy resembles the substance of metaphysicians
in that it cannot be directly experienced; you can get at phenomenal
substance only through its accidents. But there the resemblance ends,
for whereas metaphysical substance is empirically inaccessible because
it transcends experience altogether, substance of the kind Kant here
postulates is inaccessible because it is not the sort of thi~g which could
be got at in itself. Just as one could not confront the socIal structure of
a community, but only discern it in the attitudes and actions of its
members, so one cannot experience phenomenal substance, but only
grasp it through its manifestations: In o~her wo~ds,( s~bstance .is an
organising concept, concerned to relate dIfferent Items In expenence'
But it is not any less respectable because of that fact.
I am inclined to think that this part of Kant's case is not only
plausible, but plainly correct. But I confess to two serious misgivings
about it. First, any attentive reader of the Analogies notices the way in
which Kant alternates between talk of substance in the singular and
talk of substances in the plural. The third Analogy in particular will
make sense only if we are permitted to speak of phenomenal sub-
stances. But what could these substances be? In the third Analogy they
seem to be large astronomical objects like the sun. But the sun is
certainly not an eternal object with a never-ending history: it came
into existence at a particular time or during a particular period, and
will cease to exist at some time in the future. At this date the material
of which the sun is composed will be transformed into something else.
Similarly with other suggested substances, where these are identified
with familiar objects whether large or small. But to attempt the
identification is in any case mistaken, for the reason explained in the
last paragraph: to speak of substance in the sense of the first Analogy is
not to speak of an item in the world. The transition from substance to
substances accordingly seems quite unjustified.
My second misgiving arises out of the awkward fact that there is
a well-known theory in cosmology which involves a doctrine of the
continuous creation of matter, and so seems to go directly counter to
Kant's conclusions. True, this theory is not universally accepted, but it
is for all that seriously discussed, which would hardly be likely if there
were some a priori objection to it. Defenders of Kant seem to me to
have two options as regards this theory. They might in the first place
claim that the continuous creation of matter is not the continuous
creation of substance in the sense Kant intends; on this I can say only
that it looks uncommonly like it. Alternatively, they can argue that
there is after all an a priori objection to the theory of continuous
creation, though its propounders, not having considered the questions
Kant raises, are not aware of this. My own inclination would be to fall
back on this second defence, and at least to require the theory'S sup-
porters to explain how on their view the unity and continuity of time
are to be safeguarded. But I admit that I feel an awkwardness in
supposing that men as intelligent as these can make the mistake which
on this interpretation they would be making.
I must now pass to the second Analogy, which I shall again

discuss only from a limited point of view, omitting any proper enquiry
into Kant's concept of causality and concentrating on the relevance of
his arguments to the perception of time. The first Analogy, as ex-
pounded above, sought to establish a general condition for the making
of temporal judgments: the continuity and unity of time had to have
their counterpart in the world of experience. As we saw, Kant himself
connected this requirement with the perception of duration. In the
s~cond Analogy he considers another mode or aspect of time, succes-
SIon, ~nd argues that we can say that one event really precedes another
only If there are necessary connections in the experienced world. The
connections in question are not the intelligible connections for which
:ationalist philosophers like Descartes sought; all that Kant is claiming
IS that, when an event occurs, there must be some preceding event
upon which it follows according to a rule. There is no question here of
our being able to attain insight into the workings of nature; in one way
these remain as 'secret' on Kant's view as they do on that of Burne.
But in another way the experienced world as Kant sees it is altogether
different from the world of Burne, for whilst in the latter all events
are loose. and separate and anything can, in principle, precede or follow
anything else/in Kant's understanding of the scheme of things events
are tightly li~ked together, and the temporal order, so far from being
full of contingencies, is determinate down to the last detail. Unless this
were true, Kant argues, we could never say that this objectively fol-
lowed that. )
To get a grip on this at first sight extravagant argument we need
to observe at the outset that Kant is here making a specific application
of something for which he had put up a case in the first edition version
of the Transcendental Deduction, that association presupposes affin-
ity.4(In claiming that the categories are necessary for experience he
naturally had to refute the suggestion that they are nO,thing but highly
general concepts empirically arrived at) Burne had invoked the psy-
chological machinery of association with a view of showing that this
was true of the key concept of cause. According to this way of
thinking, some of our 'perceptions' introduce others in a regular
manner-there are constant conjunctions in our experience-and this
leads us, when a perception of the first sort appears, to feel very
strongly that one of the second sort will follow.lWe imagine in these
circumstances that there is a necessary link between the one event a'nd
the other, but the necessity is in fact purely subjective) Kant's objec-
tion to this theory is fundamental: he denies that the process could
even start unless necessary connections were presupposed. If the
situation were as Burne describes it-if we had to deal with nothing
but perceptions occurring in individual minds-it would be impossible
to speak of one regularly following another. That our experiences are
4 See A 112 if., 121 if.

orderly depends on the fact that the world of events is orderly, and the
world of events must be separated sharply from the sphere of private
perceptions.(Experiences do follow one another in regular sequence,
but that is because necessary connections are already built in at this
level: events of their nature have necessary ties with what precedes
and follows~ Hume succeeds in extracting causality from experience
only because it is there already, and there not as a matter of fact but as
a matter of necessity.
Before trying to elaborate this further I should like to emphasise
another relevant respect in which Kant and Hume differ fundamen-
tally. Hume builds his whole theory of mind around the occurrence of
perceptions, which he divides into the two species impressions and
ideas. Asked to account for the peculiar nature of belief-to explain
what it is to assent to or dissent from an idea, as opposed to merely
entertaining it-Hume replies that it is to contemplate the idea in
question with a special sort of feeling. So far as I know, Kant nowhere
discusses this theory of belief directly, but everything he has to say on
the closely connected subject of judgment goes to show that he must
have rejected it in entirety'(In his account of the subject the central
aim in judgment is to say what is really the case, as opposed to how
things seem to the individual experient. Judgment aspires to express
truth, and what is true holds without distinction of persons)To see the
distinctive feature of judgment as residing in a privately experienced
feeling is from this point of view to pick on something utterly irrele-
vant. For even if it were in fact the case (as it is not) that whenever
we form beliefs we experience such feelings, they clearly have nothing
to do with belief as involving truth-claims.CT0 maintain that a certain
belief is true is to claim the ability to go beyond the impressions or
feelings of individual believers and state what must be accepted by
anyone who considers the facts)Hume displays enormous ingenuity in
trying to show how on his account of the matter we could still speak
of a 'system of realities'; the fact remains, even so, that his theory is
broken-backed from the start. By contrast, Kant brings the notions of
judgment and truth into prominence from the beginning of the
Critique, and in so doing avoids the fantasy, if also some of the charm,
of his great contemporary.
Whatever the merits of these general remarks, it is certainly true
that Kant's notion of judgment has to be kept constantly in mind in
reading the second Analogy. As everyone who has looked at that
section knows, Kant puts his problem as that of how we are to pass
from subjective to objective successions: he wants to know on what
conditions we ~an s.ay t~at one thing follows another "in the object,"
as oppose~ to In my mln.d or yours. Kant says notoriously that "the
apprehensIOn of the mamfold of appearance is always successive" (B
234/ A 189), but there was no necessity from his point of view to insist
on the "always." To" get the argument started it is enough to make the
modest claim that the order of our apprehension does not necessarily
coincide with the order of actual events; to point to the obvious fact
that we sometimes apprehend successively states of affairs which we
take to be really simultaneous, and at other times believe that the order
of our apprehension is the same as, or corresponds to, the objective
order of events. Weare set a problem by the circumstances in which
our experiencing takes place: we have to sort o~t what the world is
really like from how it merely appears to us, and this problem is as
urgent for time-relations as anywhere. Hume tried to hold on to
appearances, here as elsewhere, as the only palpable realities; the so-
called real world could in his view be no more than a necessary fiction.
Not the least of Kant's merits was that he saw that this involved a
reversal of the true order of things: the subjective could become
intelligible in the light of the objective, and not vice versa. The
conception of appearance makes no sense unless we have first given
sense to the conception of reality.
The solution Kant offers to his problem is a variant of the general
doctrine advanced elsewhere in the Critique which connects the
notions of objectivity and necessity. A subjective sequence is one
which is essentially arbitrary; it is like the connecting of two ideas by
association.(An objective sequence, by contrast, is not arbitrary but
necessary; it is like the connecting of two ideas by judgment~ There is
a sense in which what is true may be said to be compulsi~e for all
thinkers, whatever the nature of the content involved; everyone who
thinks rationally is under obligation to accept it. Kant may well have
been thinking of this sort of compulsiveness in the present discussion,
but it is certainly not all he had in mind. For he wants to explicate the
notion of an objective sequence as being necessary in a further,
internal sense, namely that it is one which takes 'place in accordance
with a rule. To understand this we must begin at the point he begins
himself, with the conception of an event as occurring at a determinate
place in time.(Events are, of tp.eir nature, not self-contained, but point
both backwards and forwards in the time-series) Nor is this the mere
tautology that everything present has a past and a future. Something
happens now because of something which happened in the past; its
place in the time-series is not accidental, but is due to the occurrence
of some preceding event. Because a thing of a certain sort happened at
time t\ a thing of another sort happens at time t 2 ; there is a rule
connecting the two occurrences. It is not Kant's doctrine that rules of
this sort can be discovered a priori. All that we can know in advance of
experience is that any event will point backward to some event in the
past and forward to some event in the future. But that this much can
be said is entirely certain, for only on these terms can we understand
what it means for one thing to succeed another as a matter of objective
fact, as opposed to in our private experience.
It should be observed that Kant is not committed to the impos-
sible proposition that every objective sequence is a causal sequence; his
own instance (B 237/ A 192) of the boat seen sailing downstream
clearly precludes this. There is no rule to the effect that when boats
are seen upstream they must subsequently be seen downstream. For the
sequence to be objective what is required is that it be causally deter-
mined in a more general sense: elements in the later situation must be
what they are because of the occurrence of the earlier one. Here as
elsewhere Kant's claims are more modest than they have sometimes
been taken to be. But the fact that they are modest when considered in
detail should not disguise their radical character when considered more
broadly. That we can know a priori that there are necessary connec-
tions between events is a sufficiently startling proposition even when
all the proper qualifications have been put in. And it must be empha-
sised that Kant is in no doubt either about its truth or about the
ubiquity of its application: he believes that it would not be possible to
claim that anything really preceded or followed anything else unless
every event in the experienced world pointed forwards and backwards
in the way we have described. The time-series as a whole must be fixed
in advance, with the position of the earlier members determining that
of the later, if we are to be able to make true judgments about succes-
sion. We all know the discomfort this result caused Kant when he
came to write his moral philosophy, but we are not concerned here
with this aspect of the matter.
In the Discipline of Pure Reason (B 815/ A 787) Kant says that it
is a "peculiarity" of transcendental proofs that "only one proof can be
found for each transcendental proposition." It is somewhat curious in
view of this that he offers what seem to be six or seven separate
arguments in support of the principle of the second Analogy. But the
diversity here is perhaps misleading: at bottom Kant relies throughout
on a single main line of thought. He moves from a formal feature of
time-the fact that past must precede present and present future is
what is taken (this is the all-important point) as a single continuous
series-to its counterpart in the real world, arguing that we could not
"empirically apprehend this continuity in the connection of times" (B
244/ A 199) unless "the appearances of past time determine all exis-
tences in the succeeding time." It seems dear that this argument has a
close relationship to the main proof of the first Analogy, and indeed
the difficulty is to make any sharp separation between the two. In the
first Analogy(Kant sought to demonstrate that whatever happens must
form part of the history of something which persists through all time
without increase or diminution; his emphasis there is on the unchang-
ing subject of which all events are the history) In the second Analogy
he shifts attention from the subject to its manifestations; the point
which now preoccupies him is the specific place of events in time. But
there is the same stress in the two passages on connectibility, and in
both the conclusion is drawn that to allow exceptions to the principle
argued for would jeopardise the continuity of time. Just how close the
two come can be seen if we reflect that, instead of ruling out absolute
creation as he does in the first Analogy by arguing that it would
involve unrelatable time-series, Kant could have considered the sub-
ject in the second Analogy and declared absolute creation impossible
on the ground that the first manifestation of a new substance would
not follow on any preceding event according to a rule and could not
therefore be said to have a determinate place in time. Absolute annihi-
lation could similarly have been proscribed on causal grounds: if
something went clean out of existence we should have an event which
had no effects, a possibility which Kant believed would have fatal
results for our perception of objective succession. From the point of
view of the second Analogy the creation of a new substance would
involve what was in effect a random occurrence, whilst the annihila-
tion of some existing substance would issue in what might perhaps be
called a random nonoccurrence. Both would involve inexplicable
breaches of regular temporal sequences, and as such would constitute a
threat to the very possibility of making true judgments about th;..
objective order of events. 5 .. -
5 It is interesting to observe that there is very little in the second Analogy to

correspond to the indirect arguments of the first (see above, p. 382). In B 239-
40/ A 194-5 Kant has a paragraph which begins with the words "Let us suppose
that there is nothing antecedent to an event, upon which it must follow according
to rule," in the course of which he maintains that "we should then have only a
play of representations, relating to no object; that is to say, it would not be possible
through our perception to distinguish one appearance from another as regards
Few philosophers today are prepared to take the threat just
spoken of with entire seriousness; indeed, the prevailing view is that
conclusions such as Kant tries to establish in the Analogies must be
viewed with profound suspicion. It is not denied that experience is as a
matter of fact full of regularities, as even Hume was prepared to
admit; what is questioned is the contention that we need to provide for
total regularity, on pain of losing all ability to discriminate the real
from the imaginary. Two quite distinct factors combine to build up
and sustain this attitude. First, the widely-shared empiricist prejudice
against principles which claim to be synthetic a priori: it is felt that
only someone deplorably ignorant of or insensitive to the most ele-
mentary distinctions in philosophy could commit himself to these.
Reflection on the special character of the principlesiKant advocates in
the Analogies will perhaps go some way to dispel lhis prejudice, for
whatever account we finally give of them they can scarcely count as
ordinary truths of fact, miraculously known apart from experience)
But even if this defence succeeds, the general attitude to this part of
Kant's work is not likely to change. For there is a second factor in
operation here, namely a persistent belief that Kant is making a lot of
fuss about nothing. He confronts us, on this account, with possibilities
which he himself describes as alarming, but which can be seen, if
contemplated coolly, to be nothing of the sort. Why should there not
be an occasional random occurrence, or an occasional random non-
occurrence for that matter? Experience shows that we can take a
certain amount of disorder in our stride: the aberrant can upset us, but
need not cause the catastrophic consequences which Kant says must
result if we take it seriously. The late Dr. Waismann is said to have
believed that a hammer he once possessed went clean out of existence:
what harm is there in admitting that he may have been right? Provided
that most substances persist, the annihilation of one here and there will
make no practical difference. And if most sequences are regular, an
occasional breach of causal law can be tolerated without difficulty.
The suggestion here is that there is enough persistence and
regularity in the experienced world to enable us to accommodate the
exceptions which Kant wants to rule out dogmatically. But is there?
relations of time." But there is no attempt here to discuss the case in detail. A
later passage (B 247/ A 201) speaks briefly of what would happen "were I to posit
the antecedent and the event were not to follow necessarily thereupon": I should,
Kant claiIh~, "haye to regard t~e succession as a me:ely subjective play of my
fancy; .and If I still represented It to myself as somethmg objective, I should have
to caUlt a mere drea~." But there is no detailed discussion in this case either.
The difficulty, as Csee it, is to set any limit to the number of excep-
tions, once their possibility has been admitted. To argue that as a
matter of fact the creation and annihilation of substance are rare, and
the reign of causal law nearly if not quite universal, will not provide
the necessary security. For even if it is true (and how we could know
it is not obvious) that exceptions to Kant's principles have up to now
been few, that will not prevent their occurring with far greater
frequency in the future. How far must this process go before we have
to confess ourselves totally baffled? If the reply is made that we are
concerned with real and not merely logical possibilities, and so can
safely discount any such contingency, the question can be asked
whether without it the situation as described is free of difficulty. On
the hypothesis under consideration things are occurring-few in num-
ber, admittedly, but occurring nevertheless-which we cannot inte-
grate with the rest of our experience: events which have no ante-
cedents, events which have no consequences, happenings that come
about for no reason at all. What is there in these circumstances to
distinguish these peculiar phenomena from total illusion? If some hard-
headed person of a scientific cast of mind were to pronounce them
entirely unreal, would there be any means of answering him? To take
this line is, of course, to subscribe to Kant's principle that only what is
connectible according to law is empirically real. Alternatively, an
attempt might be made to hold on to the reality of the phenomena
whatever the consequences: the effect of this, if it were seriously
persevered with, would be to cast doubt on what had hitherto been
taken as the system of realities. We cannot, in fact, do justice at the
. same time to those happenings which conform to rule and those
happenings which do not; it is a case of choosing the one or the other.
In this respect our position is like that Kant described when he spoke
of the creation of substance involving time flowing in two different
streams. And just as in that case, we should have no reason for prefer-
ring either to the other.
I do not myself believe that those who think we could (or can)
get on with a moderately disorderly world have thought through the
consequences of their hypothesis. One merit of Kant's discussion in the
Analogies is that it makes these consequences clear.
It remains to say something about the third Analogy, the part of

