From owner-bhaskar@jefferson.village.Virginia.EDU Fri Nov 28 17:15:54 1997 Received: (from domo@localhost) by jefferson.village.Virginia.EDU (8.8.5/8.6.6) id RAA17922 for bhaskar-outgoing; Fri, 28 Nov 1997 17:14:59 -0500 X-Authentication-Warning: jefferson.village.Virginia.EDU: domo set sender to owner-bhaskar@localhost using -f Received: from ( []) by jefferson.village.Virginia.EDU (8.8.5/8.6.6) with SMTP id RAA50427 for <>; Fri, 28 Nov 1997 17:14:49 -0500 Received: from by (5.x/SMI-SVR4) id AA10214; Fri, 28 Nov 1997 15:14:27 -0700 Received: by (SMI-8.6/SMI-SVR4) id PAA20854; Fri, 28 Nov 1997 15:04:29 -0700 Date: Fri, 28 Nov 1997 15:04:29 -0700 Message-Id: <> From: Hans Ehrbar <> To: bhaskar@jefferson.village.Virginia.EDU Subject: BHA: rts2-33 Sender: owner-bhaskar@jefferson.village.Virginia.EDU Precedence: bulk Reply-To: bhaskar@jefferson.village.Virginia.EDU The Logic of Scientific Discovery 163 3. NATURAL NECESSITY AND NATURAL KINDS: THE STRATIFICATION OF NATURE AND THE STRATIFICATION OF SCIENCE In the process of the establishment of a law of nature three questions may be asked: (i) is there an empirical regularity which constitutes a prima facie candidate for a law ? (ii) is there some reason, other than the regularity, why the 164 A Realist Theory of Science predicates instantiated in the law-like statement should be conjoined ? (iii) is this reason located in the enduring powers of things and the transfactually active mechanisms of nature? If the answer to (i) is yes we have what might be called a `protolaw'.38 If the answer to (ii) is yes we have strong grounds for a law. If the answer to (iii) is yes we have a law. Typically of course the reason in question in (ii) will be provided by a model of the connection between antecedent and consequent, putative cause and putative effect. The transition from (ii) to (iii) typically occurs when a realist interpretation of the mechanism posited in the model becomes acceptable. The answers to (i) - (iii) correspond of course to three levels of criteria for law, viz. those specified by the classical empiricist, transcendental idealist and transcendental realist philosophies of science. At the Humean level laws just are empirical regularities. At the Kantian level both (i) and (ii) must be satisfied. Here we have what might be called the dual criterion theory of law.39 I have already noted its vulnerability to Humean counterattack. At the level of transcendental realism, a distinction is drawn between the empirical identifiability and the universal (transfactual) applicability of laws; and the latter is seen to be a condition of the possibility of the former. As the application of laws in open systems is justified, and presupposed by the intelligibility of experimental activity, the existence of an empirical regularity or a constant conjunction of events is now not even necessary for the ascription of a law (see Table 3.1 below). I have argued that it is only at this level that a distinction Table 3.1 Status of Constant Conjunction of Events Necessary Sufficient for Law classical empiricism Y Y transcendental idealism Y N transcendental realism N N 38 R Harre, op. cit., p. 132. 39 This theory is most clearly stated in R. Harre, op. Cit., Chap. 4. Although Harre is I think logically committed to, and may he prepared to accept, transcendental realism in the form in which it is developed here, he does not say how laws `explain away counter-instances' and so achieve universality, ibid. p. 92. The Logic of Scientific Discovery 165 between necessary and accidental sequences can be sustained. A sequence E_a.E_b is necessary if there is a generative mechanism M such that whenever E_a, E_b tends to be produced; a sequence is accidental if this is not the case. Their difference is represented in Diagram 3.4 below. Most events occur in open systems and must be conceived, as argued in 2.6 above, as `conjunctures'. Diagram 3.4 (not typed in) This is illustrated in Diagram 3.5 below. Necessity as such, like universality, is thus ascribed essentially to the activity of the mechanism; and only derivatively to some particular event sequence. For the result of the activity of the mechanism will Diagram 3.5 (not typed in) 166 A Realist Theory of Science in general be co-determined by the activity of other mechanism too.40 Now these three levels of criteria generate and are generated by different views of science. Thus whereas the classical empiricist will ask merely: (i)* is there a regularity such that whenever C then E? The transcendental idealist will ask in addition: (ii)* given a regularity, is there an explanation such that we can render it intelligible to ourselves that whenever C then E?. The transcendental realist will however, after making an essential correction, go one step further and ask: (iii) out of the plausible explanations for this regularity, is there one which correctly describes the mechanism by means of which, upon the occurrence or obtaining of C, E tends to be produced? That is to say, the transcendental realist will demand that models be tested not just for plausibility but for truth; i.e. for their adequacy in correctly describing the real generative mechanism at work (if the connection between C and E is necessary) such that when C occurs, E tends to be produced (is produced in the absence of interfering causes or the transformation of M). That real things and generative mechanisms must exist can be established by philosophical argument. It is the job of the scientist to discover which ones actually do. Given the identification of some prima facie non-random pattern in nature or protolaw the scientist thus builds up ideas of various plausible hypothetical mechanisms by the creative employment of his imagination (cf. Diagram 3.6) and subjects these ideas to rigorous theoretical criticism and empirical test. These three phases of science are of course those represented in Diagram 3.1 on page 114 above. For the transcendental realist then a model has a relationship with its subject as well as with its source. And it is within the 40 It should be remembered that transcendental realism not only warrants subjunctive and counterfactual statements (where antecedents are uninstantiated) but normic and transfactual ones (where consequents may be unrealised). This is another nail in the coffin of deductivism. For at level (iii) a law may be upheld even when P is true and Q is false; which is of course the only case, according to the principle of material implication, when a conditional is false. The moral is that falsification always depends upon the non-formal requirement that the system in which the putative counter-instance occurs be closed. The Logic of Scientific Discovery 167 nexus formed by this double articulation that new knowledge is produced. For new knowledge is doubly articulated, articulated in two dimensions (transitive and intransitive): it is a socially produced knowledge of a natural (man-independent) thing. It is this bipolarity that a model expresses in standing in two sorts of relationship: a relationship of *analogy* with its source; and a relationship of *adequacy* (when it is) with its subject matter. Diagram 3.6 (not typed in) Many philosophical problems arise from a misunderstanding of the second relationship. It is not a relationship of correspondence; the terms of the relationship are not necessarily like each other, though pictures and iconic models may play an important role in scientific thought.41 Moreover there are no general philosophical criteria for such judgements of adequacy; they are necessarily intrinsic to the particular science concerned. Analogy is one of the possible relationships that models may have with respect to their source. The existence of the first type of relationship (in the transitive dimension) is important in establishing both a constraint on the number of possible explanations42 and an indispensable means of their production. 41 See e.g. N. R Hanson, `A Picture Theory of Theory Meaning', The Nature and Function of Scientific Theories, ed. R. G. Colodny, pp. 233-73. 42 I have argued in paragraph 2 above that without such a constraint on the content of possible explanations, sorting them with respect to their plausibility, there will be an infinite number of possible explanations, even of equal simplicity. The plausibility of a possible explanation cannot be identified by purely syntactical or formal criteria alone but depends upon a complex relationship between what is so far known about the process generating the behaviour in question and established explanation patterns drawn from analogous fields. It is thus in part a function of the existing knowledge in which the predicates occurring in the possible explanations are already embedded, so that the paradoxes of confirmation, etc. that flow from the insertion of artificial predicates into already-functioning and well-connected scientific contexts cannot (at least at that point of application) arise. Cf. R. Harre, `Surrogates for Necessity', Mind 1973, pp. 355-80. 168 A Realist Theory of Science It cannot be described at all adequately as one of coherence; and here again no general philosophical criteria can be laid down for it. Science is work that requires creative intelligence, and there can be no mechanical surrogate for that. The idea of an automatic science is a will-o'-the-wisp that the philosophy of science has pursued, with damaging consequences, since Bacon's search for a `sure and certain method' that would eliminate the need for human thought, which of course inevitably entails the possibility of human error. Most science proceeds by way of a two-tiered method designed to identify invariances in nature, normally under conditions which are experimentally produced and controlled, and to explain them by reference to enduring mechanisms.43 It is in the movement from the identification of an invariance to the mechanisms and structures that account for it that the logic of scientific discovery must be found. Thus the observable reactions of chemistry, which are represented in textbooks by formula such as 2Na + 2HCl = 2NaCl + H_2, are explained by reference to the atomic hypothesis and the theory of valency and chemical bonding. The patterns which constitute the explananda of the theory of valency are needless to say by no means superficially obvious or readily available. Both the concepts and the substances and conditions had and have to be worked for, produced in the social activity of science. The theory itself sets out to describe the causal mechanisms responsible for the overt behaviour of the substances. Once its reality has been established (which justifies our assuming that chemical bonding occurs and the laws of chemistry hold outside the laboratory) and the consequences of the theory have been fully explored, the next task consists in the discovery of the mechanisms responsible for chemical bonding and valency. This has been explained in terms of the electronic theory of atomic structure. Once the reality of this explanation has been established, science moves on to the discovery of the mechanisms responsible for what happens 43 Ibid. p. 366. The Logic of Scientific Discovery 169 in the sub-atomic microcosm of electrons, protons, and neutrons; and we now have various theories of sub-atomic structure. The historical development of chemistry may thus be represented by the following schema: Stratum I 2Na + 2HCl = 2NaCl + H_2 explained by Stratum II theory of atomic number Mechanism 1 and valency explained by Stratum III theory of electrons and Mechanism 2 atomic structure explained by Stratum IV [competing theories of sub- [Mechanism 3] atomic structure] It should be noted that the historical order of the development of our knowledge of strata is opposite to the causal order of their dependence in being. No end to this process of the successive discovery and description of new and ever deeper, and explanatorily more basic, strata can be envisaged. Other sciences reveal a similar open-ended stratification. Geometrical optics is explained in terms of Young and Fresnel's wave optics; which is explained in terms of the electromagnetic theory of light; which can be explained in terms of the quantum theory of radiation.44 A general pattern of scientific activity emerges from this. When a stratum of reality has been adequately described the next step consists in the discovery of the mechanisms responsible for behaviour at that level. The key move in this involves the postulation of hypothetical entities and mechanisms, whose reality can then be ascertained. Such entities need not be smaller in size,45 though in physics and chemistry this has normally proved to be the case. The species of explanation here identified itself falls under a wider genus: in which the behaviour of individuals is explained by reference to their natures and the 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. 