Annotations to Bhaskar’s Realist Theory of Scienceehrbar/rts.pdf · Post-Renaissance Europe,...

Annotations to Bhaskar’s

Realist Theory of Science

Hans G. Ehrbar

May 12, 2005

Contents

Prefaces i0.1 Bibliographical Remark . . . . . . . . . . . . . . . . . . . . . . . . . . i0.2 Preface to the First Edition . . . . . . . . . . . . . . . . . . . . . . . . ii0.3 Preface to the 2nd edition . . . . . . . . . . . . . . . . . . . . . . . . . xiii0.4 Footnotes to the Preface . . . . . . . . . . . . . . . . . . . . . . . . . . xiii

Introduction xv

2

CONTENTS 3

1 Philosophy and Scientific Realism 11.1 Two Sides of ‘Knowledge’ . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Three Traditions in the Philosophy of Science . . . . . . . . . . . . . . 81.3 The Transcendental Analysis of Experiences . . . . . . . . . . . . . . . 19

1.3.A The Analysis of Perception . . . . . . . . . . . . . . . . . . . . 211.3.B Analysis of Experimental Activity . . . . . . . . . . . . . . . . 25

1.4 Status of Ontology and its Dissolution in Classical Philosophy . . . . . 311.5 Ontology Vindicated and the Real Basis of Causal Laws . . . . . . . . 471.6 A sketch of a critique of empirical realism . . . . . . . . . . . . . . . . 681.7 Footnotes to Chapter 1 . . . . . . . . . . . . . . . . . . . . . . . . . . 76

2 Actualism and the Concept of a Closure 822.1 Introduction: On the Actuality of the Causal Connection . . . . . . . 822.2 Regularity Determinism and the Quest for a Closure . . . . . . . . . . 902.3 The Classical Paradigm of Action . . . . . . . . . . . . . . . . . . . . . 1032.4 Actualism and Transcendental Realism: the Interpretation of Normic

Statements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1202.5 Autonomy and Reduction . . . . . . . . . . . . . . . . . . . . . . . . . 1402.6 Explanation in Open Systems . . . . . . . . . . . . . . . . . . . . . . . 158

4 CONTENTS

2.7 Footnotes to Chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . . 1692.8 Appendix: Orthodox Philosophy of Science and the Implications of

Open Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

3 The Logic of Scientific Discovery 1963.1 Introduction: On the Contingency of the Causal Connection . . . . . . 1963.2 The Surplus-element in the Analysis of Law-like Statements: A Cri-

tique of the Theory of Models . . . . . . . . . . . . . . . . . . . . . . . 2043.3 Natural Necessity and Natural Kinds: The Stratification of Nature

and the Stratification of Science . . . . . . . . . . . . . . . . . . . . . . 2223.4 The Social Production of Knowledge by Means of Knowledge . . . . . 2493.5 Objections to the Account of Natural Necessity Proposed . . . . . . . 2673.6 The Problem of Induction . . . . . . . . . . . . . . . . . . . . . . . . . 2903.7 Footnotes to Chapter 3 . . . . . . . . . . . . . . . . . . . . . . . . . . 3083.8 Appendix: Natural Tendencies and Causal Powers . . . . . . . . . . . 3163.9 Footnotes to Appendix of Chapter 3 . . . . . . . . . . . . . . . . . . . 330

4 Metaphysics and the Philosophy of Science 3314.1 Footnotes to Chapter 4 . . . . . . . . . . . . . . . . . . . . . . . . . . 348

CONTENTS 5

5 Postscript to the Second Edition 3495.1 Footnotes for the Postscript . . . . . . . . . . . . . . . . . . . . . . . . 361

Bhaskar’s Bibliography 367

6 CONTENTS

Prefaces

0.1 Bibliographical Re-mark

The first edition [Bha75] and second edi-tion [Bha78] of Bhaskar’s Realist The-ory of Science are identical until p. 250(except for a short remark at the endof the preface). The second edition hasa postscript and an updated bibliogra-phy. The paragraph numbering given be-

low follows these editions. [Bha97] is areprint of the second edition. The collec-tion [ABC+98] has reprints of chapters1 and 2, page breaks of the form <16>come from this collection. The presentpdf edition is fairly accurate concerningthe text, but many diagrams are not yettranscribed.

There is a mailing list about criti-cal realism and Bhaskar’s work, with

i

ii PREFACES

archives at http://archives.econ.utah.edu/archives/bhaskar. Between July 2,1996 and March 1998 this mailing list hasbeen discussing the Realist Theory of Sci-ence.

0.2 Preface to the FirstEdition

The preface begins be mentioning a num-ber of “problems of philosophy” that haveeluded solution this far. A little later, in10:1, Bhaskar says that this book will notsolve these problems of philosophy, but itwill prepare the ground for their solution.

6:1 It has often been claimed, and per-haps more often felt, that the problems ofphilosophy have been solved. And yet, like

the proverbial frog at the bottom of the beermug, they have always reappeared.

There was a phase in recent philosophywhen it was widely held that the problemwas the problems and not their solution. Inpractice, however, this interesting idea wasusually coupled with the belief that termi-nation of philosophical reflection of the tra-ditional kind would be in itself sufficient toresolve the problems to which, it was held,philosophical reflection had given rise.

The next paragraph gives a number ofexamples where problems of philosophycrop up in the substantive sciences.

6:2 Whatever the merits of such a view ingeneral, it is quite untenable for any philoso-pher who is concerned with science. Forin one science after another recent develop-ments, or in some cases the lack of them,have forced old philosophical problems to the

0.2. PREFACE TO THE FIRST EDITION iii

fore. Thus the dispute between Parmenidesand Heraclitus as to whether being or be-coming is ultimate lies not far from the cen-tre of methodological controversy in physics;while the dispute between rationalists andempiricists over the respective roles of the apriori and the empirical continues to domi-nate methodological discussion in economics.Sociologists are making increasing use of theallegedly discredited Aristotelian typology ofcauses. And the problem of universals hasre-emerged in an almost Platonic form instructural linguistics, anthropology and de-velopmental biology. The spectre of deter-minism continues to haunt many of the sci-ences; and the problem of ‘free-will’ is still aproblem for psychology.

⇓ These problems should have givenrise to a ferment of activity in philoso-phy, but philosophy has been held back

by lacking tools and an inadequate con-ceptual equipment.

6:3/o In this context one might have ex-pected a ferment of creative activity withinthe philosophy of science, and to a degreethis has occurred. But the latter’s capac-ity for autonomous growth is limited. Forthe critical or analytical philosopher of sci-ence can only say as much as the philosoph-ical tools at his disposal enable him to say.And if philosophy lags behind the needs ofthe moment then he is left in the position ofa Priestley forced, by the inadequacy of hisconceptual equipment, to think of oxygen as‘dephlogisticated air’;1 or, of a Winch baffledby an alien sociology.2

Question 1 In the Preface of A Real-ist Theory of Science, Bhaskar writes thatphilosophy has not kept up with modernsciences. Modern sciences are plagued

iv PREFACES

by a number of ancient unresolved “pi-losophical problems.” Philosophy has notbeen able to solve these problems becauseits framework is too limited. Bhaskarproposes in his book to expand this frame-work and thus enable philosophy to betterunderstand science and also be helpful toit. What does the great advance in phi-losophy consist of, which RB alludes tohere?

Here is Doug Porpora’s answer tothis question on the Bhaskar list, onhttp://archives.econ.utah.edu/archives/bhaskar/1996-07-26.024/msg00052.htm:

I think Bhaskar makes the crucialpoint that contemporary philosophicalor meta-theoretical discussion has “priv-ileged” epistemology over ontology. The

focus has been almost exclusively on howwe know, marginalizing consideration ofwhat has to be the case ontologically forus to know at all.

The very word “objective” is generallyassumed to be an outdated, untenable,epistemological category – as in “thereis no objective or value-neutral” point ofview. But “objective” is also an ontolog-ical category: While all our ideas aboutthe world are our own creation (episte-mologically), we still make an ontologi-cal distinction among our ideas betweenthose that refer to what is dependent onour consciousness and those that refer towhat is independent of our consciousness.Today, these two senses of “objective” areconflated at the expense of ontology.

Bhaskar intends to recover the dis-

http://archives.econ.utah.edu/archives/bhaskar/1996-07-26.024/msg00052.htm

http://archives.econ.utah.edu/archives/bhaskar/1996-07-26.024/msg00052.htm

0.2. PREFACE TO THE FIRST EDITION v

tinction. And he is going to do soby focusing on how we conceptualizethat ontologically objective quality wecall causality. Positivism conceptualizedcausality as event-regularities, a concep-tion that presupposes a certain ontology.By examining that ontology, we can de-termine whether the conception is ten-able. Already, Bhaskar is suggesting thatthis conception will not be tenable andthat a post-positivist account of causal-ity is needed, which presupposes a dif-ferent, presumably more defensible ontol-ogy. That ontology will presumably alsomake better sense of what science is ac-tually doing.

7:1

Hegel may have exaggerated when he saidthat philosophy always arrives on the scene

too late.3 Yet there can be little doubt thatour theory of knowledge has scarcely come toterms with, let alone resolved the crises in-duced by, the changes that have taken placeacross the whole spectrum of scientific (andone might add social and political) thought.In this respect our present age contrasts un-favourably with both Ancient Greece andPost-Renaissance Europe, where there wasa close and mutually beneficial relationshipbetween science and philosophy. It is truethat in the second of these periods there wasa progressive ‘problem-shift’ within philos-ophy from the question of the content ofknowledge to the meta-question of its sta-tus as such.4 This shift was in part a re-sponse to the consolidation of the Newtonianworld-view, until by Kant’s time its funda-mental axioms could be regarded as a prioriconditions of the possibility of any empiri-

vi PREFACES

cal knowledge. However, those philosophersof the present who insist upon their total au-tonomy from the natural and human sciencesnot only impoverish, but delude themselves.For they thereby condemn themselves to liv-ing in the shadow cast by the great scientificthought of the past.

7:2/o Anyone who doubts that scientifictheories constitute a significant ingredient inphilosophical thought should consider whatthe course of intellectual history might havebeen if gestalt psychology had been estab-lished in place of Hartley’s principle of theassociation of ideas; or if the phenomena ofelectricity and magnetism had come to beregarded as more basic than those of im-pact and gravity; or if sounds and smellshad been taken as constitutive of the basicstuff of reality and the rich tapestry of thevisual-tactile world had been regarded, like

a Beethoven symphony or the perfume of arose, as a mere effect of those primary pow-ers. Suppose further that philosophers hadtaken biology or economics as their paradigmof a science rather than physics; or 16th not17th century physics as their paradigm of sci-entific activity. Would not our philosophicalinheritance have been vastly different? Asthis is primarily a problem for the philoso-phy of philosophy rather than the philosophyof science, I shall not dwell on this point fur-ther here. Its significance for our story willemerge in due course.

After some preliminaries, the followingparagraphs give an interesting overview:⇓ Primary aim of this book: devel-

opment of systematic realist account ofscience, which is an alternative to pos-itivism. Other positivism critiques arevulnerable to counterattack because they

0.2. PREFACE TO THE FIRST EDITION vii

do not go all the way as Bhaskar does.8:1 The primary aim of this study is the

development of a systematic realist accountof science. In this way I hope to pro-vide a comprehensive alternative to the pos-itivism that has usurped the title of sci-ence. I think that only the position devel-oped here can do full justice to the rational-ity of scientific practice or sustain the intelli-gibility of such scientific activities as theory-construction and experimentation. And thatwhile recent developments in the philosophyof science mark a great advance on posi-tivism they must eventually prove vulnerableto positivist counter-attack, unless carried tothe limit worked out here.

⇓ Subsidiary aim is to refute posi-tivism once-and-for-all. Bhaskar doesthis by describing the conditions un-der which positivism seems plausible and

then showing that these conditions (sig-nificance of experience) are not valid gen-erally but a special case (experience is of-ten insignificant).

8:2

My subsidiary aim is thus to show once-and-for-all why no return to positivism ispossible. This of course depends upon myprimary aim. For any adequate answer tothe critical meta-question ‘what are the con-ditions of the plausibility of an account of sci-ence?’ presupposes an account which is ca-pable of thinking of those conditions as spe-cial cases. That is to say, to adapt an imageof Wittgenstein’s, one can only see the fly inthe fly-bottle if one’s perspective is differentfrom that of the fly.5 And the sting is onlyremoved from a system of thought when theparticular conditions under which it makessense are described. In practice this task is

viii PREFACES

simplified for us by the fact that the condi-tions under which positivism is plausible asan account of science are largely co-extensivewith the conditions under which experienceis significant in science. This is of course animportant and substantive question whichwe could say, echoing Kant, no account ofscience can decline, but positivism cannotask, because (it will be seen) the idea of in-significant experiences transcends the verybounds of its thought.6

The next paragraph positions thisstudy in two strands of contemporarycriticism of the positivist view that “sci-ence has a certain base and deductivestructure”: one calling attention to thesocial character and the other to thestratification of science. This study is asynthesis of both.

8:3/o This book is written in the context

of vigorous critical activity in the philoso-phy of science. In the course of this thetwin templates of the positivist view of sci-ence, viz. the ideas that science has a certainbase and a deductive structure, have beensubjected to damaging attack. With a de-gree of arbitrariness one can separate thiscritical activity into two strands. The first,represented by writers such as Kuhn, Pop-per, Lakatos, Feyerabend, Toulmin, Polanyiand Ravetz, emphasises the social charac-ter of science and focusses particularly onthe phenomena of scientific change and de-velopment. It is generally critical of anymonistic interpretation of scientific develop-ment, of the kind characteristic of empiri-cist historiography and implicit in any doc-trine of the foundations of knowledge. Thesecond strand, represented by the work ofScriven, Hanson, Hesse and Harre among

0.2. PREFACE TO THE FIRST EDITION ix

others, calls attention to the stratification ofscience. It stresses the difference between ex-planation and prediction and emphasises therole played by models in scientific thought.It is highly critical of the deductivist view ofthe structure of scientific theories, and moregenerally of any exclusively formal accountof science. This study attempts to synthe-sise these two critical strands; and to show inparticular why and how the realism presup-posed by the first strand must be extendedto cover the objects of scientific thought pos-tulated by the second strand. In this wayI will be describing the nature and the de-velopment of what has been hailed as the‘Copernican Revolution’ in the philosophy ofscience.7

⇓ But the new element Bhaskar bringsin is: epistemology presupposes an ontol-ogy. Causal laws are tendencies of things,

not constant conjunctions of events.

9:1/o To see science as a social activity,and as structured and discriminating in itsthought, constitutes a significant step in ourunderstanding of science. But, I shall argue,without the support of a revised ontology,and in particular a conception of the world asstratified and differentiated too, it is impos-sible to steer clear of the Scylla of holding thestructure dispensable in the long run (backto empiricism) without being pulled into theCharybdis of justifying it exlusively in termsof the fixed or changing needs of the scientificcommunity (a form of neo-Kantian pragma-tism exemplified by e.g. Toulmin and Kuhn).In this study I attempt to show how sucha revised ontology is in fact presupposed bythe social activity of science. The basic prin-ciple of realist philosophy of science, viz. thatperception gives us access to things and ex-

x PREFACES

perimental activity access to structures thatexist independently of us, is very simple. Yetthe full working out of this principle implies aradical account of the nature of causal laws,viz. as expressing tendencies of things, notconjunctions of events. And it implies thata constant conjunction of events is no morea necessary than a sufficient condition for acausal law.

⇓ This study focuses on science be-cause one needs a philosophy of sciencefirst, before one can have a theory ofknowledge. This also makes philosophyrelevant for science, its underlaborer.

10:1 I do not claim in this book to solveany general problems of philosophy. It is myintention merely to give an adequate accountof science. Philosophers, including philoso-phers of science, have for too long regardedthe philosophy of science as a simple substi-

tution instance of some more general theoryof knowledge. This is a situation which hasworked to the disadvantage of both philoso-phy and knowledge. If, however, we reversethe customary procedure and substitute themore specific ‘science’ (or even better ‘sci-ences’) for ‘knowledge’, considerable illumi-nation of many traditional epistemologicalproblems can, I think, be achieved. Andsome even, in so far as the ‘knowledge’ weare concerned with is that produced by ‘sci-ence’, become susceptible of definitive solu-tion. The result of this reversal will also bea philosophy which has a greater relevancethan is the case at present for scientific prac-tice. In this sense my objective could be saidto be a ‘philosophy for science’.

My comment about this passage on thebhaskar mailing list:

Bhaskar does not say here why the re-

0.2. PREFACE TO THE FIRST EDITION xi

versal will yield “considerable illumina-tion.” This is a preface, and many thingsare only hinted at. Perhaps we should,in this collective reading, try to fill insome of the gaps, i.e., inform each otherof what we are reading into the text.

Bhaskar claims that one should do the-ory of science before a more general the-ory of knowledge. Why should one? Mytake on it is the following, and I am notsure how close this is to what Bhaskarwould argue: Science is a process bywhich society acquires knowledge aboutthe world. It is “between” society andthe outside world. Much of nonscientificknowledge is communication “within” so-ciety. In order to gain knowledge aboutthe world, society has to overcome ob-stacles and acquire tools which will then

also shape its intra-social communica-tions, but which it would not be able toacquire by such communications alone.Have I made myself clear? To me, theprimacy of a theory of science within atheory of knowledge is closely related tothe primacy of the relations of produc-tion within the broader circle of all socialrelations.

Doug Porpora replied as follows:Hans, my reaction to the passage you

cite was different. I thought Bhaskarwas saying that scientific knowledge hasbeen subsumed under the larger questionof knowledge per se and that progresson that larger question has stagnatedaround such formulations as “knowledgeas justified true belief.”

So what Bhaskar is proposing is to be-

xii PREFACES

gin fresh by examining only what knowl-edge is for science. That is also a verystrategic point of entry because scien-tific knowledge is a paradigmatic form ofknowledge. Perhaps, all would agree thatscience is accomplishing at least some-thing we might call knowledge. Thus,what we learn about knowledge from sci-ence may help us know about knowledgemore generally. That is how I understoodBhaskar’s philosophical “reversal.”

⇓ Now some important remarks re-garding “underlaboring”:

For I willingly confess to Lockean motives.That is to say, I believe it to be an essential(though not the only) part of the business ofphilosophy to act as the under-labourer, andoccasionally as the mid-wife, of science.8 Ihave therefore tried in this study both to re-

late the philosophy of science to the moregeneral historical concerns of philosophy;and at the same time to indicate more pre-cisely than is usual the consequences for sci-entific practice of the methodological strate-gies implied by different philosophies of sci-ence.

⇓ This preface concludes with the re-mark that science is frail, but human ex-istence is generally frail, and science maystill be our best hope.

10:2/o We are too apt to forget the frailtyof both our science and our philosophy.There can be no certainty that they will sur-vive and flourish; or, if they do, that theywill benefit mankind. Civilisation is, likeman himself, perhaps nothing more than atemporary rupture in the normal order ofthings.9 It is thus also part of the job ofthe philosopher to show the limits of science.

0.4. FOOTNOTES TO THE PREFACE xiii

And, in this broader sense, to seek to ensurethat the Owl of Minerva takes flight beforethe final falling of the dusk.

11:1 I would like to take this opportunityto express my thanks to Alan Montefiore andRom Harre for reading earlier versions ofthis work; to Rom Harre and Hilary Wain-wright for their continual encouragement; tomany other colleagues and friends for theirhelp; and to Mrs E. Browne for typing themanuscript.

11:2 ROY BHASKAR University of Edin-burgh April 1974

0.3 Preface to the 2ndedition

11:3 This edition includes a postscript andan index. The postscript enables me to crit-

ically comment on the book. The index fills amajor lacuna in the first edition of the work.Francis Roberts and Robin Kinross helpedme to compile it.

11:4 ROY BHASKAR University of Edin-burgh September 1977

0.4 Footnotes to thePreface

1 See e.g. S. E. Toulmin, ‘Crucial Experiments:Priestley and Lavoisier’, The Journal of the His-tory of Ideas, Vol. XVIII (1957), pp. 205–20; andJ. B. Conant, The Overthrow of the PhlogistonTheory.

2 P. Winch, The Idea of a Social Science, p.114.

3 G. W. F. Hegel, Philosophy of Right, Pref-ace.

4 Cf. G. Buchdahl, Metaphysics and the Phi-

xiv PREFACES

losophy of Science, p. 25 L. Wittgenstein, Philosophical Investiga-

tions, 309.6 I. Kant, Critique of Pure Reason, Preface to

the 1st Edition.7 R. Harre, Principles of Scientific Thinking,

p. 15.8 J. Locke, Essay Concerning Human Under-

standing, Epistle to the Reader.9 Cf. M. Foucault, The Order of Things, p.

XXIII.

Introduction

In order to get away from get awayfrom the positivistic view of science, anew understanding of causal laws is nec-essary. Hume’s theory identifying causallaws with constant conjunctions of eventsis still widely accepted. Modern cri-tiques of Hume say that constant con-junctions are not sufficient for causal laws(they distinguish correlation and causa-tion). Bhaskar will argue they are not

even necessary.12:1 The aim of this book is the devel-

opment of a systematic realist account ofscience. Such an account must provide acomprehensive alternative to the positivismwhich since the time of Hume has fashionedour image of science. Central to the posi-tivist vision of science is the Humean theoryof causal laws. It is a principal concern ofthis study to develop some new argumentsand show how they relate to more familiar

xv

xvi INTRODUCTION

ones against this still widely accepted the-ory. In particular I want to argue that notonly is a constant conjunction of events not asufficient, it is not even a necessary conditionfor a scientific law; and that it is only if wecan establish the latter that we can providean adequate rationale for the former. It hasoften been contended that a constant con-junction of events is insufficient but it hasnot so far been systematically argued thatit is not necessary. This can, however, beshown by a transcendental argument fromthe nature of experimental activity.

Question 2 Bhaskar says that constantconjunctions of events are neither neces-sary nor sufficient for causal laws. Giveboth halves of this argument. Which halfis Bhaskar’s innovation?

Question 3 Bhaskar says: it can beshown by a transcendental argument fromthe nature of experimental activity that aconstant conjunction of events is not anecessary condition for a scientific law.Give this argument.

12:2 It is a condition of the intelligibilityof experimental activity that in an experi-ment the experimenter is a causal agent of asequence of events but not of the causal lawwhich the sequence of events enables him toidentify. This suggests that there is a on-tological distinction between scientific lawsand patterns of events.

⇑ Since it takes so much work on thepart of the experimenter to translate alaw into a constant conjunction of events,laws and constant conjunctions cannot beone and the same.

xvii

Obviously this creates a prima facie problemfor any theory of science.

This prima face problem is: which real-ity does a law have if it is not a constantconjunction of events?I think that it can be solved along the fol-

lowing lines:

Look at a law. Implicit in a law isalways a theory. A theory is usuallythe description of some natural mecha-nism at work. Bhaskar claims now thatsome of these mechanisms are not sim-ply as-if scenarios allowing one to orga-nize one’s thoughts but that they corre-spond to real natural mechanisms goingon in the world. Whenever we talk abouta natural necessity we refer to the work-ing of such a mechanism.To ascribe a law one needs a theory. For it

is only if it is backed by a theory, containinga model or conception of a putative causal orexplanatory ‘link’, that a law can be distin-guished from a purely accidental concommi-tance. The possibility of saying this clearlydepends upon a non-reductionist conceptionof theory. Now at the core of theory is a con-ception or picture of a natural mechanism orstructure at work. Under certain conditionssome postulated mechanisms can come to beestablished as real. And it is in the workingof such mechanisms that the objective basisof our ascriptions of natural necessity lies.

⇓ Here I propose to look at the sec-ond and third sentence first, and thenat the first sentence. Positivism, withall its emphasis on the experiment, can-not answer the following question: if ittakes so much effort by the experimenterto elicit the constant conjunctions, why

xviii INTRODUCTION

should his findings have any relevance forsomeone outside such controlled experi-mental situations? To answer this, youneed indeed the independence of the lawfrom constant conjunctions: experimentsare only relevant because the laws, whichthe experimenter has to spend so muchcare to expose, remain active outside theexperimental setup. Outside the experi-mental situation, they do not give rise toconstant conjunctions, but they are stillactive.

13:1 It is only if we make the assump-tion of the real independence of such mech-anisms from the events they generate thatwe are justified in assuming that they en-dure and go on acting in their normal wayoutside the experimentally closed conditionsthat enable us to empirically identify them.

But it is only if we are justified in assum-ing this that the idea of the universality of aknown law can be sustained or that experi-mental activity can be rendered intelligible.Hence one of the chief objections to posi-tivism is that it cannot show why or the con-ditions under which experience is significantin science. Most critics have emphasized itsdepreciation of the role of theory; this ar-gument shows its inadequacy to experience.Moreover it is only because it must be as-sumed, if experimental activity is to be ren-dered intelligible, that natural mechanismsendure and act outside the conditions thatenable us to identify them that the applica-bility of known laws in open systems, i.e. insystems where no constant conjunctions ofevents prevail, can be sustained. This hasthe corollary that a constant conjunction ofevents cannot be necessary for the assump-

xix

tion of the efficacy of a law.

The first two sentences in the nextparagraph go over the same argumentagain:

13:2 This argument shows that real struc-tures exist independently of and are oftenout of phase with the actual patterns ofevents. Indeed it is only because of the lat-ter that we need to perform experiments andonly because of the former that we can makesense of our performances of them.

⇑ Real structures (or, as Bhaskar alsocalls them, generative mechanisms) areoften out of phase with the actual pat-terns of events: this is why experimen-tal work is needed to expose them. Andthese generative mechanisms exist in-dependently of the actual patterns ofevents: this is why their knowledge is use-

ful also outside the experiment. This isBhaskar’s distinction of the real, i.e., theunderlying mechanisms, and the actual,i.e., the events that actually happen.

⇓ Then Bhaskar does some handwav-ing to derive the difference between theactual and the empirical: he says that asimilar distinction must be made betweenthe actual, i.e., that what happens, andthe empirical, i.e., that what is perceived(this is why scientists need training).

Similarly it can be shown to be a conditionof the intelligibility of perception that eventsoccur independently of experiences. And ex-periences are often (epistemically speaking)‘out of phase’ with events - e.g. when theyare misidentified. It is partly because of thispossibility that the scientist needs a scientificeducation or training.

xx INTRODUCTION

This is how he derives that the domainsof the real, actual, and empirical are dis-tinct.Thus I will argue that what I will call the

domains of the real, the actual and the em-pirical are distinct. This is represented inTable 0.1 below: –

13:3

Question 4 Re-phrase in your ownwords Bhaskar’s derivation of the separa-tion of the world into the three domains:the empirical, the actual, and the real.

⇓ Now Bhaskar goes one step further inhis theory of what laws really are: theyare not just mechanisms hanging in theair, but tendencies inherent in the waythings act. They lead to normic or trans-

factual statements about the world.

14:1 The real basis of causal laws are pro-vided by the generative mechanisms of na-ture. Such generative mechanisms are, it isargued, nothing other than the ways of act-ing of things. And causal laws must be an-alyzed as their tendencies. Tendencies maybe regarded as powers or liabilities of a thingwhich may be exercised without being man-ifest in any particular outcome. The kind ofconditional we are concerned with here maybe characterised as normic. They are notcounter-factual but transfactual statements.Nomic universals, properly understood, aretransfactual or normic statements with fac-tual instances in the laboratory (and perhapsa few other effectively closed contexts) thatconstitute their empirical grounds; they neednot, and in general will not, be reflected inan invariant pattern or regularly recurring

xxi

Table 0.1Domain of

Real Actual EmpiricalMechanisms XEvents X XExperiences X X X

sequence of events.

⇓ Back to a critique of Hume. It seemsto be a summary. A closed system is asystem which has no input from the out-side, so that constant conjunctions arepossible in it. Now I will do Bhaskar’ssecond point first because it seems eas-ier to understand for me. Weakness ofHume is: if laws are valid only in closedsystems, then why is it relevant to know

the laws for practical activity that takesplace in open systems? And how canone explain what the experimenter doeswith all the necessary preparations be-fore the constant conjunction can actu-ally be observed? Note the parentheti-cal remark Bhaskar makes in paragraph15:2 that experimental activity presup-poses open systems! To resolve this, whatone has to do is to interpret laws in a

xxii INTRODUCTION

transfactual way. But once one goes be-hind the facts, this is where one has tolocate a natural necessity in the world,not just in our brains. Perhaps one cansay: as soon as one no longer identifiesempirical regularities with laws, but saysthat empirical regularities are the man-ifestation of some transfactual tendency,then there must be something that hasthis tendency. Such a tendency cannotjust hang in empty space, it must be theway of acting of something (see precedingparagraph 14:1).

14:2 The weakness of the Humean conceptof laws is that it ties laws to closed systems,viz. systems where a constant conjunction ofevents occurs. This has the consequence thatneither the experimental establishment northe practical application of our knowledge in

open systems can be sustained. Once we al-low for open systems then laws can only beuniversal if they are interpreted in a non-empirical (trans-factual) way, i.e. as desig-nating the activity of generative mechanismsand structures independently of any partic-ular sequence or pattern of events. But oncewe do this there is an ontological basis fora concept of natural necessity, that is neces-sity in nature quite independent of men orhuman activity.

⇓ After this summary he continues theargument, goes with diagram 0.1, whichis not yet typed in here, but a copy ofthis diagram is also used in chapter 2 andtherefore can be found in the EssentialReadings p. 50.

14:3/o In science there is a kind of di-alectic in which a regularity is identified, aplausible explanation for it is invented, and

xxiii

the reality of the entities and processes pos-tulated in the explanation is then checked.This dialectic is illustrated in Diagram 0.1below.

15:1 Diagram 0.1. The Logic of ScientificDiscovery

Three steps, Bhaskar calls it dialecticbecause it leads from empirics to somereal entities which become the empiricsin the next step.

15:2 If a classical empiricist tradition inthe philosophy of science stops at the firststage, a rival neo-Kantian or transcendentalidealist tradition (discernible in the historyof the philosophy of science) stops at the sec-ond. If and only if the third step is taken anddeveloped in the way indicated above canthere be an adequate rationale for the use oflaws to explain phenomena in open systems,where no constant conjunctions prevail.

Question 5 Why is it necessary tocheck the reality of the underlying mech-anism which a theory postulates as expla-nation of certain phenomena? Is it notenough to verify that the postulated un-derlying mechanism leads to the phenom-ena to be explained?

⇓ In the remainder of this paragraphBhaskar rounds off and concludes his ar-gument. Most glaring weakness of or-thodox philosophy of science is: if onetakes Hume’s definition of causal laws se-riously, one must conclude that there areno causal laws in the sciences. Thereare no univeral empirical regularities, butphilosophers postulate those and there-fore get into troubles whenever they tryto confirm a theory or define what pro-

xxiv INTRODUCTION

cedures are scientifically rational. Or-thodox philosophy thinks the world is aclosed system (which means that every-thing has been predetermined since theBig Bang), and they ignore the fact whicheverybody knows that usually, constantconjunctions of events will only comeabout as the result of experimental activ-ity. The experimenter’s activity, like hu-man activity in general, is a causal agentwhich changes the course of the world(but this kind of activity is so relevantexactly because it leads to a display ofa natural law which is not of the experi-menter’s making!)

It is the unthinking presupposition of closedsystems together with the failure to anal-yse experimental activity (which presup-poses open systems) that accounts for the

most glaring weakness of orthodox philoso-phy of science: viz. the nonexistence in sci-ence of Humean causal laws, i.e. of universalempirical generalizations, and hence the in-adequacy of the criteria of explanation, con-firmation (or falsification), scientific ratio-nality etc., that are based on the assump-tion that a closure is the universal rule ratherthan the rare and (for the most part) artifi-cially generated exception that I contend itis. It is because our activity is (normally) anecessary condition of constant conjunctionsof events that the philosophy of science needsan ontology of structures and transfactuallyactive things.

⇓ The next paragraph is still about di-agram 0.1: How transcendental realismdiffers from empiricism and transcenden-tal idealism.

15:3 The position advanced here is charac-

xxv

terized as transcendental realism, in opposi-tion to the empirical realism common to theother two traditions. Both the neo-Kantianor transcendental idealist tradition and tran-scendental realism see the step between (1)and (2) in Diagram 0.1 as involving creativemodel building, in which plausible genera-tive mechanisms are imagined to produce thephenomena in question. But transcendentalrealism sees the need for the step between(2) and (3) also, in which the reality of themechanisms postulated are subjected to em-pirical scrutiny. Transcendental realism dif-fers from empirical realism in interpreting(1) as the invariance of an (experimentallyproduced) result rather than a regularity;and from transcendental idealism in allow-ing the possibility that what is imagined in(2) need not be imaginary but may be (andcome to be known as) real.

⇑ Here the slashed expressions re-sult/regularity and imagined/imaginaryin the diagram are significant: empiri-cism considers step (1) as a regularity,TR as a result of a causal law that maynot necessarily be regular. Transcenden-tal idealism sees (2) as necessarily imagi-nary, while TR says it is at first imaginedbut may come to be known as real.⇓ The last sentence in this paragraph

introduces a theme that will discussedfurther in the next paragraph: with-out all this one cannot explain scientificgrowth and change. I.e., the necessityof step (3) is therefore not only linkedinto the stylized facts about experimentalactivity, but also into the stylized factsabout the growth and change of science.Without such an interpretation it is impos-

xxvi INTRODUCTION

sible to sustain the rationality of scientificgrowth and change.

Question 6 Bhaskar claims that scien-tific growth and change cannot be ex-plained without transcendental realism.Why not?

⇓ This dialectic has no end.15:4/o A conception of science is argued

for in which it is seen as a process-in-motion,with the dialectic mentioned above in prin-ciple having no foreseable end. Thus whena new stratum or level of reality has beendiscovered and adequately described sciencemoves immediately to the construction andtesting of possible explanations for whathappens at that level. This will involvedrawing on whatever cognitive equipment isavailable and perhaps the design of new ex-perimental techniques and the invention of

new sense-extending equipment. Once theexplanation is discovered science then moveson to the construction and testing of possi-ble explanations for it. At each level of real-ity law-like behaviour has to be interpretednormically, i.e. as involving the exercise oftendencies which may not be realised.

⇓ Now Bhaskar introduces the conceptof epistemic fallacy. The dogma thatstatements about being can always be re-formulated so that they become state-ments about our knowledge about be-ing denies a second-order, transcendentalknowledge about the world that can begained from observing scientists at work.

16:1 Empirical realism is underpinned bya metaphysical dogma, which I call the epis-temic fallacy, that statements about beingcan always be transposed into statementsabout our knowledge of being. As ontology

xxvii

cannot, it is argued, be reduced to episte-mology this mistake merely covers the gen-eration of an implicit ontology based on thecategory of experience; and an implicit re-alism based on the presumed characteristicsof the objects of experience, viz. atomisticevents, and their relations, viz. constant con-junctions.

⇑ Let me try to explain it this way:One argument against the primacy of thequestion “what must the world be like forscience to be possible” is that everythingwe know goes through science and there-fore what we say about the world is onlya subset of what we may say about sci-ence, epistemology crowds out ontology.Bhaskar maintains that one has to lookat the ontology separately, and that theepistemic fallacy only covers up, prevents

the recognition of, an implicit empiricistontology.(These presumptions can, I think, only

be explained in terms of the need feltby philosophers for certain foundations ofknowledge.) This in turn leads to the gener-ation of a methodology which is either con-sistent with epistemology but of no relevanceto science; or relevant to science but more orless radically inconsistent with epistemology.So that, in short, philosophy itself tends tobe out of joint with science.

⇑ The sentence in parentheses is inter-esting: philosophers feel the need to baseknowledge on some certainty, thereforethey postulate atomistic events relatedby constant conjunctions. This false andunreflected ontology leads to misguidedmethodology (but he does not elaboratehere).

xxviii INTRODUCTION

Question 7 What is the epistemic fal-lacy? What are its implications for sci-ence and for philosophy?

The next paragraph explains why isit wrong to speak about the “empiricalworld.” A more detailed answer to thisquestion will be given in 27:3/o.

16:2 It is argued in Chapter 1 that thevery concept of the empirical world embodiesa category mistake, which depends upon abarely concealed anthropomorphism withinphilosophy; and leads to a neglect of theimportant question of the conditions underwhich experience is in fact significant in sci-ence. In general this depends upon an-tecedent social activity. Neglect of this activ-ity merely results in the generation of an im-plicit sociology, based on an epistemologicalindividualism in which men are regarded as

passive recipients of given facts and recordersof their given conjunctions.

Because, look in chapter 1, 27:3/o oraround there. One reason is: experiencesmay not be significant, it is the result ofprior scientific activity to sort out whichexperiences are significant. (This is rele-vant for class consciousness, also explainswhy sciences can progress in a completelydifferent direction for a long time.) Herehe speaks of epistemological individual-ism, to be distinguished from method-ological individualism, i.e., neglect of an-tecedent social activity. (Leads to an im-plicit sociology.)

16:3/o Against this it is argued thatknowledge is a social product, produced bymeans of antecedent social products; butthat the objects of which, in the social ac-

xxix

tivity of science, knowledge comes to be pro-duced, exist and act quite independently ofmen.

⇑ The “of which” construction is per-haps difficult to understand. I think hemeans that the objects whose knowledgecomes to be produced in the social activ-ity of science exist and act quite indepen-dently of men.⇓ Next he introduces the transitive di-

mension. Two criteria for adequacy of anaccount of science.These two aspects of the philosophy of sci-

ence justify our talking of two dimensionsand two kinds of ‘object’ of knowledge: atransitive dimension, in which the object isthe material cause or antecedently estab-lished knowledge which is used to gener-ate the new knowledge; and an intransitivedimension, in which the object is the real

structure or mechanism that exists and actsquite independently of men and the condi-tions which allow men access to it. Thesedimensions are related in Chapter 3. Twocriteria for the adequacy of an account of sci-ence are developed: (i) its capacity to sustainthe idea of knowledge as a produced meansof production; and (ii) its capacity to sustainthe idea of the independent existence and ac-tivity of the objects of scientific thought.

⇓ Laws do not refer to the empirical oractual but the real:

17:1 It is the overall argument of thisstudy then that knowledge must be viewedas a produced means of production andscience as an ongoing social activity in acontinuing process of transformation. Butthe aim of science is the production of theknowledge of the mechanisms of the pro-duction of phenomena in nature that com-

xxx INTRODUCTION

bine to generate the actual flux of phenom-ena of the world. These mechanisms, whichare the intransitive objects of scientific en-quiry, endure and act quite independently ofmen. The statements that describe their op-erations, which may be termed ‘laws’, arenot statements about experiences (empiricalstatements, properly so called) or statementsabout events. Rather they are statementsabout the ways things act in the world (thatis, about the forms of activity of the thingsof the world) and would act in a world with-out men, where there would be no experi-ences and few, if any, constant conjunctionsof events. (It is to be able to say this interalia that we need to distinguish the domainsof the real, the actual and the empirical.)

⇓ The next paragraph gives a sub-sidiary aim: we want to explain why em-pirical realism is so plausible and show

that it is a special case.

17:2/o Although the primary aim of thisbook is constructive, it is an important sub-sidiary aim to situate the conditions of theplausibility of empirical realism and to showit as depending upon what is in effect aspecial case. These conditions are briefly:a naturally occurring closure, a mechanis-tic conception of action and the model ofman referred to earlier. The attempt to re-duce knowledge to an individual acquistionin sense-experience and to view the latter asthe neutral ground of knowledge that (lit-erally) defines the world results in the gen-eration of an ontology of atomistic discreteevents, which if they are to be related at all(so making general knowledge possible) mustbe constantly conjoined. (Hence the pre-supposition of a closure.) On this view thecausal connection must be contingent and

xxxi

actual; by contrast I want to argue that itis necessary and real.

Conditions: naturally occuring closure,mechanistic conception of action, and themodel of man as individualistic receptorof sense-experience which is the neutralground of knowledge. Here he derives thenecessity of the implicit empiricist ontol-ogy: if events are to be related at all, theymust be constantly conjoined. What doeshe mean with last sentence in the aboveparagraph?

⇓ The next paragraph goes throughcontent of the book.

18:1 Chapter 1 establishes the necessityfor an ontological distinction between causallaws and patterns of events (see esp. 1.3)and contains a sketch of a critique of em-pirical realism (see esp. 1.6). Chapter 2 de-

velops in detail the conditions required forthe Humean analysis of laws and provides ananalysis of normic statements (see esp. 2.4).Determinism is shown to be an immenselyimplausible thesis; and the central tenets oforthodox philosophy of science—such as theprinciple of instance-confirmation (or falsifi-cation), the Humean theory of causality, thePopper-Hempel theory of explanation, thethesis of the symmetry between explanationand prediction, the criterion of falsifiability,etc.—to be manifestly untenable. Chapter 3sets out to give a rational account of the pro-cess of scientific discovery; in which both na-ture and our knowledge of nature are seen asstratified, as well as differentiated (see esp.3.3). A theory of natural necessity is devel-oped which it is claimed is capable of resolv-ing inter alia the problems of induction andof subjunctive conditionals and Goodman’s

xxxii INTRODUCTION

and Hempel’s paradoxes (see 3.6). Chapter4 rounds off the argument and summarisessome of the main themes of this study.

⇓ After this brief table-of-contents-likesummary a summary is given of the mainideas, and a brief sketch how they are pre-sented in the book. Starts with generalremarks about concepts of tendencies andpowers.

18:2 Moving towards a conception of sci-ence as concerned essentially with possi-bilities, and only derivatively with actuali-ties, much attention is given to the analy-sis of such concepts as tendencies and pow-ers. Roughly the theory advanced here isthat statements of law are tendency state-ments. Tendencies may be possessed unexer-cised, exercised unrealised, and realized un-perceived (or undetected) by men; they mayalso be transformed. Although the focus of

this study is natural science, something issaid about the social sciences and about thecharacteristic pattern of explanation in his-tory.

18:3/o If the first half of this work is con-cerned with establishing the necessity for anontological distinction between causal lawsand patterns of events and tracing the im-plications of the distinction between opensystems and closed, that is, of the differ-entiation of our world, the second is con-cerned principally with showing how sciencecan come to have knowledge of natural neces-sity a posteriori. The differentiation of theworld implies its stratification, if it is to bea possible object of knowledge for us. If gen-erative mechanisms and structures are realthen there is a clear criterion for distinguish-ing between a necessary and an accidentalsequence: a sequence Ea.Eb is necessary if

xxxiii

and only if there is a generative mechanismor structure which when stimulated by theevent described by ‘Ea’ produces Eb. If wecan have empirical knowledge of such gener-ative mechanisms or structures then we canhave knowledge of natural necessity a poste-riori. In showing how this is possible a non-Kantian ‘sublation’ of empiricism and ratio-nalism is achieved.

19:1 In the transitive process of sciencethree levels of knowledge may be distin-guished. At the first (or Humean) level wejust have the invariance of an experimentallyproduced result. Given such an invariancescience moves immediately to the construc-tion and testing of possible explanations forit. If there is a correct explanation, locatedin the nature of the thing whose behaviouris described in the putative law or the struc-ture of the system of which the thing is a

part, then we do have a reason independentof its behaviour as to why it behaves the wayit does. Now such a reason may be discov-ered empirically. And if we can deduce thething’s tendency from it then the most strin-gent possible (or Lockean) criterion for ourknowledge of natural necessity is satisfied.For example, we may discover that copperhas a certain atomic or electronic structureand then be able to deduce its dispositionalproperties from a statement of that struc-ture. We may then be said to have knowl-edge of natural necessity a posteriori. At thethird (or Leibnizian) level we may seek toexpress our discovery of the electronic struc-ture of copper in an attempted real defini-tion of the thing. This is not to put an endto enquiry, but a stepping stone to a newprocess of discovery in which we attempt todiscover the mechanisms responsible for elec-

xxxiv INTRODUCTION

tronic structure.

19:2/o In 3.5 the grounds for inductivescepticism are examined and shown to befundamentally mistaken and in 3.6 the prob-lem, which arises from the ontology of atom-istic events (and closed systems), resolved.Dynamic realist principles of substance andcausality are shown to be a condition of theintelligibility of experimental activity andthe stratification of science. Science, it is ar-gued, is concerned with both taxonomic andexplanatory knowledge: with what kinds ofthings there are, as well as how the thingsthere are behave. It attempts to express theformer in real definitions of the natural kindsand the latter in statements of causal laws,i.e. of the tendencies of things. But it is con-cerned with neither in an undiscriminatingway. It is concerned with things only in asmuch as they cast light on reasons; and rea-

sons only in as much as they cast light onthings. A realist theory of the universalsof interest to science complements the real-ist theory of scientifically significant invari-ances, i.e. invariances generated under con-ditions which are artificially produced andcontrolled.

20:1 It is the argument of this book thatif science is to be possible the world mustconsist of enduring and transfactually activemechanisms; society must consist of an en-semble of powers irreducible to but presentonly in the intentional actions of men; andmen must be causal agents capable of act-ing self-consciously on the world. They doso in an endeavour to express to themselvesin thought the diverse and deeper structuresthat account in their complex manifold de-terminations for all the phenomena of ourworld.

Chapter 1

<16> Philosophy andScientific Realism

1.1 Two Sides of‘Knowledge’

Distinguishes between transitive and in-transitive objects of science (example

1

2 CHAPTER 1. PHILOSOPHY AND SCIENTIFIC REALISM

Darwin) and in a very elegant way ar-rives at the transcendental question bygoing through all possibilities:

Starting point is an interesting “centralparadox of sciences”:

21:1/o Any adequate philosophy of sciencemust find a way of grappling with this cen-tral paradox of science: that men in theirsocial activity produce knowledge which is asocial product much like any other, which isno more independent of its production andthe men who produce it than motor cars,armchairs or books, which has its own crafts-men, technicians, publicists, standards andskills and which is no less subject to changethan any other commodity. This is one sideof ‘knowledge’. The other is that knowl-edge is ‘of’ things which are not produced bymen at all: the specific gravity of mercury,the process of electrolysis, the mechanism of

light propagation. None of these ‘objects ofknowledge’ depend upon human activity. Ifmen ceased to exist sound would continue totravel and heavy bodies fall to the earth inexactly the same way, though ex hypothesithere would be no-one to know it. Let us callthese, in an unavoidable technical neologism,the intransitive objects of knowledge.

After introducing the intransitive ob-jects of knowledge, Bhaskar remarks thatthere is also a different type of objects ofknowledge, the “transitive” ones, whichare comparable to the means of produc-tion in the production process. Transitiveobjects of knowledge are prior knowledgeand the knowledge of other things.The transitive objects of knowledge are

Aristotelian material causes.1 They are theraw materials of science—the artificial ob-jects fashioned into items of knowledge by

1.1. TWO SIDES OF ‘KNOWLEDGE’ 3

the science of the day.2 They include theantecedently established facts and theories,paradigms and models, methods and tech-niques of inquiry available to a particular sci-entific school or worker.

Bhaskar uses Darwin as an example il-lustrating the difference between intran-sitive and transitive objects. Darwin’sprocess of developing a scientific theoryis compared with a production process.Darwin produced knowledge albout theselection mechanism but he did not pro-duce the mechanism itself.The material cause, in this sense, of Dar-

win’s theory of natural selection consisted ofthe ingredients out of which he fashioned histheory. Among these were the facts of natu-ral variation, the theory of domestic selectionand Malthus’ theory of population.3 Darwinworked these into a knowledge of a process,

too slow and complex to be perceived, whichhad been going on for millions of years beforehim. But he could not, at least if his theoryis correct, have produced the process he de-scribed, the <17> intransitive object of theknowledge he had produced: the mechanismof natural selection.

Question 8 What is the central paradoxof science and how does Bhaskar proposeto deal with it?

Now Bhaskar goes through a tediouselimination argument which you may notwant to bother with. The next importanthigh point in his argument is his questionin 23:1/o ‘what must the world be like forscience to be possible?’⇓ Intransitive world without science is

possible. I.e., intransitive objects do not


depend on science.

22:1 We can easily imagine a world simi-lar to ours, containing the same intransitiveobjects of scientific knowledge, but withoutany science to produce knowledge of them.In such a world, which has occurred andmay come again, reality would be unspokenfor and yet things would not cease to actand interact in all kinds of ways. In such aworld the causal laws that science has now,as a matter of fact, discovered would pre-sumably still prevail, and the kinds of thingsthat science has identified endure. The tideswould still turn and metals conduct elec-tricity in the way that they do, without aNewton or a Drude to produce our knowl-edge of them. The Wiedemann-Franz lawwould continue to hold although there wouldbe no-one to formulate, experimentally es-tablish or deduce it. Two atoms of hy-

drogen would continue to combine with oneatom of oxygen and in favourable circum-stances osmosis would continue to occur. Inshort, the intransitive objects of knowledgeare in general invariant to our knowledge ofthem: they are the real things and struc-tures, mechanisms and processes, events andpossibilities of the world; and for the mostpart they are quite independent of us. Theyare not unknowable, because as a matter offact quite a bit is known about them. (Re-member they were introduced as objects ofscientific knowledge.) But neither are theyin any way dependent upon our knowledge,let alone perception, of them. They are theintransitive, science-independent, objects ofscientific discovery and investigation.

⇓ But science without transitive ob-jects is not possible:

22:2/o If we can imagine a world of in-


transitive objects without science, we can-not imagine a science without transitive ob-jects, i.e. without scientific or pre-scientificantecedents. That is, we cannot imag-ine the production of knowledge save from,and by means of, knowledge-like materi-als. Knowledge depends upon knowledge-like antecedents. Harvey thought of bloodcirculation in terms of an hydraulic model.Spencer, less successfully perhaps, used anorganic metaphor to express his idea of so-ciety. W. Thomson (Lord Kelvin) declaredin 1884 that it seemed to him that ‘the testof “do we understand a particular topic inphysics [e.g. heat, magnetism]?” is “can wemake a mechanical model of it?”.’4 And asis well known this was the guiding maximof physical research until the gradual dis-integration of the Newtonian world-view inthe first decades of this century. Similarly

economists sought explanations of phenom-ena which would conform to the paradigmof a decision-making unit maximizing anobjective function with given resources un-til marginalism became discredited in the1930’s. No doubt at the back of economists’minds during the period of the paradigm’shegemony was the cosy picture of a house-wife doing her weekly shopping subject toa budget constraint; just as Rutherford dis-armingly confessed in 1934, long after theparadigm was hopelessly out of date, toa predilection for corpuscularian models ofatoms and fundamental particles as ‘littlehard billiard balls, preferably red or black’.5

Von Helmont’s concept of an arche <18>was the intellectual ancestor of the conceptof a bacterium, which furnished the modelfor the concept of a virus. The biochemi-cal structure of genes, which were initially


introduced as the unknown bearers of ac-quired characteristics, has been explored un-der the metaphor of a linguistic code. Inthis way social products, antecedently estab-lished knowledges capable of functioning asthe transitive objects of new knowledges, areused to explore the unknown (but knowable)intransitive structure of the world. Knowl-edge of B is produced by means of knowledgeof A, but both items of knowledge exist onlyin thought.

⇓ And science without intransitive ob-jects? If that is considered impossible,i.e., if one is a realist, then two ques-tions must be asked. This is importantnow, and unfortunately RB does not giveenough explanation how these two ques-tions follow. First question: What mustthe world be like for science to be possi-ble? This is what he calls the philosoph-

ical as opposed to the scientific study ofthe intransitive objects of science. Onlyafterwards can one ask: What must sci-ence be like to give us knowledge of in-transitive objects (of this kind)?

23:1/o If we cannot imagine a sciencewithout transitive objects, can we imaginea science without intransitive ones? If theanswer to this question is ‘no’, then a philo-sophical study of the intransitive objects ofscience becomes possible.

The answer to the transcendental question‘what must the world be like for science tobe possible?’ deserves the name of ontol-ogy. And in showing that the objects of sci-ence are intransitive (in this sense) and ofa certain kind, viz. structures not events, itis my intention to furnish the new philoso-phy of science with an ontology. The par-allel question ‘what must science be like to


give us knowledge of intransitive objects (ofthis kind)?’ is not a petitio principii of theontological question, because the intelligibil-ity of the scientific activities of perceptionand experimentation already entails the in-transitivity of the objects to which, in thecourse of these activities, access is obtained.That is to say, the philosophical position de-veloped in this study does not depend uponan arbitrary definition of science, but ratherupon the intelligibility of certain universallyrecognized, if inadequately analysed, scien-tific activities. In this respect I am takingit to be the function of philosophy to anal-yse concepts which are ‘already given’ but‘as confused’.6

From this he derives four criteria of ad-equacy of philosophy of science.

24:1 Any adequate philosophy of sciencemust be capable of sustaining and reconcil-

ing both aspects of science; that is, of show-ing how science which is a transitive process,dependent upon antecedent knowledge andthe efficient activity of men, has intransitiveobjects which depend upon neither. Thatis, it must be capable of sustaining both (1)the social character of science and (2) theindependence from science of the objects ofscientific thought. More specifically, it mustsatisfy both:

24:2 (1)’ a criterion of the non-spontaneous production of knowledge,viz. the production of knowledge from andby means of knowledge (in the transitivedimension), and

24:3 (2)’ a criterion of structural and es-sential realism, viz. the independent exis-tence and activity of causal structures andthings (in the intransitive dimension).

(Non-spontaneous production of


knowledge means: the world does notreveal itself automatically. Science iswork.)⇓ The last paragraph gives a succinct

formulation of results.24:4 For science, I will argue, is a social

activity whose aim is the production of theknowledge of the kinds and ways of acting ofindependently existing and active things.

1.2 <19> Three Tradi-tions in the Philoso-phy of Science

⇓ (1) The first tradition is classical em-piricism: ultimate objects of knowledgeare atomistic events, natural necessity= conjunction of events. World is flat,

knowledge is map of world. Science is“automatic,” a behavioral response toconjunction of events, epiphenomenon ofnature.

24:5/o Viewed historically, three broadpositions in the philosophy of science may bedistinguished. According to the first, that ofclassical empiricism, represented by Humeand his heirs, the ultimate objects of knowl-edge are atomistic events. Such events con-stitute given facts and their conjunctions ex-haust the objective content of our idea ofnatural necessity. Knowledge and the worldmay be viewed as surfaces whose points arein isomorphic correspondence or, in the caseof phenomenalism, actually fused. On thisconception, science is conceived as a kindof automatic or behavioural response to thestimulus of given facts and their conjunc-tions. Even if, as in logical empiricism, such

1.2. THREE TRADITIONS IN THE PHILOSOPHY OF SCIENCE 9

a behaviourism is rejected as an account ofthe genesis of scientific knowledge, its validcontent can still in principle be reduced tosuch facts and their conjunctions. Thus sci-ence becomes a kind of epiphenomenon ofnature.

⇓ (2) Kant’s transcendental idealism:objects of knowledge are socially gener-ated “ideals,” i.e., human constructs im-posed on the phenomena. This discus-sion here is much too brief! Presumably,that part of nature which is not a prod-uct of mankind (things in themselves) isnot knowable.

Question 9 Both Bhaskar and Marx arecritical of idealism, but Bhaskar is alsocritical of empiricism. Define empiricismand explain why Bhaskar is critical of it.What is Marx’s stance on empiricism?

25:1 The second position received its clas-sical though static formulation in Kant’stranscendental idealism, but it is susceptibleof updated and dynamized variations. Ac-cording to it, the objects of scientific knowl-edge are models, ideals of natural order etc.Such objects are artificial constructs andthough they may be independent of particu-lar men, they are not independent of men orhuman activity in general. On this concep-tion, a constant conjunction of events is in-sufficient, though it is still necessary, for theattribution of natural necessity. Knowledgeis seen as a structure rather than a surface.But the natural world becomes a construc-tion of the human mind or, in its modernversions, of the scientific community.

⇓ (3) Transcendental Realism, which isnowadays usually called Critical Realism


(CR), uses correct elements of both:

25:2 The third position, which is advancedhere, may be characterized as transcendentalrealism. It regards the objects of knowledgeas the structures and mechanisms that gen-erate phenomena; and the knowledge as pro-duced in the social activity of science. Theseobjects are neither phenomena (empiricism)nor human constructs imposed upon thephenomena (idealism), but real structureswhich endure and operate independently ofour knowledge, our experience and the condi-tions which allow us access to them. Againstempiricism, the objects of knowledge arestructures, not events; against idealism, theyare intransitive (in the sense defined). Onthis conception, a constant conjunction ofevents is no more a necessary than it isa sufficient condition for the assumption ofthe operation of a causal law. According

to this view, both knowledge and the worldare structured, both are differentiated andchanging; the latter exists independently ofthe former (though not our knowledge of thisfact); and experiences and the things andcausal laws to which it affords us access arenormally out of phase with one another. Onthis view, science is not an epiphenomenonof nature, nor is nature a product of man.

⇓ This typology is not exhaustive,it leaves out rationalism and absoluteidealism, which are in-between states.Bhaskar is describing an extreme posi-tion here, but he will argue that this isthe only position “that can do justice toscience.”

26:1 A word of caution is necessary here.In outlining these positions, I am not offer-ing them as a complete typology, but only asone which will be of some <20> significance


in illuminating current issues in the philos-ophy of science. Thus I am not concernedwith rationalism as such, or absolute ideal-ism. Moreover, few, if any, modern philoso-phers of science could be unambiguously lo-cated under one of these banners. Nagel forexample stands somewhere along the contin-uum between Humean empiricism and neo-Kantianism; Sellars nearer the position char-acterized here as transcendental realist; andso on. One could say of such philosophersthat they combine, and when successful inan original way synthesize, aspects of thosephilosophical limits whose study we are un-dertaking. It is my intention here, in workingout the implications of a full and consistentrealism, to describe such a limit; in ratherthe way Hume did. As an intellectual ex-ercise alone this would be rewarding, but Ibelieve, and hope to show, that it is also the

only postion that can do justice to science.

⇓ Transcendental realism distinguishedfrom empirical realism, realism with re-spect to causal laws etc.

26:2 Transcendental realism must be dis-tinguished from, and is in direct oppositionto, empirical realism. This is a doctrine towhich both classical empiricism and tran-scendental idealism subscribe. My reasonsfor rejecting it will be elaborated in a mo-ment. ‘Realism’ is normally associated byphilosophers with positions in the theory ofperception or the theory of universals. In theformer case the real entity concerned is someparticular object of perception; in the lattercase some general feature or property of theworld. The ‘real entities’ the transcendentalrealist is concerned with are the objects ofscientific discovery and investigation, such ascausal laws. Realism about such entities will


be seen to entail particular realist positionsin the theory of perception and universals,but not to be reducible to them.

Question 10 What is the criterion forbeing real in transcendental realism?How does it differ from the criterion forbeing real in empirical realism?

Adequacy criteria applied to the abovethree positions:

26:3 Only transcendental realism, I willargue, can sustain the idea of a law-governedworld independent of man; and it is this con-cept, I will argue, that is necessary to under-stand science.

⇓ Classical empiricism:26:4/o Classical empiricism can sustain

neither transitive nor intransitive dimen-sions; so that it fails both the criteria of ad-equacy (1)’ and (2)’ advanced on page 24

above. Moreover in its most consistent formsit involves both solipsism and phenomenal-ism; so that neither (1) nor (2) can be up-held. In particular not even the idea of theindependence of the event from the experi-ence that grounds it, i.e. the intransitivityof events, can be sustained; and, in the lastinstance, events must be analysed as sensa-tions or in terms of what is epistemologicallyequivalent, viz. human operations.

Question 11 Bhaskar claims that Clas-sical empiricism can sustain neither tran-sitive nor intransitive dimensions. Whynot? What is his argument?

⇓ Transcendental idealism:27:1 Transcendental idealism attempts to

uphold the objectivity (intersubjectivity) offacts, i.e. (1). And, if given a dynamic gloss,


it can allow a transitive dimensions and sat-isfy criterion (1)’; so that, in this respect, itis an improvement on empiricism. Accordingto such a dynamized transcendental idealismknowledge is given structure by a sequenceof models, rather than a fixed set of a pri-ori rules. However in neither its static norits dynamic form can it sustain the intransi-tive dimension. For in both cases the objectsof which knowledge is obtained do not existindependently of human activity in general.And if there are things which do (things-in-themselves), no scientific knowledge of themcan be obtained.

⇓ Transcendental realism:

27:2 <21> Both transcendental realismand transcendental idealism reject the em-piricist account of science, according towhich its valid content is exhausted by atom-istic facts and their conjunctions. Both agree

that there could be no knowledge withoutthe social activity of science. They disagreeover whether in this case there would be nonature also. Transcendental realism arguesthat it is necessary to assume for the intel-ligibility of science that the order discoveredin nature exists independently of men, i.e.of human activity in general. Transcenden-tal idealism maintains that this order is ac-tually imposed by men in their cognitive ac-tivity. Their differences should thus be clear.According to transcendental realism, if therewere no science there would still be a nature,and it is this nature which is investigatedby science. Whatever is discovered in na-ture must be expressed in thought, but thestructures and constitutions and causal lawsdiscovered in nature do not depend uponthought. Moreover, the transcendental real-ist argues, this is not just a dogmatic meta-


physical belief; but rather a philosophical po-sition presupposed by key aspects of the so-cial activity of science, whose intelligibilitythe transcendental idealist cannot thus, any-more than the empiricist, sustain.

Now he starts arguing that the mostfundamental difference between thesethree approaches is their underlying on-tology.⇓ Three errors implied in the concept

of the “empirical world.”27:3/o Neither classical empiricism nor

transcendental idealism can sustain the ideaof the independent existence and action ofthe causal structures and things investigatedand discovered by science. It is in theirshared ontology that the source of this com-mon incapacity lies. For although transcen-dental idealism rejects the empiricist accountof science, it tacitly takes over the empiri-

cist account of being. This ontological legacyis expressed most succinctly in its commit-ment to empirical realism, and thus to theconcept of the ‘empirical world ’. For thetranscendental realist this concept embodiesa sequence of related philosophical mistakes.The first consists in the use of the categoryof experience to define the world. This in-volves giving what is in effect a particularepistemological concept a general ontologi-cal function. The second consists in the viewthat its being experienced or experienciableis an essential property of the world; whereasit is more correctly conceived as an acciden-tal property of some things, albeit one whichcan, in special circumstances, be of great sig-nificance for science. The third thus consistsin the neglect of the (socially produced) cir-cumstances under which experience is in factepistemically significant in science.


Question 12 Which three interrelatedmistakes are buried in the concept of the“empirical world.”?

⇓ For the transcendental idealist, thedifference between necessary and acci-dental sequences is not real but merelydescribes how humans understand theworld. Therefore it cannot explain howtheories are constructed and tested.

28:1 If the bounds of the real and the em-pirical are co-extensive then of course any‘surplus-element’ which the transcendentalidealist finds in the analysis of law-like state-ments cannot reflect a real difference be-tween necessary and accidental sequences ofevents. It merely reflects a difference inmen’s attitude to them. Saying that lighttravels in straight lines ceases then to expressa proposition about the world; it expresses

instead a proposition about the way men un-derstand it. Structure becomes a function ofhuman needs; it is denied a place in the worldof things.

But if one says that science can onlyhelp humans to sort out our experiences,that natural laws do not exist in realitybut only in our brains, then one cannotunderstand the actual process by whichscience is gained (experiments):But just because of this, I shall argue, the

transcendental idealist cannot adequatelydescribe the principles <22> according towhich our theories are constructed and em-pirically tested; so that the rationality ofthe transitive process of science, in whichour knowledge of the world is continually ex-tended and corrected, cannot be sustained.

⇓ From these criticisms one can seethat science cannot be ontologically neu-


tral. Some say that Bhaskar “has putontology back into philosophy.”

28:2/o To say that the weaknesses of boththe empiricist and idealist traditions lie intheir commitment to empirical realism is ofcourse to commit oneself to the impossibilityof ontological neutrality in an account of sci-ence; and thus to the impossibility of avoid-ing ontological questions in the philosophy ofscience. The sense in which every account ofscience presupposes an ontology is the sensein which it presupposes a schematic answerto the question of what the world must belike for science to be possible. Thus supposea philosopher holds, as both empiricists andtranscendental idealists do, that a constantconjunction of events apprehended in sense-experience is at least a necessary conditionfor the ascription of a causal law and that itis an essential part of the job of science to

discover them. Such a philosopher is thencommitted to the belief that, given that sci-ence occurs, there are such conjunctions. AsMill put it, that ‘there are such things innature as parallel cases; that what happensonce will, under a sufficient degree of simi-larity of circumstance, happen again’.7

One of Bhaskar’s central arguments ishowever that such constant conjunctionsare extremely rare; they occur in astron-omy, but they rarely occur spontaneouslyanywhere else. It is still important to dis-cover natural laws, not because they pro-duce constant conjunctions, but becausethey continue to be active even in envi-ronments that are not conducive to con-stant conjunctions.

⇓ Distinction between philosophicaland scientific ontology. Gives examples


of second-order conclusions: world mustbe stratified and differentiated, becausescience would not be possible otherwise.

29:1/o There are two important points toregister about such ontological beliefs andcommitments.

The first point takes up the rest of thisparagraph; the second is the next para-graph 30:1.The first is that they should only be inter-

preted hypothetically, viz. as entailing whatmust be the case for science to be possible;on which interpretation it is a contingent factthat the world is such that science can occur.It is only in this relative or conditional sensethat an account of science presupposes anontology. The status of propositions in on-tology may thus be described by the follow-ing formula: It is not necessary that scienceoccurs. But given that it does, it is necessary

that the world is a certain way. It is contin-gent that the world is such that science ispossible. And, given that it is possible, itis contingent upon the satisfaction of certainsocial conditions that science in fact occurs.But given that science does or could occur,the world must be a certain way.

Now Bhaskar brings some examples ofsuch second-order arguments:

Thus, the transcendental realist asserts,that the world is structured and differenti-ated can be established by philosophical ar-gument; though the particular structures itcontains and the ways in which it is differen-tiated are matters for substantive scientificinvestigation. The necessity for categoricaldistinctions between structures and eventsand between open systems and closed are in-dices of the stratification and differentiationof the world, i.e. of the transcendental realist


philosophical ontology.

Question 13 What does Bhaskar meanwhen he says that the world is “stratified”or “structured and differentiated”? Howcan this be shown by a philosophical ar-gument?

These distinctions are presupposed, it willbe shown, by the intelligibility of experimen-tal activity. Whenever there is any danger ofconfusion between an ‘ontology’ in the senseof the kind of world presupposed by a philo-sophical account of science and in the senseof the particular entities and processes pos-tulated by some substantive scientific theoryI shall explicitly distinguish between a philo-sophical and a scientific ontology.

The flip side of this definition of philo-sophical ontology is: the ontology one

comes up with depends on one’s conceptof science. But this does not mean thatthe ontological statements one comes upwith are in the last analysis statementsabout science. They are not. Themistaken belief that everything one saysabout ontology is really something aboutscience is called the “epistemic fallacy.”

30:1 The second point to stress is thatpropositions in ontology cannot be <23> es-tablished independently of an account of sci-ence. On the contrary, they can only be es-tablished by reference to such an account, orat least to an account of certain scientific ac-tivities. However, it will be contended thatthis essential order of analysis, viz. science→ being, reverses the real nature of depen-dency (or, we could say, the real burden ofcontingency). For it is not the fact that sci-ence occurs that gives the world a structure

1.3. THE TRANSCENDENTAL ANALYSIS OF EXPERIENCES 19

such that it can be known by men. Rather,it is the fact that the world has such a struc-ture that makes science, whether or not itactually occurs, possible. That is to say, itis not the character of science that imposesa determinate pattern or order on the world;but the order of the world that, under cer-tain determinate conditions, makes possiblethe cluster of activities we call ‘science’. Itdoes not follow from the fact that the natureof the world can only be known from (a studyof) science, that its nature is determined by(the structure of) science. Propositions inontology, i.e. about being, can only be es-tablished by reference to science. But thisdoes not mean that they are disguised, veiledor otherwise elliptical propositions about sci-ence. What I shall characterize in a momentas the ‘epistemic fallacy ’ consists in assum-ing that, or arguing as if, they are.

Question 14 What is the definition ofontology? of epistemology?

Question 15 Explain the difference be-tween philosophical and scientific ontol-ogy.

1.3 The TranscendentalAnalysis of Experi-ences

⇓ Goal is to show that empiricism failsin its most favored case: it cannot evendo justice to the role of experience in sci-ence. This is called an Achilles-heel cri-tique: criticize something where it seemsstrongest.

30:2/o The empiricist ontology is consti-


tuted by the category of experience. Whattranscendental arguments can be producedto show its inadequacy to science; and, onthe other hand, to demonstrate the intran-sitivity and structured character of the ob-jects of scientific knowledge? Now the oc-currence of experience in science would beagreed upon by all three combatants. More-over, it is generally assumed that, what-ever its other inadequacies, empiricism canat least do justice to the role of experiencein science. Now I want to argue that theintelligibility of experience in science itselfpresupposes the intransitive and structuredcharacter of the objects to which, in scientificexperience, ‘access’ is obtained. This estab-lishes the inadequacy, in its most favouredcase, of the empiricist ontology. Further Iwant to argue that, in virtue of their sharedontological commitment, neither empiricism

nor transcendental idealism can reveal thetrue significance of experience in science.

Question 16 What is empiricism?

Question 17 What is the implicit ontol-ogy of empiricism? Which transcendentalarguments suggest that the implicit ontol-ogy of empiricism cannot be right?

Question 18 Bhaskar’s Realist Theoryof Science gives an Achilles-heel critiqueof empiricism. Explain.

⇓ One will not get scientifically signifi-cant experiences if one just watches; onehas to set up experiments first:

31:1 Scientifically significant experiencenormally depends upon experimental activ-ity as well as sense-perception; that is, upon


the role of men as causal agents as well asperceivers. I will consider the two indepen-dently.

1.3.A The Analysis of Per-ception

31:2/o The intelligibility of sense-perceptionpresupposes the intransitivity of the objectperceived.

⇑ One can only make sense of percep-tion if the perceived objects are intransi-tive. This first sentence summarizes thewhole paragraph.For it is in the independent occurrence or ex-istence of such objects that the meaning of‘perception’, and the epistemic significanceof perception, lies. Among such objects areevents, which must thus be categorically in-dependent of experiences. Many arguments

have been and could be <24> deployed todemonstrate this, which there is no spacehere to rehearse. For our purposes, it is suf-ficient merely to note that both the possi-bility of scientific change (or criticism) andthe necessity for a scientific training presup-pose the intransitivity of some real objects;which, for the empirical realist at least, canonly be objects of perception.

Bhaskar says that the arguments areknown. One is change or criticism: onecannot revise or criticize one’s perceptionunless it refers to something.

Question 19 Explain why scientificchange or criticism would not be pos-sible, and scientific training would notbe necessary, if there would not beintransitive objects.


