This article was downloaded by: [Moskow State Univ Bibliote]

On: 24 November 2013, At: 12:35

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Contemporary Physics
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/tcph20

The philosophical basis of physics

a
Gerd Buchdahl
a
Cambridge University
Published online: 20 Aug 2006.

To cite this article: Gerd Buchdahl (1962) The philosophical basis of physics, Contemporary
Physics, 3:5, 382-394, DOI: 10.1080/00107516208205316

To link to this article: http://dx.doi.org/10.1080/00107516208205316

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms
& Conditions of access and use can be found at http://www.tandfonline.com/page/
terms-and-conditions
 The Philosophical Basis of Physics
by GERD BUCHDAHL
Cambridge University

Our first reaction to the very suggestion that physics is in need of a basis is
probably much like that of Sir Oliver Lodge when he wrote (in M a n and the
Universe, 1908, p. 50):
“ To say that a system does not rest upon one special fact is not to impugn its

stability. The body of scientific truth rests on no solitary material fact or group of
facts, but on a basis of harmony and consistency between facts: its support and ultimate
sanction is of no material character.”
So we might insist that physics is not in need of a basis, and that the paradigm
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

of something ‘ supported ’, either from above or below, is the wrong one. But
perhaps this is a misunderstanding; we cannot tell until we have seen this picture
in its historical setting. Perhaps that also will teach us something about the
possible justification of the picture, though we shall see that there is an important
element of insight in Lodge’s remark.

1. DEDUCTIVE AND NON-DEDUCTIVE APPROACHES

One source stimulating the search for foundations has always been the
pre-occupation with axiomatic systems. Plato already used geometry as his
primary model: scientific knowledge there comes to be defined in terms of
theorems derived from ‘ hypotheses ’; and to avoid an infinite regress he imports
the notion of a ‘ direct apprehension ’ of the primary hypotheses. Even in
those early beginnings, it is never clear whether ‘ apprehension ’ is a subject
for procedural instructions, or whether the reference is rather one to ‘ pre-
suppositions ’. Thus, when Aristotle alludes to the need for the ‘ intuition ’ of
the basic premisses, we cannot be sure whether he is drawing attention to the
need for a special (though mysterious) mental act which is itself the result of a
definite methodological enterprise, or whether he is only saying that possession
of such premisses carrying universality and necessity is an ultimate presuppo-
sition of scientific knowledge.
So far the logical aspect. But our mathematical model also had a more
direct and specific influence: under its prompting mathematics as such becomes
the ideal of knowledge. As an immediate consequence this raised the problem
of the relation of physics to mathematics. There are several traditional responses
to this question, among them the following:
( a ) The Pluto-Locke tradition: there exists a ‘ cut ’ between mathematics and
physics; physics is only a ‘ likely story ’, mere ‘ opinion ’, not yielding the
‘ certainty ’ of mathematical knowledge; itself ‘ certain ’ of course only
because abstracting from all physical attributes not in agreement with it.
(b) The Cartesian tradition: the identification of physical matter and geo-
metrical form; the ‘ mathematization of physics ’.
(c) The Galileo-Newton tradition : there is always some discoverable mathe-
matical formulation of physical facts, without the latter losing their own
physical status; ‘ physics in mathematical harness ’.
 The Philosophical Basis of Physics 383

The history of science reveals all three approaches. The works of the great
17th century methodologists and philosophers display the first brilliant sketches
defining the various methodological principles and precepts, and illustrating
the corresponding procedural routes. And because they come to it fresh and
with an open mind, their writings for ever repay renewed study; the doctrines
which they develop, and which later generations have turned into hardened
dogmas, we are here able to observe emerging and growing in their proper
context. They often therefore yield insights denied to us when concentrating
solely upon the more normal developments of later ages. Descartes’ scientific
and metascientific writing offers one of the best illustrations of the rich variety
of ideas introduced into the field of methodology and logic at that time.
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