170 A Realist Theory of Science conditions under which they act and are acted upon (see 2.3 above). Now for the transcendental realist the stratification this form of explanation imposes upon our knowledge reflects a real stratification in the world. Without the concept of real strata apart from our knowledge of strata we could not make sense of what the scientist, striving to move from knowledge of one stratum to knowledge of the next, is trying to do: viz. to discover the reasons why the individuals which he has identified (at a particular level of reality) and whose behaviour he has described tend to behave the way they do. Without this concept the stratification of science must appear as a kind of historical accident, lacking any internal rationale in the practice of science (if indeed it is not denied altogether in a reductionist and ultimately phenomenalist account of science). As it is clear that the hypothetical entities and generative mechanisms imagined for the purposes of theory-construction must initially derive at least part of their meaning from some other source (if they are to be capable of functioning as possible explanations at all) theories *must* be already understood before correspondence rules for them are laid down.46 Equally this means that the descriptive terms must have initially possessed a meaning independent of them. This enables us to see how meaning-change is possible, and indeed if the independence of predicates is denied, inevitable in the transitive process of science. Similarly we can see how knowledge of newly discovered strata may correct knowledge of less fundamental strata, as concepts and measurement techniques are refined. Now if changing knowledge of strata is to be possible the strata must not change with our knowledge of them. Thus the concept of real strata apart from our knowledge of them is necessary if both the ideas of scientific structure and scientific change, which are central to recent critical philosophy of science, are to be intelligibly sustained. More generally, acknowledgement of the real stratification of the world allows us to reconcile scientific discovery (of new strata) with scientific change (of knowledge of strata). Now the stratification of the world must be assumed by the 46 Cf. K. Schaffner, `Correspondence Rules', Philosophy of Science, Vol. 36 (1969), pp. 280-90. The Logic of Scientific Discovery 171 scientists working in any field, to be in principle unbounded. For it will always be possible for him that there are reasons, located at a deeper level, for the phenomena he has hitherto identified and described. But his knowledge may be in practice bounded by semi-permanent technical or conceptual problems or by the domain assumptions of his particular science; or by the fact that reality is itself bounded at the level knowledge of which he has attained. However, if the stratification of the world has an end, i.e. if there are `entities' which are truly ultimate - and I can see no reason for supposing this must be so - and the scientist has achieved knowledge at that level, he can never know that the level is ultimate. For it will still remain possible for him that there are reasons, located at a still deeper level, for the causes of the phenomena he has succeeded in identifying and describing. I will return to this point below. Now the only kind of necessity that holds between events is connection by a generative mechanism. But there are two other concepts of necessity applicable to the objective world order: there is the necessity implicit in the concept of a law, i.e. in the activity of a generative mechanism as such or the exercise of a thing's tendencies irrespective of their realization; and the necessity implicit in the concept of a thing's real essence, i.e. those properties or powers, which are most basic in an explanatory sense, without which it would not be the kind of thing it is, i.e which constitute its identity or fix it in its kind. The first concept of `natural necessity' is clearly derivative from the second, dependent upon the contingent feature of the system in which the thing's behaviour occurs, viz. that it be closed (see 2.4 above). I am therefore going to refer to the second as the concept of natural necessity, and the third as the concept of natural kinds. Knowledge of natural necessity is expressed in statements of causal laws; knowledge of natural kinds in real definitions. But natural kinds exist and naturally necessary behaviour occurs independently of our definitions and statements of causal laws. Now in the transition from knowledge of any one stratum to knowledge of the next, knowledge of three levels of the objective world order is progressively obtained: of relations between events, of causal laws and of natural kinds. I am going to refer to these three levels as the Humean, Lockean and Leibnizian 172 A Realist Theory of Science levels respectively. The transcendental idealist, as well as the classical empiricist, is, in virtue of his ontological commitment, restricted to the first level of knowledge of the objective world order. I shall argue that even at the Leibnizian level science remains empirical, so that the transcendental argument of Chapter 1 remains valid; and that even at that level the deductively justified prediction of events is impossible, so that the critique of philosophy of science contained in Chapter 2 continues to apply with undiminished force. Moreover I shall argue that the concepts, such as that of natural powers, that we need to render intelligible the transition to the Leibnizian level remain categorically valid even at that level. At the *Humean* level a pattern is identified or an invariance is produced. (This, we know, empirical realism cannot sustain.) We thus have a protolaw (at let us say Stratum I). This is