If changing experience of objects is to bepossible, objects must have a distinct beingin space and time from the experiences ofwhich they are the objects. For Kepler to seethe rim of the earth drop away, while TychoBrahe watches the sun rise, we must supposethat there is something that they both see(in different ways).8 Similarly when modernsailors refer to what ancient mariners called asea-serpent as a school of porpoises, we mustsuppose that there is something which theyare describing in different ways.9 The intel-ligibility of scientific change (and criticism)and scientific education thus presupposes theontological independence of the objects ofexperience from the objects of which theyare the experiences.

I think this is the concept of “referen-tial detachment” which Bhaskar uses inhis later works.

At the end, Bhaskar says, he does notthink that events and momentary statesare the primary objects of perception,which are probably processes and things,but Bhaskar does not want to argue thishere because it anticipates results still tobe derived.Events and momentary states do not of

course exhaust the objects of perception. In-deed, I do not think they are even the pri-mary objects of perception, which are prob-ably processes and things, from which eventsand states are then ‘reconstructed’.10 How-ever I do not wish to argue the point here—as it depends upon a prior resolution of theproblems of causality and induction, uponwhich their status as objects of experiencemust, at least for the empiricist, depend.11

32:1 Events then are categorically inde-pendent of experiences. There could be a


world of events without experiences. Suchevents would constitute actualities unper-ceived and, in the absence of men, unper-ceivable. There is no reason why, given thepossibility of a world without perceptions,which is presupposed by the intelligibilityof actual scientific perceptions, there shouldnot be events in a world containing percep-tions which are unperceived and, given ourcurrent or permanent capacities, unperceiv-able.

Interesting derivation of the existenceof unperceived events in our world (i.e.,the empirical is a subset of the actual).From intransitivity of the objects of per-ception follows: there would also bethings if people did not exist, i.e., if ev-erything is unperceived. From there onemust conclude that also in the presenceof people things exist which are unper-

ceived.

Therefore an ontology which definesthings that are real by their being per-ceivable is wrong.And of such events theoretical knowledge

may or may not be possessed, and may ormay not be achievable. Clearly if at someparticular time I have no knowledge of anunperceived or unperceivable event, I cannotsay that such an event occurred (as a pu-tative piece of substantive knowledge). Butthat in itself is no reason for saying that suchan occurrence is impossible or that its sup-position is meaningless (as a piece of philos-ophy). To do so would be to argue quiteillicitly from the current state of knowledgeto a philosophical conception of the world.Indeed, we know from the history of sciencethat at any moment of time there are types ofevents never imagined, of which theoretical,


and sometimes empirical, knowledge is even-tually achieved. For in the transitive pro-cess of science the possibilities of perception,and of theoretical knowledge, are continu-ally being extended. Thus unless it is dog-matically postulated that our present knowl-edge is complete or these possibilities ex-hausted, there are good grounds for holdingthat the class of unknowable events is non-empty, and unperceivable ones non-emptier;and no grounds for supposing that this willever not be so.

Another argument in favor of the ex-istence of unperceived actualities is alsothat the range of perceived things con-tinually expands, and one should not as-sume that this process of expansion hasreached its limits.

He also talkes about the unperceivableand the unknowable, about which it nev-

ertheless it is possible to make meaningfulstatements, even if one can never knowwhether they are true. How does this re-late to the main thrust of the argument?

⇓ Later he will show that the domainof actualities consists not only of eventsbut also of things.

32:2 <25> Later, I will show how the do-main of actualities, whose categorical inde-pendence from experiences is presupposed bythe intelligibility of sense-perception, may beextended to include things as well as events.

Where and how does he show that, anddoes he imply here that empiricism onlylimit itself to events?


1.3.B Analysis of Experi-mental Activity

⇓ In an experiment we are a causal agentof the sequence of events, but not of thecausal law which the sequence of eventsenables us to identify.

33:1 The intelligibility of experimental ac-tivity presupposes not just the intransitivitybut the structured character of the objectsinvestigated under experimental conditions.Let me once again focus on the empiricist’sfavourite case, viz. causal laws, leaving asidefor the moment such other objects of inves-tigation as structures and atomic constitu-tions. A causal law is analysed in empiri-cist ontology as a constant conjunction ofevents perceived (or perceptions). Now anexperiment is necessary precisely to the ex-tent that the pattern of events forthcoming

under experimental conditions would not beforthcoming without it. Thus in an experi-ment we are a causal agent of the sequenceof events, but not of the causal law whichthe sequence of events, because it has beenproduced under experimental conditions, en-ables us to identify.

⇓ Two implications of this: (1) theremust be a difference between causal lawsand sequences of events, and (2) experi-ments only make sense if the causal lawsprevail also outside the context underwhich they generate sequences of events.

33:2 Two consequences flow from this.First, the real basis of causal laws cannot besequences of events; there must be an onto-logical distinction between them. Secondly,experimental activity can only be given asatisfactory rationale if the causal law it en-ables us to identify is held to prevail out-


side the contexts under which the sequenceof events is generated. In short, the intelli-gibility of experimental activity presupposesthat a constant conjunction is no more a nec-essary than a sufficient condition for a causallaw. And it implies that causal laws endureand continue to operate in their normal wayunder conditions, which may be character-ized as ‘open’, where no constant conjunc-tion or regular sequence of events is forth-coming. It is worth noting that in general,outside astronomy, closed systems, viz. sys-tems in which constant conjunctions occur,must be experimentally established.

Question 20 How can experimental ac-tivity be used to prove that constant con-junctions of events are neither necessarynor sufficient for causal laws?

⇓ This distinction between causal laws

and patterns of events conforms with ourintuitions. Regular occurences are oftendisturbed. Everyone knows that it is easyto muck up an experiment, but nobodythinks this changes the causal laws.

33:3/o Both Anscombe and von Wrighthave recently made the point that our activeinterference in nature is normally a condi-tion of empirical regularities.12 But neitherhave seen that it follows from this that theremust be an ontological distinction betweenthe empirical regularity we produce and thecausal law it enables us to identify. Althoughit has yet to be given an adequate philosoph-ical rationale, the distinction between causallaws and patterns of events is consistent withour intuitions. Thus supposing a nuclear ex-plosion were to destroy our planet no-onewould hold that it violated, rather than ex-emplified, Newton’s laws of motion;13 just


as if something were to affect Mercury’s per-ihelion it would not be regarded as falsifyingEinstein’s theory of relativity. Similarly itlies within the power of every reasonably in-telligent schoolboy or moderately clumsy re-search worker to upset the results of even thebest designed experiment,14 but we do notthereby suppose they have the power to over-turn the laws of nature. I can <26> quiteeasily affect any sequence of events designedto test say Coulomb’s or Gay-Lussac’s law;but I have no more power over the relation-ships the laws describe than the men whodiscovered them had. In short, laws cannotbe the regularities that constitute their em-pirical grounds.

⇓ To sum up: causal laws independentof constant conjunctions of events, thisallows us to explain phenomena in open

systems.34:1 Thus the intelligibility of experimen-

tal activity presupposes the categorical inde-pendence of the causal laws discovered fromthe patterns of events produced. For, to re-peat, in an experiment we produce a patternof events to identify a causal law, but we donot produce the causal law identified. Oncethe categorical independence of causal lawsand patterns of events is established, thenwe may readily allow that laws continue tooperate in open systems, where no constantconjunctions of events prevail. And the ra-tional explanation of phenomena occurringin such systems becomes possible.

⇓ Now he brings in the other pointalready mentioned in 33:2, that sponta-neously occurring patterns of events arerare. Regular patterns of events are usu-ally generated by human activity. There-


fore both experiences and regular pat-terns of events presuppose humans, i.e.,empiricist ontology is anthropocentric.

34:2 In a world without men there wouldbe no experiences and few, if any, constantconjunctions of events, i.e. had they been ex-perienced Humean ‘causal laws’. For bothexperiences and invariances (constant con-junctions of events) depend, in general, uponhuman activity. But causal laws do not.Thus in a world without men the causal lawsthat science has now as a matter of fact dis-covered would continue to prevail, thoughthere would be few sequences of events andno experiences with which they were in cor-respondence. Thus, we can begin to see howthe empiricist ontology in fact depends upona concealed anthropocentricity.

Question 21 Explain the word“antrhropocentric.” In which two

ways is empiricism anthropocentric?

⇓ If one identifies causal laws with em-pirical regularities one makes two cate-gory mistakes: (a) One identifies eventswith experiences and (b) constant con-junctions of events with causal laws(which presupposes the ubiquity of closedsystems). This double identification pre-vents the empirical realist from seeing theconditions under which experiences aresignificant for science, (a) the perceivermust be theoretically informed, and (b)the system in which the events occurmust be closed.

34:3/o The concept of causal laws be-ing or depending upon empirical regulari-ties involves thus a double identification: ofevents and experiences; and of constant con-junctions (or regular sequences) of events


and causal laws. This double identificationinvolves two category mistakes, expressedmost succinctly in the concepts of the empir-ical world and the actuality of causal laws.The latter presupposes the ubiquity of closedsystems. Both concepts, I shall argue, areprofoundly mistaken and have no place inany philosophy of science.

The empirical realist cannot explainwhy scientific experiments take so mucheffort.This double identification prevents the em-

pirical realist from examining the importantquestion of the conditions under which expe-rience is in fact significant in science. In gen-eral this requires both that the perceiver betheoretically informed15 and that the systemin which the events occur be closed.16 Onlyunder such conditions can the experimentalscientist come to have access to those under-

lying causal structures which are the objectsof his theory. And not until the categori-cal independence of causal laws, patterns ofevents and experiences has been philosophi-cally established and the possibility of theirdisjuncture thereby posed can we appreciatethe enormous effort—in experimental designand scientific training—that is required tomake experience epistemically significant inscience.

⇓ The next paragraph summarizesthe transcendental implications of exper-imental activity: the intransitive char-acter of the objects of scientific knowl-edge; (hence by extension?) possibility ofa nonhuman world. (Secondly?,) the pos-sibility of open systems.

35:1 The intelligibility of experimental ac-tivity presupposes then the intransitive andstructured character of the objects of scien-


tific knowledge, at least in so far as theseare causal laws. And this presupposes inturn the possibility of a non-human world,i.e. causal laws without invariances and expe-riences, and in particular of a non-empiricalworld, i.e. causal laws and events without<27> experiences; and the possibility ofopen systems, i.e. causal laws out of phasewith patterns of events and experiences, andmore generally of epistemically insignificantexperiences, i.e. experiences out of phasewith events and/or causal laws.

⇓ Causal laws are therefore not empir-ical statements (they are not even state-ments about events), neither are theysynthetic a priori statements. Bhaskardoes not say what they are, he merelyrules out the only two possibilities theempirical realist considers. A more posi-

tive definition will be given in 52:1/o.

35:2/o In saying that the objects of sci-entific discovery and investigation are ‘in-transitive’ I mean to indicate therefore thatthey exist independently of all human activ-ity; and in saying that they are ‘structured’that they are distinct from the patterns ofevents that occur. The causal laws of natureare not empirical statements, i.e. statementsabout experiences; nor are they statementsabout events; nor are they synthetic a prioristatements. For the moment I merely stylethem negatively as ‘structured intransitive’,postponing a positive analysis of them untilSection 5.

1.4. STATUS OF ONTOLOGY AND ITS DISSOLUTION IN CLASSICAL PHILOSOPHY31

1.4 Status of Ontologyand its Dissolutionin Classical Philoso-phy

⇓ Our argument that experimental activ-ity only then makes sense if the worldis structured and if causal laws are notconstant conjunctions of events shows thepossibility of a philosophical ontology,which talks about the same world as thescientific ontology, and which has beendiscredited because its early attemptswere “treating a mysterious underlyingphysical realm.”

36:1 This analysis of experimentalepisodes enables us to isolate a series ofmetaphysical, epistemological and method-

ological mistakes within the tradition ofempirical realism. For if the intelligibilityof experimental activity entails that theobjects of scientific understanding areintransitive and structured then we can es-tablish at one stroke: (i) that a philosophicalontology is possible; (ii) some propositionsin it (causal laws are distinct from patternsof events, and events from experiences); and(iii) the possibility of a philosophy whichis consistent with (and has some relevancefor), i.e. which is itself ‘in phase with’,the realist practice of science. Ontology,it should be stressed, does not have as itssubject matter a world apart from thatinvestigated by science. Rather, its subjectmatter just is that world, considered fromthe point of view of what can be establishedabout it by philosophical argument. Theidea of ontology as treating of a mysterious


underlying physical realm, which owes alot to Locke and some of his rationalistcontemporaries (particularly Leibniz), hasdone much to discredit it; and to preventmetaphysics from becoming what it ought tobe, viz. a conceptual science. Philosophicalontology asks what the world must be likefor science to be possible; and its premisesare generally recognized scientific activities.Its method is transcendental; its premisescience; its conclusion the object of ourpresent investigation.

⇓ Now the epistemic fallacy is intro-duced (apart from the earlier brief men-tions in 16:1 and 30:1).

36:2/o The metaphysical mistake the ar-gument of the previous section allows usto pinpoint may be called the ‘epistemicfallacy’. This consists in the view thatstatements about being can be reduced to

or analysed in terms of statements aboutknowledge; i.e. that ontological questionscan always be transposed into epistemologi-cal terms. The idea that being can always beanalysed in terms of our knowledge of being,that it is sufficient for philosophy to ‘treatonly of the network, and not what the net-work describes’,17 results in the systematicdissolution of the idea of a world (which Ishall here <28> metaphorically characterizeas an ontological realm) independent of butinvestigated by science. And it is manifest inthe prohibition of any transcendent entities.

How does the epistemic fallacy result inthe systematic dissolution of the intran-sitive dimension?

Here is an attempt of an answer:Bhaskar’s transitive-intransitive distinc-tion not only entails that there is a worldout there independent of science (and


also independent of the existence of hu-mans), but moreover that the objects ofscientific knowledge are located in thisworld. Ther epistemic fallacy says thatscience is not about the outside world,but that in the last analysis science isonly about what people do in their heads.Someone who commits the epistemic fal-lacy will, if he or she is consistent, be ledto hold that a world independent of sci-ence does not exist.

Furthermore, Bhaskar says that theepistemic fallacy manifests itself in theprohibition of any transcendent entities.What is a transcendent entity? Is it anentity of whose existence we can only in-fer by transcendental argument?It might be usefully compared with the nat-uralistic fallacy in moral philosophy. For just

as the naturalistic fallacy prevents us fromsaying what is good about e.g. maximizingutility in society, so the epistemic one pre-vents us from saying what is epistemicallysignificant about e.g. experience in science.

What is the naturalistic fallacy inmoral philosophy? How does the natu-ralistic fallacy prevents us from asking:what is good about maximizing utility insociety?

Question 22 How does the epistemicfallacy prevent us from asking which ex-periences are relevant for science andwhich aren’t?

If you consider the epistemic fallacydifficult to understand, don’t worry, it isindeed a difficult subject. Tracing the ef-fects of the epistemic fallacy is one of the


principal subjects of this book. [Col94,pp. 76–85] has an extra section devotedto it.To show that it is a fallacy and to trace

its effects are two of the principal objectivesof this study. In showing that the intelli-gibility of experimental activity entails thatthe objects of scientific knowledge, in so faras they are causal laws, are intransitive Ihave already succeeded in the first of theseaims. For this means that a statement of acausal law cannot now be reduced to or anal-ysed in terms of a statement about anyone’sknowledge of it or knowledge in general. Onthe contrary, its assertion now entails that acausal law would operate even if unknown,and even if there were no-one to know it. Sothat knowledge ceases to be, as it were, anessential predicate of things.

How has Bhaskar shown that the epis-

temic fallacy is indeed a fallacy?Attempt at an answer: Bhaskar al-

ready showed that it is a fallacy by theargument that causal laws cannot be re-duced to empirical regularities, i.e., towhat can be known about them.⇓ The next paragraph is very dense,

brings many examples how the traditionof empirical realism commits the epis-temic fallacy:

37:1/o The epistemic fallacy is mostmarked, perhaps, in the concept of the em-pirical world. But it is manifest in the cri-teria of significance and even the problemsassociated with the tradition of empirical re-alism. Kant committed it in arguing that thecategories ‘allow only of empirical employ-ment and have no meaning whatsoever whennot applied to objects of possible experience;that is to the world of sense.’18 (For us on


the other hand if the Kantian categories wereadequate to the objects of scientific thoughtthen they would continue to apply in a worldwithout sense, and have a meaning in rela-tion to that possibility.) Similarly, the logi-cal positivists committed it when arguing, inthe spirit of Hume, that if a proposition wasnot empirically verifiable (or falsifiable) ora tautology, it was meaningless.19 Verifica-tionism indeed may be regarded as a partic-ular form of the epistemic fallacy, in whichthe meaning of a proposition about reality(which cannot be designated ‘empirical’) isconfused with our grounds, which may ormay not be empirical, for holding it. Oncethis doctrine is rejected there is no need toidentify the necessary and the a priori, andthe contingent and the a posteriori; or, toput it another way, one can distinguish be-tween natural and logical necessity, and be-

tween natural and epistemic possibility. Fur-ther there is no need to assume that the or-der of dependence of being must be the sameas the order of dependence of our knowledgeof being. Thus we can allow that experienceis in the last instance epistemically decisive,without supposing that its objects are onto-logically ultimate, in the sense that their ex-istence depends upon nothing else. Indeedif science is regarded as a continuing pro-cess of discovery of ever finer and in an ex-planatory sense more basic causal structures,then it is rational to assume that what is atany moment of time least certain epistemi-cally speaking is most basic from the onto-logical point of view.20 More generally, theepistemic fallacy is manifest in a persistenttendency to read the conditions of a partic-ular concept of knowledge into an implicitconcept of the world. Thus the problem


of induction is a <29> consequence of theatomicity of the events conjoined, which is afunction of the necessity for an epistemicallycertain base.

Question 23 What is the problem of in-duction? How is the epistemic fallacyconnected with the problem of induction?

⇓ The next paragraph gives then thesignificance of the epistemic fallacy inother traditions, namely, for rationalism.Rationalism says, in brief, that the worldticks the same way as our brains tick.

38:1/o Although the epistemic fallacy is ofmost interest to us as it is manifest in the tra-dition of empirical realism, it is worth men-tioning that a philosopher who rejected em-pirical realism might still commit the epis-temic fallacy, i.e. analyse being in terms of

knowledge, if, as in some varieties of Pla-tonism and rationalism, he were to definethe world in terms of the possibility of non-empirical knowledge of it. For the transcen-dental realist it is not a necessary conditionfor the existence of the world that science oc-curs. But it is a necessary condition for theoccurrence of science that the world existsand is of a certain type. Thus the possibilityof our knowing it is not an essential property,and so cannot be a defining characteristic, ofthe world. Rather on a cosmic scale, it is anhistorical accident; though it is only becauseof this accident that we can establish in sci-ence the way the world is, and in philosophythe way it must be for science to be possible.

Apparently Bhaskar argues here as fol-lows: the rationalist doctrine impliesthat everything about the world can beknown. But the possibility of know-


ing the world is not an essential prop-erty of the world, since we humans areonly a contingent outcome and the worldcould also have gone on without humans.Therefore it can also not be a definingcharacteristic of the world.⇓ A strong argument in favor of the

epistemic fallacy is: what can be knownto exist is part of what we can know, andtherefore must be subsumed to epistemol-ogy. Let’s see how Bhaskar refutes this:

39:1 The view that statements about be-ing can be reduced to or analysed in termsof statements about knowledge might be de-fended in the following way: ontology is de-pendent upon epistemology since what wecan know to exist is merely a part of whatwe can know.21 But this defence trades upona tacit conflation of philosophical and scien-tific ontologies. For if ‘what we can know to

exist’ refers to a possible content of a scien-tific theory then that it is merely a part ofwhat we can know is an uninteresting tru-ism. But a philosophical ontology is devel-oped by reflection upon what must be thecase for science to be possible; and this is in-dependent of any actual scientific knowledge.Moreover, it is not true, even from the pointof view of the immanent logic of a science,that what we can know to exist is just a partof what we can know. For a law may existand be known to exist without our knowingthe law. Much scientific research has in factthe same logical character as detection. Ina piece of criminal detection, the detectiveknows that a crime has been committed andsome facts about it but he does not know, orat least cannot yet prove, the identity of thecriminal.

This denies second-order conclusions,


that we know things to exit which we donot know themselves. Like a detectivewho knows that murder has been com-mitted, but does not know who did it.

In all this discussion of the epistemicfallacy I am missing the following argu-ment which seems convincing and simpleto me: the epistemic fallacy cannot betrue because knowledge is not the onlychannel which gives us information aboutthe world. We also learn about the worldby our practical activity in the world.We are able to act in the world withoutknowing why we can act in this way. ForMarx, the most fundamental human ac-tivity is production. Bhaskar is asking:what must the world be like that scienceis possible? I think it would have beenclearer if he had asked: what must the

world be like so that production is possi-ble?⇓ The thesis that to be is not to be the

value of a variable is another implicationof the epistemic fallacy:

39:2 To be is not to be the value of a vari-able;22 though it is plausible (if, I would ar-gue, incorrect) to suppose that things canonly be known as such. For if to be werejust to be the value of a variable we couldnever make sense of the complex processesof identification and measurement by meansof which we can sometimes represent somethings as such. Knowledge follows existence,in logic and in time; and any philosophicalposition which explicitly or implicitly deniesthis has got things upside down.

I guess Marx was also arguing againstthe thesis that “to be is to be the valueof a variable” with his emphasis of the


qualitative as opposed to the quantitativeaspects of things: one should not confusethe quantitative dimension of somethingwith the thing itself.⇓ Epistemic fallacy denies the possi-

bility of philosophical ontology, thereforegives implicit ontology. The false implicitontology gives a methodology irrelevantto science. That is why philosophy cameinto such disrepute!

39:3/o The metaphysical mistake theanalysis of experimental episodes pinpoints,viz. the epistemic fallacy, involves the denialof the possibility of a philosophical ontol-ogy. But if transcendental realism is correct,and ontology cannot in fact be reduced toepistemology, then denying the possibility ofan ontology merely results in the generationof an implicit ontology and an implicit re-alism. In the empirical realist tradition the

epistemic fallacy thus covers or <30> dis-guises an ontology based on the category ofexperience, and a realism based on the pre-sumed characteristics of the objects of expe-riences, viz. atomistic events, and their rela-tions, viz. constant conjunctions. (Such pre-sumptions can, I think, only be explained interms of the needs of a justificationist epis-temology, e.g. for incorrigible foundations ofknowledge.) This in turn leads to the gener-ation of a methodology which is either con-sistent with epistemology but of no relevanceto science; or relevant to science but more orless radically inconsistent with epistemology.So that, in short, philosophy itself is ‘out ofphase’ with science. Let us see how this hap-pens.

⇓ Next paragraph shows the argumentsof a philosophy which is out of phase with


science.

40:1 First, the general line of Hume’s cri-tique of the possibility of any philosophicalontology or account of being, and in par-ticular his denial that we can philosophi-cally establish the independent existence ofthings or operation of natural necessities, isaccepted. Now it is important to see whatHume has in fact done. He has not reallysucceeded in banishing ontology from his ac-count of science. Rather he has replacedthe Lockean ontology of real essences, pow-ers and atomic constitutions with his ownontology of impressions. To say that ev-ery account of science, or every philosophyin as much as it is concerned with ‘science’,presupposes an ontology is to say that thephilosophy of science abhors an ontologicalvacuum. The empiricist fills the vacuum hecreates with his concept of experience. In

this way an implicit ontology, crystallized inthe concept of the empirical world, is gen-erated. And it is this ontology which sub-sequent philosophers of science have uncrit-ically taken over. For whether they haveagreed with Hume’s epistemology or not,they have accepted his critique of ontology,which contains its own implicit ontology, asvalid.

(1) it accepts Hume’s denial that theindependence of things can be establishedphilosophically. Hume did not get rid ofontology but replaced it with his ontologyof impressions. (2) this implicit ontologyis uncritially taken over by others, eventhose who do not agree with Hume’s epis-temology.

⇓ Next paragraph describes the gener-ation of this implicit ontology in more de-


tail: instead of asking whether constantconjunctions of events do in fact occur,philosophy only asks whether our knowl-edge of the world exhausts itself in empir-ical knowledge, or whether there is alsoan a priori or theoretical component.

40:2/o Let us examine the generationof this implicit ontology in greater de-tail. In Hume’s positive analysis of percep-tion and causality experiences constitutingatomistic events and their conjunctions areseen as exhausting our knowledge of nature.Now, adopting a realist meta-perspectivethis means that such events and their con-junctions must occur in nature, if science isto be possible. But from Hume onwards thesole question in the philosophy of science iswhether our knowledge is exhausted by ourknowledge of such events and their conjunc-tions; it is never questioned whether they

in fact occur. That is, philosophy’s concernis with whether our knowledge of the worldcan be reduced to sense-experience as so con-ceived or whether it must include an a priorior theoretical component as well; not withwhether experience can adequately consti-tute the world.

⇓ Humean empiricism not only reducesknowledge to sense experience but alsoreduces the world to empirical phenom-ena. Through this double reduction itcan get an isomorphism of the two. Theresult has been a continuing ontologicaltension.

41:1 But in Humean empiricism twothings are done. First, knowledge is re-duced to that of atomistic events appre-hended in sense-experience. Secondly, theseevents are then identified as the particularsof the world. In this way our knowledge of


reality is literally identified, or at best takento be in isomorphic correspondence, with thereality known by science. From Hume on-wards philosophers have thus allowed, for thesake of avoiding ontology, a particular con-cept of our knowledge of reality, which theymay <31> wish to explicitly reject, to in-form and implicitly define their concept ofthe reality known by science. The resulthas been a continuing ‘ontological tension’induced by the conflict between the ratio-nal intuitions of philosophers about scienceand the constraints imposed upon their de-velopment by their inherited ontology. Thishas led to a nexus of interminably insolubleproblems, such as how we can reason fromone experience to another, and to a displace-ment of these rational intuitions whereby,for example, the locus of necessity is shiftedfrom the objective necessity of the natural

world to the subjective necessity of causally-determined or the inter-subjective necessityof rule-governed minds.

⇑ This is another remark relevant forinduction.⇓ Next paragraph gives examples of

this ontological tension: Hume’s realistmethodology does not fit together withhis own epistemology. Similar incon-gruity in Kant.

41:2/o Now if transcendental realism istrue, and scientists act as if the objects oftheir investigation are intransitive and struc-tured, then any adequate methodology mustbe consistent with the realist practice of sci-ence, and so inconsistent with the epistemol-ogy of empirical realism. It is instructive tolook at Hume here. One finds in the Trea-tise an eminently sensible realist methodol-ogy in almost total dislocation from, and cer-


tainly lacking any foundation in, his radicalepistemology. Thus one might be forgivenfor wondering what has become of his phe-nomenalism and the doctrine of impressionswhen Hume allows that the ‘understandingcorrects the appearances of the senses’.23 Orwhat has happened to the idea of the contin-gency of the causal connection and the prob-lem of induction when he argues that scien-tists, when faced with exceptions to estab-lished generalizations, quite properly searchfor the ‘secret operation of contrary causes’rather than postulate an upset in the uni-formity of nature.24 This is typical. Thereis a similar dislocation between Kant’s Cri-tique of Pure Reason and his MetaphysicalFoundations of Natural Science.

⇓ Consistent application of Hume’sepistemology is Mach’s methodology.“Squeezes out” possibility of any scien-

tific criticism:

42:1/o It might be argued in defence ofHume that he is concerned to show that ourrealist intuitions cannot be justified; thathis point is precisely that there is a dis-location between what can be shown andwhat must be believed (that ‘there is a di-rect and total opposition twixt our reasonand our senses’);25 and that he leaves thelatter intact. But the matter is not so sim-ple as this. Humean empiricism is not neu-tral in its consequences for scientific prac-tice. Taken consistently, it does generate amethodology; not indeed Hume’s (or New-ton’s), but Mach’s. For in the absence of theconcept of an ontological realm, the implicitrealism generated implies that whatever isexperienced in sense-experience is an eventand whatever constant conjunctions are ex-perienced are causal laws. In this way our


current knowledge fills the vacuum left bythe dissolution of the ontological realm; andin so doing it squeezes out, metaphoricallyspeaking, the possibility of any substantivescientific criticism. In the methodology ofHumean empiricism facts, which are socialproducts, usurp the place of the particularsof the world; and their conjunctions, whichare doubly social products (once qua fact,once qua event-conjunction), the place ofcausal laws. The result is the generation of aconservative ideology which serves to ratio-nalize the practice of what <32> Kuhn hascalled ‘normal science’.26 Descriptivist, in-strumentalist and fictionalist interpretationsof theory do not do away with e.g. scien-tific laws, but by reducing their ontologicalimport to a given self-certifying experience,they serve to exempt our current claims toknowledge of them from criticism.

⇓ Realism is therefore less dogmaticthan empiricism. In order to be a falli-bilist about knowledge it is necessary tobe a realist about things:

43:1 It is thus quite incorrect to supposethat realist as opposed to non-realist in-terpretations of scientific theory have con-sequences for science which are in practicemore dogmatic;27 or to suppose that the con-cept of natural necessity is a kind of sur-vival from the bad old days of scientific cer-tainty.28 On the contrary, the converse isthe case. For it is only if the working sci-entist possesses the concept of an ontolog-ical realm, distinct from his current claimsto knowledge of it, that he can philosophi-cally think out the possibility of a rationalcriticism of these claims. To be a fallibilistabout knowledge, it is necessary to be a re-alist about things. Conversely, to be a scep-


tic about things is to be a dogmatist aboutknowledge.

Question 24 Bhaskar says that hisdepth realism is less dogmatic than em-piricism. How does he make this point?

⇓ Also the condition of science between1750 and 1900 is responsible for ontolog-ical tension. This was not a period ofscientific change but of scientific consoli-dation, and philosophy became more andmore concerned with the question howknowledge can be justified, instead ofasking how knowledge can be produced.But these attempts to justify knowledgefailed; the critical realist on the otherhand can criticize scientific activity.

43:2/o Now it is not only the doctrine ofempirical realism, and philosophers’ uncriti-

cal acceptance of it, that accounts for the on-tological tension within philosophy and thedislocation of epistemology from methodol-ogy, of philosophy from science. It must beaccounted for in part by the conditions ofscience, as well as philosophy. For the pe-riod in which Humean ontology became em-bedded in philosophy (1750–1900) was, atleast in physics, a period of scientific con-solidation rather than change. The role ofphilosophy was seen more and more to bethat of showing how our knowledge is jus-tified as distinct from showing how it wasproduced, can be criticized and may cometo be changed. Thus whereas transcendentalrealism asks explicitly what the world mustbe like for science to be possible, classicalphilosophy asked merely what science wouldhave to be like for the knowledge it yielded tobe justified. It was presumed that our knowl-


edge was justified; science was not viewed asa process in motion; and doing away withontology left philosophy without any criti-cal purchase on science. The transcendetalrealist, on the other hand, allows a limitedcritical role for philosophy. For by restor-ing the idea of an ontological realm distinctfrom science, he makes it possible for us tosay that in a particular field, say social psy-chology, science is not being done, althoughas a philosopher he cannot say dogmaticallywhether or not a science of social psychologyis possible.29 (An ontological dimension is inthis way necessary not only to render intelli-gible scientific criticism, but to make possi-ble philosophical critcism of the practice ofa science.) Increasingly then it was the log-ical structure of justificatory argument thatdefined philosophy’s concept of science; andthe philosophy of science itself became a kind

of battleground for internecine warfare be-tween opposed concepts of justified belief.Moreover, when the idea of scientific cer-tainty eventually collapsed, the absence of anontological dimension discouraged anythingother than a purely voluntaristic reaction—in which it was supposed that because ourbeliefs about the world were not causallydetermined by the <33> world then theymust be completely ‘free creations of our ownminds, the result of an almost poetic intu-ition’.30

⇓ (1) There is also a strong anthro-pocentric current, and (2) the dissolutionof ontology made philosophy irrelevant.

44:1/o Behind this state of affairs thereran a strong anthropocentric current in clas-sical and subsequent philosophy,31 whichsought to rephrase questions about the worldas questions about the nature or behaviour

1.5. ONTOLOGY VINDICATED AND THE REAL BASIS OF CAUSAL LAWS47

of men. One aspect of this is the view, whichI have characterized as the epistemic fal-lacy, that ontological questions can alwaysbe rephrased as epistemological ones. Theanthropocentric and epistemic biases of clas-sical philosophy led to the dissolution of theconcept of the ontological realm, which weneed to render intelligible the transitive pro-cess of science. In this way the world, whichought to be viewed as a multi-dimensionalstructure independent of man, came to besquashed into a flat surface whose charac-teristics, such as being constituted by atom-istic facts, were determined by the needs ofa particular concept of knowledge. This ledto a barrage of problems and an impossibleaccount of science. For from now on anystructure, if it was allowed at all, had tobe located in the human mind or the scien-tific community. Thus the world was literally

turned inside out in an attempt to confine itwithin sentience. An inevitable ‘involution’in the philosophy of science occurred. With-out a concept of a reality unknown, but atleast in part knowable, philosophy could notdisplay the creative and critical activity ofscience, and ceased to be of any practicalrelevance for it. This was the price paid forthe dissolution of ontology. A philosophy forscience depends upon its reconstitution.

1.5 Ontology Vindi-cated and the RealBasis of CausalLaws

After arguing at length that there mustbe a difference between causal laws and


constant conjunctions of events, Bhaskarsets out to explain now what causal lawsare.⇓ The first paragraph uses the word

“mysterious” like Marx in the fetish-likecharacter of the commodity. We knowwhat causal laws are not: they are notsequences of events. But what are they?

45:1 In Section 3 I argued that only ifcausal laws are not the patterns of eventsthat enable us to identify them can the in-telligibility of experimental activity be sus-tained. But causal laws are, or have seemedto philosophers to be, pretty mysterious en-tities. What can it mean to say that theyhave a real basis independent of events?The answer to this question will be seento necessitate the development of a non-anthropocentric ontology of structures, gen-erative mechanisms and active things.

Bhaskar will argue that the ontology ofstructures, generative mechanisms, andactive things is really a much less mys-terious explanation of it all.

⇓ Both transcendental idealists andtranscendental realists would agree thatsome generative mechanism is at workto distinguish necessary from accidentalsequences. But the idealist locates thisgenerative mechanism only in the mind,while the realist says that the theory con-struction and change can only be ex-plained if those generative mechanismsare real.

45:2/o The ontological status of causallaws can best be approached by consideringthe divergent responses of transcendental re-alism and idealism to the problem of distin-guishing a necessary from a purely accidental


sequence of events. Both may agree, in theirmodern versions, that without some concep-tion of a generative mechanism at work noattribution of necessity is justified. For thetranscendental idealist, however, this neces-sity is imposed by men on the pattern ofevents; the generative mechanism is an ir-reducible figment of the imagination. Forthe transcendental realist, on the other hand,the generative mechanism may come to beestablished as real in the course of the on-going activity of science. Indeed he will ar-gue that it is only <34> if existential ques-tions can be raised about the objects of sci-entific theory that the rationality of theoryconstruction can be sustained. For withoutthem science would remain, as in empiricism,a purely internal process—with the familiar-ity of image replacing the reinforcement ofsensation, still lacking a rational dynamic of

change.

⇓ If it is granted that these genera-tive mechanisms are real, then the realistcan explain theory construction by exper-imental activity:

46:1 Now once it is granted that mecha-nisms and structures may be said to be real,we can provide an interpretation of the in-dependence of causal laws from the patternsof events, and a fortiori of the rationale ofexperimental activity. For the real basis ofthis independence lies in the independenceof the generative mechanisms of nature fromthe events they generate. Such mechanismsendure even when not acting; and act in theirnormal way even when the consequents ofthe law-like statements they ground are, ow-ing to the operation of intervening mecha-nisms or countervailing causes, unrealized.It is the role of the experimental scientist to


exclude such interventions which are usual;and to trigger the mechanism so that it isactive. The activity of the mechanism maythen be studied without interference. Andit is this characteristic pattern of activity ormode of operation that is described in thestatement of a causal law. It is only un-der closed conditions that there will be aone-to-one relationship between the causallaw and the sequence of events. And it isnormally only in the laboratory that theseenduring mechanisms of nature, whose op-erations are described in the statements ofcausal laws, become actually manifest andempirically accessible to men. But becausethey endure and continue to act, when stimu-lated, in their normal way outside those con-ditions, their use to explain phenomena andresistence to pseudo-falsification in open sys-tems can be rationally justified.

⇓ Is the following paragraph now theepistemological conclusion from this on-tology? Or is it a summary of what he isgoing to prove?

46:2/o Only if causal laws persist through,which means they must be irreducible to,the flux of conditions can the idea of theuniversality of a known law be sustained.And only if they have a reality distinct fromthat of events can the assumption of a natu-ral necessity be justified. On this view lawsare not empirical statements, but statementsabout the forms of activity characteristic ofthe things of the world. And their neces-sity is that of a natural connection, not thatof a human rule. There is a distinction be-tween the real structures and mechanisms ofthe world and the actual patterns of eventsthat they generate. And this distinction inturn justifies the more familiar one between


necessary and accidental sequences. For anecessary sequence is simply one which cor-responds to, or is in phase with, a real con-nection; that is, it is a real connection actu-ally manifest in the sequence of events thatoccurs.

⇓ The world consists of mechanisms,not events. These mechanism combine toproduce the flux of phenomena. In otherwords: the empirical realist says: “be-hind phenomena are atomistic events,”while the realist says: “behind phenom-ena are enduring mechanisms.” Thesemechanisms are not unknowable, butthe process of gaining knowledge aboutthem is arduous. It requires intellecual,practico-technical, and perceptual skills.

47:1 The world consists of mechanisms notevents. Such mechanisms combine to gener-

ate the flux of phenomena that constitute theactual states and happenings of the world.They may be said to be real, though it israrely that they are actually manifest andrarer still that they are empirically identi-fied by men. They are the intransitive ob-jects of scientific theory. They are quite inde-pendent of men—as thinkers, causal agentsand perceivers. They are not unknowable,although knowledge of them depends upon arare blending of <35> intellectual, practico-technical and perceptual skills. They are notartificial constructs. But neither are theyPlatonic forms. For they can become mani-fest to men in experience. Thus we are notimprisoned in caves, either of our own or ofnature’s making. We are not doomed to ig-norance. But neither are we spontaneouslyfree. This is the arduous task of science:the production of the knowledge of those en-


during and continually active mechanisms ofnature that produce the phenomena of ourworld.

⇓ Those objecting to this reconstitu-tion of ontology question first the intran-sitivity and then the structured characterof the postulated objects of scientific in-quiry:

47:2 Objections may be made to my pro-posed reconstitution of an ontological realm,which question in turn the intransitivity andthe structured character of the postulatedobjects of scientific inquiry, i.e. the ideas oftheir categorical independence from men andevents respectively. I will consider the twokinds of objections in turn.

⇓ First objection: the idea of a worldwithout men is unintelligible. Bhaskarsays it is counterfactual, but it does make

sense.47:3 Thus, it might be objected that the

very idea of a world without men is unintel-ligible because the conditions under which itis true would make its being conceived im-possible. But I can think of a world withoutmen; and I can think of a world without my-self. No-one can truly say ‘I do not exist’but that does not mean that ‘I do not exist’is unintelligible; or that it cannot be mean-ingfully, just because it cannot be truly said.It is no objection to the intelligibility of astatement that it is counter-factual. Indeedit is only because it is intelligible that we cansay that it is counter-factual.

⇓ Second objection: it is impossible tothink of a world without men. Some-how the process of thinking is insepara-ble from men, Bhaskar assumes the ar-gument is: we cannot think anything


without thinking about us thinking it orabout ourselves. Bhaskar tries to refutethis by distinguishing betwen being awareof thinking, thinking about thinking, andthinking about the thinker. I guess hemeans: every human activity is charac-terized by awareness of what we are do-ing, but this does not mean we are think-ing about what we are doing or thinkingabout ourselves.

47:4/o Someone might hold that to thinkof a world without men is not so much un-intelligible as impossible; that we must pic-ture ourselves in any picture. Now it is afact about human beings that we can dothis. But we do not have to do it, any morethan an artist must initial his work. Theidea may be perhaps that a thought mustalways contain, or at least be accompaniedby, a thought of the thinker of the thought

thinking the thought. Clearly if this wereso, an infinite regress would be impossible toavoid. However, to be aware of the fact thatI am thinking of a particular topic x, it is notnecessary for me to be thinking of that fact.Such awareness may be expressed in thought,but when it is the topic is no longer x butmy thought of x. It is possible for A to thinkepsilon and to be aware of thinking epsilonwithout thinking about thinking epsilon; andunless this were so no-one could ever intelli-gently think. Moreover it is possible for A tothink about thinking epsilon without think-ing about his (A’s) thinking epsilon. Think-ing about thinking about a particular topicmust be distinguished from thinking aboutthe thinker of the topic.32

⇓ After having said that there is noabsurdity in the supposition of a worldwithout men, he says now how important


it is to ask what a world would be like inwhich there are no men. “For example,things would still act, be subject to laws,and preserve their identity throug certainchanges.”

48:1 There is no absurdity in the supposi-tion of a world without men. Rather it is apossibility presupposed by the social activityof science. It is important to establish thisfact. For we are too liable to underestimatethe power of the pictures, often unconscious,which underpin philosophical theories. Suchpictures indeed often hold our philosophicalimagination ‘captive’.33 Our philosophy ofscience is heavily anthropocentric, which iswhy it is important to consider what it wouldbe possible to say about our world if therewere no men, given that we know that ourworld is one in which science is as a <36>matter of fact possible. For example things

would still act, be subject to laws and pre-serve their identity through certain changes.

⇓ Second objection says: there is a dif-ference between laws and constant con-junctions, however from this does not fol-low that laws are produced by real gen-erative mechanisms.

48:2/o A second kind of objection mightfocus on the structured character of the pos-tulated objects of scientific inquiry, question-ing not so much the idea itself but the inter-pretation I have given to it; and in particu-lar the explanatory value of the particularontology proposed. Thus it might be ob-jected that, while the transcendental argu-ment from experimental activity in Section3 establishing the distinctiveness of causallaws and patterns of events, is sound, theintroduction of the concept of generativemechanisms to provide a real basis for causal


laws is gratuitous.

⇓ Expands what a generative mecha-nism is: it continues to act in its normalway even if the results of its operationsare modified. And it matters even less forthis mechanism whether things are per-ceived by man or not.

49:1 What does it mean to say that a gen-erative mechanism endures and acts in itscharacteristic way? It does not mean, wehave seen, that a regular sequence of eventsoccurs or is experienced; though the occur-rence of such a sequence may, in special cir-cumstances, provide empirical grounds forthe hypothesis of the existence of the mech-anism. For the intelligibility of experimen-tal activity entails that the particular mech-anism endures and at least some mecha-nisms act through the flux of conditions thatdetermine whether they are active and co-

determine the manifest outcome of their ac-tivity. That is to say, it entails that gen-erative mechanisms endure even when inac-tive and act even where, as in open systems,there is no one-to-one relationship betweenthe causal law representing the characteris-tic mode of operation of the mechanism andthe particular sequence of events that occurs.In particular, it entails that mechanisms actin their normal way outside the closed condi-tions that enable us to experimentally iden-tify them and whether or not we do so; i.e.whether or not the results of their operationsare modified, and whether or not these re-sults are perceived by men. (In the formercase we could talk of a disjuncture betweenthe domains of the real and the actual; inthe latter case of a disjuncture between thedomains of the real and the empirical.)

⇓ Now he goes from generative mecha-


nisms to active things.

49:2/o Now the reason why the conceptof a causal law cannot itself be taken as on-tologically basic is because its analysis pre-supposes a ‘real something’ over and aboveand independent of patterns of events; andit is for the status of this real somethingthat the concept of a generative mechanismis groomed. But then does to say that a gen-erative mechanism endures and acts in itscharacteristic way mean anything more thanto say that a thing goes on acting in a certainway? As stated the reformulation is ambigu-ous. For the continuance of a form or patternof activity can be interpreted in an empiri-cal or a non-empirical way. The intelligibil-ity of experimental activity requires the lat-ter non-empirical interpretation. For it en-tails, as we have seen, that causal laws per-sist and are efficacious in open systems, i.e.

outside the conditions that enable us to em-pirically identify them. Now accepting thisnon-empirical interpretation means that ref-erence to causal laws involves centrally ref-erence to causal agents; that is, to thingsendowed with causal powers.

From the fact that generative mecha-nisms persist even if they do not haveactual results he concludes that genera-tive mechanisms must be the powers ofthings. Otherwise they could not be saidto persist without making an illicit reifi-cation of the generative mechanism. Itake it, this means, it cannot just be a se-quence of actions and reactions, becauseif it were this, then it would not start upagain after it dies down. Since it per-sists through periods in which it does notmanifest itself, even without needing a


new trigger (it is a tendency, not just apower), this indicates that it is anchoredin something that persists, i.e., in a thing.It is central to this argument that lawsare not merely the power of things, butit is the tendency of things.

Here is my definition of tendencies, li-abilities, powers: powers is somethingwhich it can do, liabilities are thingswhich can be done to it, things to whichit is susceptible. The human body is sus-ceptible to cancer or to virus infections.Tendencies are things which it has to do.Even if it is barred from doing it for awhile, afterwards it will spontaneouslystart up doing it again, and perhaps onecan say it has been trying to do it allalong.On this interpretation then the genera-

tive mechanisms of nature <37> exist asthe causal powers of things. We now havea perfectly acceptable ontological basis forcausal laws. For if it is wrong to reifycausal laws, and it is wrong to reify gen-erative mechanisms, it cannot be wrong toreify things! However, the fact that the tran-scendental analysis of experimental activityshowed that generative mechanisms must goon acting (i.e. that causal laws must be ef-ficacious) outside the closed conditions thatpermit their identification means that causallaws cannot be simply analysed as powers.Rather they must be analysed as tendencies.For whereas powers are potentialities whichmay or may not be exercised, tendencies arepotentialities which may be exercised or as itwere ‘in play’ without being realized or man-ifest in any particular outcome. They aretherefore just right for the analysis of causal


laws.34

⇓ Laws are not enduring powers,i.e., they endure between experiments inwhich they are triggered, but laws referto the continuing activity of things. Thisis why experiments are relevant for non-experimental situations.

50:1 If the analysis of causal laws (andgenerative mechanisms) is to be given by theconcept of things and not events (a possibil-ity which I have already rejected by demon-strating in Section 3 their categorical in-dependence from events), the considerationthat they not only persist but are efficaciousin open systems, which is presupposed by theintelligibility of experimental activity, entailsthat causal laws must be analysed as tenden-cies. For tendencies are powers which maybe exercised without being fulfilled or actu-alized (as well as being fulfilled or actual-

ized unperceived by men). It is by referencenot just to the enduring powers but the un-realized activities or unmanifest (or incom-pletely manifest) actions of things that thephenomena of the world are explained. Itis the idea of continuing activity as distinctfrom that of enduring power that the conceptof tendency is designed to capture. In theconcept of tendency, the concept of power isthus literally dynamized or set in motion.

⇓ Therfore not subjunctive, but normicstatements. Not statements: “if this thenthat,” but statements: “normally so andso,” not counterfactual but transfactual.

50:2/o In the full analysis of law-like state-ments we are thus concerned with a newkind of conditional: which specifies the exer-cise of possibilities which need not be mani-fest in any particular outcome. Such condi-tionals are normic,35 rather than subjunc-


tive. They do not say what would hap-pen, but what is happening in a perhapsunmanifest way. Whereas a powers state-ment says A would ψ in appropriate circum-stances, a normic statement says that A re-ally is ψ’ing, whether or not its actual (orperceivable) effects are counteracted. Theyare not counter-factuals, but transfactuals;they take us to a level at which things arereally going on irrespective of the actual out-come. To invoke a causal law is to invoke anormic conditional. A full analysis of normicand tendency statements will be providedlater. For the moment, it should be notedthat normic statements provide the correctanalysis of the nomic indicative form. Anomic statement is a transfactual statement,with actual instances in the laboratory thatconstitute its empirical grounds.

⇓ Then he talks about powers, liabili-

ties, and tendencies. The scientist is notsaying what is going to happen, but whatkind of things there are, and what theytend to do.

51:1 The world consists of things, notevents. Most things are complex objects, invirtue of which they possess an ensemble oftendencies, liabilities and powers. It is byreference to the exercise of their tendencies,liabilities and powers that the phenomenaof the world are explained. Such continu-ing <38> activity is in turn referred back forexplanation to the essential nature of things.On this conception of science it is concernedessentially with what kinds of things theyare and with what they tend to do; it is onlyderivatively concerned with predicting whatis actually going to happen. It is only rarely,and normally under conditions which are ar-tificially produced and controlled, that scien-


tists can do the latter. And, when they do,its significance lies precisely in the light thatit casts on the enduring natures and ways ofacting of independently existing and trans-factually active things.

Question 25 According to transcenden-tal realism, explanation and predictionare not the same thing. What is the roleof prediction in science?

⇓ There is nothing mysterious aboutthe concept of a generative mechanism,they are simply the ways of acting ofthings.

51:2/o There is nothing esoteric or mys-terious about the concept of the generativemechanisms of nature, which provide the realbasis of causal laws. For a generative mech-anism is nothing other than a way of act-

ing of a thing. It endures, and under ap-propriate circumstances is exercised, as longas the properties that account for it per-sist. Laws then are neither empirical state-ments (statements about experiences) norstatements about events. Rather they arestatements about the ways of acting of inde-pendently existing and transfactually activethings.

Compare that with his earlier state-ment that causal laws are held to be mys-terious. I guess what he means here is: heis coming up with a very simple explana-tion of something that otherwise can onlybe explained by a “mystery.”

What Bhaskar accords with Collier’sdefinition of generative mechanisms asthose aspects of a thing which cause itto act in a certain way.


⇓ After his negative interpretation in35:2/o, he now can give a positive inter-pretation of what structured and intran-sitive means: there are active structureswhich cause the regularities.

52:1/o It is now possible to give a posi-tive interpretation of our characterization inSection 3 of the objects of scientific investiga-tion, at least in so far as they are causal laws,as ‘structured intransitive’. ‘Structured’ inso far as it is the activities of mechanismsand causal structures, not the occurrence ofevents, that are designated in statements ofcausal law. ‘Intransitive’ in so far as themechanisms and causal structures, whose ac-tivity is designated, endure and act quiteindependently of men. To discover the in-dependently existing and transfactually ac-tive machinery of nature is not, it shouldbe stressed, the aim of an independent in-

quiry of metaphysics. Rather, it is the end towhich all the empirical efforts of science aredirected. Ontology has been vindicated notas providing a set of necessary truths abouta mysterious underlying physical realm, butas providing a set of conditionally necessarytruths about our ordinary world as inves-tigated by science. It is important to beclear about what philosophical argument canachieve. Thus as a piece of philosophy wecan say (given that science occurs) that somereal things and generative mechanisms mustexist (and act). But philosophical argumentcannot establish which ones actually do; or,to put it the other way round, what the realmechanisms are. That is up to science to dis-cover. That generative mechanisms must ex-ist and sometimes act independently of menand that they must be irreducible to the pat-terns of events they generate is presupposed


by the intelligibility of experimental activ-ity. But is up to actual experiments to tellus what the mechanisms of nature are. Here,as elsewhere, it is the task of philosophy toanalyse notions which in their substantiveemployment have only a syncategorematicuse. Thus whenever a scientist refers to athing or event, structure or law, or says thatsomething exists or acts in a certain way hemust refer to it under some particular de-scription; he is using the notion of thing,law, existence, etc. But it is the task of thephilosopher to analyse the concept as such.To argue that this task is both <39> legiti-mate and necessary is not to populate theworld with (or to suppose that there is aworld of) things without names or events-in-general.

Greater part of this paragraph is acharacterization of what philosophy can

do: it can give second-order arguemnts,but then he says somethign else: the sci-entists in their conclusions use conceptslike thing, law, existence, etc., and itis a task of the philosopher to analyzethe concept as such. This is legitimateand necessary even though “there are nothings in general.” I guess, it is an ab-straction, and the philosopher character-izes this abstraction.

⇓ “Tidies up” the analysis of experi-mental activity.

53:1 I am now in a position to tidy upmy analysis of experimental activity. Theexperimental scientist must perform two es-sential functions in an experiment. First,he must trigger the mechanism under studyto ensure that it is active; and secondly, hemust prevent any interference with the op-


eration of the mechanism. These activitiescould be designated ‘experimental produc-tion’ and ‘experimental control’. The for-mer is necessary to ensure the satisfaction ofthe antecedent (or stimulus) conditions, thelatter to ensure the realization of the conse-quent, i.e. that a closure has been obtained.But both involve changing or being preparedto change the ‘course of nature’, i.e. the se-quence of events that would otherwise haveoccurred.36 In a simple electrical experimentdesigned to illustrate say Ohm’s Law, thewiring of an electric circuit and the genera-tion of an electric current would constitute‘experimental production’; maintaining theappropriate resistance levels, ensuring thatno new magnetic field is suddenly placed inthe neighbourhood of the circuit, etc. wouldthen constitute ‘experimental control’.

I guess again, instead of a “can only”

conclusion before, he now gives a positiveand more precise interpretation. Experi-ment consists of two things: experimentalproduction (setting the antecedent, thestimulus in motion), und experimentalcontrol (to make sure that the consequentcan be realized, is not thwarted). Herehe uses the word “realization” where per-haps actualization would be better. Butone should even talk about empiricaliza-tion: it must not only occur but also beobserved.

But another question: if laws are ten-dencies of things, why do they need stim-uli?

Footnote: experimental production φchanges the state of the world α to a,which is the stimulus which tends to cre-ate b. And experimental control (there is


a typo in the third line of the footnote,he says “production” where he should say“control”) ψ excludes the intervention ofother effects, so that a can generate b, sothat the sequence a → b is a constantconjunction of events. But it is a resultof human action.

I think it is not only so in experimentalcontrol, but also in building machines, inelectric wiring: insulating the cables is ψ.⇓ says the same thing again, and adds

that this requires a degree of plasticity ofthe antecedent and ambient conditions tohuman manipulation, which is not easilywon. By “‘plasticity” he means that hu-man agency must be able to affect thosethings.

53:2 Only if the mechanism is active andthe system in which it operates is closed can

scientists in general record a unique relation-ship between the antecedent and consequentof a law-like statement. The aim of an ex-periment is to get a single mechanism goingin isolation and record its effects. Outside aclosed system these will normally be affectedby the operations of other mechanisms, ei-ther of the same or of different kinds, too, sothat no unique relationship between the vari-ables or precise description of the mode ofoperation of the mechanism will be possible.In general, experimental activity requires adegree of plasticity of the antecedent (stimu-lus) and circumambient conditions to humanmanipulation and control. Such plasticity isnot easily won. ‘Experimental design’ is asubstantial theoretical labour in itself.

⇓ Again several statements:

(1) experiment is not just putting aquestion to nature, it is a practical ques-


tion, we put it with our hands so tospeak. The last sentence here is whatthis whole paragraph is all about: the on-tological distinction between causal lawsand patterns of events depends on onlytwo premises: (a) men can interfere withnature, and (b) this “disruption” of na-ture is significant for learning about na-ture.

(2) he calls it somewhat a scandal thatempiricists “should not have untertakenan analysis of the conditions under whichexperience is significant in science.” (Bythis he apparently does not mean: whichexperiences are significant, but the ques-tion: in which kind of world are experi-ences significant in the way they are inscience.)

54:1 It has often been said, metaphorically

speaking, that in an experiment we put aquestion to nature. But it has not been saidthat the question we put is a practical one—with our hands, so to speak. The weaknessof previous analyses of experimental activityis that they have not appreciated the signifi-cance of the fact that conjunctions of phe-nomena have to be worked for practically(as well as in thought); that conjunctionsare not given to, but made by us. In animportant study, von Wright has seen this.But he has not drawn the correct conclusionfrom it: which is that, just because the ex-perimenter is a causal agent of the sequenceof events, there must be an ontological dis-tinction between the sequence he generatesand the causal law it enables him to identify.Any other conclusion renders experimentalactivity pointless. (Why generate that se-quence?) The reason for von Wright’s failure


to see this stems from his unfortunate initialassumption of (as he puts it) a ‘Tractatus-world’, i.e. a world of logically independentatomistic states of <40> affairs (which as-tonishingly he seems to regard as a harmlesssimplification);37 which precludes him fromseeing laws as anything other than condi-tional statements about atomistic states ofaffairs. It is of course something of a scan-dal that empiricists who invoke experienceas the sole ground of knowledge and scien-tific knowledge as their paradigm should nothave undertaken an analysis of the condi-tions under which experience is significantin science. It should be stressed that the re-sult that there is an ontological distinctionbetween causal laws and patterns of eventsdepends upon only two premises: (i) thatmen are causal agents capable of interferingwith the course of nature and (ii) that ex-

perimental activity, the planned disruptionof the course of nature, is a significant fea-ture of science.

⇓ The question which an experimentputs to nature is not only a practicalquestion, but it also needs sophisticatedmeasurement instruments.

54:2/o In stressing the practical compo-nent of experimental activity, it is importantnot to forget the theoretical side. In an ex-periment men put a question to nature. Butthey must put it in a language that natureunderstands, as well as in a form that makespossible an unambiguous reply. It is difficultto overestimate the importance for modernscience of the development of instrumentssuch as clocks and telescopes, which may beseen as devices designed to decipher the vo-cabulary of nature. Both the constructionand the interpretation of such instruments


depended upon theory. Hooke’s law, for ex-ample, is literally built into the constructionof spring balances.38 Experimental confirma-tion of Galilean dynamics was delayed for along time by the difficulty of measuring ‘themost fundamental magnitude of dynamics’,i.e. time. But when the Huyghens eventuallysucceeded in building such a clock in 1659 itwas only by basing it on the new dynamics(the very dynamics it was designed to vin-dicate) and in particular the theory of theisochronous curve of the pendulum.39 Sim-ilarly it has been convincingly argued thatthe development of cosmology in the early17th century was held up by the absenceof an adequate theory of telescopic vision.40

In short, experimental activity depends cru-cially upon the adequacy of the theories(sometimes referred to as ‘auxiliary’) accord-ing to which the experimental equipment is

constructed and its results interpreted.

⇓ Two furthergoing questions: (1)role of thought to understand the world(Einstein’s Gedankenexperimente). Canthat be understood without reducing theworld to a set of axioms? (2) What todo in sciences which do not allow experi-ments, i.e., are there surrogates of exper-iments.

55:1/o Two problems are raised by myanalysis of experimental activity. First, weknow that much science, of what might becalled a fundamental kind, has proceeded byway of ‘thought’ rather than by actual ex-periment. As Dijksterhuis has put it: ‘Ingeneral one has to take stories about ex-periments by Galileo, as well as his oppo-nents with some reserve. As a rule they wereperformed mentally, or they are merely de-scribed as possibilities.’41 It seems that Ein-


stein too was not averse to the occasional‘Gedankexperimente’.42 This raises the ques-tion of whether, and if so how, pure thoughtcan anticipate a law? And the problem ofhow, if it can, we then avoid the rational-ist conclusion that provided only our anx-iom base is strong enough we could deduceall the laws of nature without recourse toexperience. Secondly, we know that in manyfields, most notably history and the humansciences and in the biological sciences in as-pects of their work, experimental activity is<41> impossible. This raises the question ofwhether there are, or it is possible to devisefor them, surrogates of the experimental es-tablishment of closed systems in physics andchemistry? And here again there lurks anunacceptable rationalist implication. Bothpose prima facie problems for transcendentalrealism, which I hope to be able to resolve

at a later stage in this study.

1.6 A sketch of a cri-tique of empiricalrealism

56:1/o I have argued that the causal struc-tures and generative mechanisms of naturemust exist and act independently of the con-ditions that allow men access to them, sothat they must be assumed to be structuredand intransitive, i.e. relatively independentof the patterns of events and the actionsof men alike. Similarly I have argued thatevents must occur independently of the ex-periences in which they are apprehended.Structures and mechanisms then are real anddistinct from the patterns of events that they

1.6. A SKETCH OF A CRITIQUE OF EMPIRICAL REALISM 69

generate; just as events are real and distinctfrom the experiences in which they are ap-prehended. Mechanisms, events and experi-ences thus constitute three overlapping do-mains of reality, viz. the domains of the real,the actual and the empirical. This is repre-sented in Table 1.1 below.

56:2

The crux of my objection to the doctrine ofempirical realism should now be clear. Byconstituting an ontology based on the cate-gory of experience, as expressed in the con-cept of the empirical world and mediated bythe ideas of the actuality of the causal lawsand the ubiquity of constant conjunctions,three domains of reality are collapsed intoone. This prevents the question of the con-ditions under which experience is in fact sig-nificant in science from being posed; and theways in which these three levels are brought

into harmony or phase with one another frombeing described.

57:1 <42> Now these three levels of real-ity are not naturally or normally in phase. Itis the social activity of science which makesthem so. Experiences, and the facts theyground, are social products; and the con-junctions of events, that, when apprehendedin experience, provide the empirical groundsfor causal laws, are, as we have seen, socialproducts too. It can thus be seen that un-derlying and necessary for the implicit ontol-ogy of empirical realism is an implicit soci-ology in which facts and their conjunctionsare seen as given by nature or spontaneously(voluntaristically) produced by men. In thischapter I have outlined an answer to thequestion ‘what must the world be like forscience to be possible?’. In Chapter 3 I willask ‘what must society be like for science to


be possible?’; i.e. I shall attempt a transcen-dental deduction of certain basic sociologicalcategories from an investigation of the condi-tions for the possibility of science. The an-swer to these two questions will constitutethe interwoven themes of this work. It is im-possible to over-emphasize how closely theyare connected. For once, for example, we re-ject the doctrine that there are everywhere innature such things as spontaneously occur-ring parallel cases and see rather that in gen-eral they have to be assiduously worked forand artificially produced in the social activ-ity of science, we are forced to constitute anontology of structures distinct from events.

57:2/o For us, for the moment, it is suf-ficient merely to note that the most im-portant feature of science neglected by thedoctrine of empirical realism is that it iswork; and hard work at that. Work con-

sists, paradigmatically, in the transforma-tion of given products. Scientific change isan integral feature of science, in which whatis transformed is a part of the formally ac-credited stock of scientific knowledge. In ascientific training the object transformed isnot knowledge but man himself. But in bothcases what is transformed is itself alreadya social product. The peculiar significanceof experimental activity is that man quamaterial object (rather than simply thinkeror perceiver) exercises his causal powers totransform the natural world itself, of whichhe is also a part. Now corresponding to thedissolution of ontology in philosophy, therehas been a parallel denegation of the socialcharacter of science. In Chapter 3 I will setout to vindicate sociology in an attempt torender intelligible scientific change. This willenable me to reconstitute a transitive dimen-


sion, as complementary to the intransitiveone established here.

58:1 The concept of the empirical world isanthropocentric. The world is what men canexperience. But the couple of this concept,and from a realist meta-perspective neces-sary to sustain it, is the absence of the con-cept of the antecedent social activity neces-sary to make experience significant in sci-ence. And this has the objectionable ideo-logical consequence (from the point of viewof the practice of science) that whatever mencurrently experience is unquestionably theworld. Now it is central to the argumentof this study that the concepts ‘empirical’and ‘sense-experience’ belong quite unequiv-ocally to the social world of science. Experi-ences are a part, and when set in <43> thecontext of the social activity of science anepistemically critical part, of the world. But

just because they are a part of the world theycannot be used to define it. An experienceto be significant in science must normally bethe result of a social process of production;in this sense it is the end, not the begin-ning of a journey. But only transcendentalrealism can explain why scientists are cor-rect in regarding experience as in the last in-stance the test of theory. For it is by meansof it that, under conditions which are arti-ficially produced and controlled, skilled mencan come to have access to those enduringand active structures, normally hidden orpresent to men only in distorted form, thatgenerate the actual phenomena of our world.Empirical realism depends upon a reductionof the real to the actual and of the actual tothe empirical. It thus presupposes the spon-taneity of conjunctions and of facts. And indoing so presupposes a closed world and a


completed science.

58:2/o It is important to stress that I amnot saying that experiences are less real thanevents, or events less real than structures.This is the kind of mistake that is encour-aged by the way in which Eddington for-mulated his problem of the relationship be-tween the familiar and the scientific worlds;in which he described the situation as one inwhich there were ‘duplicates’ of every object:two tables, two chairs, two pens, etc.43 Sincethen the problem has always seemed to bethat of saying which object is real. For theordinary language instrumentalist the scien-tific object is an artificial construct;44 for thescientistic super-realist the familiar object amere illusion.45 For the transcendental real-ist however this formulation of the problem isbogus. For if there is a relationship betweenthe worlds it is one of natural generation, not

an interpretation of man. The relationshipis not between a real and an imaginary ob-ject, but between two kinds of real object,one of which is very small. The relationshipbetween electrons and tables has to be un-derstood in terms of causal connections, notcorrespondence rules. Consequents are notless real, or the statements describing themless true, in virtue of their being effects; anymore than causes, in virtue of being recon-dite, must be imaginary. In particular, thefact that the properties of everyday objects,at what has been picturesquely described asthe zone of the middle dimensions,46 can beexplained in terms of the very small (or thevery large) does not render them less realthan the entities that account for them; any-more than zinc and sulphuric acid cease toreact in a certain way when we explain their


reaction in terms of their atomic structure.

59:1/o For the transcendental realist laws,though not our knowledge of them, are cate-gorically independent of men—as thinkers,causal agents and perceivers. Transcen-dental realism can thus accommodate bothLocke’s view that there are (or may be) lawswhich are unknowable;47 and Kneale’s sug-gestion that there are (or may be) laws whoseinstances are unperceivable.48 But it allowsin addition the possibility of known laws,whose instances are perceivable, but which,when not instanced in closed systems, re-main unmanifest to men. However, my in-terpretation of these possibilities is differentfrom Locke’s <44> (and Kneale’s). For thetranscendental realist, our knowledge, per-ceptual skills and causal powers are set in thecontext of the ongoing social activity of sci-ence; and in the course of it they are contin-

ually being extended, to which process therecan be no a priori limits. Thus though itmay be necessary, to the extent that scienceis always incomplete, that at any momentof time some laws are unknowable; it is notnecessary that any particular laws are.

60:1 Locke’s mistake in failing to appreci-ate the possibility that the ‘sad experience’of chemists who ‘sometimes in vain, searchfor the same qualities in one parcel of sul-phur, antimony or vitriol, which they havefound in others’49 might come to be trans-formed in the course of the development ofscience into a knowledge of the ‘constitu-tion of their insensible parts, from which flowthose sensible qualities, which serve us to dis-tinguish one from another’50 was not a sci-entific mistake. It did not consist in his fail-ure to foresee the development of the theoryof atomic number and valency or to predict


Mendeleyeev’s predictions. His scepticismover the possibility of a scientific knowledgeof real essences was a philosophical mistake,rooted in his theory of ideas. For if all ourknowledge is acquired in perception and per-ception constitutes the world, there can beno place for an antecedent cause of knowl-edge (or of perception). But as only whatis seen as socially produced can be seen asputatively socially transformable, this leadsinevitably to an a-historical view of science.

60:2/o Locke’s error was not thereforebased on an inadequate knowledge of chem-istry. But on an inadequate concept of thetransitive dimension of science, which pre-vented him from seeing the current stateof chemistry as what it was, viz. the cur-rent state of a science; and which thusallowed him to be influenced by it intopropounding a general philosophical the-

sis about knowledge—and in particular ofcourse about the impossibility of a certainkind of knowledge, viz. of real essences.Locke’s case has a general moral. For with-out a concept of science as a process-in-motion and of knowledge as possessing (inthe sense indicated in Section 1 above) amaterial cause, it is easy to argue from thecurrent state of a science to a philosophicalthesis about knowledge. Consider, for exam-ple, the Copenhagen interpretation of Quan-tum theory. More important perhaps, theinfluence of Newtonian mechanics on 18thcentury philosophy led to a kind of stasis inthought from which the philosophy of sciencehas still to recover. Action-by-contact as aparadigm of causality, the celestial closureas a model of knowledge, gravity as the tem-plate of our ignorance all had a disastrous ef-fect. The underdevelopment of the sciences


of substance in comparison with the scienceof motion (of the time), and the form thatthe latter took, thus had, at a decisive mo-ment in the history of philosophy, throughthe generation of a static philosophical con-ception of knowledge, a permanent effect onall subsequent ‘philosophy of science’. It isin this sense that in philosophy we are stillprisoners of the scientific thought of the past.

61:1/o <45> The anthropocentric andepistemic biases of classical philosophy haveresulted in the dominance, in philosophy,of what might be styled ‘idols’ of a Baco-nian kind. These are false conceptions whichcause men to see, in philosophy, everythingin relation to themselves (cf. the concept ofthe empirical world) and their present knowl-edge. Six hundred years ago, Copernicusargued that the universe does not revolvearound man. And yet in philosophy we still

represent things as if it did. In the philos-ophy of science there must be two Coper-nican Revolutions. The first establishing atransitive dimension in which our knowledgeis seen to be socially produced, and as suchneither an epiphenomenon of nature nor aconvention of man. The second establishingan intransitive dimension, based on the re-constitution of a philosophical ontology, inwhich the world of which, in the social ac-tivity of science, knowledge is obtained isseen to be in general quite independent ofman. These Copernican Revolutions mustbe given a Copernican interpretation (forphilosophy has its Osianders too); which iswhy we need the metaphysics of transcen-dental realism, which will be vindicated byits capacity to render intelligible the under-analysed phenomenon of science.

62:1 Corresponding to the two criteria ad-


vanced on page 24 above two acid tests for aphilosophy of science may be developed:

62:2 (1) is knowledge regarded as sociallyproduced, i.e. as having a material cause ofits own kind? or is it read straight onto thenatural world or out of the human mind?

62:3 (2) are the objects of knowledge re-garded as existing and acting independentlyof men? or do they depend implicitly or ex-plicitly upon men for their existence and/oractivity?

62:4 Scientists try to discover the rea-sons for things and events, patterns and pro-cesses, sequences and structures. To under-stand how they do so one needs both a con-cept of the transitive process of knowledge-production and a concept of the intransi-tive objects of the knowledge they produce:the real mechanisms that generate the ac-tual phenomena of the world, including as a

special case our perceptions of them.

1.7 Footnotes to Chap-ter 1

1 See Aristotle, Metaphysics, 1.3.

2 See J. R. Ravetz, Scientific Knowledge andits Social Problems, pp. 116–19.

3 Cf. R. Harre, Philosophies of Science, pp.176–7.

4 W. Thomson, Notes of Lectures on Molecu-lar Dynamics p. 132.

5 See A. S. Eve, Rutherford.

6 Cf. I. Kant, On the Distinctiveness of thePrinciples of Natural Theology and Morals.

7 J. S. Mill, A System of Logic, Bk. III, Chap.3, Sect. 1.

8 Cf. N. R Hanson, Patterns of Discovery,Chap. 1.

9 Cf. J. J. C. smart, Philosophy and Scientific

1.7. FOOTNOTES TO CHAPTER 1 77

Realism, pp. 38–9.10 Cf. J. J. Gibson, The Senses Considered as

Perceptual Systems.11 Cf. M. Hollis, ‘Reason and Reality’, P.A.S.