A. Deductive approaches
“ Apprehension of first principles ” and “ mathematization of physics ”

become united in Descartes’ attempt to place the fundamental laws of dynamics,

and in particular the law of inertia (‘ Newton’s first law of motion ’) on a purely
conceptual basis, though this was of course not inconsistent with the belief in
the possibility of such laws at the same time yielding powerful explanations-as
indeed they were subsequently found capable of doing. The train of Descartes’
reasoning is briefly: given the idea of pure spatialization of matter, and its
characterization by a single additional attribute, viz. motion; defining this
motion as a ‘ state of matter ’, where a ‘ state ’ (like other ‘ properties ’) is subject
to the causal principle; adding the subsidiary definition of ‘ motion ’ as rate of
change of displacement with respect to time; and finally assuming space to have
Euclidean properties; then the first law is a direct logical consequence of these
premisses and concepts.(l)
This is quite in line with modern logical discussions of dynamics which
frequently consider its primary axioms to be implicit definitions of its key terms,
e.g. inertia, mass, time, force, energy, etc.(2) The most extreme form of such a
conceptual approach is in Eddington’s work, which interprets many of the basic
major principles of physics either as methodological prescriptions or as the out-
come of the choice of a certain mathematical system.
The feeling that the law of inertia is a ‘ fundamental ’ law, is probably due to
the fact that it may be made to appear (as in Descartes’ account) a special
‘ interpretation ’ of a version of the causal principle (“ what is real does not
change ”). Such principles form a sort of ‘ regulative component ’ of the
corresponding physical law ; they frequently function ‘ directively ’, leading to an
‘ anticipation ’ of the law, long before normal abstractive and experimental
procedures have been able to supply any ‘ constitutive support ’ for it. Galileo,
who is frequently credited with having discovered this law, certaintly never did
anything of the kind. And Newton simply gave it an experimental and
observational basis, in terms of the numerous verifiable deductions from his theory.
A note on concepts may here be in place. It is fairly easy to see that concepts
such as mass, force, energy, might be implicit ‘ functions ’ of corresponding
axioms, but this is less obvious when we generalize to include all scientific
concepts, particularly those which though being ‘ unobservables ’ we are more
often inclined to consider in rather more ‘ realist ’ terms, e.g. such substantival
entities as atoms, molecules, electrons. Nonetheless, according to some views,
to say that a certain theoretical entity ‘ exists ’, means no more than that certain
 384 Gerd Buchdahl

explanatory axioms about the properties of an otherwise unspecified ‘ something ’

are true, and furthermore, that these axioms imply certain theorems which allow
of independent and direct verification.
On the other hand, there are obvious differences between a ‘ substantival ’
type of concept such as a molecule, and ‘ adjectival ’ concepts like ‘ refractive
index ’ or ‘ gravitational force ’ which act merely as specifiable functions or
parameters of some corresponding laws which hold between a certain set of
spatio-temporally determined observables. T h e former are about certain
putatively existential entities; because of this (i.e. because of their ‘ substantival ’
nature) they do not explicitly appear in the symbolic formalism of a physical
theory. Thus, no symbol for ‘ molecule ’ (only for its mass and velocity, etc.)
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

is to be found in a quantitative version of the kinetic theory of gases. Like

Locke’s substance, being an “ unknown-I-know-what ”, it drops out of the
language of physics. Neither the ‘ non-existential ’ status of the ‘ adjectival ’
group, nor the non-representational aspect of the substantival group, deprives,
however, either group of what is often called ‘ reality ’ ; such judgements would
be purely arbitrary.(3)
B. Inverse-deductive (or ‘ retroductive ’) approaches
Descartes at times refers to the deductive approach as the ‘ synthetic method ’ ;
and he contrasts this with what he conceives to be a procedure moving in the
opposite direction, which he calls the ‘ method of analysis ’, or analytic method.
Its advantage, so he claims, lies in the fact that where applicable it actually
guides us in the discovery of ‘ causes ’ for already known physical ‘ effects ’.
(Cf. also Newton’s Opticks, Qu. 31, for the same idea). T h e deductive method
is mainly occupied with the formulation of explanatory systems through
axiomatization or with the laying of conceptual foundations for a science. T h e
analytical method emphasizes the primary need for a path leading to the discovery
of hypotheses in the light of phenomena requiring explanation. Descartes,
(like many modern writers) does not always distinguish too clearly between the
procedural aspect of the analytical method on the one hand, and its inferential
component on the other. For the present, we will concentrate on the first of
these, i.e. the actual construction of hypotheses.
Now there is no set of mechanical working-rules for the construction of
hypotheses. Descartes seems to have believed that he had discovered a universal
‘ rule of method ’, partly because what he had really discovered (or rediscovered)
was the notion of retroductive inference, according to which it makes sense to
hold that we can ‘ p r o v e ’ the hypotheses via the power of their verifiable
deductive consequences. This, whilst true, is no substitute for a method of
discovery. Rough hints exist in plenty: curve-fitting, the employment of
analogy, even the injection of regulative ideas, and countless others. Again,
by the ‘ analytic method ’ we might mean literally such things as the resolution
of complex motions into their simple components. This was the version
appearing in Newton’s Opticks, although Newton added to this “ the derivation
of the forces producing such motions ” : which is something that cannot be
obtained by the same type of ‘ resolution ’!
Still, there lies behind this notion of analysis a fundamental idea: the breaking
down of a complex of phenomena into its simple underlying operative factors.
And the fascinating thing is that Descartes actually presents his readers with an
 The Philosophical Basis of Physics 3 85

instance showing how this method may be understood to work in detail; the steps
in his discovery of the law of refraction, often called ‘ Snell’s Law’, are as follows.
Let us assume the discussion of refraction in terms of ‘ rays of light ’; mere
observation then will yield the (qualitative) information that the ray is broken
towards the normal to the interface between the media when entering the
‘ denser ’ medium (e.g. air into water). We now feed into this information an
account of the ‘hypothetical mechanism’ (and Descartes refers to it expressly as
a mere hypothesis) in terms of which we describe the passage of light. Descartes’
account here mentions an all-pervasive ether possessing a particulate structure,
the passage of light being defined as the passage of a ‘ shock-wave ’ travelling
from the luminous body to the illuminated surface. As the ether is not ‘ corn-
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

