NANCY CARTWRIGHT

IS N A T U R A L

SCIENCE

'NATURAL'

ENOUGH?:

A REPLY TO PHILIP ALLPORT

Recall the structure of Hume's Dialogues. The project is natural religion: to establish the properties of the deity - hypothesized to be omniscience, omnipotence, and benevolence - from the phenomena of the natural world. The stumbling block is evil. Demea and Cleanthes try to explain it away, with well-known arguments. Demea, for example, supposes "the present evil of phenomena, therefore, are rectified in other regions, and at some future period of existence" (Part X). Philo replies: I will allow that pain or misery in man is compatible with infinite power and goodness in the D e i t y . . . what are you advanced by all these concessions? A mere possible compatibility is not sufficient. You must prove these pure, unmixed and uncontrollable attributes from the present mixed and confused phenomena, and from these alone. (Part X)

Philo expands his argument: [ I ] f a very limited intelligence whom we sha11 suppose utterly unacquainted with the universe were assured that it were the production of a very good, wise, and powerful being, however finite, he would, from his conjecture, form beforehand a very different notion of it from what we find it to be by e x p e r i e n c e . . , supposing now that this person were brought into the world, still assured that it was the workmanship of such a divine and benevolent being, he might, perhaps, be surprised at the disappointment, but would never retract his former belief if founded on any very solid argument . . . . But suppose, which is the real case with regard to man, that this creature is not antecedently convinced of a supreme intelligence, benevolent and powerful, but is left to gather such a belief from the appearance of things; this entirely alters the case, nor wilt he ever find any reason for such a conclusion. (Part XI)

Philo is the self-avowed mystic. He believes in the wondrous characteristics of the deity, but he takes it to be impossible to defend this belief by reason and evidence. Philo's is not natural religion, but revealed. The project Allport and I discuss is natural science. He hypothesises various attributes for the laws of nature: they are "few in number", "simple", and "all-embracing". They are also supposed to be literally true. In How the Laws of Physics Lie, I noted that the laws which are the best candidates for being literally true, whether very phenomenological or far more abstract, were numerous and diverse, complicated and Synthese 94: 291-301, 1993. © 1993 Kluwer Academic Publishers. Printed in the Netherlands.

292

NANCY

CARTWRIGHT

limited in scope. "This is, however", as Allport points out, "a far cry from establishing that such a realist account is impossible" (Sect. 2). I agree. Complication and limitation in the truest laws we have available are compatible with simplicity and universality in the unknown, ultimate laws. But what is advanced by this concession? If I knew beforehand that nature were governed by a few simple, allembracing laws, I need not retract this belief in the face of the current structure of science. But supposing I have no antecedent knowledge and insist on gathering my beliefs from the world as I experience it. Then I would draw, if any, a quite different conclusion. As Allport reports, I have pointed out the failures and limitations of modern physics in deriving "the incredible richness and diversity of phenomena" from "a few rather abstract mathematical formulae". He argues, "[b]ut only if its failure can be guaranteed a priori does it follow that the laws of physics can be said to lie" (Sect. 1). My reply to Allport is to urge the project of natural science: guarantee nothing a priori, and gather our beliefs about laws, if we must have them at all, from the appearance of things. I have isolated three characteristics that Allport attributes to the fundamental laws of nature. They are: to be few in number, simple, and all-embracing. How the Laws of Physics Lie was concerned primarily with the first two. It offered a challenge to the realist. Look at the way in which the simple laws we have are used to cover the phenomena. Can we reasonably argue from that to the truth of these laws? This is the challenge that Allport takes up, and so this is the question I will focus on. But I want to stress that the assumption that the laws are all embracing is equally hard to justify from the empirical evidence. The problem is at base the venerable problem of how far and across what boundaries we can extend our inductive inferences. I have been writing about this question elsewhere, 1 and will only summarize here the alternative vision I would like to offer. Imagine for the sake of argument that our experimental models for Allport's fundamental fermions and bosons fit exactly into the 'standard model' without any of the kinds of fudging or distortion that I worried about in How the Laws of Physics Lie. As I will explain below, I have no quarrel with induction as a form of inference, so I would be willing to take this as good evidence that the standard model is precisely and exactly true of fundamental bosons and fermions in situations relevantly similar to those of our experiments. Does that make it all embracing?

IS N A T U R A L S C I E N C E ' N A T U R A L '

ENOUGH?