Vol. LXVIII (1967–8), p. 279.12 G. E. M. Anscombe, Causality and Deter-

mination, p. 22; and G. H. von Wright, Expla-nation and Understanding, pp. 60–4.

13 Cf. G. E. M. Anscombe, op. cit., p. 21.14 Cf. Ravetz’s ‘4th law of thermo-dynamics’:

no experiment goes properly the first time. SeeJ. R. Ravetz, op. cit., p. 76.

15 Cf. F. Dretske, Seeing and Knowing, Chap.3.

16 Cf. G. H. von Wright, op cit., Chap. 2.17 L. Wittgenstein, Tractatus Logico-

Philosophicus, 6.35.18 I. Kant, Critique of Pure Reason, B.724.19 See e.g. A. J. Ayer, Language, Truth and

Logic, pp. 31–41.20 A recent book, A. Quinton’s Nature of

Things, is vitiated by a failure to distinguishthese two questions. From the outset Quintontends to identify the problem of fundamental

entities with that of the foundations of knowl-edge (p. 5). This leads him to argue that ‘if allpossible evidence for the existence of theoreti-cal entities is provided by common observablesit follows . . . that the logically indispensableevidence, and thus the sense of assertions abouttheoretical entities must be capable of being ex-pressed in terms of those common observablesand thus that theoretical entities can have onlya derived and dependent existence’ (p. 285).

21 D. H. Mellor, ‘Physics and Furniture’,American Philosophical Quarterly, Studies inthe Philosophy of Science, p. 184.

22 See W. V. O. Quine, ‘Designation and Ex-istence’, Readings in Philosophical Analysis, ed.H. Feigl and W. Sellars, p. 50; Methods of Logic,p. 224; and From a Logical Point of View, Chap.1 and passim.

23 D. Hume, Treatise on Human Nature, p.632.

24 D. Hume, op. cit., p. 132. Cf. Newton’s 4thrule of reasoning in philosophy: ‘propositions in-ferred by general induction from phenomena [areto be regarded as] true . . . till such time as other


phenomena occur by which they may either bemade more accurate or liable to exceptions’, I.Newton, Principia Mathematica, Bk. III.

25 D. Hume, op. cit., p. 231.

26 T. S. Kuhn, The Structure of ScientificRevolutions, Chaps. II–IV.

27 See e.g. M. Hesse, In Defence of Objectiv-ity, p. 14.

28 See e.g. G. Buchdahl, op. cit., p. 31.

The next footnote gives premonitionsof the arguments in [Bha89].

29 The structure of such a critique would be asfollows: If the subject matter of social psychol-ogy is such that a science of social psychologyis possible and social psychologists are to haveknowledge of it, then social psychologists shoulddo phi, ψ, etc. rather than x, omega, etc. Thetranscendental realist could thus not accept thenotorious definition of economics as what econ-omists do. For him, whether or not they actu-ally do economics is at least in part a contingentquestion. Notice that the formula I have usedleaves the question of whether a science of social

psychology is possible open. This is importantbecause for the transcendental realist it is thenature of the object that determines the possi-bility of a science. Thus he can allow, withoutparadox, that there may be no humanly intelli-gible pattern to be discovered in the stars or po-litically intelligible pattern in voting behaviour.So that no science of astrology or psephology ispossible, no matter now scrupulously ‘scientificmethod’ is adhered to.

30 K. R. Popper, Conjectures and Refuta-tions, p. 192.

31 Cf. J. J. C. Smart, op. cit., pp. 149–51.32 In fact men have the capacity to be self-

conscious in two ways: first in being consciousof what they are doing; and secondly, in beingconscious of their doing it. That these two arenot equivalent is shown by the fact that in somecontexts a person may know what he has donebut not that he has done it and vice-versa.

33 L. Wittgenstein. Philosophical Investiga-tion, 115.

34 A recent antecedent of the view that causallaws should be analysed as tendencies is con-


tained in P. T. Geach, ‘Aquinas’, Three Philoso-phers G. E. M. Anscombe and P. T. Geach, pp.101ff. Important works in the recent develop-ment of the concept of powers are W. D. Joske,Material Objects, Chaps. 4 and 5; M. R. Ayers,The Refutation of Determinism, Chaps. 3–5; andR. Harre, Principles of Scientific Thinking, esp.Chap. 10.

35 I owe this term to M. Scriven, ‘Truisms asthe Grounds for Historical Explanation’, The-ories of History, ed. P. Gardiner, pp. 464ff.Scriven uses it to refer to generalizations ground-ing historical explanations which contain modi-fiers such as ‘normally’, ‘tendency’, ‘usually’, etc.My use of the term is substantially different. Butit is the nearest thing to an antecedent for thekind of conditional I am concerned with.

36 Formally we could say that in experimen-tal production by doing φ we change α to a soaltering the state that would otherwise have pre-vailed; and in experimental production by doingor being prepared to do ψ we exclude the in-tervention of elements β1 . . . βn, so allowing themechanism M set in motion by a to generate b.

The sequence a.b thus appears as a consequenceof the results of our actions. It is in this sensethat a closure is normally a human product.

37 See G. H. von Wright, op. cit., pp. 43–45.38 Cf. N. R. Hanson, Observation and Expla-

nation, p. 56.39 See e.g. A. Koyre, Metaphysics and Mea-

surement, Chap. 4.40 V. Ronchi, ‘Complexities, advances and

misconceptions in the development of the scienceof vision: what is being discovered?’, ScientificChange, ed. A. Crombie, pp. 542–61.

41 E. J. Dijksterhuis, The Mechanisation ofthe World Picture, p. 338.

42 See K. R. Popper, The Logic of ScientificDiscovery, App. XI.

43 A. S. Eddington, The Nature of the Phys-ical World, p. xi. Stebbing substituted the ideaof ‘counterparts’ for that of ‘duplicates’ in herrendering of the problem. See L. S. Stebbing,Philosophy and The Physicists, p. 60.

44 See e.g. L. S. Stebbing, op. cit., p. 66; andG. Ryle, Dilemmas, p. 80.

45 See e.g. W. Sellars, ‘The Language of Theo-


ries’, Current Issues in the Philosophy of Science,ed. H. Feigl and G. Maxwell, p. 76; and P. K.Feyerabend, ‘Explanation, Reduction and Em-piricism’, Minnesota Studies in the Philosophyof Science, Vol. III, ed. H. Feigl and A. Maxwell,p. 83.

46 M. Capek, The Philosophical Impact ofContemporary Physics, p. 294.

47 J. Locke, Essay Concerning Human Under-standing, esp. Bk. IV, Chap. III.

48 W. Kneale, Probability and Induction, pp.97–103. Kneale’s point could be strengthened byan argument to show that in the case of physicaltheories the basic entities must be unperceivable.For if they were perceivable it would seem possi-ble to ask what caused them to manifest them-selves to us as perceivable; in which case theycould not be basic. This is a general argumentin favour of a field-theoretic interpretation of ba-sic entities in physics. Cf. Dingle’s comment thatif photons could be seen they would get in theway (J. J. C. Smart, op. cit., p. 38).

49 J. Locke, op. cit., Bk. III, Chap. 6.9.50 J. Locke, op. cit., Bk. IV, Chap. 3.7.


Table 1.1Domain of

Real Actual EmpiricalMechanisms XEvents X XExperiences X X X

Note. for transcendental realism dr ≥ da ≥ de . . . (i) where dr, da, and de are the domainsof the real, the actual and the empirical respectively.For empirical realism dr = da = de . . . (ii).Comment: (ii) is a special case of (i), which depends in general upon antecedent socialactivity, and in which(a) for da = de the events are known under epistemically significant descriptions, whichdepends upon skilled perception (and thus a skilled perceiver);(b) for dr = da an antecedent closure has been obtained, which depends upon skilledexperimentation (and thus the planned disruption of nature).

Chapter 2

Actualism and the Conceptof a Closure

2.1 Introduction: Onthe Actuality of theCausal Connection

63:1 (i) ‘We have no knowledge of any-thing but phaenomena; and our knowledge

82

2.1. INTRODUCTION: ON THE ACTUALITY OF THE CAUSAL CONNECTION83

of phaenomena is relative not absolute. Weknow not the essence, nor the real mode ofproduction, of any fact, but only its rela-tions to other facts in the way of successionor similitude. These relations are constant;that is, always the same in the same circum-stances. The constant resemblances whichlink phaenomena together, and the constantsequences which unite them as antecedentand consequent, are termed their laws. Thelaws of phaenomena are all we know respect-ing them. Their essential nature, and theirultimate causes, either efficient or final, areunknown and inscrutable to us.’1

63:2 (ii) ‘To give a causal explanation ofan event means to deduce a statement, us-ing as premises of the deduction one or moreuniversal laws, together with certain singularstatements, the initial conditions.’2

63:3 (iii) ‘Since in a fully-stated D-

N [deductive-nomological] explanation of aparticular event the explanans logically im-plies the explanandum, we may say thatthe explanatory argument might have beenused for the deductive prediction of theexplanandum-event if the laws and the par-ticular facts adduced in its explanans hadbeen known and taken into account at a suit-able earlier time. In this sense a D-N expla-nation is a potential D-N prediction.’3

63:4/o (iv) ‘Criteria of refutation must belaid down beforehand: it must be agreedwhich observable situations, if actually ob-served, mean the theory is refuted.’4

64:1 (v) ‘Important though other consid-erations may be, the acid test of a theory isits predictive power.’5

64:2 It is the argument of this chapterthat there is a distinction between open andclosed systems, which most existing philos-

84 CHAPTER 2. ACTUALISM AND THE CONCEPT OF A CLOSURE

ophy of science ignores; that closed systemsare a condition of its most important doc-trines, such as those expressed in (i)–(v); andthat once the significance of this distinctionis grasped the plausibility of these doctrinescollapses.

64:3 (i)–(v) possess a family connection, inthat they all depend upon the Humean the-ory of law. This theory has often been crit-icized on the grounds that a constant con-junction of events cannot be sufficient for alaw. But most of its critics have been con-tent to allow that it is at least necessary.6

It is this notion, viz. that laws are constantconjunctions of events (plus some disputedcontribution of mind), that I intend to chal-lenge. It arises as follows: If atomistic eventsor states of affairs constitute the world then,for general knowledge to be possible, the re-lations between such events or states of af-

fairs must be constant. (This is the assump-tion that the concept of a closure is designedto mark.) On the other hand if, as I intend toargue, they are not in general constant, thenatomistic events cannot provide the only ba-sis of ontology. And the philosophical theo-ries based on the identification of causal lawswith empirical regularities plus must all beradically wrong.

64:4/o I shall use the term ‘actualism’ torefer to the doctrine of the actuality of causallaws; that is, to the idea that laws are re-lations between events or states of affairs(which are thought to constitute the objectsof actual or possible experiences).7 Behindthis idea of course lies the notion that onlythe actual (identified as the determinate ob-ject of the empirical) is real. Given it, theconstant conjunction plus analysis of lawsmust follow. In this chapter I shall not be


concerned with the ‘plus’. Moreover for con-venience I shall use the term ‘empiricism’in a generic way so as to cover the entirepost-Humean tradition of empirical realism,and in particular both its positivist and neo-Kantian wings. No harm will be done by thisusage as I am here attacking an assumption,viz. that a constant conjunction is necessaryfor a law, common to both.

65:1 The argument of this chapter is bothsimple and, I think, novel. Leaving asideastronomy, it is only under conditions thatare experimentally produced and controlledthat a closure, and hence a constant conjunc-tion of events, is possible. The empiricistis now caught in a terrible dilemma: for tothe extent that the antecedents of law-likestatements are instantiated in open systems,he must sacrifice either the universal char-acter or the empirical status of laws. If,

on the other hand, he attempts to avoidthis dilemma by restricting the applicationof laws to closed systems (e.g. by makingthe satisfaction of a ceteris paribus clause acondition of their applicability), he is facedwith the embarrassing question of what gov-erns phenomena in open systems. If he re-fuses the question, he is still left with theproblem of accounting for experimental ac-tivity, and thus the establishment of ‘laws’,however restricted, in the first place. Hisonly options here are to deny either thatmen are causal agents or that experimen-tal activity plays any role in science. For iflaws are sequences of events and men, beingcausal agents, can bring about and preventsuch sequences, there can be no rationale foraccording one rather than another sequencethe status of law. A sequence of events canonly function as a criterion for a law if the


latter is ontologically irreducible to the for-mer. And so we come back to the argumentof 1.3, where I showed how the intelligibilityof experimental activity presupposes the on-tological distinctiveness of causal laws fromthe patterns of events. But it can now beseen that not only the experimental estab-lishment but the practical application of ourknowledge depends upon this same ontolog-ical distinction. For unless causal laws per-sisted and operated outside the context oftheir closure, i.e. where no constant conjunc-tions of events obtained, science could notbe used in the explanation, prediction, con-struction and diagnosis of the phenomena ofordinary life.

65:2/o The empiricist makes mattersworse for himself by the fact that he not onlyties laws to closed systems, but typically tiesthe activities of explanation, prediction and

the identification of causes to our knowledgeof laws. A reductio ad absurdum quickly fol-lows. For to the extent that we seek to ex-plain, predict and identify the causes of phe-nomena that occur in open systems, theseactivities become impossible. And to theextent that they are necessary for our so-cial life, empiricism does. Thus there is nonecessity that we should exist. But, giventhat we do, if our social life is to be possi-ble we must ascribe causal responsibility inopen systems. And given this, the Humeantheory just cannot apply. Now I want to ar-gue both that laws apply in open and closedsystems alike; and, in a subsidiary thesis,that these other activities do not necessar-ily depend upon (though they may make useof) a knowledge of laws. From this perspec-tive the Popper-Hempel theory of explana-tion, for example, may be seen to involve a


double mistake: first, that explanation al-ways involves laws; and secondly, that lawsare or depend upon empirical regularities.

66:1/o My overall aim, it will be remem-bered, is to argue that the ultimate objectsof scientific understanding are neither pat-terns of events nor models but the thingsthat produce and the mechanisms that gen-erate the flux of the phenomena of the world.Scientists attempt to discover the way thingsact, a knowledge typically expressed in laws;and what things are, a knowledge (to bediscussed later) typically expressed in realdefinitions. Statements of laws, I have sug-gested, are statements about the tendenciesof things which may not be actualized, andmay not be manifest to men; they are notstatements about conjunctions of events, orexperiences. But in developing this theoryI do not attach any great importance to the

word or even the concept ‘law’. Rather whatis essential to the realism developed here isthe idea that the things and mechanisms ofnature, that constitute the intransitive ob-jects of scientific theory, both exist and actindependently of the conditions, normallyproduced by men, that allow men access tothem. For experimental science to be pos-sible the world must be at least partiallyopen. But if there is a real distinction be-tween open systems and closed and our in-tuitions about the rationality of science areto be preserved there must be a real distinc-tion between structures and events. In thisrespect the differentiation of phenomena stillprovides the best argument for the stratifi-cation of the world.

67:1 In isolating the special conditions un-der which a regular sequence or pattern ofevents occurs; that is, in which (to adopt the


realist mode) there is a correspondence be-tween causal laws and the pattern of events,I will be leaving it up to the epistemologistwhether he wants to sustain the universalityof laws (and inter alia the intelligibility ofexperimental activity) by postulating a cate-gorical ontological distinction between them.If this is done by the development of a non-empiricist ontology and an analysis of lawsas non-empirical and normic along the linesindicated in 1.5 above, the way is also openfor an adequate theory of natural necessityand natural kinds. On the other hand with-out this, I shall argue in Chapter 3, our in-tuitions about the lack of sufficiency of theHumean criteria for law (and the theories ofscience based on it) cannot in the last in-stance be sustained.

67:2 In showing how a closure is a con-dition of the intelligibility of empirical re-

alism my primary intention in this chap-ter is critical. For it is when confrontedwith the fact and implications of open sys-tems that the limitations of this approach toscience—with its flat ontology of undifferen-tiated experience—become most apparent.But, in dealing with the problems posed bythe largely unanalysed phenomena of opensystems, I will also be compelled to developnew and more general alternatives to thetheories, such as those expressed in (i)–(v)above, that are based on the tacit assump-tion that a closure is the universal rule in na-ture; rather than the rare exception I shallcontend it is.

67:3/o Underlying the widespread, if tacit,acceptance in philosophy of the idea of theubiquity of constant conjunctions in nature(an idea which is not confined to the empiri-cist tradition)8 and hence of the doctrine of


the actuality of causal laws is the notion thatthe universe is at rock bottom deterministic;that, in the image of Leibniz, the present isbig (in the sense of pregnant) with the fu-ture; that it, as it were, already contains itnow. It is the job of science to discover theiron laws that uniquely determine its motion.Once these laws are discovered, given only aknowledge of any complete state-description,‘nothing would be uncertain [to science] andthe future, as the past, would be presentto its eyes.’9 What accounts for the hold ofthis fantastic conception on our philosoph-ical imagination? The philosophical argu-ments for it are, taken on their own, as weshall see, pretty poor. Why then do we feelthe force of this picture? Partly no doubt be-cause many things are de facto predictable,many processes are effectively isolated andmany systems more or less closed; so that,

given that rough-and-ready regularities areeverywhere at hand, it seems plausible tosuppose that underlying them there mustbe more exact ones. Partly no doubt be-cause of an obsession with the consequencesand a neglect of the conditions of the exper-imental paradigm, the single case that thehypothetico-deductive view of science fits.Above all perhaps because of the misconcep-tion created by the celestial closure securedby Newtonian science, and in particular bythe idea that this closure embodied both amodel of phenomena and a model of sci-ence. This was a double mistake. For it wasnot the human mind, as Laplace thought,10

that gave its special perfection to astronomy.Rather it was the peculiar conditions of theplanets, and in particular the constancy ofboth their intrinsic states and the externalforces on them, that made possible the ob-


served regularities. Moreover for Newtonian,as for any other, mechanics celestial phenom-ena functioned merely as evidence that bod-ies tend to act in certain ways. The laws ofmotion, for example, describe actions whichare unobservable in principle. But the ten-dencies of the bodies to which they apply arereal; and would account for any disruption inthe established order of our solar system.

68:1/o But, it might be objected, is notthe universe in the end nothing but a giantmachine with inexorable laws of motion gov-erning everything that happens within it? Iwant to say three things: First, that the var-ious sciences treat the world as a network of‘machines’, of various shapes and sizes anddegrees of complexity, whose proper princi-ples of explanation are not all of the samekind as, let alone reducible to, those of classi-cal mechanics. Secondly, that the behaviour

of ‘machines’, including classical mechani-cal ones, cannot be adequately described, letalone understood, in terms of the ‘wheneverx, then y’ formula of regularity determinism.Thirdly, that even if the world were a single‘machine’ this would still provide no groundsfor the constant conjunction idea, or a for-tiori any of the theories of science that de-pend upon it. Regularity determinism is amistake, which has been disastrous for ourunderstanding of science.

2.2 Regularity Deter-minism and theQuest for a Closure

69:1 So far I have been content merely toidentify a closed system as one in which a

2.2. REGULARITY DETERMINISM AND THE QUEST FOR A CLOSURE 91

constant conjunction of events obtains. Butwe must now establish exactly what this en-tails. It might be thought that the idea of aclosed system could be elucidated quite sim-ply as a fragment or sector of the world ef-fectively cut off for a period of time fromnon-constant external influences. Althoughthis gives one clear sense of ‘a closed system’such a system would not necessarily satisfythe criterion of invariance implicit in the em-piricist analysis of law. For one thing con-ditions would have to be placed on the indi-viduals composing the system and the wayin which the states of the system were to bespecified. But even if this were done therewould still be no guarantee that the crite-rion of invariance would be satisfied. Theassumption that it would be depends uponthe metaphysical thesis of regularity deter-minism. This may be defined as follows: For

every event y there is an event x or set ofevents x1 . . . xn such that x or x1 . . . xn andy are regularly conjoined under some set ofdescriptions.11 That is, the world is so con-stituted that there are descriptions such thatfor every event the simple formula, ‘When-ever this, then that’ applies.

70:1 Such a thesis stands to the practice ofscience as a regulative principle. Such princi-ples are, as is well known, neither empiricallynor theoretically refutable (or confirmable).But I will contend that they are metaphys-ically so. My procedure will be to see whatthis thesis entails about the nature of theworld and about the nature of science; and toassess its adequacy in these respects in rela-tion to other possible regulative standpoints.To do so I will work out critical or test con-ditions for the thesis of regularity determin-ism; that is, conditions such that if they were


known to be satisfied and the constant con-junction formula was not vindicated the reg-ularity determinist would be bound to admithis thesis refuted. In this way I hope to showjust how restricted in its ontological presup-positions and restrictive in its methodolog-ical responses regularity determinism is. Indeveloping these limit conditions for a clo-sure I will thus be developing the conditionsunder which, on the supposition that regu-larity determinism is true, a constant con-junction of events must obtain. However, Iwill define a ‘closed system’ simply as onein which a constant conjunction of eventsobtains; i.e. in which an event of type a isinvariably accompanied by an event of typeb. Clearly the possibility of such a systemdoes not depend upon the truth of regular-ity determinism. Nor need such a system be‘closed’ in any more picturesque sense of the

word.

70:2/o Regularity determinism must bestraightaway distinguished from two otherforms of determinism: which may be called‘ubiquity’ and ‘intelligibility’ determinism.Ubiquity determinism asserts that everyevent has a real cause; intelligibility deter-minism that every event has an intelligiblecause; regularity determinism that the same(type of) event has the same (type of) cause.The concepts of ‘cause’ involved in the threedeterminisms are of course distinct. For theubiquity determinist the cause is that thing,material or agent which is productive of aneffect; for the intelligibility determinist it issimply that which renders an event intelligi-ble to men;12 for the regularity deterministit is the total set of conditions that regularlyproceeds or accompanies an event.13 Of thethree determinisms, regularity determinism


is clearly the most restrictive; and ubiquitydeterminism is more general than intelligibil-ity determinism, because it licenses no pre-sumption that the real cause of an event willalways be intelligible to men. The realist, in-telligibility and regularity concepts of causeare of course naturally associated with thetranscendental realist, transcendental ideal-ist and classical empiricist philosophies ofscience.

71:1 All three determinisms must be dis-tinguished from the idea of ‘computational’determinism. This is the supposition thatfor each characteristic or trait of any mate-rial body at any moment of time there existsat least one set of statements from which,together with the relevant antecedent state-descriptions, that trait is deducible. It isimportant to realize that because no restric-tion is placed on the statements used in the

deduction of traits, computational determin-ism is a truism. It says merely that givenany system it is possible to work out an al-gorithm for the successful computation of itstraits; or, in other words, that there is a con-sistent way of describing the development ofany system over any finite period of time.Moreover in general there will be an infinitenumber of ways of doing so. It is thereforean uninteresting truism—save perhaps to re-mind us that the notion of disorder or chaosis always relative to a particular type of or-der or class of functions;14 and that the cri-terion of deducibility is too easily satisfied tobe capable of functioning alone as a decisionrule for the selection of ‘law-like’ or ‘theo-retical’ statements and requires at the veryleast supplementation by criteria that placesome restriction on the number, type or in-


terpretation of the statements concerned.

71:2/o It is especially important to dis-tinguish regularity from computational de-terminism. For it is at least part of the in-tention of the former to assert (a) that thesame cause and effect sometimes as a matterof fact recurs and (b) that the same causeand effect could always logically recur. Forunless (a) were true instances would not fallunder it and unless (b) were true they couldnot fall under it; so that it would be at bestvacuous and at worst false. On the otherhand computational determinism is consis-tent with ‘law-like’ formulations which areso specific and detailed as to reduce thepractical likelihood of the event’s recurrencetowards zero or which mention the spatio-temporal location within which it occurredor which individuate it with a definite de-scription or a proper name (e.g. The Battle

of Edge Hill). We need not dwell on thesepossibilities here. For they have been thor-oughly explored by philosophers concernedto defend the autonomy of historiographyfrom (Positivist) science.15 For the regularitydeterminist the necessity for such formula-tions merely indicates the ignorance of thedescriber. The anti-regularity-determinist,on the other hand, may take it as a signof emergence or novelty in nature or evenof the self-determination of some agent orstructure. Unlike computational determin-ism, regularity determinism is not triviallysatisfied. It does however share with it thefeature that if there is one set of law-likestatements which satisfies it there will be aninfinite number of such sets. Hence it too re-quires supplementation by additional crite-ria, such as simplicity, intelligibility or real-ism, if it is to be capable of yielding a unique


decision procedure for the selection of laws’.

72:1/o The total cause (in Mill’s sense)of an event will normally be a complex setof conditions x1 . . . xn rather than a singleevent x. One could distinguish here be-tween the individual or component eventsor states and the total or conjunct eventor state; and refer to the case where morethan one factor is at work, following Mill,as that of ‘multiple causation’. In the sameway the consequent event will also normallybe complex, i.e. y1 . . . yn rather than simplyy; and so we could talk of a corresponding‘multiplicity of effects’. There is a genuine‘plurality of causes’ when the same effectarises from different (i.e. alternative) sets ofconditions.16 Is a plurality of causes consis-tent with regularity determinism? Note itis consistent with predictability. For given aknowledge of the conditions which actually

prevail, the effect is uniquely predictable.But it is not consistent with retrodictabil-ity. For given the effect, we cannot uniquelyinfer the cause. And this requirement isexplicit both in the Laplacean ideal, whichplaces the past on a par with the future,and the Humean definition of cause (‘in otherwords, where if the first object had not been,the second never had existed’), which makesthe cause both necessary and sufficient forthe effect.17 To satisfy this requirement theidea that every event has one and only onecause (or set of causes) must be incorporatedinto our definition of regularity determinism.This must now read: the same cause alwayshas the same effect and the same effect al-ways has the same cause; so eliminating boththe possibility of a disjunctive plurality ofcauses and of a disjunctive plurality of ef-


fects.

73:1/o Now suppose we had a system suchthat events of type a were invariably followedby events of type b. We could then say thata closure had been obtained. A closure is ofcourse always relative to a particular set ofevents and a particular region of space andperiod of time. Now supposing that at sometime t’ an event of type a was not followed byan event of type b we would have to say thatthe system was ‘open’, our criteria of open-ness just being the fact that events of typea had not been invariably followed by eventsof type b under their given descriptions; i.e.the instability, in space or over time, of actu-ally recorded empirical relationships. Shouldwe say that the system had been closed butwas now open or that it was open all along?There is nothing at stake here—it dependsentirely on the time period for which ‘the

system’ is defined: if and only if it includes t’it is open. ‘System’ here carries no indepen-dent semantic force. Either way the naturalresponse of the regularity determinist to thissituation would be to suppose that we hadleft out of our state-description an individualor variable that made a difference: that theconjunct events referred to under the samedescription ‘a’ before and after t’ were not re-ally the same in all relevant respects; in shortthat the system had been incompletely de-scribed (or enumerated). For example if thesystem was a classical mechanical one, wherethe presupposition was that mass, positionand velocity were the only relevant variables,it would be natural to suppose that a rele-vant individual had been omitted from thespecification of the overall state of the sys-tem, i.e. the total conjunct event.

74:1 To fix the point, imagine a universe


composed of a finite number of differentkinds of knives, forks and spoons. Supposethat we attempted to work out a generalrule which would enable us to predict thatwhen the knives and forks were in a certainposition (including naturally the datum ofwhether they were on the dining-room tableor in the kitchen drawer) they were invari-ably followed by another constellation of po-sitions at the next meal. We might find thisimpossible, unless we took into account thepositions of the spoons; so that we could saythat our system had been incompletely de-scribed in the former case, owing to the omis-sion of a causally relevant (in the Humeansense) variable. Of course in time we mightfind that we could only satisfy the demandsof regularity determinism by taking into ac-count further variables, e.g. the shapes andnumber of glasses at the meal; or even vari-

ables of an entirely different kind, e.g. theroom temperature—to distinguish say be-tween winter breakfasts when porridge wasthe rule and summer ones when grapefruitwas.

74:2 Subject to an important qualificationto be discussed below it can thus be seenthat regularity determinism implies a partic-ular kind of response to the phenomenon ofopen systems, i.e. to the instability of empir-ical relationships, viz. to assume that somecausally relevant individual or variable hasbeen left out of the description. On the otherhand, even if this were the case, stable em-pirical relationships might still be possible aslong as the values of the omitted variablesremained constant. A closure thus dependsupon either the actual isolation of a systemfrom external influences or the constancy of


those influences.

75:1 Assuming that a system was effec-tively isolated from non-constant external in-fluences and regularity determinism was truewould this then ensure the satisfaction of the‘whenever this, then that’ formula? Let ussuppose that we are interested in explain-ing (in the sense of Hempel and Hume) thebehaviour of some individual N, say an ele-phant. Would a knowledge of the total an-tecedent state-description enable us to pre-dict its behaviour? No—for if N is character-ized by internal structure and complexity itmay behave differently in the same externalcircumstances in virtue of its different inter-nal states. Thus what happens when I prodan elephant depends at least in part uponwhat state it is in, e.g. whether it is asleepor not; and thus to that extent the total stateof the universe, of which the elephant occu-

pies a part, will be a variable. So either theabsence or the constancy of internal struc-ture must also be a condition for a closure.And the regularity determinist now has an-other possible response to the condition ofopen-ness, namely to assume that the indi-viduals of the system have not been given asimple or atomistic enough description.

75:2 It is easy to see that an actualisolation and atomistic individuals will bepreferred, on epistemic grounds, by theregularity determinist to constant externaland internal conditions. For the regular-ity determinist has no warrant for assumingthat these conditions will remain constant.18

Whether they do or not will depend upona whole host of factors concerning which exhypothesi he has no knowledge and aboutwhich he is not therefore in a position tomake any kind of claim. Only with an ac-


tual isolation of atomistic individuals will theregularity determinist be able to categori-cally predict the future; without it, it alwaysremains on the cards that an unpredictedchange in the external circumstances of thesystem or the internal states of its individu-als will occur so as to upset an establishedregularity and so render inapplicable any hy-pothetical predictions, formulated subject totwo ceteris paribus clauses.

75:3/o I have been tacitly assuming up tillnow that the overall states of the system canbe represented as an additive function of thestates of the individual components of thesystem. This represents a third kind of re-quirement for a closure. Here again a closureis possible if the principle of organization isnon-additive, provided it remains constant;though here again the regularity deterministwill prefer the alternative of additivity on

epistemic grounds. Behind the assumptionof additivity lies of course the idea that thebehaviour of aggregates and wholes can al-ways be described in terms of the behaviourof their component parts. The assumptionsof atomicity and additivity are closely con-nected. For to say of some system that it isirreducible (in this sense) to its componentparts is presumably to say that it must beviewed as a thing in its own right, at its ownlevel (I am not concerned with the groundsfor this now.) Conversely to suppose thatit is always possible to give an atomistic de-scription of prima facie complex things is tosuppose that they can always be viewed assystems or parts of systems which can beanalysed in terms of their component parts,conceived as atomistic individuals.

76:1/o The critical conditions for a closureare set out in Table 2.1. The satisfaction of


one each of the system, individual and orga-nizational conditions is sufficient, on the sup-position that regularity determinism is true,for a closure; but not necessary for it. Ifa recorded regularity breaks down the reg-ularity determinist must assume that it isbecause one of these conditions is not satis-fied. Until now in developing the conditionsfor a closure I have been using the categories‘internal’ and ‘external’. But the categories‘intrinsic’, and ‘extrinsic’ are better in thatthey are not explicitly tied to a spatial char-acteristic and hence to things of a certaintype. Thus the category ‘intrinsic’ includessome properties of things which lie outsidetheir spatial envelope, e.g. a magnet’s field,and others which cannot be identified spa-tially at all, e.g. a person’s charm. Andit excludes others which do lie within theirspatial envelope, e.g. properties belonging to

things of another type. ‘Isolation’ must alsobe interpreted metaphorically; but there isa sense in which ‘atomicity’ must be takenliterally.

76:2 Table 2.1 Limit Conditions for a Clo-sure, i.e. for the Stability of Empirical Rela-tionships —————————————————————- Conditions (1) Epistemically(2) Epistemically for a Closure DominantCase Recessive Case —————————————————————- (A) System Isola-tion Constancy of Extrinsic Conditions (B)Individuals Atomicity Constancy of IntrinsicConditions (C) Principle of Additive Con-stancy of Non- Organisation Additive Prin-ciple —————————————————————-

77:1 Now it is easy to see that oncean actual isolation and an atomistic de-scription are set up as norms two regresses


are initiated, viz. to systems so vast thatthey exclude nothing and to individuals sominute that they include nothing. These re-gresses are typically manifest in research pro-grammes, characteristic of positivistic sci-ence, which could be dubbed ‘interactionism’and ‘reductionism’ respectively. It is clearthat they can only be halted by constitut-ing a level of autonomous being, somewherebetween the universe and atomistic individ-uals. But for the empiricist committed toregularity determinism to do so involves anenormous risk. For it means he must be pre-pared to snap the Humean link that ties thejustified performance of cognitive acts suchas the ascription of causes to a knowledge ofempirical invariances; and to say ‘yes I knowa is not invariably followed by b, yet a causedb here’.

77:2 These regresses generate notorious

paradoxes in their wake. For since in thefirst case there are at the limit no conditionsextrinsic to the system a full causal state-ment would seem to entail a complete state-description (or a complete history) of theworld. Similarly as in the second case therecan be no conditions intrinsic to the thinga causal statement entails a complete reduc-tion of things into their presumed atomisticcomponents (or their original conditions). Inthe first case we cannot make (or can atbest only make in a pragmatic way) the dis-tinction between causes and conditions; inthe second case that between individuals andvariables or between a thing and its circum-stances. In neither case do we have the keyconcept of a causal agent; i.e. the thing thatproduced or the mechanism that generated,in the circumstances that actually prevailed,


the effect in question.

77:3/o Open systems situate the possibil-ity of two kinds of possibility statements:epistemic and natural. The regularity deter-minist can accommodate the former but notthe latter. For he may allow that an eventmay be uncertain due to the describer’signorance of the complete atomistic state-description necessary to deduce it. But hecannot allow that there is a sense to a state-ment about what an individual can do inde-pendently of whether or not it will do it. Fornatural possibility statements to be possiblecondition B1 must be unfulfilled; i.e. theremust be complex things, possessing intrinsicstructure, to which the natural possibility isascribed. Without this, the distinction be-tween a power and its exercise would be in-deed, as Hume supposed, entirely ‘frivolous’.So in part the issue between the regularity

determinist and realist turns on the questionof whether there are objects not susceptibleof an atomistic analysis and in what way thisis significant for science.

78:1 A special and very important case ofthe individual conditions for a closure is thusgiven by: B1’ the absence of powers, whichis dependent upon the absence of intrinsicstructure (implied by atomicity); and B2’the constancy of powers, which is dependentupon the constancy of intrinsic structure. Ifthere are complex things then it becomes im-portant to distinguish between the subjectsand the conditions of action. For conditionsis an epistemic, not an ontological category.The conditions change, but they do not havethe power to change. Only things and mate-rials and people have ‘powers’.

78:2/o I have argued in effect in Chapter1 that for experimental science to be possi-

2.3. THE CLASSICAL PARADIGM OF ACTION 103

ble the world must be open but susceptibleto regional closures. Now corresponding tothe view of the world as open and the viewof the world as closed we have two entirelydifferent conceptions of science. The tran-scendental realist sees the various sciences asattempting to understand things and struc-tures in themselves, at their own level of be-ing, without making reference to the diverseconditions under which they exist and act,and as making causal claims which are spe-cific to the events and individuals concerned.And he sees this not as a tactic or manoeu-vre or mechanism of knowledge; but as ac-cording with the ways things really are, theway things must be if our knowledge of themis to be possible. The regularity determin-ist, on the other hand, will seek in his questfor a closure, if the very special conditionsspecified in Table 2.1 are not satisfied, to in-

corporate more and more elements into hisdescriptions and/or to break down his unitsof study into finer and finer constituents inan effort to stabilize his field. And he mustsee this merely as an attempt to vindicate tohimself the rule, which may be fairly styleda dogma, that like follows like or whenever xthen y.

2.3 The ClassicalParadigm of Ac-tion

79:1 In developing the critical conditions fora closure I have argued that epistemologicalconsiderations tilt them in favour of the casesof atomicity and additivity, i.e. cases B1 andC1 in Table 2.1 above. I now want to bring


out the concept of action implied by theseconditions, which will be seen to be as specialas the conditions themselves.

79:2 If individuals are atomistic, then allcauses must be extrinsic. And if systemsare lacking in intrinsic structure (i.e. are ex-hausted by the properties of the individu-als composing them), then there can be noaction at a distance (nor can distance be avariable in action), so all action must be bycontact. But as atomistic individuals cannotcontribute to or be affected by the action,it must consist in the communication of re-ceived properties. But as the only propertypossessed by such individuals is their posi-tion in space at a moment in time, the onlyproperty they can communicate is their mo-tion, i.e. their movement through space intime.

79:3 Thus a particular physical conception

of action, involving a corpuscularian view ofmatter and a mechanical view of causality,in which all causes are regarded as efficientand external to the thing in which the changeoccurs, is implied by the limit conditions fora closure. These form the twin pivots of whatI am going to call the classical paradigm ofaction.

79:4 Now this paradigm has three varia-tions: a physical one; a metaphysical one;and a distinctively positivistic epistemolog-ical one. It is important to keep them dis-tinct. For the prestige of the epistemologicalvariation, which underpins the Humean the-ory of laws, has stemmed in large part fromits misidentification, by Locke and Newtonamong others, with the physical one.

79:5/o On the physical concept, matteris viewed as composed of rigid corpuscleswhose motion accounts for the aggregative


and observed behaviour of things. These cor-puscles exchange momenta and redistributevelocities among themselves by impact; andthey move through space according to thestrict laws of mechanics. Action is seenas consisting in the impression of externalforces upon these corpuscles, which merelypass on their received motion by direct im-pact. Events are nothing but the displace-ment of these basic units of matter in spaceand time, they are not the transformationof pre-formed substances. Large scale eventsor macroscopic changes are merely the sur-face effects of such displacements; qualitativevariety and change are similarly the effectsof different arrangements and motions of thecorpuscles and their aggregates. The prop-erties of aggregates are essentially the sameas those of their component parts, thoughthey can manifest themselves to observers in

different, i.e. genuinely emergent ways (as inthe case of secondary qualities).

80:1 Now this physical concept of actionencourages, though it does not imply, ametaphysical concept, which was especiallyprominent in seventeenth century rational-ist thought. On it, matter is viewed as es-sentially passive and inert (which is strictlyspeaking irreconcilable with the role playedin classical mechanics by the concept of in-ertia); and causation is viewed as a processwhich is linear and unidirectional, as well asexternal and inconsistent with real novelty.In contrast to matter is mind. As matter ispassive, mind is active. And qualitative vari-ety and change, denied a place in the sphereof matter, are seen as contributions of thehuman mind.

80:2 Historically associated with this en-semble of physical and metaphysical ideas


is another distinctively positivistic epistemo-logical concept. On this concept, things areviewed as ultimately resolvable into simplequalities apprehended in sense-experience,rather than as aggregates of elementary unitsof matter in motion. The ontology is oneof atomistic (and independent) events ratherthan one of atomistic (and rigid) corpuscles.And causality is seen as the regular concom-mitance of such events, rather than the im-pression of external forces upon such indi-viduals. For the positivist, the concept ofaction is but a gloss put on our apprehensionof such sequences. It can have no applicationin a world of independent events.

80:3/o It should be noted that the episte-mological variation on the classical paradigmonly makes full sense if it is given a phenom-enalistic interpretation. For if and only ifwe start from punctiform sense-experiences

and reconstruct the world from them, is it atall plausible to suppose that the world con-sists of independent and atomistic events andstates of affairs, constituting the surrogatesof such sense-experiences, with relations thatare as external to one another as the sense-experiences that ground them are held to bedistinct. Conversely, if we reverse this pre-supposition and situate sense-experience as anatural process occurring in the world thenit always makes sense, on being told thatsome event or happening occurred, to ask:‘to what thing?’ (One could then of coursego on to explain the change, along the lines ofthe corpuscularian/mechanical programme,in terms of the arrangements and motions ofthe basic individuals.)

81:1/o On the physical concept it isthe principles of action (the strict laws ofmechanics) that explain macroscopic be-


haviour, including observed regularities; i.e.the relationship between individuals and sys-tems is one of explanation, not analysis.Now I have argued in Section 1 that theselaws cannot be plausibly construed as em-pirical regularities or constant conjunctionsof events. Moreover, if it is the behaviourof the individuals that is required to ac-count for the constant conjunctions of eventsrather than the other way round the pointof the Humean analysis of cause and theHempelian analysis of explanation is lost. Ifthe laws cannot be regarded as empirical reg-ularities, the concepts in them cannot be re-garded as given in or abstracted from expe-rience. Firstly, because we ordinarily experi-ence motion in terms of transitive verbs suchas ‘pushing’ and ‘pulling’ which cannot beexplicated ostensively.19 Secondly, becausethe concepts that figure in them demanded

a radical break with the pre-existing Aris-totelian scheme which certainly was then,and may still be now, closer to our ordi-nary way of thinking.20 Consider, for exam-ple, the radical change embodied in the prin-ciple of inertia, that only change in motionand not motion itself (i.e. change in posi-tion) requires explanation. Thus the notionthat ‘the concepts of classical physics are justa refinement of the concepts of daily life’21

must be rejected, and with it the notion thatlaws are inductive generalizations from ev-eryday experiences or at any rate descrip-tions of them or instruments for predictingtherm abandoned. For the experiences areliterally different—an idea which will onlyseem paradoxical to someone who is implic-itly misusing the category of experience todefine the world.

82:1 The key concept of atomicity is


given radically different interpretations inthe physical and epistemological variations.In the physical variation its primary iden-tification is size; in the epistemological onesimplicity. Physical atoms are (or were) nec-essarily unobservable, theoretical entities;epistemological ones were the raw data of ex-perience. The former were the ultimate en-tities of the world, the latter the basic build-ing blocks of knowledge. Physical atomswere explainers; epistemological ones justi-fiers. Physical atoms were Parmenidean in-dividuals, epistemological ones Heracliteaninstants.

82:2/o Epistemic atomicity requires thatevents should be apprehended in raw sense-experience. Hence it restricts the possiblenumbers of a cause and an effect set toone. But the constant conjunction analysisof cause requires that the cause and effect

properly so-called should be constantly con-joined. The requirements of epistemic atom-icity and empirical invariance can only bereconciled if the sequence of events is a lin-ear process and each event is a member ofa homogeneous series of determinations. Ifthis were the case then the apprehension ofthe cause event would indeed license the ex-pectation of the effect event. But this leavesthe odd assumption of the linearity of pro-cesses to be justified. In the formula ‘when-ever x then y’, it is only if ‘x’ and ‘y’ aretaken to refer to the objects of actual orpossible experiences that we have an em-piricist (in the broad sense) analysis of thecausal modality; and it is only if these ex-periences are taken to be atomistic that wehave a specifically Humean one. But it isonly if ‘x’ and ‘y’ constitute complete as wellas atomistic state-descriptions that the con-


junction between x and y will (supposingregularity determinism to be true) be con-stant. These desiderata can only be recon-ciled if all causal sequences are linear pro-cesses; an idea which it would seem extraor-dinarily difficult to defend. Finally, it shouldbe noted that whereas the constant conjunc-tions are for the most part unknown (as theregularity determinist will be bound to con-cede), the events they are supposed to con-join are regarded on the Humean analysis asbeing intuitively ascertainable!

83:1 I have dwelt at some length on thedifferences between the physical and episte-mological concepts; as well as hinting in thelast paragraph at internal difficulties in thelatter (which will be developed in Chapter3). I now want to say something about thegeneral character of the paradigm as suchand the limitations of its various ingredients.

The essential features of the classical corpus-cularian/mechanical world-view can be sum-marized as follows: –

83:2 (i) the externality of causation;

83:3 (ii) the passivity of matter, and theimmediacy of effects;

83:4 (iii) the atomicity of fundamental en-tities (whether corpuscles, events or sense-data);

83:5 (iv) the absence of internal structureand complexity;

83:6 (v) the absence of pre-formation, andof material continuity;

83:7 (vi) the subjectivity of transfor-mation and of apparent variety in nature(i.e. metaphysically, qualitative diversity andchange are ‘secondary qualities’).

83:8/o It has already been seen that atthe limit conditions for a closure the distinc-tion between intrinsic and extrinsic condi-


tions collapses. For atoms possess no intrin-sic conditions. An atom is distinguishableonly with respect to its position (or somehigher order derivative of position such asmotion, acceleration etc.) in space at a mo-ment in time. Consider now the old mech-anistic prejudice crystallized in Hobbes’ dic-tum that ‘nothing taketh a beginning fromitself’.22 Does this mean that nothing takethany part of its beginning from itself; i.e.that none of the total set of necessary andsufficient conditions for an event is intrin-sic to the thing (whether or not the trig-ger that sets it off, so to speak—what weshould ordinarily call ‘the cause’—is intrin-sic)? If it does, this is tantamount to assum-ing that the event occurs, on the satisfactionof certain antecedent conditions, whateverthe thing. It encourages, as it were, a pic-ture of a cloud of smoke every time we pour

out a drop from the bottle marked ‘ACID’,irrespective of the nature of the substanceor material onto which it is poured; or thesound of breaking glass every time a cricketball hits a stationary target at 30 m.p.h., ir-respective of whether it hits a greenhouse, asand pit or a granite wall. Such ‘explana-tions’ border of course on the fatalistic; orrather we could say that they are fatalisticwith respect to things. For if there were sub-stance or material conditions for the eventsuch that were they not satisfied it would nothave occurred, then they must be included aspart of the total cause (in Mill’s sense) of theevent. Once we allow that an event wouldhave occurred, whatever the intrinsic condi-tions, we are bound to end up denying theprinciple of material continuity. This is theprinciple that events are changes in things,never replacements of one kind of thing by an


entirely new kind of thing.23 But if all causesare extrinsic there can be no material conti-nuity through change: either because thereis no continuity (positivism) or because thereis no change (corpuscularianism).

84:1/o Ordinarily we think of the worldas consisting of things which endure throughsome but not other changes. Events we thinkof primarily as changes in things, i.e. as thetransformation of substances, rather thanthe displacement of physical masses in spaceand time. What is transformed is alreadygiven as complex and pre-formed. If it ispartially transformed, material continuity ispreserved through the change. If it is totallytransformed, we seek for a new kind of sub-stance, or level of ‘thing’, which will allow usto preserve this principle.24 A chemical atompreserves continuity through chemical reac-tions; a gene-pool through species change;

and historically, physicists have tended totreat as ‘substantive’ precisely that whichtends to be conserved, e.g. matter or energy.In our ordinary ascriptions of change then amaterial as well as an efficient cause is nor-mally essential. The classical paradigm di-rects exclusive attention to the latter. Andyet in scientific contexts it is often the searchfor hidden entities, leading to the identifica-tion of novel kinds, set off by what appearsprima facie to be a case of ex nihilo produc-tion, that is most important. (Cf. the dis-covery of the neutrino.) It is the absenceof the notion of material continuity throughchange, as it is manifest in the epistemologi-cal variation of the classical paradigm, whereit results in the generation of an ontology ofatomistic and independent events, that un-derpins the idea of the contingency of thecausal connection, which we turn to in the


next chapter after rejecting the idea of itsactuality here.

85:1 Perhaps under the influence of theclassical paradigm we tend to view eventsas happenings to passive things. But eventsare also the results of actions. Panes of glassdo not shatter without being hit by thingssuch as cricket balls; nothing could be mag-netized unless there were magnets. Thussome things must be agents as well as pa-tients; just as some conditions must be in-trinsic as well as extrinsic. Yet we feel thesource, the stimulus, the trigger is always ex-trinsic. But this is a pure prejudice. For notall efficient causes are extrinsic and not allextrinsic causes are mechanical. Thus thestructure of a field or the organization ofan environment may be the cause of whathappens within it. There is no reason whythe properties of wholes should not explain

those of their component parts. But neitheris there any reason why all things should beconceived as either wholes or parts. I am go-ing to suggest later that we are radically mis-led by spatial metaphor and imagery here.Societies, people and machines are not col-lectivities, wholes or aggregates of simpler orsmaller constituents (just as intentionality isnot an inner urge or push). In the classicalworld view it was the function of matter tooccupy space;25 so it was natural to assumethat all ‘things’ properly so-called were justmore or less highly differentiated aggregatesof matter, and so could be viewed either aswholes or parts (or as both).

85:2/o The victories of the corpuscular-ian/mechanical programme were never ascomplete as its propagandists made out.Neither gravity positing action at a dis-tance nor magnetism in which distance was


a variable in action could be assimilatedto the paradigm. The fact that Newtoncould not find a contact explanation for grav-ity provided the basis for much early criti-cism of his work; and the theory of a con-necting medium—the aether—has its post-relativistic advocates even today.26 Paradox-ically the very fact that Newton could notfind an explanation for gravity strengthenedthe positivist variant of the paradigm—forit gave a degree of credibility to the beliefthat science eschews ‘hypotheses’.27 Never-theless the prestige of Newtonian mechan-ics was such that from Hume’s time on-ward scientific explanation came to be iden-tified with mechanical explanation; and areduction of all other branches of learningto mechanics was loudly proclaimed by Hel-vetius and its other propagandists to be athand. Of course we know that after me-

chanics had been fully developed and ap-plied successfully in domain after domainthere was bound to come a time when, un-der the stimulus of internal inconsistenciesand irrefutable counter-instances, the prin-ciples of mechanics would themselves haveto be explained non-mechanistically. More-over we know as a matter of fact that theNewtonian system has been replaced and notjust subsumed;28 and that in one respect af-ter another the classical world view has beenabandoned by modern physics.29 But giventhis, it is still important to ask (in view ofits perennially attractive features) whetherit provided a logically coherent schema forthe conceiving of fundamental explanationsin physics or indeed any other science.

86:1/o Now both Boscovitch and Kantargued, anticipating later developments inphysics, that contact explanations could not


be fundamental; but that they themselvesrequired explanation in terms of forces—ofattraction and repulsion—acting at a dis-tance.30 For, they argued, since action bycontact proceeds by compression, the bod-ies involved cannot be absolutely rigid. Ifthey are, then there cannot be any trans-fer of motion. For a transfer depends uponthe individuals being deformed to some ex-tent. But basic individuals cannot be de-formed. For to be deformed is to lose aproperty once possessed, but atoms have noproperties to lose. Thus the most funda-mental interactions cannot be by contact be-tween corpuscles or atoms. But could therebe non-mechanical, i.e. non-contact, interac-tions between atoms? If there could, atomsmight still be the ultimate ‘things’. Now anysuch action would of necessity have to be ata distance. But aside from spatio-temporal

location (and its derivatives), the only at-tribute such ‘atoms’ could possess would bethe power of affecting things at a distance.In short ‘atoms’ would be playing a purelynominal role. Such atoms would not be‘atoms’ at all, but potentials or bare pow-ers; that is, point-centres of mutual influencedistributed in space.31

87:1 So far I have not discussed one impor-tant aspect of our normal concept of action;and that is the idea that things possess pow-ers and liabilities32 to do and suffer thingsthat they are not actually doing and suffer-ing and that they may never actually do orsuffer. It remains true to say of a Boeing727 that it can (has the power to) fly 600m.p.h. even if it is safely locked up in itshangar; just as it would remain true to sayof a person that he could (would be liable to)get hurt if he happened to be in the way of


a herd of stampeding buffaloes, even if as amatter of fact we knew he had no intentionof ever going to the Prairies. The elucidationof this concept will occupy us much later on,so I will be brief with it here. It is suffi-cient for our purpose to note that it dependsnot only upon the idea of the complexity andpre-formation of things, which we found tobe necessary to understand change, but uponthe further idea of the stratification of theirproperties.

87:2/o Now if things are at any momentof time complex and pre-formed then it al-ways makes sense to suppose that they mighthave behaved in ways that they did not. Butthe ‘might’ here is susceptible of a purelyconditional analysis, viz. as meaning merelythat they would have behaved in those ways,if the actual conjunction of intrinsic andextrinsic conditions had in fact been dif-

ferent. The potentiality involved remainspurely epistemic: it is still predicated es-sentially of events and only derivatively ofthings. To say that a thing has a powerto do something is, by contrast, to say thatit possesses a structure or is of such a kindthat it would do it, if the appropriate con-ditions obtained. It is to make a claim firstand foremost about the thing; and only sub-sidiarily, if at all, about events. It is to saysomething essentially about what the thingis, and only derivatively about what it willdo. It is to ascribe a natural possibility tothe thing, whose actualization will dependupon the flux of conditions. An old Austin 7can go 105 m.p.h. if it is towed by a Jaguar;but it does not possess the power to do so.And yet, if it did, the two events if conceivedin themselves (without reference to the in-trinsic structure of the thing) would be the


same.

88:1 The ascription of powers differs fromthe simple ascription of complexity to thingsin that it presupposes a non-conventionaldistinction between those properties of thething which are essential to it and thosewhich are not. The essence of hydrogen is itselectronic structure because it is by referenceto it that its powers of chemical reaction areexplained; the essence of money is its func-tion as a medium of exchange because it is byreference to this that e.g. the demand for itis explained. Not all properties of a thing areequally important because it is by referenceto some but not others that its causal pow-ers are explained. In general it is these thatconstitute its identity and allow us to talk ofthe same thing persisting through change.

88:2/o Now Locke was wrong to construethe fundamental distinction on which the

corpuscularian/mechanical world view wasbased, viz. between the manifested qualitiesof things and the configuration and motionof their parts (by which the former were ex-plained), as one between two types of qual-ities. For, in the first place, though someproperties of the explanatory stratum, such asolidity and motion, could plausibly be mod-elled on the properties of observed things(so that one might say their ideas ‘corre-sponded’), others, such as a point mass or africtionless surface, had no such analogues.Thus Locke did not sufficiently appreciatethat their concepts could not be read directlyonto experience, but rather had to be pro-duced as a result of the theoretical work ofscience (when they might subsequently cometo inform and direct experience). Secondly,Locke should have articulated the distinctionas one between the causal powers that mate-


rial things possess in virtue of their hypoth-esized internal structure and the manifesta-tion of those powers, including their mani-festation as the qualities of observed thingsin sense- experience. If he had done this, itwould have obviated the need to find an im-mediate reflection in sense-experience of thehypothesized basic or ‘primary’ properties ofthe world—a move which provided the gristfor Berkeley’s attack on his ontology andhas been the staple diet of phenomenalistsever since. At the same time, it would haveboth left open the possibility of a qualitativeaccount of primary properties and removedthe necessity to view the latter as ultimate.This would also have undermined the basisof his scepticism over a knowledge of realessences. Locke’s pessimism about the possi-bility of future technological and theoreticaladvance is really the inverse of his philosoph-

ical under-statement, implied by his theoryof ideas, of the conceptual advances made bythe corpuscularian/mechanical world-view.For unless a philosophy of science acknowl-edges the existence of a past and the possi-bility of a future, it cannot pay tribute to theachievements of the present.

89:1/o Science has travelled far since Ba-con could take it as axiomatic that ‘natureknows only mechanical causation, to the in-vestigation of which all our efforts shouldbe directed’ and Boyle roundly declare that‘the phenomena of nature are caused bythe local motion of one part of matter hit-ting against another’. Philosophy unfortu-nately has lagged behind. We have seenhow the two basic interactions of the New-tonian system contributed—the one by itssuccess, the other by its failure—to the gripof positivistic epistemology. Paradoxically,


the very breakdown of the system coincidedwith, and indeed contributed to, a revivalof positivistic-sensationalist and operationistthought. For one thing the new ‘fundamen-tal entities’ seemed both event-like and sta-tistical in character. And the apprehensionof scientific change seemed merely to under-line the necessity for a subjectivist ontol-ogy. (Later I shall show how, on the con-trary, scientific change provides in fact thebest possible argument for the objectivityof things.) As for the programme of reduc-tion, historically associated with the corpus-cularian/mechanical world-view, two ques-tions raised by it survive the eclipse of thatworld-view: First, whatever the nature ofthe basic entities postulated by physics atany time, how are we to understand the sta-tus of apparently emergent things and prop-erties? (Are people, for example, metaphys-

ically ‘secondary qualities’?) And how arewe to understand the phenomena of diversityand change? Secondly, what does it meanto say that the subject matter of one sciencecan be ‘reduced to’ (in the sense of ‘explainedin terms of’) that of another? How does sucha reduction proceed? And how, if at all, doesit affect the ontological status of the entitiesof the reduced science?

90:1 Finally, something should be saidabout the significance of the psychologicalstudies of Piaget, Michotte and others forthe classical paradigm.33 These studies showthat we ordinarily experience mechanicalcausation, i.e. the displacement of physicalmasses in space and time, in terms of tran-sitive verbs such as ‘pushing’ and ‘pulling’which cannot be explicated ostensively; butrather embody an intensional relationshipbetween cause and effect. Such verbs take


an objective complement: We understand‘pushing’ as ‘pushing away’, ‘pulling’ as‘pulling towards’. Jack does not just fall,he falls down; and the ‘down’ is an essentialpart of what the ‘falling’ means. Ink bottlesdo not only get knocked, they get knockedover; doors do not only get slammed, theyget slammed shut. And this is how we cometo understand the meaning of these verbs.34

Now it seems to me that this raises pointsagainst Hume, though not against Newton.In the first place, it undermines Hume’s psy-chological account of the genesis of our ideaof ‘connection’. For it is now not regu-lar comcommitances, but completed move-ments, transitively understood, that providethe source of the latter. Secondly, it showsthat the concepts in terms of which scien-tists come to understand motion could nothave been given in or abstracted from sense-

experience (a) because our ordinary conceptscannot be explicated ostensively; (b) be-cause Newtonian ones cannot either; and (c)because the concepts are radically differentanyway. Thirdly, it shows the poverty of theHumean account of our understanding of thebasic interactions of mechanics. Because wehave to ask why for three hundred years sci-entists and philosophers found this paradigmso compelling. And if today we must resistthis compulsion, it is as well to be aware ofthe source of its power.


2.4 Actualism andTranscendental Re-alism: the Inter-pretation of NormicStatements

91:1 In Section 2 the critical conditions fora closure have been developed and in Sec-tion 3 the concept of action implied by themhas been brought out. Now two questions ofgreat significance may be asked of any clo-sure:

91:2 (i) are the conditions for the closureuniversally satisfied or is the antecedent ofthe law-like statement for which the closureis defined instantiated in some open system?

91:3 (ii) were the conditions for the clo-sure artificially produced or did they occur

naturally or spontaneously, i.e. without theactive intervention of men?

91:4/o One could usefully distinguish herebetween ‘universal’ and ‘restricted’ closures;and between ‘artificial’ and ‘spontaneous’ones. Only a universal closure is consistentwith the empiricist concept of a law as a uni-versal empirical regularity. For to say thatthe antecedent is instantiated in an open sys-tem is just to say, according to the criterionof Section 2 above, that given the antecedentthe consequent fails to materialize on at leastone occasion in the space-time region forwhich the system is defined. In general ifa closure has been artificially established itcannot also be universal.35 It is of course pre-cisely the ubiquity of open systems in naturethat makes necessary an experimental ratherthan a merely empirical science. Once this isaccepted, the idea of invariance over space-

2.4. ACTUALISM AND TRANSCENDENTAL REALISM: THE INTERPRETATION OF NORMIC STATEMENTS121

time must give way to the idea of invarianceunder experiment as a criterion of the em-pirical basis of a science. Moreover, strictlyspeaking, the invariance is that of a result,not a regularity. In general the result willbe invariant to space and time, but not overthem. On the other hand, it is clear thatif the notion of laws as universal empiricalregularities is retained, then the same logicthat led to the regress of interactionism willlead to the demand for a closure of all in-teracting systems until—if everything is as-sumed to be in interaction—we have whatmight be called a ‘global’ or ‘Laplacean’ clo-sure. (Such a slide can only be avoided if itis supposed that a non-interacting eternallyclosed system can be found—without this af-fecting anything in the system.)

92:1 Now confronted with the instantia-tion of the antecedents of laws in open sys-

tems, i.e. in systems where their consequentsare not invariably realized, the empiricistmust abandon either the laws or his conceptof them, viz. as universal empirical general-izations. For whatever is empirical must beactual. And in open systems laws if they areto be actual cannot be universal; and if theyare to be universal cannot be actual. So hemust say either that they are not laws; orthat laws are not universal; or that laws arenot empirical. The first position, which maybe characterized as ‘strong actualism, was ineffect adopted by Mill in his doctrine of lawsas ‘unconditional sequences’.36 The troublewith it is that there are no unconditional se-quences known to science. The second po-sition, which may be characterized as ‘weakactualism’, involves restricting the applica-tion of laws to closed systems. This maybe done by making the satisfaction of a ce-


teris paribus clause a condition of the law’sapplicability. The trouble with it is that itleaves unanswered the question of what gov-erns phenomena in open systems. Moreoverit cannot provide a rationale for either theexperimental establishment or the practicalapplication of our knowledge (ironically inview of its sponsorship by self-styled ‘em-piricists’ and ‘pragmatists’). The third po-sition is that of transcendental realism. Itrejects the idea, common to both forms of ac-tualism, that laws are empirical statementsor statements about events. Instead, it re-gards them as normic or transfactual state-ments that apply in open and closed sys-tems alike. On this view, closures are impor-tant in the experimental establishment of ourknowledge. But they do not affect the onto-logical status of laws. On the contrary, thetranscendental realist asserts, it is just be-

cause the things to which laws are ascribedgo on acting in their normal way indepen-dently of whether or not a closure obtainsthat the scientific investigation of nature ispossible.

93:1 The empiricist, when confronted withthe phenomena of open systems, i.e. thenon-availability of universal closures, is facedwith the trilemma of choosing one of theforms of actualism (which involves eitherpreserving his philosophical integrity at theexpense of science or abandoning his in-tegrity to justify science) or succumbing totranscendental realism. My strategy will beto argue that weak actualism is not a gen-uine alternative and if pushed must collapseinto one of the other two.

93:2 One way of describing these optionsis in terms of their different responses tothe identifying mark of an open system, viz.


the non-realisation of the consequent, giventhe instantiation of the antecedent of a law-like statement. For the strong actualist thismeans that the statement must be false, forthe weak actualist it may be inapplicable,for the transcendental realist it can be bothapplicable and true. It must be false for thestrong actualist because a law-like statementasserts the invariance of the conjunction be-tween antecedent and consequent. It may beinapplicable rather than false for the weakactualist, if the ceteris paribus clause, sub-ject to which it is regarded as being formu-lated is not satisfied. It can be both appli-cable and true for the transcendental realist,if it correctly describes the working of a gen-erative mechanism and the mechanism wasreally at work in that instance. Moreoverfor the transcendental realist the statementcan be known to be both applicable and

true, namely if the statement has been in-dependently verified (e.g. under experimen-tally closed conditions) and there is no rea-son to suppose that the nature of the thingpossessing the tendency whose operation isdescribed in the law has changed.

93:3/o The weak actualist is immediatelyfaced with a problem here. For although thelaw-like statement may be inapplicable, viz.if the CP condition was not satisfied, ratherthan false, viz. if it was satisfied, there isno way on actualist lines that he can de-cide between these alternatives. The wayin which this is normally settled is to seeif the consequent is realized; if the conse-quent is not realized this means that the CPclause was not satisfied. But this involvesusing the law (thus presupposing both itstruth and applicability—the latter in virtueof the satisfaction of the explicitly mentioned


antecedent conditions) as a criterion of thestability of the circumambient conditions.Hence any attempt to use the stability ofthe circumambient conditions as a criterionof the applicability of the law is viciouslycircular—as the law’s applicability would bealready presupposed in the test for the sta-bility of the conditions. The situation inwhich the weak actualist finds himself hasbeen expressed as follows: –

94:1 When a prediction turns out to befalse, the situation as regards the generallaws used in making it is indeterminate: itcannot be known with certainty whether oneor all the general laws have been discon-firmed or whether the ceteris paribus con-dition has not been fulfilled.37

94:2 It might be thought that the situationwould be improved if an independent meansof verification for the law was available. But

supposing a law were experimentally verifiedits use in open systems would presuppose ageneral principle sanctioning the applicabil-ity of laws when their consequents were unre-alized. But this is precisely what is in ques-tion here and what both forms of actualismdeny.

94:3 Now the strong actualist can onlyjustify the retention of a law-like statementwhose antecedent is instantiated in an opensystem as a temporary proxy or stand-in forthe yet to be discovered unconditionally uni-versal statement. But can the weak actu-alist do any better? For we are bound toask him: if laws are restricted to closed sys-tems what governs or accounts for phenom-ena in open ones? His options here are lim-ited: either nothing does or something does.The former entails complete indeterminism.But the latter sets the weak actualist on


the road to strong actualism. For to sup-pose that something accounts for the phe-nomena and to hold that current laws are in-applicable only makes sense on the assump-tion that open systems may be eventuallyclosed. So it seems that the incomplete ornon-atomistic descriptions that we currentlycall ‘laws’ must be replaceable in time bycomplete atomistic ones which (given onlythat regularity determinism is true) will afterall be both strictly universal and still empir-ical. So that the weak actualist too comes toregard present ‘laws’ as temporary stand-insfor the Laplacean hour.

94:4/o The trouble with weak actualism isthat it is prepared to acknowledge the fact ofopen systems without generating the meansfor science to cope with it; that is, it is pre-pared to differentiate but not to stratify re-ality, thus removing the possibility that in

our ascription of laws we are referring to away of acting or a level of structure that isnot confined to closed systems. The neces-sity to view the satisfaction of the CP clauseas a condition of a law’s applicability van-ishes once we realize that it is precisely akey function of the concept of law to applytransfactually, in open and closed systemsalike. The satisfaction of the CP clause is,on the other hand, a condition for a deci-sive test situation (its verification dependingnecessarily upon the applicability of ‘auxil-iary’ or bridge laws). But the truth of anynormic statement is in general determinedquite independently of, and antecedently to,its explanatory and other uses in open sys-tems.

95:1 Given only a knowledge that the an-tecedent is instantiated and the absence ofspecific reasons for supposing that the ten-


dency is no longer possessed by the thingwe can then be justifiably sure that the ten-dency is being exercised or as it were in play;although only if we have grounds for suppos-ing the system closed does that certainty li-cense the prediction of its fulfilment. The ci-tation of a law presupposes a claim about theactivity of some mechanism but not aboutthe conditions under which the mechanismoperates and hence not about the results ofits activity, i.e. the actual outcome on anyparticular occasion. This will in general beco-determined by the activity of other mech-anisms too. Indeed it is precisely because itis non-committal about the nature of the cir-cumambient conditions that a statement oflaw does not in general justify a claim aboutevents, let alone experiences.

95:2 Strong actualism regards the appear-ance of open systems as a mark of ignorance

and initiates interactionist and reductionistregresses in an attempt to overcome it. Weakactualism acknowledges the de facto exis-tence of open systems but then proceeds tofence them off from science. For strong andweak actualism alike, open systems fall out-side the pale of science. In this way empiri-cism understates its potential scope of appli-cation. Lacking from both forms of actual-ism is the concept of generative mechanismswhich endure, so that the laws they groundcontinue to prevail, in open and closed sys-tems; so making possible the scientific under-standing of things and structures which ex-ist and act quite independently both of ourdescriptions and the exercise of our causalpowers.

96:1 Braithwaite falls into the same trapas weak actualism by arguing that a ten-dency statement is a conditional with an un-


specified antecedent.38 For if it is unspecifiedwe cannot know when to apply it. It is in factvital to distinguish the explicit conditions inthe protasis of the law-like statement fromthe unknown conditions that the CP clausemay be required to cover. The satisfactionof the former is a condition for the appli-cability of a law. But neither a knowledge(strong actualism) nor the stability (weak ac-tualism) of the latter can be a condition forthe applicability of a law. There are threereasons for this. First, it is in principle im-possible to specify all the conditions that theCP clause may be required to cover. Indeed,if one could do so, there would be no needfor the CP clause in the first place. Sec-ond, as has been seen, the satisfaction of theCP clause cannot normally be verified inde-pendently of the actualization of the conse-quent; hence to make it a condition for the

applicability of the law is circular. Thirdly,as the satisfaction of the CP clause is time-dependent (being trivially satisfied instanta-neously), acceptance of it as a condition forthe law’s applicability generates absurd andtotally counter-intuitive results. For exam-ple, on it a law may be applicable for everyfive-minute interval in a day, but not for theday overall. The proper place of the phrase‘other things being equal’ is not as part ofthe protasis but at the tail-end of the state-ment as a reminder that, because the sys-tem in which the thing’s behaviour occursmay not be closed, the tendency postulatedin the statement may not be actualized.

96:2/o Satisfaction of the CP clause is nota condition for the applicability of a law. Itis, however, a condition for the actualiza-tion of the tendency designated in the state-ment (for which it is sufficient, although not


strictly necessary). And from this are de-rived its two main roles: first and foremost,as a signal of the normic nature of the propo-sition being expressed, as a reminder thatthe tendency designated may not be actu-alized; and secondly and derivatively, as awarning to historicists and pseudo-falsifiers,cautioning the former that the prediction ofthe tendency is not deductively justified andthe latter that if the tendency is unfulfilledthe statement should not – on that groundalone – be held to have been falsified. Thusthe CP clause does not place a condition onexplanation, for one can explain an event interms of tendencies when the latter are neverrealized. Rather it places a condition on pre-diction and falsification.

97:1 This account needs qualifying in twoways. First, if we distinguish between theconstancy of intrinsic and extrinsic condi-

tions (as suggested in Section 2 above) andbetween the constancy of more and less im-portant intrinsic ones (as suggested in Sec-tion 3) then the constancy of intrinsic struc-ture is a condition for the applicability ofa law. Tendencies are only possessed, andhence can only be exercised, as long as thenature of their possessor remains unchanged.But this does not vitiate my account. Forlaw-like behaviour is predicated essentially ofthings, which are typically referred to in theprotasis. There is a real asymmetry, whichis reflected in the structure of law-like state-ments, between the intrinsic structure or es-sential nature of a thing (which in generalconstitutes its identity or fixes it in its kind)and the conditions under which it acts inthat a change in the former but not the latterleads to a change in the thing’s tendencies,


liabilities and powers.

97:2 Secondly, I have said that the CPclause functions as a reminder and a warn-ing. But such reminders and warnings areonly necessary as long as law-like statementscontinue to be formulated and thought of inthe actualist mode. If there were no histori-cists or pseudo-falsifiers there would be noneed for reminders to them. Hence a fullyrealist philosophy of science could in princi-ple dispense entirely with the CP clause (atleast in this aspect of its work).39 For what-ever is conveyed by ‘This happens CP’ can beequally well conveyed by ‘This tends to hap-pen’. (To add CP to this statement would beto qualify the tendency, not its fulfilment.)This is not a shallow, equivocal, sloppy ormean formulation; but the logical form of allthe laws of nature known to science.