to be explained by reference to the circumstances and nature of the thing whose behaviour is described. The scientist never doubts for a moment that something is generating the effect in question. His problem is: what is? That is, why does x behave the way it does, viz. B, in conditions C_1...C_n? The first step in the scientific explanation of B is to ascribe a power (or liability) of x to B, i.e. to do (or suffer) phi. This is to say, very roughly, that x does phi in virtue of its nature N.47 The next step is thus to investigate N (defining Stratum II). This involves inter alia creative model-building and rigorous empirical-testing (cf. Diagram 3.1). As a result of this investigation we may say x comes to do phi in virtue of its having a certain constitution or intrinsic structure, e.g. genetic constitution atomic structure or electric charge. Now it is contingent that x has the nature (e.g. constitution or structure) that it has. But given that it has, It is necessary that it behaves the way it does. One criterion of this is our capacity to deduce the tendency to 47 Given the B is law-like and allowing for open systems we must say: x tends to do phi in virtue of its nature N. A discussion of the rather complex relationship between tendencies and powers must be postponed to the appendix to this chapter. For the moment they may be regarded as a class of powers whose exercise is normically qualified. But this is not a complete analysis. For a power may be exercised when the behaviour is not law-like, so that it would be wrong to attribute a tendency. The logic of power ascriptions, their role in science and the ontological status of powers will be discussed below. The Logic of Scientific Discovery 173 B from N. This is the *Lockean* level of knowledge. Note that at this level it is still contingent that the thing has the structure that it has. Now at the third *Leibnizian* level possession of that structure or constitution comes to be regarded as defining the kind of thing that x is. Now it is necessary that x has the structure it has if it is to be the kind of thing it is. It is no longer contingent that hydrogen is a gas with a particular atomic structure; rather anything possessing that structure is hydrogen. That is, the criterion for the application of the concept `hydrogen' ceases to be the lightest gas and become instead possession of that structure. At this level the only contingent questions are whether and where things of a given kind exist.48 But note contingency still lies in the flux of the circumstances in which things act, so that events are still not deductively predictable. That is, the `contingency' of events deriving from open systems applies even at the Leibnizian level, so that laws must still be formulated as tendencies (whatever their scope of application). Moreover it is important to see that knowledge at the Leibnizian level is, or may be, attained empirically. We may discover, quite empirically, that the most important explanatory property or real essence of hydrogen, identified as the lightest gas, is its atomic structure; and then attempt to express this discovery in a real definition of hydrogen. Once more the importance of viewing science as a process in motion is clear. For if we stay at any one level, phase or moment of science the idea that a definition may be arrived at empirically will appear absurd. If it is accepted, however, the reason why the laws of nature cannot be deduced a priori from self-evident axioms becomes clear. For the axiomatic base of a science at any moment of time, at any stratum of reality, is something that has had to be worked for, produced, as part of the irreducibly empirical process of science. Scientists attempt to discover what kinds of things there are, as well as how the things there are behave; to capture the real 48 The second question is both distinct from the first and important. Because it raises the question of the range or scope of application of the statements expressing the tendencies of the individuals concerned. It cannot be assumed that all tendencies will be spatio-temporally universal; for individuals and kinds may be transformed in time and bounded in space. A law may of course be universal (transfactually applicable) within its range and restricted in this way. 174 A Realist Theory of Science essences of things in real definitions and to describe the ways they act in statements of causal laws. The real essences of things are their intrinsic structures, atomic constitutions and so on which constitute the real basis of their natural tendencies and causal powers. Thus there is no conflict between explanatory and taxonomic knowledge. Rather, at the limit, they meet in the notion of the real essences of the natural kinds, whose tendencies are described in statements of causal laws. At the Leibnizian level statements of law are substitution instances of necessary truths about the individuals to which the refer. For any individual which did not behave in that way would not be an individual of that kind. They may thus be regarded as analytic truths. But they are arrived at in the transitive process of science a posteriori, by empirical means. Thus a fully dynamic philosophy of science must take seriously the question `how is analytic knowledge arrived at a posteriori possible ?' To this question I will return in paragraph 5 below The situation at the Humean level is rather like that faced by the citizens of Koenigsberg who knew, from experience, that there was no way of crossing each of the town's seven bridges just once.49 See Figure 3.1 below. At the Lockean level this fact is deducible from the topology of Koenigsberg, given Euler's theorem. At the Leibnizian level, there is a necessary truth about a certain physical set-up, whether or not there is a town called `Koenigsberg' or any town at all to which it applies. At the Leibnizian level, Mendeleyeev was able to deduce from his Periodic Table, interpreted as dealing with atomic number and valency, the properties of several new elements. But it remained contingent whether, and if so where, there were elements in the world to which his predictions applied. Certain chess games have Figure 3.1. The Seven Bridges of Koenigsberg (not typed in) 