Kant's case which has received least attention from commentators.
Here Kant tries to do for coexistence what he had, to his own satisfac-
tion at least, already done for succession in the second Analogy. It
would be impossible, according to the argument of the latter, for us to
say that one event is really prior or posterior to another unless, in
general, events had necessary connections, the occurrence of one at a
determinate point of time necessitating. the occurrence of another at
some subsequent point. Irreversibility of perception is on this view a
sign of objective succession, but itself needs to be explained by some-
thing more fundamental, which turns out to be the ubiquity of causal
connections between earlier and later members of the temporal series.
Similarly in the third Analogy Kant starts from the reversibility of
perceptions as empirical evidence of coexistence, but argues that two
things can be judged to coexist only if a further and deeper condition
is fulfilled, namely that they should stand in mutual causal interaction.
The perception of coexistence is thus possible only if we can know a
priori an important and surprising truth about the world of experi-
enced fact.
The first difficulty in assessing this argument is to know what
Kant is talking about. In the second Analogy he was concerned with
the succession of events; here he speaks not about events being simul-
taneous but of substances coexisting. We have already seen that serious
problems are involved when Kant passes from the singular to the plural
in his discussion of substance: the first Analogy seems to argue for the
existence of a single continuing substance, and makes no provision for
this to exist in separate bits. 6 Moreover, the substances of the third
Analogy are perceived to coexist in space; they are, in fact, familiar
objects like the moon and the earth. Things of this sort are, of course,
only relatively persistent; they cannot, as we have seen, fulfill the
function in our knowledge of temporal relations which Kant assigns to
substance in the first Analogy. Nevertheless, it seems clear enough that
it is of them that Kant is thinking in this part of his work: he wants to
show that physical things are in dynamical interaction, and to maintain
that this is a necessary condition of our being able to make any
judgments about real coexistence.
How do we in fact know that the earth and the moon coexist?
We can look first at the earth and then at the moon, or we can look
first at the moon and then at the earth, and the order of our percep-
tions will be without effect on their content. But this, according to
Kant, would not suffice to show that the two are really coexistent. For
• 6 'Yhat m~kes this still more curi~us is that Kant was strongly opposed to
atomIs~ I~ phYSIcal the~ry. !;Ie thought It a mere prejudice to assume that matter
must eXist In packets WhICh dIffered only in size. See e.g. B 215/ A 173 ff.
something of the kind might be true if we each lived in a world of his
own private experience, in which everything would be what it seemed
to be and nothing could be said about objective dates. Just as the
perception of succession demands the universal operation of the cate-
gory of causality as regards successive members of any temporal series,
so the perception of coexistence demands that there shall be no
temporal series which are wholly self-contained. It cannot, for in-
stance, be the case that what we may perhaps call the life-histories of
the earth and the moon are each determined throughout by causal law,
but nevertheless remain entirely without influence on one another. For
if this situation obtained we should once more be in the position of
having separate temporal orders with no means of bringing them into
relationship with one another; in these circumstances we should not be
able to operate a unitary temporal system. The fact is, however, that
we do take our temporal system to be unitary, and must therefore
accept as true whatever is necessary for it to be so. In Kant's eyes this
means that we are committed to the category of reciprocity as well as
to the categories of substance and causality.
It might be thought that this argument evidently claims too
.much. Many very different kinds of event are thought of as happening
at the same time; many different kinds of substance, in the loose sense
of 'substance' used in the third Analogy, are taken to be coexistent. If,
to take an instance, Mr. Harold Wilson coexists with the Taj Mahal,
must we suppose them to be in thoroughgoing causal interaction? It
should be noticed, however, that Kant says only that "each substance
. . . must contain in itself the causality of certain determinations in
the other substance" (B 259/ A 212; my italics). I take this to mean
that the Taj Mahal need not affect the whole of Mr. Wilson's life-
history, nor vice versa; the influence of the one on the other need not
be significant, provided it is real. That it is real, though slight, insofar
as Mr. Wilson and the Indian monument are both physical bodies,
would be generally admitted. The point Kant wants to add is that it
must be real.
The argument here becomes altogether more plausible if we
observe the emphasis placed in the third Analogy on substances
coexisting in space; the causal interaction Kant postulates is clearly
between objects in a physical universe, united by, for example, gravita-
tional force. But though this illustrates what Kant was after, I do not
think it necessarily exhausts it. Just as in the first Analogy the notion
of continuing substance can, but need not, be illuminated by referring
to indestructible matter, so here reference to the dynamical commu-
nity of objects in a gravitational system is helpful but not compulsive.
If scientists have abandoned the conception of matter as Kant himself
understood it, the argument of the ~rst Analogy is not invalidated, for
it requires only that there be something in the experienced world
which persists through all change; as was emphasized earlier, it is for
scientists to say what form it will take. Similarly in the third Analogy.
It could be that physical objects have to be thought of in ways which
were not suspected in Kant's day, and that the principles which unite
them in a single physical system are very different from what men
thought them then. But even if this is true it does not alter the situation
in essentials, for there will still be an a priori reason, connected with
our perception of time, for supposing that nothing in the universe can
be totally independent of anything else. How the concept of reci-
procity applies must be found out empirically; that it applies can be
shown from the principles of the critical philosophy.
Is this the beginning of Naturphilosophie? Historically no doubt
it is. We all know that the Critique led on to the Metaphysical First
Principles of Natural Science, and that the latter was the starting-point
for the wild speculations of the Opus Postumum. That Kant, for all his
suspicion of metaphysics, had a taste for the constructivism fashionable
in his later years could scarcely be denied. Nor is it easy to refute the
suggestion that he might never have started on this slippery slope if he
had not believed himself to have a good case in the Analogies. But
however regrettable the later steps in this progress, the fact that they
occurred cannot in itself discredit its beginning. I suggest that the
argument of the Analogies deserves attention for its own sake, and that
its conclusions, which are both clearly stated and closely reasoned,
cannot be set aside for any general reason, as that they involve a claim
to intellectual intuition (which they do not) or conflict with the plain
truth of empiricism. If they are to be refuted at all, they must be
refuted on their own ground and in their own terms. And the critics
must tell us how the problems about continuity, succession and
coexistence which Kant raises are to be· solved, if they are not to be
solved along Kantian lines.
THE SECOND ANALOGY AND THE
PRINCIPLE OF INDETERMINACY*
Lewis White Beck
In classical physics, given a specification of the relevant parameters of

an event El and an appropriate law L, it should be possible to predict
the parameters of an event E 2, in a force-free field, with unlimited
exactitude and certainty. For example, if £1 is the movement of a body
of mass m with velocity v at time tl and position PI, the laws of
mechanics tell us when E 2, the passing of this body through P2' will
occur.
In quantum physics this is said to be impossible when dealing
with subatomic events. It may be impossible for one or both of two
reasons. First, it may be impossible to determine all the parameters of
the two events with sufficient exactitude to foretell precisely where
event E2 will occur or to know whether E2 did occur at the point
predicted; the maximum exactitude attainable is precisely determined.
Or, second, the law relating El and E2 may be only probabilistic, so
that E1 and the law L are not sufficient conditions for the exact predic-
tion of ~, even assuming the parameters of E1 and ~ are precisely
known.
Kant's theory of causation is designed to give epistemological
defence of the claim sketched in the first paragraph. Just as the
development of non-Euclidean geometries has either modified or re-
futed Kant's theory of mathematical knowledge, it may be thought
that the new development in physics known as the principle of
indeterminacy forces a major revision in his theory of physical causa-
tion if it qoes not, indeed, render it wholly indefensible. The purpose
of this paper is to examine the relation between Kant's theory of
causation which was fitted to the Newtonian physics and the features
.. Reprinted by permission of the publisher from Kant-Studie71 (1966),

pp. 199-205.
89
of indeterminacy or uncertainty which are integral parts ~f ~odern

physics. More specifically, .we shall ask whether th.e .prmclple. of
indeterminacy, as sketched III the second paragraph, IS mcompanble
with the Second Analogy of Experience, the putatively a priori
synthetic judgment that "all alterations take place in conformity with
the law of the connection of cause and effect". While there is a clear
contradiction between them, we shall find that the Analogy, properly
interpreted and supplemented, is required for the establishment of the
principle of indeterminacy itself.
II
In the Second Analogy, Kant is talking of empirical objects like

ships and houses, not of the particles of microphysics. But there is
every reason to believe that he would gladly extend his principle into
the most recondite parts of microphysics, and it would be a poor
defence of his principle to say that it was meant to apply only to
middle-sized objects. In the range of middle-sized objects the conse-
quences of the principle of uncertainty are so minute as to be un-
noticeable; hence in spite of Kant's Newtonism, it would be entirely
possible to draw a line between the region of objects where the Second
Analogy is applicable and the region where the principle of indeter-
minacy is applicable. But since the Analogy was meant to be one of the
foundation stones of physics and to be of universal and necessary
application, little of the Analogy would be left. It would then appear
to be a mere empirical fact that some events are causally related, but
not others, when, in fact, the Analogy is meant to be a priori and the
principle of indeterminacy is intimately involved in the whole theo-
retical structure of quantum theory.
Rather than trying to defend the Analogy by limiting its claims, I
wish to defend itl by showing that it is required in arguments designed
to establish the indeterminacy principle itself, and that the limits of the
application of the indeterminacy principle vis-a-vis the application of
the Analogy are not dependent just upon physical and statistical laws
but have an epistemological basis.
1 This paper is not meant to provide a general defence of the Second Anal-
ogy. My purpose is to show only that the function it served in the epistemological
foundations for the Newtonian theory is involved also in establishing t;he evidence
for the non-Newtonian theory; that is, I propose to show only that the need for
the Second Analogy is not reduced by the success of the indeterminacy principle,
and that the scope of the Analogy is not thereby limited.
The Second Analogy 91
The principle of indeterminacy holds that the most complete
knowledge we can have of E1 is not a sufficient condition for the
prediction that E2 will occur, but only that the knowledge of £1 gives
a probability that E2 will occur. Hence E1 could occur without E 2, and
E2 could occur without or before E 1. If the Second Analogy is correct,
however, and if E1 and E2 could occur one without the other, there is
no way in which we could determine which of them occurs first when,
in fact, both do occur.
F or the Second Analogy argues that since all representations are
successive even when they represent simultaneous or permanent states
of affairs, there must be something other than the successiveness of the
representations that serves as evidence for the successiveness of the
respective states of affairs. The event of representation R1 being
succeeded by R2 is not a sufficient condition for saying that the states
of affairs represented, Sl and S2, are successive, or that there is an event
(the transition from Sl to S2) instead of a continuing complex state of
affairs in which Sl and S2 are ingredients manifested one after the
other.
This analysis of the meaning of and conditions for the assertion
that an event occurs other than the transition from R1 to R2 is the
central point in what I consider to be Kant's principal answer to Hume
(though Hume is not mentioned) at A 195-6 = B 240-1. There Kant
tells us that the Humean theory that the causal principle is abstracted
from similar event-sequences is incorrect because the distinction be-
tween event-sequences and mere sequences of representations (which
is all Hume has a right to claim to know) itself requires the principle
that the object of one representation must precede the object of the
other. But this principle is equivalent to the causal principle itself.
Hence in supposing that we observe sequences of events and then
come to know by generalization that the earlier event is the cause of
the later, Hume put the cart before the horse. We do not know that
we are cognizing events except when we know that events are caus-
ally related in a way in which simultaneous states of affairs are not
causally related.
The difference between Hume and the proponents of the inde-
terminacy principle is this. Hume supposes that in event-sequences ~1-
~, E/-E2', etc., events like E1 will always be followed by events hke
Ez, though he can give no reason why this should be so or how we can
know it if it is so. The indeterminist denies this supposition, and hence
denies the principle of uniformity and the ideal of perfect predict-
ability. But their resemblances are greater than their differences, for
both assume that we can determine the difference between the follow-
ing cases: a) the series RCR2 when the R's are representations of
events and when their sequence is evidence of the event-sequence E 1-
E2; and b) the series R1-R2 when the R's are diverse representations of
permanent or simultaneous states of affairs so that the R-series is not
evidence for the E-series. Hence both Hume and the indeterminist are
committed, if Kant is correct, to the Second Analogy in establishing
the occurrence of an event-series. That Hume thinks the event-series is
itself causal and that the modern physicist thinks it is not are secondary
to their fundamental agreement that events can be distinguished from
continuing states of affairs revealed seriatim.
It would appear also that the differences between Hume and
Kant are much less significant than those between Kant and the
modern physicist, since the former two in fact agree that the event-
series is causally deterministic and they dispute only about our way of
knowing this (whether it be-a posteriori or a priori). It would seem
that Kant and the physicist cannot agree at all, since the former holds
that events can be recognized as events only if they are causally
related, and the latter holds that there are events not causally related
but only statistically related yet recognizable as events.
The question then arises, how does the physicist know that E2
temporally follows E1? More fundamentally, how does he know that
his representations R1 and R2 are representations of events, if the
events in question are not causally related? For if they are events, and
if the Second Analogy is correct, they must be causally related 2. If it is
asserted that they are events, and denied that they are causally related,
obviously the Second Analogy must be denied.
III
Let us examine the experimental situation, to be sure idealized

and simplified, which would give evidence for the principle of indeter-
minacy. We set up a clock on which we can read times t 1, t 2, etc.,
• 2 This is not stricdy true, since the Analogy (being regulative, not constitu-
tIve) does not tell us what events are causally related with each other, nor does it
guar~ntee .that a pair of events picked out by induction are in fact a causal pair.
But If £.. IS an event, the Analogy tells us that there is some other event which is
causally related to it, and it tells us how to proceed in finding that other event.
F~r t.he purl?ose of s!mplificati.on in this ex,?osition, I ignore this fact, since the
prmciple of indeterminacy demes that there IS any other event E which is related
to £.. as a classical cause.
when the hand points to positions C1, C 2, etc. We call the successive
positions of the hands the clock-series. Whenever we observe a certain
flash of light. (In classical theory, we should always see another flash at
naturally of a scintillation counter), we set the clock at C 1 and say the
flash occurred at t 1. We then find that when the clock is at C2 and
when we say the time is t 2, in some fraction of the cases we see another
flash of light. (In classical theory, we should always see another flash at
t 2; it is an empirical fact that we do not.) We interpret the flash at t1 as
evidence for a subatomic event E1 and the flash at t2 as evidence for
another subatomic event E 2, and deny that E1 is the cause of ~ because
it is not perfectly correlated with it.
The question then is, in the light of the Second Analogy, why do
we say _.E1 and E2 are events? The flashes of light are like Kant's
representations; they must be successive if they occur at all. For all we
know, the flash like that at t2 could occur before the flash like that at
t 1; and if it did, reasoning according to' the Second Analogy would
show that what was represented by the flashes (viz., the subatomic
states of affairs) were not events at all. In that case it is utterly trivial
to say that they are not causally related. The indeterminist means to
assert far more than that, namely, that they are events not causally
related.
It is here that the Second Analogy must be called upon by. the
indeterminist. The clock, as a middle-sized object, must have a fixed
order in its readings so that the reading C1 must occur before the
reading C2, regardless of whether the flashes F land F 2 which are
usually simultaneous with C1 and C 2 respectively are invariably associ-
ated with the clock-series. The decision that the subatomic states of
affairs we call E1 and ~ are events not causally related to each other
depends upon a prior decision that the states of affairs we call events in
the clock-series are causally related and hence unvariably associated
with each other in a fixed order, and upon empirical fact that the F-
series is not invariably correlated with the C-series. But we need still
another principle (which in some cases is empirical, in others a postu-
late) to associate the E-series with the F -series.
That an additional principle is needed is easily shown by a simple
example. Suppose I have a clock which is set to read C1 when I see a
cannon firing in the distance, and I find that I always hear the explo-
sion when the clock-event is C 2. This does not indicate that there is an
objective sequence of events: flash of light, then sound of explosion,
instead of a complex situation involving simultaneous light and sound
(like Kant's house, in contrast to his ship). We need to assert another
postulate in order to make sure that the sequence of events simultane-
ous with successive clock-events is evidence for a sequence of events
said to be correlated with them. We need, in other words, to postulate
a specific connection between the F -series and the E-series.
This postulate is: The temporal relation between the clock event
C1 and the state of affairs E1 giving rise to a report Flat C 1 is the
same as that between C2 and E2 when F 2 is made at C2.
This postulate is not fulfilled in the example of the sight and
sound of the cannon; hence the sequence of F's is not evidence of a
sequence of E's in the firing of the cannon. Nor is it fulfilled in the
example of the discovery of different reaction times of different
astronomical observers. In these examples, we use our knowledge of
the states of affairs giving rise to the reports at C1 and C2 to determine
the difference between the temporal relations of the states of affairs
and the F 1 and F 2 occurring at C1 and C2 respectively. (Thus we use
our knowledge of astronomical regularities to measure psychological
reaction-times, and our knowledge of the nature of explosions and of
the speed of light to determine the velocity of sound.)
If the postulate is not fulfilled but we can measure the differences
between the time-interval F 1-E1 and the time-interval F 2-E2, we can
generalize the postulate and still use it for establishing the objective
order in the E-series; the postulate as stated is merely the limiting case.
The complete generalization of the postulate in the special theory of
relativity is needed in order to correct errors which arise from neglect
of the speed of light in a signal arriving at C that an event E has
occurred elsewhere. In its complete generality, however, the postulate
is only a postulate, not directly testable empirically, but acceptable as a
decision made for the sake of giving an order to the E-series which is
independent of the position and movement of the observers with
different clocks.
In the experiment on the unobserved subatomic events, we
cannot show that our postulate is fulfilled by any independent observa-
tion of the date of the members of the E-series, for any attempt to do
so again involves us in the correlation of C-series, F -series, and E-series.
Hence this postulate (perhaps, however, in a more generalized, but
equally a priori form) must be, assumed if we are to reason from the
sequence of F -events to the sequence of E-events. When we assume
the Second Analogy in order to fix the sequence of the C-events and
thereby the sequence of F -events, and when we assume the postulate in
order then to serialize the objective states of affairs as E-events, it
remains a merely empirical question whether the E-events are invari-
ably related to each other or not.
Kant thought that they were, for he had no reason to doubt it in
1781; it is now denied on good empirical grounds. But it is denied only
after the Second Analogy has done its work in setting up a temporal
order among C-events and after the postulate has done its work in
synchronizing E-events with F -events which have already been syn-
chronized empirically with C-events. Therefore I suggest that there
are good epistemological grounds for regarding our knowledge of
indeterminacy as parasitic upon our knowledge of causal determinacy.
Without the causal determinacy of middle-sized objects, as asserted in
the Second Analogy, I do not see how we could get the evidence we
have for non-causal relations among microscopic objects.
In conclusion, I wish to point out an analogy (in the ordinary,
not the Kantian, sense of this word) between some of the "transcen-
dental apparatus" of the Kantian philosophy and the apparatus used in
a scientific laboratory. It has often been pointed out how the "forms of
intuition" resemble "colored spectacles" in their function of determin-
ing the content of our sensibility; this analogy may be of some
pedagogic value (but it has also dangers), but it does not cut very
deeply. On the other hand, it may be noticed here how the flashes
synchronized with clock-readings function like the representations
which are the raw data with which Kant begins. They are and must be
successive; but what they mean can be inferred to be successive only
by means of the Second Analogy and a postulate which involves the
formal features of signals. Scientific instruments like clocks and rulers
are constructed in order to do things to raw data very much like what
forms and categories and schemata do to or· with the contents of
sensibility in Kant. There is an analogy between instruments which
reduce the "rhapsody of sensations" to data in a scientist's notebook
and the transcendental syntheses which makes it possible to "spell out
the appearances so that they may be read as experience". Since Kant
wishes to avoid psychologism, but necessarily uses language borrowed
from psychology, he sometimes obscures his arguments and purposes,
and seems to be describing how we in fact think in the ordinary affairs
of life when what he says is more lucid and defensible when inter-
preted as an account of what a sophisticated scientist would do in
reasoning from his data to the objects of experience he is interested in
constructing logically. We do not normally think by rule but, at best,
in accordance with them; but when one reasons from clocks and
flashes of light to unobserved events occurring somewhere else, knowl-
edge of the construction and function of the instruments used plays a
constitutive role in the establishment of empirical facts. The termi-
nology of "forms of intuition" suggests much less about the truths of
physical geometry than discussion of the construction of rulers and
clocks; discussion of the ways in which successive flashes of light and
the movement of a clock hand indicate or fail to indicate successive
events generating these flashes provides a lucid and sophisticated
account of scientific procedure which is, I think, relatively free from
the dubious psychological assumptions (e.g., that of the non-simul-
taneity of representations) which underlie Kant's exposition of the
Second Analogy.
I do not propose that such analogies get to the root of Kant's own
meaning; after all, something very like his Analogies is required at the
next lower level in determining the conditions under which we can see
whether the flashes of light and the readings on the clock are simul-
taneous or not. But if we decide to shift to a protocol language of
physics instead of using a phenomenalistic protocol language and build
our science on the former rather than on the latter, the rules, forms,
and concepts by which Kant moved from the latter to a scientific
world picture are, as it were, brought out of the "mind" and put into
"instruments" and rules for their use. They thereby become more
readily inspectable and corrigible. No complete parallelism between
the Kantian movement from representations to objects and the move-
ment outlined here from physicalistic protocol statements to state-
ments about objects can perhaps be maintained. It would be worth-
while, both from the standpoint of a better understanding of Kant and
a better understanding of the rules of scientific construction, to see
where the parallelism obtains and where it breaks down. The present
essay is a very preliminary and tentative exploration of this way of
freeing the Kantian philosophy of science from its exclusive concern
with the problems of Newtonian physics and its apparent dependence
upon a psychology of faculties. A full exploration of these possibilities
would be the lifework of another Kant.
ON THE KANTIAN FOUNDATION
OF SCIENCE AND MATHEMATICS*
S.Korner
By taking the results of scientific investigations seriously a philosophy