97:3 I want to turn now to consider in

more detail the character of normic state-ments. A full analysis of the logic of ten-dency statements must, however, be post-poned until Chapter 3.

97:4/o On the view of science advancedhere, power and tendency statements arecategorical rather than, as maintained byHume and Ryle, hypothetical. Hypotheti-cals provide the empirical grounds for our as-criptions of powers and tendencies, but theydo not capture their meaning. Tendenciesare roughly powers which may be exercisedunfulfilled. They are thus well adjusted tocope with open systems. If a system is closedthen a tendency once set in motion must befulfilled. If the system is open this may nothappen due to the presence of ‘offsetting fac-tors’ or ‘countervailing causes’. But theremust be a reason why, once a tendency isset in motion, it is not fulfilled; in a sense in


which it would be dogmatic to postulate thatthere must be a reason why the tendency isset in motion. Once a tendency is set in mo-tion it is fulfilled unless it is prevented.

98:1 The following is my interpretationof the mode of application of lawlike state-ments. Such statements, when their ini-tial conditions are satisfied, make a claimabout the activity of a tendency, i.e. aboutthe operation of the generative mechanismthat would, if undisturbed, result in the ten-dency’s manifestation; but not about theconditions in which the tendency is exer-cised and hence not about whether it willbe realized or prevented. Because the op-eration of the generative mechanism doesnot depend upon the closure or otherwiseof the system in which the mechanism op-erates, the mode of application of law-likestatements is the same in open and closed

systems; what does differ is the inferencethat can be drawn from our knowledge ofthe applicability of the statements in the twocases. Notice that although the applicationof a normic statement warrants a subjunc-tive conditional about what would have hap-pened if the system were to have been closed,the full force of its meaning cannot be un-derstood or captured in this way. It has tobe interpreted categorically and indicativelyto the effect that a generative mechanismwas really at work; which helps to accountfor, though it does not completely determine,whatever actually happened.

98:2/o The ‘thing’ which possesses thetendency is not necessarily the same ‘thing’as that whose behaviour is recorded in thelaw-like statement. Indeed it is character-istic of science to postulate novel entitiesas the bearers of the tendencies and powers


manifest in the behaviour of observed things.The class of ‘things’ is far wider than that of‘material objects’: it includes fluids, gases,electronic structures, fields of potentials, ge-netic codes, etc.; so we must try to divest theconcept of its normal material object conno-tations. The idea of a tendency exercisedunfulfilled seems strange if we think of or-dinary material objects such as tables andchairs. People provide in this respect a bet-ter model for the entities discovered and in-vestigated by science. There is nothing mys-terious about tendency ascriptions to people.We know what it is like to be in a situa-tion where we tend to lose our patience ortemper and we know what it is like keepingit. Tendencies exercised unfulfilled; shown,perhaps, but unrealized in virtue of our self-control.

99:1/o Now when a tendency is exercised

unfulfilled two things are not in doubt: (a)that something actually happens, towardsexplaining which the exercise of the tendencygoes some way; and (b) that something is re-ally going on, i.e. there is a real generativemechanism at work, which accounts for theinfluence of the factor the tendency repre-sents in the generation of the event. In thecase of (a) there are two conceptual traps.The first is to think of the exercise of thetendency unfulfilled as an action without re-sults, rather than as an action with modifiedresults. Something does happen; and thetendency, as one of the influences at work,helps to explain what. The second is tothink of it as if it were an action fulfilled,i.e. in terms of its fulfilment. It is a mis-take to think of the exercise of a tendencyin terms of the imagery, metaphors or de-scriptions appropriate to its fulfilment. Yet


Mill in his unofficial doctrine of tendenciesin effect does this when he argues that ‘al-though two or more laws interfere with oneanother, and apparently frustrate or modifyone another’s operations, yet in reality allare fulfilled, the collective effect being theexact sum of the causes taken separately’.40

Mill’s mistake here is to suppose that when-ever a tendency is set in motion the effectmust be in some sense (or in some realm) oc-curring (as if every time we ran fast we hadto be in some way winning). But Geach (andfollowing him Ryan) in ridiculing this posi-tion make the converse mistake of supposingthat whenever no effect (of a given type) oc-curs, nothing can be in motion or really go-ing on.41 But here Mill is right and Geach iswrong. Balaam’s ass is pulled in two ways;we do just manage to keep our tempers; themarket equilibrium is explained in terms of

an exact balance of buying and selling; whenthe beam finally collapses it is due to thereal cumulative effect of the woodrot. Mill’smistake is to think of the exercise of the ten-dency under the description of its fulfilment,as if Balaam’s ass, in order to be pulled twoways, had actually to go in both directions.Geach’s mistake is to suppose that becauseneither tendency is fulfilled neither tendencycan be in play. In other words, they bothmake the mistake of seeing the fulfillment ofa tendency a condition of its exercise.

100:1 Let me stress that the scientist’s sit-uation is such that he is never in any doubtthat the given an effect something is produc-ing it; his doubts is only over what is. Nowclearly this does not mean that he is commit-ted to a realist interpretation of every theory;what it does mean is that as a theorist histask remains essentially incomplete until he


has produced a theory which correctly de-scribes the mechanisms by means of whichthe effect in question is produced. It is inthis light that other possible interpretationsof normic statements must be considered.

100:2/o It is misleading to think of normicstatements as ‘idealizations’ or ‘abstrac-tions’. For both concepts conceal a crucialambiguity as to the object idealized or ab-stracted from, in which the superior realityof events or experiences is tacitly assumed.The conception of the generative mechanismor structure that backs a normic statementneed not be ‘idealized’ or ‘abstract’ in rela-tion to really existing or the reality of exist-ing structures. Once the necessity for a re-definition of the objects of a science as struc-tures rather than events is accepted then theconcept of an idealization must be used inrelation to the reality of that object. And

it cannot be assumed that all theoreticalstatements are idealizations in this sense. Amodel of the intrinsic structure of an atomor a DNA molecule or the solar system isnot necessarily more perfect than the intrin-sic structure of a real atom, DNA moleculeor solar system. The standard of perfectionis not set by men. Of course if one takes ‘the-oretical’ as a synonym for ‘unreal’ (or at anyrate ‘less real’) normic statements will ap-pear as ‘ideal’ in that the tendency they des-ignate or mechanism they describe is rarelyif ever manifest in unmodified form;42 andas ‘abstract’ in that they select from whatis in open systems a mesh of influences andcross-influences just one as the focus of at-tention.43 But to think like this is to fallinto the error of supposing that events aremore real than the structures and mecha-


nisms that generate them.

101:1 Scriven makes a similar mistakein contending that normic statements are‘guarded generalizations’.44 The only thingone need be ‘guarded’ about in using anormic statement is the assumption that thetendency whose activity is designated in thenormic statement will be realized. If suchstatements have been independently andwell confirmed (under experimentally closedconditions) then we may be completely andrationally confident in using them. Such con-fidence is expressed in, rather than weak-ened by, our willingness to use the CP clauseagainst naive actualist objections on theirbehalf. It is only if one tacitly views law-likestatements as in the final analysis empiricalgeneralizations that one will feel that (be-cause in asserting a law-like statement one isasserting the realization of the consequent),

if one cannot be sure of the realization of theconsequent then one can only assert the law-like statement ‘guardedly’. But of course inasserting a normic statement one is not as-serting the realization of the consequent; butthe operation of a mechanism irrespective ofits results (which it is precisely the functionof the normic statement to be non-committalabout).

101:2/o Both these ideas depend upon animplicit recognition that reality is differen-tiated in a way that classical empiricism ig-nores and so requires something more of sci-ence than it provides. But the possibilityopened up by this recognition is constrainedby a continuing commitment to empirical re-alism. It is this which prevents the acknowl-edgement that reality is not only differenti-ated but stratified too. Once the stratifica-tion of the world is grasped it is possible to


see how our knowledge can be both univer-sally applicable and rarely (empirically) in-stantiated; and so to resolve Poncare’s prob-lem that ‘on the one hand, [laws] are truthsfounded on experiment and approximatelyverified so far as concerns isolated systems.On the other hand, they are postulates ap-plicable to the totality of the universe andregarded as rigorously true’.45 Normic state-ments speak of structures not events, thegenerator not the generated. In assertinga normic statement one is not making aguarded or idealized statement about an em-pirical reality. Rather one is making a state-ment, which may be ‘guarded’ or ‘idealized’in its own right, about a different level of re-ality. Normic statements are not second bestkind of empirical generalizations. They arenot empirical statements at all.

102:1 Two further misinterpretations of

normic statements must be guarded against.Normic (or transfactual) statements are notcounterfactual statements. They legitimatethe latter; and, like them, are only validat-able in relation to an antecedently and inde-pendently established body of theory. Butwhereas to say that a statement is a counter-factual is just to say that the conditions spec-ified in the antecedent do not obtain; in thecase of a normic statement these conditionsmay obtain, and if they do (and the state-ment has been independently verified) it canthen be interpreted quite straightforwardlyas a statement about what is really goingon though in a perhaps unmanifest way. (Inthe case of counterfactuals antecedents areby definition unsatisfied; in the case of trans-factuals it is contingent whether consequentsare realized.) It is only if the CP clause is re-garded as a component of the protasis that


it is plausible to interpret a normic state-ment as making, in its open systemic uses,a counter-factual claim. This is a postion,most naturally associated with weak actu-alism, that has been argued against above.Normic statements have also sometimes beenjustified as ‘averages’ or ‘rough approxima-tions’; or alternatively as elliptical probabil-ity statements. Both ideas involve a con-fusion of epistemic and natural possibility.For, on the one hand, I may be quite cer-tain about the activity of a natural mecha-nism on a particular occasion but incapableof any judgement about the outcome; and,on the other, I may be sure that some ruleof thumb will hold though quite uncertainabout the reasons why.

102:2/o I have argued that in open sys-tems consequents may be unrealized but thatdespite this we may know that a law is ap-

plicable (a mechanism is at work) if we knowthat its antecedent has been instantiated andit has been in dependently verified. But bothantecedents and consequents are events inopen systems. Is there not an asymmetryhere? Am I not placing a higher demandon antecedents than consequents? Ontologi-cally no; but epistemically yes. For a mech-anism may be set in motion and because ofthe complexity or opacity of the conditionsunder which this happens the describer maynot know that it has been set in motion; sothat a fortiori he cannot know that the law itgrounds is applicable. To explain an event byinvoking a law I must have grounds for sup-posing that a mechanism is at work; but themechanism may be at work, given that itsstimulus and other conditions are satisfied,without my knowing it. Some fields may be


incapable of detection.

103:1 In section 2 the critical conditionsfor a closure were developed and in section3 the concept of action implied by them wasbrought out. In both cases their restricted-ness was noted. In this section a realist ac-count of laws has been counterposed to theactualist account and its superiority clearlydemonstrated. Once we are persuaded of thevery special conditions presupposed by actu-alism and the possibility of an alternative,what havoc must we make of the doctrinesof orthodox philosophy of science?

103:2 In nature, constant conjunctions arethe rare exception; not, as supposed by ac-tualism, the universal rule. And in generalit requires human activity to generate them.To invoke a law I must have grounds forsupposing that the antecedent conditions aresatisfied, so that the mechanism designated

is active. But it is only if I have groundsfor supposing that the system in which themechanism acts is closed that the predictionof the consequent event is deductively justi-fied. With this in mind let us return to thetheories expressed in statements (i)-(v) onpages 634 above. It is only under conditionsof a closure that given the antecedent, thededuction of the consequent event is possi-ble, so that the conditions for the Popper-Hempel theory of explanation are satisfied(ii) or those for the symmetry between ‘ex-planation’ and p‘rediction’ obtain (iii). Itis only then that ex ante criteria of refu-tation can be laid down for a theory (iv)or that it makes sense to judge a theoryby its predictive success (v). For it is onlythen that the resemblances and sequencesbetween phenomena, that Mill identified and


so confused with laws, are constant (i).

103:3/o It is contingent whether some en-during thing or mechanism is activated. Andthough, given this, it is necessary that a cer-tain tendency should be ‘in play’, it is con-tingent, upon the occurrence of a closure,whether the consequent of the law-like state-ment is realized. In short, to know that lawis effective I do not need to be in a positionto predict any event (and, it might be added,vice versa).

104:1 Now once we have grasped the ubiq-uity of open systems in nature we will be in abetter position to understand the embarrass-ment with which textbooks in the philosophyof science gloss over their failure to producea single law or explanation which satisfy thecriteria they so laboriously develop and de-fend; a fact which bears eloquent witness tothe non-ability of universal closures of any

epistemic significance. We will also be in abetter position to understand not just thisfailure, but their absurdity, when they seekto apply these same criteria to fields such ashistory and the human sciences, where theconditions for even a restricted closure (of anon-trivial kind) are not naturally and can-not be experimentally satisfied, and wherethe concept of action implied by these crite-ria is patently inapplicable.

104:2 For a closure one each of the sys-tem, individual and organizational condi-tions must be satisfied. Reflection on theconditions set out on page 76 above and theconcept of action implied by them (see (i)-(vi) on page 83) shows the patent absurdityof trying to apply the constant conjunctionformula to the domain of social life. Con-sider the conditions for a closure as appliedto e.g. the category of persons. Remem-


ber that people are individuals, which meansthat they are complexly structured and pre-formed in different ways, so that they willrespond differently in the same external cir-cumstances (i.e. to the same stimulus). Re-member too that they are subject to a con-tinuing flow of contingencies, none of whichcan be predicted with deductive certainty.And, without calling into question the ap-plicability of the classical paradigm (with itsassumption that the stimulus conditions foraction are always extrinsic), that they are en-gaged in activities such as writing and cook-ing, bar billiards and chess, which cannot beplausibly analysed in terms of atomistic com-ponents. In short, where the subjects, condi-tions or forms of action are characterized bystructure, diversity or change, the Humeantheory of the actuality of causal laws, andipso facto the theories of science that are

based on it, just cannot apply.

104:3/o Conversely, it is just because thevery special conditions for a closure aresometimes satisfied in physics and chemistry(though they are not normally possible inthe other natural sciences—from cosmologyto biology) that accounts for the prima fa-cie plausibility of these theories there. Butthe transcendental analysis of experience al-low us to turn the tables on actualism andempiricism here. For it is not given conjunc-tions of events (or experiences) but struc-tures which are normally out of phase withthe patterns of events (and experiences) thatemerge from it as the true objects of scien-tific understanding. This raises the questionof whether there are analogous structures atwork in fields other than physics and chem-istry. If there are, we must bear in mindthat it would not even be plausible to mis-


construe them as empirical generalizations.On the other hand, if we continue to confuselaws and empirical generalizations we shallnever be able to identify them.

2.5 Autonomy and Re-duction

105:1 Laws we already know do not describethe patterns of events. But how do theystand to the world of our everyday actionand of perceived things?

105:2 Reflect for a moment on the world aswe know it. It seems to be a world in whichall manner of things happen and are done,which we are capable of explaining in vari-ous ways, and yet for which a deductively-justified prediction is seldom, if ever, possi-ble. It seems, on the face of it at least, to be

an incompletely described world of agents. Aworld of winds and seas, in which ink bottlesget knocked over and doors pushed open, inwhich dogs bark and children play; a criss-cross world of zebras and zebra-crossings,cricket matches and games of chess, mete-orites and logic classes, assembly lines anddeep sea turtles, soil erosion and river banksbursting. Now none of this is described byany laws of nature. More shockingly perhapsnone of it seems even governed by them. Itis true that the path of my pen does not vio-late any laws of physics. But it is not deter-mined by any either. Laws do not describethe patterns or legitimate the predictions ofkinds of events. Rather it seems they mustbe conceived, at least as regards the ordinarythings of the world, as situating limits andimposing constraints on the types of action

2.5. AUTONOMY AND REDUCTION 141

possible for a given kind of thing.

106:1 Laws then not only predicate ten-dencies (which when exercised constitute thenormic behaviour) of novel kinds (or of famil-iar things in novel or limit situations); theyimpose (more or less absolute) constraintson familiar things. In this section I wantto reconcile these aspects of laws by argu-ing that familiar things are comprehensiveentities which may be controlled by (or sub-ject to the control of) several different prin-ciples at once; and that they may be said tobe agents. Laws ascribe possibilities whichmay not be realized and impose necessitieswhich constrain but do not determine; theyascribe the former to novel kinds and imposethe latter on familiar things. These featurescannot be explained away as an imperfectionof knowledge; but must be seen as rooted inthe nature of our world. They are therefore

inconsistent with the thesis of regularity de-terminism which underpins the doctrine ofthe actuality of causal laws, and to which Imust now return.

106:2/o So far I have discussed regular-ity determinism merely as an epistemologi-cal thesis to the effect that our knowledge ofthe world can be cast in a certain form. Butof course this presupposes that the world issuch that our knowledge of it can be castin that form. To deal with regularity deter-minism I must thus draw out this ontologi-cal presupposition; i.e. cast the thesis itselfin ontological form. The main work for thishas already been done. For I have alreadyshown in sections 2 and 3 that regularity de-terminism makes a claim about what wouldhappen (and the way it would happen) if cer-tain highly restrictive conditions were sat-isfied. These were, it will be remembered,


conditions such that if we knew they weresatisfied and the constant conjunction for-mula was not vindicated, the regularity de-terminist would be bound to admit his thesisrefuted. Now regularity determinism’s on-tological claim is simply that the world issuch that these conditions are satisfied andhis thesis is not refuted. Now of course be-cause we can never know that these condi-tions are satisfied we can never refute regu-larity determinism in this way. But I havealso asserted that regularity determinism ismetaphysically refutable. How can this bedone? In the only way open to transcenden-tal realism: that is by showing that if theworld were as claimed by regularity deter-minism science would be impossible. But asscience is possible (which we know, becauseas a matter of fact it occurs) the world mustbe such that either the critical conditions are

not satisfied and/or the constant conjunctionformula is abrogated. In short, the ontolog-ical untruth of regularity determinism is acondition of the possibility of science.

107:1 Close to the appeal of determinismlies the following error: to think that be-cause something happened and because itwas caused to happen, it had to happenbefore it was caused. Now if we take de-terminism to assert that all events are de-termined before they happen and conceivetheir determination as lying in the satisfac-tion of antecedent sufficient conditions forthem then we have a picture of a chain ofantecedent sufficient conditions for eventsstretching back infinitely into the past (as-suming that conditions can be analysed asevents or vice-versa). So if we ask how longis an event determined before it actually hap-pens the answer must be at any (i.e. at ev-


ery) time before it happens. And so if wenow take cause in the ordinary sense, wehave the result that every event is deter-mined before it was caused (or made) to hap-pen. At play here are of course two conceptsof cause: qua causal agent (cause1) and quaantecedent condition (cause2). I am goingto argue that the former is irreducible tothe latter and essential to science. To saythat something is determined before it hasbeen caused to happen is either to say thatit can be known before it has been caused1

to happen (epistemic determinism) or that ithas been caused2 before it has been caused1

to happen (ontological determinism). Theformer depends upon a closure, the latterdepends upon the critical conditions for itbeing satisfied. Now I want to argue thatat any (and every) time the world consistsof things which are already complexly struc-

tured and pre-formed wholes; which may besimultaneously constituted at different lev-els and simultaneously controlled by differ-ent principles. It is because things cannot bereduced to the conditions of their formationthat events are not determined before theyare caused to happen. This fact accountsfor both the temporal asymmetry of causesand effects and the irreversibility of causalprocesses in time. And it is because thingscannot be reduced to atomistic componentsthat when events are caused to happen it isby the thing which acts (i.e. the agent), theevent being produced in the circumstancesthat prevail.

107:2/o Now I want to argue that deter-minism is ontologically false (it is not truethat events are determined before they arecaused to happen, whether in a regularly re-curring or non-recurring way) and epistem-


ically vacuous (there are no significant de-scriptions that satisfy the formula of regu-larity determinism). This has the method-ological corollary that the search for such de-scriptions is likely to be unrewarding. (Andhere once again it is necessary to counter-pose the investigation of complex preformedthings to the search for the complete atom-istic state-descriptions that it is supposedwould enable us to predict their behaviour.)The only sense in which science presupposes‘determinism’ is the sense in which it pre-supposes the ubiquity of causes1 and hencethe possibility of explanations. And the onlysense in which it presupposes ‘regularity de-terminism’ is the sense in which it presup-poses the ubiquity of causes1 for differencesand hence the possibility of their explana-tion. But it is probably better not to use ‘de-terminism’ in this way (nb. cause1 is not the

same as cause2). Now any refutation of reg-ularity determinism as an ontological thesismust depend upon establishing the auton-omy of things, in the sense of the impossi-bility of carrying out the reductions implicitin the vital conditions B1 and C1 of Table2.1 on page 76 (their being a clear asymme-try, for the realist, between the subjects andthe condition of action, and the constancyalternative being recessive). It is here thatI will pitch my attack. Thus I am not go-ing to argue that if the critical conditionswere satisfied the constant conjunction for-mula would not be vindicated. Rather, I amgoing to argue that the critical conditionscould not be satisfied in any world contain-ing science. The question of whether or nothistory would repeat itself is one that neednot detain us here. A nagging doubt mayremain: surely, it might be felt, in the (very)


last instance regularity determinism must betrue. But this is not so. For once we haveestablished an ontology of structures thereis no earthly reason why events should [haveto] be constantly conjoined. There are in-deed principles of indifference (as we shallsee in Chapter 3). But they do not apply,nor is there any reason why they should, toevents, states-of-affairs and the like.

108:1/o In establishing the autonomy ofthings I will follow the normal procedure oftranscendental realism; that is, I will firstanalyse some more or less underanalysed fea-ture of science and then ask what the worldmust be like for this feature to be possible.The feature I am concerned with are two as-pects of scientific laws, viz: –

109:1 (i) their normic and non-empiricalcharacter; and

109:2 (ii) their consistency with situations

of dual (and multiple) control.

109:3 I will argue that for these featuresto be possible the world must be composedof agents. Agents are particulars which arethe centres of powers. In an incompletely de-scribed world of other agents powers must beanalysed as tendencies. And laws are noth-ing but the tendencies or ways of acting ofkinds of thing. By an agent 1 mean sim-ply anything which is capable of bringingabout a change in something (including it-self). A hydrogen atom is, in virtue of itselectronic structure, an agent. For it pos-sesses the power to combine with an atomof chlorine to produce, under suitable con-ditions, a molecule of hydrochloric acid. Itshould perhaps be said at the outset that Iam not going to refer to quantum mechan-ics in my argument. It seems to me to bealways a mistake, in philosophy, to argue


from the current state of a science (and espe-cially physics). In general, I have refrainedfrom scoring points against determinism andactualism which turn on the inaccuracy (orimprecision) of our descriptions or the inde-terminacy of our measures. This is becausethey do not in general raise important on-tological questions. It is debatable whetherquantum mechanics does—but if it in factrequires a reinterpretation of the category ofcausality in fundamental physics it will notbe in the Humean direction and can onlystrengthen the anti-determinist’s hand.

109:4/o I have already discussed (i) atsome length so I will be brief with it here.Contrast the law of conservation of energy orof mass action with a simple empirical gener-alization like ‘all pillar-boxes are red’ or ‘allblue-eyed white tom cats are deaf’. Whereasthe latter, at least so long as they remain

unattached to any theory, could be defeatedby a single counter-instance, the truth of theformer is consistent with almost anythingthat might happen in the world of materialobjects and human beings. For they do notattempt to describe this world; i.e. they can-not be interpreted as undifferentiated em-pirical generalizations. Rather they mustbe interpreted as principles of theories—ofphysics and chemistry—which tell us some-thing about the way things act and interactin the world. As such they specify condi-tions which we presume are not contravenedbut rather continually satisfied in the count-less different actions and interactions of theworld, including those of which we have di-rect experience. And they are manifest incertain impossibilities, e.g. that of buildinga perpetual motion machine. Neverthelessthey are principles for which any test would


require not only fine measurement but closedconditions. As such they are not normallyempirically manifest to us or actually sat-isfied. (For the scientist this feature ap-pears as a difference between the real or cor-rected and the actual or measured values ofthe variables he is concerned with.) Thuswe could say that relative to these vantagepoints, viz. of experience and actuality, theseprinciples specify levels of deep structure or(metaphorically) place conditions on the in-ner workings of the world.

110:1 Now it might be said that laws, suchas those of mechanics or electricity, do notdescribe the world as such, but only those as-pects or parts of it which fall within their do-main, i.e. the mechanical or electrical aspectsof it. But this concedes my point. For onecan only say which aspects are mechanicalor electrical by reference to the antecedently

established laws of mechanics and electric-ity, and such aspects are real. Clockworksoldiers and robots do not more nearly ob-serve the laws of mechanics than real peo-ple. Rather their peculiarity stems from thefact that if wound up and left alone their in-trinsic structure ensures that for each set ofantecedent conditions only one result is pos-sible. But outside the domain of a closurethe laws of mechanics are, as Anscombe hasput it, rather like the rules of chess; the playis seldom determined, though nobody breaksthe rules’.46

110:2/o Closely connected with this fea-ture of laws is their consistency with situ-ations of ‘dual control’. A game of cricketis only partially controlled by the rules ofcricket, language-using by those of gram-mar. Chemical reactions are only partiallycontrolled by Dulong and Petit’s law, black


bodies behave in all kinds of ways that arenot specified by the Stefan-Boltzmann law.Coulomb’s law does not completely describethe action of charged particles, or Fara-day’s law all that happens to an electrode.Similarly the ‘boundary conditions’ for thelaws of mechanics, the domain within whichthey apply, are controlled by the operatingprinciples defining a machine.47 Laws leavethe field of the ordinary phenomena of lifeat least partially open. They impose con-straints on the type of action possible for agiven kind of thing. But they do not saywhich out of the possible actions will actu-ally be performed. They situate limits butdo not dictate what happens within them. Inshort, there is a distance between the laws ofscience and the ordinary phenomena of theworld, including the phenomena of our ac-tual and possible experience. And it is with

the investigation of this distance that I amhere concerned.

111:1 To say that laws situate limits butdo not dictate what happens within themdoes not mean that it is not possible to com-pletely explain what happens within them.The question ‘how is constraint without de-termination possible?’ is equivalent to thequestion how ‘can a thing, event or processbe controlled by several different kinds ofprinciple at once?’ To completely accountfor an event would be to describe all thedifferent principles involved in its genera-tion. A complete explanation in this senseis clearly a limit concept. In an historicalexplanation of an event, for example, we arenot normally interested in (or capable of giv-ing an account of) its physical structure.

111:2 In deciding to write ‘!’ on this pieceof paper I select the conditions under which


the laws of physiology and physics are to ap-ply. So that it is absurd to hold that the lat-ter might account for my ‘!’; or that it mighthave been predicted in the basis of a knowl-edge of a physical state-description prior tomy writing it. On the other hand my neuro-physiological state and the physical condi-tions must be such that I can write it; theycould prevent it (e.g. if I were suddenly tofall asleep or be propelled into orbit aroundthe moon). There is a space between thelaws of physics and physiology and what Ido within which deliberation, choice and vol-untary behaviour have room to apply. Thetheory of complex determination, in situat-ing persons as comprehensive entities whosebehaviour is subject to the control of severaldifferent principles at once, allows the possi-bility of genuine self-determination (subjectto constraints) and the special power of act-

ing in accordance with a plan or in the lightof reasons.

112:1 Human freedom, on this view, if itexists, would not be something that some-how cheats science (as it is normally con-ceived) or, on the other hand, somethingthat belongs in a realm apart from science;but something whose basis would have to bescientifically understood. As freedom wouldbe ana1ysed as a power of men and scienceis, for us, non-predictive there is nothing in-consistent or absurd about such an asser-tion; any more than to say that purposeful-ness in animals, which is no doubt not thesame as intentionality in men, has (still) tobe scientifically understood. I suggest thatonly the theory of complex determination iscompatible with agency; and that there areno grounds for assimilating intentional ac-tion to the classical paradigm or supposing


that intentionality is not a real attribute ofmen. However, this is peripheral to my mainconcerns here. Dogs cannot fly or turn intostones, but they can move about the worldand bark in all kinds of ways. To deny thelatter possibility is as absurd as to deny theformer necessity. But the reasons why theybehave in canine ways is an open questionfor a putative science of animal ethology toanswer.

112:2 The difference between laws of na-ture and empirical generalizations is anal-ogous to the difference between the rulesof cricket and a television recording of theactual play on some particular occasion.Whether or not Boycott scores a century isnot determined by the rules of cricket; butby how he bats and how the opposition play.Now it is clearly necessary for the intelligi-bility of the idea of dual (or multiple) control

that the higher-order level is open with re-spect to, in the special sense of irreducible to,the principles and descriptions of the lower-order level. It is easy to see why this mustbe so. For it is the operations of the higher-order level that control the boundary con-ditions of the lower-order level, and so de-termine the conditions under which the lawsof that level apply. It is the state of theweather that determines, in England, whenand where the rules of cricket can apply; thestate of the conversation that determines theways in which we can express ourselves inspeech; the state of the market that deter-mines the use of machines, the use of ma-chines that determines the conditions underwhich certain physical laws apply. The useof machines is thus subject to dual control:by the laws of mechanics and those of eco-nomics. But it is the latter that determine


the boundary conditions of the former.

113:1 It follows from this that the opera-tions of the higher level cannot be accountedfor solely by the laws governing the lower-order level in which we might say the higher-order level is ‘rooted’ and from which wemight say it was ‘emergent’. Now an his-torical explanation of how a new level cameto be formed would not, it is important tosee, undermine this principle. Let us supposethat we could explain the emergence of or-ganic life in terms of the physical and chemi-cal elements out of which organic things wereformed and perhaps even reproduce this pro-cess in the laboratory. Now would biologistslose their object of inquiry? Would livingthings cease to be real? Our apprehension ofthem unmasked as an illusion? No, for in asmuch as living things were capable of actingback on the materials out of which they were

formed, biology would not be otiose. For aknowledge of biological structures and prin-ciples would still be necessary to account forany determinate state of the physical world.Whatever is capable of producing a physicaleffect is real and a proper object of scien-tific study. It would be the task of biolo-gists to investigate the causal powers of liv-ing things in virtue of the exercise of whichinter alia they brought about various deter-minate states of the physical world. Livingcreatures qua causal agents determine theconditions under which physical laws apply;they cannot therefore already be manifest inthe latter. Sentience determines the condi-tions of applicability of physical laws, butit is also subject to them. If the elementsof the lower-order are real then so must bethe causes that determine the conditions oftheir operation, i.e. the comprehensive enti-


ties formed out of them. If black bodies arereal then so are physicists, if charged parti-cles are real then so are thunderstorms. Inshort, emergence is an irreducible feature ofour world, i.e. it has an irreducibly ontolog-ical character.

113:2/o Reflect once more on the distinc-tiveness of laws of nature and empirical gen-eralizations. The laws of nature leave theconditions under which they operate open,so the field of phenomena is not closed: it issubject to the possibility of dual and mul-tiple control, including control by humanagents. What I can do is constrained by theoperation of natural laws. But I can hackmy way all over the physical world, defeat-ing empirical generalizations. I can interruptthe operations or break the mechanism of amachine and so falsify any prediction madeon the basis of its past behaviour. But I can-

not change the laws that governed and soexplained its mode of operation. And I cancome, in science, to have a knowledge of suchnormic and non-empirical statements; andperhaps in time begin to recognize analogousprinciples at work controlling my own be-haviour (marking the site of a possible psy-chology).

114:1 I have argued that complex objectsare real (and that the complexity of ob-jects is real); and that the concept of theiragency is irreducible. Complex objects arereal because they are causal agents capableof acting back on the materials out of whichthey are formed. Thus the behaviour of e.g.animate things is not determined by phys-ical laws alone. But that does not meanthat their behaviour is not completely de-termined: only that an area of autonomy ismarked out which is the site of a putatively


independent science. And because the formsof determination need not fall under the clas-sical paradigm this in turn situates the pos-sibility of various kinds of self-determination(including the possibility that the behaviourof men may be governed by rational princi-ples of action).

114:2 From the normic and non-empiricalnature of laws and their consistency with sit-uations of dual control I conclude that theworld is a world of agents incompletely de-scribed. Laws neither undifferentially de-scribe nor uniquely govern the phenomenaof our world. And this is accounted forby the fact that it is an incompletely de-scribed world of agents which are consti-tuted at different levels of complexity and or-ganization.48 However it might be objectedhere that all I have shown is that the lawsthat we currently possess do not describe the

world as we currently know it. And thatI have not shown that if we were in factable to reduce (apparently) complex thingsto complete atomistic state-descriptions thatwe would be unable to predict future phys-ical states of the world without referring tocomprehensive entities and principles of be-haviour special to them. The final stage ofmy argument against actualism must thusconstitute a critique of strong actualism inwhich the incoherence of the programme ofreduction it envisages for science is demon-strated.

114:3/o It is important to be clear aboutthe different senses of ‘reduction’. Thereare three distinct ways in which a sciencemight be said to be ‘reducible’ to a morebasic one, which ought not to be confused.There is first the idea of some lower-orderor microscopic domain providing a basis for


the existence of some higher-order propertyor power; as for example, the neurophysi-ological organization of human beings maybe said to provide a basis for their power ofspeech. There is secondly the idea that onemight be able to explain the principles of thehigher-order science in terms of those of thelower-order one. This depends upon beingable to undertake at least a partial transla-tion of the terms of the two domains; thoughit is conceivable that they may retain sub-stantially independent meanings and overlaponly in some of their reference states. Sucha ‘reduction’ may of course result in modi-fications of the laws of the higher-order do-main.49 There is finally the sense in whichit is suggested that from a knowledge of thestates and principles of the lower-order sci-ence we might be able to predict behaviourin the higher-order domain. It is important

to see that it is to this claim that the strongactualist is committed, if he is to eliminatecomplex behaviour in favour of its atomisticsurrogates. It depends not only upon theestablishment of a complete parallelism be-tween the two domains, but upon a closure,i.e. the attainment of a complete atomisticstate-description of all the systems withinwhich the events covered by the descriptionsof the higher-order science occur.

115:1/o Now it is especially important tokeep the second and third senses distinct.For though it is clear that we can explainthe principles and laws of chemistry in termsof those of physics or of classical mechanicsin terms of quantum mechanics, we cannotpredict physical and chemical events such asthe next eruption of Vesuvius on the ba-sis of that knowledge alone. For that wewould need an antecedent complete atom-


istic state-description, i.e. a closure, as well.Now the strong actualist, claiming that theworld is in the end closed, must, unless heis to limit himself merely to a dogmatic re-assertion of this claim, presumably map outa strategy for the sciences to attain such aclosure. The fact that a successful reductionin science does nothing in itself to achieveempirical invariances is something of a blowto the programme (as distinct from dogma)of strong actualism. But even if it did thereis an even more damaging objection at hand(which carries a more general moral for allthose who see in ‘reduction’ the hope of the‘less developed’ sciences). For every histor-ically successful reduction of one science toanother has depended upon the prior exis-tence of an established corpus of scientificprinciples and laws in the domain of the re-duced science. It is easy to appreciate why

this must be so: for without the specifica-tion of some already more or less clearly de-marcated and well charted domain no pro-gramme of reduction could possibly get towork. But this means that as a means of dis-covery, i.e. of achieving such a body of knowl-edge reductionism must fail. For it presup-poses precisely what is to be discovered.

116:1/o I still have not refuted strong ac-tualism as a possible account of the world.This I shall now do by arguing that it is in-consistent with any world containing science,and thus in any world in which science ispossible. The only way of reconciling exper-imental activity with the empiricist notionof law is to regard it as an illusion; that is,to regard the actions performed in it as sub-sumable in principle under a complete atom-istic state-description. In principle this ap-plies not only to experimental activity but


to all scientific activity (including theory-construction) in as much as it involves phys-ical effects. Now this has the absurd conse-quence that the apparent discovery of nat-ural laws depends upon the prior reductionof social to natural science. Or to put it an-other way, in an actualist world there wouldbe no way of discovering laws which did notalready presuppose a knowledge of them. Soa closed world entails either a completed orno science. But as ‘completion’ is a processin time the former possibility is ruled out:so a closed world entails the impossibilityof science. But as science occurs the worldmust be open. This is not the reason whythe world is open (though it is the reason formy justified belief that it is). Rather it is be-cause the world is open that science, whetheror not (and for how long) it actually occurs,is possible. In an open world all laws must

be of normic form; and this is quite indepen-dent of our knowledge of them. In short, thecomplexity of agents and the normic char-acter of laws are irreducible ontological fea-tures of the world; that is, they are necessaryfeatures of our world established as such byphilosophical

117:1 It is relatively easy to show that all(and not just scientific) action depends uponour capacity to identify causes in open sys-tems. For all action depends upon our ca-pacity to bring about changes in our phys-ical environment. Hence we must belongto the same system of objects (nature) onwhich we act. But we not only act on it,in the sense of bringing about changes thatwould not otherwise have occurred; we acton it purposefully and intentionally, i.e. soas to bring about these changes (as the re-sults and consequences of our actions) and


knowing that we are acting in that way. Thisdepends upon our being able to identify fea-tures of our environment as the objects of ourcausal attention and as part of the system towhich causality applies. Thus we must be ca-pable of identifying and ascribing causes inour environment, and knowing ourselves as acausal agent among others. Unless we coulddo this, we could not act intentionally at all.Thus all human action depends upon our ca-pacity to identify causes in open systems (towhich of course the Humean theory cannotapply).

117:2 I suggested earlier that human free-dom is not only compatible with science, buthad to be scientifically understood. This isimportant because it is inter alia a precon-dition for science. For science to be possiblemen must be free in the specific sense of be-ing able to act according to a plan e.g. in

the experimental testing of a scientific hy-pothesis. Human freedom is not somethingthat stands opposed to or apart from science;but rather something that is presupposed byit. The idea that freedom is opposed to orapart from science stems from the empiricistconception of scientific experience as consist-ing in the passive observation of repeated se-quences rather than in the active interven-tion of men in the world of things in an en-deavour to grasp the principles of their be-haviour. Men are not passive spectators of agiven world, but active agents in a complexone.

117:3/o The view of the world as open andthe view of the world as closed lead to to-tally different conceptions of science. Thelaws of nature, which are painstakingly un-covered by the theoretical work of sciencesupplemented wherever possible by experi-


mental investigation, do not seek to describethe myriad phenomena of the world, the con-tents of a biscuit tin or the junk in thebuilder’s yard. They do not seek to tracethe path of a squirrel, predict which rafter asparrow will light on or how many buns thevicar will have for tea.50 They can indeedcome to explain such things in a certain way,but only on the condition that they are notinterpreted as describing them.

2.6 Explanation inOpen Systems

118:1/o The fact that closed systems are apresupposition of the actualist account of sci-ence is reflected (a) in the absence of a the-ory of their establishment and (b) in the ab-sence of a clear contrast between pure and

applied phases of scientific activity or, per-haps better, between science and its uses. Itis with the second that I will be concernedhere. Now consistency with our conceptionof the objects of science as the mechanismsthat produce phenomena, not the phenom-ena they produce (which must now be seenas both complex and differentiated), meansthat we must carefully distinguish betweentwo moments of the scientific enterprize (in-terpreted broadly): the moment of theory,in which closed systems are artificially es-tablished as a means of access to the en-during and continually active causal struc-tures of the world; and the moment of itsopen-systemic applications, where the resultsof theory are used to explain, predict, con-struct and diagnose the phenomena of theworld. Actualism cannot sustain this dis-tinction; or, if we confront it with it, show

2.6. EXPLANATION IN OPEN SYSTEMS 159

how the practical application of our knowl-edge is possible in open systems. This de-pends upon precisely the same ontologicaldistinction as is necessary to sustain the in-telligibility of experimental activity, namelythat between causal laws and the patterns ofphenomena, the mechanisms of nature andthe events they generate, the domains of thereal and the actual. In this way actualism’sassumption of an undifferentiated reality ismirrored in the assumption of an undifferen-tiated science.

119:1 It is because of this ontological dis-tinction that theory is never disconfirmed bythe contrary behaviour of the uncontrolledworld, where all our predictions may be de-feated. Meteorology provides an instructiveexample here. We can have very little confi-dence in the ex ante predictions of weatherforecasters, because of the instability of the

phenomena with which they have to deal.But we can place a great deal of rationalconfidence in their ex post explanations. Forthe law-like statements they use to retrodictthe antecedent events and states by meansof which they both explain what actuallyhappened and excuse their forecasts of itare not meteorological laws. So that me-teorology is in this sense not a theoreticalscience. Rather, mentioning general physi-cal variables, they are physical laws whichhave been confirmed quite independently oftheir use to explain and predict the weather.Thus meteorology, like engineering, standsto physics and chemistry as an applied toa pure science, using the experimentally-established results of the latter. (I am notruling out the possibility that there may beirreducibly meteorological principles.)

119:2 Now it is characteristic of open sys-


tems that two or more mechanisms, perhapsof radically different kinds, combine to pro-duce effects; so that because we do not knowex ante which mechanisms will actually beat work (and perhaps have no knowledge oftheir mode of articulation) events are not de-ductively predictable. Most events in opensystems must thus be regarded as ‘conjunc-tures’. It is only because of this that itmakes sense to talk of a stray bullet or anunhappy childhood affecting ‘the course ofhistory’. And it is only in virtue of thisthat laboratory closures can come to be es-tablished. The importance of experimentalactivity in natural science, conceived as aspecific kind of conjunctural occurrence, al-lows us to stress that the predicates ‘natu-ral’, ‘social’, ‘human’, ‘physical’, ‘chemical’,‘aerodynamical’, ‘biological’, ‘economic’, etc.ought not to be regarded as differentiating

distinct kinds of events, but as differentiat-ing distinct kinds of mechanisms. For in thegeneration of an open-systemic event severalof these predicates may be simultaneouslyapplicable.

120:1 The skills of an applied and a purescientist are characteristically different. Theapplied scientist must be adept at analysinga situation as a whole, of thinking at severaldifferent levels at once, recognizing clues,piecing together diverse bits of informationand assessing the likely outcomes of variouscourses of action. The pure scientist, on theother hand, deliberately excludes, whereasthe applied scientist seeks always to accom-modate, the effects of intervening levels ofreality. Though he is unafraid of flights ofdaring (always risky for the practical man),he holds fast to his chosen objects of inquiry.The applied scientist is an instrumentalist


and a conservative, the pure scientist a real-ist and (at the highest level) a revolutionary.Keynes had the rare gift among economistsof knowing both how to make money andhow money is made.51

120:2/o I said in section 1 that the ac-tivities of explanation, prediction and theidentification of causes not only do not pre-suppose a closure, but they do not neces-sarily involve, though they may make useof laws. There are two points here. First,there is a difference in general between sci-entific and lay explanations. That this is sois entailed by one of the most obvious fea-tures of science, namely the prolonged periodof scientific education and training a novicemust normally undergo before he is consid-ered capable of ‘scientific explanation’. Thishas a rationale in the real stratification ofthe world and a real effort, which is science

needed to penetrate it. Needless to say, how-ever, that stratification cannot justify anyparticular institutionalized form or any so-cial division e.g. between scientists and non-scientists (the educator and the educated)arising from the latter. Secondly what pri-marily distinguishes scientific from lay expla-nations of events is not their structure butthe concepts that figure in them. Thus therole played by laws in the scientific expla-nation of events, a role which is played viathe invocation of the concept of the mecha-nism at work in the generation of the event,which is the function of the citation of thelaw (and which will be discussed further inthe next chapter), is paralleled in lay expla-nations by other kinds of normic statementssuch as platitudes, truisms, assumptions ofrationality or more crudely or vaguely formu-lated law-like statements. Moreover, there


is a case, which I am now going to exam-ine in some detail, in which both scientificand lay explanations have exactly the sameform and in which they do not involve normicstatements at all. This is the transitive verbmodel to which I have already alluded in sec-tion 3 above.

121:1 ‘Why is the door open?’—‘BecauseTania pushed it open’. The door is open be-cause Tania pushed it [open]’ is a paradigmcausal explanation, accomplished withoutreference to laws, by the redescription of theexplanandum event in terms of its cause. Itis informative—there are other reasons whythe door might be open. But it is alsologically necessary; i.e. the explanation isdeductive—if Tania pushed the door open, itmust be open. In this ‘Tania pushed the dooropen’ differs from ‘Tania pushed the doorhard’. ‘Tania pushed the door hard’ may

explain why the door is open but it does soonly contingently. On the other hand ‘Taniaobserved the door open’ cannot explain whythe door is open because there is no conceiv-able way in which observing can bring abouta change in the object concerned (viz. thestate of the door). Now the role of the verb‘push’ in ‘Tania pushed the door open’ is tolink the A-sequence and the B-sequence inDiagram 2.1 by supplying an interpretationof the latter, so that the door’s movementcan be seen as the result of a continuousaction sequence. Note that though ‘Taniamoved up to the door and then the doormoved away’ is a true description it does notmean the same as ‘Tania pushed the dooropen.’

121:2 A Sequence T —–¿–+ — Door — BSequence +———¿—-


Diagram 2.1

121:3/o Now transitive verbs such as‘pushing’, ‘pulling’, ‘knocking’, ‘twisting’,‘binding’, ‘squeezing’, ‘holding’, ‘forcing’,‘driving’, ‘turning’, ‘stimulating’, ‘produc-ing’, ‘generating’, ‘bringing about’, ‘making’,etc. lie at the root of our notion of cause.52

When something is cited as a cause it is beingviewed as that element, paradigmatically anagent, in the total situation then prevailingwhich, from the point of view of the cause-ascriber, ‘so tipped the balance of events asto produce the known outcome’.53 Now theimportance of the transitive verb model isthat it accounts for both the large numberof ordinary causal explanations which are de-ductive (or become so with the addition of asuitable objective complement, perhaps tac-itly understood) and the basic interactions ofclassical mechanics; i.e. the fact that action-

by-contact was not itself felt to be in needof explanation. In neither case is there refer-ence to laws or any other general statements.‘Juanita made Xara push the door open’,‘The mixture made him sick’, ‘He drove hiswife to despair’, ‘The sergeant forced himto pull the trigger’, ‘The elephant crashedinto the juggernaut’, ‘The first billiard ballsmacked into the second’, ‘The irate posi-tivist knocked his ink bottle over’, ‘The psy-choanalyst suggested he open the window’these are the primaeval explanation forms.It has been suggested that it is the fact thatsomething is subject to human manipulationor control that accounts for our identifica-tion of it as the cause.54 But apart from ob-vious counter-examples, it is clear that wecould only know ourselves as causal agentsin a world of other causal agents and thatour notion of cause takes in the possibility


of a world without men. It is because menare agents, not because ‘other agents’ haveaffinities with men, that the concept of causewould still find application in such a world.

122:1 Now if most events in open sys-tems are conjunctures, i.e. are to be ex-plained as the results of a multiplicity ofcauses, to the extent that basic causal ex-planations are involved, one would expecta modification of the transitive verb modelto be necessary, corresponding and similarto that which required a restatement of thenomological model in normic form. This isso. For if a single influence was responsi-ble for the outcome the event could be seen,as in Diagram 2.1, as the simple pure lineardisplacement of its cause (and deducibilitywould be preserved). To the extent how-ever that more than one factor is at workthe event will have to be seen as a kind of

‘condensation’ or ‘distillation’ of its compo-nent causes.

122:2/o I now want to illustrate this bylooking at a fairly typical piece of histori-cal narrative. This will also enable me toidentify some more general characteristics ofexplanation in open systems. In the piece ofnarrative that follows I underline obviouslycausal notions.

123:1 This pressure from the LabourParty, with its great influence on the in-dustrial workers, combined with the attitudeof President Wilson himself, slowly propelledLloyd George in the direction of the formu-lation of war aims. Hindered as he was bythe obligation of earlier agreements with theEuropean allies, he ensured that his decla-ration, made on the 5th January 1918, wasonly in the vaguest terms. It was, how-ever, not incompatible with the much more


specific Fourteen Points enunciated indepen-dently by the American President a few dayslater, and appealed by the German Govern-ment as a basis for peace negotiations at thetime of the armistice in November.’55

123:2/o The first thing to notice aboutthis piece of historical narrative is its de-centralized focus, allowing the emergence, ina series of redescriptions of the event con-cerned, of a picture of the conjuncture orbalance of forces in which it occurred andin terms of which it is explained. The eventis in fact known under three different de-scriptions: Ea, Lloyd George’s formulationof his war aims; Eb, his vague formulationof these war aims; and Ec, his vague yetcompatible (with the Fourteen Points) for-mulation of these war aims. Secondly, theindispensable role that causal notions playin both indicating the key variables which

brought about the event and in renderingintelligible their efficacy can be seen. Whydid Lloyd George formulate his war aims?Because of pressure from the Labour Partyand from President Wilson. Here we imaginethe event as if it were a simple displacement.But now the simple displacement is modi-fied by the effect of another factor, viz, hisprevious obligations, and so Lloyd Georgeformulates his war aims vaguely. The eventbecomes a condensation of the different ex-planatory linkages. Thirdly, each of theseindividual linkages could in principle be lo-cated within some interpretative schema ortheoretical structure. But it is simple dis-placements, transitively understood, and therole that causal notions play in them that ex-plains the peculiar efficacy of what Dray hascalled ‘continuous series’.56 Finally, the non-unified ontology of the explanation should be


noted. The industrial proletariat and Pres-ident Wilson’s attitude co-exist within thesame explanation. The pattern of the expla-nation is illustrated in Diagram 2.2.

124:1 A possible misunderstanding mustbe avoided and a possible puzzle allayed.The physical action causal notions usedin the explanation of such an event areof course employed metaphorically. LloydGeorge is not literally propelled. In thisway they stand in for what some would sayare trivial, though I would prefer to say are(as yet) inadequately understood, processes.Now this differs from the kind of criticismthat I directed against the action-by-contactparadigm when I argued in section 3 thatthough it may provide the source of our con-cept of causality, it cannot provide an ad-equate model for the understanding of ul-timate physical actions. For doors do really

get pushed open and it is perfectly legitimateto talk in this way. What is illegitimate is toregard corpuscles as acting like doors. (Forif the door was a corpuscle it could not re-tain its shape—it would have to be bent tobe ‘opened’.) A puzzle may arise about pre-cisely what event is being explained in oursimple historical explanation, when the sameevent is referred to under three different de-scriptions. But the puzzle dissolves when itis realized that the phrase ‘the event whichoccurred (in si at tj)’ is essentially syncate-gorematic; that is to say that it refers onlyon the basis of some prior description of theevent concerned. And it is precisely the func-tion of the notion of an event to generate re-descriptions of events as specified under theiroriginal descriptions in their explanation. Inthis way it also acts as a possible signpost


into the language of theory.

124:2

graph commented out

Diagram 2.2

125:1 I have taken a simple historical ex-planation because it illustrates some moregeneral features of explanation in open sys-tems. The pattern of explanation, evenwhere well-developed scientific theory can bebrought to bear on an event, is substan-tially the same. In general as a complexevent it will require a degree of what mightbe called ‘causal analysis’, i.e. the resolutionof the event into its components (as in thecase above). These components will then re-quire theoretical redescription, so that thetheories of the various kinds of mechanismat work in the generation of the event canbe brought to bear on the event’s explana-tion. The next step will consist in retrodic-

tion from redescribed component events orstates to the antecedent events or states ofaffairs that could have produced them. Tothe extent that for each determinate effectthere is a plurality of possible causes retro-diction alone cannot be decisive. And so itwill need to be supplemented by indepen-dent evidence for the antecedents until wehave eliminated from the total set of possiblecauses all but the one which, together withthe other factors at work, actually producedthe effect on the occasion in question. Thefour stages in the explanation of an open-systemic event may therefore be summarizedas follows: (i) causal analysis (or resolution)of the event; (ii) theoretical redescription ofthe component causes; (iii) retrodiction vianormic statements to possible causes of thecomponents; (iv) elimination of alternative


causes.

125:2/o Now it is particularly importantto beware of the supposition that if we haveachieved such a complete explanation of anevent (normally of course we will only beinterested in one or two of the influencesat work) this would put us in a positionwhereby we could have predicted it. For thedifferent levels that mesh together in the gen-eration of an event need not, and will notnormally, be typologically locatable withinthe structures of a single theory. In generalthe normic statements of several distinct sci-ences, speaking perhaps of radically differentkinds of generative mechanism, may be in-volved in the explanation of the event. Thisdoes not reflect any failure of science, but thecomplexity of things and the multiplicity offorms of determination found in the world.The idea that a complete explanation of an

event entails a potential prediction of it de-pends upon the possibility of the reductionof the various sciences to a single level anda complete description of all the individualsat that level; i.e. it depends upon the ideaof an antecedent closure. Now it is not thatthis represents an unreasonable ideal for sci-ence; but rather that it constitutes a conjec-ture about the nature of the world which isin fact false and which, if acted upon, couldhave the most deleterious effects on science.If science is to be possible the world must beopen; it is men that experimentally close it.And they do so to find out about structures,not to record patterns of events.



1 J. S. Mill, Auguste Comte and Positivism, p.6.

2 K. R. Popper, The Logic of Scientific Dis-covery, p. 59.

3 C. G. Hempel, Aspects of Scientific Expla-nation, p. 366.

4 K. R. Popper, Conjectures and Refutations,p. 39, n. 3.

5 J. Gibbs and W. Martin, Status, Integrationand Suicide, p. 197.

6 See e.g. N. R. Hanson, Observation and Ex-planation, p. 45.

7 M R. Ayers, The Refutation of Determin-ism, p. 6 and passim uses the term ‘actualism’to refer to the doctrine that only the actual ispossible. The connection between the two con-cepts will become clear in due course.

8 Leibniz’s pre-established harmony of mon-ads may be usefully compared with Hume’s con-

stant conjunctions of atomistic events.

9 P. S. de Laplace, A Philosophical Essay onProbabilities, p. 4. Positivists still pay obeisanceto this concept of knowledge. See e.g. Brodbeck’scharacterisation of it, in an echo of Laplace, as‘perfect knowledge’ (M. Brodbeck, ‘Methodolog-ical Individualism: Definition and Reduction’,Readings in the Philosophy of the Social Sci-ences, p. 289); and Hempel’s wry admission thatclassical deterministic (i.e. Laplacean) systemsconform ‘best’ to his model of explanation asdeductive subsumption under universal laws (C.G. Hempel, Aspects of Scientific Explanation, p.351). It is but a short step to identifying suchsystems with theories. (See e.g. M. Brodbeck,op. cit., p. 288; C. G. Hempel, ibid; and R. Rud-ner, The Philosophy of Social Science, p. 91).

10 P. S. Laplace, ibid.

11 The concept ‘event’ functions here syncate-gorematically. Its purpose is, in context, to gen-erate the appropriate redescriptions of the eventsconcerned.

12 See e.g. W. Kneale, Probability and Induc-


tion, p. 60.13 See e.g. J. S. Mill, A System of Logic, Vol.

I, Bk. III, Chap. 3, Sect. 3, or A. J. Ayer, Foun-dations of Empirical Knowledge, Chap. 4, Sect.17. As has been frequently pointed out, by Milland Ayer among others, this concept does notaccord well with our normal usage; so in prac-tice the Humean tends to modify it in the direc-tion of the intelligibility concept by making thecause an individually critical factor in a jointlysufficient set. On the other hand, to the extentthat the intelligibility theorist is committed tothe doctrine of empirical realism he must relyon a background of empirical generalizations tojustify his citation of the cause.

14 Cf. e.g. E. Nagel, The Structure of Science,p. 334.

15 See e.g. P. Gardiner, The Nature of His-torical Explanation; W. Dray, Laws and expla-nation in History; C. B. Joynt and N. Rescher,‘The problem of Uniqueness in History’, Historyand Theory, Vol. 4; and M. Scriven, op. cit.

16 Cf. M. Bunge, Causality, p. 122.17 Cf. Newton’s 2nd Rule of Reasoning in Phi-

losophy: to the same natural effects we must, asfar as possible, assign the same natural causes’,ibid. And Hume: ‘the same cause always pro-duces the same effect, and the same effect neverarises but from the same cause’, A Treatise onHuman Nature, p. 173. The rationale for this re-quirement lies in the counter- intuitive nature ofthe implication that the future be better knownthan the past. Moreover given the logical re-versibility of the connective and the classicalconcept of time a disjunctive plurality of causescould be transformed into a disjunctive pluralityof effects so as to produce a radical indetermin-ism in nature, i.e given x then y1 or y2 . . . oryn!

18 Regularity determinism does not make aclaim about the constancy of conditions. Itsclaim concerns the constancy of the conjunctionbetween conditions, whether conditions shouldhappen to be changing or not.

19 Cf. J. R. Aronson, ‘Explanation with-out Laws’, The Journal of Philosophy, Vol. 66,


(1969), pp. 541–57.20 See e.g. A. Koyre, op. cit., Chaps 1–221 W. Heisenberg, Physics and Philosophy.22 T. Hobbes, On Human Nature, reprinted

in Body, Man and Citizen, ed. R. S. Peters.23 Cf. M. Bunge’s genetic principle or prin-

ciple of productivity: ‘nothing can arise out ofnothing or pass into nothing’, op. Cit., pp. 25–6. Cf. also Kant’s 1st Analogy of Experience,op. cit.; and W. H. Walsh, ‘categories’, Kant-studien Band 45 (1954), reprinted in Kant, ed.R. P. Wolff, pp. 54–70.

24 Cf. S. Korner, ‘Substance’, P.A.S. Supp.Vol. 38 (1964), pp. 79–90.

25 Cf. M. Capek, op. cit., p. 54.26 see e.g. I. Dirac, Is there an aether?’, Na-

ture 168, pp. 906–727 Cf: ‘Hitherto we have explained the phe-

nomena of the heavens and our sea by the powerof gravity, but we have not yet assigned the causeof this power.... To us it is enough that gravitydoes really exist, and act according to the lawswe have explained, and abundantly serves to ac-count for all the motions of the celestial bodies,

and of our sea.’ I. Newton, op. cit, Book III,General Scholium.

28 See e.g. P. K. Feyerabend, ‘Problems ofEmpiricism’, Beyond the Edge of Certainty, ed.R. G. Colodny, esp. pp. 168–70, and T. S. Kuhnop. cit., esp. pp. 99–103.

29 See e.g. M. Capek, op. cit., Pt. II.30 R. J. Boscovitch, A Theory of Natural Phi-

losophy, pp. 10–13 and 19–68 and I. Kant, TheMetaphysical Foundations of Natural Science.

31 Cf. R. Harre, op. cit., p. 308.32 By a ‘liability’ I mean simply what Hobbes

called a ‘passive power’.33 See J. Piaget, The Child’s Perception of

Physical Causality, and A. Michotte, The Per-ception of Causality.

34 Cf. J. R. Aronson, op. cit., pp. 551–5.35 However a closure might be both univer-

sal and artificial if a generative mechanism hadendured as a latent potentiality of nature untilawakened by science under experimentally con-trolled conditions or if it had never been acti-vated in its experimental range. But to the ex-tent that the sciences are concerned with struc-


tures that not only exist but act independently ofthem (and so explain what goes on in the worldoutside the laboratory) the first possibility willbe exceptional; and to the extent that they areconcerned with the conditions under which thesestructures act the second possibility will.

36 J. S. Mill, A System of Logic, Vol. I, p. 378.

37 E. Grunberg, ‘The Meaning of Scope andExternal Boundaries of Economics’, The Struc-ture of Economic Science, ed. S. R. Krupp, p.151.

38 R. B. Braithwaite, Scientific Explanation,p. 302.

39 There is another possible use for a ceterisparibus clause, viz. as a protective device in theearly stages of a science’s development. This willbe considered in Chapter 3.

40 J. S, Mill, op. cit., Bk. III, Chap. 10, Sect.5.

41 P, T. Geach, ‘Aquinas’, Three Philoso-phers, G. E. M. Anscombe and P. T. Geach, p.103; and A. Ryan, The Philosophy of John Stu-

art Mill, pp. 65–6.42 See e.g. E. Nagel, op. cit., p. 493.43 Cf. Weber’s concept of an ‘ideal type’ as

a one-sided exaggeration of an aspect of ‘con-crete’, i.e. empirical, reality. See e.g. M. Weber,Methodology of the Social Sciences.

44 M. Scriven, op. cit., p. 466.45 H. Poincare, Science and Hypothesis, p. 98.46 G. E. M. Anscombe, op. cit., p. 21.47 Cf. M. Polanyi, ‘The Structure of Con-

sciousness’, The Anatomy of Knowledge, ed. M.Grene, p. 321.

48 Cf. M. Bunge, The Myth of Simplicity,Chap. 3; and M. Polanyi, The Tacit Dimension,Chap. 2.

49 Cf. P. K. Feyerabend, ‘Explanation. Re-duction and Empiricism’, Minnesota Studies inthe Philosophy of Science Vol. III, eds. H. Feigland G. Maxwell, pp. 28–95.

50 A caricature of such an empiricism exists insome of the early experiments conducted underthe august auspices of the Royal Society. Thefollowing is an example: ‘1661, July 24: a cir-cle was made with a powder of unicorn’s horn,

2.8. APPENDIX: ORTHODOX PHILOSOPHY OF SCIENCE AND THE IMPLICATIONS OF OPEN SYSTEMS173

and a spider set in the middle of it, but it im-mediately ran out several times repeated. Thespider once made some stay upon the powder’,C. R. Weld, History of the Royal Society, Vol.I, p. 113. Among the items of allegedly scien-tific interest collected by the Society were ‘theskin of a moor, tanned with the beard and hairwhite’ and ‘an herb which grew in the stomachof a thrush’, ibid, p. 219. Quoted in P. K. Fey-erabend, ‘Problems of Empiricism’, op. cit., p.156.

51 See R. F. Harrod, The Life of John May-nard Keynes.

52 Cf. H. L. A. Hart and A. M. Honore, Cau-sation in the Law, Chap. 2, Sect. 2.

53 M. Scriven, ‘Causes, Connections and Con-ditions in History’, Philosophical Analysis andHistory, ed. W. H. Dray, p. 248.

54 D. Gasking, ‘Causation and Recipes’, Mind1955, pp. 479–87.

55 H. Pelling, Modern Britain 1885–1955, p.77.

56 W. H. Dray, Laws and Explanation in His-tory, pp. 66ff.

2.8 Appendix: Ortho-dox Philosophy ofScience and the Im-plications of OpenSystems

127:1 It may be felt that I have dealt rathersummarily in Chapter 2 with some of themost hallowed doctrines of received philos-ophy of science; so I want here to turn toa more detailed examination of them in thelight of the phenomenon of open systems.

127:2 The structure of orthodox philos-ophy of science is based squarely on theHumean theory of the actuality of causallaws. But it is convenient to give Nicod’scriterion,1 which presupposes and implies itco-equal status. I shall formulate them as


two principles:

127:3 P1, the principle of empirical-invariance, viz. that laws are or depend uponempirical regularities; and

127:4 P2, the principle of instance-confirmation (or falsification), viz. that lawsare confirmed (or falsified) by their in-stances.

127:5/o Post-Humean philosophy of sci-ence has called into question only the suffi-ciency, not the necessity of these principles;i.e. it has left the ontology implicit in themintact. Thus in the most advanced recentpositions theory is regarded as irreducible;and as supplying at least part of the groundsfor laws. The significance of such modifica-tions will be considered in the next chapter.Here, as in the body of Chapter 2, I will notdistinguish between philosophers who regardP1 and/or P2 as necessary and sufficient and

those who regard them as merely necessary.Once more no harm will be done by this con-flation as all my objections here turn on thelack of necessity of these principles and ipsofacto of the theories they sustain. It is forthe sake of explanatory convenience and toavoid repetition that I formulate and discussthem in their ‘necessary and sufficient’ form.

128:1 Both P1 and P2 depend upon a clo-sure, and hence upon the assumption of asimple undifferentiated reality. P1 gives riseto the truth-functional concept of naturalnecessity. This is the idea that the logicalstatus of laws can be explicated, at least inpart, by the formula “for all x, fx inplies gx,”where the predicates ‘f’ and ‘g’ are definedextensionally or are at least given some defi-nite empirical interpretation e.g. by means ofcorrespondence rules. The definition may beostensive or operational;2 and if ostensive,


either sensationalist3 or physicalist.4 P1 issusceptible of descriptivist (Mach)5 and in-strumentalist (Ryle)6 interpretations; and ofclassical empiricist and transcendental ide-alist ones. P2 is susceptible of inductivist(Carnap)7 and falsificationist (Popper)8 in-terpretations; and of positivist and conven-tionalist ones.

128:2/o I am going to use the term ‘de-ductivism’ to refer to the ensemble of theo-ries erected on the basis of P1 and P2. Mychoice of the term ‘deductivism’ may notseem an altogether happy one in view of thefact that it is meant to cover philosopherswho have regarded themselves, as ‘induc-tivists’ and ‘instrumentalists’, as opposed to‘deductivism’. However, I do not think thatthis is a serious difficulty. For there is no wayin general of getting an inductive policy go-ing without appealing to an antecedently for-

mulated lawlike statement. (The exceptionis provided by the pure Humean case wherethe events are intuitively ascertainable atom-istic instants, and each event is a member ofa linear series.) For one sequence to give sup-port to another the antecedent events mustbe alike in relevant respects. But to talk ofthe relevance of the ‘respects’ already pre-supposes a tentative (conjectured) or con-firmed law. Hence the inductivist in theorymust be a deductivist in practice. Similarlya rule of inference can always be recast asthe major premise of a syllogism. That isto say, an inference ticket (a ‘season ticket’,as Ryle revealingly calls it)9 remains validonly as long as some empirical generalizationis true. Thus the instrumentalist questionsonly the descriptivist interpretation of theuse of laws, not their logical form.

129:1 I will first set out the overall struc-


ture of deductivism before examining itscomponents, individually and collectively, inthe light of open systems.

129:2 Underpinning deductivism is the ac-tualist thesis that laws are relations betweenevents or states of affairs. If the world con-sists only of atomistic events or states-of-affairs then for a general knowledge of it tobe possible their relations must be constant.And so we have P1 the principle of empirical-invariance

129:3 (1) laws are or depend upon con-stant conjunctions of events or states of af-fairs (which constitute the objects of actualor possible experiences). As such it gener-ates the familiar Humean theory of causal-ity. This theory is susceptible of differentinterpretations, viz.

129:4 (2)’ as a theory of what we mean bysaying ‘X causes Y’, viz. that the events as

specified under their descriptions X and Yare regularly conjoined; and

129:5 (2)” as a theory of how we must beprepared to justify the claim that X causesY, viz. by showing that the event as spec-ified under these descriptions are regularlyconjoined.

129:6 In the case of both theories it ispossible to substitute the weaker require-ment that the events be specified underthese or some other set of descriptions. Theweaker variants could be indicated by thesubscript1.

129:7 We then have a theory of explana-tion, that

129:8 (3) events are explained by subsum-ing them under one or more universal laws;i.e. by deducing them from a set of one ormore universal laws, together with a state-ment of their initial conditions. This has be-


come known as the Popper-Hempel theory ofexplanation. Popper was the first to restateit in modern times10 and Hempel has beenits most systemic advocate and defender.11

It is convenient to divide the theory into tworequirements:

130:1 (3a) a deducibility requirement, viz.that the explanandum be deducible from theexplanans; and

130:2 (3b) a covering-law requirement,viz. that the explanans contain at least oneuniversal law.

130:3 Next, a theory of prediction to theeffect that

130:4 (4) events are predicted by deducingthem from a set of universal laws togetherwith a statement of their initial conditions.(3) and (4) together give rise to the theorythat

130:5 (5) explanation and prediction are

symmetrical (in the sense of (iii) on page 63above).

130:6 On this view their difference liesmerely in the fact that in ‘explanation’ theexplanandum event lies in the past and in‘prediction’ in the future.

130:7 Then we have a theory of the expla-nation of laws, namely

130:8 (6) laws are explained by subsum-ing them under or deducing them from moregeneral, abstract or inclusive statements.Such statements may be regarded as theo-retical principles or hypotheses. They maybe interpreted descriptively, instrumentallyor as fictions. I leave aside consideration ofthe various theories of theories until Chapter3.

130:9 Next a theory of the explanation oftheories and even sciences, viz.

130:10 (7) theories and sciences are ex-


plained by deductively subsuming them un-der more basic or general ones. The ex-plained theory or science is then said to havebeen ‘reduced’ to the explaining one.12 Thuswe have a theory of the development of sci-ence, viz.

130:11 (7)* science develops monisticallyor in a linear fashion so as to leave meaningand truth-value unchanged.

130:12/o Theories (7) and (7)* are re-jected by many philosophers, most notablyPopper, committed to other components ofthe deductivist structure. As Feyerabendhas pointed out these theories generate theirown restrictive methodology, embodied inthe conditions that

131:1 (7)** theoretical innovations shouldbe consistent and meaning-invariant with re-spect to established, i.e. currently accepted,

theory.13

131:2 These conditions tend inevitably tohave a conservative effect. And they maybe regarded as rationalizing the practice ofwhat Kuhn has called ‘normal science’.14

131:3 I do not intend to discuss theories(7)* and (7)** in any detail here. That willbe done in Chapter 3. But their connec-tion through thesis (7) with the doctrine ofactualism and their consequent dependenceupon the presupposition of a closure shouldbe clear.

131:4 According to theses (3), (6), (7) and(7)*, the explanation of events, laws, the-ories and sciences all partake of essentiallythe same ideal ‘deductive-nomological’ form.And they share this also (through theories(2), (4) and (5)) with the activities of pre-diction and the identification of causes. Sci-entific knowledge then must consist (in part


or in whole) of a deductive structure. Butwhich one? Which out of all possible deduc-tive structures is the best (or in Popperianterms, ‘least worst’)?

131:5 Here P2 gets to work. Thus we havethe theories that:

131:6 (8) laws, theories and sciences aredirectly or indirectly confirmed or corrob-orated by their instances (which constitutethe objects of actual or possible experiences);and

131:7 (9) laws, theories and sciencesare directly or indirectly falsified by theircounter-instances (which constitute the ob-jects of actual or possible experiences).

131:8 Unless the meaning of theoreticalterms is reduced to ostensively defined in-stances, as in Machian descriptivism, suchcriteria cannot be sufficient, but they arenormally posited as at least necessary. (8)

and (9) have sometimes been interpreted ina conventionalist way. But as what is re-garded as ‘conventional’ is only the decisionto accept a report as being genuinely ‘ob-servational’ (and so capable of furnishing agenuine instance or counter-instance of a pu-tative law), a conventionalist interpretationdoes not affect the status of the principle it-self.15

132:1 Finally, following on from (9), wehave a maxim of scientific practice to the ef-fect that:

132:2 (10) scientists should, in formulat-ing their theories, state quite unambiguouslythe empirical conditions under which theyare prepared to reject them (cf (iv) on pp.63–4 above). This may also be taken as acriterion of what it is to be ‘scientific’ and‘unscientific’, viz.