pressible’, the wave moves with ‘ infinite speed’. This is still difficult to
understand, for the concepts are not sufficiently clear. We shall therefore dig
still more deeply: to this end we must consider the passage of any natural power
be it a wave or a particle.
So far all is qualitative. Descartes now ‘ tightens ’ his account by intro-
ducing a geometrical model to derive the law of refraction quantitatively.
In other words, he develops a conceptual model to which he then applies those
general laws of motion which he had already developed elsewhere and which
he believed to be of very general application. His model does not exactly
reproduce the (already hypothetical) structure; the reason is that with the
absolute freedom of genius he considers the matter in terms of the motion of a
particle, of whose laws he has a knowledge which he lacks concerning the
hypothetical wave motion. The derivation is not ‘ pure ’ : there is (as in most
‘ real ’ scientific situations) considerable ‘ fitting ’ of the account to the desired
result; we need not reproduce it in any further detail. Yet the general procedure
is clear: the introducing of the hypothetical mechanism; the search for the
fundamental operative process, the tightening of this in terms of a quantitative
model, reasoning round which ultimately leads us back to the desired ‘ law ’
accounting for the qualitative and quantitative relationships. There are here
the essential ingredients of a genuine ‘ motion ’ from ‘ effect ’ to ‘ cause ’ : the
tentative introduction of a theoretical schema (making it once more possible to
‘ reason deductively ’) ; the rapid shuffling backwards and forwards between
phenomena and explanation; the ’side-reference to external principles; the
employment of models not necessarily isomorphic with the putative hypothetical
or theoretical structure.
The statement that this was an instance of a ‘ motion ’ from effect to cause,
needs some qualification. There is ‘ motion ’ in that direction only in the sense
that we began with the set of obskved pairs of angles of incidence and refraction,
rising from this to a consideration of the qualitative behaviour of light rays; and
we ended with the formula-‘ the hypothesis ’ in accordance with which the
values of the angles of incidence and refraction are computable. However,
there was clearly no step-by-step ascent in mechanical fashion from the explican-
dum to the hypothesis! Less still should the phrase ‘ motion from effect to
cause ’ mislead us into thinking that here is a method which guarantees a deduction
of the explanation from the explicandum. (Scientists often speak of a certain
law, e.g. the law of constant proportions, as having been inferred from certain
experimental results, a phrase which frequently causes much misunderstanding.)
But as already mentioned, Descartes’ notion of analysis also covered-in addition
C.P. 1B
 3 86 Gerd Buchdahl

to the procedural aspect-the idea of an inference (non-deductive !) from effect

to cause; and this idea is still with us: it goes under various names; we have
called it ' retroductive inference ', the better-known ' inductive inference ' being
another member of the same species. And we shall now turn to the problems
raised by the second, the inferential, aspect of analysis; problems which will
involve more explicitly the contentious notion of a ' philosophical basis '.

2. A LOGICAL PROBLEM : IS THERE A N INDUCTIVE INFERENCE ?

T h e fundamental logical operation involved in retroduction (as well as in-
duction) is the affirmation of the antecedent, given the consequent. Of course,
if we should prove that some given or presumptive construction alone would
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

account for a set of data, this would give us formal validity. Likewise, if an
hypothesis can be shown to imply a consequence contrary to fact, it is thereby
formally invalidated. But between these logical cornerstones there lies the
vaguer though more substantial region of positive affirmation. If the hypothesis
has led to verified data; and if it has not been falsified; and if it fits into a more
general network of other non-falsified hypotheses; in short if it comes up to the
standards of procedural performance discussed in section I, above: are we then
' entitled ' to ' infer ' that this hypothesis holds universally ?
At the outset, we must recognize that in recent times vital objections have
been raised against the whole conception of non-deductive inference as such:
objections which if sustained would go a long way towards showing that our
question was misconceived, even senseless, and with it the whole baneful notion
of a ' basis '. Moreover this suggestion is supported by the consideration that
no satisfactory answer to our question has ever really been proposed. These
objections occur in two versions, the first of which is due to Karl Popper, the
second to Gilbert Ryle, both representing a movement of considerable contem-
porary thought. (4)
(1) Since retroductive and inductive inferences are formally invalid, we cannot
thus define at all the form taken by scientific reasoning. Instead we need
to replace this by a formally valid inference schema, which involves a falsifica-
tion procedure, according to which a scientific test can yield an assured result
only if (and in the sense that) it disconjirms a proposed hypothesis. It is then
only a ' convention ' to hold on to an hypothesis so long as it has not been
disconfirmed; but as such, non-negative tests have no logical significance for
a verification-procedure. (5)
(2) This is a slightly more extreme version of the first, and holds that we should
disabuse ourselves of the idea that scientists at work ever pass on to the
business of making an inference of any such kind (excepting deductive
explications). Rather, they merely show that certain hypotheses work by
showing them working, i.e. through leading to verifiable consequences. We
can show our hypothesis working-i.e. observe it as ' work-in-progress ',
but we do not in addition pass on to any sort of inference: ' . . . therefore it
is true ' . ( 6 )
I am not sure that either of these objections to non-deductive inference are
really serious. Let us consider them briefly in turn. T h e first and the
methodological prescription which it entails, seem to possess a certain plausibility
for many scientists. This is due, I think, only to a lack of realisation of the
 The Philosophical Basts of Physics 387