293

That depends on the extent of 'relevantly similar' - and that is a matter for hard scientific investigation, not a priori metaphysics. That is the reason I am so concerned with the successes and failures of basic science in treating large varieties of situations differing as much as possible from our experimental arrangements. We are inclined, following the mechanical philosophy of the seventeenth and eighteenth centuries, to believe in fundamental particles, and to assume: (i) that everything that exists in nature is nothing but a collection of these fundamental particles in complex interaction; (ii) that the fundamental particles behave in the same ways with respect to each other wheresoever they appear; and (iii) that all the properties that objects have are reducible to the properties of collections of fundamental particles studied in physics. Given the first doctrine, it is the third doctine that makes the claim that particles always behave in the same way plausible. For what could distinguish the situations where they behave in one way from those in which they behave in another? The view that there are distinguishing features - many of them, perhaps even ones we refer to in everyday life - gets derided as emergentism. But this is, I think, to mistake how the domain of properties studied by physics gets set. Here is one caricature: we begin with an interest in motions - deflections, trajectories, orbits. Then we look for the smallest set of properties that is closed (or, closed enough) under prediction. That is, we expand the set until we get all the factors that are causally relevant to our starting factors, and then everything causally relevant to those, and so forth. To succeed does not show that we have got all the properties there are. 2 One can of course argue against this possibility by pointing to successful, or nearly successful, cases of property reduction. But again the question of natural versus revealed science arises. The multitude of failures are certainly compatible with the truth of reductionism, but they do not argue for it. The overall programme I want to urge is a careful and detailed philosophical study of the evidence about the boundaries of relevance for the standard model, or for any of our other fundamental taws. We have to allow for the possibility that they are true but not universal; exact but limited in range. I turn now more directly to Allport's discussion. A number of the claims that Allport makes are ones on which he clearly expects us to be in agreement, most notably in Section 2. More surprising perhaps is that I agree with almost all that Altport says, dropping a phrase here

294

NANCY

CARTWRIGHT

and there, usually fight up to the last sentences of each section. The discussion of Allan Franklin's studies of experiments on weak interactions in Section 4 is a good example. Atlport maintains that Franklin "clearly demonstrates that experimental pure science is not always led by the nose by theory" (Sect. 4). This sounds like an endorsement of Ian Hacking's well-known doctrine that experiment "has a life of its own", 3 which I have always taken to provide sympathetic or collateral support for the structure of science pictured in How the Laws of Physics Lie. Allport argues to the contrary, for two related reasons: (i) the experiments described by Franklin led to the dismissal of deeply held theoretical views; and (ii) the adjustment of theory to one set of specific experimental findings "always has many empirical consequences in other sectors of the theory" (Sect. 4). I will discuss the latter first. Allport continues: "If one were content with phenomenological laws in each of these fields, then this unity of explanation would be lost and the consequent predictions, due to the fundamental connection of these different phenomena, would be missed" (Sect. 4). I agree. He concludes: "With such experiments the covering laws can indeed be evaluated empirically". 4 Here I disagree (at least if evaluated means 'judged true or false'.) The point of these theory-sceptical works is not that they deny that covering laws provide very, very precise - and novel - predictions in at least some realms of nature. Rather, the point is to query how that bears on the truth of these covering laws. Franklin's studies give a nice account of the structure of reasoning that led to the rejection of the hypothesized charge/parity symmetries. But I do not see how to turn Franklin's account into an answer to the Duhemian challenge to show why these detailed experimental results, and the back and forth "interplay between theory and experiment" (see Allport, Sect. 4), are a reasonable basis for judging the truth or falsehood of the theory rather than its power to organize the phenomena, to summarize and turn up buried analogies, and to produce precise predictions. It seems to me that Allport's considerations often do not discriminate finely enough between this Duhem-rooted view and what is sometimes called the 'Baconian view', which is so attached to direct inference from observation that it finds no role for theory at all. Perhaps it will help to pursue the similarities of How the Laws of Physics Lie with van Fraassen's The Scientific Image in order to highlight the differences with Allport. Van Fraassen admits both highly precise

IS N A T U R A L S C I E N C E ' N A T U R A L '

ENOUGH?