132:3 (10)* to be ‘unscientific’ is not to be


prepared to state such conditions or havingdone so to revise them ex post facto.16

132:4 Most received philosophy of scienceis based on a core extracted from theories(1)-(10).* These theories all share one greatweakness: they all presuppose a closure. If,as I have argued, the world is in fact openthen they must all be more or less drasti-cally revised and in some cases completelyrejected. I have already argued against (1)in Chapter 2 (especially section 4) above so Iwill not discuss it separately here. It will beremembered that the criterion of open-ness isthe non-invariance of empirical relationships.Now clearly if the law-like statements whoseantecedents are instantiated in open systemsare interpreted as invariant empirical regu-larities they must be regarded as false. Butthis means that there can be neither laws,because there are no invariant empirical reg-

ularities; nor theories, because they are con-tinually being falsified; so that neither ex-planation nor prediction can be given a ra-tional basis. I have examined two actual-ist responses to this predicament (viz. weakand strong actualism) and showed how nei-ther can sustain the concept of laws applyingtransfactually, viz. in open and closed sys-tems alike, that we need to render intelligi-ble both the experimental establishment andthe practical application of our knowledge.

132:5/o It is important to keep the de-ducibility and covering-law requirements, asexpressed in (3a) and (3b) distinct. For ei-ther can be non-trivially satisfied withoutthe other. I have already shown in 2.6 howone can have explanation in terms of a net-work of normic statements (which may bestrictly universal in the sense of space-time-invariant) without the event being deducible.


Of course the sense in which the covering-law‘covers’ in this case is different. In the sameway but from another aspect, the deducibil-ity requirement may be violated in the de-velopment of science although all the state-ments involved are universal. An exampleof this is given by the way in which New-ton’s theory both explained and correctedKepler’s and Galileo’s laws.17 In these cases(3b) is satisfied but not (3a). On the otherhand our atavistic causal explanation ‘Taniapushed the door open’ is deductive, thoughno laws are involved. It is also possible forthe deducibility requirement to be satisfiedby statements mentioning named individu-als or specific space-time regions.18 It shouldbe noted that one can have deductive ‘ex-planations’ of events under transient empir-ical regularities but not deductive explana-tions of their explanations: for the space-

time restriction cannot itself be derived froma strictly universal law.

133:1/o Thesis (9) implies that all law-likestatements whose antecedents are instanti-ated in open systems are false. It is there-fore as stated quite useless as a decision rulefor choosing between different law-like state-ments (or as an ‘organon of criticism’).19

One way of dealing with this would be toallow theory a role in grounding laws. Butwith an unchanged ontology this is bound tobe ultimately unsatisfactory. Of course if wepossess a good theory it is irrational to re-linquish it in the face of recalcitrant facts—without a better one. But our justificationfor holding on to the theory must be that itmight eventually be able to explain them (bysuitable modifications, refinements or devel-opments) or show that they are not after allfacts (i.e. that the statements used to state


them are untrue). If it could never explainand/or correct them this justification wouldcollapse. Of course the sufficiency of Nicod’scriterion must be disputed: the grounds fora law or theory cannot be exclusively em-pirical. Theory must supply some idea of a‘connection’, without which it would be im-possible to tell necessary from accidental se-quences. But in an open world Nicod’s crite-rion cannot be necessary either: the groundsfor a law or a theory cannot be undifferenti-atedly empirical. For the conditions mustnormally be carefully controlled so that ahypothesis about the connecting mechanismcan be put to a fair test.

134:1 If it is wrong to regard law-likestatements and theories as being falsifiedby the non-occurrence of their consequentsin open systems, it is equally wrong to re-gard them as being confirmed or corrobo-

rated by their occurrence in an undifferen-tiated way, i.e. independently of the contextin which the putatively falsifying/confirminginstance occurs. Theses (9) and (8) musttherefore be restated so as to place a restric-tion on the system in which a genuinely fal-sifying/confirming instance occurs, viz. thatit be closed.

134:2 Not all evidence is equal; or rathernot all evidence is evidence for or againsta law. In general it takes a closed systemto furnish evidence capable of falsifying orconfirming a law. And within the class ofclosed systems, experimentally closed onesare preferable.

134:3 For experimentally we can test andre-test a greater number and variety of sub-junctive conditionals of the form ‘if x were totake on a certain value, then yj would occur’by instantiating their antecedents. Whereas


outside the laboratory we are restricted toobserving whatever sequential performancesnature is obliging enough to put on.

134:4 Similar considerations apply to the-sis (10). One cannot lay down hard andfast criteria spelling out beforehand whichobservable situations ‘if actually observedmean the theory is refuted’ (see (iv) on pp.63–4 above). For one can never know be-forehand whether the system will be actu-ally closed. On the other hand the closureof the system is not a part of the observ-able situation; so that it cannot be incorpo-rated into the criterion of scientificity with-out destroying it. The judgement that thesystem is closed can only be made ex postafter we have observed (and theoretically as-sessed) the observable situation.

134:5 Theses (3) and (9) postulate a syn-tactical identity between explantion, predic-

tion and falsification in that, taken together,they imply a correct prediction explains andan incorrect prediction falsifies. They de-pend upon the assumptions that it is possibleto give a purely syntactical account of scien-tific activities and that these activities al-ways occur in the context of an antecedentlygiven closure. Recent philosophy of sciencehas clearly demonstrated the poverty of theformer; it is with the incorrectness of the lat-ter assumption that I am here concerned.

135:1 Now explanation in open systems,failing the attainment of an antecedent clo-sure, normally requires, as I have pointed outin 2.6 above, retrodiction; that is the infer-ence from present effects to prior (perhapshidden, perhaps just unrecorded) causes, viathe application of normic statements. Nowthe significance of this activity is that it pre-supposes a non-conventional division of the


class of law-like statements into those whichare and those which are not capable of func-tioning in this way, i.e. into those which areaccepted (for transfactual application) andthose which are still under test. Now oncewe allow this the postulated symmetries be-tween explanation and prediction and expla-nation and falsification break down. And itbecomes important to distinguish betweentwo kinds of prediction conflated in deduc-tivism’s syntactical account of science: prac-tical predictions of categorical form whichare rarely made in science but which areimportant in some of its practical applica-tions in open systems and about which theapplied scientist can never be deductivelycertain; and test predictions of hypotheti-cal form made under effectively closed condi-tions in order to test a theoretical hypothesis

or putative law.

135:2/o It is easy to see why the expla-nation/falsification symmetry collapses oncewe allow the legitimacy of retrodiction. Forthe activity of retrodiction presupposes thetruth and applicability of the law used; thepossibility that it is false is ruled out a pri-ori. Now the intelligibility of falsification de-pends upon the idea that the would-be fal-sifier has independent grounds for the oc-currence of the initial conditions. If thelegitimacy of retrodiction is denied a vi-cious regress back to sense-experience en-sues. Thus suppose we have a law-like state-ment of the form S1 ‘whenever events of typeE0 occur events of type E1 occur’. For S1 tobe used to explain the occurrence of E1 ina way which is consistent with the idea ofits being subject to falsification independentgrounds for E0 are required, say G0. But


the connection between E0 and G0 is itself acontingent causal one, which may be repre-sented by the hypothesis S, ‘whenever eventsof type G0 occur events of type E0 occur’.Hence we need independent grounds for theoccurrence of G0, say G’0, if our use of S2, isto be consistent with the idea of its falsifiabil-ity. But as G’0 stands in a contingent causalrelationship we need independent groundsfor it too, and so on.... There is of courseonly one connection with E0 which, beingnon-causal and non-contingent, does not re-quire independent grounds, namely immedi-ate sense-experience. Thus insistence on in-dependent grounds for the initial conditionsof an explanation, which is an inevitable con-sequence of the idea of its susceptibility tofalsification, inevitably leads to the require-ment that the initial conditions be appre-hended in sense-experience; in which case the

event could have been predicted. The root ofthe trouble here is that the causal relation-ship is taking too much strain: it is requiredboth to be contingent (and as such to besubject to falsification) and to explain; func-tions that it cannot combine without viciousregress to sense-experience. Once we distin-guish between open and closed systems, how-ever, this regress can be avoided. For we mayallow that events may be explained in openand closed systems alike, but that law-likestatements may only be falsified under ef-fectively closed conditions (where deductivetest predictions are possible).

136:1/137 Now the point of the explana-tion/prediction symmetry thesis is vitiatedin open systems. For we can give excellentexplanations, in virtue of the transfactualapplicability of our independently validatedknowledge, where we are incapable of any


predictions (save perhaps of the most imme-diate or the most tentative sort). Moreoverthe kinds of statements involved in the twoactivities are radically different: explanationproceeding by way of normic, and predictionby way of empirical, statements. An empir-ical generalization typically merely general-izes the problem to be explained, whereas anormic statement locates it in the context ofan explanatory theory. On the other hand,normic statements may be inferior predictorsto the most crude generalizations or rules ofthumb. Further, the occurrence of the eventitself may be a practically necessary condi-tion of our knowledge of the former state ofthe system, as in the case of the collapse ofa bridge or an airplane crash, or even of thekind of system with which one is concerned,as in the case of the sudden onset of uncon-trollable hysteria.20 Again, the intelligibility

of much practical science of an exploratorykind, such as prospecting for oil, dependsupon the existence of a radical asymmetrybetween explanation and prediction. Be-cause of such difficulties defenders of thesymmetry thesis have been forced to mod-ify it so that it requires only that were we tobe in possession of all the information avail-able at the time of the explanation then wecould have predicted it.21 Now I have arguedin 2.6 that it is possible to give a completeexplanation of an event without thereby be-ing in a position to deduce it, namely if thedifferent generative mechanisms at work areof radically different kinds; so that the refor-mulated symmetry thesis is either false or,if deducibility is built into the definition of‘explanation’, uninterestingly tautologous.22

One further point on prediction. We are onlydeductively justified in predicting an event if


the system is closed. But there is no way ofknowing in advance (at the only time whena prediction is relevant) whether the condi-tions for a closure will in fact be satisfied.Hence the probability of an event’s occur-rence can never be 1.

138:1 The undifferentiated ontology of re-ceived philosophy of science results in thevery damaging view expressed in (v) on page64 above, viz. that the acid test of a theory isits predictive power. On this view the moreaccurate a theory’s predictions—no matterof what or where—the more worthwhile it isretaining. Coupled with permissiveness overthe use of the CP clause, such a position pro-vides a powerful rationale for scientific con-servatism of any school. It has been usedas such by Heisenberg in physics and Skin-ner in psychology, by Friedman in economicsand Osiander in astronomy. But armed with

our concept of a complex and differentiatedreality we can see what is wrong with it. For,on the one hand, there will always be morethan one hypothesis capable of saving anygiven set of facts, so independent tests forthem will always be necessary; and, on theother, it is only under closed conditions thatsuch tests can be decisive. Consistency withthe facts is neither necessary nor sufficientfor a theory.

138:2/o Popper does not seem to see theconnection between the criteria of explana-tion and rationality he expouses and the his-toricist view of social science he condemns.That there is a connection is clear. For if weknow that power corrupts and regard thisas a true, if trivial, empirical generalizationthen given only a knowledge of the initialcondition that N is powerful we can predictwith deductive certainty that N will be cor-


rupted. On the other hand if we cannot knowwhen the initial condition is satisfied the lawcannot be applied and so is quite useless foreither explanatory or social engineering pur-poses. If Popper is committed to thesis (3)in social science he is committed to thesis(4) and so to the historicist view of scienceas the prediction of events (savoir, pour pre-voir). Popper nowhere denies the applicabil-ity of thesis (8) to social science.23 But heequivocates between a conception of histori-cism as the view that the aim of the socialsciences is the prediction of future events24

and the view that its aim is to make un-conditional historical prophecies.25 But theauthors he attacks did not make uncondi-tional historical prophecies: Hegel in factmade no predictions, and Marx only condi-tional ones. Popper’s real argument is notagainst predictability as such in social sci-

ence, but against the predictability of cer-tain kinds of social events, viz. large-scalesocial changes and their consequences. It isan argument against certain theories of so-cial becoming which he interprets historis-tically. Indeed one could almost say it isan argument for an a-historical form of his-toricism (in which the laws involved are re-garded as strictly universal) against an his-torical one (in which they are regarded asspatio-temporally restricted). Historicism isin general invalid in all its forms. And itis invalid in both natural and social science.And for exactly the same reasons. Inciden-tally, this is true for the refutation of his-toricism that Popper derives from the logicalimpossibility of predicting the precise effectsof future knowledge in as much as any suchprediction depends upon a knowledge of thatknowledge.26 For its effects on nature are no


less predictable than its effects on men.27

139:1/o I turn now to the central unify-ing theory of explanation, viz thesis (3) andthe theories of causality, viz. (2)’ and (2)”,that underlie it. Consider once more theparadigm of the kind of context in whicha causal claim is made. ‘Why is the dooropening?’ ‘Because Tania’s pushing it’. Nowit is certainly not the case that in saying‘Tania’s pushing it caused the door to open’we mean that every time Tania pushes itthe door opens. For there are times whenit is locked and times when she must turnthe door knob too. (Nor equally is it thecase that every time the door opens is it be-cause Tania pushes it.) But neither couldwe produce any universal law which wouldshow why the door opened in this particu-lar case. Generally, theories (2)’ and (2)”are only plausible if in the case of (2)’ we

mean to imply and in the case of (2)” wehave grounds for supposing that a closurehas been obtained of the system in which theevents occur. Without this we cannot pos-sibly mean nor can we reasonably be com-mitted to showing that the events which weclaim are causally connected are so in virtueof being constantly conjoined. In short theHumean theory cannot be a general theoryof causality but at best a theory of what isinvolved in the making of causal claims inclosed systems and where the events are sep-arately identifiable atomistic instants.

140:1/o Davidson has proposed a sophis-ticated reformulation of the Humean theory.‘It is an error’, he says, ‘to think that noexplanation has been given until a law hasbeen produced. Linked with [it] is the ideathat singular causal statements necessarilyindicate, by the concepts they employ, the


concepts that will occur in the entailed law.Suppose a hurricane which is reported onpage 5 of Tuesday’s Times causes the eventreported on page 13 of Wednesday’s Tribune.Should we look for a law relating events ofthese kinds? It is only slightly less ridicu-lous to look for a law relating hurricanesand catastrophies. The laws needed to pre-dict the catastrophy with precision would, ofcourse, have no use for concepts like hurri-canes and catastrophies’.28 Davidson’s anal-ysis of ‘singular causal statements’ is as fol-lows: ‘“A caused B” is true if and only ifthere are descriptions of A and B such thatthe sentences obtained by putting these de-scriptions for “A” and “B” in “A caused B”follows from a true causal law’.29 The objec-tion to this suggestion is that it places a re-quirement on the verification of causal claimswhich is (a) impossible, (b) useless, and (c)

unnecessary. For, on it, ‘A’ and ‘B’ stand in,in the original causal claim, for the completeatomistic state-descriptions that would formthe antecedents and consequents of such acausal law. But it is presumably only be-cause (at least from a Humean viewpoint)we have not got and perhaps cannot get suchstate-descriptions that we make the originalcausal claim in the first case. Davidson holdsthat such claims are defended by ‘producinga relevant law or giving reasons for believ-ing such exists’.30 How? then would one setabout defending a claim about the causes ofthe French Revolution? By giving groundsfor believing that a true ‘neurological, chem-ical or physical law’31 exists? But such laws,were they to be known, would cover any setof historical causes of the French Revolution;in short, they would not enable us to discrim-inate between true and false causal claims at


the level we are concerned with; that is, atthe level at which the initial causal claim ismade. Neurological laws are consistent withany social event and so cannot be possiblyused to defend specific causal claims involv-ing people. Worse still, in the end the onlyway of defending the belief that a Humeancausal law exists covering some particularcase will be by appealing to the truth of regu-larity determinism. Hence the defense of themost specific causal claim becomes an ex-pression of faith in the philosophical dogmathat the world is so constituted that the sim-ple formula ‘same cause, same effect’ every-where applies. On the other hand, once wedeny the premise that a causal claim entails aHumean causal law, and hence a closed sys-tem, these absurdities can be avoided; andwe can then allow that causal claims are de-fended, where they require justification in

terms of general statements at all, by an ap-peal to the normic statements of the level atwhich the original causal claim was made.

141:1 Hempel makes a similar mistake inarguing that singular causal statements ofthe ‘q because p’ (man bites dog) type ‘claimby implication’, ‘tacitly presuppose’, or ‘as-sert by implication’ the existence of coveringlaws.32 Now it is not clear from his accountwhether he regards this as part of the mean-ing of the original causal claim (cf. thesis(2)’ above) or merely as indicating the wayin which it is to be justified (cf. thesis (2)’above). If the former we have the absurditythat ‘man’ and ‘dog’ change their meaningwhen they are used in the making of a causalclaim. If the latter a confusion between whatit is necessary to know to ascribe causes andwhat is necessary for our knowledge of theascription of causes to be possible, viz. that


certain underlying laws hold. The latter isneither necessary nor sufficient for the for-mer.

141:2/o The non-availability of Humeancausal laws is undoubtedly an embarrass-ment to the modern Humean. Inevitably hefalls back on the idea that our explanationsare sketches to be filled out in the fullness oftime (cf. strong actualism) or that they aresubject to an implicit ceteris paribus clause(cf. weak actualism). In addition he may re-lax the stringencies of the deductive modelby allowing other categories of explanation:such as statistical, elliptical and partial for-mulations.33 Statistical explanations, how-ever, spoil the point of the deductive model:for, as Scriven has put it, ‘they abandonthe hold on the individual case’.34 Moreoverthere is no a priori reason to suppose that theworld is not statistically open.35 As for the

other categories: all explanations, contextu-ally speaking, are elliptical; and the deduc-tivist must show how partial explanationscan be universally applicable.

142:1 At the methodological level one ofthe most unfortunate consequences of thespell of actualism is the blurring of the realdifferences that exist between the various sci-ences: both in their subject matters and thedegrees to which they have achieved knowl-edge of them. The experimental scienceshave been able as a result of theoretical en-deavour and technical ingenuity, to carveout a chunk of the uncontrolled world anduse it as an object of inquiry. The non-experimental scientist has no such easy ac-cess. Now the view of the world as closedsets him off in the wrong direction—for itsets him looking for a complete descriptionof a given field. A view of the world as open


can, on the other hand, concentrate his en-deavours enormously. For it means that allhe has to do is to identify and describe (inways to be considered later) some interest-ing and significant object of inquiry, withoutsupposing that this will enable him to makedeductively successful predictions.

1 After J. Nicod, Foundations of Geometryand Induction, p. 219. See C. G. Hempel, As-pects of Scientific Explanation, p. 11.

2 P. W. Bridgman, The Logic of ModernPhysics, esp. pp. 1–25.

3 Physics cannot be regarded as validly basedupon empirical data until [light] waves have beenexpressed as functions of the colours and othersense-data’, B. Russel, Mysticism and Logic, p.109.

4 See e.g. O. Neurath, ‘Universal Jargon andTerminology’, P.A.S. Vol. 41, pp. 127–48.

5 ‘The communication of scientific knowledgeinvolves description: that is, the mimetic repro-duction of facts in thought, the object of whichis to replace and save the trouble of new experi-

ence. This is all that natural laws are’, E. Mach,Popular Scientific Lectures, p. 192.

6 G. Ryle, The Concept of Mind, esp. pp. 116–22. Cf. also C. S. Pierce, Collected Papers, Vol.2, p. 354; F. P. Ramsey, The Foundations ofMathematics, pp. 194–7; and S. Toulmin, ThePhilosophy of Science, Chap. 3.

7 R. Carnap, The Logical Foundations ofProbability or ‘Testability and Meaning’, Read-ings in the Philosophy of Science, eds. H. Feigland M. Brodbeck, pp. 47–92. —

8 K. R. Popper, The Logic of Scientific Dis-covery, esp. Chap. 4 or Conjectures and Refuta-tions, Chap. 10.

9 G. Ryle, op. cit., p. 117.10 K. R. Popper, The Logic of Scientific Dis-

covery, p. 59.11 See esp. C. G. Hempel, op. cit., Chap. 12.12 See E. Nagel, op. cit., Chap. 11.13 p. K. Feyerabend, op. cit., p. 164.14 T. S. Kuhn, op. cit. Chap. II.15 Poincare is widely regarded as the founder

of conventionalism in the philosophy of science.For a conventionalist interpretation of Popper


see I. Lakatos, ‘Criticism and the Methodology ofScientific Research Programmes’, Criticism andthe Growth of Knowledge, eds. I. Lakatos andA. Musgrave, pp. 104ff.

16 See e.g. K. R. Popper, Conjectures andRefutations, Chap. I and passim.

17 See e.g. P. Duhem, The Aim and Structureof Physical Theory, Chaps. 9–10.

18 Cf. A. Donagan, ‘The Popper-Hempel The-ory Reconsidered’, Philosophical Analysis andHistory, ed. W. H. Dray, pp. 136–8.

19 K. R. Popper, Objective Knowledge, p. 21and passim.

20 Cf. M. Scriven, ‘Explanation, Predictionand Laws’, Minnesota Studies in the Philosophyof Science, Vol. III, eds. H. Feigl and G. Maxwell,esp. pp. 176–90.

21 See e.g. A. Grunbaum, ‘Temporally Asym-metric Principles, Parity between Explanationand Prediction and Mechanism versus Teleol-ogy’, Philosophy of Science Delaware Seminar,Vol. I, ed. B. Baumrin, p. 73.

22 The interesting question is then of courseshifted to that of whether complete explana-

tions must be ‘deductive’. Most of the earlyobjections to the deductive model turned onthe non-availability of generalisations connect-ing events like the cracking of radiators or mis-sile failures. Although this was no doubt encour-aged by the way in which its advocates presentedit, it was somewhat disingenuous of its criticsnot to realise the possibility of sophisticated re-formulations of it. Mandelbaum, for example,has argued correctly that such events must beregarded as complex and analysed into compo-nents (M. Mandelbaum, ‘Historical Explanation:The Problem of Covering Laws’, History of The-ory, Vol. I, pp. 229–42). However he still seesexplanation as depending upon a knowledge oflaws (which he interprets in the Humean way)covering the component events (ibid., p. 241);and given this, the complex event itself still re-mains deductively predictable. I have argued, bycontrast, that the laws covering the componentsare normic and that they may involve referenceto radically different kinds (so that they cannotbe incorporated within a single theory). Hencethe complex event, even though completely ex-


plained, may not be deducible.

23 On the contrary he repeatedly emphasisesthe essential similarity in the logical form of thenatural and the social sciences. According tohim what is peculiar to the latter is its subjectmatter, and to history its interest in the partic-ular (see K. R. Popper, The Poverty of Histori-cism, p. 143). But the pattern of explanation isthe same.

24 See e.g. ibid., p 3.

25 See e.g. K. R. Popper, Conjectures andRefutations, p. 339.

26 K. R. Popper, Poverty of Historicism, pp.vi-vii.

27 Indeed one could even go so far as to saythat the latter depends upon the former; thatis, that it is only in so far as the effects of fu-ture knowledge are unpredictable on nature thanthey are unpredictable on men, in as much asall human action has a physical aspect but theconverse is not the case. A social closure pre-supposes a natural one but not vice versa; therecould be a nature without men, but not men

without a nature.28 D. Davidson, ‘Actions, Reasons and

Causes’, Journal of Philosophy 1963, reprintedin ed. A. White, Philosophy of Action, p. 92

29 Ibid, p. 92. See also D. Davidson, ‘CausalRelations’, Journal of Philosophy 1969, p. 701.

30 D. Davidson, ‘Actions, Reasons andCauses’, p. 92.

31 D. Davidson, ibid., p. 93.32 C. G. Hempel, op. cit., pp. 360, 362 and

363.33 C. G. Hempel, op. cit., pp. 376ff and pp.

415ff.34 M. Scriven, ‘Truisms as the Grounds for

Historical Explanation’, p. 465.35 The question of the closure of systems is

distinct from that of their statistical or non-statistical properties. The latter turns on the de-ducibility of single as distinct from mass events;whereas the former turns on the stability of em-pirical relationships. Clearly statistical systemscan be open or closed.

Chapter 3

The Logic of ScientificDiscovery

3.1 Introduction: Onthe Contingency ofthe Causal Connec-tion

143:1 In Chapter 2 I assumed the existenceof a body of knowledge and asked how it

196

3.1. INTRODUCTION: ON THE CONTINGENCY OF THE CAUSAL CONNECTION197

could be applicable to the world. My partic-ular concern was to establish its universality(transfactuality). I now want to turn to thequestion of how such knowledge, given thatit is transfactually applicable to the world,comes to be produced; and in particular tothe question of how law-like statements cometo be established as necessary. My concernshifts here then from the synchronic to thediachronic aspects of science, and in partic-ular to the question of how, in the social ac-tivity of science, natural necessity comes tobe ascribed. In the course of this chapter Iwill consider to what universality and neces-sity is properly ascribed, and what must bethe case for these ascriptions to be possible.

143:2 In order to show how the conceptof natural necessity is possible I will needto turn from a critique of the ontology ofclosed systems to a critique of the ontology

of atomistic events that implies it; and hencefrom a critique of the idea of the actualityof the causal connection to a critique of theidea of its contingency. In Chapter 4 I willask what accounts for the assumption of theatomicity of the events conjoined that entailsa closed system and generates, in its wake, ahost of philosophical problems.

143:3/o The connection between my con-cerns in this and the preceding chapter isclear. For once an ontology of atomisticevents is constituted, it follows that, for gen-eral knowledge to be possible, events must bealways conjoined (under appropriate descrip-tions) and never connected.1 That is, orderin the world must consist of an unfailing orinvariant order of the co-existence of eventsin space and their succession in time. Con-versely once it is appreciated that events,though caused (and consisting in transfor-

198 CHAPTER 3. THE LOGIC OF SCIENTIFIC DISCOVERY

mations), are very rarely conjoined, it canbe seen why order in the world must bepitched at a level categorically distinct fromevents. Now I have argued in effect thatwe produce conjunctions to discover con-nections and apply connections in a worldof non-conjunctions; so that events, thoughrarely conjoined, are sometimes connected.In this chapter I want to consider the na-ture of the connection that holds betweenevents (when it does) and the nature of thenecessity implicit in the concept of law. Iwill thus be shifting my attention from thedifferentiation of the world as such to thenature of the stratification that, if we areto render intelligible the experimental es-tablishment and practical application of ourknowledge, it implies. Science attempts, Iwill argue, in its essential movement, to cap-ture the stratification of the world. In or-

der to describe this movement I will needto reconstitute the other dimension of theCopernican Revolution in the philosophy ofscience, viz. the transitive (or sociological)dimension in which men come, in their socialactivity, to acquire knowledge of the endur-ing and transfactually acting mechanisms ofnature, in virtue of which some but not othersequences of events are necessarily connectedand some but not other statements are uni-versally applicable. The idea that there areno necessary connections between matters offact occupies an analogous position in under-pinning the doctrine of the contingency ofthe causal connection, as the idea that thereare always descriptions for events such thatthe formula ‘whenever this, then that’ ap-plies does in underpinning the doctrine ofits actuality. And I will argue that just asfor science to be possible the world must be


open; so there must be necessary connectionsbetween matters of fact, if science is to bepossible.

144:1/o In Chapters 1 and 2 I have shownhow the intelligibility of the activities of theexperimental establishment and the practi-cal application of our knowledge presupposesthe categorical independence of causal lawsfrom the patterns of events, and how causallaws must be given an ontological basis in theenduring and transfactually active mecha-nisms of nature. Modern transcendental ide-alist philosophies of science, which are per-haps more influenced by Wittgenstein thanKant, stop at what is in effect the secondstage of a dialectic or process of discovery inscience, by refusing to allow (or inadequatelyinterpreting) the possibility of a realist inter-pretation of theory.

145:1 Thus there is in science a charac-

teristic kind of dialectic in which a regular-ity is identified, a plausible explanation forit s invented and the reality of the entitiesand processes postulated in the explanationis then checked. This is the logic of scientificdiscovery, illustrated in Diagram 3–1 below.If the classical empiricist tradition stops atthe first step, the neo-Kantian tradition seesthe need for the second. But it either deniesor does not draw the full (transcendental re-alist) implications of the third step. If andonly if the third step is taken can there be anadequate rationale for the use of laws to ex-plain phenomena in open systems (where noconstant conjunctions prevail) or for the ex-perimental establishment of that knowledgein the first place.

Question 26 Describe in detail thethree steps in scientific discovery.


145:2 Diagram 3.1. The Logic of Scien-tific Discovery

145:3/o Just as transcendental realism dif-ferentiates itself from empiricism by inter-preting the first stage of the dialectic as theinvariance of a result rather than that of aregularity, so it differentiates itself from tran-scendental idealism in its interpretation ofthe second stage. Both transcendental real-ism and idealism see the move from (1) to(2) as involving creative model-building, inwhich plausible generative mechanisms areimagined to produce the phenomena in ques-tion. But whereas for transcendental ideal-ism the imagined mechanism is imaginary,for realism it may be real, and come to be es-tablished as such. What is imagined may bereal; but what is imaginary cannot. ‘Imag-inary/real’ marks an ontological watershed;‘imagined/ known to be real’ an epistemic

one. Now what is imagined at t1 may comeat t2 to be known to be real. And for tran-scendental realism the move from (2) to (3)involves experimental production and con-trol, in which the reality of the mechanismspostulated in the model are subjected toempirical scrutiny. For transcendental re-alism that some real things and generativemechanisms must exist can be establishedby philosophical argument (their existence,and transfactual activity, is a condition ofthe possibility of science). But it is contin-gent and the job of substantive science todiscover which ones actually do. That is, itis the task of science to discover which hy-pothetical or imagined mechanisms are notimaginary but real; or, to put it the otherway round, to discover what the real mech-anisms are i.e. to produce an adequate ac-


count of them.

146:1 Science is a process-in-motion. It in-volves three distinct stages, which cannot beomitted or collapsed into one another with-out doing tremendous violence to our under-standing of science. But these stages cannotbe identified with moments of chronologicaltime; they are phases of science. It shouldbe noted that the move from (1) to (2) justbecause it involves the postulation of novelentities and processes cannot be given a de-ductive interpretation. But given this it canonly be justified in a non-pragmatic way ifwe hold out the possibility of a realist inter-pretation of some of the hypothetical entitiesetc. invoked to explain the behaviour. Suchan interpretation can in turn only be justifiedempirically if it is set in the context of theongoing social activity of science. Thus it isin the planning of future experiences rather

than in the ordering of present ones or thememory of past ones that our rational andempirical ‘faculties’, ‘whose unkind and ill-starred divorce’ Bacon saw as responsible forall the confusion in ‘the affairs of the humanfamily’,2 are most productively combined.

146:2/o It is only, I shall argue, if we al-low the possibility of the move from (2) to(3) that we can, in the end, uphold the legit-imacy of the move from (1) to (2). Moreoverit is only if we begin to see science in termsof moves and are not mesmerized by termi-nals that we can give an adequate account ofscience. In this respect much philosophy isstill in the same position as a Martian tryingto discover what trams are but able only toobserve them in open-air museums with chil-dren scrambling over them. It is the task ofthe philosophy of science to capture science’sessential movement, not to guess its eventual


destination.

147:1 Recent work in the philosophy of sci-ence has established (i) the fact of scientificchange and (ii) the poverty of a purely de-ductivist analysis of explanation. In this wayit has done much towards the establishmentof a conception of science as a critical socialactivity. The case for transcendental real-ism can, however, be strengthened by con-sidering the limitations of this work. Forunless these two insights are taken togetherand a new ingredient is added to the existingphilosophical mix they are, I think, vulner-able to positivist counter-attack. This newingredient must be in the field of ontology.The argument of Chapter 1 enables us to seewhy this is so. For the logical empiricismagainst which recent philosophy of sciencehas reacted contained not only an accountof science, but (implicitly) an account of re-

ality, of the world known by science. And itis in this unacknowledged ontological legacythat the weaknesses of both developmentslie. My aim in this chapter and the next isto pinpoint these weaknesses. And to showin particular why and how an adequate non-empiricist account of science, capable of ac-commodating the facts of scientific changeand structure, requires an ontology of thekind outlined in Chapter 1 and elaboratedin Chapter 2. Indeed, recent philosophy ofscience illustrates very well the kind of ‘on-tological tension’ that can occur when a fun-damental objection is made to a philosophi-cal theory without simultaneously question-ing that theory’s ontology. The general dif-ference between recent philosophy of scienceand transcendental realism could be summedup by saying that whereas recent philosophyhas asked merely what are the conditions of


the possibility of individual experience andfound an answer in the intersubjective worldof science, transcendental realism asks in ad-dition for the conditions of the possibility ofthe social activity of science, finding an an-swer in the intransitive world of things.

148:1 I will need in this chapter not onlyto show the necessity for the philosophicalontology of transcendental realism, but alsoto begin the development of the philosophi-cal sociology that I argued in 1.6 is presup-posed by any theory of science. Scientificdevelopment, I have argued so far, consistsin the transformation of social products, an-tecedently established items of knowledge,which may be regarded as Aristotelian ma-terial causes. Certain implications flow fromthis conception. First, that men never con-struct their knowledge from scratch. Itstands to them always as a given product,

a social transmit;3 which they must them-selves reproduce or partially transform. TheCopernican Revolution in the transitive di-mension of the philosophy of science thus hasthe profound implication that man never cre-ates, but only changes, his knowledge, withthe cognitive tools at his disposal. Secondly,what is to be changed, has first to be ac-quired. And what is acquired consists alwaysof an ensemble of theoretical and empiricalideas, so that knowledge can never be anal-ysed out as a function of individual sense-experience. Once this is grasped the groundsfor the atomistic ontology that generates theidea of the contingency of the causal connec-tion collapse.

148:2 Science then is an ongoing social ac-tivity which pre-exists any particular genera-tion of scientists and any particular momentof consciousness. Its aim is the production


of the knowledge of the independently exist-ing and transfactually active mechanisms ofnature. Corresponding to the criterion de-veloped in the intransitive dimension of thephilosophy of science, viz. the conceivabil-ity of a world without men, we thus have acriterion in the transitive dimension, namelythe inconceivability of knowledge withoutantecedents.

3.2 The Surplus-element in theAnalysis of Law-like Statements:A Critique of theTheory of Models

148:3/o It has often been held that a con-stant conjunction of events is not a sufficientcondition for a causal law. This may be be-cause it is regarded as incapable of sustain-ing the intuitively obvious and importantdifference between necessary and acciden-tal sequences or in Johnson’s time-honouredterminology between universals of law’ and‘universals of fast’.4 Or it may be becauseit is regarded as incapable of licensing what

3.2. THE SURPLUS-ELEMENT IN THE ANALYSIS OF LAW-LIKE STATEMENTS: A CRITIQUE OF THE THEORY OF MODELS205

it is intuitively felt causal laws do licence,namely counter-factual conditionals.5 It isnever seriously denied that we feel, and sci-entists act as if, some but not other se-quences of events are ‘necessarily connected’;so that we must possess the concept. Whatthe radical empiricist, in the form of Hume,denies is: (a) that there is any objective basisfor this distinction, i.e. that it correspondsto any real difference between the two se-quences of events; and (b) that there is anyjustification, apart from habit or custom, forour ascriptions of natural necessity and ac-cident.6 Most philosophers since Hume haveattempted to show how he was wrong in (b)without objecting to (a). I want to arguethat Hume was wrong in (a); and that it isonly if we can establish this that we can showwhy he was wrong in (b) also.

149:1 The radical empiricist challenge to

philosophers then is to provide an alterna-tive account of the ‘surplus-element’7 in theanalysis of law-like statements; that is, thatelement over and above the (presumed) con-stant conjunction that explains our ascrip-tions of necessity; and which will show how,and the conditions under which, a distinctionbetween necessary and accidental sequencesand the assertion of counter-factuals canbe rationally justified. The usual responseto this challenge consists in the attemptto locate the surplus-element in the state-ment’s ‘explanation’, and more particularlyin the ‘theory’ which explains it. Howeverthe terms ‘explanation’ and ‘theory’ cover agamut of philosophical positions, which mustnow be considered.

149:2/o The deducibility of a law-likestatement from a set of higher order state-ments is often regarded as a criterion of


‘explanation’.8 However if deducibility isthe only criterion for explanation and thesource of the surplus-element is its expla-nation there will be an infinite number ofsurplus-elements for any statement. Henceany statement can be said to be law-like onan infinite number of grounds!9 Deducibil-ity alone cannot explicate the distinction be-tween necessary and accidental or nomic andnon-nomic universals. Moreover additionalcriteria such as simplicity can only reducethe number of possible explanations for astatement which has already been identifiedas law-like. But they cannot be used to saywhich statements are law-like and so possessthe surplus-element. For even if there werea simplest explanation for every statement,there are no absolutely simple explanations.Thus such criteria can at best be used to ex-plain why we choose one explanation rather

than another, but not why one statementrather than another is regarded as law-like.10

150:1/o Of course it might be objectedthat when everything is explained all factualstatements will be law-like. But what wouldcount as an explanation then? Could it beanything other than an inexplicable constantconjunction of events, as in the case of Mill’sunconditional laws?11 If it could not, we areback with Hume, and have done nothingto allay the sting of the radical empiricistchallenge. If it could, some alternative non-Humean analysis of the ultimate or highest-order laws must be given which will showhow they, as uniquely qualified ‘explainers’,do possess a genuine surplus-element. Weare thus faced with the following dilemma:either explanation is achieved by subsump-tion under higher-order laws in which casethe problem is merely shifted, for a surplus-


element must be found for them if they are toqualify as ‘laws’; or an alternative analysis of‘explanation’ must be given, which does notidentify the explanans with a further set oflaws, and so provides room for the locationof a surplus-element in the analysis of laws,within the context of their explanation, atany one level.

151:1 It might be thought that it is inthe capacity of the law-like statement toyield successful predictions that the sourceof the surplus-element lies. But this will notdo without an analysis of the ‘capacity’ or‘power’. For the Humean it is the past andactual successes of the statement that count,not its potential ones. And these can at bestexplain, not justify, the surplus-element. Itis the surplus-element that must provide ourinductive warrant, if we have one: ratherthan the other way round. Moreover even an

accidental generalization is capable of yield-ing correct predictions, viz. as long as theconditions that account for it persist. Thissuggests that, even if we were to possesssome general inductive warrant, predictivesuccess alone could not differentiate neces-sary from accidental sequences or license theassertion of counterfactuals.

151:2/o It seems clear that if we are toget any further in our search for the surplus-element the idea of purely formal differen-tiae must be abandoned. Inductive consid-erations prove no better than deductive ones.For accidental generalizations may be induc-tively confirmed, just as they may be deduc-tively explained. In practice then the non-radical empiricist, if he is not to concede thegame, is forced to re-examine the accountof science that seems to render any non-Humean conclusion impossible. The funda-


mental fact about science that has been miss-ing from the discussion so far is the existenceat any moment of time of an antecedentlyestablished body of theory. And it is herethat the non-radical empiricist attempts tolocate the surplus-element. But can ‘theory’do what experience and deducibility fail todo, i.e. provide a rational ground for our as-criptions of natural necessity? The answerclearly depends upon the extent to whichthe former contains components irreducibleto the latter. And the onus is on the philoso-pher who attempts to locate the surplus-element in the systematic organization ofour knowledge or the capacity of a theoryto explain many different laws12 or to pre-dict novel kinds of facts13 to show how theirconcept of theory escapes Humean analysis.Goodman’s notion of entrenchment,14 for ex-ample, functions in exactly the same way as

Hume’s notion of custom and can no morejustify our attributions of necessity than thelatter could.

152:1 In short, unless theory contains el-ements irreducible to experience and truth-functional operations on it there is no basisfor a non-Humean theory of natural neces-sity.15 Thus the possibility of the latter de-pends upon some terms of the theory not be-ing explicitly defined in terms of experienceand/or some statements of the theory notbeing deductively connected and/or someideas of the theory being non-propositionalin logical (or non-sentential in linguistic)form. These establish the possibilities ofintensional relationships between predicates,non-deductive (e.g. analogical) relationshipsbetween ideas and non-propositional (e.g.iconic) ideas respectively as potential sourcesof necessity. It is the second of these that has


been most thoroughly explored; and it is toCampbell’s initial formulation of the theoryof models that I now turn.

152:2/o On Campbell’s view a theorymust contain not only a ‘dictionary’ corre-lating some, but not all, of the theoreticalconcepts with empirical terms but a ‘model’for the hypotheses or theoretical statementsof a theory, by means of which its hypothet-ical subject matter may be imagined to belike in some, but not all, respects the realempirical subject matter of some field whichis already known.16 On this view the surplus-element just is the model. Thus what distin-guishes Boyle’s law from a merely accidentalgeneralization is, according to Campbell, thecorpuscularian model informing the kinetictheory of gases. By means of this model gasmolecules are imagined to be, in certain re-spects, like billiard balls bouncing off each

other and exchanging their momentum byimpact. And it is in our prior understandingof this that the necessity of the gas laws ul-timately lies. Notice that for Campbell it isnot the mere availability of a theory or eventhe organization that the theory makes pos-sible (e.g. the fact that Boyle’s law, Charles’law and Graham’s law are all deductive con-sequences of the kinetic theory) but the in-terpretation theory explains.17

153:1 As a critique of the deductivist viewof the structure of scientific theories, as typ-ified by Mill, Duhem and Hempel, Camp-bell’s case is a strong one. The deductivist,he says, merely exhibits ‘the dry bones ofscience from which all the spirit has de-parted”.18 His project is to revitalize it. Hesees the driving force of science as the ex-ploitation of analogies in the conquests ofnew fields, without which neither theory nor


the range of facts could grow or the languagein which to state them develop.19 But is hiscase unanswerable? How does it fare whenfaced with the challenge of radical empiri-cism? Is it capable of providing an adequateaccount of the surplus-element in the analy-sis of law-like statements? To answer thesequestions we must look more closely at theterms of the modelling relationship whichis intended to provide the basis for a non-Humean theory of natural necessity.

153:2/o Now essential to Campbell’s cor-rection of the deductivist view of explana-tion is the idea that for the explanation ofa range of phenomena say Ea to have oc-curred the relationship between the theoryTa which explains the phenomena and fromwhich the latter is deducible must be sup-plemented and informed by another relation-ship. This is a relationship of analogy not

deduction; and it is by means of it that werender Ta intelligible to ourselves. See Dia-gram 3.2 below. According to Campbell theentities and processes postulated at Ta areunknowable; i.e. they do not constitute partof the phenomenal world described by sci-ence. Although we cannot know what pro-duces Ea we can imagine it to be like some-thing we do know. Such an act of imagina-tive daring need not be totally arbitrary. Forit is possible to conceive of principles of ana-logical, just as there are principles of deduc-tive or inductive reasoning. Only when wehave constructed a model can we be said tohave achieved scientific understanding. Thatis, not just saved the facts, preferably withelegance and economy, but explained them.Using the analogy provided by Eb a real orempirical phenomena can thus ‘enliven’ theabstract theoretical relationships from which


Ea is deduced. And Eb does this by standingin for or representing (in the sense of the Ger-man ‘darstellung’) the unknown causes ofEa. Explanation thus involves, centrally, thesubstitution in our imagination of a real orempirical relationship for an unreal or theo-retical one.20 This is Campbell’s debt to em-piricist ontology: a debt that it summed upby his tacit acceptance of the concept of theempirical world. For on his theory Ta cannotbe, or come to be known as, real; though it isat any moment of time, and perhaps forever,unperceivable to us. For him theoretical en-tities, such as molecules, can only be said tobe ‘real’ by analogy with material objects.

154:1 Diagram 3.2 Campbell’s ‘Theory’

154:2 Campbell does not deny thatdeducibility is necessary for explanation,merely that it is sufficient. His theory maythus be regarded as providing an alternative

shave to Occam’s razor. Tyndall formulatedthe criterion for the selection of explanationsimplicit in Campbell’s theory as follows: ‘askyourself whether your imagination will ac-cept it’.21 Now such a criterion is clearly ca-pable of selecting a theory within a givenmetaphysical schema, such as that providedby the classical mechanical world-view. Butit is not capable of judging between differ-ent schemas, when it is precisely the natureor the limits of the imagination that is inquestion. To take an obvious example: Aris-totelian and Galilean dynamics are in con-flict over whether when a stone falls to theearth, the earth should be conceived as fixed(Aristotle, Ptolemy and Tycho Brahe) or asmoving (Copernicus, Giordano Bruno, Ke-pler and Galileo). Now, try as you may, thereis no neutral way of conceiving the fallingof the stone.22 Our imagination, although


not fixed, is either Aristotelian or Galilean.Tyndall’s criterion cannot help us to decidebetween the competing frameworks, becausewhat it in question is the nature of the con-cept in terms of which any motion has to beunderstood.

155:1 There is a similar break involved inthe transition from Newtonian to Einsteiniandynamics. Part of the trouble with currentmicro-physics is that our imagination can-not accept it, and yet we have every reasonto believe it to be true. If Tyndall’s criterionwere acted upon it could have effects on sci-entific practice as conservative and dogmaticas the consistency and meaning-invarianceconditions of classical empiricism. A new sci-entific ontology or a fundamental change inscientific concepts may transform our con-ception of what is plausible. At such timesin the history of science it becomes necessary

for the scientist to stand Tyndall’s criterionon its head, and dizzily ask himself whetherhe can continue to accept his imagination.

155:2/o Although its inadequacy to dealwith fundamental scientific change is mostevident, Tyndall’s criterion is no less inade-quate to deal with the continuing processesof conceptual micro-adjustment, in whichour imagination is continually modified andextended, that are a part and parcel of theprocess of ‘normal science’. More gener-ally, it is always legitimate for scientists toask and sometimes possible for them to an-swer, questions about whether gases are re-ally composed of molecules or whether theearth really moves. Such questions cannotbe rephrased as questions about the plausi-bility of our conceptions. This would be, interms of Diagram 3.1, to reduce phase (1) tophase (2). Rather the normal procedure in


science is if we have a plausible conceptionto go on to ask whether it is true, which isto ask whether the entities and processes itpostulates are real, or only fictional. Plausi-bility is a prima facie criterion for a theoret-ical explanation. But is is neither sufficient,nor in the last instance necessary.

156:1 How does Campbell’s theory fare asa response to the challenge of radical empiri-cism? According to it, the surplus-elementin the analysis of law-like statements is themodel at the heart of the theory that ex-plains it. But for Campbell the model can-not prompt questions about the reality of theabstract entities and processes postulated inthe theory. For theoretical entities are bydefinition unperceivable and hence, given thefundamental equation of empiricist ontology,viz. real = empirical, cannot exist. Mod-els function then not as knowledge-extending

but as essentially pragmatic devices, servic-ing the needs of the understanding. Theoryinvolves a journey from one set of experi-ences Eb to another Ea. Because of this it isalways possible for the radical empiricist toask whether the journey is really necessary.Moreover, even if a way could be found ofshowing that some model is necessary, therewould seem to be no way of justifying thechoice of any particular one (given that theidea that its necessity could be demonstrateda priori is rejected as being inconsistent withthe fact of scientific change).

156:2/o To this it may be contended thatmodels are necessary not only as concep-tual crutches for the tender-minded and asheuristic devices for the young (which theradical empiricist may graciously concede)but for a theory’s growth and development,and in particular (so as not to beg the is-


sue by positing non-empiricist criteria of de-velopment) for the generation of facts em-pirically relevant for the theory but whichwould not have been forthcoming withoutit.23 But this only pushes the argument backa stage further. In a completed science mod-els would be dispensable. For, as Duhem hasput it, ‘to explain is to strip reality of the ap-pearances in which it is wrapped as in veilsin order to see this reality naked and faceto face’.24 When we have done this, whatmore can there be to do? The objection that‘explanations are practical context-bound af-fairs’25 either is covered by the heuristic roleallowed to models or depends upon the in-completeness of science, in which case theirnemesis is merely (if perhaps indefinitely)postponed.

157:1 We are thus forced inexorably backto a particular conception of reality, the only

‘world’ that Campbell’s account of sciencecontains: the world of Mach and Hume. Insuch a world causality is bare and invariantconjunction; and scientific knowledge con-sists, for its part, in ‘description, that is themimetic reproduction of facts in thought, theobject of which is to replace and save thetrouble of new experience’.26

157:2 Suppose now that arguments are ad-vanced to show that no science can ever becomplete in the requisite sense. Science stillremains, on the Campbellian conception, apurely internal process, locked in a dosedcircle of thought. Science is still a creatureof custom and habit, the only difference be-ing that the habit is now one of the imag-ination, rather than sensation. In virtue oftheir shared ontology Campbell is closer toMach and Tyndall to Occam than one mightthink.27 In neither case can the possibility


of major conceptual revisions be accommo-dated or the mechanism of scientific discov-ery be displayed.

157:3/o Let us apply to Campbell’s the-ory the litmus test for the adequacy of an ac-count of science developed in Chapter 1. Canit sustain the idea of the applicability of theconcept in question, viz. that of necessaryconnection, in a world without men? Theanswer is obvious. In the case of Campbell,as of Hume, there is still no difference, inde-pendent of men, between a necessary and anaccidental sequence of events. The Camp-bellian can at best talk of a nomically neces-sary statement; he cannot talk of a nomicallynecessary sequence. The attempt to locatethe surplus-element in the analysis of law-like statements in the imagination of men isa failure.

158:1 For transcendental realism the

surplus-element distinguishing a law-likefrom a non law-like statement is the conceptof the generative mechanism at work produc-ing the effect in question. Such mechanismsexist and act independently of men; so thatthe necessity can be properly ascribed to thesequence. Moreover as the world is open notall events will be connected by a generativemechanism; so that the transcendental real-ist can sustain a concept of natural accident.

158:2 Only a real difference between nec-essary and accidental sequences can justifyour distinguishing law-like from non-law-likestatements. Hence one cannot deny Humeanconclusion (b) (on page 149 above) with-out objecting to Humean conclusion (a), andthus to the ontology that implies it.

158:3/o Nowhere is the anthropocentric-ity of post-Humean philosophy more evidentthan in the notion that natural necessity


must be sought in the behaviour or natureof men. And nowhere is the displacementof rational intuitions more obvious than theattempt to locate structure in the imagina-tion of men. ‘Connection’ is, as Chisholmhas remarked, an ‘ontological category anda source of embarrassment to empiricism’.28

But it is not an irreducible one. For its basislies in the generative mechanisms of naturewhich connect events as cause and effect andwhich exist as the powers of things. Thusto assert a counterfactual is not to make ameta-statement29 (which would be to makea statement about its grounds), but to makea statement about the way some thing wouldhave behaved (exercised its tendencies, lia-bilities or powers) had the conditions in factbeen different. Theory is not an ellipticalway of referring to experience,30 but a wayof referring to hypothesized inner structures

of the world, which experience can (in waysto be explored in paragrpah 3 below) confirmor falsify. We are not locked in a dosed circleof thought; because there are activities, viz.perception and experimentations by meansof which under conditions which are delib-erately generated and carefully controlled,relatively independent cross-bearings on theintransitive objects of thought can be ob-tained. Such activities are not independentof thought, but their results are not impliedby them either.

159:1 Campbell’s achievement is to haveseen that scientific theory cannot be identi-fied with a deductive system erected on thebasis of a single set of experiences. But hemade two mistakes. He too, like the em-piricists, missed the essential point that sci-ence is essentially developing; so that the hy-pothetical mechanisms of yesterday may be-


come today’s candidates for reality and to-morrow’s phenomena. But behind this fail-ure also lay an inadequate intransitive di-mension, and in particular the absence of theconcept of objects apart from our changingknowledge and possibilities of perception ofthem. Campbell’s theory has been extendedin two ways. Some have rectified his firstmistake but not his second, and viewed sci-ence as a sequence of models, an unfoldingprocess of shifts in intellectual fashion. Oth-ers have developed his theory in a realist way.Harre, for example, has drawn attention tothe role of the existential questions promptedby the creative use of analogies in the devel-opment of science.31 By way of concludingmy discussion of Campbell’s theory I wantto sketch out such a dynamized realist ver-sion of it.

159:2/o In Diagram 3.3 below the dotted

lines now stand for relationships of deductionand the continuous lines for relationships ofanalogy (to indicate their reversed relativeimportance). Ta has come to be establishedas real, and in this case also is perceivable. Inthe course of this process facts Ea have beencorrected and now become facts E′

a. Ta nowprovides one of the sources for a new modeldesigned to explain phenomena Ea. And theprocess of checking its reality (which will al-most certainly modify our conception of it)has begun. Needless to say there will in gen-eral be more than one model for Ea. Thestate of chemistry c.1930 provides an illus-tration of the model. Ta is Prout’s hypoth-esis and Tα the theory of sub-atomic struc-ture. Ea → E′

a, consists in the eliminationof the impurities that dogged the verifica-tion of Prout’s hypothesis for over a cen-tury. And the new model might be the Bohr-


Rutherford model of atomic structure; whichconceived as a hypothesis about the internalstructure of atoms is, we now know, false.The source of such models may lie eitherin some general conceptual scheme (such asatomism in chemistry) or some other scienceor proto-science (such as the wave modelsof light and sound in particle physics). Thesubject of such models is the unknown butknowable intransitive structure of the world.It is by means of the experimental testing ofthe hypotheses suggested by already existingknowledge that new knowledge comes to beproduced.

159:3 Diagram 3.3 A Dynamic Realist De-velopment of Campbell’s Theory

160:1 The problem of the surplus-element,and Hume’s challenge, has another aspect.This turns on the question of what warrantwe have for distinguishing between cases of

genuine and pseudofalsification, and hencefor invoking the CP clause in defence of gen-eralizations in the former case. This callsinto question the necessity of deducibility,not just its sufficiency in the explanation oflaws.

160:2/o Science needs a concept of pseudo-falsification for three reasons, two of whichare epistemic and one of which is ontolog-ical. Firstly, because a theory may not beat present sufficiently refined or developed tocope with anomalous counter-instances; thatis to say, every theory needs a ‘protective-belt’ for its development.32 Secondly, be-cause the ‘facts’ may be wrong: either inthe simple sense that they are misrepresenta-tions of the phenomena or more profoundlybecause they depend upon false or inad-equate observational theories.33 As is wellknown, every new theory is faced with innu-


merable anomalies and counter-instances ofthese kinds. They form in a sense the staplediet of normal science. A successful theory isone which, like Newton’s, though it never re-solves them all and generates new ones in theprocess of their resolution, ‘turns each newdifficulty into a victory for its programme’.34

Thirdly, science needs a concept of pseudo-falsification because a countervailing causeor interfering agent may be at work gener-ating the ‘counter-instance’. It is only un-der closed conditions, as we have seen, thata theory can be given a fair test or that acrucial experiment—Bacon’s ‘instance of thefingerpost’35—becomes possible.

161:1 The problems of the necessity anduniversality of law are indeed inextricablylinked, but not in the way Hume thought.For if the surplus element in the analysis oflaw-like statements is the concept of a gen-

erative mechanism at work and this conceptis irreducible to that of a sequence of eventsthen it is quite rational to uphold an ontolog-ical distinction between cases of genuine andpseudo-falsification (in which, as exemplifiedby the case of Prout’s hypothesis referred toabove, our epistemic distinctions too may begrounded). For we may readily allow thatthe generative mechanism in virtue of whichnatural necessity is ascribed is not under-mined by the instability of the conditions un-der which it operates. So that if a law hasbeen confirmed under closed conditions andthere is no reason to suppose that the gener-ative mechanism at work in those instanceshas ceased to operate, the law that the con-cept of the mechanism grounds may be sup-posed to continue to apply outside the condi-tions under which it was confirmed, whetheror not the consequent of the statement hap-


pens to be realized.

161:2/o By now it would, I think, be gen-erally agreed that models play some cogni-tive role in science and that there is a fea-ture about such models which renders themirreducible to the experiences that they arein some way intended to embroider or ex-plain. (This feature is, I have argued, typi-cally an idea of a mechanism which would, ifit were real, generate the phenomena in ques-tion.) But the representatives of the threetraditions in the philosophy of science differradically in their interpretations of the sta-tus and role of such models, and of the irre-ducible concept that constitutes its essentialcore.

162:1 The classical positivist view is thatit is merely a heuristic device (Duhem,Hempel and Brodbeck). This is liable toencourage the view that the rationale for

distinguishing necessary from accidental se-quences is solely pragmatic; that it is, as itwere, a question of our greater attachmentto the former (Quine),36 or of the deeper en-trenchment of their predicates in our con-ceptual system (Goodman). Similarly it en-courages the idea of the CP clause as a de-vice that can be relaxed or invoked, switchedoff or on, according to whether or not we areprepared to forego the falsified law-like state-ment. This view carries the implication ofcourse that the use of the CP clause is boundto be more or less arbitrary or dogmatic.37

And this in turn creates the Kuhn-Popperproblem of the functions of dogma.

162:2 The concept of the generative mech-anism may be given a firmer status, and thedistinctions it grounds a better rationale, byseeing its function as concerned essentiallywith the development of science. Protec-


tion from pseudo-falsification then becomesprotection from too easy or too early falsifi-cation; that is, before the full potentialitiesof the theory have been developed (Lakatosand Feyerabend). This view allows that ourknowledge is structured—that it contains, asit were, layers of different age. The condi-tions of knowing are here explicitly distin-guished from the conditions of being. Butpositivism still provides the underlying ac-count of the world. And because of thisthe rationale of the concept of the genera-tive mechanism, which forms the heart oressential core of the theory, is still more orless pragmatic, still science- or knowledge- orman-dependent.

162:32/o The third position consists incoming to see not just our knowledge butthe world itself as structured and differenti-ated. According to this conception, which

is that of transcendental realism, science isconcerned neither with the incessant accu-mulation of confirming facts (or the inces-sant search for falsifying ones), nor evenwith its own growth and development, butrather with the understanding of the differ-ent mechanisms of the production of phe-nomena in nature. Thus it allows that un-der certain conditions the concept of the gen-erative mechanism at work may be given arealist interpretation as a representation inthought of the transfactually active causalstructures of the world. The possibility ofsuch an interpretation supplements internalconsistency and contextual plausibility as aconstraint on the possible forms of theoreti-cal advance; and it constitutes the ultimategoal of all theory construction.

163:1 Now empirical realism generates thefollowing dilemma: Either theoretical enti-


ties refer ultimately to experience, in whichcase they can be eliminated. Or theoreti-cal entities constitute experience (in wholeor in part), in which case they cannot beeliminated, but must, given the equation ofempirical realism, constitute the world (inwhole or in part). Now as long as an ontol-ogy based on the category of experience isretained there can be no grounds indepen-dent of man for ascribing necessity to somebut not other statements. On the first hornthis generates the problem of what justifiesour belief that the future will resemble thepast, or the unobserved the observed, i.e. theproblem of the induction. But on the secondhorn it generates the problem of what jus-tifies the assumption of intellectual confor-mity. And, on this horn, scientific change, oreven dissent, actually constitutes (in wholeor in part) a breakdown in the uniformity of

nature!

3.3 Natural Necessityand Natural Kinds:The Stratificationof Nature and theStratification of Sci-ence

163:2 In the process of the establishment of alaw of nature three questions may be asked:

163:3 (i) is there an empirical regularitywhich constitutes a prima facie candidate fora law?

163:4/o (ii) is there some reason, otherthan the regularity, why the predicates in-

3.3. NATURAL NECESSITY AND NATURAL KINDS: THE STRATIFICATION OF NATURE AND THE STRATIFICATION OF SCIENCE223

stantiated in the law-like statement shouldbe conjoined?

164:1 (iii) is this reason located in the en-during powers of things and the transfactu-ally active mechanisms of nature?

164:2 If the answer to (i) is yes we havewhat might be called a ‘protolaw’.38 If theanswer to (ii) is yes we have strong groundsfor a law. If the answer to (iii) is yes wehave a law. Typically of course the reason inquestion in (ii) will be provided by a modelof the connection between antecedent andconsequent, putative cause and putative ef-fect. The transition from (ii) to (iii) typi-cally occurs when a realist interpretation ofthe mechanism posited in the model becomesacceptable.

164:3/oo The answers to (i)–(iii) corre-spond of course to three levels of criteriafor law, viz. those specified by the classical

empiricist, transcendental idealist and tran-scendental realist philosophies of science. Atthe Humean level laws just are empirical reg-ularities. At the Kantian level both (i) and(ii) must be satisfied. Here we have whatmight be called the dual criterion theory oflaw.39 I have already noted its vulnerabil-ity to Humean counterattack. At the levelof transcendental realism, a distinction isdrawn between the empirical identifiabilityand the universal (transfactual) applicabilityof laws; and the latter is seen to be a condi-tion of the possibility of the former. As theapplication of laws in open systems is jus-tified, and presupposed by the intelligibilityof experimental activity, the existence of anempirical regularity or a constant conjunc-tion of events is now not even necessary forthe ascription of a law (see Table 3.1 below).I have argued that it is only at this level that


a distinction between necessary and acciden-tal sequences can be sustained. A sequenceEa.Eb is necessary if there is a generativemechanism M such that whenever Ea, Eb

tends to be produced; a sequence is acciden-tal if this is not the case. Their differenceis represented in Diagram 3.4 below. Mostevents occur in open systems and must beconceived, as argued in 2.6 above, as ‘con-junctures’. This is illustrated in Diagram3.5 below. Necessity as such, like universal-ity, is thus ascribed essentially to the activ-ity of the mechanism; and only derivativelyto some particular event sequence. For theresult of the activity of the mechanism willin general be co-determined by the activityof other mechanisms too.40

164:3 Table 3.1 Status of Constant Con-junction of Events Necessary Sufficient forLaw classical empiricism Y Y transcenden-

tal idealism Y N transcendental realism NN

165:1 Diagram 3.4 (not typed in)


166:1 Now these three levels of criteriagenerate and are generated by different viewsof science. Thus whereas the classical em-piricist will ask merely:

166:2 (i)* is there a regularity such thatwhenever C then E?

166:3 The transcendental idealist will askin addition:

166:4 (ii)* given a regularity, is there anexplanation such that we can render it in-telligible to ourselves that whenever C thenE?.

166:5 The transcendental realist will how-ever, after making an essential correction, goone step further and ask:

166:6 (iii) out of the plausible explana-


tions for this regularity, is there one whichcorrectly describes the mechanism by meansof which, upon the occurrence or obtainingof C, E tends to be produced?

166:7 That is to say, the transcendentalrealist will demand that models be testednot just for plausibility but for truth; i.e.for their adequacy in correctly describing thereal generative mechanism at work (if theconnection between C and E is necessary)such that when C occurs, E tends to be pro-duced (is produced in the absence of inter-fering causes or the transformation of M).That real things and generative mechanismsmust exist can be established by philosoph-ical argument. It is the job of the scientistto discover which ones actually do. Giventhe identification of some prima facie non-random pattern in nature or protolaw thescientist thus builds up ideas of various plau-

sible hypothetical mechanisms by the cre-ative employment of his imagination (cf. Di-agram 3.6) and subjects these ideas to rigor-ous theoretical criticism and empirical test.These three phases of science are of coursethose represented in Diagram 3.1 on page114 above.

166:8/oo For the transcendental realistthen a model has a relationship with its sub-ject as well as with its source. And it iswithin the nexus formed by this double ar-ticulation that new knowledge is produced.For new knowledge is doubly articulated, ar-ticulated in two dimensions (transitive andintransitive): it is a socially produced knowl-edge of a natural (man-independent) thing.It is this bipolarity that a model expressesin standing in two sorts of relationship: a re-lationship of analogy with its source; and arelationship of adequacy (when it is) with its


subject matter. Many philosophical prob-lems arise from a misunderstanding of thesecond relationship. It is not a relation-ship of correspondence; the terms of the re-lationship are not necessarily like each other,though pictures and iconic models may playan important role in scientific thought.41

Moreover there are no general philosophi-cal criteria for such judgements of adequacy;they are necessarily intrinsic to the particu-lar science concerned. Analogy is one of thepossible relationships that models may havewith respect to their source. The existence ofthe first type of relationship (in the transitivedimension) is important in establishing botha constraint on the number of possible ex-planations42 and an indispensable means oftheir production. It cannot be described atall adequately as one of coherence; and hereagain no general philosophical criteria can

be laid down for it. Science is work that re-quires creative intelligence, and there can beno mechanical surrogate for that. The ideaof an automatic science is a will-o’-the-wispthat the philosophy of science has pursued,with damaging consequences, since Bacon’ssearch for a ‘sure and certain method’ thatwould eliminate the need for human thought,which of course inevitably entails the possi-bility of human error.


168:1/o Most science proceeds by way ofa two-tiered method designed to identify in-variances in nature, normally under condi-tions which are experimentally produced andcontrolled, and to explain them by referenceto enduring mechanisms.43 It is in the move-ment from the identification of an invarianceto the mechanisms and structures that ac-count for it that the logic of scientific dis-


covery must be found. Thus the observablereactions of chemistry, which are representedin textbooks by formula such as 2Na + 2HCl= 2NaCl + H2, are explained by referenceto the atomic hypothesis and the theory ofvalency and chemical bonding. The pat-terns which constitute the explananda of thetheory of valency are needless to say by nomeans superficially obvious or readily avail-able. Both the concepts and the substancesand conditions had and have to be workedfor, produced in the social activity of science.The theory itself sets out to describe thecausal mechanisms responsible for the overtbehaviour of the substances. Once its real-ity has been established (which justifies ourassuming that chemical bonding occurs andthe laws of chemistry hold outside the labo-ratory) and the consequences of the theoryhave been fully explored, the next task con-

sists in the discovery of the mechanisms re-sponsible for chemical bonding and valency.This has been explained in terms of the elec-tronic theory of atomic structure. Once thereality of this explanation has been estab-lished, science moves on to the discovery ofthe mechanisms responsible for what hap-pens in the sub-atomic microcosm of elec-trons, protons, and neutrons; and we nowhave various theories of sub-atomic struc-ture. The historical development of chem-istry may thus be represented by the follow-ing schema:

169:1 Stratum I 2Na + 2HCl = 2NaCl +H2 explained by Stratum II theory of atomicnumber Mechanism 1 and valency explainedby Stratum III theory of electrons and Mech-anism 2 atomic structure explained by Stra-tum IV [competing theories of sub- [Mecha-


nism 3] atomic structure]

169:2 It should be noted that the historicalorder of the development of our knowledge ofstrata is opposite to the causal order of theirdependence in being. No end to this processof the successive discovery and descriptionof new and ever deeper, and explanatorilymore basic, strata can be envisaged. Othersciences reveal a similar open-ended strati-fication. Geometrical optics is explained interms of Young and Fresnel’s wave optics;which is explained in terms of the electro-magnetic theory of light; which can be ex-plained in terms of the quantum theory ofradiation.44

169:3/o A general pattern of scientific ac-tivity emerges from this. When a stra-tum of reality has been adequately describedthe next step consists in the discovery ofthe mechanisms responsible for behaviour at

that level. The key move in this involvesthe postulation of hypothetical entities andmechanisms, whose reality can then be ascer-tained. Such entities need not be smaller insize,45 though in physics and chemistry thishas normally proved to be the case. Thespecies of explanation here identified itselffalls under a wider genus: in which the be-haviour of individuals is explained by refer-ence to their natures and the conditions un-der which they act and are acted upon (see2.3 above).

170:1 Now for the transcendental realistthe stratification this form of explanationimposes upon our knowledge reflects a realstratification in the world. Without the con-cept of real strata apart from our knowledgeof strata we could not make sense of what thescientist, striving to move from knowledge ofone stratum to knowledge of the next, is try-


ing to do: viz. to discover the reasons whythe individuals which he has identified (ata particular level of reality) and whose be-haviour he has described tend to behave theway they do. Without this concept the strat-ification of science must appear as a kind ofhistorical accident, lacking any internal ra-tionale in the practice of science (if indeedit is not denied altogether in a reductionistand ultimately phenomenalist account of sci-ence).

170:2 As it is clear that the hypotheticalentities and generative mechanisms imaginedfor the purposes of theory-construction mustinitially derive at least part of their meaningfrom some other source (if they are to be ca-pable of functioning as possible explanationsat all) theories must be already understoodbefore correspondence rules for them are laiddown.46 Equally this means that the descrip-

tive terms must have initially possessed ameaning independent of them. This enablesus to see how meaning-change is possible,and indeed if the independence of predicatesis denied, inevitable in the transitive processof science. Similarly we can see how knowl-edge of newly discovered strata may correctknowledge of less fundamental strata, as con-cepts and measurement techniques are re-fined. Now if changing knowledge of stratais to be possible the strata must not changewith our knowledge of them. Thus the con-cept of real strata apart from our knowl-edge of them is necessary if both the ideasof scientific structure and scientific change,which are central to recent critical philos-ophy of science, are to be intelligibly sus-tained. More generally, acknowledgementof the real stratification of the world allowsus to reconcile scientific discovery (of new


strata) with scientific change (of knowledgeof strata).

170:3/o Now the stratification of the worldmust be assumed by the scientists workingin any field, to be in principle unbounded.For it will always be possible for him thatthere are reasons, located at a deeper level,for the phenomena he has hitherto identifiedand described. But his knowledge may be inpractice bounded by semi-permanent tech-nical or conceptual problems or by the do-main assumptions of his particular science;or by the fact that reality is itself boundedat the level knowledge of which he has at-tained. However, if the stratification of theworld has an end, i.e. if there are ‘entities’which are truly ultimate—and I can see noreason for supposing this must be so—andthe scientist has achieved knowledge at thatlevel, he can never know that the level is ul-

timate. For it will still remain possible forhim that there are reasons, located at a stilldeeper level, for the causes of the phenomenahe has succeeded in identifying and describ-ing. I will return to this point below.

171:1 Now the only kind of necessity thatholds between events is connection by a gen-erative mechanism. But there are two otherconcepts of necessity applicable to the objec-tive world order: there is the necessity im-plicit in the concept of a law, i.e. in the activ-ity of a generative mechanism as such or theexercise of a thing’s tendencies irrespectiveof their realization; and the necessity implicitin the concept of a thing’s real essence, i.e.those properties or powers, which are mostbasic in an explanatory sense, without whichit would not be the kind of thing it is, i.ewhich constitute its identity or fix it in itskind. The first concept of ‘natural necessity’


is clearly derivative from the second, depen-dent upon the contingent feature of the sys-tem in which the thing’s behaviour occurs,viz. that it be closed (see 2.4 above). I amtherefore going to refer to the second as theconcept of natural necessity, and the third asthe concept of natural kinds. Knowledge ofnatural necessity is expressed in statementsof causal laws; knowledge of natural kinds inreal definitions. But natural kinds exist andnaturally necessary behaviour occurs inde-pendently of our definitions and statementsof causal laws.