extreme puritanism it involves. It is certainly true that we attempt to test an

hypothesis by pushing outward and thereby determining the frontiers beyond
which it may be found to break down. But that is not all this view really entails.
For the scientist, when he embarks on new tests extending over a novel range,
with the possibility that his hypothesis might suddenly fail, would certainly
take it that a non-negative test had implications for the range over which the
hypothesis is in fact believed to hold! Whereas it is a consequence of the
doctrine here in question, that a positive test (absence of falsification) contributes
strictly nothing towards any possible verification, towards any rational belief
in the strength of the hypothesis; for to do so we should have to assume that such
tests when repeated under identical conditions would yield similar (or statistically
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

predictable) results ; i.e. that the situation in question is putatively universal,

which is just the assumption which the present view was intended to avoid;
indeed, to make unnecessary.
Similar comments apply to the second objection. The proposition: ‘ The
theory works since it can be shown to work ’ is not tautologous; its successful
application on given individual occasions is taken to imply something concerning
its universal applicability.
That these doctrines when considered closely lead to situations which appear
to conflict with the normal logical instincts of scientists, is perhaps no refutation.
On the other hand, the fact that its plausibility for the scientist depends upon
the surreptitious re-introduction of the very concept whose contentious nature
had led to its attempted elimination (the concept of inference to a universal or
lawlike connection) suggests that we should pursue further the problems
surrounding it, in order to give ourselves the chance at least of facing clearly
the logical complications involved.
The question then is: how do we ‘ know ’ that a certain test under identical
conditions would yield identical results ? ‘ Know ’ is here put in quotation-
marks in order to remind ourselves that we are no longer concerned with the
usual every-day use of ‘ knowing ’. In other words, our enquiry now concerns
the ‘ basis ’ of the concept of ‘ uniformity’ or ‘ law’. It is important to
emphasize the conceptual nature of this enquiry because this fact is seldom
realized; we act as though we were concerned with the question: how can we
know that a certain law is true; or perhaps even “ necessary and eternal ”
(Galileo’s words). The answer is: by observation, experiment and theo-
rizing, and certain additional procedures of conceptualization which we have
briefly touched on in the first section. And this entails the consequence that
the conclusions thus arrived at carry a built-in corrigibility-proviso, according
to which-as Newton put it in his fourth ‘ Rule of Reasoning ’ in the PrinCipia
(Bk. iii)-
“ we are to look upon propositions inferred by general induction from phenomena
.
as accurately or very nearly true . . till such time as other phenomena occur by which
they may either be made more accurate or liable to exceptions ”.
Clearly our question i s not to guess what further data will go to support or
upset a given law ‘ inferred by general induction from phenomena ’, but rather,
what is meant by ‘inductive inference’, by the very conception of ‘ law’,
relative to individual phenomena, and what general considerations about the
world are presupposed in our attempt to make sense of the concept so
defined?
 388 Gerd Buchdahl

3. INDUCTIVE
INFERENCE

(1) Newton’s ‘ Rules ’ and the ‘ conventionalist ’ doctrine.

We have tried to separate the two main aspects of scientific thinking, the
procedural and the inferential; the latter will occupy us for the remainder of this
essay. Although Descartes had flirted with the problem of retroductive inference,
on the whole he had been content to fall back on a solution in terms of his ‘ dog-
matic ’ conception of the ‘ mathematization of physics ’. But it was the marriage
of the Galileo-Newton approach to the Plato-Locke tradition that was to put the
critical problem into a sharper focus. Newton’s pronouncements on induction
in the ‘ Rules of Reasoning ’ for the Principia, as well as passing references in
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

letters to some of his correspondents, mean that in science we must ‘ deduce

principles from the phenomena ’ in order that thereafter they be ‘ made general
by induction ’. This, whilst it helped to formulate the matter, did nothing to
explain the significance.(7) Perhaps this fact is itself of some importance. Such
logical formulations would then have to be understood as ‘ procedural gestures ’
(cf. Ryle’s ‘ work-in-progress ’ doctrine), in accordance with which hypotheses
formulated in the light of certain phenomena requiring explanation are to be
understood to possess a generalizing function, a fact expressed by rendering them
in universal form, or using them with universal import. T o say that these are
‘ procedural gestures ’ would then mean that it does not make sense to demand a
‘ justification ’ of these gestures; like social customs, they have social (e.g.
grammatical) purposes; and these at most require causal explanations. It is as
though we always demanded a justification for or explanation of any general prin-
ciple in terms of which we happen to account for some particular state of affairs.
This, so we then should be told by the ‘ conventionalists ’, would be something
of a philosophical joke, for general principles are simply introduced as ‘ instru-
ments ’ for explanations into the general social scientific context ; whilst we can
properly ask for explanations of individual cases, it is socially improper to ask
the same question concerning explanations of the explanations,-at least ; in
any ultimate sense.
These conventionalist positions seem very sophisticated, if only because they
appear to beg the very question at issue. T o consider the analogy last mentioned;
the argument would succeed here only if explanatory laws could be considered
exclusively through the use of the philosophical paradigm of ‘ reason-giving
devices ’, thereby treating them as social tools ; whereas there are counter-
instantial paradigms, according to which an explanatory law may be considered
(for instance) as a novel discovery, capable of falsification; and here (making in
fact use of the conventionalist mode of argument itself) it would be perfectly
proper to require a justification. This reflection once again does not constitute
a logical refutation. Nor is this serious since ‘ logical refutations ’ in philosophi-
cal disputes proverbially lack sustained force! O n the other hand, it suggests
two things: we should persist in taking the problems raised by the notions of
‘ law ’ and ‘ inference ’ seriously; but we should also endeavour at the same time
to learn the lesson that a ‘ solution ’ of the resulting conflicts will have to lean
heavily on the fact that the philosopher has a position of ‘ relative autonomy ’ in
terms of the paradigms through which he is going to consider his philosophical
material.
 The Philosophical Basis of Physics 389