such as Kant's exposes itself not only to philosophical criticism, but
also to objections which stem from scientific investigation. It is to be
expected that the objections which are levelled against it from both
inside and outside, will supplement one another and thereby be relevant
both to the philosophy itself and to the sciences it examines. In this
paper I want to examine Kant's doctrines of the justification of
synthetic a priori judgments in mathematics and physics in the light of
some later results in these sciences. My procedure will in both cases be
the same: I shall first try to show that Kant's assumptions about the
possibility of these sciences are incompatible with their present state. I
shall then trace this incompatibility to a presupposition which Kant
does not justify and which is not capable of justification. I start with
some critical remarks on the transcendental method and conclude with
some observations on those of its features whose value remains unim-
paired.
I. A difficulty in Kant's notion of transcendental knowledge. The

result of any transcendental deduction is a transcendental judgment,
that is, according to Kant, a synthetic a priori judgment "by which we
know that-and how-certain representations (intuitions or concepts)
can be employed or are possible purely a priori."l A transcendental
judgment is thus a synthetic a priori judgment which has as its object
another such judgment. Now it is clear that a judgment about another
judgment can be of the same type as its object or of a different type.
Thus there are analytic judgments about empirical judgments, e.g.,
when it is correctly asserted that an empirical judgment logically
implies another one; and there are empirical judgments about analytic
• Reprinted by permission of the author and publisher from Kant-Studien

(1966), EP' 463-473, translated from the German by Petra von Morstein.
lB 80. Cpo B 25, B 117.
97
judgments, e.g. when it is correctly asserted that some mathematicians
avoid the application of the tertium non datur in inference. According
to the Kantian classification of judg.ments, a judgment that expresses
transcendental knowledge must either be synthetic a priori or analytic,
since Kant's definition of transcendental knowledge excludes the possi-
bility that it is synthetic a posteriori.
Here we encounter a difficulty. If a transcendental judgment
were analytic, it could certainly establish what can be deduced from
our philosophical presuppositions of whatever kind they might be, but
it could never justify these presuppositions. An analytic judgment can
be the result of a logical, but not of a transcendental, deduction. If, on
the other hand, a transcendental judgment were synthetic a priori, then
it would itself, according to Kant, need a justification. To justify
synthetic a priori judgments by means of transcendental synthetic a
priori judgments would thus lead to an infinite regress. Although Kant
is well aware of the importance of his distinction between transcen-
dental and other a priori judgments, he does not concern himself with
this difficulty. By and large he seems to hold that transcendental
judgments are synthetic a priori judgments about synthetic a priori
judgments.
That transcendental judgments, whether they are analytic or
synthetic a priori, cannot fulfill their task of justifying synthetic a
priori knowledge was realized by Fries, and later by Leonard Nelson.
These philosophers follow Kant in making a sharp distinction between
synthetic a priori judgments that need a justification and the judgments
that justify them; and they believe that a way out of the dilemma is to
be found by considering the latter not as a priori, but as a posteriori,
namely as belonging to an empirical introspective psychology, the
object of which is synthetic a priori knowledge. However, I do not
wish to discuss here this interesting approach (which is in some
respects reminiscent of Hilbert's programme for a metamathematical
proof theory), especially since the attempts of Fries and Nelson are
meant-apart from minor modifications-to justify the same synthetic
a priori principles as Kant's transcendental deduction.
Obscurities and even mistakes in a proof or justification do not,
of course, imply that the propositions in question, insufficiently proved
or justified though they may be, are false; nor do they imply that these
propositions cannot be correctly proved or justified. However, if such
a justification or proof is not available the question arises whether there
may not be good reasons for rejecting the allegedly secure propositions
as false or at least for restricting their generality. As regards Kant's
On the Kantian Foundation 99
transcendental dedqction reasons for such doubts are nowadays not
hard to find, especially if we consider that the tasks of a transcendental
deduction are given by a metaphysical exposition of concepts, i.e., by
an exposition which, for each concept, "contains that which exhibits
the concept as given a priori."2 In the Transcendental Aesthetic the
exposition rests on the assumption that in mathematics objects are
determined as "entirely pure", and in the Transcendental Analytic it
rests on the assumption that in physics objects are determined as "at
least partly pure".3 Kant has only one mathematics and only one
physics in mind, so that what these sciences define as their pure or
partly pure objects is unambiguously determined. The development of
mathematics and physics after Kant, and in particular the variety of
non-monomorphic geometrical and set-theoretical axiomatic systems,
as well as the recent relativity theory and quantum mechanics, force
every philosopher who takes Kant's work seriously to examine the
extent to which Kant's metaphysical exposition, and, consequently, his
transcendental deduction of the principles of mathematics and the
sciences can be upheld in the light of their later results. The following
remarks are concerned with this problem.
II. The doctrine of the Transcendental Aesthetic and the bifurcation

of mathematical theories. Like all his predecessors, Kant held that the
axioms and theorems of mathematical theories are true in reality, that
their truth is in principle accessible to mathematical investigation, and
that conceptual thinking reflects the non-conceptual or pre-conceptual
mathematical reality completely and unambiguously. Godel's incom-
pleteness theorem (1931) and Church's insolvability theorem (1936)
show at least that the philosophical theses about the accessibility of
mathematical truth, along with the theses that all classes of mathe-
matical problems are in principle solvable and that mathematical reality
is completely reflected in our thinking, must be revised or formulated
more precisely. However, I shall avoid these questions, not only
because they are discussed very frequently, but a!so because they do
not concern the core of Kant's doctrine. I shall even neglect here
Kant's central conception of mathematical reality as a reality which is
only empirical but not transcendentaJ.4
From the point of view of the contemporary philosophy of
2B 38.
3Bx.
4 B 44,52.
mathematics, Kant's most important doctrine is his thesis that the
axioms and theorems of mathematics are synthetic a priori judgments.
This contradicts the theories of Leibniz, Frege and Russell according
to which the axioms and theorems are logically true, that is, true in
every possible world, including the actual one; and it also contradicts
the frequently assumed but philosophically not fully clarified doctrine
that the axioms and theorems are true "only in a possible world", i.e.
not "true" in the strict sense of the word. Those mathematical dis-
coveries which are most important for the philosophical evaluation of
Kant's philosophy of mathematics and its philosophical rivals are what
I shall here call "metamathematical bifurcation theorems", in particular
the theorem that both Euclidean and non-Euclidean geometries are
internally consistent, and Cohen's theorem (1963) according to which
both Zermelean and non-Zermelean set-theory are internally consis-
tent. The former contains, besides the usual axioms of set-theory, the
=
continuum hypothesis (~Il 2x.); the latter contains instead of the
continuum hypothesis its negation as a postulate.
It should be emphasized that this bifurcation of set-theory, which
is not the only one, also affects arithmetic. For if "natural number" is
defined in terms of "set" then different set-theories imply different
number-concepts. In addition to this distinction between different
number-concepts in classical mathematics there is the distinction be-
tween the concept of a number in intuitionistic mathematics and that
in classical mathematics, since a natural number which belongs as an
element to a merely potential infinity is in principle distinct from a
natural number which is an element of an infinite totality.
It is easy to see that the metamathematical bifurcation theorems
are more in line with the Kantian doctrine than with the Leibnizian.
Thus the proof of the independence of Euclid's fifth postulate, i.e. that
it can without contradiction be replaced by its negation, shows that, in
Kant's sense, it is synthetic. That it is a priori, i.e. "independent of
experience and even of all impressions of the senses",5 is not in
question. The assumption that a decision between Euclidean and non-
Euclidean geometry is possible on the basis of observation rests on a
misunderstanding. For, at best, one can choose in this way between
physical theories which contain either Euclidean geometry, or else a
non-Euclidean geometry, plus specific physical hypotheses concerning
measurement of length and time, etc. Einstein recognized this quite
clearly.
I) B 2.
Although Kant did consider the possibility of a non-Euclidean
geon:etry, h~ did not in fact doubt that Euclidean geometry is true of
phYSIcal realIty. A further reason for this conviction was that New-
tonian physics presupposes Euclidean g~ometry. After the discovery
o.f no~-Eucli~ean geo~etries and, especially, their successful applica-
tIOn III phYSICS, one IS faced not only with the question which
geometry can be applied to reality, but also with the question, what is
meant by the applicability of a geometry to reality? At least it can no
longer be claimed that the view that Euclidean, or any other, geometry
describes the structure of "space" needs no justification.
Here, as Nelson 6, Bernays and others have realized, one must
come to grips particularly with Felix Klein's thesis according to which
"spatial intuition is £:irst of all something imprecise which we idealize in
the so-called axioms for the purpose of mathematical treatment".7
Nelson agrees with Klein that the results of observation "are valid only
within certain limits of precision and under certain conditions." He
objects, however, that every idealization presupposes an ideal which
"cannot be taken from observation precisely because it is supposed to
form the norm for correcting observation". (loc. cit.) And from this
he infers that this ideal is a pure intuition in Kant's sense.
However, from the fact that an ideal is not taken from observa-
tion, it does not follow that there is only one ideal and that different
geometries do not represent different ideals of spatial intuition. It is
advisable here to distinguish between judgments and principles, as
Kant himself occasionally does, e.g. when he talks of regulative prin-
ciples. Judgments stand in logical relations to one another, e.g. the
relations of compatibilIty and incompatibility, and they are true or
false. Principles, which include not only judgments but also e.g. rules,
also stand in logical relations to one another, but need not be true or
false. Thus one can do justice to the possibility of different geometries
or of different ideals or idealizations by regarding the axioms of a
geometry as principles which, though synthetic (deniable without
contradiction) and a priori (compatible with any possible judgment of
observation), are yet neither true nor false, but only "true in a
possible, but not actual, world". Anyone who wants to dispute this and
6 Bemerkungen uber die Nicht-Euklidische Geometrie und den Ursprung

der mathematischen Gewissheit, in: Abhandlungen der Fries'schen Schule, Neue
Folge Band I. (1905/1906).
7 Uber die Arithmetisierung der Mathematik, in: Nachr. von der K. Gesell-
schaft der Wissenschaften zu G(ittingen, 1895.
shares the opinion that philosophy of mathematics must not ignore the
results of mathematics, has to put forward a counterargument, as
Nelson has attempted to do. He cannot rest his case on the transcen-
dental deduction of Euclidean geometry, since such a transcendental
deduction does not justify the truth of the synthetic a priori principles,
but rather presupposes it without justification.
Concerning the set-theoretical bifurcation theorems, in particular
Cohen's theorem of the independence of the continuum postulate,
similar conclusions follow. These are, moreover, completely general,
since any mathematical theory can be formulated in classical set
theory. One could object here that, while the bifurcation of geometry
at Euclid's fifth postulate is analogous to the bifurcation of set theory
at the continuum postulate (at the axiom of choice or elsewhere),
nonetheless, the concepts and propositions of set theory and arithmetic
(like "set", "one set contains another", "two", "one plus one is two")
are themselves empirical and exact and do not have to be replaced by
idealizations in order to be accessible to mathematical treatment.
However, this objection is not justified: since empirical attributes are
always defined at least in part by means of examples and counter-
examples of their application or of the application of their definientia,
the extensions that correspond to them, i.e., empirical sets, are not
exact, but admit borderline cases. Thus the number of such a set
cannot always be unambiguously determined, e.g., whether or not it
contains only one element, whether it is a couple, a triple, etc. When
we have very large sets or when mathematical induction comes into
play, there can be no doubt about the necessity for substituting
mathematical idealizations for empirical sets. In passing we might note
that the objection that elementary empirical concepts like "between"
do not need a geometrical idealization, since they are themselves
already exact, may be dealt with in a similar way.
One must distinguish clearly the positivist view that principles of
mathematics are meaningless propositions or conventions from the
thesis that they are synthetic a priori principles which are true only in
a possible world. The positivist view rests on a dichotomy of all
meaningful propositions into logical and empirical ones; on the basis of
this dichotomy many principles which are meaningful are, with no
good reason, called "meaningless" and thereby put on a level with
genuinely meaningless propositions, such as "Identity is the mother of
the colour green".
The thesis that mathematical propositions do not describe sense
exp.erience but idealize it .can be strengthened by further analysis
whIch takes note of the dIfference between ordinary empirical lan-
guage on the one hand and the language of mathematics and the
physical s~iences on- the other .. These sciences are embedded in a logico-
mathematIcal framework whIch consists of the first order predicate-
calculus with identity, enlarged as required by additional logico-
mathematical concepts and postulates, e.g. by means of the theory of
real numbers and thus by differential and integral calculus. The
employment of the predicate-calculus forces us to replace inexact by
exact concepts; the employment of the theory of identity forces us to
replace the non-transitive relation of perceptual distinguishability by
the transitive relation of mathematical equality. By enlarging the logico-
mathematical framework beyond elementary logic one imposes further
restrictions on empirical con~epts and judgments. In order to achieve a
better understanding of the modifications which the empirical material
undergoes when it is systematized by the usual logico-mathematical
theories, it is necessary to make expliCit the relevant characteristics of
the logical structure of our ordinary empirical language. Only when
this is done, can one understand the nature of the idealization by
comparing the idealized empirical concepts and propositions with the
idealizing mathematical concepts and propositions. A more detailed
discussion of these points would be inappropriate here. 8
Applying mathematics to sense experience, then, does not yield a
description of sense experience, nor does it, as Kant thought, yield a
description of the spatio-temporal structure which, qua pure intuition,
underlies sense experience. To apply mathematics is to identify some,
though not necessarily all, propositions of a mathematical theory with
corresponding, but not logically equivalent, empirical propositions
in certain more or less clearly defined contexts, for certain more or less
clearly defined purposes. Schematically, the application of a mathe-
matical proposition m, which belongs to a mathematical theory M, to a
result of observation which is expressed by the empirical proposition e,
is to be understood as follows: In a given context and for a given
purpose m (e.g. a proposition about a Euclidean triangle) is identifiable
with e (e.g. a proposition about a triangle formed by light-beams).
Such a proposition, which expresses something about a mathematical
and an empirical prop--?sition, is itself empirical.
What remains ofmathematical intuition if we accept this concep-
tion of mathematical principles as synthetic and a priori, but as "true in
a possible world" only? Weare faced with an uncomfortable dilemma.
8 See e.g. Deductive Unification and Idealization, British Journal for the
Philosophy of Science (Vol. XIV, 1964), and Experience and Theory (London,
1966).
If two non-equivalent mathematical theories have two possible worlds
as their objects, then they are completely compatible with each other.
And if two non-equivalent mathematical theories have the same pos-
sible world as their object then they are wholly incompatible with
each other. Yet it is clear that different mathematical theories can agree
with one another to a greater or lesser extent, and one can justly
demand that this partial agreement or its absence be explained.
The required explanation is easily found, if one investigates the
relation between mathematics and sense experience. For, if two mathe-
matical propositions which belong to different mathematical theories
are identifiable with the same empirical proposition with respect to the
same contexts and purposes, then it is precisely this "co-identifiability"
that they have in common. The notion of co-identifiability of mathe-
matical propositions which belong to different mathematical theories
can easily be extended to different mathematical theories. It can then
meaningfully be said that such theories have the same object, insofar as
they are co-identifiable with the same sense experiences, or more
precisely, with the same empirical propositions which express these
sense experiences. Thus the content of applied mathematics is not sense
experience as described by mathematics, but, according to the explana-
tion just given, sense experience as idealized by, and identified with,
mathematics. It should be noted that co-identifiability has to be clearly
distinguished from isomorphism: two isomorphic theories which are
applicable to different perceptual objects are not co-identifiable, and
two non-isomorphic theories, e.g., the geometry of the circle and of a
regular polygon with very many sides, may be co-identifiable.
Kant did not contemplate the possibility of a bifurcation and the
consequent diversity of mathematical theories. Historically, this is
understandable. The possibility of this diversity of mathematical the-
ories nowadays follows trivially from their existence-ab esse ad posse.
In addition, however, Kant also failed to consider the difference
between exact mathematical and inexact empirical judgments, and this
is, perhaps, more remarkable. For it is at least plausible to suggest that
some of his predecessors took this difference seriously. We may call to
mind Plato's contrasting of sense experience with the Forms, and his
doctrine of p.i.(h.~L'> which relates them to one another. "Geometry",
says Kant, "is a science which determines the properties of space
synthetically and yet a priori."9 And he assumes quite generally that
mathematics determines spatio-temporal experience not only syntheti-
cally and yet a priori, but also uniquely. This unwarranted result of
9B 40.
the metaphysical exposition of space and time is an essential premiss of
the transcendental deduction of the mathematical principJes. But since
this premiss stands in need of a justification which, at least at the
present stage of mathematics, is not available, the philosophy of
mathematics is forced not so much to' reject Kant's solution of his
problem as to alter the problem itself.
III. The doctrine of the. Transcendental Analytic and quantum