171:2/o Now in the transition from knowl-edge of any one stratum to knowledge of thenext, knowledge of three levels of the objec-tive world order is progressively obtained:of relations between events, of causal lawsand of natural kinds. I am going to referto these three levels as the Humean, Lock-

ean and Leibnizian levels respectively. Thetranscendental idealist, as well as the classi-cal empiricist, is, in virtue of his ontologicalcommitment, restricted to the first level ofknowledge of the objective world order. Ishall argue that even at the Leibnizian levelscience remains empirical, so that the tran-scendental argument of Chapter 1 remainsvalid; and that even at that level the deduc-tively justified prediction of events is impos-sible, so that the critique of philosophy ofscience contained in Chapter 2 continues toapply with undiminished force. Moreover Ishall argue that the concepts, such as that ofnatural powers, that we need to render intel-ligible the transition to the Leibnizian levelremain categorically valid even at that level.

172:1 At the Humean level a patternis identified or an invariance is produced.(This, we know, empirical realism cannot


sustain.) We thus have a protolaw (at letus say Stratum I). This is to be explainedby reference to the circumstances and natureof the thing whose behaviour is described.The scientist never doubts for a moment thatsomething is generating the effect in ques-tion. His problem is: what is? That is, whydoes x behave the way it does, viz. B, in con-ditions C1 . . . Cn?

172:2/o The first step in the scientific ex-planation of B is to ascribe a power (or li-ability) of x to B, i.e. to do (or suffer) phi.This is to say, very roughly, that x does φin virtue of its nature N.47 The next stepis thus to investigate N (defining StratumII). This involves inter alia creative model-building and rigorous empirical-testing (cf.Diagram 3.1). As a result of this investiga-tion we may say x comes to do φ in virtueof its having a certain constitution or intrin-

sic structure, e.g. genetic constitution atomicstructure or electric charge. Now it is contin-gent that x has the nature (e.g. constitutionor structure) that it has. But given that ithas, It is necessary that it behaves the wayit does. One criterion of this is our capacityto deduce the tendency to B from N. This isthe Lockean level of knowledge. Note that atthis level it is still contingent that the thinghas the structure that it has.

173:1 Now at the third Leibnizian levelpossession of that structure or constitutioncomes to be regarded as defining the kindof thing that x is. Now it is necessary thatx has the structure it has if it is to be thekind of thing it is. It is no longer contin-gent that hydrogen is a gas with a partic-ular atomic structure; rather anything pos-sessing that structure is hydrogen. That is,the criterion for the application of the con-


cept ‘hydrogen’ ceases to be the lightest gasand become instead possession of that struc-ture. At this level the only contingent ques-tions are whether and where things of a givenkind exist.48 But note contingency still liesin the flux of the circumstances in whichthings act, so that events are still not de-ductively predictable. That is, the ‘contin-gency’ of events deriving from open systemsapplies even at the Leibnizian level, so thatlaws must still be formulated as tendencies(whatever their scope of application). More-over it is important to see that knowledge atthe Leibnizian level is, or may be, attainedempirically. We may discover, quite empir-ically, that the most important explanatoryproperty or real essence of hydrogen, identi-fied as the lightest gas, is its atomic struc-ture; and then attempt to express this dis-covery in a real definition of hydrogen. Once

more the importance of viewing science asa process in motion is clear. For if we stayat any one level, phase or moment of sciencethe idea that a definition may be arrived atempirically will appear absurd. If it is ac-cepted, however, the reason why the laws ofnature cannot be deduced a priori from self-evident axioms becomes clear. For the ax-iomatic base of a science at any moment oftime, at any stratum of reality, is somethingthat has had to be worked for, produced, aspart of the irreducibly empirical process ofscience.

173:2/o Scientists attempt to discoverwhat kinds of things there are, as well ashow the things there are behave; to capturethe real essences of things in real definitionsand to describe the ways they act in state-ments of causal laws. The real essences ofthings are their intrinsic structures, atomic


constitutions and so on which constitute thereal basis of their natural tendencies andcausal powers. Thus there is no conflictbetween explanatory and taxonomic knowl-edge. Rather, at the limit, they meet inthe notion of the real essences of the nat-ural kinds, whose tendencies are describedin statements of causal laws.

174:1 At the Leibnizian level statementsof law are substitution instances of necessarytruths about the individuals to which the re-fer. For any individual which did not behavein that way would not be an individual ofthat kind. They may thus be regarded asanalytic truths. But they are arrived at inthe transitive process of science a posteriori,by empirical means. Thus a fully dynamicphilosophy of science must take seriously thequestion ‘how is analytic knowledge arrivedat a posteriori possible?’ To this question I

will return in paragraph 5 below

174:2/o The situation at the Humean levelis rather like that faced by the citizens ofKoenigsberg who knew, from experience,that there was no way of crossing each of thetown’s seven bridges just once.49 See Figure3.1 below. At the Lockean level this fact isdeducible from the topology of Koenigsberg,given Euler’s theorem. At the Leibnizianlevel, there is a necessary truth about a cer-tain physical set-up, whether or not there is atown called ‘Koenigsberg’ or any town at allto which it applies. At the Leibnizian level,Mendeleyeev was able to deduce from hisPeriodic Table, interpreted as dealing withatomic number and valency, the propertiesof several new elements. But it remainedcontingent whether, and if so where, therewere elements in the world to which his pre-dictions applied. Certain chess games have


only one possible solution. But it remainscontingent whether they are ever played.

174:3 Figure 3.1. The Seven Bridges ofKoenigsberg (not typed in)

175:1 The concept of powers has played akey role in our analysis of science’s transitionfrom knowledge of one stratum to knowledgeof the next. To ascribe a power is to saythat a thing will do (or suffer) something,under the appropriate conditions, in virtueof its nature. This is not, as is so oftenclaimed, a pseudo-explanation50 or a purelyverbal formula.51 Rather it is an indicationof work to be done. Moliere’s doctors in LeMalade Imaginaire have often been ridiculedfor speaking of opium as possessing a ‘dormi-tive virtue’. But in doing so they left openthe possibility of an investigation, at somefuture date, into the nature of opium with-out committing themselves to what would

doubtless have been, for them, a rash conjec-ture at the time. Moreover it is far preferableto the Humean alternative, viz. that when-ever men smoke opium they fall asleep. Forin the first place, the latter is untrue. Sec-ondly, it is less informative. It might be acomplete accident that everyone in the denis asleep: the powers formula rules this out.It says that there is something about opiumin virtue of which when men smoke it theytend to fall asleep. The connection is nec-essary. But it is only a tendency. Thirdly,the Humean formula is regulatively useless.The powers statement is by contrast quitesuggestive. For it indicates the need foran investigation into the chemical propertiesof opium and the way they induce sleep inmen.52 In context, it constitutes an open ad-mission of ignorance. The Humean, on theother hand, must pretend that once he has


his generalization there is nothing more tobe known. And if he should stumble upona higher-order generalization this can be, forhim, only accidentally related to the origi-nal one: there is no inner logic connectingthe two, or rationale by which science movesfrom the first to the second.

175:2/o To this it might be objected thatthe concept of powers does not figure in thediscourse of science. This is true. And thereason for it is of course that the scientist,unlike Moliere’s doctors, is never just con-tent to ascribe a power but moves immedi-ately to the construction of possible expla-nations for it with the paradigms and otherinstruments of thought at his disposal. Thatis his job. (Sometimes, however, when weare completely at a loss we do just ascribe apower.) The concept of powers is introducedprecisely to describe this normally instanta-

neous (or simultaneous) and unselfconsciousresponse of the scientist to the identificationof protolaws; it represents, if you like, an at-tempt to reconstruct the internal rational-ity of the inter-strata move. The concept ofpowers is not intended to figure in the dis-course of science, but in the discourse of thephilosophy of science (which is the former’srational reconstruction).

176:1 It should perhaps be stressed herethat the stages of my rational reconstructionof the process of scientific discovery repre-sent phases of scientific activity; they cannotbe identified with moments of chronologicaltime. Thus most scientific work must oc-cur, for reasons I will bring out in paragraph4, in the context of a research programmedesigned to show that on the supposition ofthe mechanism M the field of phenomena canbe rendered intelligible. Thus the identifica-


tion of a protolaw normally depends uponthe prior existence of a conjecture or a hy-pothesis of a mechanism intended to functionas a possible explanation for the presumedprotolaw.53

176:2/o To ascribe a power is to say thatthere is something about the thing, whichmay be unknown, in virtue of which it be-haves the way it does. The grounds for theascription of a power must thus be strongerthan the mere occurrence of a regularity. Forwe must possess some reason to suppose theconnection necessary (though in the limit-ing case this may just be the invariance ofan experimentally produced result). It is be-cause it indicates the power-ascriber’s beliefin the existence of a reason, located at thenext highest level of inquiry (in the natureof the thing), whether or not the reason iscurrently known, that the concept of pow-

ers, in pinpointing an essential moment inthe transition from knowledge of one stra-tum to knowledge of the next, plays sucha key developmental or strata-bridging role.In this way, a powers statement is a promis-sory note cashed in the development of sci-ence, a schematic explanation filled out inthe growth of our knowledge.54

177:1 It is worth noting that the structureof a powers ascription is well adjusted to ac-commodate both falsification (obviously, asthe hypothesized reasons may be subjectedto independent tests) and meaning-change(less obviously). If meaning change is tobe possible, some elements of meaning mustremain constant through the change. Nowif ‘x does B’ is analysed as ‘x is of such anature N [defining Stratum II] that it willdo φ in conditions C1 . . . Cn [defining Stra-tum I]’ we may allow that the meaning of


‘phi’ remains constant while the meaning ofthe N component changes between theories(and vice-versa). This applies even in thecase of simple descriptive observational pred-icates such as ‘blue’. For ‘x is blue’ may beanalysed as ‘x looks blue [defining StratumI] in virtue of its reflecting light of a certainwavelength [Stratum II]’. The simple theorythat things look blue because they are bluemay then be replaced by the scientific the-ory that they tend to look blue in normalcircumstances because they reflect light ofwavelength 4400A. Subsequently we may al-low the latter to define the scientific use of‘blue’; in which case of course it is no longercontingent that blue surfaces reflect light ofthat wavelength.

177:2 Now although the concept of pow-ers serves this essential developmental func-tion, it cannot be reduced to it. For when

we have climbed up to Stratum II, we can-not throw away the ladder, so to speak. Topursue the analogy, the ladder is a rope,not a wooden one. For to make a powersstatement is to make a categorical statementabout the nature of the thing situated atthe level to which we have climbed. It is tomake a statement about possibilities whichare possessed by the thing quite categori-cally, whether they are known (or actualized)or not. Dogs do not lose their power to barkwhen we understand how they do so, just asglass does not cease to be brittle when weknow its molecular structure.

178:1 The ontological bases of powers justare the properties that account for them;i.e. the natures in virtue of which they areascribed. Now in the transitive process ofscience such natures may come to be qual-itatively described. When this happens it


will of course initiate a search for the higher-order entities and mechanisms that accountfor them. But how does it happen?

178:2 In general, at any one level, indi-viduals must be identified and their normicbehaviour described. Now for a qualitativedescription of a thing or a dispositional ac-count of its behaviour it must be present tothe scientist’s senses and he must be ableto describe it correctly, i.e. as being of thekind or type that it is. This will normallydepend upon two kinds of work: practical(experimental and technical) work, in whichthe scientist’s causal and perceptual powersare augmented (the latter with the aid of theconstruction of sense-extending equipment,such as microscopes); and theoretical work,in which the scientist’s conceptual and de-scriptive powers are augmented. It is theaim of the former to produce the object, i.e.

to render the thing or behaviour directly ac-cessible to the scientist’s senses (so that itbecomes the possible object of an act of im-mediate demonstrative reference). And it isthe aim of the latter to produce the conceptof the object, so that the scientist is capableof an adequate description of it. Both arenecessary for a qualitative description.

178:3 It should be noted that the twokinds of criteria, viz. demonstrative andrecognitive, are distinct. For my incapac-ity to identify the chromosome structure bypeering down an electron-microscope doesnot mean that it is not a possible objectof an act of demonstrative reference. It ispresent to my senses, whether I recognizeit or not. Conversely to render it accessi-ble to my senses is an independent labour(itself only possible if some concept of it ispossessed), requiring great ingenuity, just as


experimental production and control does,when we are concerned with the descriptionof the law-like behaviour of some thing al-ready identified. The production of the ob-ject and the production of its concept arethus independent tasks, each essential to aqualitative description of a thing or accountof its behaviour. The thing must be thereand I must know what kind of thing it is, i.e.how to describe it; in general this will involvea theoretical redescription of it.

178:4/o Now it is important to realizethat though the production of the objectand the production of its concept are dis-tinct, the judgement that the object has beenproduced itself depends upon a tacit theoryof vision and the instruments according towhich its range is extended. The case of theelectron microscope illustrates this very well.In general it is the function of such back-

ground or auxiliary theories to specify theconditions under which an object of the ap-propriate type may be said to be present tothe senses. In this sense they constitute, asit were, the criteriology of empirical science.

179:1 It is clearly essential to the theoryof scientific development proposed here thatimagined entities may come to be establishedas real. Now an entity may be ‘theoreti-cal’ either in the sense that its existence isopen to doubt (theoretical1) or in the sensethat it cannot be directly perceived, eitherunaided or with the help of sense-extendingequipment (theoretical2). The same distinc-tion applies in the case of behaviour. Nowan entity (or mode of behaviour) may betheoretical1 at t1 and perceived and ade-quately described at t2, so that it then ipsofacto ceases to be theoretical1. The existenceof bacteria, initially conceived as minute hos-


tile micro-organisms, and molecules, initiallymodelled on material objects, came to be es-tablished in this way. This is typical of sci-ence and shows once more the importance ofviewing it as a process in motion.

179:2/o But if an entity cannot be per-ceived, i.e. is theoretical2, does this meanthat it cannot be known to exist, so thatit must be theoretical1? If this were thecase all theoretical2 entities would indeed behypothetical, and our knowledge would benecessarily confied to the domain of observ-able things, even if this were now regardedas an expanding class. Fortunately this con-clusion does not follow. For theoretical2 en-tities may be known to exist indirectly, viz.through the perception of their effects. Theparadigm here is the case of the detectionof radio-active materials by a geiger counter,of electricity by an electroscope, of a mag-

netic field by a compass needle. That thereis a difference between the cases of detectionand perception is clear. In the case of detec-tion the thing can be individuated only indi-rectly, i.e. via the spatio-temporal frameworkor through its effects on particular things; itcannot be the object of an act of immediatedemonstrative reference. Whatever the men-tal imagery we use to think of a magneticfield it can be present to us only through itseffects. On the other hand my incapacityto identify a bacterium under a microscopeas being of a particular type, or even as be-ing a bacterium at all, does not mean thatit is not present to my senses; and so ca-pable of functioning as the object of a pos-sible act of immediate demonstrative refer-ence, although ex hypothesi I am incapableof intentionally performing it.

180:1 It should be stressed that in the de-


tection case that something does exist pro-ducing the effect is not in question. Noris the fact that it exists and acts indepen-dently of its detection. To say ‘electricity iswhat electricity does’55 is to collapse pow-ers to their exercise. Electricity is not whatelectricity does; but what it can do. Themode of reasoning employed in inferring theexistence of causal agents through the osten-sion of their effects is thus perfectly proper.Hence though it is correct to say that whenwe cannot qualitatively describe the causewe know less about it than when we can(given that in the latter case we know thething’s causal powers as well) it is not trueto say that there is a cause is less certain. Itis just that in the detection case what we canknow about a thing is limited to its causalpowers.

180:2/o Now there are two possibilities

here. One is that there is a nature, suscepti-ble in principle to a qualitative description,as yet unknown, which is the bearer of itscausal powers. The other is that the na-ture of the thing just is its causal powers,as in the case of physical field theories. Atany moment of time a science may have toput down its ultimate entities just as pow-ers to produce effects, e.g. to affect observersand equipment, possible observers and pos-sible equipment, material things, in certainways. About such entities all the scientistknows is their powers. It always remainspossible that he will be able to achieve aqualitative description of them, and he muststrive to do so. On the other hand, it is alsopossible that such entities are their powers.The scientist can never dogmatically elimi-nate one of these alternatives in advance. Ifthere is a frontier to possible knowledge of


the world the scientist can never know whenhe has reached it. But whatever is responsi-ble for the world as manifest must possesscausal powers which are continually beingexercised; it must be co-extensive with spaceand continuous with time. It must be struc-tured and complex; it cannot be atomistic orevent-like. The concept of a field of poten-tial seems closest to meeting these require-ments.56 However it seems to me there is noreason in principle why there should not bestrata of fields (of perhaps radically differentkinds), forever unknown to us. It should benoted that only the identification, not the ex-istence of fields depends upon the existenceof material things in general. Here again theorder of dependence in being is opposite tothe order of dependence of our knowledge ofbeing. The ontological order is distinct from

the epistemic one.

181:1/o The general thrust of my argu-ment in Chapter 2 was against reduction-ism. How does this square with my empha-sis on strata of knowledge? It will be re-membered that I did not deny the possibil-ity of an explanatory reduction but stressed(a) the need for a well-defined reductans (sothat a reduction could not in general be ameans of acquiring knowledge of a higher-order or less fundamental stratum); and (b)that a reduction left the reality of the higher-order entities intact, at least in as much asthey were causal agents capable of actingback on the materials out of which they areformed (see 2.5 above). It is clear that Iwas there taking possession of causal pow-ers, and hence existence in time, as the mostgeneral criterion of reality. There is an asym-metry between space and time here. For


powers must be possessed and exercised intime, but they need not be localized at anypoint in space. Relations, for example, suchas that of spin (in physics) and marriage en-dure through time and have causal effects.But they have no position in space. Now ingeneral a reduction is possible because theentities in terms of which the behaviour ofthe thing is explained occupy a different vol-ume of space, either larger or (more usually)smaller. Thus the possibility of a reduc-tion implies in general that the individualsof the different kinds cannot be said to oc-cupy the same place at the same time andone not be part of the other. This gives us ageneral criterion which imposes limits on re-gresses of strata, i.e. upon the possibility ofa sequence of (explanatory) reductions. Forone could define a branch of science as a se-ries of theories within which this criterion

is satisfied. On it, quantum mechanics andchemistry would belong to the same branch.But electromagnetism and mechanics, neu-rophysiology and psychology and (it will beargued) psychology and sociology would be-long to different branches.

182:1 Changes of things are explained interms of unchanging things. If there areultimate entities they must be unchanging.Atoms have already been disqualified as pos-sible ultimate entities (see 2.3 above). Soultimate entities must be powers; that is, in-dividuals characterized solely by what theycan do. For if one could describe the chang-ing states or conditions in virtue of whichtheir powers were exercised they could notbe ultimate (unchanging). In the last in-stance to be is just to be able to do. Butthis does not rule out the possibility of ascience of cosmology (which would be con-


cerned with the distribution in space andredistribution in time of the ultimate enti-ties) or of irreducibly historical branches ofscience in which the ultimate entities wereAristotelian or even Strawsonian individuals.The transformation of the principles govern-ing such things would in general have to beconceived as conjuncturally determined opensystemic events (see 2.6 above). In this waya complex thing such as a person (or a so-ciety) could come to be the cause of its owntransformation.

182:2 Now it is because we are ourselvesmaterial things that our criteria for estab-lishing the reality of things turn on the ca-pacity of the thing whose existence is indoubt to bring about (or suffer) changes inits material constitution or the constitutionof some material thing. Space, for exam-ple, might be regarded quite abstractly just

as any system of relations in which objectsstand to one another. And we can conceivethe possession and exercise of causal pow-ers in time in ways, and at levels, foreverunknowable to men. We can never knowwhere we stand absolutely in the chain ofbeing. Despite this cosmic incapacity sci-ence has succeeded in identifying strata ofreality. Now a scientist never doubts for amoment that there are reasons for the be-haviour he has identified and described. Itis in the search for such reasons, at a deeperlevel of reality, at present known to him onlythrough its effects, that the essence of scien-tific discovery lies. This search necessitatesthe construction of both new concepts andnew tools. But, as what is produced mustpossess a material cause, the scientist standsfor his essential task, in two systems of so-cial relationships, depending necessarily on


the work of others.

183:1 I have argued that the concept ofnatural necessity is the concept of a realgenerative mechanism at work, a conceptwhich is applicable to the world quite inde-pendently of men. And it is in virtue of theirconnection by such a mechanism, of whichknowledge may be attained in the social ac-tivity of science, that necessity is properlyascribed to some but not other sequences.In paragraph 5 I will analyse and criticizesome objections to this concept of naturalnecessity and the related concept of naturalkinds. But I want to deal here with the fol-lowing basic objection to the account I haveproposed: If, as I have contended, at eachstratum or level of reality an entity is identi-fied and its behaviour is described what pos-itive advantages does this account have over

the traditional empirical realist ones?

183:2 I think it has at least four substan-tial advantages. First, it reveals the essen-tial movement of science. Second, it allowsroom for the location of a surplus-element,reflecting a difference independent of men, inthe analysis of law-like statements at any onelevel. Third, it alone is capable of sustainingthe ideas of the necessity and universality oflaws, which are necessary for the rationalityof theory-construction and the intelligibilityof experimental activity. Finally, it alone iscapable of accommodating the possibility ofthe existence of entities and the necessaryphase of the knowledge of entities which can-not be analysed as substances with qualities,but must be conceived as powers to produceeffects, powers which are possessed and maybe exercised quite independently of their de-tection. (Needless to say, these advantages


are not independent of each other.)

183:3/o Science never stops still for a mo-ment. At whatever level we look, it alwaysinvolves something more than the empiricalrealist concedes. For example, if we considerthe phase of the identification of a protolaw(which seems prima facie most susceptible toempirical realist analysis), we find the cate-gorical clause implicit in a powers ascription,representing the scientist’s instantaneous re-sponse to this situation, indicating his beliefin the existence of a reason, located at thenext highest level of inquiry, for the predi-cates being conjoined. Only the powers con-ceptual system is capable of giving an ac-count of the internal rationality of science,by which it moves from knowledge of onestratum to knowledge of the next, so display-ing the actual historical development of thesciences as something other than a sequence

of accidents.

184:1 Now it is our knowledge of the rea-sons at Stratum II for the behaviour at Stra-tum I that warrants our designating the be-haviour as necessary. But the reasons for thebehaviour at Stratum II cannot be collapsedinto the behaviour at Stratum I or an inter-pretation or model of that behaviour consis-tently with the intelligibility and rationality(respectively) of theory-construction or thepossibility of empirical test. Nor can suchreasons be glossed simply as more fundamen-tal regularities,57 if they are to be subjectto experimental confirmation (or corrobora-tion) .

184:2 I have already shown in detail thatthe empirical realist account of laws, andhence the ontology that underpins it, is de-fective. Laws, I have argued, cannot be in-terpreted as conjunctions of events, but must


be analysed as tendencies of things. If sci-ence is to be rendered intelligible the worldmust be seen as one of persisting things, ofdiffering degrees of structure and complex-ity, to which powers and tendencies are as-cribed; it cannot be reconstructed as a worldof atomistic events apprehended in sense-experience. Briefly, to summarize my ac-count of laws: To invoke a law I must havegrounds for supposing a generative mecha-nism at work. These comprise: (a) indepen-dent grounds, preferably under experimen-tally closed conditions, for the mode of op-eration of the mechanism; (b) grounds forthe satisfaction of the antecedent (or stimu-lus) conditions for the operation of the mech-anism on the particular occasion in ques-tion; and (c) the absence of specific groundsfor supposing a breakdown or transformationof the mechanism in that case. Generative

mechanisms, I have argued, must be anal-ysed as the ways of acting of things; and theiroperations must be understood in terms ofthe exercise of tendencies and causal pow-ers. Tendencies may be possessed unexer-cised, exercised unrealized, and realized un-perceived (or undetected) by men.

184:3/o Finally, the empirical realist can-not deal with the case of entities which justare their powers or about which all we knoware their powers. He thus rules out dog-matically, tout court the possibility of a cer-tain kind of entity and a necessary phase ofknowledge. In virtue of this he is no moreable to make sense of the frontiers of knowl-edge, than show the mechanism by which sci-ences if it can and when it does, will advance.

3.4. THE SOCIAL PRODUCTION OF KNOWLEDGE BY MEANS OF KNOWLEDGE249

3.4 The Social Produc-tion of Knowledgeby Means of Knowl-edge

This section is omitted in the reprint ofthis chapter in [ABC+98].

185:1 The basic conception of scientificactivity that I have been concerned to ad-vance here is that is (consists in or involves)work. Science, I have argued, must beconceived as an ongoing process of trans-formation, continually or essentially in mo-tion, in an attempt to capture (i.e. pene-trate and describe) the stratification of theworld. The logical structure of work is Aris-totelian. It depends, in particular, upon theco-presence in any given productive episodeof both a material and an efficient cause.

Science operates on given materials, includ-ing pre-existing theory and antecedently es-tablished facts, with given materials, i.e. bymeans of an ensemble of intellectual andtechnical tools (including among the formerparadigms, models, metaphors and analo-gies), producing new theories and facts.

185:2 Science is produced by the imagi-native and disciplined work of men on whatis given to them. But the instruments ofthe imagination are themselves provided byknowledge. Thus knowledge is producedby means of knowledge. The objects from,and by, which knowledge is generated arethus always themselves social products (asis the knowledge generated). Thus scienceas a process is always entirely intrinsic to‘thought’. However, by perception and ex-periment access to objects, viz. things andcausal structures, existing independently of


thought may be obtained.58 And of such ob-jects knowledge may be achieved. Science isnot an epiphenomenon of nature, for knowl-edge possesses a material cause of its ownkind. But neither is nature a product ofman, for the intelligibility of the scientific ac-tivities of perception and experiment presup-poses the intransitive and structured charac-ter of the objects of knowledge, viz. that theyexist and act independently of the operationsof men and the patterns of events alike.

186:1 Thus science, I have argued, presup-poses the ontological independence and thepossible disjuncture of the domains of thereal, the actual and the empirical at everystratum or level of reality. At each stra-tum scientists attempt to identify the enti-ties responsible for what happens at the lessfundamental stratum (their point of depar-ture) and describe their normic behaviour.

But knowledge of existence, I have argued,cannot be identified with demonstration ofit. Causal powers, for example, can only beknown, not shown to exist. Hence if, as Ihave suggested there are grounds for suppos-ing, the ultimate entities in any one branchof science are bare powers, they must nec-essarily be undemonstrable. However, undercertain conditions, some states of things maybe perceived, unaided or with the help ofsense-extending equipment; and some causesmay be demonstrated indirectly, i.e. throughthe ostension of their effects. But for an ex-istential or a dispositional claim to be con-firmed or corroborated the states (or effects)and behaviour must be recognised or iden-tified as being of the asserted type. Hencein general two kinds of criteria, viz. demon-strative and recognitive, must be satisfiedfor such a claim to be granted. Because


the theoretical and technical conditions un-der which such claims are made (and criteriaelaborated) are themselves developing, ourknowledge may be extended; and becausethey may be falsified if the criteria are notmet (or revised) our knowledge may be cor-rected.

186:2 The paradoxical air of talkingof the correction of knowledge vanishesonce the demand for extra-theoretical truthand intertheoretical synonymity is rejected.Progress, I shall contend, can be shown tohave occurred but only from some particularposition, some specific vantage point, as itwere, in theoretical time.

186:3/o Science is explanatory, not simplydescriptive. Explanation is achieved by ref-erence to enduring mechanisms. Such mech-anisms exist as the powers of things and actindependently of the conditions that enable

us to identify them. Thus there is a di-rect link between the dynamic realist thesisthat the things and causal structures of na-ture not only exist but act independently ofmen and the conception of science as a so-cial activity sui generis in which both thefacts and the conjunctions that, when at-tainable, provide the empirical grounds forcausal laws are seen as social products. Inclassical empiricism, in a subtle interchange,these ideas are crossed: so that facts andtheir conjunctions appear as naturally givenand things and causal structures as experi-ences of men. (In transcendental idealism,the former is seen as in part imposed by,and the latter as unknowable to, men.) Nowthe identification of the conditions of (knowl-edge of) being with the conditions of experi-ence in empirical realism leaves ‘theory’ witha very uncertain status. For it must be ei-


ther reduced to, or grounded a priori in somenecessary condition of, experience; so that itis either reducible or immutable. For tran-scendental realism theory is both irreducibleand mutable. It is always there and liableto change, as part of our socially innate in-tellectual endowment. It is this endowmentthat we must draw upon as we attempt todeepen our knowledge of the way things areand act in the world; and in so doing we cancontinually add to and modify it. The exis-tence of this stock, as a layered structure, isa necessary feature of any human cognitivesituation; so knowledge can never be seen asa function of individual sense-experience.

187:1/o The necessity for a scientific train-ing shows that knowledge is a social productand cannot be conceived as purely individualacquisition. For it always stands to the in-dividual as something that must be acquired

to be used (for scientific work). That sci-ence is ongoing implies that some individu-als do so. Knowledge shares a feature com-mon to many social products then: namelythat though it exists only in virtue of hu-man activity, it is irreducible to the acts ofmen. For any cognitive act to be possiblethere must be a material cause; some knowl-edge established, given to us, already pro-duced. No sum of individual cognitive actscan yield knowledge, for the first member ofthe series would already presuppose it. Ex-perience is, on the other hand, susceptible toa purely individualistic analysis; ‘mass expe-rience’ is clearly derivative and analysableas ‘the experiences of masses of individuals’.It can thus be seen that underpinning em-pirical realism is an epistemological individ-ualism. That knowledge is not analysablein terms of individual experiences does not


imply that it is not analysable in terms ofexperience. But that the latter is the casecan be seen by reflecting upon the consid-eration that the antecedent cognitive situ-ation of the individual would have at thevery least to contain one theoretical conjec-ture, viz. that there were experiences of oth-ers. Assuming that the category of experi-ence was allowed to apply intersubjectively,if all terms were explicitly (ostensively or op-erationally) defined all truths would be an-alytic and all falsehoods contradictory, as aclaim about the facts would be implicit inthe meaning of every theoretical conjecture,so that the point of appealing to experiencewould be lost; and theory would be incapableof growing and developing. In short, if an-tecedently established knowledge is to be ca-pable of functioning as a material cause, thelayered stock must contain some terms not

completely definable in terms of experience;i.e. it must consist of a web of empirical andtheoretical ideas.

188:1/o Recent work establishing that sci-ence has a transitive (or sociological) dimen-sion and some facts about its nature hasbeen widely regarded as shocking. That sci-ence is a social activity which shares manyof the characteristics of, and does not ex-ist in isolation from others; that it dependsupon a whole complex of institutions, someof which have little interest in knowledge‘for its own sake’; and that in particularcircumstances its fortunes can depend uponthe accidents of particular men raises seri-ous moral and political questions. Moreoverthere are some particularly disturbing fea-tures about current science. One need onlyinvoke, from the recent literature, the epi-thets of entrepreneurial science and shoddy


science, reckless science and dirty science,59

government science60 and mob science,61 re-pressive science, Stalinist science and theiranodyne anarchist science62 to appreciatethis. However these problems do not flowfrom the social character of science per se,i.e. the mere existence of a transitive dimen-sion, but from the present character of its so-cial character. The realization that sciencehas social problems could only be shocking ifone had been tacitly viewing it, in the styleof Hume, as a kind of behavioural responseto the stimulus of given facts and their con-junctions. This is the positivist concept ofbehaviourist or automatic science. It is aconcept which can itself be used to disguiseembarrassing facets and rationalize the prac-tice of a science.

189:1 Recognition of the transitive dimen-sion implies that scientific beliefs can no

longer be distinguished by their content. Forexperiences and the facts they generate mustnow be viewed as socially produced andwhat is socially produced is socially change-able. There are no absolutely privilegedstatements. The application of the cate-gory ‘empirical’ becomes relative and theory-dependent. Hence it cannot be used, with-out a degree of circularity, to establish thescientificity of one class of statements withrespect to another. Knowledge, viewed as atransitive process, has no foundation—onlya structure in time. The sciences have his-tories, which like all histories are character-ized by both continuity and change; and inwhich, as in all histories, certain events standout retrospectively as especially significant,e.g. the discovery of oxygen, the publicationof The Origin of the Species, the Michelson-Morley experiment. (Later we shall have


to inquire into the grounds for their signifi-cance.)

189:2 Now the fact that scientists donot possess a special attitude or a superiormorality does not mean that science does nothave a rationality of its own. Nor does thefact that scientific beliefs cannot be distin-guished by their content imply that scientificactivities cannot be distinguished by theirstructure or their aim. There are two er-rors here: the first is to suppose that sci-ence is not a social activity in the fullestsense (exactly what this entails we have yetto see). The second is to suppose that itis not equally a social activity, quite unlikeany other, sui generis: namely a social ac-tivity whose aim is the production of theknowledge, with the cognitive tools at itsdisposal, of the enduring and transfactuallyactive mechanisms of the production of phe-

nomena in nature.

189:3/o In paragraph 3 I described scienceas the systematic attempt to capture thestratification of the world. Only the conceptof a real stratification allows us to sustainthe idea of scientific progress, in a way whichis both non-inductivist and consistent withthe possibility of scientific change. Knowl-edge of new strata does not dissolve, thoughit may occasion a correction of, knowledgeof old strata. Nor does it render the lessfundamental strata illusory. We do not needthe metaphor to which so many writers havefound it necessary to resort in order to rec-oncile progress with change: viz. that of anasymptotic approach to the truth.

190:1 Two fallacies must be most assidu-ously avoided. The first is to suppose thatscience grows but does not change. Thesecond is to suppose that science changes


but does not grow. It is the fact that sci-ence grows, i.e. that in the transitive pro-cess of science new strata and dimensions(or branches) of reality are discovered, thatmeans that scientific change can be accom-modated as a fact of history without sac-rificing the idea of scientific progress. Butprogress can only be shown to have occurredfrom some substantive theoretical stand-point or position. There is no Archimedeanpoint outside theoretical time. But knowl-edge changes as it grows. For knowledge at anew level may lead to a revision, correctionor modification of knowledge at the previ-ous level. For what is explained is never the‘pure’ phenomena, but always the phenom-ena read in a certain way: i.e. facts. Thescientist seeks to describe the mechanismsgenerating the phenomena; but the resultsof his activity belong to the social world of

science, not the intransitive world of things.Does this mean that it is wrong to talk of thescientist explaining events describing mecha-nisms, etc.? No: provided we remember thatwhat is explained in any concrete scientificepisode is always the event known under aparticular description. This does not meanthat the event is, or that we must think of itas if it were its description. On the contrary,the ontological independence of the event isa condition of the intelligibility of its descrip-tion. But here, as elsewhere, it is the task ofphilosophy to analyse concepts, such as thatof an event, which can only be used syncat-egorematically in science.

190:2/o Scientific activity is continuous.This has meant that ‘refutations’ have nor-mally taken the form of ‘replacements’. Nowit has been pointed out that cases of both‘inconsistency’ and ‘meaning-change’ can be


drawn from the history of the sciences. Forexample, Newtonian physics corrected Ke-pler’s and Galileo’s laws;63 and the conceptsof ‘mass’ employed in classical dynamics andthe theory of relativity are radically differ-ent.64 Now given that these are the most ob-vious ways in which scientific changes occur,great care must be taken about the way inwhich they, and their relationship, are for-mulated. Kuhn, Feyerabend, and others,have claimed that theories may be so rad-ically different in meaning as to be literally‘incommensurable’. To this there is the ob-vious objection that if they were literally in-commensurable, i.e. shared no elements ofmeaning in common, it is difficult to see howscientists could have had grounds for prefer-ring one to another. It is clear that at themoment of ‘falsification’, when one theory isreplaced by another, some elements of mean-

ing must be shared in common.65 But thatthe subsequent divergent development of thetheories may result in their eventually be-coming ‘incommensurable’. So that ‘incon-sistency’ and ‘incommensurability’ refer todistinct moments of the scientific process.66

(Something similar must be true of the nor-mal process of education.)

191:1/o Theory without experiment isempty. Experiment without theory is blind.But in the historical development of the sci-ences experiment and theory are often outof step. Michelson and Morley did not seetheir experiment as a refutation of the other,and Michelson never in fact relinquished hisbelief in it.67 On the other hand Prout’s hy-pothesis could not be vindicated until theinvention of physical techniques of chemicalseparation, a century after its formulation.It is not just the experimental results but


what is done with them that counts. On theother hand, the nonfulfilment of experimen-tal expectations can always be explained interms of the deficiency of experimental tech-niques.

192:1 Scientific activity is itself differen-tiated into periods or better ‘phases’ (soas not to identify them with chronologi-cal time): viz. into (a) phases of discoveryand/or change and (b) phases of r applica-tion. Both are necessary. I use these charac-terizations in preference to the emotive andsomewhat misleading terms ‘revolutionary’and ‘normal’. (a) consists in the produc-tion of the knowledge of a new stratum orlevel and/or the radical revision of knowl-edge at the current one. This is often pre-ceded by the hint or glimpse of a new levelor by a crisis induced by the proliferationof anomalous facts of a particularly disturb-

ing kind. The discovery of X-rays illustratesboth these facets. Indeed Lord Kelvin ini-tially thought that Roentgen had devised anelaborate hoax.68 (b) consists in the appli-cation of the discovery and/or change to ac-count for (and perhaps correct) currently es-tablished facts and generate new ones.

192:2/o Needless confusion has been en-gendered by the failure to distinguish mod-els, theories, paradigms, etc. Very roughly,a theory is a model with existential commit-ment; that is, a model conceived, and meantto be taken, as true; i.e. a model in which theentities posited and mechanisms describedare conceived as real. It is relatively easyfor the scientist to invent models, but muchmore difficult for him to construct theories.There were several models of the aether, butnever a satisfactory theory of it. Diagram 3.7illustrates a schema for the development of


science. A general conceptual scheme (ab-breviated here to G.C.S.) or metaphysicalframework, such as that provided by atom-ism, ‘begets’ (logically, not temporally) a re-search programme, such as that associatedwith the attempt to explain phenomena byreference to the primary qualities of matter.The research programme in turn generatesa theory and/or a sequence of theories ei-ther intended for different fields (or differentstrata) and/or in competition with and re-placing each other. At the centre of theory-construction is the process of model-buildingand technical innovation required for the em-pirical testing of the various models. Theselevels are never in complete harmony. Somehypotheses, seemingly necessitated by thefacts, are always out of line with the theorygenerating the facts. Gravity, for example,could never be assimilated to the corpuscu-

larian metaphysical paradigm.

193:1/o Puzzles or problems are the con-crete working data of the scientist. His im-mediate task is their clarification and res-olution. For this he must use the toolshe has at his disposal: established results,facts, and theories, promising hypothesesand half-tried (or forgotten) ideas, the avail-able formal and technical equipment andusually some methodological (or heuristic)paradigm, concretely embodied perhaps insome exemplary piece of work. It is be-cause such paradigms are shared that thereare intersubjective criteria for assessing thescientific adequacy of his work. The scien-tist’s work normally takes place within a def-inite institution, a disciplinary matrix andis governed by what some continental writ-ers have called the problematic of his sci-ence (that is, roughly the structured field


within which alone meaningful questions canbe asked or problems posed, expressing thedominant theoretical concerns of the time).The research programme may have its ownmethodological paradigm, such as Newton’sPrincipia or Durkheim’s Suicide. Now ifthe problems generated by work at phase (a)cannot be resolved by the material availableto the scientist in his own field he must nec-essarily draw on another. This provides therationale for paramorphic model-building, arole for analogies and metaphors (the mod-els of discursive thought), leading to existen-tial novelty (e.g. ‘what are the punctuationmarks in a genetic code?’). Incidentally, thisalso helps to explain why theoretical inno-vations are often made by individuals origi-nally working in fields adjacent to the fieldthey innovate, as in the case of Dalton (ameteorologist); or with a strong ‘prejudice’

taken from an adjacent field, as in the case ofPasteur who was convinced, in opposition toorthodox opinion at the time, that fermenta-tion could not be caused by chemical agentsalone but had to be explained by dissym-metrical forces associated with the activityof living organisms.69

193:2 Diagram 3.7. The Internal Struc-ture of Science

193:3 Nb. so called ‘non-normal’ sciencecorresponds to phase (a) viz. of discoveryand change: in it models are invented andsubjected to empirical tests with the aim oftheory construction. so called normal’ sci-ence corresponds to phase (b) viz. of ap-plication: in it puzzles generated by phase(a) are resolved. the structure of theory-construction: models + facts + theories +techniques [within G.C.S. + Res. Prog.] –¿new theories. the structure of puzzle-solving:


facts + theories + techniques + method-ological (and heuristic) paradigms –¿ reso-lution of puzzles. critical science has no for-mal structure but may call into question anylevel, including established theories, success-ful research programmes and even the G.C.S.or metaphysical framework, leading to thereplacement or development of theory, re-search programme or G.C.S. (e.g. Mach, Ein-stein).

194:1/o Conceiving science as work read-ily lends itself to Aristotelian schematiza-tion.70 The material cause is antecedentlyestablished knowledge, facts and theories;the efficient cause is the methodologicalparadigm or generative theory at work in thetheoretical and experimental activity of men;the formal cause new knowledge, facts andtheories; and the final cause knowledge ofthe enduring and transfactually active mech-

anisms of nature. Now from the perspectiveadvanced here an event, such as the discov-ery of oxygen, is significant not just becauseit refuted the phlogiston theory of combus-tion, but because it constituted a decisivemoment in the transformation from one wayof doing chemistry, viz. that associated withthe theory of elective affinity, to another, viz.that represented by post-Daltonian atomicchemistry; that is, because it constituteda transformation in the ongoing activity ofchemistry.

195:1 Knowledge does not exist in a thirdworld.71 Rather, it exists in our world, em-bedded in the scientific community. Withoutmen there would be no knowledge, only itstraces. In this sense it depends upon men.But though it exists only in virtue of thethoughts and actions and products of men,it is irreducible to them. For though it would


not exist without the activity of some men,its pre-existence is a necessary condition ofthe activity of anyone. It is a public mixthat always antedates the individual. Nowit is not necessary that science should con-tinue, i.e. be ongoing. It is contingent thatit is. But given this men must reproduce (ormore or less transform) the knowledge thatis given to them. Men do not construct theirknowledge: they reproduce or transform it.This is another way of saying that any knowl-edge that there is must possess a materialcause. Now in general for scientific activityto be continuous all levels depicted in Dia-gram 3.7 must be represented. The troublewith social science, for example, is not that ithas no (or too many) paradigms or researchprogrammes; but rather that it lacks an ad-equate general conceptual scheme.

195:2/o I said in 1.6 that an adequate phi-

losophy of science would depend not onlyupon the development of an adequate philo-sophical ontology, but upon the developmentof an adequate philosophical sociology too.This must consist in an answer to the ques-tion: what must society be like if science (asa specific kind of social activity) is to be pos-sible? It must satisfy the desiderata of beinga structure irreducible to but present only inits effects. Society can only be known, notshown, to exist. It exists only in virtue ofthe intentional activity of men but it is notthe result (or the cause) of their intentionalactivity. Sociology and psychology thus con-stitute distinct branches of science, in thesense of the criterion developed in paragraph3 above. Sociology is not concerned withmasses of individuals or mass behaviour; butwith the persistent relationships between in-dividuals. Such relationships would not exist


without their relata but they do not dependfor their efficacy upon any particular relata,any particular named individuals.

196:1 Now the autonomy of sociology andpsychology accords well with our intuitions.Thus we do not suppose that the reasonwhy the war is fought is the soldier’s rea-son for fighting it, just as the reason whythe bar of chocolate is wrapped need not bethe chocolate wrapper’s reason for wrappingit (though it depends upon the latter). I donot have to know the laws of supply and de-mand to buy a mackintosh or to know thedeep structure of language in order to use it.The deep structure of language may indeedimpose limits (like natural structures) uponthe kinds of speech acts I can perform but itdoes not determine what I say. This concep-tion of social science thus preserves the sta-tus of human agency, but does away with the

myth of creation (logical or historical), i.e.the possibility of a methodologically individ-ualist reduction. It is not necessary that thatsociety should continue. But if it is to do sothen men must reproduce (or more or lesstransform) the structures (languages, formsof economic and political organization, sys-tems of belief, cultural and ethical norms,etc.) that are given to them. The Newto-nian revolution in sociology consists in com-ing to see that it is not necessary to explainsociety as such; but only the various struc-tures responsible for different societies andtheir changes. The problem of how men re-produce any particular society belongs to alinking science of social psychology. As soconceived, society may be regarded as an en-semble of powers which exist, unlike otherpowers, only as long as they are exercised;and are continually exercised via (i.e. in the


last instance through) the intentional actionof men.72

196:2/o Established facts are social prod-ucts. Understanding their logic may helpus to clarify the relationships between menand society and men and the world. Herethe metaphor of a reading may be used.Its adequacy depends upon the existence ofboth a given language and an independenttext. (But the metaphor is misleading in oneway: in that the text of nature exists inde-pendently of any language.) It is this thatmakes possible talk of a correct, rather thanjust commutatively successful, reading. Mennever create this language. For it alwayspreexists them. But it exists as an actual,i.e. ‘living’, language only in virtue of, andchanges with, their uses of it. Thus if soci-ety is represented by the model of a languageit may be regarded as a structure which is

always there; which men must reproduce orpartially transform; but which would not ex-ist without its ‘functionaries’. It is method-ologically incorrect to search for an efficientcause of society, though society depends nec-essarily upon the efficient activity of men.But a reading depends upon antecedent so-cial activity; the acquisition of a language bythe reader. It is in this sense that the factsalways depend upon social activity. In ex-perience the skilled scientist reads the worldas if it were a text in an attempt to under-stand the mechanisms of the production ofphenomena in nature. But his own readingdepends upon the mechanisms of the repro-duction and transformation of language, ofknowledge and of society.

197:1/o To return now to science: it iscontingent that science is continuous butgiven that it is men must reproduce (or


more or less transform) the knowledge thatis given to them. The condition that sci-ence be continuous is equivalent to the con-dition that all knowledge possesses a mate-rial cause. Hence the criterion of adequacy inthe transitive dimension of the philosophy ofscience is that the account of science shouldbe capable of sustaining the concept of it asan ongoing social activity. It is here thatdyadic theories of falsification, that is, the-ories that conceive falsification as consistingin a confrontation between a single theoryand a set of facts fail. If science is to be con-tinuous, refutations must be replacements;which means that always more than one the-ory must be involved. But, related to this,is an even more basic objection to fallibilismas such. For the refutation of any theorypresupposes the acceptance of the refutingobservation statement. If everything is con-

ditional nothing can be. If all knowledge is(equally) conjectural, no statement can berefuted. Of course it is always possible thatthe scientist is mistaken in any particular be-lief (and a good scientist is continually aliveto this possibility). But in order to demon-strate a mistake some proposition must beasserted (some theory accepted and frame-work worked within). In order to learn fromour mistakes we must know that (and when)we are mistaken. Lacking from fallibilism, asfrom classical empiricism, is the key conceptof knowledge necessarily possessing a mate-rial cause: antecedently established knowl-edges; science’s means of production. It isnot necessary that a scientist works withinan particular framework or accept any par-ticular theory; but it is necessary that heworks within (accepts) one.

198:1 Underpinning empirical realism is a


model of man in which men are seen as sen-sors of given facts and recorders of their con-stant conjunctions: passive spectators of agiven world rather than active agents in acomplex one. This model plays a role at leastas important as that played by the classicalparadigm of action and the celestial closurediscussed in Chapter 2. Together they forma complementary triangle (see Diagram 3.8)

198:2 Diagram 3.8. Three Sources of Em-pirical Realism

198:3/o Implicit in empirical realism is aconflation between a ground of knowledge,viz. experience, and the world. If experienceis to be capable of playing the role tradition-ally assigned to it of grounding our knowl-edge (in whole or in part) then the items ofwhich it is composed must be perfectly sim-ple and atomistic; i.e. insusceptible to fur-ther analysis or justification. But if it is to

define the world then the world must be sim-ilarly composed: of atomistic and discreteevents (or momentary states) independent ofeach other. If knowledge is to have its foun-dations in experience and experience is to de-fine the world then both the ultimate itemsand objects of knowledge must be atomisticand independent of each other. This createsthe problem of what grounds we can havefor moving from the observed to the unob-served, or from the actual to the possible(and thence to the counter-factual). Falli-bilism, which shares this model, can no lessescape this problem. For the refutation oftheory T at time t1 by an observation state-ment is consistent with its corroboration bythat statement after t1 unless we are justifiedin moving from the observed to the unob-served, and from the actual to the possible.(Unless induction is justified or nature is uni-

3.5. OBJECTIONS TO THE ACCOUNT OF NATURAL NECESSITY PROPOSED267

form we can never know that a ‘mistake’ isa mistake; so we can never put our mistakesbehind us.)

3.5 Objections to theAccount of NaturalNecessity Proposed

199:1 Having outlined the principal advan-tages of my account of natural necessity andnatural kinds (on pp. 183–5 above), I nowwant to consider some objections to it. InChapter 4 I will consider the conditions ofthe plausibility of these objections.

199:2 The chief Humean counter-arguments may be put in the form of threetheses: –

199:3 (i) there can be no, or at least no

knowledge of, necessary connections betweenmatters of fact;

199:4 (ii) if there were necessary correc-tions between matters of fact they wouldhave to be known a priori; so science couldnot be empirical;

199:5 (iii) men are never directly awareof any causal power or agency or necessaryconnections between matters of fact, so theseconcepts cannot be justified by experience(though they may be explained by it; or are,for the neo-Kantian, imposed upon it).

199:6 The argument for thesis (i) is typ-ically constructed as follows: there is noth-ing inconsistent about the supposition thatthe cause of a phenomenon, say putting akettle of water on the stove and heating it,should not be accompanied by the effect inquestion. It is conceivable that water mightfreeze instead of boil when it is heated. Now


thesis (i) is, as stated, highly ambiguous. Itis not clear whether it is an ontological or anepistemological thesis (this ambiguity is ofcourse explicit in the way I have formulatedit); whether the ‘necessity’ is logical or non-logical; and whether the ‘matters of fact’ areevents and states of affairs or the statementsdescribing them. Before returning to the ar-gument, we must see exactly what is at stakein it.

200:1 Now, it will be remembered, thatfor the transcendental realist to say that asequence Ea.Eb is necessary is to say thatthere is a generative mechanism at work suchthat when Ea occurs Eb tends to be pro-duced (is produced in the absence of inter-fering causes). If there is such a mechanismthe sequence is necessary; and its necessityis quite independent of any knowledge of it.To analyse the necessity of the connection in

terms of our knowledge of the necessity of theconnection would be to commit the epistemicfallacy (see 1.4 above). There is a real differ-ence, quite independent of men, between thefact that when I heat the kettle of water itboils and the fact that it boils when the timeis half-past two or the colour of my socks isblue. The necessary connections that bindsome but not other events together (whichare the enduring mechanisms of nature) arequite independent of our knowledge of them.

200:2 Statements clearly belong to theepistemic not the ontological order; and log-ical connections hold only between state-ments, not between events and states of af-fairs. Hence the prima facie absurdity ofthose who, in attempting to refute Hume,try to establish that nomic necessity is, ormay be, a species of logical necessity.73 Nat-ural necessity is not logical necessity. Natu-


ral connections hold between things, events,states of affairs and the like; logical connec-tions between propositions. Moreover therecould be a world without propositions, inwhich the concept ‘logical connection’ hadno application. The laws of logic are notfeatures of the world, nor are they imposedupon it. Rather, we must say: the world issuch that changes in it can be consistentlydescribed.

200:3/o Neither natural necessity norknowledge of natural necessity can be iden-tified with logical necessity. But our capac-ity to deduce the Wiedmann-Franz law fromDrude’s theory of electrical conductivity mayserve as a criterion of our knowledge of thenecessity the theory describes. I suggestedin paragraph 3 above that three levels ofknowledge of the objective world order canbe distinguished in the development of sci-

ence; so that statements can be classified asdefinitions, deductive consequences of truetheories and simple protolaws according tothe position they occupy (at any moment oftime) in the development of our knowledge.Hence the deducibility of a tendency from anature may serve as a criterion at the Lock-ean level for our knowledge of natural neces-sity, just as a correct definition may serveas a criterion at the Leibnizian level for ourknowledge of natural kinds. But whetheror not a sequence of events is necessary isquite independent of the logical status of theproposition used to express it; which is afunction of the way it is described in the con-text of our knowledge; which in turn may beshown to have a certain rationale in the de-velopment of science.

201:1 Some causal statements expressingnecessary connections are logically necessary


and some are logically contingent.74 For theHumean, however, logical and natural ne-cessity are easily confused. For given theisomorphic relationship between knowledgeand the world assumed in empirical realismand restricting our knowledge of nature tothe protolegal phase of science (see page 172above) he naturally comes to regard relation-ships between events as characterizable inthe same kind of way as the statements ex-pressing their relationships are at that phasetypically, though not invariably, character-ized; namely as contingent. But it is intothis very same trap that defenders of the en-tailment view of natural necessity fall.

201:2 I shall construe thesis (i) as an epis-temological claim to the effect that knowl-edge of necessary connections between eventsis impossible. And I will attempt to refuteit by arguing that unless there were nec-

essary connections between some (but notother) events, science would be impossible;and that in science the most stringent cri-teria for knowledge of natural necessity maybe satisfied.

201:3 Unless there were necessary connec-tions between matters of fact neither con-firmation nor falsification would be possi-ble. For without them no confirmation in-stance adds any probability whatever to anyinductive instance.75 On the other hand forit to be rational to reject what is falsifiedit must be assumed that a hypothesis whichhas been false in the past will not suddenlybecome true in the future.76 Whether theconclusions of inductive arguments are weak-ened to probability judgements or it is deniedthat science is inductive in nature there mustbe necessary connections between matters offact. Such necessary connections are pro-


vided by enduring mechanisms. Moreover, ifexperimental science is to be possible, theremust be necessary connections between somebut not other events. This implies a dynamicprinciple of indifference: to the effect thatmechanisms not only endure but are trans-factually active. Neither their enduring northeir transfactual activity is in need of ex-planation.

202:1 Unless there were necessary connec-tions between matters of fact we could haveno knowledge, even particular knowledge (inas much as this depends upon inferences be-yond what is immediately observed), of theworld. For science to be possible then theworld must consist of enduring and transfac-tually active mechanisms; and there must benecessary connections between some but notother matters of fact.

202:2 Natural mechanisms are of course

nothing other than the powers or ways ofacting of things. Thus, if science is to be pos-sible, there must be a relationship of naturalnecessity between what a thing is and whata thing can do; and hence between what athing is and what it tends to do, in appro-priate conditions. The deducibility of a ten-dency from a nature thus constitutes a crite-rion for our knowledge of natural necessity.Events are necessarily connected when nat-ural tendencies are realized.

202:3/o With this in mind, let us returnto a detailed examination of the argumentfor thesis (i). Is it conceivable that watershould not boil when it is heated? Now itmight be said straightaway that it is incon-ceivable to suppose that water might not boilwhen it is heated. Since anything that didnot boil when it was heated could not prop-erly be said to be ‘water’ at all. That is


that, in Lockean terminology, ‘boiling whenheated’ specifies part of the nominal essenceof water; or we could say with Putnam that‘water’ functions as a ‘law-cluster concept’.77

Now the strength of this reply should not beunder-rated. I have no doubt that we shouldordinarily say something on these lines. In-deed, unless we have some criteria for thecorrect application of the term ‘water’ thereis no reason why we should use it to referto substances which as a matter of fact boilwhen heated rather than to say desk lampsor Saturday afternoons (which do not boilwhen heated). And such criteria would beat least in part dispositional; appearances,notoriously, can be misleading. Litmus pa-per that does not turn red when dipped intoacid, a metal that does not conduct electric-ity, or petrol that does not explode whenignited could not be said to be ‘litmus pa-

per’, ‘a metal’ or ‘petrol’ respectively; sincethe point of referring to the particulars con-cerned in those ways would be gone.78 Amagnet that could not magnetize, a fire thatcannot burn or a pen that can never writewould not be ‘magnets’, ‘fires’ and ‘pens’ atall. Things must satisfy certain criteria forthem to be (correctly identified as) the kindsof things they are. By far the most impor-tant of such criteria are those that dependupon their powers to affect other bodies (aclass which may be extended, analytically, toinclude their powers to affect observers un-der specified conditions in certain standardways).

203:1 Such a reply will not however satisfythe Humean (particularly if he believes thatdefinitions are merely matters of conventionand cannot express empirically ascertainedtruths about kinds of things). More to the


point it will not satisfy the scientist: for, ac-cepting that ‘boiling when heated’ specifiespart of the nominal essence of water, i.e. thecriteria for the identification of a substanceas ‘water’, he will want to know what it isabout water in virtue of which it boils whenit is heated. That is, he will set out to con-struct an explanation, in terms of the molec-ular and atomic structure of water, fromwhich he can deduce its tendency to boilwhen it is heated. Now it is clearly inconsis-tent with this explanation to suppose thatwater might freeze, blush shyly or do any-thing else rather than boil when it is heated.That is, if the explanation is correct watermust boil when it is heated.

203:2/o Suppose however we came acrossa stuff which in all other respects looked andbehaved like water but which did not boilwhen it was heated. Assuming standard con-

ditions and a closed system (so as to elimi-nate the possibility of intervening causes) itwould seem that we have the following alter-natives:

204:1 (a) our explanation was false;

204:2 (b) the fact that it was intended toexplain, viz. that water boils when heated,was false

204:3 (c) the particular concerned hadbeen wrongly identified: it was not a sam-ple of water after all;

204:4 (d) the particular concerned hadchanged; so that it had ceased to be waterby the time it was heated.

204:5 Now the Humean asks us to imag-ine, and inductive scepticism requires thatit be possible, that the cause event occursand the effect event fails to materialize. Letme call this the critical situation. Now Iwant to argue that, given only that possi-


bility (a) is ruled out, so that we have a cor-rect explanation, the critical situation is im-possible; that is, it is not possible that thecause event occurs and the effect event failsto materialize—in our example, that wateris heated and does not boil.

204:6 Let me show this. If the explana-tion is correct water must boil rather thanfreeze when it is heated (though of coursethe converse is not the case); so possibility(b) is ruled out. Consider (c), the misiden-tification of the particular concerned. Nowin this case it is not true to say that wa-ter did not boil when it was heated. Forwhat did not boil was not water but onlysomething which looked, and perhaps other-wise behaved, like it, say ‘nwater’. Finallyconsider (d), a change in the particular con-cerned: what was water when it was put intothe kettle at time t1, ceased to be water by

the time it froze at t2. Here again it is nottrue to say that water did not boil when itwas heated. For by the time it froze it hadbecome something else, say ‘retaw’. Hencegiven only the possibility of a realist interpre-tation of the entities postulated in the expla-nation, the conditions for inductive scepti-cism cannot be satisfied. If there is a real rea-son, located in the nature of the stuff, inde-pendent of the disposition concerned, watermust tend to boil when it is heated (thoughin an open world any particular predictionmay be defeated). The stratification of na-ture thus provides each science with its owninternal inductive warrant

204:7/o Now it might be objected that Ihave omitted from my list of alternatives thepossibility of the explanation, though correctup to time t1, subsequently breaking down.But this possibility equally does not satisfy


the requirements of the critical case. For,now at Stratum II (defining the Leibnizianlevel of the particular movement of sciencewith which we are here concerned), noth-ing which did not possess the molecular andatomic structure that water has been discov-ered to possess could be said to be ‘water’.So, here again, it would not be water thatwas freezing. A stuff remains water only solong as its nature (or real essence) remainsunchanged. (Of course scientists could makea taxonomic change, but this does bear uponthe argument.

205:1 It is of course possible that the na-ture of some particular will be transformed:in which event, scientists will search both foran underlying substance or quasi-substancewhich preserves material continuity throughchange (e.g. a gene pool through specieschange, an atom in chemical reactions, en-

ergy in microphysics) and for the agent ormechanism which brought about the change.The principles of substance and causality areinterdependent and complementary. Thingspersist (and continue to act in their normalway) unless acted upon; and their changesare explained in terms of the action of per-sisting (and transfactually active) things.If science is to be possible changes mustbe transformations, not replacements; andtransformations must be effected by the ac-tions of causes (causal agents). Things can-not pass clean out of existence or events hap-pen for no reason at all. These are ideals ofreason. But if science is to be possible ourworld must be such that they hold. This en-tails that it must be a world of enduring andcontinually acting things. It is o course truethat it is impossible to prove that cases of exnihilo production and miracles cannot ever


happen. All we can say is that they cannotbe known to happen. For it always remainspossible for the scientist that what appearsto be a case of an ex nihilo production ora miracle at time t1 can come eventually att2 to be explained in terms of the transfor-mation of real things and the action of realcauses upon them.

205:2/o I have argued that provided wehave a correct explanation the critical sit-uation cannot occur; that, for example, aslong as the particular stuff remains water itmust tend to boil when heated. But it mightbe urged if, as I have acknowledged, thenature of some particular may be changeddoes this not open the floodgates of induc-tive scepticism once more? The answer is no:for there is a big difference between wonder-ing whether some particular will be so actedupon by real causes in its environment that

its nature (in this case, molecular structure)will be transformed, so that it ceases to bean individual of that kind, and wonderingwhether, while remaining an individual ofthat kind, it will cease to behave in the waythat it has tended to behave in tr the past.The point is even clearer if we generalise it,so raising the questions of the boundariesof kinds and of the scope of application oflaws. The difference is between wonderingwhether water will cease to exist; and won-dering whether, while continuing to exist, itwill stop boiling (in exactly the same circum-stances) when it is heated.

206:1 It might be objected that while whatI have said clearly covers case (d), viz. thatof a particular changing, I have not takenthe possibility of case (c), viz. that of a par-ticular being misidentified, of nwater beingmistaken for water, seriously enough. What


is to prevent us continuously misidentifyingparticulars in just this way? Now just as par-ticulars may be transformed, so they maybe misidentified. But the situation the in-ductive sceptic asks us to imagine only getsof the ground if we assume that the rele-vant particulars have been correctly identi-fied. The problem of induction is the prob-lem of what guarantee we have that the un-observed will resemble the observed, or thefuture the past; it is not the problem of whatguarantee we have that we have correctly ob-served the observed or correctly describedthe past. The suggestion that what I havehere may in fact be a piece of lead piping isirrelevant to the question of what warrant Ihave for assuming that water will continueto boil when heated or for supposing thatthere is a necessary connection between wa-

ter boiling and its being heated.

206:2/o Nevertheless despite this irrel-evance to our present concern scepticismabout particular knowledge can and shouldbe met. It might be met in the followingway: Any argument in which the case forthe general misidentification of particularsis stated itself presupposes the capacity toidentify certain particulars, namely words astokens of a type and hence possessing a cer-tain standard meaning in a given context.Hence no argument for the general misiden-tification of particulars can be consistentlystated. If this argument does not carry con-viction try to imagine a world in which we(a) systematically (b) at random misidenti-fied (α) some particulars (β) all particulars(a’) all the time (i) some of the time. Aworld in which we systematically misiden-tified some given class of particulars (such


as books as saucers and vice versa) wouldjust be a world in which objects had differentnames. But a world in which our misidenti-fications were haphazard or universal is notcoherently conceivable. It makes no sense tosay that a particular has been misidentifiedunless one is prepared to say in what respectit has been misidentified. This itself presup-poses the capacity to identify the particularas of a certain type. Of course our capac-ity to identify particulars presupposes theextended or dynamic principle of substanceenunciated above, namely that things per-sist and continue to act unless acted upon,and hence in this way it presupposes theexistence of necessary connections betweenmatters of fact. It is up to the criteriologyof empirical science to determine whether aparticular has been misidentified or a per-ceptual report is nonveridical. The point is,

however, that if science it to be an ongoingconcern it cannot persistently demand andpersistently return negative verdicts.

207:1 It might be objected to my refuta-tion of thesis (i) that I have not consideredthe possibility that the explanation, whichgives each science at any moment of time itsown inductive warrant, is incorrect. Now itis of course always possible that we are mis-taken in our explanation of why water mustboil when heated; that our description of themechanism in virtue of which it does so iswrong. But this is a general condition ofall knowledge; it does not bear on the argu-ment of thesis (i), which concerns the specialdifficulty of knowledge of necessary connec-tions between matters of fact. I have alreadyargued against the idea that all knowledgeis conjectural on the grounds that refuta-tions presuppose acceptances (progress re-


quires a material cause). But whether ornot my account of the transitive dimension ofthe philosophy of science is accepted, refuta-tions presuppose necessary connections be-tween matters of fact.

207:2/o I have argued that scepticismabout change, about our capacity to identifyparticulars and about the possibility of non-conjectural knowledge as such are all distinctfrom the special kind of scepticism involvedin thesis (i), which is scepticism about thepossibility of knowledge of necessary connec-tions between matters of fact. I have shownhow the second and third forms of scepti-cism, though irrelevant to thesis (i), may beaverted. But how can Heraclitean scepticismbe countered? Changes in things, I have ar-gued, are explained in terms of unchangingthings. The world is stratified. We need onlyworry about whether atoms will cease to ex-

ist when tables and chairs do; we need onlyworry about whether electrons will cease toexist when atoms do. It is contingent thatthe world is such that science is possible. Butgiven that it is the dynamic principles of sub-stance and causality that I have formulatedmust be true of it.

208:1 Three further forms of Heracliteanscepticism are possible in which we could beinvited to imagine that our world is replaced(a) by a totally different one; (b) by one inwhich the principles of substance and causal-ity no longer held; and (c) by one in whichscience ceased to be possible. I shall arguethat the replacements envisaged in (a) and(b) are impossible, but that I am precludedby my own premises from saying anythingabout (c).

208:2/o In (a) it is supposed that ourworld could be replaced by a totally differ-


ent one; but to which, once it had come intobeing, inductive techniques could be reap-plied. Now this is not an intelligible suppo-sition, not only because scientific continuitywould be lost during the replacement (so itwould make little sense to talk of reapplyinginductive techniques), but because there isno possible way in which such a replacementcould be affected save by the action of realcauses.79 In (b) it is supposed that our worldmight be replaced by one to which the princi-ples of substance and causality do not apply.Now although the existence of our world iscontingent, given that it exists the supposi-tion that it might be replaced in this way isnot an intelligible one. Transcendental real-ism demands that we reason from the effect,science, to the condition of its possibility, viz.a world of enduring and transfactually activemechanisms. So we can rest assured that

long after mankind has perished things willpersist and continue to interact in the worldthat we once lived in. This leaves us with(c), about which I have said my premisespreclude me from speaking. But a moment’sreflection shows that (c) is devoid of interestfor us. It is an empty counterfactual. Forwe know as a matter of fact that our worldis one in which science is possible. Hence toassert the possibility of a world without sci-ence is merely to reassert the contingency ofthe circumstance that makes a study of theconditions of the possibility of science possi-ble.

209:1 I have established that we can have(and that science actually possesses) knowl-edge of necessary connections between mat-ters of fact. And I have shown how inductivescepticism proper, namely that arising fromthe assumption of the possibility of what I


have called the critical situation, viz. theoccurrence of the cause event and the non-occurrence of the effect, can be allayed, viz.by the provison of an adequate explanation;and how the other forms of scepticism of-ten confused with inductive scepticism canbe countered. I now turn to theses (ii) and(iii) which the Humean uses to bolster hiscentral contention.

209:2 Thesis (ii) alleges that if there werenecessary connections between matters offact they would have to be known a pri-ori, so that science could not be empirical.It is clear that this argument trades on atacit conflation of logical and natural neces-sity and the identification of the resultantconcept with that of the a priori. To refuteit, I will have to show how knowledge of thenatures or real essences of things, which Ihave argued ground our ascriptions of natu-

ral necessity, can come to be attained empir-ically; that is, how a posteriori knowledge ofnatural necessity is possible.

209:3/o As there is some misunderstand-ing about the role of the concept of essence(and, as we shall see, the nature of defini-tion) in science, some preliminary termino-logical clarification is necessary. The nomi-nal essence of a thing or substance consists ofthose properties the manifestation of whichare necessary for the thing to be correctlyidentified as one of a certain type. Thereal essences of things and substances arethose structures or constitutions in virtueof which the thing or substance tends tobehave the way it does, including manifestthe properties that constitute its nominalessence. Science, I have argued, seeks toexplain the properties of things identifiedat any one level of reality by reference to


their intrinsic structures, or the structuresof which they are an intrinsic part (defin-ing the next level of inquiry). Thus thedispositional properties of say nickel, e.g.that it is magnetic, malleable, resistant torust, melts at 1445 degrees C and boils at2900 degrees C are explained, in the con-text of post-Daltonian atomic theory, by ref-erence to such facts about its intrinsic struc-ture as that its atomic number is 28, itsatomic weight is 58.71 and its density is 8.90.The atomic constitution of nickel is its realessence. But it was discovered a posteriori,in the transitive process of science. And it it-self constituted an explanandum of the nextphase of scientific inquiry.

210:1 In general to classify a group ofthings together in science, to call them bythe same name, presupposes that they pos-sess a real essence or nature in common,

though it does not presuppose that the realessence or nature is known. Thus we arejustified in classifying alsations, terriers andspaniels together as different varieties of thesame species dog because we believe thatthey possess a common genetic constitutionwhich, despite their manifest sensible differ-ences, serves to differentiate them from themembers of the species cat. A chemist willclassify diamonds, graphite and black carbontogether because he believes that they pos-sess a real essence in common, which maybe identified as the atomic (or electronic)structure of carbon, of which these are al-lotropic forms. To classify a thing in a par-ticular way in science is to commit oneselfto a certain line of inquiry. Ex ante therewill be as many possible lines of inquiry asmanifest properties of a thing, but not allwill be equally promising. Thus if one’s con-


cern is to account for the manifest proper-ties of cucumbers it is clearly preferable toclassify a 12 in. long green cucumber un-der the sortal universal ‘cucumber’ ratherthan under the universals ‘green’ or ‘12 in.long’. Not all general terms stand for nat-ural kinds or taxa; because not all generalfeatures of the world have a common expla-nation. Carbon and dogs constitute naturalkinds; but tables and chairs, red things andblue, chunks of graphite and fuzzy dogs donot. The justification of our systems of tax-onomy, of the ways we classify things, of thenominal essences of things in science thus liesin our belief in their fruitfulness in leadingus to explanations in terms of the generativemechanisms contained in their real essences.Not all ways of classifying things are equallypromising because not all sets of propertiesindividuate just one and only one kind of

thing.