( 2) The Newton-Locke tradition and the development of Logical Atomism.

The Newtonian approach of ‘ physics in mathematical harness ’, deprived
of the Cartesian solution, seemed to question why nature should be-in Galileo’s
words-“ written in the language of mathematics ”; a language whose model-
relation is that of logical entailment, of necessity? ; a language which operates
with functional relations represented by the general form y = f(x), interpreted as
a rule ‘ rigidly determining values of y , given values of x ’.
Here John Locke was the first to discuss critically the logical issues involved
in the Newtonian corpus of knowledge (in his celebrated Essay Concerning Human
Understanding, 1690). His solution (if it can be called such) is that we have
‘ real knowledge ’ only of ‘ mathematical truths ’, and that such knowledge
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

applies to ‘ real things ’ solely because we autonomously select only those features
of the physical world which agree with the mathematical ‘ archetypes ’. (Essay,
iv. 4, 6). This doctrine which has had a far-reaching influence down to our
own days, was epitomized in Kant’s dictum that every ‘‘ systematic study of
nature contains only as much science properly speaking as there is in it mathe-
matics applied to its data ” (Metaphysical Foundations of Science, Preface.(*)).
For Locke, this view had certain scepticist consequences. For it seemed to
follow that all ‘ knowledge ’ based on the empirical observation of material bodies
was “but judgement and opinion, not knowledge and certainty” (Essay, iv. 12.10).
“ Certainty and demonstration are things we must not, in these matters pretend

to ”, he wrote (Essay, iv. 3. 26); and his terminology suggests (and his argument
implies) that this is because information gathered through ‘ observation and
experiment ’ does not come up to the ideal of yielding necessary connexions ’
found in the ‘ demonstrations ’ of the mathematicians.
This again is a ‘ philosophical complaint ’, and not a ‘ scientific ’ one. Locke
was fully aware of the success of Newton’s Principia. His preoccupation is
with epistemological questions ; he struggles to hammer out an adequate con-
ception of ‘ scientific knowledge ’, a struggle which is brought on by the conflict
that he seems to find between the paradigm of mathematical knowledge on the
one hand, and what he calls experience ’ on the other.
The origins of this conflict lie in the Lockean conception of ‘ experience ’
as being equivalent to the perception and combination of ‘ ideas ’. These ideas
he views each as being quasi-self-contained logical ‘ atoms ’ ; they “ enter the
mind by the senses simple and unmixed” (Essay, ii. 2. 1.) Hence, complex
ideas (and most of our ideas, including ‘ generalizations ’, are complex) must be
compounded, the result of a ‘ subjective ’ synthesis.
Though based on small psychological evidence, this story is a myth for the
following doctrine :
( a ) All knowledge is based on experience. This entails that the character
of the known is essentially determined by the character of the ‘ process ’
of knowing, an important mark of which is that the knower is able to
consider the known as it is presented ‘ in itself ’, apart from all relations
to other knowable objects (or ideas). This doctrine of the ‘ presentational
immediacy ’ of the object of knowledge, together with the premisses just

t cf. Galileo’sreference to the laws of nature being‘necessaryandeternal’,whichwe might

here use, not with respect to the problem of verification of laws, but the meaning of the
concept.
 390 Gerd Buchdahl

stated, at once yield the ‘ atomicity ’ of all ideas as such. (Atomistic

ideas of course were a commonplace to the 17th century ‘ corpuscularians’.)
(b) I t is a consequence of (u) that relating our simple ‘ ideas ’ (of objects) into
complex ones of substances involves a subjective (‘ mental ’) factor, an
element of spontaneity on our part. (Locke calls it “ voluntariness ”.)
( c ) T h e relations whereby we unite our ideas in experience do not exhibit
the necessitarian character required by the argument from science, by
our conception of law; e.g. of causal law.
T h e answer to a possible objection to this conclusion will make the last point
clearer. For it may be said that surely it is ‘ in experience ’ that we find the
qualities of objects united for us ;the unity can hence be hardly of our own making.
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