295

scientific knowledge (about observable phenomena) and the crucial role of theoretical facts. For him there may well be no way to provide a sensible picture of the world that contains just the observable phenomena in all their precise details without also containing theoretical entities behaving in ways describable by what we would happily call theoretical laws. Still, he maintains, we are entitled to believe only in the observable facts and not in the theoretical. This is because he supposes that the observable is epistemicatly privileged; its accessibility is taken for granted. Then he asks: What mode of inference can transfer this epistemic accessibility from what is observable to what is not? An 'Aufbau' would do, but we no longer have hope that the theoretical can be constructed from the observable. What else then? The hypotheticodeductive method commits the fallacy of affirming the consequent; theories of confirmation purporting to lay out logical relations between evidence and hypotheses have been generally abandoned; subjective probabilities work better, but it's hard to see how they provide us with justified beliefs; and so forth. How does Atlport help us to answer van Fraassen? I am more permissive than van Fraassen. We agree that abstract theory is important for gaining genuine scientific knowledge (though I may be more inclined than he is to restrict that to a few special domains in which physics has been successful). But unlike van Fraassen I allow scientific knowledge to extend far beyond the observable. I think this arises chiefly from two differences between us: (i) I cannot identify any reasonable criterion that makes the directly observable, and only the directly observable, so especially knowable; and (ii) I focus not so much on what entities we believe in as on what causally or nomologically relevant features we attribute to them, since even quite observable objects can have highly theoretical properties ascribed to them, from being subject to a force of 17 newtons to having an obsessional neurosis. It seems to me that however stringent we make our standards of certainty and meaningfulness, we can on occasion ascribe one of these properties no less reliably and with no less settled signification than properties distinctly registered in our perceptual apparatuses. In Nature's Capacities and Their Measurement 5 ! describe the Stanford Gravity Probe-B, which may, I argue, enable us to ascribe with great confidence a spin-orbit coupling induced precession to a small number of gyroscopes spinning deep in space, and thereby test the general theory of relativity. I do not know whether Allport agrees with

296

NANCY CARTWRIGHT

my claims about the chain of reasoning that leads to such single-case ascriptions. But if I am correct in these claims, it puts us in a position to envisage simple inductive generalizations to laws relating highly theoretic properties, a kind of inference that is not available given van Fraassen's starting points. The structure of the argument is essentially that of Clark Glymour's bootstrap methodology: 6 the data from the experiment combined with our background knowledge entail an instance of the taw under study. The question is, which law? How far up the ladder of generality can we go? We begin with a very detailed law: any highly homogeneous fused quartz gyroscope placed in space near the earth, electrostatically suspended by a layer of superconducting material, will experience a certain amount of precession due to spinorbit coupling. Presumably some regularity of this kind would obtain if only enough details of the experimental design were filled in; that, after all, is what the experimenters are at pains to ensure. A positive experimental result will clearly count as a positive instance of this law, and will count strongly in its favour. But this law is a long way from the equations of the general theory of relativity that we are aiming to test. Can it be deduced from them, in conjunction with the requisite auxiliaries? If so, holding fixed the auxiliaries, the experimental result can be seen as an inductive instance of a part of the content of the general theory: it is what that aspect of those very abstract laws amounts to in this very concrete situation. This is the gist of Adolf Grianbaum's generic-specific account of explanation, which I discuss in How the Laws of Physics Lie. The account transforms an argument with hypothetico-deductive structure into a case of induction. Much more must be said about why the handful of results that the Gravity Probe-B will collect constitute a strong inductive base, 7 but the point here is just to remind ourselves that once we admit 'theoretical' features into our data base, we have available traditional forms of support that are closed to van Fraassen. Again, then, the question: Is the highly detailed 'phenomenological' law just alluded to indeed a deductive consequence of the general laws (given true descriptions of the experimental arrangement), or not? I think not. I find it instructive to break the reasoning process into two chunks: (i) the derivation of the behaviour in an abstract diagrammatic model (the simulacra of the simulacrum model of How the Laws of Physics Lie), a process elaborating Schiff's original observation that a spinning gyroscope should experience a shift in its precession due to