211:1 The distinction between real andnominal essences should not be confusedwith that between real and nominal defi-nitions. Real definitions are definitions ofthings, substances and concepts; nominaldefinitions are definitions of words. (Nom-inal essences are the properties that serve toidentify things). Real definitions, in science,are fallible attempts to capture in words thereal essences of things which have alreadybeen identified (and are known under theirnominal essence) at any one stratum of re-ality. As so conceived, they may be true orfalse (not just or even—more or less useful).The atomic weight of copper is 63.5. It wouldbe wrong to claim that it was 53.4 or alter-natively that 63.5 was the atomic weight oftin. Of course this fact was discovered a pos-teriori; but it may now be said to constitute


part of the real definition of copper. If thereal essence of copper consists in its atomic(or electronic) structure, its nominal essencemight consist in its being a red sonorousmetal, malleable and a good conductor ofelectricity etc. Something that did not sat-isfy these properties could not properly besaid to be ‘copper’. But conversely just be-cause the word ‘copper’ in science has a his-tory, and at any moment of time a use, thenominal essence of copper cannot suddenlybe designated by the use of ‘reppoc’ or ‘tin’.Nominal definitions in science cannot there-fore be conceived as stipulative, arbitrary ormatters of convention. Although there is asense in which any other symbol could havebeen used to refer to copper; given this usageand that history ‘copper’ cannot be replacedby ‘bronze’ or ‘female’ for no reason at all.80

Changes in the definitions of words in ongo-

ing social activities require justification.

211:2/o On the view advanced here sci-ence consists in a continuing dialectic be-tween taxonomic and explanatory knowl-edge; knowledge of how the things there arebehave. It aims at real definitions of thethings and structures of the world as well asstatements of their normic behaviour. Thesource of the failure to see this is the ontol-ogy of empirical realism which reduces thingsto qualities, taxa to classes, enduring and ac-tive mechanisms to constant conjunctions ofindependent and atomistic qualia.81 Now ifthe world consists only of qualia and qualiaare independent of one another then the par-ticular names that we give to qualia cannotmatter and all qualia will appear on a par.On this conception, predicates must be inde-pendent of one another and classification is


ultimately arbitrary.

212:1 Now just as it is a mistake to as-sume that science is concerned with any andall behaviour it is a mistake to assume thatit is concerned with any and all things. Sci-entists do not seek to describe the behaviourof or to classify common objects like tablesand chairs, though the laws of physics andthe principles of scientific taxonomy (e.g. theidentification of a table as an oak one) maybe brought to bear on them. Now from thefact that tables have no real essence it doesnot follow that carbon has none. Electronsare not related in the same way as games. Aresemblance theory of universals works bestfor the complex Strawsonian individuals ofordinary life. But the universals of inter-est to science are real: they are the gener-ative mechanisms of nature which account,in their complex determination, for the phe-

nomena of the world, including (upon anal-ysis) the genesis and behaviour of ordinarythings. The dialectic of explanatory and tax-onomic knowledge must thus be formulatedas follows: science is concerned with the be-haviour of things only in as much as it castslight upon their reasons for acting and henceupon what kinds of things there are; and sci-ence is only concerned with things of a par-ticular kind, in as much as they constitutethe reason for some pattern of normic be-haviour and thus themselves become an ap-propriate object of inquiry.

212:2/o The importance of taxa in sciencemay be expressed by saying that what is non-accidentally true of a thing is true of a thingin virtue of its essential nature. A thingacts, or at least tends to act the way it is.It should be stressed that the difference be-tween a thing which has the power or tends


to behave in a certain way and one whichdoes not is not a difference between whatthey will do, since it is contingent upon theflux of conditions whether the power is evermanifested or tendency exercised. Rather, itis a difference in what they themselves are;i.e. in their intrinsic natures A copper vaseremains malleable even if it is never pressedout of shape. It is contingent whether anelectric current is ever passed through a cop-per wire. But it is necessary, given its elec-tronic structure, that it be a good conductorof electricity. We know how things will be-have, if certain conditions materialize, if weknow what the things are. But we can onlyknow what things are a posteriori, via theempirical process of science.

213:1 This view may be contrasted withthe idea that scientists are not concernedwith questions such as ‘what is energy?’ or

‘what is an atom?’ but only with ques-tions of the kind ‘how can the energy of thesun be made useful?’ or ‘under what con-ditions does an atom radiate light?’82 Pop-per’s ‘methodological nominalism’ seems tobe based on the idea that to suppose thatthings have essences is to suppose that itis possible to give explanations which are‘ultimate’ in the sense that they are insus-ceptible in principle of further explanation(which is what he calls ‘essentialism’).83 Al-though Locke may have held this view, is itcertainly no more a necessary feature of theconcept of real essence than it is a necessaryfeature of the concept of behaviour to sup-pose that because a thing can be described asbehaving in a certain way the behaviour it-self cannot be subject to further explanation.It is clear that to suppose that things havereal essences is not to suppose that the real


essences of those things cannot be explainedin terms of more fundamental structures andthings.

213:2/o Two other arguments sometimesinvoked against the concept of real essencesshould be mentioned. The first dependsupon the assumption that differences in na-ture are continuous, not discrete; that ‘Godmakes the spectrum, man makes the pigeon-holes’;84 so that ‘genera, species, essences,classes and so on are human creations’.85 Ican find no possible warrant for such an as-sumption. Taken literally, it would implythat a chromosome count is irrelevant in de-termining the biological sex of an individual,that the class of the living is only conven-tionally divided from the class of the dead,that the chemical elements reveal a con-tinuous gradation in their properties, thattulips merge into rhododendron bushes and

solid objects fade gaseously away into emptyspace. The second involves the belief thatto suppose that there are natural kinds isto suppose that these kinds are fixed, andis in particular to rule out the possibility ofa mechanism of evolution.86 Again, this iscompletely unwarranted. For natures maychange; and whether, and if so the ways inwhich they do, are matters for substantivescientific investigation. No spectrum existsbetween men and apes but that does not pre-clude the possibility of a mechanism of evo-lution (involving a whole sequence of ‘miss-ing links’). What happens in such cases isthat biologists posit a novel entity, a genepool, as the underlying continuant throughthe species’ change. The objection is onlyvalid at the level of ultimate physical en-tities since necessarily if such entities exist


they must be enduring.

214:1/o Scientists attempt to discover thereal essences of things a posteriori, and toexpress their discoveries in real definitionsof the natural kinds. From a description ofthe nature of a thing its behavioural tenden-cies can be deduced. When such tendenciesare realized the events describing the stim-ulus or releasing conditions for the exerciseof the tendency and its realization may besaid to be necessarily connected. Thus sci-entists can come to possess knowledge of nec-essary connections between events as a resultof an a posteriori process of discovery. Scien-tists are not content to collect conjunctionsof events. Rather they try to discover thenatures of things. Given this, no problem ofinduction can arise. Since it is not possiblefor a thing to act inconsistently with its ownnature and remain the kind of thing it is.

That is, a thing must tend to act the way itdoes if it is to be the kind of thing it is. Ifa thing is a stick of gelignite it must explodeif certain conditions materialize. Since any-thing that did not explode in those circum-stances would not be a stick of gelignite butsome other substance. Now given the satis-faction of the criteria for the identificationof a substance, say water, and the recording,preferably under experimentally closed con-ditions, of its most significant and suggestivebehavioural properties, scientists move im-mediately to the construction and testing ofpossible explanations for the protolaws iden-tified. But if there is an explanation, locatedm the nature of the stuff or the system ofwhich the stuff is a part, whether or not itis known by men, water must tend to boilwhen it is heated. It is the real stratificationof nature that justifies induction in science.


It is not we that impose uniformities uponthe world, but nature that makes induction(properly circumscribed) a rational activityfor men.

215:1 The third Humean counter-argument is that we are never directlyaware of any necessary connection betweenmatters of fact or causal power or agencyso that these concepts cannot be justifiedby experience. Thesis (iii) thus completesa triangle, whose other sides are theses(ii) and (i). It could be argued that weare sometimes directly aware of necessarilyconnected sequences (see 2.3 above), andthat we are sometimes directly aware ofthe exercise of causal powers (though thepowers themselves can only be known, notshown, to exist; i.e. we are never directlyaware of causal powers as such).87 It seemsclear that we are aware of ourselves as

causal agents in a world of other causalagents; and that unless we were so aware wecould not act intentionally, or come to knowourselves as causal agents at all. (Projec-tive explanations of our idea of necessaryconnection are clearly anthropocentric.)However for the transcendental realist thisis incidental. For, for him, the status ofthe concept of necessary connection is clear:it has been established, by philosophicalargument, as applicable to some but notother sequences of events as a necessarycondition of the social activity of science. (Itshould be stressed that this does not meanthat any particular science has correctlyidentified, let alone adequately described,the necessary sequences: it is a condition ofthe possibility of science.) Thus the conceptof natural necessity does not have to bejustified in terms of or traced back to its


source in sense-experience; though theremust be a scientific explanation of how wecome to possess the concept.

215:2 That science has a posteriori knowl-edge of necessary connections between mat-ters of fact is a proposition that can be givenno further justification.

3.6 The Problem of In-duction

215:3/o In the concluding section of thischapter I intend to argue that traditional ap-proaches to the problem of induction fail; toreveal a crucial ambiguity in the formulationof the problem; and to show how transcen-dental realism can resolve it. In doing so Iwill be bringing together my critiques of theideas of the actuality and contingency of law;

and I will relate my resolution of the problemof induction to the problem of universals. Ishall argue that the condition of the intelli-gibility of the traditional problem of induc-tion is an ontology of atomistic events andclosed systems; but that in our world induc-tive reasoning may be shown to have a ratio-nal place.

216:1 The traditional problem of induc-tion is the problem of what warrant wehave for reasoning from particular instancesto general statements (induction proper) orfrom observed to unobserved or past to fu-ture instances (eduction). Now it is clearthat unless we (sometimes) have some suchwarrant nothing can be justified, shown tobe mistaken or called into doubt: memorycannot be relied upon, a mistake demon-strated or grounds for a sceptical conclusiongiven. (Why, for instance should the fact

3.6. THE PROBLEM OF INDUCTION 291

that my senses have deceived me in the pastbe a ground for believing that they will doso in the future?) Indeed complete scep-ticism about induction seems literally un-thinkable.88 So pervasive a feature of our so-cial life is inductive-type reasoning that itseems patently unsatisfactory to be told thatit is just a contingent fact about the worldthat induction is successful.89 If inductive-type reasoning is necessary, then it seemsincumbent upon us to ask what the worldmust be like for it to be possible; and whatmust have been assumed (inter alia aboutthe world) for the problem to have remainedintractable. The answer to the first ques-tion will constitute a set of synthetic a pri-ori truths about the world; the answer tothe second a set of synthetic a priori truthsabout received philosophy of science

216:2/o The standard responses to the tra-

ditional problem of induction are of course:(i) to deny that science is inductive in na-ture (e.g Popper); (ii) to justify inductioninductively (e.g. Black, Braithwaite); (iii) tostrengthen the premises of inductive argu-ments, so that they became in effect en-thymematic deductive arguments (e.g. Mill);(iv) to weaken the conclusions of induc-tive arguments to probability judgements(e.g. Carnap); (v) to justify induction prag-matically or vindicate it (e.g. Reichenbach,Salmon); (vi) to dissolve the problem, i.e. toclaim that it is a pseudo-one (e.g. Strawson,Edwards). The objections to (ii)-(v) are wellknown; they all in one way or other beg thepoint at issue which is:

217:1 (A) the problem of what warrant wehave for supposing that the course of naturewill not change.

217:2 (A) must of course be distinguished


from:

217:3 (B) the problem of what warrant wehave for believing some proposition, state-ment or theory true.

217:4 For to say that the special theoryof relativity refuted Newtonian mechanics isnot to say that the course of nature changed:it is to say that Newtonian mechanics was (atleast in one respect) wrong all along. I willleave aside considerations pertaining to ques-tion (B) and the rationale for distinguishingit from question (A) until Chapter 4.

217:5/o Popper claims to have solved theproblem of induction, by accepting Hume’sconclusion that induction cannot be justi-fied but denying that science is inductive innature. According to Popper science pro-ceeds by the refutation of bold conjectures(general statements) by their deductive con-sequences (singular relatively observational

statements). However, for a relatively obser-vational statement to refute a general law-like statement (or theory) it must be pre-supposed that the course of nature will notchange so that the experimental and obser-vational context in which the refuting obser-vation statement is true ceases to be true.Popper would reply to this objection as fol-lows: ‘. . . there is a logical asymmetry:one singular statement—say about the peri-helion of Mercury—can formally falsify Ke-pler’s laws; but these cannot be formallyverified by any number of singular state-ments. The attempt to minimize this asym-metry can only lead to confusion’.90 Butthe decision to accept the singular statementabout the perihelion of Mercury as falsify-ing Kepler’s laws presupposes that in ex-actly the same circumstances Mercury’s per-ihelion would behave in exactly the same


way. The asymmetry is there alright. Butwhat warrant is there in Popper’s systemfor supposing that nature is uniform so thatits course will not change, in the way Humeand Goodman invite us to imagine, so thatour best-falsified theories (astrology, Marx-ism, psycho-analysis, Newtonian mechanics)become true? Whatever the merits of Pop-per’s philosophy of science, his claim to havesolved the logical problem of induction ismanifestly untenable and based on a confu-sion of problems (A) and (B)

218:1 (vi) also fails, for a number of rea-sons. First, it seems possible to imagineworlds in which induction would be unsuc-cessful not just ‘counter-inductive’ worlds,in which the unexpected always happens,but capricious worlds, for which no kind ofrule could be formulated.91 Secondly, evena straightforwardly counter-inductive world

would be incapable of sustaining scientific orsocial life. For the unexpected is a poten-tially infinite class. No inductive rule couldbe operationalized in a counter-inductiveworld. ‘Expect the unexpected’ can only beapplied ex post, after the unexpected has ac-tually happened. The fact that induction is(sometimes) successful places a constraint onthe world in which we live. For in some con-ceivable worlds induction would be unsuc-cessful or unoperationisable. Hence it wouldseem reasonable, and indeed necessary, toisolate the conditions that must obtain forinduction to be successful. If there are neces-sary conditions for the success of induction,they must constitute the missing ‘justifica-tion’. Thirdly, it is clear that not all induc-tive arguments are equally good. To appealto induction as an institution does not helpus to decide between good and bad induc-


tive arguments (any more than appeal to thelaw helps us to decide between good and badlaws). Finally, induction is not always justi-fied. If it is not always justified there must beconditions under which it is justified, aboutwhich the approach represented by (vi) hasnothing to say.

218:2 Induction, I have said, is not al-ways justified. In general induction is onlyjustified if we have some reason other thanpositive instances for the generalization con-cerned. Two kinds of reasons are distinguish-able: –

218:3/o (Alpha) a plausible model or hy-pothesis of a mechanism by means of whichwe can render it intelligible to ourselves thatwhen Ea then Eb.

92

219:1 (β) knowledge of the mechanismwhich given Ea generates Eb.

219:2 Induction is only justified if the gen-

eralization concerned is a law of nature.93 Ofcourse ex ante when inductive reasoning oc-curs we do not know whether the sequence isnecessary. But it is justified if it is necessary.And to justify it we need only have groundsfor supposing that it is necessary. Now I havealready argued that (α), though represent-ing an important moment in the process ofscientific discovery, carries too little ontolog-ical bite to justify the assumption of a lawof nature (see paragraph 2 above). The gen-erative mechanisms of nature are of coursenothing other than the powers or ways ofacting of things.

219:3 Now eduction, or inference fromparticular instances to other particular in-stances, is only justified if in addition thesystem in which the consequent event oc-curs is closed. The crucial ambiguity in theformulation of the problem of induction to


which I referred earlier now becomes clear.It turns on the question of whether the gen-eralization referred to (or in eduction as-sumed) is an empirical or a normic state-ment, a statement about the conjunctions orevents or the tendencies of things, a state-ment about actualities or possibilities. Nowa belief in the uniformity of nature is quitemisplaced if it is a belief in the invarianceof patterns of events (or experiences). Forthe non-invariance of their patterns is, I haveshown in Chapter 2, a condition of the possi-bility of science. A belief in the uniformity ofnature is only rational if it is a belief in theinvariance of structures. The eventsequentialpast is an unreliable guide to the future. In-stead what we require, and in small measureactually possess, is a knowledge of the invari-ant tendencies and natures of things (though

this does not legitimate predictions).

219:4/o Induction is only justified if thegeneralization is a law of nature and educ-tion is only justified if the system is closed(so that the tendency designated in the lawstatement must, given the occurrence of theantecedent, be realized). Induction is jus-tified because nature is stratified. Now wedo not need to know what the structuresare to know that nature is stratified (We donot need to know what the explanation is toknow that there is an explanation.) We knowthat nature is stratified because its strati-fication is a condition of the possibility ofscience-in-general. And we know that sci-ence is possible in general because it in factoccurs. To know that induction is justifiedwe do not need to know what any particularexplanation is.

220:1 Now if we know what the correct


explanation is we do not need to reason in-ductively. And when we need to reason in-ductively we do not know what the correctexplanation is. Given then that our knowl-edge that nature is stratified is secured a pri-ori how can we justify any particular piece ofinductive reasoning? By giving grounds forsupposing that there is an explanation, lo-cated in the nature of things (at the nexthighest rung of reality), for the generaliza-tion claimed. It is the possibility of an ad-equate explanation, in terms of real invari-ances from which the behaviour concernedcan, normically understood, be deduced thatmust justify any piece of inductive reasoningin science. Thus it is the possibility of thesatisfaction of a deductive criterion that jus-tifies induction in science. But we know thatthe world must be such that this is so It isthe structure of the world that makes induc-

tion, when it is, a rational activity for men.

220:2 I have of course already argued thatthe conditions for inductive scepticism onlyobtain if we deny the possibility of (β) above.If we allow (Beta), i.e. a realist interpretationof the entities postulated in scientific the-ory, then we do have a reason independentof the facts identified at any one level of re-ality as to why one but not another sequenceof events must be forthcoming (if the systemis closed). Now as the argument for induc-tive scepticism turns on the alleged impossi-bility of knowledge of necessary connectionsbetween matters of fact and I have demon-strated how we can (and do) come to havesuch knowledge a posteriori in science, it re-mains only for me to examine the conditionsof the plausibility of the traditional problemof induction and to show how the galaxy ofproblems in its wake can be rationally re-


solved.

220:3/o The problem of induction arises ifwe restrict the grounds for a generalizationto its instances, i.e. if we accept Nicod’s cri-terion of the evidence for a law. It is resolvedin two steps: (α)* by allowing a model of agenerative mechanism or structure to sup-ply the missing reason that the coherence ofscientific practice demands, and in particu-lar to provide a crucial part of the groundsfor a law; and (β)* by allowing that undercertain conditions, i.e. if certain criteria aresatisfied, such models held out in the scien-tific imagination as plausible representationsof the real mechanisms of nature may cometo be established as real. Mechanisms areenduring; they are nothing but the powersof things. Things, unlike events (which arechanges in them), persist. Their persistencedoes not need explaining. Space and time

are causally inert: they possess neither lia-bilities nor powers. Now step (β)* involvesexperimentation. It is condition of the intel-ligibility of experimental activity that causalstructures not only persist but act indepen-dently of the patterns of events. Thus theworld is open, and the laws of nature mustbe analysed as the tendencies of things. Thedynamic realist principles of substance andcausality to which I have been working maythus be stated as follows: the world con-sists of enduring and transfactually activethings (substance) which endure and act intheir normal way unless acted upon (causal-ity). Effects presuppose both continuantsand causes. They must occur in things andbe brought about by things (other than po-sition in space or moment in time). On theother hand only effects need explaining.

221:1/o The condition of the intelligibil-


ity of the problem of induction is an ontol-ogy of atomistic events and closed systems.For without closed systems there is no reasonfor the past to resemble the future and with-out atomistic events there is reason why itshould. The grounds for the atomistic ontol-ogy of empirical realism disappear when werealise that sense-experience is neither theonly ground nor the only source of knowl-edge; and that it is analysable neither inpurely atomistic terms94 nor as a happeningto passive men.95 The grounds for the ac-tualist ontology of closed systems disappearwhen we realize that in general, outside as-tronomical contexts, they need to be exper-imentally established. In place of the ontol-ogy of experience and atomistic events con-stantly conjoined, transcendental realism es-tablishes an ontology of complex and activestructures and things. In place of the con-

trast (and unbridgeable gulf) between ourparticular and our general knowledge of theworld, transcendental realism allows knowl-edge both of things and of their powers orways of acting. In place of the analysis oflaws as constant conjunctions of events, tran-scendental realism analyses laws in terms ofthe tendencies of things which may be exer-cised unrealized and realized unperceived bymen. Science becomes a social activity, dif-ficult and discriminating; not an automatic,individualistic affair. Science is explanatorynon-predictive.

222:1/o I now want to show how the re-placement of the empirical realist ontology ofatomistic events and closed systems by thetranscendental realist ontology of persistingand transfactually active things allows us toresolve the problems and paradoxes associ-ated with the problem of induction. This


problem, I have argued, only arises if wedeny the possibility of a reason, located inthe enduring nature of some thing, for thebehaviour concerned. In its sharpest formit may be expressed as follows: if all predi-cates refer ultimately to experience and ex-periences are independent of each other, asthey must be if they are to ground (in part orin whole) our knowledge of the world, thenpredicates must be independent of one an-other. There can then be no reason for ex-pecting one rather than another set of ex-periences so that for all we know predicatesmay become associated in entirely new ways.Thus there is no reason why cabinet minis-ters should not suddenly start bearing figsor Mancunians disintegrate when exposed tothe sun (the problem of induction); no rea-son why emeralds examined after A.D. 2000should not turn out to be blue or blue things

become green next Christmas (Goodman’sparadox);96 no reason why the sighting ofa black raven should confirm the proposi-tion that all ravens are black better thanthe sighting of a red herring or a white shoe(Hempel’s paradox);97 no reason to supposethat if I had gone for a walk in the rain fiveminutes ago I would in fact have got wet (theproblem of subjunctive conditionals). Nowthere are two ways of meeting these absurdi-ties. The first is to hold that the paradoxesand problems stem from the insertion of arti-ficial predicates and fanciful conjectures intoalready functioning and well-connected sci-entific contexts, for which no positive reasoncan be given. The trouble with this line of re-sponse is that it is still vulnerable to the ob-jection that there is no ground, independentof custom or convention or past practice ormob psychology, for expecting one sequence


of events rather than another. And besidesthe nature of the ‘connection’ predicates aresupposed to enjoy is unclear. The secondis the transcendental realist line. This lineholds that there are objective connections inthe nature of things, which may be identi-fied as enduring mechanisms, which bind orlink some but not other events and states ofaffairs. I will now sketch the transcendentalrealist resolution of these problems.

223:1 It is physically impossible for a cab-inet minister to bear figs; that is, nothingwhich bore figs could properly be said to bea cabinet minister at all. Desk lamps cannotfly or walk about the room, just as Man-cunians do not disintegrate when exposedto the sun. A particular must tend to be-have in certain ways if it is to be of thekind that it is. On an ontology of things thegeneral problem of induction cannot arise,

though there may be specific problems ofidentification and special reasons for expect-ing change. Things persist. They are natu-ral endurers and their changes are explainedin terms of unchanging things. What is therationale for this resolution? The scientificexplanation of scientifically significant be-haviour is in terms of invariant principles ofstructure. Thus the scientist assumes thatthere is something about metals (their pos-session of free electrons, perhaps) in virtueof which it is not possible for them not toconduct electricity. Their possession of freeelectrons is the invariant principle of struc-ture. There is something about cabinet min-isters (their genetic constitution, perhaps) invirtue of which it is not possible for them tobear figs; just as, if Socrates is a man hemust die.

223:2/o On an ontology of things Good-


man’s paradox cannot arise. Now either ‘allemeralds are green’ is law-like or it is not.If it is not the Goodmanesque alternative‘all emeralds are grue’ is equally admissible.For it is then ex hypothesi purely accidentalthat all emeralds happen to be green. Onthe other hand, to suppose that ‘all emer-alds are green’ is law-like is to suppose thatthere is a reason, located in its crystallinestructure of chemical composition, why itdifferentially reflects light the way it does.Now given that structure, emeralds must, tonormal observers under standard conditions,look green. So anything which looked bluecould not possess that structure, and hencewould not be an emerald at all. Now ofcourse occasionally we may have grounds forsupposing that a particular and even a kindwill cease to exist, i.e. be transformed into adifferent thing or kind (or even into an en-

tirely different kind of thing or kind). Thus agenuine Goodman-type problem could arise.However it would be a specific problem, it-self presupposing the existence of both a con-tinuant and a cause. Moreover no predicatesuch as ‘grue’ could ever be admissible to sci-ence. Since the mere passage of time cannotconstitute a cause. It would have to be a co-incidence that emeralds examined after A.D.2000 looked blue Dates can be at best onlyproxy causes.

224:1 Hempel’s paradox may be resolvedquite simply once the significance of his intu-ition, viz. that propositions about shoes andherrings are irrelevant to the truth of propo-sitions about ravens, is grasped. If laws arestatements about things and there must besome reason other than instances for accept-ing them, then Hempel’s paradox may be re-solved as follows: If ‘all ravens are black’


is law-like there must be a reason, locatedin the nature of ravens (not in the natureof black), why ravens are black. ‘All ravensare black’ is a truth about ravens, not aboutcolour. Hence the contrapositive ‘all non-black-things are non-ravens’ has no bearingon it. The logical subject of a law of natureis a (natural kind of) thing. Hence there isa logical asymmetry built into its structure,reflecting the site of the mechanism desig-nated, in virtue of which its terms are notequivalent and contraposition is prohibited.To put this another way: the mechanismthat, to use Strawson’s term, ‘collects’98 redunder herring or white under shoe is eitherentirely different from the mechanism thatcollects black under ravens or else where asin the shoe case the ‘connection’ is entirelyaccidental there is no mechanism involved at

all.

224:2/o The problem of subjunctive con-ditionals is easily and rationally resolved onan ontology of things. To assert a law ofnature is to ascribe a possibility to a thing—a possibility which is possessed by the thing,and has a real basis in the enduring nature ofthe thing, whether it is exercised or not. Toassert a subjunctive conditional is just to saythat the possibility possessed by the thingwould have been exercised, had the condi-tions in fact been different. I would have gotwet alright, rain being what it is.

225:1 The source of these problems lies inthe reduction of things to qualities and laws,which are statements about things, to con-junctions of events. This is reflected mostsharply in the failure to sustain the ideaof the necessity of law. But side by sidewith these well-known problems is a less well-


known set (due to the tacit assumption, byalmost all philosophers of science, of closedsystems), which turn on the failure of the ac-tualist ontology of empirical realism to sus-tain the idea of the universality of law. Lack-ing from the former set is a criterion fordistinguishing necessary from accidental se-quences (depending upon a concept of thestratification of the world); lacking from thelatter set is a criterion for distinguishingopen from closed systems (depending upon aconcept of the differentiation of the world).For empirical realism all sequences are acci-dental and the world is closed; for transcen-dental realism some sequences are necessaryand the world is open.

225:2/o Let us briefly note these homo-logues of the well-known problems on theuniversality axis. Corresponding to theproblem of induction we have the problem

of what justifies the assumption that lawswill continue to hold outside the laboratory.This is resolved by allowing (or rather see-ing that it is a condition of the intelligibil-ity of experimental science) that things en-dure and continue to act in their normal wayoutside as well as inside the laboratory (asthey will do in the future as in the past)unless, as may sometimes happen, they arethemselves transformed. Corresponding tothe problem of subjunctive (and counterfac-tual) conditionals we have the problem ofnormic (and transfactual) ones. Correspond-ing to the paradoxes of confirmation, para-doxes of falsification. (Laws and theories arestraightaway falsified by any open-systemicinstance, just as they are straightaway con-firmed by any contrapositive instance, if weregard laws as empirical statements.) Corre-sponding to the problem of justifying the use


of hypothetical entities in theory construc-tion (which Hempel has called ‘the theoreti-cian’s dilemma’)99 we have the problem ofjustifying the use of the CP clause in theoryapplication (which we could call ‘the engi-neer’s dilemma’, or the problem of the ap-plied scientist’s excuse). All these problemscan be rationally resolved by an account ofscience which sees it as an attempt to pen-etrate ever deeper into the nature of thingsand to describe more adequately the thingsof nature.

226:1 The Humean analysis of laws is afailure: it does too little and too much. Thecausal contingent is neither contingent noractual, but necessary and real.

226:2 The intelligibility of perception pre-supposes that objects persist, in space andtime, independently of our perception ofthem. The intelligibility of experimental ac-

tivity presupposes that they act, in spaceand time, independently of the patterns ofevents they generate. Now the use of generalterms in identifying these objects presup-poses that they fall into natural kinds. Butit is not possible to say anything in generalabout the number of kinds there are or aboutthe numbers in any particular kind. Now thethings posited by science in its investigationsmay be quite recondite and abstract with re-spect to our ordinary experience. It is wrongto think of them as necessarily like materialobjects—they may be powers, forces, fieldsor just complex structures or sets of relation-ships. Their metaphysical character, whichjustifies us labelling them as ‘things’ to marktheir insusceptibility to analysis as ‘events’or ‘experiences’, lies in their persistence andtransfactual activity. This entails that theypersist even when they do not act, and act


in their normal way in the flux of condi-tions that co-determine the actual outcomeof their activity. Things, as so conceived,must be complex and structured; in virtue ofwhich possibilities may be ascribed to themwhich may be unexercised or exercised unac-tualized or actualized unperceived by men.On this account of science the actual is seenas an instance of the possible; and a normicmood is added to the hierarchy of condition-als marking the space of possibilities exer-cised but unactualized.

226:3/o On the account of laws advancedhere they cannot be identified with constantconjunctions of atomistic events or regardedas reporting correlations between either in-dependent or equivalent variables. On thecontrary, they must always be grounded insome conception of an explanatory mecha-nism and ascribed, as tendencies, to specific

kinds of things. This is consistent with theview of ordinary things as subject to dual(and multiple) control, perhaps by princi-ples of relatively different kinds. Laws donot describe the patterns of events. Rather,we could say, they describe the normic be-haviour of novel kinds and impose con-straints on familiar things. Ordinary thingsmay be conceived, metaphysically, as com-pounds. This allows us to make sense of theindividuality of historical particulars; just asthe conception of ordinary events as ‘con-junctures’ (see 2.6 above) allows us to makesense of the uniqueness of historical events.

227:1 If the ordinary things of the worldare compounds then it is natural that theyshould share nothing in common except re-semblances. But just as only some eventsare significant in science, although all inprinciple may be explained by it, so with


things. Ordinary things have a genesis andtheir changes may be rationally explained(in terms of continuants and causes) by ref-erence to the exercise of the tendencies ofthings which share a common identity, i.e.which fall into a natural kind. Scientificallysignificant generality does not lie on the faceof the world, but in the hidden essences ofthings.

227:2 How can this be shown? Either clas-sification is arbitrary or it is not. If it is non-arbitrary it must be based on a relationshipof resemblance (similarity) or identity. If itis only based on the assumption of a rela-tionship of resemblance there is no rationalefor the stratification of science. On the otherhand if it is based on an assumed theoreticalidentity then we do have a rationale for themove from manifest behaviour to essentialnature that we have seen lies at the heart of

rational theory-construction in science. Tostress, nothing can be said about the numberor variety of real universals there are. Butit is clear that ‘table’ and ‘red’ are not realuniversals; and ‘gene’ and ‘molecule’ are.

227:3/o A similar trilemma may be ap-plied to our explanatory knowledge of theworld. Either explanation is arbitrary or itis not (arbitrariness is suggested by the prob-lem of induction or any of the paradoxesdiscussed above). If it is non-arbitrary theground for the explanation is either imposedby men or it exists in the world. If it is im-posed by men we are left without any ra-tionale for experimental activity, the processof testing human constructions against theworld. Predicates are not independent ofeach other and classifications are not arbi-trary in science because there are necessaryconnections in the world and things fall into


natural kinds.

228:1 I can now return to the questionI asked at the beginning of this chapter.To what, in our ascription of laws, is ne-cessity and universality properly ascribed?The answer is to the transfactual activityof things, i.e. to enduring mechanisms atwork. For these ascriptions to be possible theworld must be composed of enduring mech-anisms which act independently of men; sci-ence must be an ongoing social activity; andmen must be (in the sense indicated in 2.5above) free.

228:2 Now it is because we are mate-rial things, possessed of the senses of sightand touch, that we accord priority in verify-ing existential claims to changes in materialthings. But scientists posit for these changesboth continuants and causes, some of whichare necessarily unperceivable. It is true that

‘that a flash or a bang occurs does not en-tail that anything flashes or bangs. “Letthere be light” does not mean “let some-thing shine” ‘.100 But a scientist can neverrest content with effects: he must searchfor causes; and causes reside in or consti-tute things. Charged clouds, magnetic fieldsand radio stars can only be detected throughtheir effects. But this does not lead us todeny their existence, any more than we canrationally doubt the existence of society or oflanguage as a structure irreducible to its ef-fects. There could be a world of electronswithout material objects; and there couldbe a world of material objects without men.It is contingent that we exist (and so knowthis). But given that we do, no other posi-tion is rationally defensible. It is the natureof the world that determines which aspectsof reality can be possible objects of knowl-


edge for us.101 But it is the historical devel-opment of the various sciences that deter-mines in what manner and to what degreethese possibilities are taken up by men.


1 This is the ontological form of Hume’s doc-trine that events ‘seem conjoined, but never con-nected’. See D. Hume, An Enquiry ConcerningHuman Understanding, p. 74.

2 F. Bacon, Novum Organum.

3 To borrow Toulmin’s useful concept. See S.Toulmin, Human Understanding, Vol. I, p. 158and passim.

4 W. E. Johnson, Logic, Vol. III, Chap. 1.

5 R. Chisholm, ‘The Contrary to Fact Con-ditional’, Mind 55 (1946), reprinted in Readingsin Philosophical Analysis, eds. H. Feigl and G.

Maxwell, pp. 482–97.6 (a) and (b) correspond of course to Hume’s

two definitions of ‘cause’. See D. Hume, Treatise,p. 172 and Inquiry, pp. 76–7.

7 I owe this term of G. Buchdahl, op. cit., p.27 and passim.

8 See e.g. R. B. Braithwaite, op. cit., Chap. 8;C. G. Hempel, op. cit., Chap. 12; and E. Nagel,op. cit., Chap. 4.

9 The Jesuit mathematicism Clavius demon-strated this fallacy in Osiander’s apologetic pref-ace to Copernicus’ De Revolutionibus. Osianderhad argued, as Galileo was later invited to be-fore the Inquisition, that the helio-centric theorywas merely a mathematically adequate represen-tation of the facts of planetary motion that madeno claim to be true. Clavius pointed out that itwas never a good argument in favour of a theorythat it ‘saved the appearances’, as a true resultcould be derived from any number of absurd orfalse premises. (Cf. J. Losee, An Historical In-troduction to the Philosophy of Science, pp. 44–5.) Indeed even if we exclude all premises whichwe know to be false or which are not explicitly


defined there will still be an infinite number ofsets of premises from which the facts can be de-duced, provided we allow for the introductionof artificial predicates such as Hesse’s ‘tove’ (M.B. Hesse, Models and Analogies in Science, p.30), of which place- and time-dependent predi-cates such as Goodman’s ‘grue’ (N. Goodman,Fact, Fiction and Forecast, p. 74) merely forma special class. Hence deducibility cannot pro-vide a sufficient criterion for choosing one setof premises rather than another (the source ofGoodman’s paradox) or for justifying one state-ment rather than another as law-like.

10 This is of course a very poor best. For(i) the simplest of any small number of explana-tions is not necessarily the best (cf. M. Bunge,The Myth of Simplicity, pp. 51–134); (ii) therewill still be an in principle infinite number ofequally simple explanations, if we restrict our-selves to formal or syntactical criteria alone (cf.J. J. Katz, The Problem of Induction and its

Solution, Chaps. 4 and 5).11 J S. Mill, A System of Logic, Vol. I, p. 378.12 E. Nagel, op. cit., pp. 64–5.13 I. Lakatos, op. cit., p. 116.14 N. Goodman, op. cit., pp. 92–122.15 For, as Craig’s theorem shows, if it does

not the theoretical component is then completelyeliminable. See W. Craig, ‘The Replacement ofAuxiliary Expressions’, Philosophical Review 65(1956), pp. 35–55.

16 N. R. Campbell, The Foundations of Sci-ence, esp. Chap. 6.

17 Ibid. pp. 126–40.18 N. R. Campbell, What is Science?, p. 99.19 N. R. Campbell, Foundations, pp. 132–7;

and M. B. Hesse, op. cit., pp. 35–43.20 N. R. Campbell, op. cit., pp. 243–56.21 J. Tyndall, ‘Scientific Uses of Imagination’,

Fragments of Science for Unscientific People, p.131.

22 P. K. Feyerabend, ‘Problems of Empiri-cism, Part II’, The Nature and Function of Sci-entific Theory, ed. R. G. Colodny, p. 317.

23 See e.g. M. B. Hesse, op. cit., pp. 35ff. Cf.


also P. K. Feyerabend, ‘Problems of Empiricism’,op. cit., pp. 173ff.

24 P. Duhem, op. cit., p. 7.

25 M. Scriven, ‘Truisms’, p. 450.

26 ‘This’, says Mach, ‘is all that natural lawsare’, op. cit., p. 192.

27 Indeed one might be tempted to see the dif-ference as merely one of taste or temperamentas when Duhem compared the ‘rolling drums’,‘pearl beads’ and ‘toothed wheels’ of the me-chanical models of English physicists such asMaxwell, Kelvin and Lodge with his own Carte-sian conception of an axiomatic electricity. Seeop. cit., pp. 70–1.

28 R Chisholm, op. cit., p. 496.

29 See e.g. E. Nagel, op. cit., p. 75; or S. Toul-min, op. cit., p. 185.

30 Ibid., p. 185.

31 R. Harre, op. Cit., esp. Chaps. 2–3.

32 I. Lakatos, op. cit., pp. 134–8.

33 As Feyerabend has put it: a theory maybe in trouble only because of ‘the backwardnessof the observational ideology’. See ‘Problems of

Empiricism, Part II’, pp. 292ff.34 P. S. de Laplace, The System of the World,

Bk. III, Chap. II.35 F. Bacon, op. cit., Bk. II, Aphorism

XXXVI.36 ‘Any statement may be held true come

what may, if we make drastic enough changeselsewhere in the system’, W. V. O. Quine, Froma Logical Point of View, p. 43.

37 See e.g. K. R. Popper, Logic of ScientificDiscovery, p. 42 and pp. 80–2; and T. W. Hutchi-son, The Significance and Basic Postulates ofEconomic Theory, pp. 40–6. I have of courseargued (in 2.4 above) on quite distinct realistgrounds that once the irrationality of pseudo-falsification is granted the CP clause becomessuperfluous.

38 R Harre, op. cit., p. 132.39 This theory is most clearly stated in R.

Harre, op. Cit., Chap. 4. Although Harre is Ithink logically committed to, and may he pre-pared to accept, transcendental realism in theform in which it is developed here, he does notsay how laws ‘explain away counter-instances’


and so achieve universality, ibid. p. 92.40 It should be remembered that transcenden-

tal realism not only warrants subjunctive andcounterfactual statements (where antecedentsare uninstantiated) but normic and transfactualones (where consequents may be unrealised).This is another nail in the coffin of deductivism.For at level (iii) a law may be upheld even whenP is true and Q is false; which is of course theonly case, according to the principle of mate-rial implication, when a conditional is false. Themoral is that falsification always depends uponthe non-formal requirement that the system inwhich the putative counter-instance occurs beclosed.

41 See e.g. N. R Hanson, ‘A Picture Theory ofTheory Meaning’, The Nature and Function ofScientific Theories, ed. R. G. Colodny, pp. 233–73.

42 I have argued in paragraph 2 above thatwithout such a constraint on the content of pos-sible explanations, sorting them with respect totheir plausibility, there will be an infinite numberof possible explanations, even of equal simplicity.

The plausibility of a possible explanation cannotbe identified by purely syntactical or formal cri-teria alone but depends upon a complex relation-ship between what is so far known about the pro-cess generating the behaviour in question and es-tablished explanation patterns drawn from anal-ogous fields. It is thus in part a function of theexisting knowledge in which the predicates oc-curring in the possible explanations are alreadyembedded, so that the paradoxes of confirma-tion, etc. that flow from the insertion of artifi-cial predicates into already-functioning and well-connected scientific contexts cannot (at least atthat point of application) arise. Cf. R. Harre,‘Surrogates for Necessity’, Mind 1973, pp. 355–80.

43 Ibid. p. 366.44 Cf. M. Bunge, op. cit., p. 38.45 Cf. G. Schlesinger, ‘The Prejudice of

Micro-Reduction’, B.J.P.S., Vol. 12, pp. 215–24.46 Cf. K. Schaffner, ‘Correspondence Rules’,

Philosophy of Science, Vol. 36 (1969), pp. 280–90.

47 Given the B is law-like and allowing for


open systems we must say: x tends to do φin virtue of its nature N. A discussion of therather complex relationship between tendenciesand powers must be postponed to the appendixto this chapter. For the moment they may beregarded as a class of powers whose exercise isnormically qualified. But this is not a completeanalysis. For a power may be exercised whenthe behaviour is not law-like, so that it wouldbe wrong to attribute a tendency. The logic ofpower ascriptions, their role in science and theontological status of powers will be discussed be-low.

48 The second question is both distinct fromthe first and important. Because it raises thequestion of the range or scope of application ofthe statements expressing the tendencies of theindividuals concerned. It cannot be assumedthat all tendencies will be spatio-temporally uni-versal; for individuals and kinds may be trans-formed in time and bounded in space. A law mayof course be universal (transfactually applicable)within its range and restricted in this way.

49 Cf. M. Hollis and E. J. Nell, Rational Eco-

nomic Man, Chap. 7.50 See e.g. A. Flew, An Introduction to West-

ern Philosophy, p. 49; or E. Nagel, op. cit., p.37.

51 See e.g. L. Kolakowski, Positivist Philoso-phy, p. 34.

52 Cf. R. Harre, Principles of Scientific Think-ing, pp. 274–5.

53 Cf. K. R. Popper, Conjectures and Refu-tations, Chap. 5. However the protolaw itselfwhen it finally emerges, pari passu with its ex-planation after the limitations and modificationsnecessitated by the experimental process, maybe in a form far more complex and refined thanthat in which it was originally conceived (cf. S.Korner, Experience and Theory, passim). Thenormal response to a (genuine) counter-instanceis modification within a continuous research pro-gramme, rather than (as is implied by naive fal-sificationism) the complete abandonment of theoriginal conjecture and its replacement by a to-tally different one (cf. I. Lakatos, op. cit.).

54 Cf. R Harre, op. cit., p. 275.55 See A. J. Ayer, The Fundamental Ques-


tions of Philosophy.

56 See R. Harre and E. H. Madden, ‘NaturalPowers and Powerful Natures’, Philosophy, Vol.45 (1973), esp. pp. 223–30.

57 See e.g. P. Achinstein, Law and Explana-tion, pp. 13ff.

58 As the immediate objects of perception arenormally assumed to be short events or momen-tary states, perceptual access to things presup-poses a resolution of the problem of induction(to be discussed in paragraph 6 below).

59 J. R. Ravetz, op. cit., esp. pp. 47–59.

60 N. Chomsky, ‘Objectivity and LiberalScholarship’, American Power and the NewMandarins, pp. 23–129.

61 K. R. Popper, ‘Normal Science and ItsDangers’, Criticism and the Growth of Knowl-edge, eds. I. Lakatos and A. Musgrave, pp. 51–8.

62 P. K. Feyerabend, ‘Against Method’, Min-nesota Studies in the Philosophy of Science, Vol.IV, eds. M. Radner and S. Winokur, pp. 17–130.

63 See e.g. P. Duhem, op. cit., Chaps. 9 and10, or K. R. Popper, ‘The Aim of Science’, Ob-

jective Knowledge, Chap. 5.64 See P. K. Feyerabend, ‘Problems of Em-

piricism’, op. cit., pp. 168–71, and T. S. Kuhn,op. cit., pp. 100–1.

65 I have already suggested in paragraph 3above the role that the concept of powers mightplay here.

66 It is his failure to see this that I thinkleads Feyerabend into error. For he wants to say(a) that there is bound to come a time whenthe ‘alternatives’ do not share a single state-ment (including observation statement) in com-mon, yet (b) we could still ‘choose’ between thetheories, viz. in terms of the uninterpreted sen-tences that the scientists testing them would bemotivated to produce in observational contexts,op. cit., pp. 214–5. Now what is objectionableabout this suggestion is not only that such unin-terpreted sentences could never provide groundsfor a choice (cf. e.g. D. Shapere. ‘Meaning andScientific Change’, Mind and Cosmos, ed. R. G.Colodny, p. 61) but also the idea that we couldever be in a position to make such a choice. Forthis involves the hypostatisation of a whole his-


torical process of meaning-change and its encap-sulation in a single notional moment of judge-ment. In this way it involves a relapse backto the pre-relativistic notion that we can makejudgements outside some particular theory andsome particular position in theoretical time.

67 Cf. I. Lakatos, op. cit., pp. 159–65.68 Cf. T. S. Kuhn, op. cit., p. 59.69 See e.g. M. J. Mulkay, The Social Process

of Innovation, p. 12.70 This has been noted by J. R. Ravetz,

op. cit., pp. 116–18. But his use of Aristotle’sschema differs substantially from mine.

71 See K. R. Popper, Objective Knowledge,esp. Chaps. 3 and 4.

72 For the concept of powers continually ex-ercised we have of course groomed the conceptof tendency.

73 See e.g. A. C. Ewing, The Fundamen-tal Questions of Philosophy, pp. 159–81 and B.Blanshard, Reason and Analysis, Chaps. 11–12. Cf. also N. Maxwell, ‘Can there be nec-essary connections between successive events?’,B.J.P.S. Vol. 19 (1967), pp. 1–25; and M. Fisk,

‘Are there Necessary Connections in Nature?’Philosophy of Science, Vol. 37 (1969), pp. 385–404.

74 Cf. ‘Tania pushed the door open’ logicallyimplies ‘the door opened’. As Davidson has putit: ‘the truth of a causal statement depends uponwhat events are described; its status as ana-lytic or synthetic depends upon how they aredescribed’, op. Cit., p. 90.

75 Cf. M. Fisk, op. cit., p. 390.76 Cf. R. Harre, Surrogates for Necessity’, p.

380.77 H. Putnarn, ‘The Analytic and the Syn-

thetic’, Minnesota Studies in the Philosophy ofScience, Vol. III, eds. H. Feigl and G. Maxwell,p. 376.

78 Cf. E. H. Madden, ‘Hume and the FieryFurnace’, Philosophy of Science 1971, p. 66.

79 It is of course inconceivable that a fun-damental entity or entities should act inconsis-tently with its (their) nature. Hence in thelast (non-Laplacean) instance everything is as itmust be.

80 For a discussion of the history of ‘copper’


see M. Crosland, Historical Studies in the Lan-guage of Chemistry and R. Harre and E. H. Mad-den, Causal Powers, Chap. 1.

81 To use Goodman’s very useful term. SeeN. Goodman, The Structure of Appearance, p.130 and passim. Goodman himself attributes theterm to C. I. Lewis, Mind and the World Order.

82 K. R. Popper, The Open Society and ItsEnemies, Vol. I, p. 32.

83 K. R. Popper, Conjectures and Refuta-tions, Chap. 3, esp. p. 102.

84 A. Flew, op. cit., p. 450.

85 Ibid., p. 449.

86 S. Toulmin, op. cit., pp. 135–6.

87 E. H. Madden and p. Hare do not clearlydistinguish powers from their exercise in theircriticism of this Humean argument in ‘The Pow-ers That Be’, Dialogue 1971, pp. 12–31.

88 Cf. K. Campbell, ‘One Form of Scepticismabout Induction’, Analysis, Vol. 23, pp. 80–3,reprinted in ed. R. Swinbourne, The Justificationof Induction, pp. 144–8.

89 Strawson, An Introduction to Logical The-

ory, Chap. 9, Pt. II, esp.

90 K. R. Popper, op. cit., p. 41, n. 8.

91 M. Hollis, ‘Reason and Reality’, P.A.S.1967–8, esp. pp. 282ff.

92 See e.g. G Harman, ‘Enumerative Induc-tion as inference to the Best Explanations, Jour-nal of Philosophy 65 (1968), pp. 529–33.

93 The stringency of this requirement maybe relaxed in the kind of way indicated in 2.6above when we move from scientific contexts tothe rough-hand-ready generalizations of every-day life.

94 See e.g. M. Vernon, The Psychology of Per-ception.

95 See esp. J. J. Gibson, The Senses Consid-ered as Perceptual Systems.

96 See N. Goodman, Fast, Fiction and Fore-cast, pp. 73–80.

97 See e.g. C. G. Hcmpel, op. cit., Chap. 1.

98 P. Strawson, Individuals, pp. 167ff and pas-sim.

99 C. G. Hempel, op. cit., Chap. 8.

100 Cf. P. Strawson, op. cit., p. 46 (my em-


phasis).

101 Cf: ‘In the Newtonian world and in New-tonian science . . . the conditions of knowledgedo not determine the conditions of being; quitethe contrary, it is the structure of reality thatdetermines which of our facilities of knowledgecan possibly (or cannot) make it assessible tous. Or, to use an old Platonic formula: in theNewtonian world and in Newtonian science, it isnot man but God who is the measure of things’,A. Koyre, ‘The Influence of Philosophical Trendson the Formulation of Scientific Theories’, TheValidation of Scientific Theories, ed. P. G. Frank,p. 199.

3.8 Appendix: Natu-ral Tendencies andCausal Powers

229:1 I have argued that there is a pair ofontological distinctions, presupposed by theintelligibility of experimental activity andscientific training, at any one stratum orlevel of reality, between structures and theevents they generate (and things and thechanges that occur in them) and the experi-ences men have of them. These distinctionsmay be conveniently expressed by the for-mula Dr ≥ Da ≥ De,

1 where the specialcase Dr = Da = De, assumed to be sponta-neously satisfied by empirical realism, has infact to be worked for in the social activity ofscience. The possibility of Dr 6= Da impliesthat not all events, and that of Da 6= De

3.8. APPENDIX: NATURAL TENDENCIES AND CAUSAL POWERS 317

that not all experiences, are equally signifi-cant epistemically (see 1–6 above). Now thepostulation of this novel non-Humean on-tology of structures and generative mecha-nisms raises the question of how they oughtto be analysed. Possibility Da 6= De im-plies that they cannot be reduced to qual-ities; Dr 6= Da that they cannot be anal-ysed as dispositions. I have suggested thatstructures and generative mechanisms mustbe analysed as the tendencies and powersof enduring and transfactually acting things.In this appendix I want to clarify these con-cepts, and to defend the transcendental real-ist ontology from some possible misconcep-tions.

229:2/o A ‘tendency’ as I have been usingthe term up till now is just a power whichmay be exercised unrealized, a power normi-cally qualified. This qualification is necessi-

tated of course by the fact of open systems.I will call this concept of tendency the pri-mary concept. But there is another concept,in principle distinct from this, in which itfunctions so as to pinpoint the enduring ori-entations, rather than the possibility of thetransfactual activity, of things. On this con-cept a tendency is something more than apower. It depends upon distinguishing fromwithin the class of actions naturally possi-ble for a thing (which constitute the total-ity of its powers), in virtue of its being thekind of thing that it is, those which are typ-ical, usual or characteristic of that thing asdistinct from others of its kind. It is thefunction of this second concept to individ-uate natures within kinds, species r withingenera, individuals within classes, etc. It ismade possible by the fact that some com-plex structured objects reveal, in virtue of


their pre-formed structure, what I am goingto metaphorically characterize as an ‘onto-logical preference’ for some but not other ofthe natural possibilities open to them. Athing possesses powers in virtue of its fallinginto a natural kind tendencies in virtue of itsbeing one of a type within that kind. All men(living in certain kinds of societies) possessthe power to steal; kleptomaniacs possess thetendency to do so.

230:1 The distinctiveness of the two con-cepts of tendency is clear. To say Taniapushed the door open completely explainswhy the door is open and implies that shecan do it, i.e. has the power to do it. Butto say that she tends to push the door openis to say something more; which cannot beanalysed as when she exercises her power topush the door open, it tends to open (whichis just to normically qualify the exercise of

her power). There is a real difference in thekind of behaviour, and the state of the thingwhose nature is referred to, in the two cases.Men but not dolphins, can (i.e. possess thepower to) smoke; but some men are non-smokers. To attribute a tendency (in the sec-ond sense) is not just to normically qualifythe exercise of a power; but to say that someof the intrinsic enabling conditions of a rela-tively enduring kind for the power’s exerciseare satisfied; that the thing is predisposed ororiented towards doing it, that it is in some-thing of a state or condition to do it.

230:2/o In distinguishing in this way be-tween tendencies and powers we are able toavoid the dilemma of supposing either thatall behaviour is law-like or that some eventsare uncaused. For a power may be exer-cised and completely explain an event whenthe behaviour is not law-like (typical) and so


cannot be seen as the exercise of a tendencyin the second sense (which I will henceforthdenote as tendency2). Particular circum-stances may account for the exercise of apower; whereas they must be invoked to ac-count for the non-realization of a tendency2,if the conditions for its realization are satis-fied.

231:1 To say that a thing has the powerto do 0 is to say that it will do phi, in theappropriate circumstances, in virtue of itsnature (e.g. its intrinsic structure or geneticconstitution). I have suggested in paragraph3 above that in indicating the existence of areason, at the next highest level of reality, forthe manifest behaviour a power statementconstitutes a schematic explanation which isfilled out in the growth of science. To ascribea power is to make a statement about pos-sibilities which may not be actualized and

which are possessed by the thing whether ornot they are known by men; so powers can-not be reduced to their exercise or our igno-rance.

231:2 Now if powers are possessed bythings which act in open systems their exer-cise must be normically qualified; and theymust be seen as tendencies1. This is theconcept of tendency I have been using sofar. In order to apply any tendency1 ornormic statement we must know when theantecedent or stimulus conditions for themechanism it designates are satisfied. Butthis does not warrant the prediction of thetendency’s1, fulfilment, i.e. the consequent’srealization; which depends upon the systembeing closed, and in particular upon the non-intervention of countervailing causes.

231:3 If things are complex and pre-formed they may be in a relatively endur-


ing state or condition to exercise some butnot other of their powers. This is the con-cept of tendency2—the concept of tendencynormally employed in ordinary life. Whensuch things act in open systems the exerciseof tendencies2, will of course also have to benormically qualified. To say that a thing, X,has a tendency2, to do φ is thus to say :

231:4 (i) X has the power (or liability) todo (or suffer) phi;

231:5 (ii) X is in an enduring conditionto do phi, i.e. it is predisposed or orientedtowards doing phi;

231:6 (iii) X will do phi, given an appro-priate set of circumstances, in virtue of itspredisposition, in the absence of intervening(or countervailing) causes.

231:7/o That is, it is to say that X can doit, that most (or the most important) of theintrinsic enabling conditions for it are satis-

fied and that when the other conditions aresatisfied it will do it unless it is prevented.

232:1 Notice that a tendency2 statementpresupposes a power statement but the con-verse is not the case. For a thing must notonly possess the power but be in somethingof a condition to exercise it. Now lt mightbe thought, in view of this, that thoughboth tendencies, and powers designated pos-sibilities, tendencies2 designated less remoteones, possibilities closer to or in the processof realization. But this metaphor must beinterpreted very carefully. For the greater‘proximity’ of the tendency’s realization isstill a purely non-epistemic one. For I maybe absolutely certain that the stimulus orother conditions will be unsatisfied or thatcountervailing causes will intervene so thatthe tendency2 will be unexercised or if exer-cised unrealized respectively. Knowledge of


a tendency, is quite distinct from judgementabout the tendency’s2, realization.

232:2 It is a mistake to think of offset-ting causes as relatively short-run retardingbarriers which are sooner or later overcomeby the superior staying power of the ten-dency. This is to tacitly conflate epistemicand natural possibilities. Equilibria maylast for ever. A related mistake is to iden-tify the ‘lawfulness’ implicit in a tendency2,statement with the regularity or repeatabil-ity of a syndrome of behaviour. For sometendencies2, (to explode to commit suicide)have of necessity only a single manifestationand cannot be fulfilled twice.

232:3 A tendency2 may be revealed in ac-tion, even when it is unexercised; in whichcase our grounds for attributing it must beindirect. If it is not shown at all, i.e. hasno overt expression then our grounds for at-

tributing it must be based entirely on ourtheoretical knowledge of the thing. In gen-eral, the attribution of tendencies2, requiresmore about things to be known than is thecase in the attribution of powers. For wemust know that the enabling conditions forthe power’s exercise are satisfied and hencewhat they are.

232:4/o Tendency2 ascriptions may bemerely generic. There may be no uniqueset of conditions under which a tendency2

is exercised; and the way in which it is re-alised, if it is, may depend upon the partic-ular circumstances in which its exercise oc-curs. A kleptomaniac’s tendency is not nor-mally fixed on a particular shop and it isnot normally knowingly exercised under theeye of the law. (If it is, he ceases to be akleptomaniac and becomes something else.)Similarly the way in which Fido expresses his


generic tendency2 to bark may depend uponthe particular factor that excites it. He maybark viciously at an intruder but convention-ally at the postman, fearfully at the moonbut affectionately at another dog, arrogantlyat a cat and playfully at an old shoe.

233:1 Both (ii) and (iii) play an impor-tant role in indicating a characteristic ex-planatory problem and the appropriate di-rection of research. Thus the problem in-dicated in (iii) is not: what accounts forthe fulfilment of tendencies, given that theconditions for their fulfilment are realized?But rather, what accounts for their nonful-filment in these circumstances? (The for-mer needs no special explanation.) And ittells us what to look for: namely, the pres-ence of interfering causes. It should be notedthat a formulation which only included ref-erence to the circumstances would, if the

tendency2 were unrealized, set us off in thewrong direction, on a search for some ex-ceptionless generalization which covered thatcase too; so providing an empty and ever-expanding redescription rather than an ex-planation. Generality in nature lies in thingsnot conditions. Explanation in terms ofthe powers/tendencies conceptual networkavoids the circularity and triviality of expla-nations conforming to the Popper-Hempelmodel, where an event cannot be explainedother than by or without subsumption undergeneralizations. The former is always insuf-ficient and the latter mostly impossible. Soroom must be found for the concept of a gen-erative mechanism, activated under closedconditions, explaining regularities when theyoccur; and the concept of an agent bringingabout and so explaining an event in a givensituation without this being interpreted as


involving a claim to a regularity.

233:2/o Offsetting causes are often as-sumed to be always extrinsic. But the causeof a failure of a car to move when the gearis m neutral is not something distinct fromand extraneous to the mechanism responsi-ble for its normal motion. Science is neverconcerned with just listing causes but seeksalways to relate them, even when seeminglyopposed or contradictory, to common struc-tures. Now intrinsic offsetting causes mayor may not directly interfere with the oper-ation of the mechanism responsible for thesatisfaction of the intrinsic enabling condi-tions. If they do then we must say that thetendency2 is no longer possessed, as condi-tion (ii) ceases to be satisfied. But not allintrinsic offsetting causes are like that. Thusa person may possess a reason for acting ina certain way and not act in that way under

appropriate circumstances if, at the time, hepossesses in addition a set of overriding ormore compelling reasons for not acting inthat way. The idea of a perfect balance ofopposed causes is expressed in the theoreti-cal concept of an equilibrium. .

234:1 It is the specific role of (ii), I sug-gest, to indicate the existence of a level ofactivity within the thing which is ensuringor has ensured the satisfaction of the intrin-sic enabling conditions for phi. That is, itdirects our attention to the mechanisms re-sponsible for those relatively enduring statesof a thing that distinguish the possessionof a tendency2 from that of a mere power.The tendency2 statement is action-orientedby contrast with the simple power statementin two respects then: firstly, in indicating thepast or present activity of generative mech-anisms responsible for the persisting orien-


tations of things- and secondly, in indicatingthe possibility of their transfactual activity,i.e. the exercise of these tendencies irrespec-tive of their fulfilment. A simple power state-ment is, on the other hand consistent withcompletely quiescent or dormant things orthings which have a level of activity suffi-cient only for the . retention of that power.

234:2 Now in an important sense it is notwhether the tendency2 is realized or not thatconstitutes the scientifically interesting thingabout it (for this will always be to a greateror lesser extent circumstantially determined)but the reason for the relatively enduring ori-entation that differentiates the possession ofa tendency, from that of a power. Thus inthe case of the kleptomaniacs, the scientistwill be concerned with the reasons for histendency2, whether he suppresses it or not.Of course this need not be the chief concern

of social policy. But this merely underscoresthe latter’s distinctiveness from science. Forscience is concerned with the behaviour ofthings only in as much as it casts light ontheir reasons for acting and hence upon whatthey are.

234:3/o The underlying action indicatedin the tendency2 statement may be continu-ous or sporadic; and the onset of a tendency2

may be gradual or sudden. Moreover someof the mechanisms responsible for the pos-session of tendencies2, e.g. those posited inFreudian theory, may have ceased to oper-ate long before the time of their manifes-tation. The grounds for our attribution ofthe tendency must then of course lie en-tirely in our evidence for the pre-formed in-trinsic structure rather than in our evidencefor the present higher-order activity of thething. Such tendencies2 are latent It should


be stressed that we cannot assume that theunderlying action occurs in systems whichare themselves anything other than open;just as we cannot assume that underlyingthe action of those mechanisms are anythingother than more basic ones.

235:1 Some tendencies2 are powers whichare held in abeyance; and are straightawayexercised when the impediments to their ex-ercise are removed. Other tendencies2 arepowers which require for their exercise theactive stimulus of other things. One might inthe former case talk of ‘releasing rather than‘stimulus’ conditions. Releasing or stimulusconditions may be intrinsic or extrinsic tothe thing possessing the tendency2. Spon-taneity, self-determination and acting for areason all fall under the former paradigm. Itis better to distinguish between powers andliabilities than between active and passive

things. But within the category of extrinsiccauses we must remember to include pullsas well as pushes, structural relationships aswell as the momenta of other things.

235:2 Both components (ii) and (iii) of theanalysis of tendencies2 encourage a furtherinvestigation of powers: viz. into those re-sponsible for the persisting orientations ofthings and into those whose exercise is re-sponsible for the non-realization of knowntendencies2, when the conditions for theirrealization are satisfied. To sum up then atendency2 statement says there is a level ofactivity, perhaps unknown, intrinsic to thething, such that it is predisposed to performan action of a certain type. Its chief func-tion is to indicate the existence of a level ofactivity within the thing such that it is ori-ented towards some rather than other of thenatural possibilities open to it. In this way


it leads us to a more precise specification ofthe natures of particular things (or groups)within kinds.

235:3/o I now want to turn to some pos-sible misconceptions arising out of my useof the primary concept of tendency (whichI shall not denote by a subscript) and theconcept of power. The intelligibility of ex-perimental activity and the possibility of thepractical application of our knowledge de-pends, I have argued in Chapter 2, upon theassumption of the ontological independenceof causal laws (which I have analysed as thetendencies of things) from the patterns ofevents. That science can job back and forthfrom open to closed systems depends entirelyupon .the consideration that the natures ofthings and the generative mechanisms of na-ture persist and act in open and closed sys-tems alike. That is to say, the intelligibility

of scientific practice requires that we viewthem theoretically as at any moment of timeunaffected by the closure or otherwise of thesystems in which they occur. Clearly thisdoes not mean that natures and generativemechanisms must be viewed as eternal oras completely unaffected by the actions ofthings or their operations. Powers may waneand disappear. Mechanisms may be transformed. Such transformations must them-selves be analysed as events in open systems;and are in principle explicable in terms of anetwork of normic statements. The chang-ing is explained in terms of the unchanging,but the unchanging need only be relativelyso. Some fields may be irreducibly histori-cal. But the very move, which I have arguedis essential to our understanding of science,from talking of the occurrence of events tothe things and structures that account for


them presupposes that they are relativelypersistent. A non-enduring thing would bea mere event, a totally affected mechanismincapable of production.

236:1/o Now it is vital to distinguish theinitial conditions whose instantiation war-rants the application of a normic or tendencystatement, i.e. of a causal law, from thecomplete atomistic state-description that theempirical realist needs to sustain his analy-sis of laws. For on the transcendental realistview a law is applicable if the mechanism itdesignates is set in motion, i e. if the condi-tions explicitly mentioned in the protatis ofthe law-like statement are satisfied, whetheror not the system is closed i.e. whether ornot what are sometimes referred to as theboundary conditions’ are satisfied, and hencewhether or not the consequent is realized.The ‘boundary conditions’ for the transcen-

dental realist are conditions for the experi-mental testing, not the applicability of a law.To say that laws are applicable if and only iftheir ‘boundary’ as well as their initial condi-tions are satisfied is useless because it is menthat ensure the satisfaction of the boundaryconditions. And this activity is only intelli-gible if laws continue to apply in open sys-tems, where the so-called ‘boundary condi-tions’ are not satisfied. We isolate systemsto understand mechanisms; but the mecha-nisms endure and act quite independently ofour activity.

237:1 The ordinary things or standardparticulars of the world are metaphysicallycompounds, just as ordinary events are con-junctures. The philosophy of science hasnoted, quite correctly, that the objects of sci-entific thought are ‘ideal’ or abstract withrespect to such things and events. But the


transcendental realist sees such objects asreal. For him the world is composed of realthings and generated by real mechanisms. Itis the world itself, not our thought of it, thatis abstract and ideal. The world consists ofchemical elements, electrons, quarks; its phe-nomena are generated by actions, reactions,forces, fields. The received tradition in thephilosophy of science has seen such thingsas inventions of men designed to gloss, andaccount for deviations in, our familiar expe-riences. But transcendental realism, startingfrom the premise of the contingency of ourown experience, sees nature as real; and sci-ence as our persistent effort to understandit.

237:2 In paragraph 3 I argued that theconcept of powers, in combining a specificattribution of behaviour with an unspecificattribution of structure, plays a key strata-

bridging role. For in hinting at the perhapsto be discovered reason [at Stratum II] forbehaviour [at Stratum I] it represents an im-portant moment in the process of scientificdiscovery. Its structure reflects this, in in-corporating both a characterization at thescience s point of departure and a-referenceto its intended destination. To say that Xhas the power to do φ is to say that it will doφ in the appropriate circumstances in virtueof its nature (e.g. structure or constitution);2

that is to say it will do it in virtue of its beingthe kind of thing that it is.

237:3/o Now to ascribe a power is to sup-pose that there is a real basis for the posses-sion of that power independent of whetherthe power is exercised or not.3 But it is notpossible to say whether that real basis canbe qualitatively described or not. Some realbases, such as the molecular structure ac-


counting for the fragility of glass, can bequalitatively described; others, such as thestructure of a gravitational or magnetic field,can not. Hence it is contingent whetherwe can know about a thin anything otherthan its causal powers; and it is contingent,when we can not, whether a thing is any-thing other than its causal powers. NowMackie has objected to the concept of pow-ers on the ground that if it is to form the ba-sis for a new metaphysics of science it mustviolate ‘Hume’s principle that there can beno logical connections between distinct ex-istences’.4 In support of this he asks ‘whatwould be the point of showing that opiumcontains morphine which (despite its name)is only contingently related to sleep, if weknew already that opium contained an in-trinsic power whose presence entailed theproduction of sleep?’5 Now I have of course

already argued that if science is to be possi-ble there must be necessary connections be-tween some but not other events; and thatsuch necessary connections are provided bythe enduring mechanisms that bind some butnot other events together and that exist asthe powers of things. Now the knowledge ofthe chemical nature of opium which allows usto deduce its tendency to produce soporificeffects in men is not a priori; but discovereda posteriori, in the ongoing process of sci-ence. Once we know a thing’s nature then wecan deduce its tendencies, how it will behaveCP. But typically in science when we ascribepowers we do not know a thing s nature butonly believe that it is in virtue of the thing’snature, whatever it is, that it does what itdoes (i e that its behaviour is necessary, notaccidental). A power-statement then acts asan imperative for scientists to find, and as a


temporary place-holder for, that explanationwhich, by capturing the essence of the thing,will allow the most stringent possible crite-rion for our knowledge of natural necessityto be satisfied.

238:1 When we know what a thing is weknow what it will tend to do, if appropriatecircumstances materialize.

3.9 Footnotes to Ap-pendix of Chapter 3

1 Dr = domain of the real; Da = domain of theactual; and De = domain of the empirical in thesense specified in 1.6 above.

2 Cf. R. Harre, Powers’, B.J.P.S. Vol. 21(1970), p. 85 and Principles of Scientific Think-ing, p. 270. See also D. Armstrong, DispositionalProperties’, Analysis 22 (1961–2), pp. 44–46 and

Belief, Truth and Knowledge, esp. pp. 11–16.3 Cf. D. Armstrong, A Materialist Theory of

Mind, p. 86.4 J. L. Mackie, Truth, Probability and Para-

dox, p. 137.5 Ibid., pp. 143–4.

Chapter 4

Metaphysics and thePhilosophy of Science

239:1 If science is to be possible men mustpossess certain essential powers. Amongthese is the power of affecting the sequences

of states and events in the world in the senseof bringing about effects which but for theiraction would not have been realized. In this

331

332 CHAPTER 4. METAPHYSICS AND THE PHILOSOPHY OF SCIENCE

way men contribute to the universal mael-strom of existence. More specific to men istheir power to initiate and prevent changein a purposeful way. The possession of thispower seems to stem from the fact that menare material things with a particular degreeof neuro-physiological complexity which en-ables them to monitor and control their ownactions. Foremost among the powers nec-essary for science and, as far as we know,distinctive of men is their power of inten-tional action, which enables them to act self-consciously on the world: that is not just tomonitor and control their performance, butto monitor the monitoring of their perfor-mance: to plan, to act and so to make ananticipatory commentary come true.

239:2 Among the powers falling withinthis genus perhaps the most basic and cer-tainly the most studied is the power to ac-

quire and use language. The latter impliesat the very least that we are material thingswith a point of view in space and an exis-tence in time, so that we must be able tocommunicate with each other from differ-ent spatio-temporal locations. And the for-mer implies at the very least that we mustbe able to communicate with each other onthe basis of the possession of differing cogni-tive equipment, so that reindividuation andrecharacterization must be possible and thedefinitions of terms must be at least partiallyopen. However language is by no meansthe only vehicle of thought; nor is language-using our only intellectual skill. Pictures, di-agrams and iconic models play, it has beenargued, an indispensable role in scientificthought.

239:3/o Now endowed with our ensembleof intellectual powers we are able not just to

333

describe the patterns of events but by imag-ining structures (which may come to be es-tablished as real) to grasp the mechanismsof their production.

240:1 Closely related to these powers isour capacity to design, manufacture and usetools. This inter alia enables us to act at adistance; and in the course of science it hasenormously extended our powers of percep-tion and detection. More generally, it hasenormously increased our powers of inter-vention in and control over, as we say, ‘thecourse of nature’.