mechanics. Since the relation of Kant's philosophy of science to the
present stage of the sciences is in many respects analogous to the
relation of his philosophy of mathematics to the present stage of
development in mathematics, I can be more brief here. An essential
difference between the synthetic a priori principles of mathematics and
those of the sciences, according to Kant, is that the former are per-
ceptual, i.e., they describe the world of experience, whereas the latter
are non-perceptual, but nevertheless applicable to the world of experi-
ence since the objects of experience are constituted by the applicability
of those principles to sense impressions. Kant's insight that concepts
which are neither empirical nor logical or mathematical need not for
this reason be empty, and that principles which are neither empirical
nor logical or mathematical can be meaningful, may also be easily
confirmed in the light of contemporary physics.
Consider for example the Second Analogy of Experiencelo-"All
alterations take place i~ conformity with the law of cause and effect"
which contains "an indispensable law of empirical representation of the
time-series", namely that "the appearances of past time determine all
existences in the succeeding time, and that these latter, as events, can
take place only insofar as the appearances of past time determine their
existence in time, i.e. determine them according to a rule."11 This
means that the laws of physics are deterministic laws of the form y =
f (Xl' .. Xm t); moreover, they must be regarded as differentiable. 12
The laws of Newtonian physics are of this sort.
An analysis of orthodox quantum mechanics, on the assumption
that it is free of contradiction, shows that the allegedly indispensable
causal law is in fact a synthetic a priori principle: that it is a priori is
not in question; that it is synthetic follows from the fact that it is
denied in quantum mechanics without contradiction. But this demon-
stration is so simple only because it rests mainly on the fact that the
lOB 232.
11 B 242.
12 See e.g. B 248.
principle of causality is, in accordance with contemporary physics and
in opposition to Kant, taken to be dispensable from the scientific point
of view, since it can be replaced by a corresponding statistical prin-
ciple.
However, this revision of Kantian philosophy which is, in my
opinion, unavoidable, does not imply that the principle of causality is
false. For as we have seen in the case of the synthetic a priori principles
in mathematics, a principle need not be either true or false. Certainly,
the principle of causality, which as both Hume and Kant realized, does
not describe any actual intuition, cannot be interpreted as a description
of some merely possible intuition. But one can understand its logical
status and its role in physics, if one recalls Kant's distinction between
regulative and constitutive principles and considers the law of causality
as a regulative principle for the construction of physical theories, i.e.,
as a norm which was followed by Newton and the classical physicists
when they constructed their theories, but was not followed by Heisen-
berg, Born, and Pauli. Such a principle is used in the construction of
physical theories, but is tied to experience more loosely than are the
propositions of the physical theory itself. For what is confirmed or
refuted by experience is certainly not the norm according to which
causal or statistical theories are constructed, but rather the theories
themselves. It follows that one cannot accuse those heterodox physi-
cists, who, like Einstein and Schrodinger, wanted to replace a statistical
quantum mechanics by a causal theory, of being irrational, since it is
conceivable that one day a physical theory constructed accor<i:ing to
the regulative principle of causality might be more adequate to experi-
ence than the purely statistical contemporary theories.
In a similar way one can approach the other Analogies of
Experience. Thus the first Analogy, namely, the principle of the
conservation of substance, in Kant's and Newton's sense of material
substance, is certainly a priori. And it is synthetic because the principle
of the conservation of mass can, as is well known, be replaced without
contradiction by the principle of the conservation of mass-energy.
Here again we are dealing with a synthetic a priori principle which,
despite Kant's opinion, is not indispensable and is either false or a
regulative principle. The positivist view that metaphysical principles
are meaningless can in fact be often refuted merely by showing that
they are regulative principles. 13
13 On this point see, e.g., On Philosophical Arguments in Physics in: Ob-

servation and Interpretation (London 1957, New York: Dover Publications, 1960).
It is the task of the transcendental deduction to justify the results
of the metaphysical exposition. The task consists in proving that the
Categories, e.g. the a priori concepts of substance and causality, are
objectively valid. The principle of the transcendental deduction says
that the Categories "must be recognised as a priori conditions of the
possibility of experience, whether of the intuition which is to be met
with in it or of the thought."14 Since it was a result of the transcen-
dental exposition in the Analytic that any object of experience is,
among other things, necessarily a substance and a member in a causal
chain, it must be shown that the first and second Analogies of experi-
ence express conditions of the possibility of experience. N ow it
became clear in the course of the development of post-Kantian physics
that obi.ects of experience can be subsumed under Categories other
than the Kantian ones, e.g. under. the Category of mass-energy and not
of material substance, or placed in a statistical, not a causal, order.
Thus the result of the metaphysical exposition concerning the Anal-
ogies of experience is a premiss which is essential for their transcen-
dental deduction, but which itself stands in need of justification-
a justification which, at least in the present stage of the sciences, is not
available. Like the contemporary philosophy of mathematics, the con-
temporary philosophy of science is forced not so much to reject the
Kantian solution of the problem, but rather to alter the problem itself.
IV. Results of the transcendental method which remain valid. Accord-

ing to Kant's metaphor taken from jurisprudence the quaestio juris of
the transcendental foundations of mathematics and the sciences pre-
supposes an answer to the quaestio facti about their conceptual posses-
sions. If the preceding arguments are correct, then the inadequacy of
the answer to the quaestio facti implies that the quaestio juris, as Kant
posed it, does not arise: that is, cadit 1uae~tio. Fr?m this it does. not, of
course, in the least follow that Kant s phIlosophIcal approach IS to be
rejected entirely. On the contrary, we m?st put hi~ quaestio facti .and
his quaestio juris anew for the mathematICS and SCIences of our tIme.
The metaphysical exposition would show that in mathematics and in
the sciences we possess synthetic a priori principles which, however,
do not describe reality. The answer to the quaestio juris as to how such
principles are possible in mathematics would be that ~hey a~e P?ssible
as idealizations of empirical judgments and that theIr applIcatIon to
experience consists in identifying empirical judgments in certain con-
14 B 126.
texts and for certain purposes with their idealizations. The a~swer to
the quaestio juris as to how such principles are possible in the natural
sciences would be that they are possible as regulative principles for the
construction of theories and that their connection with sense experi-
ence is only indirect, since the theories which are constructed accord-
ing to them can be tested by experiment and observation, but the
principles used in their construction cannot themselves be tested in this
manner.
Beyond this we may, at least in the light of the present-day
scientific knowledge, say that Kant is mistaken in asserting that the
Categories and synthetic a priori principles suggested in the Critique of
Pure Reason are indispensable. However, a weaker thesis which fol-
lows from this assertion appears to be correct, namely that some
Categories and synthetic a priori principles are indispensable for sci-
ence. From the correctness of this thesis the incorrectness of both the
positivist and rationalist philosophy of mathematics and science fol-
lows, since both deny the possibility of synthetic a priori principles.
The axioms of mathematics are not analytic, as Hume and Leibniz
thought, and science, as even most modern positivists admit, cannot do
without theoretical concepts, i.e., a priori concepts, nor without
theoretical propositions, i.e., synthetic a priori principles. However,
the possibility of a transcendental deduction of Categories and prin-
ciples is suspect, since such a deduction must presuppose the unchange-
ability of the concepts which it tries to justify. Since retrospectively,
we find no such unchangeability, it would seem incautious to project,
even partially, the present stage of science or everyday thinking into
the future.
THE SIMPLICITY OF THE SOUL*
Jonathan Bennett
The second Paralogism argues that the soul is noncomposite, or

lacks parts, because "th~action of [the soul] can never be regarded as
the concurrence--of-severaLthings acting" (A 351). Although not an
obviously nourishing philosophical topic, this is less jejune than it
looks. Kant's argument that 'The soul is noncomeosite' (S) when
properly understood does express an a priori truth; that the latter is
"eIl!£ty" in a special way; and that it is peculiarly liable to be misunder-
stood ,as having~tent. Each part of his argument raises matters worth
discussing.
Here is Kant's reason for thinking that S says something true. If
something is to be viewed as composite, some mind must unite its
constituent parts by an "intellectual synthesis" -must combine them in
the judgment that they somehow go together. My basic notion of
compositeness is that of a number of elements united by my judgment:
it pre-requires myself, my synthesizi!lg self, to judge that the elements
areSo interrt£ltedas ~ compose a whole. I therefore cannot bring this
notion to bear upon myself:
~
Although the whole of the thought could be divided and distributed among
many subjects, the subjective "I" can never be thus diy'id~d and distributed,
and it is this "I" that we presuppose in all thinking (A 354).
This argument restricts "What is the soul like? " to "What can I regard
my soul as being like?" Before discussing that vital restriction, let us
see how Kant operates within it.
An ordinary statement describing something as noncomposite has
implications of the form: "If ... , my experience will, be thus and
not so," e.g., by implying that if the thing is hammered it won't split.
Kant insists that the truth expressed by S has no such implications:
• Reprinted by permission of the author and publisher from the Journal of

Philosophy (1967), pp. 648-660.
109
[It] tells us nothing whatsoever in regard to myself as an object of experi-
ence . . . It concerns only the condition of our knowledge; it does not
apply to any assignable object (A 356).
S gets its truth from the fact that I can be aware of a composite only if
I stand in judgment over all its parts. This, however, is a fact about
"the condition of [myJ knowledge," not about its content:
The judgment "I think" . . . is the vehicle of all concepts . . . It can have
no special designation, because it serves only to introduce all our thought,
as belonging to consciousness (A 341).
Kant's view about S, then, is grounded in something more general: my

inability to apply 'composite' to myself does not license me to apply
'noncomposite' instead, because it is one instance of my inability to
apply to myself-to my "thinking subject"-any concepts at all. In the
following, 'categories' can for· present purposes be read as "conceEts":
[S] does, indeed, occupy itself with an object of experience, but only in that
aspect in which it ceases to be an object of experience (B 427).
We can thus say of the thinking "I" (the soul) ... that it does not know
itself through the categories, but knows the categories, and through them
all objects, ... through itself (A 401-2).
The subject of the categories cannot by thinking the categories acquire a
concept of itself as an object of the categories. For in order to think them,
its pure self-consciousness, which is what was to be explained, must itself
be presupposed (B 422).
There is a build-up there: from what can be experienced to what can

be known to what can be thought.
What makes this worth arguing, Kant thinks, is that someone
who accepts S for the right reasons might nevertheless mishandle it:
one is tempted to "conclude from the transcendental concept of the
subject, which contains nothing manifold, the absolute unity of this
subject itself," despite the fact that we "possess no concept whatso-
ever" of it (A 340). This temptation generates "rational psychology"
-the invalid derivation of substantive conclusions about the soul from
true premises about "the .condition of our knowledge." In the follow-
ing passage 'the category of substance' could as well read "the concept
of noncompositeness":
Rational psychology owes its origin simply to misunderstanding. The unity

of conSCIOusness ... is here mistaken for an intuition of the subject as
The Simplicity of the Soul 111
object, ~nd. the category of substance is then applied to it. But this unity is
only umty In thought, by which alone no object is given (B 421-422).
II
Are there empirical propositions that one might infer from S? I

doubt it; but S does have consequences that one might think are em-
pirical.
Consider a piece of rational psychology, closely related to the
second Paralogism, by Kant's principal target. In his sixth Meditation,
Descartes says:
When I..consider the mind, that is to say, my self in so far only as I am a

thinking thing, I can distinguish in myself no parts; I apprehend myself to be
a thiQg single and entire. Although the whole mind may seem to be united
to the whoI.e..:body, yet if a foot, an arm, 'or any other part of the body is cut
off, I know that my mind is not thereby dimimshed (Kemp Smith, Descartes'
Philosophical Writings, p. 261).
Descartes's move from "my mind" to "the mind" is not (as I alleged in
a Note! in 1965) a simple non sequitur: it concerns that basic first-
person orientation which I have yet to discuss. Another point-missing
objection: "Your mind may have parts that you have failed to distin-
guish." Descartes, I think, would reply that nothing could count as
discovering that one's mind is composite because such a discovery
would have to have the form "I observe that these elements are related
thus and so." This answer is implicit in his reply to a different possible
objection:
Nor can [the mind'~J faculties of willing, sensing, understanding, etc. be

spoken of as being its pa~s; for jt is one and the same mind which willsi
which senses, which understands (Descartes, loco cit.).
Descartes clearly has in mind something like the Ka,!1}ian basis for S,
and apparently credi!s S with implications of a kind which Kant
rightly denies to it: "ILa foot is cut off. . . ."
Does Descartes think it impossible that some loss of skill, mem-
ory, etc. should accompany every physical amputation? That would be
an uninteresting mistake; but Descartes probably does not make it, and
1 "A Note on Descartes and Spinoza," Philosophical Review, LXXIV, 3 (July

1965): 380.
certainly it is not implied, or even apparently implied, by the Kantian
basis for S. What Descartes is declari~g i!TIp-ossible, I think, is that some
part o~_his mind should continue to exist in_ as~oci~~ion with the
amputated limb while the rest remained associated with the rest of his
body. Anyway, I shall use the name 'Descartes' for a possible philos-
opner who takes that view and mishandles it in ways that make him a
suitable foil for Kant. I shall invent my Descartes ambulando. Initially,
we know only that he adduces S as proving that a mind cannot divide
into two.
Two what? If we must say "two half-minds," I am lost. Fortu-
nately, though, we do not need "half-minds" in order to deal with the
main issue or at least with a centrally relevant one. If Descartes denies
only that a mind could split into two whole minds, we can still get a
Kantian purchase on his position.
Consider the Smith phenomenon. Smith's body is halved; each
half becomes a complete human body; each of these qualifies fully as
the body of a person (as being associated with a mind); there is
overwhelming evidence that two distinct persons (or minds) are
involved; and in respect of each of these-call them Smith l and Smith2
-there is maximum evidence that he is mentally as well as physically
continuous with Smith in the ways that would ordinarily count toward
his being (identical with) Smith.
That last clause cannot be contracted to '. . . maximum evidence
that he is Smith'; for Smith l and Smith2 , being distinct from each
other, cannot both be identical with Smith. We may think of them
both as claiming to be Smith, and as completing the form "Before I lost
half my body, 1 ... " with fragments of Smith's biography; but, since
identity is transitive, we presumably couldn't endorse all these claims
just as they stand. Logical issues arise here, 2 but I needn't discuss them,
for my use of the Smith phenomenon is not threatened by any deep
problems in identity-logic. We are concerned solely with whether
Smith's mind has d!:!.~d to form the mind§_of Smith l and Smith2 , and
that emphasis on divisio!:1 also puts identity-logic in its place: the fission
of a~, if it could happen, would involve the concept of identity in
the same way (whatever that is) as the fission of an amoeba.
Tl1eKantian basis for S is obviously consistent with a human
body's __heing divided and its regenerat~<L hal~es' presenting onlookers
with certain patterns of ~inguistic and other behavior. If Descartes is to
2 See A. N. Prior, "'Opposite Number'," Review of Metaphysics, XI, 2