IS N A T U R A L SCIENCE ' N A T U R A L ' ENOUGH.9

297

the space-time curvature predicted in the vicinity of the earth by the general theory of relativity; and (ii) the elaborate fitting of this diagrammatic model to a real gyroscope spinning in real space. The full details of this latter will have taken more than 20 years of effort by a large team of experimental physicists and technicians. At both stages approximation and fudges and totally fictitious 'idealizations' of the kinds described in How the Laws of Physics Lie occur. 1 wilt not look at the details here because I doubt that Atlport would deny that these fudges are used. Rather, he feels differently about their significance: '[T]he current existence of anomalies, dubious mathematical rigor, 'ad hoc manoeuvres', and deficiencies in predictive power may simply iIlustrate a transitional 'economy with the truth'" (Sect. 1). But as I have been urging, that defence of realism may go a long way against a kind of Baconianism that scorns theory altogether. It is very much what is at stake in the debate with modern Duhemians. A pale blue dove goes no way to confirming the hypothesis 'All doves are grey'; but it does not disprove the usefulness of this claim as a starting point for the construction of models for different varieties of doves. Given this range and variation in colour among doves and the differences between that range and variation and those of other species of birds, this claim may provide not only the best and most efficient general starting point for the construction of models, but indeed a very powerful one. Coupled with local knowledge 8 it may provide highly detailed profiles of different varieties of doves in different regions; and given our general background theories of evolution and ecology, the mutual constraints between the covering law in question ('All doves are grey') and our local knowledge, in analogy to the 'interplay between theory and experiment' in the standard model, can lead to very successful - and very unsuccessful - novel predictions. This returns us to the rejection of charge/parity conservation on the basis of the experiments described by Franklin. I make here the same point as throughout: the rejection of theory on the basis of experiment seems to be indifferent between the realist and the Duhemian. Has the theory been shown false - thereby admitting its candidacy for truth in the first place - or, rather, has it been shown to be uncombinable in any neat, reasonable and non-ad hoc way with the other claims we so far wish to cling to? The real difference between Allport and me, I suspect, is that he sees the Duhemian position as an unpleasant 'half-way house' which he cannot take seriously, and which he continually drops out of con-

298

NANCY CARTWRIGHT

sideration. Either nature is random or it is governed by a few simple and elegant principles - symmetries being so far the very nicest of these that we have imagined. The alternative - that there are tens of thousands of patches, cut up in no particularly logical way, exhibiting tens of thousands of different regularities of countless different forms - is not a genuine contender for Allport, I believe. Perhaps he wonders: Why should it be thus when it might be unified? But why should it be unified? Why indeed should it be any one way rather than another? If we are the builders of the laws of nature, I can see an answer. But not if 'God' is. There is one place where Allport does offer an argument that confronts the Duhemian position directly. That is in his claim at the end of Section 9 that "cosmology can only be possible as a science if it is assumed that the development of the entire universe has been governed in the main by a finite set of all-embracing laws" (Sect. 8). We develop our laws of physics, Allport argues, by studying terrestrial phenomena; yet it is these laws that we must use to produce cosmological models. What else do we have? As an anti-Duhemian attack, this remark is well aimed. It surely counts more against me than against van Fraassen or Duhem himself, since I am in general sceptical of the range of phenomena for which anything like our current taws of physics will produce good models. I am not surprised that physics has trouble with feathers and with fluid flow; I wait to see the vaunted deterministic treatments of dice throws; and I would positively bet against the quantum mechanics of macroscopic bodies. 9 Independent of my views and of the dispute between Duhemianism and realism, Allport broaches a crucial issue here, for which we so far have only the thinnest of philosophical accounts. How does the 'interplay between theory and experiment' work in cosmology, or even the interplay between theory and concrete fact, experimental or not? Compare the attempts to produce various grand unified theories. What facts control these theories? Precious few. Nevertheless, the constructions are highly precise, and progress, both positive and negative, is possible, in large part because we are already committed to theories constrained by certain kinds of symmetries, which we replace, if at all, reluctantly and by other symmetries. The symmetry principles themselves are supposed to be borne on a larger set of empirical results, but again the number and variety, given the richness of the phenomena of the universe, are not great, and the structure of the support is not

IS N A T U R A L S C I E N C E ' N A T U R A L ' E N O U G H ?

299

clearly enough laid out to settle the issue between the Duhemian and the realist. Is the ladder of support 'straight up' so to speak, or is it 'dog-legged'? Do we adopt symmetry principles in some domains and then carry them across as constraints in the study of new domains? Of course that would be an unproblematic strategy if we already knew that 'the entire universe has been governed in the main by a finite set of allembracing laws'. In that case the symmetries, or any other principles, gathered in one set of domains would count as a genuine empirical constraint when carried to a quite different one. Return to cosmology. Perhaps it is only possible because we are determined to build our models a certain way - determined either by aesthetics, as Altport hopes to avoid, or by the need to constrain ourselves somehow in the face of the freedom allowed by the thinness of the data. Before closing I have two brief further remarks. First, Allport likes taws; I like causes. Allport argues that in causality "there is something a little too anthropomorphic and suggestive of analogies with acts of will" (Sect. 3). I think that in the concept of law there is a little too much of God. We try to finesse the issue with possible worlds, fictive regularities, and ceteris paribus clauses. But in the end the concept of a law does not make sense without the supposition of a law-giver. In Nature's Capacities and Their Measurement I explain why ! find the problems with causation far less troubling than the problems with laws. Second, Allport argues that it is often hard to distinguish laws and entities, using virtual particles as an example. If the point is that we often talk about particles which we believe ~n only because of highly abstract theory or which we can only come to understand or learn more about via this abstract theory, I agree. For these particles, our degree of belief can be no greater than our degree of belief in the abstract theory. I did not want to claim that particles are always more welt known or knowable than theories. Rather my claim was that there is a kind of explanation - causal explanation - that particles can provide that is closed to laws, and that for this kind of explanation, inference to the best explanation is valid. I suspect, however, that Atlport may be making a quite different point I do not yet understand. In the end Allport turns to aesthetics. He does not want in any way to rest the case for unifying covering taws on questions of beauty, since he takes this not to provide a defence of a sufficiently realist perspective. Nevertheless he does take questions of aesthetics seriously as well. In a recent article comparing my views with those of Sir Karl Popper, 1°