This is however not Locke’s point. For he would certainly grant as much but
add that ‘ in experience ’ qualities are only united ‘ here-and-now ’, whereas
the unity which we mean by an ‘ object ’ entails a combination which holds
universally and with necessity (what Locke calls ‘ the real essence ’)-once more
a conceptual matter (Essay, iv. 12. 9). And this kind of unity is certainly not
‘ given ’ but is “ made voluntarily ” (Essay, ii. 12. 2), and hence though certainly
‘ real ’ is not ‘ objective ’. As he puts it at iv. 3. 14 of the Essay:
“ Let our complex idea of any species of substances be what it will, we can hardly,

from the simple ideas contained in it, certainly determine the necessary coexistence of
any other quality whatever. Our knowledge in all these enquiries reaches very little
farther than our experience.”
This, then, is the background to the problem with which philosophers since
the 17th century have been wrestling, and I think we understand it better by
tracing it back to its origins. What is the foundation of that relation (called by
the 17th and 18th century philosophers ‘ necessary connection ’, though we
might want to generalize this) in virtue of which we transcend the given moment
of ‘ presentational immediacy ’ (to use Whitehead’s term) ; more generally even:
what makes it possible-in the sense of logical justification-to speak of ‘ external
objects ’, of ‘ objective ’, lawlike connections ?

4. THE“ FOUNDATION
OF SCIENTIFIC
INFERENCE
”: SOME HISTORIC SOLUTIONS.
It would be possible to approach such questions in various ways.
(1) Investigate the general philosophical nature of such questions. For instance
we might pursue the consequences of arguments involving considerations of
philosophical paradigm-cases (cf. end of section 111) and reflect on the nature
and power of this type of philosophical treatment. For instance: what kind
of considerations, and what sort of answer do conceptual questions deserve,
once we have severed them (as we have found necessary) from their material
basis ?(9)
(2) Investigate the nature, meaning and logic of concepts like ‘induction ’,
‘ inference ’, ‘ law ’, ‘ necessity ’, etc.
(3) Ask what generates the idea of an objective causal connection, and what is
the general justification of this idea.
It is only during our own century that a conscious start has been made on
the exploration of the first two of our questions ;whilst the remaining question set
the tone in terms of which most philosophical systems of the 17th and 19th
centuries, from Hume through Kant, have debated this matter. Nor should
we think that their perspectives were necessarily fashioned by intellectual
 The Philosophical Basis of Physics 391

obtuseness. I n many cases the solutions ’ bear a surprising similarity to one

another once the temporal forms have been stripped away. Which explains the
importance for us of a renewed study of some of these earlier powerful and
penetrating attempts.
For instance, a great deal of time and ingenuity has been devoted in recent
times to the investigation of the logical nature of lawlike statements. Should
we think of these as empirical generalizations or as general schemata, ‘ methods
of representation ’ functioning perhaps like ‘ inference tickets ’ permitting the
making of predictions ?(lo)Or again, are lawlike statements a class of ‘descriptive
statements ’, or are they more like statements of an entailment relation, indicative
of some kind of necessity?(ll) How to explain the peculiar feature of ‘ unfulfilled
hypotheticals’ ? (12) When we say that the evidence supports the lawlike
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

hypothesis, is the relation involved between evidence and hypothesis characteri-

zable at all as deductive or inductive, or is it some quite different, as yet unspeci-
fied logical relation?(13) These are some of the formulations that have been
fashionable in fairly recent times. They are still the old questions dressed up in
new clothes: though this may help by suggesting new hints for treatment. But
in any case, the three questions listed above are after all closely related, and need
joint investigation. Thus, you cannot ask ‘what does or can one mean by
necessary connection ’, without also asking how one would justify such a thing.
Per contra, some celebrated philosophical versions (cf. the Kantian opus) have
justified these concepts by showing that an analysis of their meaning could be
obtained by bringing to light what was involved in thinking about science in
general.
So whilst our philosophical activity has certainly grown more like an apparently
verbal enquiry in recent times, it is not really that. It is still the old task of
discovering a meaning for, and a justification (or the impossibility of such a
justification) of a basis. Let this be our excuse, then, in concluding these notes
with a very brief account of what is still the most subtle and the most penetrating
treatment which our problem has received by civilized man, namely in the hands
of Hume and Kant.
Hume (in his Treatise of Human Nature, 1739) postulated from the start
that we do have this idea of causal, or necessary, connection. Its justification
cannot lie among the relations as they exist between objects in the external world:
this follows once more from their atomicity, and the consequent subjectivity or
spontaneity of our ‘ connecting ’ activity. We are forced to make, or at any
rate, we do make, inferences,-certainly in Newton’s sense of spontaneous
inductions ’. But whence shall we collect the ‘ necessity ’ of the connection
that would guarantee the possibility of the inductive inference ? Hume’s exciting
and revolutionary escape from this dilemma was that it is not the necessary
connection that has to be supplied to justify the inference but rather that the
making of the inference and the presence of the necessary connections are one
and the same thing; the latter, instead of being a presupposition of the former,
is necessarily and spontaneously given together with it. I n other words,
necessarily connecting is a universal characteristic of the inductively reasoning
animal. (Treatise, i. 3 . 14).
Now this answer is surprisingly similar to the one found in Kant’s Critique
of Pure Reason, (1781). Kant simply sharpens the contrasts; he sees more clearly
than his predecessors that the atomicity of ‘ ideas ’ is not a quasi-psychological
 392 Gerd Buchdahl