49 Cf. M. Hollis and E. J. Nell, Rational Economic Man, Chap. 7. The Logic of Scientific Discovery 175 only one possible solution. But it remains contingent whether they are ever played. The concept of powers has played a key role in our analysis of science's transition from knowledge of one stratum to knowledge of the next. To ascribe a power is to say that a thing will do (or suffer) something, under the appropriate conditions, in virtue of its nature. This is not, as is so often claimed, a pseudo-explanation50 or a purely verbal formula.51 Rather it is an indication of work to be done. Moliere's doctors in Le Malade Imaginaire have often been ridiculed for speaking of opium as possessing a `dormitive virtue'. But in doing so they left open the possibility of an investigation, at some future date, into the nature of opium without committing themselves to what would doubtless have been, for them, a rash conjecture at the time. Moreover it is far preferable to the Humean alternative, viz. that whenever men smoke opium they fall asleep. For in the first place, the latter is untrue. Secondly, it is less informative. It might be a complete accident that everyone in the den is asleep: the powers formula rules this out. It says that there is something about opium in virtue of which when men smoke it they tend to fall asleep. The connection is necessary. But it is only a tendency. Thirdly, the Humean formula is regulatively useless. The powers statement is by contrast quite suggestive. For it indicates the need for an investigation into the chemical properties of opium and the way they induce sleep in men.52 In context, it constitutes an open admission of ignorance. The Humean, on the other hand, must pretend that once he has his generalization there is nothing more to be known. And if he should stumble upon a higher-order generalization this can be, for him, only accidentally related to the original one: there is no inner logic connecting the two, or rationale by which science moves from the first to the second. To this it might be objected that the concept of powers does not figure in the discourse of science. This is true. And the reason for it is of course that the scientist, unlike Moliere's doctors, is never just content to ascribe a power but moves 50 See e.g. A. Flew, An Introduction to Western Philosophy, p. 49; or E. Nagel, op. cit., p. 37. 51 See e.g. L. Kolakowski, Positivist Philosophy, p. 34. 52 Cf. R. Harre, Principles of Scientific Thinking, pp. 274-5. 176 A Realist Theory of Science immediately to the construction of possible explanations for it with the paradigms and other instruments of thought at his disposal. That is his job. (Sometimes, however, when we are completely at a loss we do just ascribe a power.) The concept of powers is introduced precisely to describe this normally instantaneous (or simultaneous) and unselfconscious response of the scientist to the identification of protolaws; it represents, if you like, an attempt to reconstruct the internal rationality of the inter-strata move. The concept of powers is not intended to figure in the discourse of science, but in the discourse of the philosophy of science (which is the former's rational reconstruction). It should perhaps be stressed here that the stages of my rational reconstruction of the process of scientific discovery represent phases of scientific activity; they cannot be identified with moments of chronological time. Thus most scientific work must occur, for reasons I will bring out in paragraph 4, in the context of a research programme designed to show that on the supposition of the mechanism M the field of phenomena can be rendered intelligible. Thus the identification of a protolaw normally depends upon the prior existence of a conjecture or a hypothesis of a mechanism intended to function as a possible explanation for the presumed protolaw.53 To ascribe a power is to say that there is something about the thing, which may be unknown, in virtue of which it behaves the way it does. The grounds for the ascription of a power must thus be stronger than the mere occurrence of a regularity. For we must possess some reason to suppose the connection necessary (though in the limiting case this may just be the invariance of an experimentally produced result). It is because it indicates the power-ascriber's belief in the existence of a reason, located at the next highest level of inquiry (in the nature of the thing), whether 53 Cf. K. R. Popper, Conjectures and Refutations, Chap. 5. However the protolaw itself when it finally emerges, pari passu with its explanation after the limitations and modifications necessitated by the experimental process, may be in a form far more complex and refined than that in which it was originally conceived (cf. S. Koerner, Experience and Theory, passim). The normal response to a (genuine) counter-instance is modification within a continuous research programme, rather than (as is implied by naive falsificationism) the complete abandonment of the original conjecture and its replacement by a totally different one (cf. I. Lakatos, op. cit.). The Logic of Scientific Discovery 177 or not the reason is currently known, that the concept of powers, in pinpointing an essential moment in the transition from knowledge of one stratum to knowledge of the next, plays such a key developmental or strata-bridging role. In this way, a powers statement is a promissory note cashed in the development of science, a schematic explanation filled out in the growth of our knowledge.54 It is worth noting that the structure of a powers ascription is well adjusted to accommodate both falsification (obviously, as the hypothesized reasons may be subjected to independent tests) and meaning-change (less obviously). If meaning change is to be possible, some elements of meaning must remain constant through the change. Now if `x does B' is analysed as `x is of such a nature N [defining Stratum II] that it will do phi in conditions C_1 . . . C_n [defining Stratum I]' we may allow that the meaning of `phi' remains constant while the meaning of the N component changes between theories (and vice-versa). This applies even in the case of simple descriptive observational predicates such as `blue'. For `x is blue' may be analysed as `x looks blue [defining Stratum I] in virtue of its reflecting light of a certain wavelength [Stratum II]'. The simple theory that things look blue because they are blue may then be replaced by the scientific theory that they tend to look blue in normal circumstances because they reflect light of wavelength 4400A. Subsequently we may allow the latter to define the scientific use of `blue'; in which case of course it is no longer contingent that blue surfaces reflect light of that wavelength. Now although the concept of powers serves this essential developmental function, it cannot be reduced to it. For when we have climbed up to Stratum II, we cannot throw away the ladder, so to speak. To pursue the analogy, the ladder is a rope, not a wooden one. For to make a powers statement is to make a categorical statement about the nature of the thing situated at the level to which we have climbed. It is to make a statement about possibilities which are possessed by the thing quite categorically, whether they are known (or actualized) or not. Dogs do not lose their power to bark when we understand how they do so, just as glass does not cease to be brittle when we know its molecular structure. 54 Cf. R Harre, op. cit., p. 275. 178 A Realist Theory of Science The ontological bases of powers just are the properties that account for them; i.e. the natures in virtue of which they are ascribed. Now in the transitive process of science such natures may come to be qualitatively described. When this happens it will of course initiate a search for the higher-order entities and mechanisms that account for them. But how does it happen ? In general, at any one level, individuals must be identified and their normic behaviour described. Now for a qualitative description of a thing or a dispositional account of its behaviour it must be present to the scientist's senses and he must be able to describe it correctly, i.e. as being of the kind or type that it is. This will normally depend upon two kinds of work: practical (experimental and technical) work, in which the scientist's causal and perceptual powers are augmented (the latter with the aid of the construction of sense-extending equipment, such as microscopes); and theoretical work, in which the scientist's conceptual and descriptive powers are augmented. It is the aim of the former to produce the object, i.e. to render the thing or behaviour directly accessible to the scientist's senses (so that it becomes the possible object of an act of immediate demonstrative reference). And it is the aim of the latter to produce the concept of the object, so that the scientist is capable of an adequate description of it. Both are necessary for a qualitative description. It should be noted that the two kinds of criteria, viz. demonstrative and recognitive, are distinct. For my incapacity to identify the chromosome structure by peering down an electron-microscope does not mean that it is not a possible object of an act of demonstrative reference. It is present to my senses, whether I recognize it or not. Conversely to render it accessible to my senses is an independent labour (itself only possible if some concept of it is possessed), requiring great ingenuity, just as experimental production and control does, when we are concerned with the description of the law-like behaviour of some thing already identified. The production of the object and the production of its concept are thus independent tasks, each essential to a qualitative description of a thing or account of its behaviour. The thing must be there and I must know what kind of thing it is, i.e. how to describe it; in general this will involve a theoretical redescription of it. Now it is important to realize that though the production of The Logic of Scientific Discovery 179 the object and the production of its concept are distinct, the judgement that the object has been produced itself depends upon a tacit theory of vision and the instruments according to which its range is extended. The case of the electron microscope illustrates this very well. In general it is the function of such background or auxiliary theories to specify the conditions under which an object of the appropriate type may be said to be present to the senses. In this sense they constitute, as it were, the criteriology of empirical science. It is clearly essential to the theory of scientific development proposed here that imagined entities may come to be established as real. Now an entity may be `theoretical' either in the sense that its existence is open to doubt (theoretical_1) or in the sense that it cannot be directly perceived, either unaided or with the help of sense-extending equipment (theoretical_2). The same distinction applies in the case of behaviour. Now an entity (or mode of behaviour) may be theoretical_1 at t_1 and perceived and adequately described at t_2, so that it then ipso facto ceases to be theoretical_1. The existence of bacteria, initially conceived as minute hostile micro-organisms, and molecules, initially modelled on material objects, came to be established in this way. This is typical of science and shows once more the importance of viewing it as a process in motion. But if an entity cannot be perceived, i.e. is theoretical_2, does this mean that it cannot be known to exist, so that it must be theoretical_1? If this were the case all theoretical_2 entities would indeed be hypothetical, and our knowledge would be necessarily confied to the domain of observable things, even if this were now regarded as an expanding class. Fortunately this conclusion does not follow. For theoretical_2 entities may be known to exist indirectly, viz. through the perception of their effects. The paradigm here is the case of the detection of radio-active materials by a geiger counter, of electricity by an electroscope, of a magnetic field by a compass needle. That there is a difference between the cases of detection and perception is clear. In the case of detection the thing can be individuated only indirectly, i.e. via the spatio-temporal framework or through its effects on particular things; it cannot be the object of an act of immediate demonstrative reference. Whatever the mental imagery