(December 1957): 196-201.
be in~eresting. he must allow that the phenomenon might occur while
denymg that It could be evid_<:.nce t~a--Illind~ad divided: "No one
coul~ know that his mind had divided; therefore nothing is properly
descnbable as the fission of a mind." Postponing discussion of this
inference, consider. its premise. Is it true that I cannot intelligibly
suppose that my mind might divide into two?
Well, I can make sense of this:-My body is divided into two,
leaving me with a half-body which grows into a complete one; the
other half develops similarly, and qualifies as the body of a person who
claims continuity with the pre-fission JFB; yet I remain confident
throughout of my continuity with the pre-fission JFB. In describing a
Smith phenomenon with myself as subject, it seems, I must identify
myself _with-one-of-the post-fission people, and speak army knowing
that there is someone else who .J!tc.
Descartes will say: "Just so! All you can describe is a division of
your body accompanied, perhaps, by the creation of a second mind
which resembles }:.ours and is associated with the half of your body
that y-ou-you-Iose." This is not the abstract, symmetrical, logical
point that JFB l and JFB 2 cannot both be identical with JFB. Des-
cartes's point is rather that my description of a Smith phenomenon
with myself as subject must be asymmetncal; whence he concludes
that a Smith phenomenon could not occur without a relevant, and
indeeq crucial, asymmetry. He does not say: "Smith l and Smith2 may
both be continuous with~h; but we may not_ identif~ bo~~-and
probably oy.ghtn't to identify either of them with-him. ' He says,
rather: "Only one~c;;a!1 be continuous with Smith, and that one is
identical with him." -
. . This, though not absurd, is '.UQ..ng. To cash it with a contentful
asymI11etr.y we must divorce Descartes's position from the only prem-
ises that ,§Ypport.it.
We might cash it as follows. If Smith l and Smith2 were observed
carefully enough, one would have to find_evid~~~e that one of them
was an impostor, betr;ryed by a shifty look; or that one of them was
honestly deceived, as shown by curious memory. gaps. But Descartes's
premises obviously~n't imply this. To say that there must be ~uch
observable asymmetries is just to say that a stricts~s_e_QLtbe physlcal-
behaviO'fal-Smitllphenomenon could not occur.
Descanes's asymmetry claim, then, must say something about the
form in which one could experience the Smith phenomenon as subject.
But what content of this sort could it have? Descartes may say: "I have
told you. Necessarily, if you undergo a Smith phenomenon only one
of the two post-fission minds will be yours-and you will know which
it is." To this I reply: 'lJ!ho will know? Suppose that Descartes puts his
view to Smith just before the phenomenon occurs, and that afterwards
both Smith} and Smith2 have the thought: "Descartes was right! It is
quite clear that I am Smith." Wouldn't their both thinking him right
imply that he was wrong?
Descartes may protest, "But they couldn'~ both . have that
tllQJ!ght"; but he has no arguments to_~up-p-ort him. He is certainly not
supported. by the fact that in describing what-a-Smith phenomenon
would be like "on tbe inside" I cannot stay "on the inside" of both the
·f·post-fi~ion ~ple, i.e., by the trivial fact that I canno~n~elligibly
sup~se myself to be two people at once.
Kant is right and Descartes wrong; but I don't deny that the case
is queer. Apart from its wild improbability, and from the demands it
makes for conceptual modifications, the Smith phenomenon puts us
under pressure at a deeper level still. Suppose that Smith knows in
advance that plans have been made for Smith} and Smith 2 to have lives
of, respectively, happiness and misery; and then consider what his
attitude can be to this fact. Equipoised calm, because the two prospects
cancel out? That doesn't make sense. Elation because he might become
Smith}, mixed with depression because he might become Smith2 ? That
concedes everything to Descartes. Indifference, because he is due for
extinction and' won't be involved at all? Perhaps-but can one regard
that as obviously, straightforwardly, the attitude that best fits the
facts?
III
I have pitted Kant against Descartes on an agreed battleground,

namely, tbe first-person case, with a shared assumption that this is the
right place to start. The assumption might be challenged: "The
Kantian basis for S shows that my view of myself must be blinkered;
but it doesn't follow that my thinking about myself must be cramped,
for I can always listen to what others say about myself and modestly
accept it. We might conclude, by majority vote, that every mind is
composite." Kant would find this too glibly dismissive.
The argument might go like this. Kant: "How can you envisage
accepting that your mind is composite if you cannot view it as
c~mp~te?" Objector: "I understand 'composite' and 'mind' as they
occur in a common language, and I understand how someone else's
mind can justifiably becalled 'c.?~~~site'. My mind is just one among
others, and I could know that it too is composite-my liberating
membership of a community of minds enables me to enlarge my view
of myself through others' views of me and my views of others." Kant:
"What you call liberation is really abd~cation from your intellectual
autonomy. You are not giving due weight to the fact that everything
you are entitled to say-including the claim that yours is not the only
mind-rests ultimately on facts about your mind. You are not entitled
to say that your mind is just one among others."
I give Kant that speech on the strength of an important and (I
submit) valuable.feature of his thought, namely his adoption of what I
shall call methodological solif.?.sism (MS). The rationale of MS is a
sequence of near-tautologies. (1) Any theoretical problem presents
itself to me as a problem only if it presents itself as my problem. (2)
To solve a theoretical problem of mine, I must decide what it is
correct or justifiable for me to believe about something or other. (3)
In justifying my beliefs I can appeal only to data that I have, and what
appeal I make depends wholly on my intellectual handling of those
data. All these would look sterile but for the emphasis on certain
pronouns; but (1) to (3), selectively stressed, do suggest a philosophi-
cal program. It is an unfashionable program, I know; but are the above
trivialities robbed of their truth by their italics?
MS consists in approaching philosophical problems in ways
which are recommended by (1) to (3) above. It generates certain
techniques, prominent among which is an insistence on t~he_ first-person
~st in the philosophy of mind, i.e., upon testing "There could be a
sentient creature which . . ." by asking "Could J know myself to be
su~h a creature?" Kant's respect for this test explains his initial
sy~ with the second Paralogism. The connection can be seen
here:
It is obvious that, if I wish to represent to my~elf a thinking being, I must

put myself in his place, and thus substitute, as it were, my oW!1_sJlbject for
the object I am seeking to consider .... , and that we demand the absolute
unity of the subject of a thought, only because otherwise we could not say J
"!think" (A 35 3- 35 4) .
---
A little earlier, he gives a similar underpinning to the Paralogisms as a
whole:
[I am entitled to say] that that which thinks must, in all !=ases, be constituted
as the voice of self-consciousness declares it to be".£onstitul~d in my own self.
The reason is this: we must assign to things, necessarily and a priori, all the
properties which consti~te the conditions under which alone we think
them. Now I cannot have any representation whatsoever of a thinking be-
ing, through any otl:!.er experience, but only through self-consciousness.
Objects of this kind are, therefore, nothing more than the transference of
this consciousness of mine to other things, which in this way alone can be
represented as thinking beings (A 346-347).
Kant goes too far. It is one thing to value the first-person test, and
quite another to disqualify any statement that fails it. We sometimes
credit creatures with conscious states that they cannot accompany by
an "I think"; are such statements shown to be wrong just because,
spectacularly, they fail the first-person test? (It is arguable that Kant
demands self-consciousness not for all conscious creatures but only for
those which make judgments. I think he is committed to the stronger
view; but the weaker one raises analogous problems, and my ensuing
remarks apply, mutatis mutandis, to it as well.)
Kant might reply: "No-they are not shown to be wrong, but
are shown to be only ways of codifying facts about behavior. What
the insight underlying MS shows is that 'mind', 'conscious', etc. don't
apply univocally to creatures without self-consciousness and to hu-
mans. To say that something is conscious but not self-conscious is to
say something totally different from, wholly dependent on, and two
levels less basic than, anything one can say about one's own mind."
This reply would be rash. To show why, I must take a small detour.
Sometimes in philosophy one wishes to make an essentially first-person

point, and to distinguish it from autobiography. Thus one might say
"My mind is noncomposite, and aren't you also prepared to say 'My
mind is noncomposite'?", trying to generalize about the first person.
This procedure is vaguely described by Kant's expressions "I put
myself in his place" and "the transference of this consciousness of mine
to other things." (These expressions do not explain the procedure. To
do that, one would have to tackle the neglected problem of elucidating
the language in which-extraordinarily-we discuss with one another
the philosophical problem about other minds.) But the notion of
"putting myself in his place" does not describe, even vaguely, the
making of specific, contingent, test-passing judgments about the minds
of others. These are the topic of my detour. The first-person test
allows that someone else could have a headache, but what does Kant
say about my judging that someone else does have a headache?
He ought to see a problem here. His methodological solipsism,
and such statements as that "Thinkin bein s, as such, can never be
found by us among out€lr app€larances" (A 357), forbid 1m to ta e or
granted the ways in which we do-or even the fac~_!..bat we do-apply
mental predicates to others. Yet Kant is wholly inattentive to this as to
all aspects of the notion of an embodied mind. 3 Strawson notes one
significant sentence: "[The soul's] permanence during life is, of course"
ev~~!).t, since_the thinking being (as man) is itself likewise an Qbject of
the _outer senses" (B 415). But there is no evidence that Kant has duly
weighed the vital phrase "the thinking being (as man)"; and that
particular sentence is embedded in, and notably contributes to, a
muddle. In one place, Kant does remark that under certain circum-
stances "the thesis that only souls . . . think, would have to be given
up; and we should have to fall back on the common expression that
men think" (A 359-60). But since the circumstances are not actual or
even possible ones, the remark is irrelevant to any third-person judg-
ments that we do or could make. Other apparently relevant sentences
occur in A 362-364, but to assess these I should need to understand the
surrounding discussion of the third Paralogism.
Sometimes, indeed, Kant is notably evasive:
~ as thinking, am an object of inner sense, and am called "soul". That

which is an object of theouter senses is called "body" (A 342).
- ------ -
The second sentence's lack of any personal pronoun looks like a

delib~ate attempt to avoid any reminder of the problem about how
the soul is connected with the body.
Still, what could Kant say, consistently with his MS, about con-
tingent, test-passing, third-person statements about minds? Almost
anything. MS implies a partial program for seeking an analysis of third-
person statements, but makes no prejudgment about what must be
found. In stressing what I can know given my data it does not imply,
for example, either (a) tnat I could be self-conscious even if there
were no other minds or (b) that lcowaoe -self-conscIous even if I had
no perc;pt~ of ~!:l.- outer w~rld. For what the fact is worth: I am
inclined to accept (a), but only because the arguments so far adduced
against it seem to me invalid; and I reject (b) on the strength of a
thoroughly MS-type argument against it. MS do~s perhaps imply that
some of my first-person statements are in some way more basic than
any of my third-person statements; but that says almost nothing.
The detour is complete. We are confronted by (1) first-person

judgments, (2) test-passing third-person judgments, and (3) test-
failing iudgments; and we are considering a methodological solipsist
:1 See my Kant's Analytic (New York: Cambridge, 1966) pp. 56, 143,222.
who, seeing that none of (3) could be expressed as members of (1),
avoids dismissing them all as false by saying that they aren't what they
seem. The members of (3) -he says-are not really about minds at all
but about behavior-patterns, and statements expressing them use
mental predicates with meanings other than those they have in state-
ments expressing (1) and (2). But this implies an entitlement to
seperate (1) and (2) from (3), and that won't do at all. It· isn't clear
that a methodological solipsist can avoid behaviorism about (3), but
the reasons for thinking that he can't are also reasons for thinking that
he can't avoid behaviorism about (2) either. If he is to isolate (3) by
saying that they admit of a purely behavioral analysis, then he must
show-not assume-that behaviorism is not the whole story about (2).
I don't deny that this can be shown; but attempts so far have satisfied
few but their authors, and Kant in particular doesn't even make one. [I
have encountered a tendency to associate the line separating (1) and
(2) from (3) with the line separating creatures that use language from
ones that don't, and to infer that (3) is therefore on a different footing
from (2). But even if that association held-which it doubly fails to
do-it is not obvious that it would save (2) from behaviorism:
language is linguistic behavior, and the question of what it is to
understand what another says, like every epistemological question, can
be answered thoroughly and in depth only by an MS approach.]
I conclude that Kant is entitled to his initial sympathy with the
second Paralogism: there are sound reasons for exploring what might
be called the "logic of the first-person case." But those reasons imply
that there are problems about (2) test-passing judgments about the
minds of others and (3) test-failing judgments. Kant fails in his obliga-
tion to face up to those problems-m the case of (2)---.Qy evasion, and in
the case oj GilLan implied dismissiveness to which he has not earned
thuight.
The fourth part of my paper is merely a coda. It seeks to widen
the historical setting, to amplify some of the foregoing remarks, and-
finally-to add a suggestion about the analysis of (2) test-passing third-
person judgments.
IV
Of all great modern philosophers, Spinoza is probably the least

MS-oriented. He should, but does not, ask "How, if my philosophy is
true, can I know that it is?". He does not, and cannot, give a remotely
satisfactory account of self-consciousness. His system's denial of any
special status to ~I" is as damaging as its more notorious dethronement
of "now." These charges point to some of the gravest weaknesses in
Spinoza's remarkable edifice. But if his work loses depth by its freedom
from MS, it also gains scope. In pardcular, it makes room, as the
Cartesian and Kantian contexts do not, for worth-while questions
about (3) judgments that fail the first-person test.
In Part II of the Ethics, for example, the material between
propositions 13 and 14 implies that the notion of an individual body is
neither sharp nor absolute: an ordinary human body, for instance, is a
complex configuration that we can regard as "one body" only because
its parts interact in ways that give it a sufficient degree of unity. (This,
together with Spinoza's psychophysical parallelism, yields proposition
15: "The idea which constitutes the actual being of the human mind is
not §imple, but compounded of a great number ~- .-
...
of ideas," which
---'
tramples down the second Paralogism and· even Kant's limited sym-
pathy therewith. But that is bx:-the way.) That suggests that the
everyday concept!l!llQP-p-ortuJ!ism whereby we count bodies might
sometimes fail us because some physical phenomena remain obstinately
borderline between "one _body" and "two bodies." And that raises a
further question: could it ever be correct to say "This. organic
configurationjs_p-recisely borderline between being a si~g~ _creature
with a mind and being a pair of .cre~tur~s with a mind_each"? An
affirmative answer clearly fails the first-person test; but the question
still seems to me an open and an interesting one, and I am grateful to
Spinoza for raising it.
Objection: "A borderline case between 'an organism' and 'a
colony of organisms' need not also be borderline between 'a mind' and
'a colony of minds'. Spinoza thinks otherwise, but only because of his
wholly implausible psychophysical parallelism. Aren't you implicitly
congratulating him upon that?" No, I reject Spinoza's thesis that
everything extended has its mental counterpart, related to it as the
human mind is to the human body. But that thesis is rooted in two~
good insights: (a) our non-first-person judgments about what minds
t~d what they are like must be closely linked to some of our
identifications an~Lgescriptions of bodies; and (b) on a scale of known
kinds of body, from inorganic through to human, we cannot make a
precise absolute, nonarbitrary cut at the point where mentali~ comes
.in..Jhese defensi e t ese III 0 unrestncte psy:c ophysical'
paralkfu.m. simply ~y_Spinoza's rationaJistic passion for depth, definite~
nesS0l"nd theoretic tidiness, which turns (a)'s "closely linked to some"
into "wholly paralleled by all,'; and (b)'s "we cannot make a precise
etc. cut" into "there is no difference of kind between the two ends of
the scale." This is not a defence, but a representation of Spinoza's
position as consisting in some valuable insights which-understandably
but not inevitably-he has extravagantly inflated.
This picture at least gives us hope of learning something from the
visible core of Spinoza's philosophy of mind. A borderline between
"one" and "two"-to return to my example-can be sought in possible
organic configurations to which mental predicates could be plausibly
applied. To be led by Spinoza to the thought of a borderline between
"one mind" and "two minds," one need not swallow his psychophysi-
cal parallelism whole.
What one does need, though, is freedom from the inhibitions
induced by the first-person test.
Spinoza bought this freedom from those inhibitions at too high a
price. One's attention to (3) test-failing judgments may be paid for not
by ignoring (1) the first-person case but merely by failing to relate
(3) fully to it. This is unsatisfactory too, but it improves on Spinoza
and may be the best we can do in the meantime. Work done on (3) in
this spirit will be thoroughly, if provisionally, behavioristic (I use this
term throughout in a very broad sense). Pending a satisfactory relating
of (1) with (2) and of both with (3), we shan't know whether
behaviorism is fully adequate for (3), nor shall we know whether its
adequacy would isolate (3) from all our other judgments about minds.
But that can be borne with, so long as we know-as we surely do-
that a wholly behaviorist approach to (3) can yield results which are
not seriously wrong in detail.
Provisional behaviorism also lets us proceed with the analysis of
mental concepts as used in (2), but here we must tread more deli-
cately.
The case for thinking that behaviorism is the whole story about
(2) stems from a persuasive and usually helpful view about the
relationship between what I can mean and what data I could have. The
most persuasive case against it depends upon a denial that behaviorism
will do for the first-person case, together with a denial that mental
terms are systematically ambiguous as between (1) and (2). In short,
the unsolved problem that some of us have takes this form: we cannot
reconcile our views about meaning in general with our views about the
meanings of (1) and (2) in particular. I suspect that our views about
meaning in general must yield, or broaden; but I have nothing useful to
say about that. What I do want to suggest is a way of reducing the
area of conflict.
Briefly, the suggestion is that behaviorism may be adequate for all
our distinctions between mental states. Even if no set of hypotheticals
about behavior etc. can entail "He is in pain," such a set might entail
"If he feels anything, he feels pain" and. "He is not wholly absorbed in
a feeling of mild euphoria." This is not to propose that we partly
separate (2) from (1) by conceding behaviorism for one aspect of the
former, but rather to suggest that behaviorism is adequate for that
aspect of both.
A wholly behaviorist account of (1) seems unacceptable. I think
there is more to my being in pain than just my behaving and being
disposed to behave in certain ways, even if I cannot intelligibly
elucidate 'more':
The very fact that we should so much like to say: "This is the important
thing"-while we point privately to the sensation-is enough to show how
much we are inclined to say something which gives no information (Witt-
genstein, Philosophical Investigations § 298).
4~
Perhaps; but it also shows that behaviorism is felt to be inadequate for
the first-person case. For those who feel this, the great unsolved
problem is, precisely, to "give information" about the inadequacy or to
show that what they now "like to say" does after all "give informa-
tion" that isn't captured by a behaviorist analysis. (Their position is
intellectually disreputable-I used 'feel' advisedly-and I venture to
confess my unargued sympathy with it only because 1 think that many
others are similarly inclined. Only they could be interested in my
proposal.) The desire to say "This [pointing privately] is my pain,"
then, solves no problem; but it locates a problem, and so it should be
distinguished from the desire to say "It is because of the nature of this
[pointing privately] that I count as being, specificallYt in pain. tt The
first desire need not be accompanied by the second. To give expression
to the first alone would be to say: "My being in pain is my having a
conscious state such that I (am disposed to) behave thus and so." The
phrase 'conscious state' merely marks the point where resistance to
behaviorism begins, but if all such resistances could be expressed in
that form then something would be gained. For then we should have
only the .~n.itary p~oblem of ~aking bette!..:!han:-be~a,vi?rist sen~e of
the notion of "conscious state as such; and the elUCIdation of dIffer-
entiae-of statements- attributing this rather than that conscious state
to any creature-could proceed along wholly and non provisionally
behavioristic lines. The second half of 'conscious state such thatt
probably need not ~(! construed causally; but i~ it is to d~ any work at
all in my suggested formula, then the suggestIon mustn t be that any
two conscious states differ only in the behavior etc. in which they are
manifested. Room must be left for "inner" differences, which can't be
described behavioristically. That, however, leaves the suggestion stand-
ing: it may be that such "inner" differences play no part-or only a
part depending entirely on their association with behavior etc.-in
applications of mental predicates to anyone, even oneself.
Teachers sometimes lead into these matters by saying "How do
you know that his feelings are not utterly belied by his outward
appearance?", or by saying "How do you know that, although he is
outwardly responsive, he is not all blank within?" If my tentative
suggestion is right, these are not simply less and more radical introduc-
tions to a single problem.
EXISTENCE, PREDICATION, AND
THE ONTOLOGICAL ARGUMENT*
Jerome Shaffer
Hume said, "There is no being . . . whose non-existence implies a