fact but a philosophical postulate which is forced upon us since without

it we cannot begin to talk of ‘ knowledge ’ at all-a surprising inversion of the
old complaint! His argument was that if you consider the concept of relation
as given together with the concept of the related (i.e. if both were viewed as
‘ given from without ’), this would have the logical consequence of automatically
introducing the sceptical doubt regarding the required necessitarian character
of the relation in general; Kant’s hidden premiss here being that only what is
contributed by the logical self (really, the logical forms of thought) is exempt
from surprises ; and is so contributed necessarily, then necessarily exempt !
Making a virtue of the results already achieved by Locke and Hume, he hence
considered the subjectivity of the concept of relation as the only guarantee of its
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

being such as it presents itself; and the thing is deprived of its arbitrariness by
showing that without the relation we could not speak of a scientific experience
at all; not a very surprising conclusion since Kant has so contrived his analysis
that what his sceptical doubt has taken away from the observational data, his
definition of knowledge must subsequently put back again.
Moreover, the ‘ necessitarian ’ character of the concept of relation is made
good also in a second sense by the argument that the character of the combination-
being ‘ spontaneous ’-would be wholly ‘ subjective ’ unless we conceive it in a
‘ determinate way ’. But that we so conceive it, is actually entailed by an analysis
of the character of even the simplest experience: say of the content of our thought
which asserts ‘ This is so ’: or, ‘ This follows that ’; or, This is simultaneous
with that ’. Thus, in the very conception of an objective state of affairs ’ (e.g.
of a ship going down a stream) there is contained the thought that your ideas-
antecedently conceived atomistically, and ‘ loose and separate ’-in the required
act of combination must be considered as related in a determinate order; the
model for this order being a ‘ rule ’ or ‘ function ’, the paradigm for which is
once again the formula of mathematics. (As with Locke and Hume, this ‘ act
of synthesis ’ is, I think, a philosophical myth, which, when translated into a
logical terminology, can be seen to yield some important results.) The specific
form of such a principle expressing a ‘ determinate order ’ would then depend
upon the type of experimental proposition involved. Thus, the statement
X is followed by Y ’ carries the built-in categorical form expressed in the
principle that ‘ everything that happens presupposes something which determines
it in accordance with a rule ’, the so-called principle of causation.?
This was however a costly answer. Remember that we began with the
complaint that whilst we could ‘ observe ’ united qualities in the moment of
presentational immediacy, we required a justification of the assumed feat of
universal and necessitarian connection,-or better : of this concept of connection;
a concept which entailed transcendence beyond the ‘ specious present ’. But
now we find that in order to guarantee the absolute tightness of the argument to
the ‘ reality ’ of the relation, the synthesis-Kant’s necessary synthesis-must

7 It will be noted that when thus stated, the principle is not inconsistent with the
statistical aspect of contemporary quantum physics, since the form and nature of the ‘ rule ’
is left open. It was an historical accident that Kant thought that the unique model for such
a rule would be Newton’slaws of motion. The correct question hence becomes :what should
be the preferred paradigms for causal action? In this way we can see some of the com-
temporary perplexities concerning causation and indeterminism might receive some
enlightening re-interpretations.
 The Philosophical Basis of Physics 393

take place, and is an absolute necessity, even in the moment of presentational

immediacy! In that case, however, we are surely left once more without an
answer to our earlier question, unless it can be shown that the concept of the
‘ universal ’ involved in the case of the ‘ specious present ’ is identical with that
needed by the supposed feat of ‘ transcendence ’. Now this is certainly not the
case, nor-in justice to Kant-did he claim that it was; on the contrary he
insisted that there was no such identity.
His solution was that the original problem of induction had been shown to
evaporate under the heat of his own analysis. In its place he had put, on the one
hand, the theory just sketched, which amounted to the proposition that there is a
sense in which forms of scientific reasoning have no basis-unless the spontaneous
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