we use to think of a magnetic field it can be present to us only through 180 A Realist Theory of Science its effects. On the other hand my incapacity to identify a bacterium under a microscope as being of a particular type, or even as being a bacterium at all, does not mean that it is not present to my senses; and so capable of functioning as the object of a possible act of immediate demonstrative reference, although ex hypothesi I am incapable of intentionally performing it. It should be stressed that in the detection case that something does exist producing the effect is not in question. Nor is the fact that it exists and acts independently of its detection. To say `electricity is what electricity does'55 is to collapse powers to their exercise. Electricity is not what electricity does; but what it can do. The mode of reasoning employed in inferring the existence of causal agents through the ostension of their effects is thus perfectly proper. Hence though it is correct to say that when we cannot qualitatively describe the cause we know less about it than when we can (given that in the latter case we know the thing's causal powers as well) it is not true to say that there is a cause is less certain. It is just that in the detection case what we can know about a thing is limited to its causal powers. Now there are two possibilities here. One is that there is a nature, susceptible in principle to a qualitative description, as yet unknown, which is the bearer of its causal powers. The other is that the nature of the thing just is its causal powers, as in the case of physical field theories. At any moment of time a science may have to put down its ultimate entities just as powers to produce effects, e.g. to affect observers and equipment, possible observers and possible equipment, material things, in certain ways. About such entities all the scientist knows is their powers. It always remains possible that he will be able to achieve a qualitative description of them, and he must strive to do so. On the other hand, it is also possible that such entities are their powers. The scientist can never dogmatically eliminate one of these alternatives in advance. If there is a frontier to possible knowledge of the world the scientist can never know when he has reached it. But whatever is responsible for the world as manifest must possess causal powers which are continually being exercised; it must be co-extensive with space and continuous with time. It must be structured and complex; it cannot be atomistic or event-like. The concept of a field of potential seems 55 See A. J. Ayer, The Fundamental Questions of Philosophy. The Logic of Scientific Discovery 181 closest to meeting these requirements.56 However it seems to me there is no reason in principle why there should not be strata of fields (of perhaps radically different kinds), forever unknown to us. It should be noted that only the identification, not the existence of fields depends upon the existence of material things in general. Here again the order of dependence in being is opposite to the order of dependence of our knowledge of being. The ontological order is distinct from the epistemic one. The general thrust of my argument in Chapter 2 was against reductionism. How does this square with my emphasis on strata of knowledge? It will be remembered that I did not deny the possibility of an explanatory reduction but stressed (a) the need for a well-defined reductans (so that a reduction could not in general be a means 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 as they were causal agents capable of acting back on the materials out of which they are formed (see 2.5 above). It is clear that I was there taking possession of causal powers, and hence existence in time, as the most general criterion of reality. There is an asymmetry between space and time here. For powers must be possessed and exercised in time, but they need not be localized at any point in space. Relations, for example, such as that of spin (in physics) and marriage endure through time and have causal effects. But they have no position in space. Now in general a reduction is possible because the entities in terms of which the behaviour of the thing is explained occupy a different volume of space, either larger or (more usually) smaller. Thus the possibility of a reduction implies in general that the individuals of the different kinds cannot be said to occupy the same place at the same time and one not be part of the other. This gives us a general criterion which imposes limits on regresses of strata, i.e. upon the possibility of a sequence of (explanatory) reductions. For one could define a branch of science as a series of theories within which this criterion is satisfied. On it, quantum mechanics and chemistry would belong to the same branch. But electromagnetism and mechanics, neurophysiology and 56 See R. Harre and E. H. Madden, `Natural Powers and Powerful Natures', Philosophy, Vol. 45 (1973), esp. pp. 223-30. 182 A Realist Theory of Science psychology and (it will be argued) psychology and sociology would belong to different branches. Changes of things are explained in terms of unchanging things. If there are ultimate entities they must be unchanging. Atoms have already been disqualified as possible ultimate entities (see 2.3 above). So ultimate entities must be powers; that is, individuals characterized solely by what they can do. For if one could describe the changing states or conditions in virtue of which their powers were exercised they could not be ultimate (unchanging). In the last instance to be is just to be able to do. But this does not rule out the possibility of a science of cosmology (which would be concerned with the distribution in space and redistribution in time of the ultimate entities) or of irreducibly historical branches of science in which the ultimate entities were Aristotelian or even Strawsonian individuals. The transformation of the principles governing such things would in general have to be conceived as conjuncturally determined open systemic events (see 2.6 above). In this way a complex thing such as a person (or a society) could come to be the cause of its own transformation. Now it is because we are ourselves