co~tradiction'',1 and Kant said, "The predicate of existence can . . .
be rejected without contradiction".2 In making these claims, Hume
and Kant intended to bring out a peculiarity in assertions of existence,
for they both would admit that assertions that something was (or was
not), for example, round might turn out to be self-contradictory,
whereas assertions that something exists (or does not exist) could
never turn out to be self-contradictory. Now if this is a genuine
peculiarity of assertions of existence, then it follows that any proof
that something exists because its nonexistence implies a contradiction
will be invalid.
A famous example of an argument which purports to prove the
existence of something by showing that its nonexistence implies a
contradiction is the Ontological Argument, which purports to show
that it follows from a particular concept of God that such a being
exists, and therefore that the assertion of the nonexistence of God is
self-contradictory. Most philosophers have agreed with Hume and
Kant that the Ontological Argument is invalid, although it has recently
been defended. 3 My own view is that the argument is basically
unsound, but I find the standard criticisms totally unconvincing. In this
paper I shall show what I take to be the faults in the standard criticisms
and then go on to show what I take to be the proper criticism of the
argument.
The many versions of the Ontological Argument have in com-
• Reprinted by permission of the author and publisher from Mind (1962),

pp. 307-325.
1 Dialogues Concerning Natural Religion, Part IX.
2 Critique of Pure Reason, trans. Norman Kemp Smith, Macmillan, 1953,
p.504.
3 Norman Malcolm, "Anselm's Ontological Arguments," Philosophical Re-
view,69 (1960),41-62.
123
mon the following feature: a definition of "God" is given from which,
by the use of certain premises, the conclusion, "God exists", is de-
duced. The Ontological Argument has frequently been attacked by
casting doubt upon the acceptability of these premises. I wish to avoid
such controversies. I am interested in the move from a definition to an
existential statement. Therefore I shall use an argument which brings
attention to bear just on that move. This will be the specim,en under
discussion:
Let the expression, "God", mean "an almighty being who exi~ts and is
eternal". Therefore "God is an almighty bein?, who exists and is eternal" is
true by definition, and that entails "God exists'.
This argument purports to demonstrate that "God exists" is a tautol-

ogy, true by definition. It is not necessary to show that the expression,
"God", really does mean what it is here defined to mean, for the
definition is purely stipulative. But given that meaning, it is argued,
anyone who denies that God, in the sense laid down, exists has contra-
dicted himself. Only the laws of logic are required to show that he has
contradicted himself.
The following short objections to this argument will not do. (1)
The question is begged from the start by using a proper name, "God".
No, for "God" is not used here as a logically proper name. I have
simply introduced an expression into discourse, an expression which is,
grammatically, a proper noun. (2) Even if we grant the conclusion,
"God exists", that does not imply that there is a God or that something
is a God. Why not? In ordinary discourse the expressions are used
interchangeably in most contexts. It must be shown, if it is true, that
the implication does not hold here. (3) Tautologies only tell US about
our use of language, about the meanings of our terms, not about what
actually is the case. This objection begs the question. Proponents of
the argument claim that here is a tautology which tells us about what
actually is the case, namely that God exists. (4) But then with suitable
definitions we could "prove" the existence of a number of things
which we know perfectly well do not exist. No, for the things we
know not to exist would necessarily be different from the things
picked out by our definitions. (5) The argument involves a non
sequitur, for the premise is about a word but the conclusion is about
something different, a thing. No, the definition of the word allows us
to prove, by substitution, that the conclusion is a tautology.
The traditional attack on the Ontological Argument consists in
Existence, Predication, and the Ontological Argument 125
trying to show that the definition is inadmissible because it is ill-
formed. Kant argued that definitions can consist only of strings of
predi:~tes and since "exists" is not a predicate it cannot be a part of the
defimtIOn. Others have argued that "exi.sts" is a purely formal element,
present in any definition of a thing, and therefore could not be used to
show that some particular thing exists as opposed to anything else. I
shall, in Part I, discuss these arguments in detail and show that none of
them succeeds in establishing any impropriety in the definition of
"God" I have proposed. In Part II, I shall show that the argument,
although formally correct, does not do what the religious expect it to
do. The agnostic will still be able to raise his doubt and the atheist will
still be able to affirm his disbelief, no contradiction arising in either
case, even if each accepts the specimen argument under consideration.
u IExists' is not a predicate"
The Ontological Argument purports to show that God must exist

because the condition that He exists is a part of the definition of the
kind of thing He is. Kant argued that it could not be the case that
existence was a defining feature of God or of anything else, because
"exists" is not, as he put it (p. 504), a "real" or "determining" predi-
cate (he admitted that, grammatically, "exists" is a predicate).
What is a "real" predicate? Kant defines it as something "which
is added to the concept of the subject and enlarges it" (ibid.). This is a
most unfortunate definition for Kant to use, however, since it leads to
contradiction with another important doctrine of his, that existential
propositions are alway s synthetic (ibid. ). Synthetic judgments are
those which "add to the concept of the subject a predicate which has
not been in any wise thought in it" (p. 48), and if existential judg-
ments are always synthetic then "exists" must be a predicate which
adds to the concept of the subject, in short, a "real" predicate as
defined above. But even without the difficulties this definition of a
"real" predicate raises within the Kantian system, it is not a very help-
ful definition, for it represents predicating something of a subject as
revising the concept of the subject (by enlarging that concept),
something only philosophers of a Reconstructionist bent do very
often. But I shall say more of Kant's misrepresentation of predication
below.
What argument does Kant give for holding that "exists" is not a
"real" predicate, that is, not a predicate which adds something to the
concept of the subject? He argues that if "exists" were a real predicate,
then in asserting that something exists we would be altering our
concept of that something, thereby ending up with a different concept
from the one we started with. Since we now have a new and different
concept, we will have failed to assert existence of the original subject.
Thus if "exists" were a predicate, "we could not, therefore, say that
the exact object of my concept exists" (p. 505). But since we
obviously can say that, we cannot be adding anything to the concept
of the subject when we say that the subject exists, and therefore "ex-
ists" cannot be a real predicate.
It is astonishing that this argument has stood up for so long and is
still commended by philosophers, e.g. by Malcolm (op. cit. p. 44). For
the argument, if sound, shows that nothing could be a real predicate.
Suppose I wish to say that something is red, where "red" is intended as
a real predicate. In asserting that the thing is red, I would be adding to
my concept of the thing, and hence would be unable to say that the
object as originally conceived is red, that "the exact object of my
concept" is red. The argument which shows that "exists" is not a
"real" predicate also shows that nothing could be one.
The difficulty here lies in an incomplete picture of predication.
Kant seems to think that when I say that so-and-so is such-and-such, I
must be doing one of two things: either I am extracting the concept of
such-and-such from the concept of so-and-so (an analytic judgment)
or else I am revising my concept of so-and-so by adding to it the
concept of such-and-such (a synthetic judgment). Now which of
these two things am I doing when I say that so-and-so exists? Noticing
that existential propositions are often justified by an appeal to experi-
ence (p. 506), Kant decides that they cannot be cases of extracting the
concept of existence from the concept of the subject. So they must be
cases of revising concepts; thus, "all existential propositions are syn-
thetic". But this really will not do either, since to revise one's concept
of the subject is simply to change the subject of the proposition.
Hence in other places Kant concludes that "exists" is not a predicate at
all. Kant's vacillation here comes from an overly narrow account of
predication. To say that so-and-so is such-and-such is sometimes
neither to analyse the concept of so-and-so nor to revise it, but, to put
it roughly, to say something about the object conceived of. This use of
language is obviously not peculiar to existential propositions. The
sentence, "Crows are black", may be used to express a proposition not
Existence, Predication, and the Ontoiogit'ai Argument 127
about the concept of crows but about crows, and when so used would
create as much difficulty for Kant, given his account of predication, as
"Crows exist". There may well be important differences between
" . . . are bl ac k" an d" ... eXIst,
. "but K ant f al·1 s to bnng
. them out
by this line of argument.
Philosophers have tried to express what they took to be the truth
in Kant's claim that "exists" is not a real predicate without appeal to
that obscure notion of a real predicate. For example Malcolm restates
Kant's argument in this way:
Suppose tha~ t~o royal councillors, A and B, were asked to draw up sepa-
rately deSCrIptIOns of the most perfect chancellor they could conceive, and
that the descriptions they produced were identical except that A included
existence in his list of attributes of a perfect chancellor and B did not. (I
do not mean that B put nonexistence In his list.) One and the same person
could satisfy both descriptions. More to the point, any person who satisfied
A's description would necessarily satisfy B's description and vice versa
(p.43-44).
While Malcolm admits that this is not a "rigorous" argument, and

"leave(s) the matter at the more or less intuitive level", he thinks it
does show that "exists" is very different in character from the expres-
sions which go to make up a description or list of attributes of
something. But I cannot see how it goes to show that at all. For it
seems to me false that any person who satisfies B's description neces-
sarily satisfies A's. Could not a nonexistent person, say Merlin, satisfy
B's description but not A's? It cannot be said that Merlin fails to satisfy
B's description (B might have had Merlin in mind when he drew up
the list), unless the notion of satisfying a description is such that only
real things, existent beings, can be said to satisfy a description. But if
one uses this rather technical notion, the argument loses its intuitive
appeal.
Subject-predicate statements as really hypotheticals

Some philosophers have attempted to bring out the special fea-
ture of "exists" which debars it from appearing in a definition in the
following way. If we take "Crows are black" as a typical affirmative
subject-predicate statement and "~rows are not blac~" as a typic~l
negative one, then they would claIm that th~ affirmatl.ve stat~me~t IS
equivalent to the hypothetica~; "If there eX1s~s anythmg ~hIch ~s a
crow, then that thing is black , and the negatIve statement IS eqUlva-
lent to the hypothetical, "If there exists anything which is a crow, then
that thing is not black". Now if we hold that existential statements are
of the same form, then "Crows exist" would be equivalent to "If there
exists anything which is a crow, then that thing exists", and "Crows do
not exist" would be equivalent to "If there exists anything which is a
crow, then that thing does not exist". But the former hypothetical is a
tautology whereas "Crows exist" obviously is not a tautology, and the
latter hypothetical is not readily intelligible whereas "Crows do not
exist" is perfectly clear. Therefore it is most implausible to claim that
existential statements are of the same form :.ts the typical subject-
predicate statements, that is, implausible to construe "exists" as a
predicate.
I do not· find this argument very compelling. It is a mistake to
think that the hypothetical expresses the meaning of the typical subject-
predicate statement. Notice that the hypothetical must be put in the
form, "If there exists anything which is . . .". If we allow the
antecedent clause to range over nonexistent as well as existent things,
then the hypothetical, "If anything is . . . , then it exists", is not
tautological and the hypothetical, "If anything is . . . , then it does
not exist", is perfectly clear. And therefore the main reason for saying
they cannot be equivalent to categorical existential statements has
disappeared. But if we take the antecedent in the strong existential
sense, then the claim that such hypotheticals express the meaning of
subject-predicate statements breaks down. Consider the following sub-
ject-predicate statement: "Unicorns are proper subject-matter for
mythologists." It is not equivalent to "If there exists anything which is
a unicorn, then that thing is proper subject-matter for mythologists",
for the former is true but the latter is false. On the other hand,
"Unicorns are proper subject-matter for zoologists" is false, but "If
there exists anything which is a unicorn, then it is proper subject-
matter for zoologists" is true. And these are not just odd cases. There
are many other things which it would be true to say of unicorns, if
they existed, e.g. that they would obey the laws of physics, be sought
after by zoos, be mentioned in books which describe the species of
animals, etc., but which are not true of them since they do not exist;
and there would be many things which it would be false to say of them
if they existed but which are not false since they do not exist. So the
purported equivalence of subject-predicate statements to hypotheticals
does not hold. The reason is evident· enough. If a thing exists, then
given the way the world is it will have certain features and lack others,
so to say something about a thing is not the same as to say what the
thing would be like if it existed. Of course a thing keeps its defining
characteristics whether it exists or not, so analytic subject-predicate
statements will entail analytic hypotheticals, but in general the equiva-
lence does not hold. Nor does it hold if we understand the hypotheti-
cal as that of material implication, for "Unicorns do not have horns" is
false but the parallel hypothetical, "If there exists anything which is a
unicorn, then it does not have a horn", is true if taken materially,
which shows that they cannot be equivalent.
Since it is a popular view that subject-predicate statements are
equivalent to hypotheticals, it is important to see how someone might
come to such a view. One way would be through the Kantian restric-
tion of subject-predicate statements to statements about the concept of
the su~ject. If this were correct, then the kind of counter-example I
used, where the thing referred to by the concept turns out, given the
nature of the world, to have some feature could not arise. All subject-
predicate statements would be tautologous, either by analysis of the
concept of the subject or by revision of the concept of the subject,
and in this limited case the equivalence would hold.
But there is a line of argument more plausible to the modern
mind which would yield the same result. Suppose one thought that for
a series of words to express a meaningful subject-predicate statement,
the grammatical subject must refer to some existent or set of existents.
Then there would be no difference between attributing a property to a
thing and attributing a property to the thing if it exists. Again my
counter-examples could not arise. And philosophers have held such a
view. Ryle once said:
How can we make propositions about Mr. Pickwick or sea-serpents, given

that they do not exist? We cannot. For a proposition is only about some-
thing when something in fact answers to the designation in it. And nothing
· . "M r. P'IC k WIC
answers to t he pseu d0 deSlgnatlOn . k" or " t hose sea-serpents.
" Sa
And Broad holds, "Dragons do not exist. . . cannot be about

dragons; for there will be no such things as dragons for it to be about".4
Since such propositions are not meaningless, they must be about
something, however, and philosophers of this persuasion have offered
various candidates for the subject of statements about nonexistents,
Sa G. Ryle, "Imaginary Objects", Aristotelian Society, Supplementary Vol-

ume XII, 1933, p. 27.
4 C. D. Broad, Religion, Philosophy, and Psychical Research, New York,
1953, p. 182.
urging that they are really about beliefs, books, paintings, propositional
functions, properties, inscriptions, or story-tellers.
I do not see why statements cannot be made about nonexistents.
We can dream about them, think about them, and describe them, just
as we can wait for them, hope to have them, and look for them. We
can mention them, allude to or direct attention to them, and make
reference to them. One thing we cannot do, of course, is to point to
them, and someone who thinks of mentioning, alluding or referring as
a substitute for pointing will be puzzled as to how we can point to
what does not exist. But if we have not fallen prey to this overly
narrow conception of what it is to mention something, then we will
not be puzzled about how we can mention something nonexistent.
The most modest proposal along this line is that of Strawson,
who abandoned his earlier position that the use of a referring expres-
sion in cases where the object referred to is nonexistent is "a spurious
use"5 in which we either "pretend to refer, in make-believe or in
fiction, or mistakenly think we are referring when we are not referring
to anything" (p. 40). His modified view is that the "primary" use of
referring expressions occurs when the speaker believes the expression
to refer to some existent, but that such expressions may be used to
"refer in secondary ways, as in make-believe or in fiction". 6 But I take
it that Strawson would still wish to say that unless we refer to an
existent we do not succeed in expressing a subject-predicate "state-
ment", that is, an assertion which admits of truth or falsity. And if he
is right about this, then it becomes more plausible to claim that subject-
predicate statements are equivalent to hypotheticals (I do not wish to
suggest that Strawson would claim they are equivalent).
But here again it seems perfectly obvious to me that statements
which are true or false can be made about nonexistent things. After all,
unicorns do have horns, giants are two-legged, and Mr. Pickwick is a
most benevolent gentleman. Nor am 1 engaging in make-believe or
story-telling when 1 assert these things. My grounds for such assertions
are quite different from my grounds for claiming that my neighbour
is a most benevolent gentleman, but that does not detract from the
truth of such assertions.
To summarize, the claim that subject-predicate statements are
I) P. F. Strawson, "On Referring," reprinted in Essays in Conceptual Analysis

ed. A. Flew, London, 1956, p. 35. '
6lbid. footnote (added to original article), p. 40. See also Strawson's "A
Reply to Mr. Sellars", Philosophical Review, 1954, p. 229.
equivalent to hypotheticals fails, although perhaps a case could be
made for their equivalence with hypotheticals in those cases in which
the grammatical subject refers to some existent or set of existents. If so,
we might introduce a technical sense of "subject-predicate statement"
for those statements which are translatable in the way that "Crows are
black" is translatable into "If anything exists which is a crow, then that
thing is black". For reasons which differ in each case, none of the
fOllowing would be subject-predicate statements: "Crows are plenti-
ful", "Crows scatter during storms", "Crows vary greatly", "Crows
change over the centuries", and "Crows live in our harns". "Crows
exist" would not be a subject-predicate statement either, since the
requirement that the subj ect must refer to some existent would yield
the result that if it were a subject-predicate statement it would have to
be true. 7 No wonder it would only come to the trivial "If anything
exists which is a crow, then it exists", if construed as a subject-predi-
cate statement in the technical sense we have given that term. But none
of the peculiarities of "exists" which are brought out by saying that it
cannot be an element in a subject-predicate statement, in the sense
specified, go to indicate any impropriety in framing a definition in
which "exists" appears, such as the definition of "God" given above.
UExists" as a universal predicate
When faced with the claim that if "exists" is taken as a 'predicate

positive existential assertions become tautological and negative existential
assertions not readily intelligible, most philosophers have decided that
"exists" cannot be taken as a predicate. But some have boldly accepted
the consequence that "exists" is a trivial predicate, predicable of
everything conceivable. Hence the Ontological Argument becomes
harmless, for "exists" is a necessary predicate not only of God but of
everything conceivable. As recent supporters of this view have put
it:
Every conception involves the predicate "exists". Thus not only God's
essence but every essence implies existence. 8
7 Cf. P. F. Strawson, Introduction to Logical Theory, London, 1952, pp. 190--