generation of such forms on the part of the creative scientist be counted as such.
On the other hand, he simply offered his readers the consideration that the
discovery and verification of actual individual laws (as contrasted with the
postulation of the concept of lawlikeness) is purely a matter of application of
the methods of the scientists; methods involving as a matter of fact the procedures
of axiomtization and systematization, and the application of regulative ideas,
which we met with in Section I-the whole thing becoming thereby, it will be
seen, a slightly circular procedure.
The passage of this episode in the history of the problem is fairly typical of
other first rate philosophical problems, or of the same problem in different
epochs ; except that the Hume-Kant case-study offers us the supreme example
both of magnificent brilliance of performance yielding novel insights for all time
(achievements as well as failure being all the more obvious to us given our know-
ledge of hindsight), and also a clear picture of the mode whereby a relentless
pursuit of a problem when genuinely metaphysical comes to evaporate in a
peculiar way.
In the sequel one may discern one or two pronounced tendencies emerging
from the Hume-Kant ruin. There is on the one hand a strong feeling that it
isn’t so much that the problem of induction (and the related one of ‘ necessary
connexion ’) is insoluble, but rather that it is bereft of any genuine significance.
The most one will concede is that there is a definite meaning to the concept of
law, adding perhaps, like the Wittgenstein of the Tractatus, that
“ If there were a law of causality, it might run: ‘ There are laws of nature ’. But
of course one can not say this: it shows itself. In the terminology of Hertz one might
say: Only lawlike connexions are thinkable.” (op. cit., 6. 36-6. 361).
It will be seen that this rather Kantian passage (written over 40 years ago)
already contains the germ of the ‘ procedural ’ or ‘ work-in-progress ’ doctrines,
assuming that we can make sense of the notion of lawlike form. On the other
side, there are still many who persist in the less spectacular and somewhat tedious
task of elucidating the principles and modes of a justification of induction,
though deliberately avoiding for this purpose the concept of necessary connection,
and largely operating with some version of the idea of ‘uniformity’.(14) The
first group has a tendency to subside into more or less historical accounts of the
various forms of successful scientific argumentation; (15) the second aims at
formalization and quantification of certain types of argument, and particularly
emphasises the related problems of probability and statistical reasoning.
Between these, is anything of philosophical (as contrasted with technical)
importance still left?
 3 94 Gerd Buchdahl

I think it is possible that by a renewed study of some of the significant themes

and undertones among the episodes which have here come under review we may
begin to perceive new trends, and new hints for proper ways to tackle some of the
problems which we have sketched and which (it may be thought) have only been
lost or temporarily abandoned. These have simply needed stripping of their
contemporary and quite accidental formulations or fortuitous and encumbering
preoccupations with matters no longer of interest in our own day. Thus we
have seen that only a purely conceptual study has any relevance for the discussion
of the philosophical basis of scientific reasoning though (as the conventionalist
doctrine implied) we cannot sever this entirely from procedural analyses.
Secondly, we must understand that the resulting problematic concepts of induc-
Downloaded by [Moskow State Univ Bibliote] at 12:35 24 November 2013

tive inference, necessary connection, etc. emerge from subjective analyses of

structures (e.g. laws) which must be regarded as embedded in or related to
certain paradigm situations. This means, that such analyses are not ‘ God-
given ’, but introduced by us in the capacity of autonomous logicians.
This notion of ‘ autonomy ’ which emerges is of course familiar to us already;
for we found that it was the key-concept in the central arguments of the philoso-
phers from Locke to Kant. It seems likely therefore that a systematic applica-
tion of the notion might yield satisfactory formulations through which an alter-
native to the conventionalist position could be envisaged; an alternative which
would at least have the advantage of confirming the instinctive feeling of the
scientist, that the forms i n which he reasons (inductive and retroductive) are
‘ natural forms of thought ’; that truth is the issue, and not just ‘ procedural
rules ’.
REFERENCES
(1) BUCHDAHL, GERD,1962, Descartes’ Anticipation of a ‘Logic of Scientific Discovery’,
in A. C. Crombie (ed.), The Structure of Scientific Change (forthcoming).
(2) BUCHDAHL, 1951, Science & Logic: Brit. J. Phil. SOC.,2, 217-235.
(3) BUCHDAHL, 1959, Sources of Scepticism inAtomicTheory, Brit.J.Phil.Soc.,lO, 120-1 34.
(4) BUCHDAHL, 1960, Convention, Falsification & Induction, Aristot.Soc.,Sup.Vo1.113-130.
(5) POPPER, KARL,1959, The Logic of Scientific Discovery, Chs. 1-5, 10.
(6) RYLE,GILBERT, 1957, S. Koerner (ed.), Observation and Interpretation, 165-170.
(7) BUCHDAHL, 1956, Inductive Process & Inductive Inference, Aust.J.Phil., 34, 164-181.
(8) BUCHDAHL, 1961, The Image of Newton and Locke in the Age of Reason.
(9) BUCHDAHL, 1957, Science & Metaphysics, in D. Pears (ed.), The Nature of Metaphysics,
61-82.
(10) TOULMIN, S. E., 1953, The Philosophy of Science, 3.
(11) KNEALE,WILLIAM, 1949, Probability and Induction, 17-18.
(12) BRAITHWAITE, R. B., 1953, Scientific Explanation, 9.
(13) WAISMANN, F., 1951, “ Verifiability ”, in A. G. N. Flew (ed.), Essays in Logic and
Language, First Series, 116-144.
(14) STEBBING, L. S., 1930, A Modern Introduction to Logic, 21.
(15) HANSON,N. R., 1958, Patterns of Discovery, 2 4 .

The Author:
First studied engineering, and after some years, spent as a structural design engineer,
he succumbed to his interest in general philosophical matters by acquiring graduate and
postgraduate qualifications in philosophy. Between 1948-58 he built up the School of
History and Philosophy of Science in the University of Melbourne. He has been Visiting
Lecturer in Philosophy of Science at Oxford University. At present he is Lecturer in
Philosophy of Science, and in charge of History and Philosophy of Science and the
Whipple Science Museum, University of Cambridge, as well as Assistant Lecturer in
Moral Science, King’s College, Cambridge.