240:2 In all these activities—involvingperceiving, manipulating, discoursive andpictorial thought—the role of the nature ofmen as mechanisms capable of second-ordermonitoring and feed- back is vital. In virtueof it our diverse perceptual, manipulativeand cognitive performances become possi-

ble objects of self- criticism; and inter aliathrough our capacity to make public ourmonitored commentaries upon these moni-tored performances open to the criticism ofothers. Simple rudimentary perception, suchas sniffing, and basic actions, such as rais-ing one’s arm, which depend in general onlyupon a first-order monitoring, are in them-selves relatively insignificant in science. Thechemist who sneezes while conducting an ex-periment is aware of himself in a certain way,but not in the way that the physiologist un-derstands all sneezes, including his own. Forthe latter knows the mechanism of sneezing,while the former merely experiences its ef-fects. (The chemist, unlike the layman, mayhowever be aware of the cause, in the senseof antecedent or stimulus conditions for themechanism, in this case).

240:3/o Scientific perception is non-simple


then in the sense that it must normally betheoretically-informed. And most scientificactions are non-basic (in Danto’s sense)1 inthat they consist in doing things to bringother things about (i.e. phiing to ψ),2 that isinitiating sequences of events whose outcomeis both distant and planned. It is throughour acquired skills of perception that wecome to be in a position to formulate propo-sitions concerning the behaviour of things, toidentify and describe the flux of events. Butit is through our manipulative powers that,by interfering with the course of nature (thisflow of events), we are able to check the re-ality and study the operation of the hypo-thetical generative mechanisms that in thescientific imagination we picture as respon-sible for their behaviour.

241:1 In such a move we disrupt the se-quences of events to identify the underlying

causal laws. Its significance - the significanceof experimentation—lies in the considerationthat it gives us access to causal structuresthat exist and act independently of the ex-periment. In the same way the significanceof perception lies in the consideration that itgives us access to things that exist indepen-dently of it. And in a sense scientifically-relevant perception depends upon an an-terior disruption too—of the sequence ofcommon-sense experiences in a scientific ed-ucation or training.

241:2 Now the crux of my objection to thedoctrine of empirical realism should be clear.It is that it cannot sustain the intelligibilityof perception and experimental activity; andthat in positing a correspondence (or even inits positivist form an identity) between:

241:3 (a) sense-experiences, which are ingeneral only made significant by the trans-

335

formation of antecedent common sense, andtheir objects, viz. events and the states ofthings, as expressed in the concept of theempirical world; and

241:4 (b) constant conjunctions, which arein general only made possible by human ac-tivity, and causal laws, as expressed in theidea of the actuality of causal laws

241:5 it makes impossible both scientificchange (at least in our descriptions of possi-ble objects of experience) and the scientificexplanation of things existing and acting inopen systems (that is, in systems where in-variant conjunctions of events have not beenfound or made to prevail). In this way em-pirical realism comes to seriously understatethe critical significance and scope of applica-tion of science.

241:6/o (a) depends upon the possibil-ity of identifying or characterizing the same

thing or phenomenon (or explaining thesame event) in a different way, as well as fromdifferent spatio- temporal locations. (b) de-pends upon the intelligibility of the idea ofcausal structures existing and acting whereno empirical regularities prevail. Both de-pend upon the non-identity and possible dis-juncture of the terms of the pairs; that is,upon the possibility of one varying with-out the other. Thus we must have a con-cept of facts as social products and what isdescribed and/or explained as independentof men. And we must have a concept ofthe mechanisms that generate phenomena ir-reducible to the phenomena they generate.And so two dimensions must be establishedin the philosophy of science: a transitive di-mension, in which experiences and conjunc-tions of events are seen as socially produced;and an intransitive dimension, in which the


objects of scientific thought are seen as gen-erative mechanisms and structures which ex-ist and act independently of men.

242:1 Now empirical realism is character-ized by the absence of an intransitive dimen-sion and of a transitive dimension with re-spect to experience. But I have already ar-gued that this merely results in the genera-tion of an implicit ontology, based on the cat-egory of experience and an implicit sociology,based on the category of men: that is, in thegeneration of an ontological atomism and anepistemological individualism (at least withrespect to experience). I have further sug-gested in 3.4 above that underpinning em-pirical realism is a particular conception ofmen (at least with respect to experience):in which men are seen as passive sensors ofgiven facts and recorders of their given con-junctions. This model of man constitutes,

together with the celestial closure and theclassical paradigm of action, the analogicalgrammar’3 of empirical realism, the scien-tific substance that lends plausibility to itsmetaphysics, that gives credence to its philo-sophical form.

242:2 If there is a basic or fundamentallevel of knowledge, at which what is ex-pressed is certain and given independentlyof any human activity and reflects (and ulti-mately constitutes) the world, then its con-stituents must be atomistic. For if theywere not atomistic they would be them-selves susceptible of analysis, and so requirejustification. Hence it is the characteris-tics of the concept of knowledge to whichempirical realists have subscribed, that ac-counts for the disastrous ontology that wehave examined. It is the desire for incorrigi-ble foundations in a level which constitutes

337

the world that generates the ontology ofatomistic events and closed systems respon-sible for the problems and errors analysed inChapters 2 and 3. Against this I have arguedthat knowledge depends upon knowledge-like antecedents and that the world is inde-pendent of men. But if it is the requirementsof an incorrigible ground for knowledge ofthe world in empirical realism that generatesthe implicit ontology of empirical realism, itis the model of man necessary to sustain theincorrigibility of this ground that forms thelynchpin of the tradition; and hence explainsin the last instance the doctrines and prob-lems we have been criticizing and examin-ing This is of course as one would expectgiven the epistemological bias and anthro-pocentricity of our philosophical thought.

243:1 Empirical realism is faced with thefollowing trilemma: either (i) our knowledge

is determined (in part or in whole) by expe-rience or (ii) it is fixed a priori as a necessarycondition of experience or (iii) it is free (inthe sense of unconstrained by experience).4

I said at the outset that part of my inten-tion was a philosophy for science. What,then, are the implications of these three po-sitions for scientific practice? It is clear thatif knowledge is regarded as justified in termsof given experience we have the makings ofwhat is in effect a conservative ideology, inwhich the current experiences of a science arerationalized in being thought of as natural,given or implied by the very nature of thingsthemselves. If (ii) is seriously considered itwould tend to have the same conservative ef-fect. It is less obvious that (iii) would too—until we remember that science is an insti-tutionalized activity in which if to be freeis to be unconstrained by the possibility of


critical experience a self-perpetuating dogmamay ensue. Examples could be drawn fromthe history of physics and chemistry and eco-nomics to show how empirical realism hasfunctioned in this way as a conservative ide-ology for science.

243:2/o A philosophical system may serveto rationalize the practice of a science in an-other way, viz. through its own substantivescientific analogies and the correspondenceor resonance they find in the science. Thusempirical realism readily finds an echo inmechanistic explanations in physics and psy-chology and reductionist programmes in so-ciology and biology. One would expect sucha link to be particularly strong in the caseof the social sciences, if only because of thegreater sensitivity on the part of social sci-entists to the philosophy of science—a sensi-tivity partly explained by the apprehension

of their own underdevelopment and partlyby the consideration that science in generaland social science in particular is also partof their own field of inquiry. The resonancebetween Skinner and Hume or Marshall andMill is too clear to need further remarking.

244:1 In addition to the effect of its con-cept of knowledge and its own substantivescientific analogies a philosophy of sciencemay of course also, and most obviously, ra-tionalize the practice of a science throughits own conception of the methods of scienceand the appropriate objects of scientific in-quiry. If I am right in my argument themethodology of empirical realism is not thatof science but if it were to be acted upon itwould have the most deleterious effects onits practice.

244:2 Now if empirical realism dependsheavily for its plausibility upon its analogical

339

grammar, it could be asked upon what sub-stantive scientific analogy does transcenden-tal realism depend? The answer is, I think,none. All philosophical argument that is notexplicitly transcendental depends, I think,upon more or less strained scientific analo-gies or tacitly presupposed substantive the-ories. But a transcendental argument whosepremise is explicitly stated need not dependupon any particular theory, other than thosebound up in the activity that is its objectand which is its task to explicate.

244:3 The central argument of this study,establishing an ontological distinction be-tween causal laws and patterns of events (theindependence of the domains of the real andthe actual, the irreducibility of structuresto events), has turned on the possibility ofexperimental activity. Now as it is clearthat experimental activity is impossible in

the social sciences and at the very least de-void of the same significance in psychologyas it possesses in physics and chemistry, Iwant to round off my argument by consider-ing whether something analogous to the con-trolled investigation of nature, making pos-sible the experimental confirmation and fal-sification of theories, might be possible inthe social sciences and psychology and otherfields where experimental activity is impos-sible or more or less seriously circumscribed.

245:1 Three points are clear. First, thatthere is a general problem of confirmation (orcorroboration) and falsification in the non-experimental sciences. Second, that thoughwe can assume that there are explanationsfor social and psychological phenomena (andunder social and psychological descriptions),we cannot assume that the social or psy-chological sciences have got anywhere near


them. Thirdly, that any adequate solution tothe methodological problems (and in partic-ular problems of confirmation and falsifica-tion) of the non-experimental sciences mustdepend upon a more adequate conception ofnatural science than that which has so farinformed discussion of them.

245:2 An awareness of the general prob-lem of confirmation and falsification in thenon-experimental sciences is shown by thefollowing quotation:

245:3 The linguist . . . is studying one fun-damental factor that is involved in perfor-mance, but not the only one. This idealiza-tion must be kept in mind when consideringthe problem of confirmation of grammars onthe basis of empirical evidence. There is noreason why one should not also study the in-teraction of several factors that are involvedin complex mental acts and that underlie ac-

tual performance, but such a study is notlikely to proceed vary far unless the sepa-rate factors are themselves reasonably wellunderstood.5

245:4 Now if I am right in arguing that thesignificance of experimental activity in nat-ural science is that it gives us access to en-during and transfactually active structuresand that it is only under closed conditionsthat confirmation and falsification of theoryis possible, then we are in a better positionto see that the central problem of the psy-chological and social (and other non- experi-mental) sciences is that of devising (or recon-structing) an analogous procedure of inquiryand selectively empirical confirmation (andfalsification) and to appreciate the great gulfthat must separate them, in the absence ofsuch a procedure, from the sciences of na-

341

ture.

245:5/o In the case of psychology theagent’s capacity to give a commentaryon intentional action might provide anexperiment- analogue, though one cannotrule out the possibility that new conceptswill have to be given by the investigator tothe agent (even e.g. in the case of the identifi-cation of emotions). Much social science canbe seen to depend upon attempted real def-inition of forms of social life which have al-ready been identified under certain descrip-tions and are known by the agents who par-ticipate in the social activities concerned un-der those or other descriptions. Social struc-tures, unlike natural structures, cannot ex-ist independently of their effects. Thus realdefinitions of concepts such as capitalism,democracy, power, love can only be Justi-fied by their capacity to render intelligible

a certain domain of phenomena. I suggestthat they are falsified by their incapacityto explain in a non-ad-hoc way a range ofphenomena that takes on a special signifi-cance for the agents that participate in theforms of social life they define. Thus it wasthe mass unemployment of the 1930’s thatdemonstrated the inadequacy of the neo-classical system and provided the motor forthe Keynesian innovation which showed howan unemployment equilibrium was possible.Clearly I have no space to defend or elabo-rate these suggestions here. It is sufficientfor our purposes merely to note the prob-lem: what are the enduring and transfactu-ally active ‘mechanisms’ of the sciences of so-ciety and man? Transcendental realism con-ceives the various sciences as unified in theirmethod but specific to (or differentiated with


respect to) their particular objects.

246:1 Now once generative mechanismsare seen to be the objects of scientificthought, it can be seen that four questionscan be asked of any generative mechanismG: –

246:2 (i) is the mechanism enduring?

246:3 (ii) is the mechanism operating?

246:4 (iii) are the results of the activity ofthe mechanism unaffected by the operationsof others (of either the same or different

246:5 (iv) are the results of the activityof the mechanism perceived or otherwise de-tected by men?

246:6/o Now if one assumes, as the ac-tualist does, that laws are empirical invari-ances (universal empirical generalizations)one cannot sensibly ask these questions.Two consequences should be noted. First,a neglect of the conscious human activity

that is necessary for the generation of event-invariances under significant descriptions,that is of the activity involved in (ii)-(iv).Thus the stimulus and enabling conditionsfor the operation of the mechanism must besatisfied; the mechanism must be isolatedand the flux of conditions held constant orotherwise controlled; and skilled observersmust be present to perceive or detect it. Andsecondly, neglect of the possibility of non-enduring mechanisms, of laws which thoughuniversal and normic in form (i.e. transfac-tually applicable) are themselves bounded inspace and restricted in time. On the tran-scendental realist ontology, the descriptionof what the world must be like if science isto be possible, the classical principle of indif-ference (or invariance) applies only to struc-tures, not to events. But structures maythemselves be transformed; and so concepts

343

of diversity and change, like that of struc-ture, squeezed out by the implicit undiffer-entiated ontology of empirical realism, maycome to occupy as significant a place in on-tology as the concept of indifference.

247:1/o I have stressed throughout thisstudy the diversity of phenomena and theautonomy of the various sciences (both fromone another and with respect to commonsense), the opacity of their concepts and thestrangeness of the objects with which theyhave to deal. Natural processes are indepen-dent of man but man is himself in, and incontinual interaction with, nature. Natureitself is diversified and complex. But I doubtwhether our concept of nature can be under-stood in isolation from our concept of man;and I doubt whether we understand the lat-ter. The scope of this enquiry has how everbeen strictly philosophical; and its conclu-

sions apodeictic. I have shown the struc-tured and intransitive character of the ob-jects of scientific enquiry to be a conditionof the intelligibility of experimental activ-ity and the social nature of knowledge to bea condition of the intelligibility of scientifictraining. (The laboratory and the classroomare the two most under analysed, and yetthe two most obvious, sites of science). Ihave isolated the conditions of the plausibil-ity of the doctrine of empirical realism andshown then to be very special. A conceptionof both nature and our knowledge of natureas differentiated and stratified has been ad-vanced. And I have developed two criteriafor the adequacy of any account of science:viz. its capacity to sustain the possibility of aworld without men and the impossibility ofknowledge without antecedents. These es-tablish the necessity for both of what I have


called the intransitive and transitive dimen-sions of the philosophy of science: any ac-count of science that does not view knowl-edge as socially produced and the objects ofknowledge as independent of men must beruled out as a possible account of science.

248:1 If science is to be possible the worldmust be one of enduring and transfactuallyactive mechanisms; and society must be astructure (or ensemble of powers) irreducibleto but present only in the intentional ac-tion of men. Science must be conceived asan ongoing social activity; and knowledgeas a social product which individuals mustreproduce or transform, and which individ-uals must draw upon to use in their owncritical explorations of nature. Science is aprocess in motion, continually on the movefrom manifest behaviour to essential nature,from the description of things identified at

any one level of reality to the constructionand testing of possible explanations and thusthe discovery of the mechanisms responsi-ble for them. This process necessitates theconstruction of both new concepts and newtools (or the resurrection or refinement of oldones). The aim of science is the discoveryof the mechanisms of the production of phe-nomena in nature; and it proceeds by way ofa dialectic of taxonomic (or descriptive) andexplanatory knowledge, in which the con-flicting principles of empiricism and rational-ism can be reconciled, a dialectic which hasno foreseeable end.

248:2/o In order to render intelligible sci-entific change and to reconcile it with theidea of scientific progress we must have theconcept of an ontological realm, of objectsapart from our descriptions of them. We canthen allow, for example, that theory Ta is

345

preferable to theory Tb, even if in the termi-nology of Kuhn and Feyerabend it is ‘incom-mensurable’ with it if theory Ta can explainunder its descriptions almost all the phe-nomena p1 . . . pn that Tb can explain underits descriptions Bp1 . . . Bpn plus some signifi-cant phenomena that Tb cannot explain. Wecan speak in this way in the meta-languageof philosophy, and we must speak so if weare to retain the idea of scientific progresswithout falling back on the idea of certainfoundations of knowledge or theory-free ex-perience. It is the intuition of this neces-sity that accounts, I think, for the readinesswith which some philosophers of science haveembraced Tarski’s theory of truth.6 But thistheory cannot help us to resolve the problemposed by the apprehension of the general rel-ativity of our knowledge: viz. that wheneverwe speak of things or of events etc. in sci-

ence we must always speak of them and knowthem under particular descriptions, descrip-tions which will always be to a greater orlesser extent theoretically determined, whichare not neutral reflections of a given world.Epistemological relativism, in this sense, isthe handmaiden of ontological realism andmust be accepted. Now this does not meanthat it is impossible to communicate betweendifferent theoretical or conceptual schemesor that a scientist cannot know the same ob-ject under two or more different descriptions.To show the difference between say Newto-nian and Einsteinian dynamics and that thelatter is an advance on the former a scien-tist must be capable of doing so. Similarlythough there is no guarantee of successfulcommunication between the adherents of twodifferent conceptual schemas, there is no in-evitability about failure (It is difficult to un-


derstand the concept of total failure.) Epis-temological relativism insists only upon theimpossibility of knowing objects except un-der particular descriptions. And it entailsthe rejection of any correspondence theoryof truth. A proposition is true if and onlyif the state of affairs that it expresses (de-scribes) is real. But propositions cannot becom pared with states of affairs; their rela-tionship cannot be described as one of corre-spondence. Philosophers have wanted a the-ory of truth to provide a criterion or stampof knowledge. But no such stamp is pos-sible. For the judgement of the truth of aproposition is necessarily intrinsic to the sci-ence concerned. There is no way in whichwe can look at the world and then at a sen-tence and ask whether they fit. There is justthe expression (of the world) in speech (or

thought).

249:1/o Transcendental idealists are fondof saying that either knowledge must con-form to objects or objects conform to knowl-edge: that either how we speak must be afunction of things or things must be a func-tion of how we speak.7 But this dichotomyis bogus. Science is an activity, a process inthought and nature which attempts to ex-press in thought the natures and constitu-tions and ways of acting of things that existindependently of thought. Thought has a re-ality not to be confused or identified with thereality of its objects: knowledge may changewithout objects and objects change with-out knowledge. There is no correspondence,no conformity, no similarity between objectsand thought. Thoughts are only like otherthoughts objects (including thoughts) simi-lar to or identical with other things. Things

347

exist and act independently of our descrip-tions but we can only know them under par-ticular descriptions Descriptions belong tothe world of society and of men; objects be-long to the world of nature. We express [ourunderstanding of] nature in thought.

250:1 Science, then, is the systematic at-tempt to express in thought the structuresand ways of acting of things that exist andact independently of thought. The world isstructured and complex and not made formen. It is entirely accidental that we exist,and understand something about our bit ofit. It is important to avoid the epistemicfallacy here. This consists in confusing theontological order with the epistemic order,priority in being with priority in decidingclaims to being, the question of what has rel-atively underived (or independent) existencewith the question of what entitles us to re-

gard some kinds of statements as groundsfor other kinds of statements, etc. In par-ticular the question of what is capable of in-dependent existence must be distinguishedfrom the question of what must be the casefor us to know that something is capable ofindependent existence Thus electrons couldexist without material things; but we couldnot know this proposition, let us say P, un-less there were material things. The truth-conditions for our knowledge of P are not thesame as the truth-conditions for P. Therecould be a world without men; but therecould not be knowledge without antecedents.



1 See A. C. Danto, ‘Basic Actions’, A.P.Q. Vol.2 (1965), reprinted in The Philosophy of Action,ed. A. White, pp. 43–58.

2 Cf. G. H. von Wright, op. cit., pp. 66ff.3 To borrow Buchdahl’s useful concept. See

G. Buchdahl, op. cit., p. 3 and passim.4 Radical conventionalists adopt a mixture of

the first and third horns in the sense of allow-ing that we are free to decide what constitutesexperience.

5 N. Chomsky, ‘Problems of Explanation inLinguistics’, Explanation in the Behavioural Sci-ences, eds. R. Borger and F. Cioffi, pp. 427–8. Ihave of course already argued against the use ofthe concept of ‘idealisation’ to refer to generativestructures in 2.4 above.

6 See, e.g. K. R. Popper, op. cit., p. 224.7 See e.g. H. Schwyzer, Thought and Real-

ity: The Metaphysics of Kant and Wittgenstein’,Philosophical Quarterly, 1973, p. 205.

Chapter 5

Postscript to the SecondEdition

251:1 In this postscript I wish to clarifyambiguities in two key terms in the text;comment on what I now see as its principalweaknesses; and indicate some of the ways in

which I think further work is necessitated.

251:2 Two terms, viz ‘law’ and ‘cause’,are used in a systematically ambiguous waythroughout the book. In general I think that

349

350 CHAPTER 5. POSTSCRIPT TO THE SECOND EDITION

the context always determines which usage isintended. But my failure to distinguish themexplicitly may have been a source of gratu-itous misunderstanding.

251:3/o The term ‘law’ is customarily usedto refer both to statements of law and towhat such statements designate. A distinc-tion between the two is, of course, implicit inthe basic distinction of the work, that is be-tween the transitive and intransitive dimen-sions in the philosophy of science. Yet theterm ‘law’, together with the terms describ-ing their characteristics (‘normic’, ‘transfac-tual’, etc.), is used indifferently in the textto refer to both. A critic determined to becaptious may find confusion here where nonein fact exists.1 Were I to rewrite the book Iwould restrict my use of the term ‘law’ to theconcept in the intransitive dimension, alwaysqualifying its use to denote the concept in

the transitive dimension by some term suchas ‘statement’.

252:1/oo Secondly, I use the term ‘cause’to refer both to the antecedent event, con-dition or agent which triggers a mechanismand to the mechanism (and a fortiori the lawit grounds) itself. Thus I talk both of men ascausal agents of sequences of events and ofthe causal laws to which their activity maysometimes give them access, but of whichthey are not of course (in general) the agents.This duplex use of the concept of causalityat a crucial state in the argument may haveconfused some readers. Thus I do not ar-gue (on pp. 33ff), as e.g. Sharpe supposes,‘that the causal agency involved in the sci-entist’s intervention cannot itself be simplyreduced to sequence’.2 Rather, I argue thatcausal laws cannot, if we are to render in-telligible the significance of and necessity for

351

the scientist’s experimental activity, be re-duced to sequence. My main criticism is ofthe Humean account of laws. Only in a sub-sidiary argument (see e.g. pp. 66, 117) do Iobject to the Humean analysis of the agencyinvolved. The former implies the latter, butthe converse is not the case (as is shownby the possibility of Davidsonian anomalousmonism).3 The importance of this is thatmany philosophers, particularly those influ-enced by hermeneutical thought, such as vonWright and Apel, have recently drawn atten-tion to our causal activity (in experimenta-tion) in science, taking this as a criticism ofthe empiricist account of causality, withoutrecognising its significance (cf. p. 54 above):viz in yielding a definitive criticism of theempiricist account of laws, and hence of thewhole wretched ontology of empirical real-

ism.4

255:1 In general the work has been sub-ject to three kinds, or levels, of criticism: (I)of the transcendental method employed; (II)of the particular arguments used; and (III)of the results obtained. I shall restrict myremarks to (I) and (II), since if the argu-ments are valid and the method sound, theresults must follow. It is, however, worthmentioning that many critics of, or scepticsabout, (I) and/or (Il) have been willing toconcede that transcendental realism of thesort elaborated here, or something very likeit, ‘has been implicit in the thinking [activ-ity] of most creative scientists from Galileo toour own time’.5 Now, if this is the case, andtranscendental realism does successfully re-capture what Bachelard has called the ‘diur-nal’ philosophy of the scientists, i.e. the phi-losophy implicit in their spontaneous prac-


tice. And if it is also the case that thetheses of transcendental realism depart rad-ically from the ‘nocturnal’ philosophy of thephilosophers, forged in the schools of empir-ical realism, to which scientists too tend toreturn when they reflect upon their practice,then the problem arises of deciphering themeaning, and explaining the mechanism, ofthis discrepancy.6 To this problem I will re-turn.

255:2/o A good example of the kind ofcriticism directed against transcendental re-alism at levels (I) and (II) is provided ina recent book by Ruben.7 Ruben arguesthat though reality is indeed, as I claim,structured and intransitive, there cannot beany interesting non-question-begging validdeductive arguments for this conclusion. Tothis end he seeks to demonstrate that myargument for the structured nature of real-

ity is invalid, and that my argument for itsintransitive nature itself presupposes what Iam trying to prove.

256:1 Ruben claims that my contentionthat the experimental scientist produces asequence of events is false. For, though heproduces the antecedent a, and thus indi-rectly produces the consequent b, it does notfollow, given that a occurs, that he producesthe sequence a, b.8 For it might well be thecase, according to Ruben, that whenever aoccurs (however produced) so does b. Thatis, I have not eliminated the possibility ofempiricist reconstructions of causal laws assequences.

256:2/o But of course I have. For ourcausal activity is as much a necessary con-dition for the realisation of the consequentb, given that a occurs, as it is (in the con-text of an experiment) for the occurrence of

353

the antecedent. I explicitly designate (onp. 53 above) the former activity as ‘exper-imental control’ and the latter as ‘experi-mental production’. Experimental produc-tion is necessary to ensure the operation ofthe mechanism; experimental control for theoccurrence of a closure. My argument is thatwithout our causal activity, given a, b maynot, and in general will not, occur. Patently,if it is the case that our causal activity isnecessary for the realisation of the conse-quents of laws, they just cannot be glossed,without absurdity, as empirical regularities.That is, if it is not the case that whenevera then b then in making a claim about acausal law, we cannot, if we are to sustainthe intelligibility of the experimental estab-lishment and transfactual application of ourknowledge, be making a claim about a se-quence of events. Instead we must be con-

strued as making a claim about somethingthat bears only a contingent relationship tothe actual world (including that significantsubset of it produced by human work). Thisclaim, I have argued, is about the operationof a tendency or the working of a mechanismirrespective of its actualisation in any partic-ular outcome. And if that is the case, thenall the theories based on the flawed princi-ple of empirical-invariance—from the consis-tency condition of monistic historiographyof science to Feyerabend’s ‘dadaism’,9 fromthe Popper-Hempel theory of explanation tostandard (empirical realist) analyses of coun-terfactuals (see p. 158 above), from the 2ndAnalogy to Bachelard’s cogitamus10—mustall be radically wrong.

257:1 It is not clear to me why Rubenshould take it as an objection to transcen-dental arguments, and not say mathemati-


cal proofs, that ‘in a deductively-valid ar-gument nothing can appear in a conclusionwhich was not already, at least covertly, inthe premisses’11. Still, if it is the case thatphilosophy is, as I have claimed it can be,a conceptual science, then, like any science,it ought to be able to tell us something wedid not already know: it ought to be ableto surprise us. Philosophy does so when it(for the first time) makes explicit what is al-ready presupposed by the activities in whichwe engage; or when, to put it another way,it shows the conditions of their possibility.

257:2/o Now Ruben contends that I as-sume in my argument (on p. 31) that sci-entific change consists in changing theoriesabout the unchanging, and do not (and can-not, without circularity) prove it.12 Here it isimportant to distinguish the concept ‘scien-tific change’ which can function in a neutral

way between different philosophical theoriesfrom the premiss, established as a result ofthe analysis into the conditions of the pos-sibility of change (and criticism) in science,‘scientific change (and criticism) is only pos-sible on the condition that there are (rela-tively) unchanging objects, existing and act-ing (relatively) independently of the chang-ing theories of which they are the objects’.From this premiss (only if Q then P) to-gether with the minor premiss, viz that sci-entific change (and criticism) occurs (P), myconclusion (Q) does indeed follow. In thissense it is a touchstone of the validity of anyargument, transcendental or otherwise, thatthe conclusion be ‘implicit’ in the premiss.But the interest of a transcendental argu-ment clearly does not lie in the formal deriva-tion of the conclusion, which is trivial; but inthe production of the knowledge of the ma-

355

jor premiss(i.e. in the analysis). It is in thisthat the essence of conceptual discovery inphilosophy lies. Now such knowledge cannotbe held to be already contained in the an-tecedently existing concept which figures inphilosophical discourse as ‘scientific change’,as is shown precisely by the existence of non-realist explications of it. (It would requirethe drastic expedient of some absolute ide-alist thesis of the identity of opposites, orperhaps a Platonic theory of anamnesis toavoid this conclusion.)

258:1 But what are my grounds for themajor premiss in this case? It is clear howon the assumption of the independent ex-istence of nature the possibility of conflict-ing, differing, changing, clashing and incon-sistent descriptions all become readily intel-ligible. But to leave the matter there wouldbe to overlook ‘just those ‘alternative neo-

idealist interpretations of scientific change. . . according to which there is no neu-tral world “shared” by different theories orparadigms’.13 So, to ‘complete’ (on whichmore anon) the argument, let us considerthem. Such theories posit either ‘incommen-surability’ or ‘Kuhn-loss’. Now it is pre-cisely a condition of the intelligibility of in-commensurability (non-inter-translatabilityof the terms of the rival theories) that thereexists a field of real objects with respectto which the rival theories are incommensu-rable. (As I have remarked elsewhere no-onebothers to say that the rules of cricket andfootball are incommensurable.) In ‘Kuhn-loss’, on the other hand, if it is total, thereare no objects in common, so that Ruben’sobjection is sustained. But now no sensecan be given to the concept ‘scientific change(and criticism)’. For total Kuhn-loss in-


volves neither transformation nor discursiveintelligence, but an archetypal, intuitive un-derstanding constructing its world in a sin-gle synthetic act14-a possibility from whichit is significant that even Feyerabend hasbaulked.15

258:2/o I do not think that any of the ob-jections so far directed at the main theses ofthe book hold. Moreover I think the tran-scendental method employed is, in general,sound. It is, however, certainly the case thatthe book contains no adequate defence, ormeta-philosophical justification, of the lat-ter.16 This is not the place to make good thisomission, which I now regard as the mostserious in the work. A sequel, to be pub-lished shortly, will treat the matter fully. Inthe meantime some brief comments on thequestion ‘how is philosophy possible?’ are in

order.

259:1 On the conception of philosophy atwork in this book both the ultimate pre-misses and the immediate conclusions ofphilosophical considerations are contingentfacts, the former (but not the latter) be-ing necessarily social and so historical. It isonly in this relative or conditional sense thatphilosophy can establish synthetic a prioritruths (truths about the world investigatedby science). Philosophy operates by the useof pure reason. But it does not operate bythe use of pure reason alone. For it exercisesthat reason always on the basis of prior con-ceptualisations of historical practice, of somemore or less determinate social form.

259:2/o Now philosophy as so conceived,can tell us that it is a condition of the possi-bility of scientific activities phi, ψ, etc. thatthe world is structured X and differentiated

357

Y. But it cannot tell us what structures itcontains or the ways in which it is differenti-ated. These are entirely matters for substan-tive scientific investigation. Scientific activ-ity is a contingent, hisorically transient af-fair. And it is contingent that the world isas described by X, Y, Z. But given φ and ψ,X and Y must be the case. A ‘deduction’,or demonstration, of this necessity, (whichmay be termed ‘transcendental’) and whichis represented by the major premiss of a tran-scendental argument when set out in its triv-ial ‘deductive’ form, will normally consist oftwo parts, a straightforward ‘positive’ partin which it is shown how e.g. X makes φ in-telligible; and a ‘negative’ part in which itis shown how absurd, incoherent, counter-intuitive or counterfactual results flow fromthe failure to sustain X, typically expressedin the form of one or more theories that im-

plicitly or explicitly deny it.17 (Thus if lawsare identified with empirical regularities wehave the absurdities that men, in their ex-perimental activity, create or even changethe laws of nature; and that either outsideclosed systems nothing governs phenomena,so that nature becomes radically indetermin-istic, or that as yet science has discovered nolaws!) Misunderstandings about the inten-tions of transcendental arguments typicallystem from the failure to appreciate the crit-ical contexts in which they are developed-against already existing philosophical theo-ries. Thus it is certainly the case that thereis no way of demonstrating the uniquenessof the conclusion of such an argument inadvance of every possible philosophical the-ory.18 But the transcendental considerationis not deployed in a philosophical vacuum: itis designed to replace or situate, an existing


theory; and may come, in time, to suffer asimilar fate. Moreover both the acceptabilityand acceptance of some piece of philosoph-ical reasoning will depend upon the accept-ability and acceptance respectively of the mi-nor premisses concerned. Someone who de-nies that our knowledge is experimentally es-tablished and practically applied and thatscience develops in time need be bound bynone of the results of this book. Further itshould perhaps be stressed that I have notdemonstrated that transcendental realism isthe only possible theory of science consistentwith these activities; only that it is the onlytheory at present known to us that is consis-tent with them.

260:1/o Some implications of this concep-tion of philosophy should be noted. First,by making the possiblity of philosophical dis-course contingent upon the actuality of par-

ticular social practices it provides, at leastschematically, a way of reconciling transcen-dental and historical analyses of human ac-tivities such as science.19 Secondly, philos-ophy cannot anticipate the form of a suc-cessful scientific practice; that is to say theminor premisses of the arguments it usesmay have to be developed afresh in the caseof each specific science. Does this mean,then, that transcendental realism can do nomore than paint its grey on grey? No. Forthough I have developed my arguments inthis book mainly from a consideration of theexperimental sciences of physics and chem-istry, ideologies, derived from defective con-ceptions of those sciences, weigh, like a deadhand, heavily on the shoulders of many ofthe other sciences, and particularly of coursethe proto-sciences of society and man. Phi-losophy can perform at least two tasks here.

359

First, freeing these sciences from the intel-lectual grip of theories secreted by the flatundifferentiated ontology of empirical real-ism, it can set the terms for a more ratio-nal appraisal of the real problems they face.For instance, in the absence of spontaneouslyoccurring, and given the impossibility of ar-tificially creating, closed systems in the hu-man sciences, their criteria for the rationalconfirmation and rejection of theories cannotbe predictive and so must be exclusively ex-planatory.20 But besides this negative func-tion, another more positive possibility opensup for philosophy too, For, it shares an affin-ity with social science in that they both seek,as at least part of their project, to identifyand describe the conceptions of agents en-gaged in social practices. Given this, thepossibility is bound to arise of posing forsocial practices other than science transcen-

dental questions of the form ‘what must bethe case for φ to be possible?’ where ‘φ’ de-notes some characteristic activity as concep-tualised in experience. The conclusion of theargument would be a statement of the con-ditions of possibility of the particular socialactivity concerned. However, in oppositionto a neo-Kantian stream of thought, tran-scendental realism allows that such condi-tions are real and subject to historical trans-formation, so that the resultant hermeneu-tics is contingently critical.21 If this formalanalogy between philosophical and social sci-entific discourse should prove fruitful thenphilosophy could live down Hegel’s jibe (andBradley’s that it merely produces bad rea-sons for what we believe on instinct) andhelp to act not just as the underlabourer,but as the midwife of a science, or group of


sciences.

261:1/o The book aimed to present a sys-tematic account of science. I believe the ac-count to be substantially correct. But it wasan error to imply that the account could alsobe complete. For if philosophy is indeed, asI have been arguing, a conceptual sciencewhich takes as its premisses human activi-ties, situated in time and subject to trans-formation, then there is a sense in whichphilosophy’s work can never be completed.Philosophy is in history too.

262:1 Philosophy is not independent of thevarious sciences, but neither is it reducible tothem. Now this book aimed to be a philos-ophy for the sciences (and against the ide-ologies that threaten them). Throughout itI have stressed the close connection that ex-ists between the transcendental realist on-tology of enduring and transfactually active

structures and a conception of scientific ac-tivity as work; and between the denegationof ontology in empirical realism and the viewof facts, and constant conjunctions, existingquite independently of men. But the per-ceptive reader will have noted an asymetryin the development of the argument. Forwhile the ontology of empirical realism hasbeen explained in terms of a certain concep-tion of man, no explanation of the latter hasbeen given. It is in this area that I think themost pressing problems for the further de-velopment of the Copernican Revolution inthe philosophy of science lie.

5.1. FOOTNOTES FOR THE POSTSCRIPT 361

5.1 Footnotes for thePostscript

1 See e.g. C. Whitbeck, ‘Review of A RealistTheory of Science’ (hence-force R.T.S.), Philo-sophical Review July 1977, p. 115. It is, how-ever, going too far when this reviewer takes me totask for distinguishing a tendency from a powerin two different and incompatible ways withinthe space of two pages (ibid, p. 118n.8) when onthose very pages (pp. 230–1) I explicitly distin-guish two concepts of tendency (tendency1 andtendency2), devoting considerable attention toan elucidation precisely of their differences! Car-oline Whitbeck’s captiousness is surpassed onlyby that of John Krige who, writing in RadicalPhilosophy 12 (Winter 1975), p 39, representsme as attempting to deduce the development ofchemistry (presumably on pp. 168–9) from oneinorganic reaction, when it is clear that the reac-tion cited, that of hydrochloric acid and sodium,is given only as an example of the kind of reac-tion explained by the theory of atomic number

and valency. Krige then goes on polemically toask how I would fit alchemy and phlogiston intomy schema. But my schema is designed to il-lustrate the concept of the real stratification ofthe world. And it is dear that I can only dothis by invoking some or other scientific ontol-ogy. Now it is a fact unfortunate though it mayseem to some romantically-inclined philosophers,that there is at present no scientific ontology thatincludes the three principles (salt, sulphur andmercury), phlogiston and electrons. Each newscientific breakthrough redraws, or situates thepossibility of redrawing, the contours of its ter-rain of reality. So that from each new stand-point in theoretical time the history of the sci-ence (its positive and negative contributions tothe present) looks different, is rewritten, a factthat Bachelard attempted to register with hisconcept of ‘recurrence’ (see e.g. G. Bachelard, LeRationalisme Applique, P.U.F., Paris 1949, p. 2& passim. Cf. also, of course, T. S. Kuhn, whomKrige cites approvingly, The Structure of Scien-tific Revolutions, p. 137 ff.). Now it follows fromthe syncategorematic character of philosophical


discourse, stressed in the book, that I cannotgive a concrete illustration of the philosophicalontology of trancendental realism, without mak-ing use of some or other scientific ontology. Butthat does not imply a philosophical commitmentto it. A history of chemical theories represent-ing the development of thought (a concept in thetransitive dimension) not the stratification of theworld (a concept in the intransitive dimension),would of course include alchemy, phlogiston andmuch else besides.

As I have cited Whitbeck and Krige as cap-tious critics it is perhaps only fair to respond totheir more serious points. Whitbeck finds a diffi-culty in the concept of the non-actual real, con-tending that possibilities and mechanisms ‘mustbe actual in order to explain the behaviour whichmay occur’ (op. cit., p. 117). Now the non-actualreal includes both real possibilities (and latentmechanisms), and their unfulfilled exercise. It isnot the case that a tendency has to be actualisedin an event to help to explain it (see above pp.99–100). But neither is it the case that scienceis concerned only with the explanation of events,

as distinct from possibilites. Moreover events,when they are explained, are typically explainedas instances of the possible. Real possibilitiesmay or may not have an actual basis: fragility,for us, does; gravity, for Newton, did not (cf. p.180 above).

Krige argues that if it is discovered that somenaturally occurring pattern of events repeatedlyreveals the presence of a specific tendency ‘theclaims made by orthodoxy philosophy of sciencegain respectability’ (op. cit., p. 38). Now whileit is clear that some systems, such as biologicalones, are more nearly closed (reveal a greater de-gree of regularity of behaviour, or recurrence ofsyndromes) than others, such as social ones, ifthe law-like statements still have to be analysedas tendencies and not as invariant conjunctionsof events, as orthodox philosophy of science re-quires, then it is not a whit better off. And if, onthe other hand, they do not, then we are deal-ing with a naturally closed system, which tran-scendental realism can of course allow (see p. 91above). But orthodox philosophy of science nowfaces the problem of what governs phenomena,


and would license the application of laws, out-side such systems (see p. 65 above).

Whitbeck also claims that I do ‘not suffi-ciently appreciate how little can be said . . .about the general character of the real world’(op. cit., p. 117), declaring the task I have un-dertaken to be beyond human capabilities (ibid.p. 118). Now, as in 1.4, I argued that every the-ory of science, or epistemology, whether it likesit, admits it or not, presupposes an ontology, theonus is surely on any critic to show how a theoryof knowledge can do without one. It is that en-deavour, though not without effects (see e.g. pp.41–4 above), that I have tried to show is truly be-yond human capabilities. Knowledge cannot beprized apart from its form, and cut loose fromassumptions about what the world must be likefor it to have the form that in some theory ofknowledge it is claimed to possess. From now onthe boot is on the other foot: there is no escapefrom ontological commitment.

Finally, Krige reckons (op. cit. pp. 38–9) thatI fail to describe the mechanisms that wouldmake science ‘a liberating force and not an op-

pressive ideology’. Allez les rouges! Now in thisbook I have attempted to sketch some of theways in which traditional philosophical concep-tions of science can function as ideologies. ButA Realist Theory of Science does not purport tobe a political economy, or sociology of science.Obviously, such work is both legitimate and nec-essary.

2 R. A. Sharpe, ‘Review of R.T.S.’, Philosoph-ical Quarterly July 1976, p. 284.

3 See e.g. D. Davidson, ‘Psychology as Philos-ophy’, Philosophy of Psychology, ed. S. Brown,Macmillan, London 1974, reprinted in The Phi-losophy of Mind, ed. J. Glover, Oxford Univer-sity Press, Oxford 1976 pp. 102–3. (For a cri-tique of the Davidsonian defence of Humean the-ory see pp. 140–1 above).

4 Perhaps it is this which accounts forSharpe’s failure to appreciate why ‘the connec-tion between the underlying mechanism and theobserved phenomena’ cannot, as the Humeansupposes, ‘succumb to an analysis of cause interms of sequence’ (loc. cit). As most readersof the book have fully appreciated, it is precisely


because of the existence in nature of two kindsof systems, viz open and closed ones, and thefact that human activity is in general necessaryfor the latter, that laws, if they are to be uni-versal and operate independently of human ac-tivity, cannot be analysed as sequences at all.(See e.g. the reviews by R. Harre, Mind Octo-ber 1976, pp. 627–30; S. Korner, T.L.S. 11.4.75,p. 397; W. Outhwaite, Social Studies of Science,February 1976, pp. 123–7). Incidentally Sharpequite erroneously attributes acceptance of a ‘likecause like effect’ principle to me (op. cit. p. 285).(For an explicit rejection see p. 77 above.) As Irepeatedly emphasise I am committed to (rela-tive) invariance only at the level of structures,not events (see e.g. p. 219). Still this does leaveopen the fair question of ‘what entitles us to use. . . generality as a criterion of the existence ofa causal (nomic) connection?’ (loc. cit.). Now iftwo objects have the same essential nature theymust behave the same way (see 3.5 above). Butthis condition derives from the identity of theirnatures. Each object must behave as it does be-cause of its essential nature. But it is contingent

that two objects have the same essential nature;just as it is contingent how many other objectsdo (cf. p. 226 above).

On the realist theory of universals developedhere, there is no puzzle about generality. Invo-cation of the latter as a criterion for the exis-tence of normic behaviour derives from the as-sumption of the identity of powerful particularsclassified together as members of a natural kind,which in turn expresses a science’s commitmentto a certain line of enquiry (cf p. 210 above).Sharpe is also worried about how ‘the suppo-sition that there is an underlying structure invirtue of which substances have the features theydo . . . [turns] the relation from one of analyt-icity into physical necessity,’ (loc. cit.). But ifmy transcendental argument for the intransitiveand structured nature of the objects of scientificenquiry (in 1.3) is correct then Sharpe has in-verted the problem of the relationship betweenphysical and logical necessity. For whether ornot a relationship of natural necessity obtains inany given case is quite independent of men, andhence prior to the question of the logical status


of the propositions we use to express it (changesin which can, indeed, be shown to have a certainrationale) (cf. pp. 200–1 above). I cannot give asense to the notion of logical necessity in nature,unless it be taken as referring to that aspect ofnature which consists in the spatio-temporallysituated actions of men.

5 S. Korner, loc. cit.

6 See my forthcoming ‘Philosophies as Ide-ologies of Science’, Essays in Marxist-Philosophy(Provisional title) eds. J. Mepham & D. Ruben,Harvester Press, Hassocks, 1978.

7 D. H. Ruben, Marxism and Materialism,Harvester Press, Hassocks, 1977.

8 ibid, p. 131.

9 See my ‘Feyerabend and Bachelard: TwoPhilosophies of Science’, New Left Review 94(Nov.-Dec. 1975). p. 46.

10 ibid, p. 54.

11 D. H. Ruben, op. cit., p. 101.

12 ibid, p. 100.

13 loc. cit.

14 cf. I. Kant, Critique of Judgement, Hafner,

New York, 1972, pp. 249–58.15 Thus note the overlapping domain marked

D in the second diagram on p. 178 of his AgainstMethod, New Left Books, London, 1975.

(For the record I should add that the par-ticular argument that Ruben cites (on pp. 31–2above) I take only as establishing the intransitiv-ity of the objects of experience. The intransitiv-ity of the objects of knowledge is established bythe analysis of experimental activity, not simplesense-perception. For me, the objects of knowl-edge are disjoint from the objects of experience;and the conditions of possibility of knowledgeand experience are not the same.)

16 A paper published in the same year ‘Formsof Realism’, Philosophica 15 (1), 1975 pp. 99–127merely touches on the subject.

17 cf. W. H. Walsh, Kant’s Criticism of Meta-physics, Edinburgh University Press, Edinburgh,1975, pp. 102–6.

18 S. Korner, Categorical Frameworks, Black-well, Oxford, 1970, p. 72.

19 Cf. e.g. S. B. Barnes, Interests and theGrowth of Knowledge, Routledge and Kegan


Paul, London, 1977, chap. 1.20 See my ‘On the Possibility of Social Scien-

tific Knowledge and the Limits of Naturalism’,Journal for the Theory of Social Behaviour 8 (1)March 1978

21 See my forthcoming The Possibility of Nat-uralism, Harvester Press, Hassocks, 1978.

Bhaskar’s Bibliography

263:1 Nb. This bibliography contains onlybooks and articles referred to in the text andpostscript plus a few others.

263:2 P. Achinstein, Laws and Explana-tions, Clarendon Press, Oxford 1971.

263:3 G. E. M. Anscombe, Causality andDetermination, Cambridge University Press,1971.

263:4 Aristotle, Metaphysics, ed. W. D.

Ross, Oxford 1924.

263:5 D. Armstrong, ‘Dispositional Prop-erties’, Analysis 22 (1961–2).

263:6 — A Materialist Theory of Mind,Routledge and Kegan Paul, London 1969.

263:7 — Belief, Truth and Knowledge,Cambridge University Press, 1973.

263:8 J. R. Aronson, ‘Explanation With-out Laws’, The Journal of Philosophy, Vol.

367

368 BHASKAR’S BIBLIOGRAPHY

66 (1969).

263:9 J. L. Austin, Sense and Sensibilia,Clarendon Press, Oxford 1962.

263:10 A. J. Ayer, Language, Truth andLogic, 2nd ed., Victor Gollancz, London1962.

263:11 — Foundations of EmpiricalKnowledge, Macmillan, London 1963.

263:12 — The Fundamental Questions ofPhilosophy, Weidenfeld and Nicolson, Lon-don 1973.

263:13 M. R. Ayers, The Refutation of De-terminism, Methuen, London 1968.

263:14 G. Bachelard, Le Rationalisme Ap-plique, P.U.F., Paris 1949

263:15 F. Bacon, Novum Organum, ed. T.Fowler, Oxford 1889.

263:16 S. B. Barnes, Interests and theGrowth of Knowledge, Routledge and Kegan

Paul, London 1977.

263:17 J. Bennet, ‘Real’, Mind 1966.

263:18 R. Bhaskar, ‘Forms of Realism’,Philosophica 15(1), 1975.

263:19 — Feyerabend and Bachelard:Two Philosophies of Science’, New Left Re-view 94 (Nov.-Dec. 1975).

263:20 — On the Possibility of Social Sci-entific Knowledge and the Limits of Natu-ralism’, Journal for the Theory of Social Be-haviour 8 (1), March 1978.

263:21 — ‘Philosophies as Ideologies ofScience’, Essays in Marxist Philosophy (Pro-visional title), ed. J. Mepham & D. H.Ruben, Harvester Press, Hassocks 1978.

264:1 — The Possibility of Naturalism,Harvester Press, Hassocks 1978.

264:1 B. Blanshard, Reason and Analysis,Open Court, La Salle, Illinois 1962.

264:1 R. J. Boscovitch, A Theory of Nat-

369

ural Philosophy, M.I.T. Press, Cambridge,Mass. 1966.

264:2 R. Boyle, Collected Works, ed. T.Birch, London 1672.

264:3 R. B. Braithwaite, Scientific Expla-nation, Cambridge University Press, 1953.

264:4 P. W. Bridgman, The Logic of Mod-ern Physics, Macmillan, New York

264:5 M. Brodbeck, ‘Methodological Indi-vidualism: Definition and Reduction’, Read-ings in the Philosophy of the Social Sci-ences, ed. M. Brodbeck, Macmillan, NewYork 1969.

264:6 G. Buchdahl, Metaphysics and thePhilosophy of Science, Basil Blackwell, Ox-ford 1969.

264:7 M. Bunge, Causality, Harvard Uni-versity Press, 1959 (Meridian 1963).

264:8 — The Myth of Simplicity, Prentice

Hall, New Jersey 1963.

264:9 — ‘What are Physical TheoriesAbout?’, A.P.Q. Studies in the Philosophyof Science, ed. N. Rescher, Blackwell, Oxford1969.

264:10 E. A. Burtt, The MetaphysicalFoundations of Modern Physical Science,2nd ed., Routledge and Kegan Paul, London1932.

264:11 K. Campbell, ‘One form of scep-ticism about induction’, Analysis Vol. 23.Reprinted in The Justification of Induction,ed. R. Swinburne, Oxford University Press,1974.

264:12 N. R. Campbell, The Foundationsof Science, Dover 1957 (Physics, The Ele-ments, Cambridge 1919).

264:13 — What is Science?, Dover 1953.

264:14 M. Capek, The PhilosophicalImpact of Contemporary Physics, van


Norstrand, New York 1961.

264:15 R. Carnap, ‘Testability and Mean-ing’, Readings in the Philosophy of Science,eds. H. Feigl and M. Brodbeck, New York1953.

264:16 — The Logical Foundations ofProbability, University of Chicago Press,1950.

264:17 R. Chisholm, ‘The Contrary-to-Fact Conditional’, Mind 1946. Reprinted inReadings in Philosophical Analysis, eds. H.Feigl and G. Maxwell, New York 1949.

264:18 N. Chomsky, ‘Objectivity and Lib-eral Scholarship’, American Power and theNew Mandarins, Penguin, Harmondsworth,1969.

264:19 N Chomsky, ‘Problems of Explana-tion in Linguistics’, Explanation in the Be-havioural Sciences, eds. R. Borger and F.

Cioffi, Cambridge University Press, 1970.

265:1 A. W. Collins, ‘Explanation andCausality’, Mind 1966.

265:2 J. B. Conant, The Overthrow of thePhlogiston Theory, Cambridge, Mass. 1950.

265:3 N. Copernicus, De Revolutionibus,trans. C. G. Wallis, Great Books of the West-ern World, Vol. 16, Encyclopaedia Britan-nica, Chicago 1952.

265:4 I. Copi, ‘Essence and Accident’,Journal of Philosophy 1954. Reprinted inUniversals and Particulars, ed. M. J. Loux,Anchor Books, New York 1970.

265:5 W. Craig, ‘The Replacement ofAuxiliary Expressions’, Philosophical Re-view Vol. 65 (1956).

265:6 M. Crosland, Historical Studies inthe Language of Chemistry, Heinneman,London 1962.

265:7 A. C. Danto, ‘Basic Actions’, A.P.Q.

371

1965. Reprinted in The Philosophy of Ac-tion, ed. A. White, Oxford University Press,1968.

265:8 D. Davidson, ‘Actions, Reasonsand Causes’, Journal of Philosophy 1963.Reprinted in The Philosophy of Action, ed.A. White, Oxford University Press,’1968.

265:9 — ‘Causal Relations’, Journal ofPhilosophy 1967.

265:10 — ‘Psychology as Philosophy’,Philosophy of Psychology, ed. S. Brown,Macmillan, London 1974. Reprinted in ThePhilosophy of Mind, ed. J. Glover, OxfordUniversity Press, l976.

265:11 E. J. Dijksterhuis, The Mechani-sation of the World Picture, trans. C. Dik-shoorn, Clarendon Press, Oxford 1961.

265:12 I. Dirac, ‘Is there an aether?’ Na-ture 168.

265:13 A. Donagan, ‘The Popper-Hempel

Theory Reconsidered’, Philosophical Analy-sis and History, ed. W. H. Dray, New York1966.

265:14 W. H. Dray, Laws and Explanationin History, Oxford University Press, 1957.

265:15 F. Dretske, Seeing and Knowing,Routledge and Kegan Paul, London

265:16 P. Duhem, The Aim and Structureof Physical Theory, trans. P. Weiner, NewYork 1962.

265:17 A. S. Eddington, The Nature of thePhysical World, London 1929.

265:18 A. S. Eve, Rutherford, Cambridge1934.

265:19 A. C. Ewing, The Fundamen-tal Questions of Philosophy, Routledge andKegan Paul, London 1931.

265:20 B. A. Farrel, ‘On the Design of aConscious Device’, Mind 1970.

265:21 P. K. Feyerabend, ‘Explana-


tion, Reduction and Empiricism’, MinnesotaStudies in the Philosophy of Science Vol III,eds. H. Feigl and G. Maxwell, University ofMinnesota Press, Minneapolis

265:22/o — ‘Problems of Empiricism’,Beyond the Edge of Certainty, ed. R. G.Colodny, Prentice Hall, New Jersey, 1965.

266:1 — ‘Problem of Empiricism II’, TheNature and Function of Scientific Theories,ed. R. G. Colodny, University of PittsburghPress, 1970.

266:2 — ‘Against Method’, MinnesotaStudies in the Philosophy of Science Vol IV,eds. M. Radner and S. Winokur, Universityof Minnesota Press, 1970.

266:3 — Against Method, New LeftBooks, London 1975.

266:4 M. Fisk, ‘Are there necessary con-nections in nature?’, Philosophy of Science

Vol. 37 (1969).

266:5 A. Flew, An Introduction to West-ern Philosophy, Thames and Hudson, Lon-don 1971.

266:6 M. Foucault, The Order of Things,Tavistock, London 1970.

266:7 D. Gasking, ‘Causation andRecipes’, Mind 1955.

266:8 P. Gardiner, The Nature of Histor-ical Explanation, Oxford University Press,1952.

266:9 P. T: Geach, ‘Aquinas’, ThreePhilosophers, G. E. M. Anscombe and P. T.Geach, Blackwell, Oxford 1961.

266:10 J. Gibbs and W. Martin, Status,Integration and Suicide, University of Ore-gon Press, Eugene, Oregon 1964.

266:11 J. J. Gibson, The Senses Consid-ered in Perceptual Systems, George Allen

373

and Unwin, London 1966.

266:12 N. Goodman, Fact, Fiction andForecast, 2nd ed., Bobbs-Merill, New York1959.

266:13 — The Structure of Appearance,2nd ed., Bobbs-Merill, New York 1966.

266:14 A. Grunbaum, ‘Temporally Asym-metric Principles, Parity between Explana-tion and Prediction and Mechanism versusTeleology’, Philosophy of Science DelawareSeminar Vol 1, ed. B. Baumrin, New York1963.

266:15 E. Grunberg, ‘The Meaning ofScope and External Boundaries of Eco-nomics’, The Structure of Economic Science,ed. S. R. Krupp, Prentice Hall, New Jersey,1966.

266:16 S. Hampshire, Thought and Ac-tion, Chatto and Windus, London 1959.

266:17 N. R. Hanson, Patterns of Discov-

ery, Cambridge University Press, 1963.

266:18 — ‘A Picture Theory of TheoryMeaning’, The Nature and Function of Sci-entific Theories, ed. R. G. Colodny, Univer-sity of Pittsburgh Press 1970.

266:19 — Observation and Explanation,Allen and Unwin, London 1972.

266:20 G. Harman, ‘Enumerative Induc-tion as Inference to the Best Explanation’,Journal of Philosophy 1968.

266:21 R. Harre, ‘Powers’, B.J.P.S. Vol. 21(1970).

267:1 — Principles of Scientific Thinking,Macmillan, London 1970.

267:2 — Philosophies of Science, OxfordUniversity Press, 1972.

267:3 — ‘Surrogates for Necessity’, Mind1973.

267:4 — ‘Review of A Realist Theory ofScience’, Mind, October 1976. R. Harre and


E. H. Madden, ‘Natural Powers and PowerfulNatures’, Philosophy Vol. 45 (1973).

267:5 — Causal Powers, Blackwell, Ox-ford, 1975.

267:6 R. Harre and P. Secord, The Expla-nation of Social Behaviour, Blackwell, Ox-ford 1972.

267:7 R. F. Harrod, The Life of John May-nard Keynes, Macmillan, London 1951

267:8 H. L. A. Hart and A. M. Honore,Causation in the Law, Oxford UniversityPress 1959.

267:9 G. F. W. Hegel, Philosophy of Right,trans, and ed. T. H. Knox, Oxford 1942.

267:10 W. Heisenberg, Physics and Phi-losophy, Allen and Unwin, London

267:11 C. G Hempel, Aspects of ScientificExplanation, Free Press, New

267:12 M. B. Hesse, Models and Analogiesin Science, University of Notre Dame Press,

Indianapolis 1966.

267:13 — In Defence of Objectivity, Ox-ford University Press, 1973.

267:14 T. Hobbes, ‘On Human Nature’.Reprinted in Body, Man and Citizen, ed. R.S. Peters, Macmillan, New York 1967.

267:15 M. Hollis and E. J. Nell, RationalEconomic Man, Cambridge University Press,1975.

267:16 M. Hollis and E. J. Nell, RationalEconomic Man, C.U.P., 1974.

267:17 D. Hume, A Treatise on HumanNature, ed. L. A. Selby-Biggs, ClarendonPress, Oxford 1967.

267:18 An Enquiry Concerning HumanUnderstanding, ed. A. Selby-Biggs, Claren-don Press, Oxford 1962.

267:19 T. W. Hutchinson, The Signifi-cance and Basic Postulates of Economic The-

375

ory, 2nd ed. New York 1968.

267:20 W. F. Johnson, Logic, Cambridge1921.

267:21 W. D. Joske, Material Objects,Macmillan, London 1967.

267:22 C. B. Joynt and N. Rescher, ‘TheProblem of Uniqueness in History, Historyand Theory Vol. 4.

267:23 I. Kant. On the Distinctivenessof the Principles of Natural Theology andMorals, 1764.

267:24 — Critique of Pure Reason, trans.N. Kemp Smith, Macmillan, London 1970.

267:25 — Critique of Judgement, trans. J.Bernard, Hafner, New York 1972.

268:1 — Metaphysical Foundations ofNatural Science, trans. J. Ellington Bobbs-Merill, New York 1970.

268:2 J. J. Katz, The Problem of Induc-tion and its Solution, University of Chicago

Press, Chicago 1962.

268:3 W. Kneale, Probability and Induc-tion, Clarendon Press, Oxford 1949

268:4 L. Kolakowski, Positivist Philoso-phy, Penguin, Harmondsworth; 1972.

268:5 S. Korner, ‘Substance’, P.A.S.Suppl. Vol. 35 (1964).

268:6 — Experience and Theory, Rout-ledge and Kegan Paul, London

268:7 — Categorical Frameworks, Black-well, Oxford 1970.

268:8 — ‘Review of A Realist The-ory of Science’, Times Literary Supplement11.4.1975.

268:9 A. Koyre, ‘The Influence of Philo-sophical Trends in the Formulation of Sci-entific Theories’, The Validation of Scien-tific Theories, ed. P. G. Frank, Beacon Press,Boston 1959.

268:10 — Metaphysics and Measurement,


Chapman and Hall, London 1968

268:11 J. Krige, ‘Review of A Realist The-ory of Science’, Radical Philosophy 12 (Win-ter 1975).

268:12 T. S. Kuhn, The Structure of Sci-entific Revolutions, 2nd ed, University ofChicago Press, 1970.

268:13 I. Lakatos, ‘Criticism and theMethodology of Scientific Research Pro-grammes’, Criticism and the Growth ofKnowledge, eds. I. Lakatos and A. Musgrave,Cambridge University Press, 1970

268:14 P. S. de Laplace, A PhilosophicalEssay on Probabilities, New York 1951.

268:15 — The System of the World, Lon-don 1809.

268:16 D. Lecourt, Marxism and Episte-mology, New Left Books, London

268:17 C. I. Lewis, Mind and the World

Order, New York 1929

268:18 J. Locke, An Essay Concerning Hu-man Understanding; ed. A. C. Fraser, Dover1959.

268:19 J. Losee, An Historical Introduc-tion to the Philosophy of Science OxfordUniversity Press, 1972

268:20 E. Mach, Popular Scientific Lec-tures, Chicago 1894.

268:21 J. L. Mackie, Truth, Probabilityand Paradox, Clarendon Press,

268:22 E. H. Madden, ‘Hume and theFiery Furnace’, Philosophy of Science

268:23 — ‘Scientific Explanations’, Re-view of Metaphysics 1973

268:24 E. H. Madden and P. Hare, ‘ThePowers That Be’, Dialogue 1971

268:25 M. Mandelbaum, ‘Historical Ex-planation: The Problem of Covering Laws’,

377

History and Theory Vol. I.

269:1 N. Maxwell, ‘Can there be necessaryconnections between successive events?’,B.J.P.S. Vol. 19 (1967). Reprinted in TheJustification of Induction, ed. R. Swinburne,Oxford University Press,

269:2 D. H Mellor, ‘Physics and Furni-ture’, A.P.Q. Studies in the Philosophy ofScience, ed. N. Rescher, Blackwell, Oxford1969.

269:3 A. Michotte, The Perception ofCausality, New York 1963.

269:4 J. S. Mill, A System of Logic, 8thed. London 1961.

269:5 — Auguste Comte and Positivism,2nd ed., London 1866.

269:6 M. J. Mulkay, The Social Process ofInnovation, Macmillan, London

269:7 E. Nagel; The Structure of Science,

Routledge and Kegan Paul, London 1961.

269:8 O. Neurath, ‘Universal Jargon andTerminology’, P.A.S. Vol. 41.

269:9 I. Newton, Principia Mathemat-ica, 2nd ed., University of California Press,Berkeley 1947.

269:10 J. Nicod, Foundations of Geome-try and Induction, trans. P. Weiner, London1934.

269:11 W. Outhwaite, ‘Review of A Real-ist Theory of Science’, Social Studies of Sci-ence, February 1976.

269:12 J. Passmore, ‘Explanations in Sci-ence, History and Everyday Life’, Historyand Theory, Vol. II.

269:13 — Man’s Responsibility to Nature,Duckworth, London 1974.

269:14 W. Pelling, Modern Britain 1885–1955, Nelson, Edinburgh 1960.

269:15 J. Piaget, The Child’s Perception


of Physical Causality, N.J. 1960.

269:16 — Structuralism, Routledge andKegan Paul, London 1971.

269:17 C. S. Pierce, Collected papers Vol.II, eds. Hartshorne and Weiss, Cambridge,Mass. 1932.

269:18 H. Poincare, Science and Hypoth-esis, trans. G. B. Halsted, New York 1903.

269:19 M. Polanyi, Personal Knowledge,Routledge and Kegan Paul, London

269:20 — The Tacit Dimension, Routledgeand Kegan Paul, London 1967.

269:21 — ‘The Structure of Conscious-ness’, The Anatomy of Knowledge, ed. M.Grene, Routledge and Kegan Paul, London1969.

269:22 K. R. Popper, The Poverty of His-toricism, Routledge and Kegan Paul, Lon-don 1957.

269:23 — The Open Society and Its En-

emies, Vol. I, Routledge and Kegan Paul,London 1962.

269:24 — The Logic of Scientific Discov-ery, Hutchinson, London 1959.

269:25 — Conjectures and Refutations,Routledge and Kegan Paul, London

269:26/o — ‘Normal Science and Its Dan-gers’, Criticism and the Growth of Knowl-edge, eds. I. Lakatos and A. Musgrave, Cam-bridge University Press, 1970.

270:1 — Objective Knowledge, ClarendonPress, Oxford, 1972.

270:2 H. Putnam, ‘The Analytic and theSynthetic’, Minnesota Studies in the Philos-ophy of Science Vol. III, eds. H. Feigl and G.Maxwell, Minneapolis 1962.

270:3 R. A. Putnam, ‘Seeing and Observ-ing’, Mind 1969.

270:4 W. V. O. Quine, ‘Designation andExistence’, Readings in Philosophical Anal-

379

ysis, eds. H. Feigl and W. Sellars, New York1949.

270:5 — Methods of Logic, 2nd ed., Lon-don 1962.

270:6 — From a Logical Point of View,Cambridge, Mass. 1953.

270:7 — Ontological Relativity and OtherEssays, New York 1969.

270:8 A. Quinton, The Nature of Things,Routledge and Kegan Paul, London 1973.

270:9 F. P. Ramsey, The Foundations ofMathematics, Routledge and Kegan Paul,London 1931.

270:10 J. R. Ravetz, Scientific Knowledgeand Its Social Problems, Clarendon Press,Oxford 1971.

270:11 V. Ronchi, ‘Complexities, advancesand misconceptions in the development ofthe science of vision: what is being discov-ered?’, Scientific Change, ed. A. Crombie,

New York 1963.

270:12 D. H. Ruben, Marxism and Mate-rialism, Harvester Press, Hassocks 1977.

270:13 R. Rudner, The Philosophy of So-cial Science, Prentice Hall, New Jersey, 1966.

270:14 B. Russell, Mysticism and Logic,Allen and Unwin, London 1962.

270:15 A. Ryan, The Philosophy of JohnStuart Mill, Macmillan, London 1970.

270:16 G. Ryle, The Concept of Mind,Penguin, Harmondsworth, 1963.

270:17 — Dilemmas, Cambridge Univer-sity Press, 1964.

270:18 W. Salmon, ‘Foundations of Scien-tific Inference’, Mind and Cosmos, ed. R. G.Colodny, Pittsburgh 1966.

270:19 K. Schaffner, ‘CorrespondenceRules’, Philosophy of Science Vol. 36

270:20 M. Schlesinger, ‘The Prejudice of


Micro-Reduction’, B.J.P.S. Vol. 12.

270:21 H. Schwyzer, ‘Thought and Real-ity: The Metaphysics of Kant and Wittgen-stein’, Philosophical Quarterly 1973.

270:22 M. Scriven, ‘Truisms as theGrounds for Historical Explanation’, Theo-ries of History, ed . P. Gardiner, Free Press,New York 1959.

270:23 — ‘Explanation, Prediction andLaws’, Minnesota Studies in the Philosophyof Science Vol. III, eds. H. Feigl and G.Maxwell, Minneapolis 1962.

271:1 — ‘Causes, Connections and Condi-tions in History’, Philosophical Analysis andHistory, ed. W. H. Dray, New York 1966.

271:2 W. Sellars, ‘The Language of The-ories’, Current Issues in the Philosophy ofScience, eds. H. Feigl and G. Maxwell, NewYork 1961.

271:3 — Science, Perception and Reality,

Routledge and Kegan Paul, London 1963.

271:4 D. Shapere, ‘Meaning and Scien-tific Change’, Mind and Cosmos, ed. R. G.Colodny, Pittsburgh 1966.

271:5 R. A. Sharpe, ‘Review of A RealistTheory of Science’, Philosphical Quarterly,July 1976.

271:6 B. F. Skinner, Beyond Freedom andDignity, Jonathan Cape, London 1972.

271:7 J. J. C. Smart, Philosophy and Sci-entific Realism, Routledge and Kegan Paul,London 1963.

271:8 L. S. Stebbing, Philosophy and thePhysicists, London 1938.

271:9 P. Strawson, An Introduction toLogical Theory, Methuen, London 1952.

271:10 — Individuals, Methuen, London1959.

271:11 R. Taylor, Action and Purpose,

381

Prentice Hall, New Jersey, 1966.

271:12 W. Thomson (Lord Kelvin), Notesof Lectures on Molecular Dynamics, 1884.

271:13 S. Toulmin, The Philosophy of Sci-ence, Hutchinson, London 1953.

271:14 — ‘Crucial Experiments: Priestleyand Lavoisier’, The Journal of the History ofIdeas, Vol. XVIII (1957).

271:15 — Human Understanding Vol. I,Clarendon Press, Oxford 1973.

271:16 J. Tyndall, ‘Scientific Uses of Imag-ination’, Fragments of Science for Unscien-tific People, New York 1872.

271:17 M. Vernon, The Psychology of Per-ception, Penguin, Harmondsworth, 1962.

271:18 W. Wallace, Causality and Scien-tific Explanation Vol I, University of Michi-gan Press, 1972.

271:19 W. H. Walsh, ‘Categories’, Kant-Studien Band 45 (1954). Reprinted in Kant

ed. R. P. Wolff.

271:20 W. H. Walsh, Kant’s Criticism ofMetaphysics, Edinburgh University Press,1975.

271:21 M. Weber, The Methodology of theSocial Sciences, Free Press 1949.

271:22 C. R. Weld, History of the RoyalSociety Vol. I, London 1848.

271:23 C. Whitbeck, ‘Review of A Real-ist Theory of Science’, Philosophical Review,July 1977.

271:24 P. Winch, The Idea of a Social Sci-ence, Routledge and Kegan Paul, London1959.

271:25/o L. Wittgenstein, TractatusLogico-Philosophicus, trans. D. Pears andB. Mc.Guiness, Routledge and Kegan Paul,London 1961.

272:1 — Philosophical Investigations,trans. G. E. M. Anscombe, Blackwell, Ox-


ford 1953.272:2 G. H. von Wright, Explanation and

Understanding, Routledge and Kegan Paul,London 1971.

Bibliography

[ABC+98] Margaret Archer, Roy Bhaskar,Andrew Collier, Tony Lawson,and Alan Norrie, editors. Crit-ical Realism: Essential Read-ings. Routledge, London andNew York, 1998.

[Bha75] Roy Bhaskar. A Realist Theoryof Science. Leeds Books Ltd.,84 Woodhouse Lane, Leeds, LS28AB, U.K., 1975.

[Bha78] Roy Bhaskar. A Realist Theory ofScience. Harvester Wheatsheaf,London and New York, secondedition, 1978.

[Bha89] Roy Bhaskar. The Possibility ofNaturalism: A Philosophical Cri-tique of Contemporary HumanSciences. Harvester-Wheatsheaf,London and New York, secondedition, 1989.

383

384 BIBLIOGRAPHY

[Bha97] Roy Bhaskar. A Realist Theory ofScience. Verso, London and NewYork, reprint of the 1977 secondedition, 1997.

[Col94] Andrew Collier. Critical Re-alism: An Introduction to RoyBhaskar’s Philosophy. Verso,London, New York, 1994.

Annotations to Bhaskar’s Realist Theory of Scienceehrbar/rts.pdf · Post-Renaissance Europe,...

Documents

Transcript of Annotations to Bhaskar’s Realist Theory of Scienceehrbar/rts.pdf · Post-Renaissance Europe,...