191.
8 G. Nakhnikian and W. Salmon, "'Exists' as a Predicate," Philosophical
Review, 1957, p. 541.
And Hume suggested a doctrine very much like this when he said:
To reflect on any thing simply, and to reflect on it as existent, are nothing

different from each other. That idea, when conjoined with the idea of any
object, makes no addition to it. Whatever we conceive, we conceive to be
existent. 9
Thus attributions of existence to anything I conceive become tauto-

logical, and denials of existence not readily intelligible.
Now whatever the intrinsic merits of this view, it is a most
embarrassing one for Hume to hold. For he also wishes to maintain
that "whatever we conceive as existent, we can also conceive as non-
existent",lO and thus that "the non-existence of any being is as clear
and distinct an idea as its existence".l1 Since Hume uses "conceive to
be existent" and "conceive as existent" interchangeably,12 these two
doctrines are flatly contradictory, for the first implies that whatever
we conceive we cannot conceive as non-existent and the second states
that whatever we conceive we can conceive as non-existent. Nor can
we save Hume by interpreting the second to mean that whatever we
can conceive we can believe to be non-existent, for if I cannot con-
ceive of a thing except as "existent, then surely I cannot believe in its
non-existence.
Perhaps Hume was writing carelessly when he said that "what-
ever we conceive, we conceive to be existent", and expressed his real
thoughts more precisely when he said, "whatever the mind clearly
conceives includes the idea of possible existence."ls But if he did seri-
ously hold the first view, there are many ways in which he might have
come to it. The reason he does give,
Since w~ neve~ rememb~r any idea or impression without attributing exist-

ence to It, the Idea of eXIstence . . . must be the very same with the idea of
the perception or object. 14
confuses the existence of the conception with the existence of what is

conceived, the realitas formalis of the idea with its realhas objectiva.
9 David Hume, A Treatise of Human Nature, ed. L. A, Selby-Bigge, Ox-

ford, 1888, p. 66.
10 David Hume, Dialogues Concerning Natural Religion Part IX.
1~ David Hume, Enquir.y ,Concerning the Human Unde~standing"ed.lL. A.
Selhy-Blgge, Oxford, 1902, p. 164.
12 Treatise, op. cit. pp. 66-67.
IS Ibid. p. 32.
14 Ibid. p. 66.
But he might have argued, in line with his notion of ideas as pictures,
that we cannot form the picture of a thing as non-existent, and there-
fore cannot conceive such a thing. Also his attack on the abstract idea
of existence as distinguishable and separable from the ideas of particu-
lar obj~Ctsl5 .and his connected claim that in making judgments that
s?methmg eXIsts we only entertain the idea of the thing in an especially
lIvely and forceful way16 lead him to say that the idea of existence
cannot be something over and beyond the things we conceive, and
therefore would incline him to say that whatever we conceive, we
conceive to be existent.
Such a doctrine leaves most unclear what negative existential
judgments could be, and yet it is obviously most important to be able
to giv~. such judgments sense. At one point, when Hume concerns
himself with negative existential judgments, he abandons this doctrine,
interpreting them as judgments in which the idea of the object is
conjoined with the idea of non-existence,17 although in another place
he flatly rejects such an interpretation without giving what he takes to
be the correct account. 18 So far as I can see, the contradiction here
was one which Hume was never able to eliminate. 19 Nor, so far as I
can see, is it possible to reconcile the two. If the definition of a substan-
tive must include the notion that the thing exists (which is what I take
Hume's doctrine to mean), then that the thing exists follows from the
definition and is necessarily the case. To deny that the thing exists is to
contradict oneself just as certainly as to deny that the thing possesses
any other defining characteristic is to contradict oneself. And if all our
conceptions of things include existence as necessary properties of the
things, then no denials of existence will make sense.
Hume was mistaken in thinking that whatever we conceive of we
must conceive to be existent. For suppose it is a necessary feature of a
chimera that it be not only a she-monster of a particular sort but an
imaginary she-monster. Then it would be a necessary statement that
chimeras do not exist; anyone who held that chimeras exist would
contradict himself.. Could one conceive of a chimera? I do not see why
not. But one would be conceiving of it as non-existent.
15 Ibid. p. 623.
16 Ibid. p. 86 and passim.
17 Ibid. p. 15.
18 Ibid. p. 96, footnote.
19 J. A. Passmore, in his illuminating book, Hume's Intentions (Cambridge,
1952), diagnoses the difficulties which appear here as arising from Hume's attempt
to produce "a logic in which the only links are psychological" (p. 27).
I do not wish to suggest that philosophers were mistaken in
thinking that "exists" is in many ways different from grammatically
similar expressions, for it obviously is. To take Hume's point, for
example, if I wish to picture an animal, it will make a difference
whether I picture it as yellow or not; but one cannot make the same
kind of sense out of speaking of picturing it as existent or non-existent.
It does not follow from this that whatever I picture I picture as
existing, but Hume is certainly right in thinking that I cannot repre-
sent the existence or non-existence of the thing by adding to my
picture in an exactly parallel way to the way in which I represent the
yellowness or non-yellowness of the thing by adding to the picture. It
requires some special convention to indicate that what is pictured is
pictured as, say, imaginary. (Comic strip creators have special conven-
tions for showing this, for example by encircling it and linking it by a
stream of small circles to someone's head to show that he is just
imagining it.) To make this point is to bring out a difference between
"exists" and predicates like "is yellow". Further differences between
"exists" and other predicates are brought out in the arguments for the
slogan, " 'Exists' is not a predicate". What must be shown, however, is
that these differences bear relevantly on the issue, whether existential
statements can be true by definition. I have been concerned to argue,
in this section, that no differences have been noted which rule out
existential statements which are true by definition.
II
Until funher arguments are offered, it seems reasonable to hold

that there is nothing logically improper in so defining the expression,
"God", that "God exists" is a tautology and "God does not exist" self-
contradictory. In fact it seems to me that the definition I have given
expresses a concept of God (i.e., as necessarily existing) which many
people actually accept (just as it is a common conception of Satan that
he merely happens to exist). I wish to show in this section that this
concept of God can give no support to the religious. I shall argue that
no matter what its content, this concept of God is still simply a
concept. What must be shown, and what cannot be shown just by an
analysis of the concept, is that there actually exists something which
answers to the concept. Even if we have here the concept of an object
which necessarily exists, a further question remains whether any
existent meets the specifications of the concept. The difficulty lies in
showing that this further question makes sense, for I have admitted
that "God exists" is a necessary statement, analytically true, and
therefore it looks as if there could be no further questiQn. But that is
an illusion. It must however be dispelled.
As a first step, I wish to point out that the concept of God is
hardly unique in its capacity to generate a tautological existential
statement. For one thing, we could invent new tautologies. Suppose we
introduce the word, "particular", to mean "object which exists", and
the word, "nonentity", to mean "object which does not exist". Then
to bring out the difference between these two words we might
properly say, tautologically, "Particulars exist and nonentities do not
exist". Nor, for another thing, do we have to invent such words, for
we already have many words with existential notions included in their
meanings. The following sentences all have tautological uses: "Exis-
tents exist", "Fictitious objects do not exist", "Members of extinct
species existed once but no longer exist", "Hallucinatory objects do
not exist", "Historical persons have at some time existed". I do not
suggest that these sentences can never be used in non-tautological
ways, but I do suggest that they may be used tautologically in those
circumstances in which we wish to emphasize that such concepts
include as a necessary feature, as a defining element, notions of exis-
tence or non-existence.
As a preliminary to seeing what these tautological existential
claims come to, let us examine the relation between expressions of the
form, "A's exist", and of the form, "There are A's". Take the
tautology, "Fictitious objects do not exist". One might think that
"Fictitious objects do not exist" means the same as "There are no
fictitious objects". But a moment's thought will show that this is
incorrect, for although the former is true the latter is false. There are
fictitious objects, many of them-Alice's looking glass, Jack's bean
stalk, Wittgenstein's beetle, to mention only a few. So "Fictitious
objects do not exist" does not mean, the same as "There are no ficti-
tious objects." Similarly "Particulars exist" does not mean the same as
"There are particulars". In general, given a tautology of the form, "A's
exist", we cannot deduce from it, "There are A's", nor from a
tautology of the form, "A's do not exist", can we deduce "There are
no A's". And specifically, given the tautology, "God exists", we
cannot deduce from it, "There is a God". The statement, "God
necessarily exists, but there is no God", is not self-contradictory.
As it stands, the situation is most paradoxical. For in many of its
ordinary uses, "A's exist" is equivalent to "There are A's". If I raise a
question about the existence of pearls as large as my fist, it usually does
not matter whether you say, "Yes, such pearls exist", or "Yes, there are
such pearls". So if there is a way of saying that certain things exist
which does not mean that there are such things, then this must be
explained.
Now 1 have misinterpreted the situation somewhat. 1 have spoken
as if it were important whether we used the form "A's exist" rather
than "There are A's". But this is not the case. "There are A's" is
perhaps more resistant to being treated as a tautology but it is still
possible to frame tautologies of the form, "There are A's". "There is
what there is" and "There are what there are", are tautologies, and one
could imagine situations in which one might say them in this tautologi-
cal way. If 1 define "particulars" as "whatever entities there really
are", then "There are particulars" is a tautology, namely, "There are
whatever entities there really are." And if 1 define "a God" as "what-
ever divine being there is", then "There is a God" is tautological too.
So there is nothing distinc~ive about the forms "A's exist" or "There
are A's". They both function in very similar ways. But we do have
both forms of expression, and that is most convenient. For it allows us
to formulate a further existential question, using the alternative form,
when we are presented with a tautological existential assertion. Thus if
someone says, tautologically, "There are particulars [i.e. there are
whatever objects there are} ", we can avoid the danger of formal
contradiction in asking, "I grant that there are particulars, but do
particulars exist?"
What 1 am claiming is that if we are given a tautological existen-
tial assertion like "Particulars exist" or "God exists", the existential
question is not settled. Just as the tautology, "Fictitious objects do not
exist", leaves open the question whether there are fictitious objects, so
the tautology, "God exists", leaves open the question whether there is
a God. But what is this further question? How paradoxical it seems to
deny that "once one has grasped Anselm's proof of the necessary
existence of a being a greater than which cannot be conceived, no
question remains as to whether it exists or not". 20
It is tempting to try to resolve the paradox in accordance with
Aristotle's principle that "there are several senses in which a thing may
be said to 'be' ". Then to say that fictitious objects do not exist would
be to say that fictitious objects lacked, say, spatio-temporal existence,
whereas to say that there are fictitious objects would be to say that
20 Malcolm, op. cit. p. 52.

they had some oth~r kind of existence-hence no contradiction, since
in a sense fictitious objects exist and in a sense they do not. But appeals
to the systematic ambiguity of "exists" will not work in all cases, for
we may deny that there are A's in precisely the same sense of "be" that
we claim tautologically that A's exist. For example, it will be tautologi-
cally true that particulars exist in precisely the same sense of "exist",
say, temporal existence, that it might be true that there are no par-
ticulars.
A more promising line of argument consists in showing that a
tautological existential claim is quite different from a non-tautological
existential claim. How are we to explain the difference? Suppose we
say that a tautological existential assertion consists in attributing to the
subject a special property, the property of necessary existence. We
could explain this property by saying, Ii la Malcolm, that a being
which has this property is such that· it is senseless to speak of its non-
existence or of its coming into existence or going out of existence or of
the existence of anything else as a condition of its existence (pp. 44-
59). Now this account will not do. First, the attempt to explain the
necessity of the statement by postulating a special property commits us
to an infinite regress of properties, for presumably this special prop-
erty might not be one which a being just happens to have but one
which it necessarily has and which it is senseless to speak of its not
having, and thus by similar reasoning we are led to necessary necessary-
existence, etc. And it is most unclear what these properties could be Of
how we could distinguish them. But, secondly, it is not clear what this
property of necessary existence is, if this is any more than a way of
saying that the existential proposition is necessary. Am I making any-
thing clearer when I say that squares, which are necessarily four-sided,
have the special property of necessary four-sidedness? A defining
property is not a special kind of property. So the tautological character
of the existential assertions I have been discussing cannot be explained
by postulating a special predicate, necessary existence. Their tautologi-
cal character arises from nothing but the definition we have stipulated
for the subject term. But then we are still left with our puzzle: how is
it possible to say that A's necessarily exist but there may be no A's?
I wish to consider one further attempt to remove the paradox,
one suggested by some remarks by Carnap on a somewhat different
issue:
If someone wishes to speak in his language about a new kind of entities, he

has to introduce a system of new ways of speaking, subject to new rules;
we shall call this procedure the construction of a framework for the new
entities in question. And now we must distinguish two kinds of ques~ions of
existence: first, questions .. : within the framewor~; we call the~ tnternal
questions; and second, questions concernmg the eXIstence or reahty of the
framework itself, called external questions. 21.
Carnap goes on to explain that "external questions" concern nothing

but "whether or not to accept and use the fonns of expression for the
framework in question", a purely practical question to be answered in
terms of expediency and fruitfulness (p. 23). To apply this distinction
to our paradox, given the basic definitions and rules of a particular
religiou~ language it will be a necessary statement that God exists
(although not, perhaps, a necessary statement that the Devil exists),
but we may ask the further question, Is there a God? meaning, Is this
language a useful one?
Now waiving objections we may have about the vagueness of
talking of the "fruitfulness" of a set of expressions, it still remains the
case that "Are there any A's" will not always be identical with asking
if the language is fruitful, for it may be very fruitful to talk of perfect
pendulums, frictionless pulleys, the ideal society, Euclidean points, and
the economic man, even if it is perfectly clear that there do. not exist
such things. So the question whether there are such things cannot be
identical with the question whether it is in some sense fruitful to use
expressions which refer to them; a fiction or fayon de parler may turn
out to be useful and fruitful.
What lies at the hean of the puzzle about the Ontological
Argument is the fact that our concepts have two quite different as-
pects, marked by the familiar philosophical distinction of intension and
extension. A word like "horse" has a panicular meaning and is logically
connected with other words like "animal"; its corresponding concept,
the concept of a horse, has a particular content and is connected with
other concepts like the concept of an animal. It is this intensional
feature of words and their corresponding concepts which makes
cenain assertions like "A horse is an animal" tautological. But words
and concepts are also applicable to things. It turns out to be the case
that there have existed, do now exist, and will exist entities such that it
is true of each of them that it is a horse, true of each of them that the
concept of a horse applies to it. And this fact we may express by
saying that the word "horse" or the concept of a horse has extension.
21 R. Carnap, "Empiricism, Semantics, and Ontology", Revue lmernationale