material things that our criteria for establishing the reality of things turn on the capacity of the thing whose existence is in doubt to bring about (or suffer) changes in its material constitution or the constitution of some material thing. Space, for example, might be regarded quite abstractly just as any system of relations in which objects stand to one another. And we can conceive the possession and exercise of causal powers in time in ways, and at levels, forever unknowable to men. We can never know where we stand absolutely in the chain of being. Despite this cosmic incapacity science has succeeded in identifying strata of reality. Now a scientist never doubts for a moment that there are reasons for the behaviour he has identified and described. It is in the search for such reasons, at a deeper level of reality, at present known to him only through its effects, that the essence of scientific discovery lies. This search necessitates the construction of both new concepts and new tools. But, as what is produced must possess a material cause, the scientist stands for his essential task, in two systems of social relationships, depending necessarily on the work of others. The Logic of Scientific Discovery 183 I have argued that the concept of natural necessity is the concept of a real generative mechanism at work, a concept which is applicable to the world quite independently of men. And it is in virtue of their connection by such a mechanism, of which knowledge may be attained in the social activity of science, that necessity is properly ascribed to some but not other sequences. In paragraph 5 I will analyse and criticize some objections to this concept of natural necessity and the related concept of natural kinds. But I want to deal here with the following basic objection to the account I have proposed: If, as I have contended, at each stratum or level of reality an entity is identified and its behaviour is described what positive advantages does this account have over the traditional empirical realist ones? I think it has at least four substantial advantages. First, it reveals the essential movement of science. Second, it allows room for the location of a surplus-element, reflecting a difference independent of men, in the analysis of law-like statements at any one level. Third, it alone is capable of sustaining the ideas of the necessity and universality of laws, which are necessary for the rationality of theory-construction and the intelligibility of experimental activity. Finally, it alone is capable of accommodating the possibility of the existence of entities and the necessary phase of the knowledge of entities which cannot be analysed as substances with qualities, but must be conceived as powers to produce effects, powers which are possessed and may be exercised quite independently of their detection. (Needless to say, these advantages are not independent of each other.) Science never stops still for a moment. At whatever level we look, it always involves something more than the empirical realist concedes. For example, if we consider the phase of the identification of a protolaw (which seems prima facie most susceptible to empirical realist analysis), we find the categorical clause implicit in a powers ascription, representing the scientist's instantaneous response to this situation, indicating his belief in the existence of a reason, located at the next highest level of inquiry, for the predicates being conjoined. Only the powers conceptual system is capable of giving an account of the internal rationality of science, by which it moves from knowledge of one stratum to knowledge of the next, so displaying the actual 184 A Realist Theory of Science historical development of the sciences as something other than a sequence of accidents. Now it is our knowledge of the reasons at Stratum II for the behaviour at Stratum I that warrants our designating the behaviour as necessary. But the reasons for the behaviour at Stratum II cannot be collapsed into the behaviour at Stratum I or an interpretation or model of that behaviour consistently with the intelligibility and rationality (respectively) of theory-construction or the possibility of empirical test. Nor can such reasons be glossed simply as more fundamental regularities,57 if they are to be subject to experimental confirmation (or corroboration) . I have already shown in detail that the empirical realist account of laws, and hence the ontology that underpins it, is defective. Laws, I have argued, cannot be interpreted as conjunctions of events, but must be analysed as tendencies of things. If science is to be rendered intelligible the world must be seen as one of persisting things, of differing degrees of structure and complexity, to which powers and tendencies are ascribed; it cannot be reconstructed as a world of atomistic events apprehended in sense-experience. Briefly, to summarize my account of laws: To invoke a law I must have grounds for supposing a generative mechanism at work. These comprise: (a) independent grounds, preferably under experimentally closed conditions, for the mode of operation of the mechanism; (b) grounds for the satisfaction of the antecedent (or stimulus) conditions for the operation of the mechanism on the particular occasion in question; and (c) the absence of specific grounds for supposing a breakdown or transformation of the mechanism in that case. Generative mechanisms, I have argued, must be analysed as the ways of acting of things; and their operations must be understood in terms of the exercise of tendencies and causal powers. Tendencies may be possessed unexercised, exercised unrealized, and realized unperceived (or undetected) by men. Finally, the empirical realist cannot deal with the case of entities which just are their powers or about which all we know are their powers. He thus rules out dogmatically, tout court the possibility of a certain kind of entity and a necessary phase of 57 See e.g. P. Achinstein, Law and Explanation, pp. 13ff. The Logic of Scientific Discovery 185 knowledge. In virtue of this he is no more able to make sense of the frontiers of knowledge, than show the mechanism by which sciences if it can and when it does, will advance. --- from list --- .