300

NANCY CARTWRIGHT

Altport argues (p. 56) that it is Popper's "lofty vision" as opposed to my "more practical" and "bottom up" perspective that will "communicate the excitement of pure research and the beauty of the theoretical vision that motivates it". That I think depends entirely on where one finds beauty. Popper and Allport, it seems, find beauty in "the mastery of the whole world of experience, by subsuming it ultimately under one unified theoretical structure" (Allport here is quoting Gerald Holton). I follow Gerard Manley Hopkins and look elsewhere: "Glory be to God for dappled things . . . . All things counter, original, spare, strange; Whatever is fickle, freckled (who knows how?)" 12

NOTES 1 Cf. Cartwright (1991a, t991b, forthcoming). 2 Though it may show that nature has complicated problems of consistency to solve. We do not, for instance, want to find colour patches in regions from which the taws of motion have dictate~ the disappearance of all matter and energy. This is a problem I worry about in Cartwright (1979). 3 Hacking (1983). 4 How the Laws of Physics Lie is a descendant of Duhem ([1906]/1962). So, too, is van Fraassen (1980). 5 Cartwright (1989). 6 Glymour (1980). 7 I give my own account in Cartwright (1992). 8 Or better: "When we couple it with local k n o w l e d g e . . . " . For an account of the active role of the scientist in constructing approximations, see Ramsey (1992). 9 Though of course I will not be surprised at the introduction of quantum models for a few selected phenomena, especially in devices like SQUIDs where the chains of cause and effect leading up to the phenomena are themselves modelled only in quantum mechanics. lo AIlport (1991). 11 Gerand Manley Hopkins: 1990, 'Pied Beauty'.

REFERENCES Allport, P. P.: 1991, 'Still Searching for the Holy Grail', New Scientist 5, 55-56. Allport, P. P.: 1993, 'Are the Laws of Physics 'Economical with the Truth", Synthese 94(2). Cartwright, Nancy: 1979, 'Do Token-Token Identity Theories Show Why We Don't Need Reductionism?', Philosophical Studies 36, 85-90. Cartwright, Nancy: 1989, Nature's Capacities and Their Measurement, Oxford University Press, Oxford.

IS NATURAL SCIENCE ' N A T U R A L ' ENOUGH?

301

Cartwright, Nancy: 1991a, 'Fables and Models', The Aristotelian Society Suppl. LXV, 55-68. Cartwright, Nancy: 1991b, 'Can Wholism Reconcile the Inaccuracy of Theory with the Accuracy of Prediction?', Synthese 89, 3-13. Cartwright, Nancy: 1992, 'Aristotetian Natures and the Modern Experimental Method', in John Earmen (ed.), Inference, Explanation & O~her Philosophical Frustrations, University of California Press, Berkeley, pp. 44-71. Cartwright, Nancy: forthcoming, 'How We Relate Theory to Observation', in P. Horwich (ed.), a volume of essays in honour of Thomas Kuhn. Duhem, Pierre: [1906]t1962, Aim and Structure of Physical Theory, Atheneum, New York. Glymour, Clark: 1980, Theory and Evidence, Princeton University Press, Princeton. Hacking, Ian: 1983, Representing and Intervening, Cambridge University Press, Cambridge. Ramsey, Jeffry: 1992, 'Towards an Expanded Epistemology for Approximations', in David Hull, Mickey Forbes, and Kathleen Okruhlik (eds.), PSA 1992, Vol. I, Philosophy of Science Association, East Lansing, MI, pp. 154-64. Van Fraassen, Bas: 1980, The Scientific linage, Clarendon Press, Oxford. Department of Philosophy, Logic and Scientific Method London School of Economics and Political Science Houghton Street London WC2A 2AE England