de Philosophie, § 11, January 1950, pp. 21-22.
In making assertions about the extension of a concept there are typical
forms of expression which we use: " ... exist", " ... are non-
existent", "There are ... ", "There are no ...", " ... are plentiful",
" . . . are scarce,
" " ... are extmct,
. ""... are myth 0 ogIca
i ' 1", " ...
are found in Africa", etc. That such expressions are typically used in
~ssertions about the extension (or lack thereof) of particular concepts
IS what is correctly brought out in the slogan, "'Exists' is not a
predicate". But the typical use is not the only use. Since any statement,
with suitable definition, can be true by virtue of the meanings of the
terms, sentences with existential expressions can be used to express
tautological statements. The very same sentence which is typically
used to make a claim about the extension of the concept may instead be
used to make a claim about the intension of the concept. We cannot
tell by the form of the expression how the expression is being used.
"Particulars exist", when asserted tautologically, is used to make a
claim about the meaning of the word, "particulars", and therefore
cannot be used to make a claim about the extension of the term. Simi-
larly, if someone uses the sentence, "God exists", tautologically, he
tells us only that being an existent is a logical requirement for being
God. If, on the other hand, someone asserts, "God exists", non-
tautologically, then he claims that the term, "God", has extension,
applies to some existent. In the case of the Ontological Argument the
only valid conclusion is an intensional statement about the meaning of
the concept of God. A fortiori the conclusion cannot be about
whether anything exists to which the concept applies. The prima facie
plausibility of the Argument comes from the use of a sentence inten-
sionally when the typical use of that sentence is extensional. In this
way it conceals the illicit move from an intensional
,
to an extensional
statement.
It looked as if the familiar distinction between intension and
extension stood in danger of breaking down in the case of existential
tautologies. But we have seen that this is not the case. For even when
we have an existential tautology like "Particulars exist" or "God
exists", it still remains an open question whether the concept of
particulars or the concept of God has application, applies to any
existent. What is settled at one level is not settled at another level. It IS
important to see that we can go on to settle the question at the other
level, too, for we can make it a priori true that the concept has
application. For example, let the expr~ssion, "the concept of God",
mean "a concept which has application and applies to a being such
that ... ". Then by ~efinition the concept of God has application; the
statement, "The concept of God has application", is now a tautology,
given the definition. But nothing is gained by such a manoeuvre. We
have given the expression, "the concept of God", a meaning; we have
framed a concept, namely the concept of the concept of God, and this
concept makes certain statements tautologically true. Yet we can still
raise the extensional question, Does this concept refer to any existent?
At this level the extensional question would be whether there actually
is a concept of God such that this concept has extension, and there is
such a concept only if there actually is a God. So making the condition
of having application or extension a necessary condition for being a
concept of God still leaves open the question, concerning that concept,
whether it has extension. Nothing has been settled except the meaning
of a certain expression.
Why is it that extensional assertions cannot be tautological?
Because they do not merely tell us what the requirements are for being
an A but, starting with these requirements, tell us whether anything
meets these requirements. Even if it is a conceptual requirement that
the thing exist in order to be an instance of the concept, that in no way
settles whether the requirement is met. And if we make it a tautology
that the requirement is met, by framing a concept of a concept, then
we are left with the open question whether the newly framed concept
has extension. That is what is true in the thesis that no "existential"
proposition can be analytic. But we must remember that an "existen-
tial" proposition can turn out to be an intensional proposition, and
therefore tautological.
Since much of what I have claimed depends upon the legitimacy
of the intension-extension distinction, I wish to consider, finally, two
threats to this distinction. The first concerns the so-called intensional
object. When I conceive of an object, think about it, describe it, make
a painting of it, long for it, look for it, and expect to find it, it may
nevertheless be the case that the object does not exist, that the concept
has no extension. But it is tempting to say that there must be something
such that I conceive of it, think about it, describe it, etc., tempting to
say that the object in some sense exists. And thus it is tempting to say
that the mere fact that there is a concept of some object entails that the
object in some sense exists. Well, even if one says that, it is obviously
not the sense in which the religious usually wish to say that God exists
nor the sense in which the atheist wishes to deny that God exists. They
disagree about whether anything answers to that concept of an object,
not about whether that concept is a concept of an object.
A second, and more troublesome, threat to the intension-exten-
sion distinction aris~s when we try to apply the distinction to certain
conceJ?ts. We seem quite clear that the concept of a horse does have
extensIOn and that the concept of a unicorn does not have extension,
and that these are contingent facts. But now suppose we ask whether
the concept of a number has extension. If we hold that the concept
ultimately has as its extension things in the world, then it still remains a
~on~ngent fact that the concept has extension. But suppose we are
mchned to say that the concept has extension simply because, as we all
know, there are (infinitely) many numbers. Surely it is not a contin-
gent fact that there are (infinitely) many numbers. So if this fact leads
us to say that the concept of a number has extension, then it will be a
necessary proposition that the concept of a number has extension and,
given the concept of a number, we can say a priori that the concept
applies to (infinitely) many things.
What makes this case puzzling is that we have no idea what
would count as establishing that the concept of a number has extension
or that it does not have extension. We can investigate whether the
concept of a number is a legitimate one, clear and self-consistent; we
can note its logical connection with other mathematical concepts; and
we can frame propositions which state these connections, even propo-
sitions like "There exists a number which is even and prime". But what
would count as showing that the concept, over and above its inten-
sional content, has extension as well? Where would one look for traces,
signs, evidences, intimations, or testimonies of the existence of num-
bers? Would we not say of someone who did think such a search
sensible that he had misconceived the nature of numbers? Nothing
would count as showing that the concept of numbers had extension
over and above its intensional content, and this is to say that the notion
of extension does not apply here. The most that could be said is that
numbers are intensional objects.
The same thing must be said for the existence of God. The most
that the Ontological Argument establishes is the intensional object,
God, even if this intensional object has the attribute of existence as an
intensional feature. To establish that the concept of God has extension
requires adducing some additional argument to show that over and
above its intensional features, over and above the content of the
concept (or the meaning of the word, "God"), the concept of God
has extension as well. This additional argument will of necessity have
to be an a posteriori argument to the effect that certain evidences make
it reasonable to think that some actual existent answers to the concept.
We are thus led to the result that the Ontological Argument of itself
alone cannot show the existence of God, in the sense in which the
concept is shown to have extension. And this is just as the religious
wish it to be. They do not conceive of God as something whose being
expresses itself entirely in the concepts and propositions of a language
game. They conceive of Him as something which has effects on the
world and can in some way be experienced. Here is a crucial respect in
which His status is meant to be different from that of the numbers.
The concept of God is a concept which might have extension. But
some further argument is required to show whether it does or not.
Concepts are like nets. What they catch depends in part upon
how we construct them and in part upon what is outside the net.
Suppose I produce a net for catching fish one-millionth of an inch
long. Of such a net we are entitled to say, "This net catches fish one-
millionth of an inch long", and what shows that this statement is true is
nothing but the construction of the net. Does the net catch anything?
It catches fish one-millionth of an inch long. Still, a question remains.
Shall we ever find such fish in our net? For those who hunger for such
fish, the existence of the net does not in any way show that what they
hunger for shall be given unto them.
KANT AND THE
COSMOLOGICAL ARGUMENT*
Peter Remnant
It is frequently said that Kant has shown that the cosmological argu-
ment for the existence of God presupposes the validity of the ontologi-
cal argument, to the embarrassment of all those who accept the valid-
ity of the former but deny the validity of the latter.
However, it has been recently maintained by several writers that
Kant is guilty of a simple logical howler in the course of his demon-
stration-in the words of one of these writers, that Kant's reasoning "is
a piece of sheer though, of course, unconscious sophistry"l.
I shall begin by stating Kant's argument. Kant represents the
.cosmological argument as comprising two parts: first, the argument
that since something exists therefore a necessary being exists; second,
the argument that if anything is a necessary being then it must be an
infinitely perfect being, that is, God. He does not, as Father Johnston
says he does, attempt to reduce the cosmological argument to the
ontological argument. What he does do is argue that the second part of
the cosmological argument presupposes the ontological argument. I
shall restrict myself to what Kant has to say in defence of this con-
tention.
If we accept the first part of the cosmological argument as sound
and draw the conclusion that a necessary being exists we must then
show that this necessary being possesses the properties which theolo-
gians attribute to God. But, although the cosmological argument
begins with an appeal to experience, it "uses this experience only for a
single step in the argument, namely, to conclude the existence of a
• Reprinted by permission of the author and publisher from the Australasian

Journal of Philosophy (1959), pp. 152-155.
1 T. A. Johnston, S.J., "A Note on Kant's Criticism of the Arguments for
the Existence of God", this Journal, 21 (1943), p. 11. Cf. J. J. C. Smart, "The Exist-
ence of God", in Flew and MacIntyre, New Essays in Philosophical Theology,
pp. 36-37. Can sophistry be unconscious? In view of the prevailing tone of Father
Johnston's paper the concession must be taken for what it is worth.
143
necessary being. What properties this being may have, the empirical
premiss cannot tell us. Reason therefore abandons experience alto-
gether, and endeavours to discover. from mere concepts what prop-
erties an absolutely necessary being must have . . . "2 Reason
concludes that only an ens realissimum could be a necessary being. But
to draw such a conclusion is, Kant says, tantamount to carrying out
the ontological argument.
This he seeks to show by setting the matter out formally. We
wish to prove that a necessary being, any necessary being, must be an
ens realissimum. Now if it is the case, as we hope to be able to estab-
lish, that any necessary being will be an ens realissimum, then, convert-
ing per accidens, some entia realissima will be necessary beings. "But
one ens realissimum is in no respect different from another, and what is
true of some under this concept is true also of all." (A608, B636).
Therefore every ens realissimum must be a necessary being. This
means that "the mere concept of the ens realissimum must carry with
it the absolute necessity of that being; and this is precisely what the
ontological proof has asserted . . ." (I bid.)
It should be clearly noted that "A necessary being must be an ens
realissimum" is not, according to Kant, ~he conclusion of the first part
of the cosmological argument. It is what still remains to be proved
·after completion of the first part of the argument. Failure to grasp this
point has led another writer to attribute to Kant the bizarre argument
that, since the conclusion of the cosmological argument is the converse
of the conclusion of the ontological argument and "since the truth or
falsity of the convertend entails the truth or falsity of the converse, it
follows that a being which must exist is perfect (the conclusion of the
argument from contingency) is no truer than its convenend, a perfect
being must exist (the conclusion of the ontological argument)."3
The cosmological argument then, according to Kant, presupposes
the ontological argument in the sense that the cosmological argument
depends for its final step on the principle that any necessary being
must be a perfect being; this principle entails the proposition that any
perfect being is a necessary being; and not only is the latter principle
• 2 Kant, ~ritique of Pure Reason, A606, B634. All references are to the Kemp
SmIth translatIon. Father J<;>hnston .h~s som.e harsh things to say ~bout the quality
of contemporary scholarshIp, but It IS curIOUS to find anyone stIll relying as he
does, on the old Meiklejohn translation. '
3 Gerard Smith, S. J., Natural Theology, Metaphysics II. N. Y., Macmillan,
1951, p. 69. Father Smith does not raise any objection to this argument as such but'
denies that the one conclusion is the converse of the other. '
Kant and the Cosmological Argument 145
the conclusion of the ontological argument, but the ontological argu-
ment constitutes the only even plausible attempt so far to establish the
truth of this proposition. It should be noted, however, that even given
the conclusion of the ontological argument we cannot deduce from it
the principle needed by the cosmological argument without the aid of
the further principle that whatever is true of some necessary being is
true of all. In these exalted realms it is difficult to know which prin-
ciples are self-evident and which need proof.
So much for Kant's demonstration. On what grounds do Father
Johnston and Professor Smart reject it as fallacious? Here is Father
Johnston's criticism:
"When he [Kant] goes through the process of converting the proposi-

tion 'Necessary being is infinitely perfect being', he takes it for granted that
it is by this very proposition that the existence of necessary being is estab-
lished, entirely neglecting the fact that the existence of necessary being has
already been established by the previous step in the argument. When we say
'Necessary bein£ is infinitely perfect', we are speaking of a being already
known to exist.' 4
This is the whole of what Father Johnston has to say with regard to
the fallacy in Kant's reasoning; he does not say why he thinks that
Kant is taking it for granted that the existence of necessary being is
established by the proposition in question. But it would be curious if
Kant had forgotten already what he had written on the previous page,
where the problem is clearly stated-Given that we have proved the
existence of necessary being, how can we prove that this is the ens
realissimum?-and particularly curious since, as Smart points out, the
conversion of "Any necessary being is infinitely perfect" to "Some
perfect being is necessary" is valid only if a necessary being exists.
Smart's criticism adds nothing substantial to that of Father John-
ston: Kant has forgotten that the existence of a necessary being has
already been proved; he has been taken in by the ambiguity of proposi-
tions of the form "All As are Bs", which may mean either "There are
As, and they are all Bs", or "If there are any As then they are all Bs";
Kant "has taken himself and other people in by using 'all' sometimes in
the one way and sometimes in the other". 5
I can see no force whatever in these criticisms-there is no reason
to believe that Kant has forgotten or been taken in by anything in the
4 Johnston, op. cit., p. 15. Italics in the original.

5 Smart, op. cit., p. 37.
course of his demonstration. Granting for the sake of argument that
the first part of the cosmological argument has proved the existence of
a necessary being he then asks: "How can it be proved that a being of
this kind must be supremely perfect?" and he replies: "To prove that a
necessary being must be a perfect being amounts to proving that a
perfect being must be a necessary being-and this is just what the
ontological argument, and no other known argument, is intended to
do." It is irrelevant that in this case we already know that there is a
necessary being-the ontological argument is not designed to prove
simpliciter that there is a necessary being, and it is not assisted in
arriving at its own proper conclusion by the knowledge that there is a
necessary being.
That this is the line of Kant's reasoning is borne out quite clearly
by what he says in his introductory remarks to the discussion of the
three theological arguments:
" ... If we take the issue as being that which is here stated, namely,
first, that from any given existence ... we can correctly infer the existence
of an unconditionally necessary being; secondly, that we must regard a
being which contains all reality . . . as being absolutely unconditioned, and
that in this concept of an ens realissimum we have therefore found the con-
cept of a thing to which we can also ascribe absolute necessity-granting all
this, it by no means follows that the concept of a limited being . . . is for
that reason incompatible with absolute necessity." (A588, B616.)6
6 Kemp Smith gives "reality" as the last word of the quotation. This is ob-
viously a slip on his part-the German reads "Notwendigkeit".
Suggested Further Readings
Bennett, Jonathan, Kant's Analytic (Cambridge University Press,

1966).
Bird, Graham, Kant's Theory of Knowledge (Routledge & Kegan Paul,
1962).
Cassirer, H. W., Kant's First Critique (Macmillan, 1955).
Copleston, Frederick, A History of Philosophy, Vol. VI (Doubleday,
1963).
Korner, Stephan, Kant (Penguin, 1955).
Paton, H. j., Kant's Metaphysic of Experience, 2 vol. (Macmillan,
1936).
Smith, N. Kemp, A Commentary to Kant's Critique of Pure Reason
(Macmillan, 1918). .
Strawson, P. F., The Bounds of Sense (Methuen, 1966).
Walsh, W. H., Reason and Experience (Clarendon Press, 1947).
Weldon, T. D., Introduction to Kant's Critique of Pure Reason (Ox-
ford University Press, 1945).
Wolff, R. P., ed., Kant (Harvard University Press, 1963).
The reader's attention is also drawn to the excellent article by

M. j. Scott-Taggart, Recent Work on the Philosophy of Kant, in the
American Philosophical Quarterly (July 1966).
---~-----~--~~-- -~--
ISSUES
aristotle's AND WALSH
ethics INTERPRETATIONS & SHAPIRO
berkeley's
CRITICAL ENGLE
principles of
STUDIES & TAYLOR
human knowledge
STUDIES
hobbes's IN THE BAUMRIN
leviathan PHILOSOPHY
OF THOMAS HOBBES
STUDIES
human IN THE SESONSKE
understanding PHILOSOPHY & FLEMING
OF DAVID HUME
STUDIES
limits of RADCLIFF
OF MILL'S
liberty ON LIBERTY
~,,"-~,~<~-
. meta- STUDIES
SESONSKE
~itations
IN
DESCARTES & FLEMING
---
I, mill's TEXT SMITH
utilitarianism ~ AND CRITICISM & SOSA
I

Reflections On Kant'S Critique of Pure Reason Penelhum Macintosh

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Reflections On Kant'S Critique of Pure Reason Penelhum Macintosh

Uploaded by

Copyright:

Available Formats

the first

Critique of Pure Reason

BERKELEY'S PRINCIPLES OF HUMAN

THE FIRST CRITIQUE:

WADSWORTH PUBLISHING COMPANY, INC.

The idea of a series of Studies in Philosophical Criticism devel-

Lewis White Beck

Lewis White Beck

SUGGESTED FURTHER READINGS 147

Kant's Critique of Pure Reason, though one of the greatest philosophi-

• Reprinted by permission of the author and publisher from Journal of the

12 See E. W. Beth, The Foundations of Mathematics (Amsterdam, 1959) on

14 On tbe First Grounds of tbe Distinction of Regions of Space. The argu-

16 Frank P. Ramsey, The Foundations of Mathematics (London, 1931), 115-6.

coalitions. Kant's attempt to discover this common, but false, funda-

18 Critique of Pure Reason A 273 = B 327.

21 See my Commentary on Kant's Critique of Practical Reason (Chicago,

antithesis is that there is only a "causality of nature." Both of these

27 Critiq1!e of Pract!cal Reason, 98 (V, 95); Critique of Pure Reason, B xxix.

28 Critique, of Judgment § 77; Critique of Practical Reason, 103 (V, 100).

30 Studies in the Philosophy of Kant (New York, 1965),52 n.

What is a 'necessary truth', and what is a 'necessary proposition'?

• Reprinted by permission of the author and publisher from Mind (1958)

Kant held, at least in effect, that Leibniz's definition of 'necessary

What we need in this area is in the first place a division of

It is fair to say that a great deal of Kant's admitted failure to make

Kant Leibniz Kemp Smith

These need some further comment.

. . • Reprinted from Kant's Critique of Pure Reason, second edition, by per-

2 Kant realized the unclearness of Erscheinung and tried to do something

he wants to use a wider term he prefers Wahrnehmung, which Kemp

The whole of this story is probably best thought of by the

physiology and he would have referred a questioner to Haller's text-

In his brilliant essay on the Foundations of Arithmetic, Frege observes

2 Bertrand Russell, History of Western Philosophy and Its Connection with

6 Se~ Hannann Paul, Deutsches W orterbuch, Vierte Auflage Halle Saale

9 Regulae ad directionem ingenii, rule 3. Cf. L. J. Beck, The Method of

14 J. Chr. Adelung, Auszug aus dem grammatisch-kritischen Worterbuch

17 See Regulae, especially rules 5 and 6,

18 J. H. Lambert, Anlage zur Architectonic, oder Theorie des Ersten und

21 Cf. Prolegomena §§ 7-8; also A 19. As pointed out by H. J. Paton (Kant's

. 22 Witness also ~ant's emphasis on the phenomenal character of space and

26 See E. W. Beth, "Kants Einteilung der Urteile in analytische und syn-

Our interpretation of the role of the notion of intuition in Kant is

In recent years there has been widespread use of arguments described

a scandal to philosophy and to human reason in general that the existence

The transcendental deduction (along with the Refutation of Idealism)

• Reprinted by pennission of the author and publisher from the Journal of

4 When I speak of "the sceptic" I do not mean to be referring to any person,

5 R. Carnap, "Empiricism, Semantics and Ontology", Appendix A in M ean-

(1) We think of the world as containing objective particulars in a single

He emphasizes that this is a remark "about the way we think of the

( 6) Obj ects continue to exist unperceived.

Only on this understanding of the sceptic is there any plausibility in

6 See J. O. Urmson, "Some Questions Concerning Validity" is Essays in

(2) If we think of the world as containing objective particulars in a single

And again, necessarily:

(3) If we can reidentify particulars then we have satisfiable criteria on the

Strawson's argument actually stops here, thus showing that he regards

Either this is a suppressed premiss of Strawson's argument or it is what

9 Ibid., pp. 35-36.

Of any sentence that appears to say that it is logically impossible to know

It is a necessary (logical, or conceptual) truth, not a contingent one, that