Natural Necessity, Objective Chances and Causal Powers
C. Behan McCullagh
There is an interesting convergence among several recent theories of causation. They describe causes as events of a kind which increase the objective chance of events of other kinds, which are their effects. The theories I have in mind are those of David Lewis (1986, 1994), D.H. Mellor (1995), and Peter Menzies (1996). They attribute various other properties to causes, but all agree that this is an important and necessary fact about them. Thus, dropping a crystal wine glass onto a hard floor can be said to have caused it to shatter because it increased the probability of its shattering at that time, since whenever a piece of crystal glass is struck by a hard object, it will very probably shatter.
Such generalizations relating cause and effect can be understood as being true of something because of one of its properties. For instance one can say: if something has the molecular structure of crystal glass, then if it is hit by a hard object it will very probably shatter. The generalization relating cause and effect can then be seen as a disposition of the thing whose property makes it true. In this case it amounts to saying that things which have the molecular structure of crystal glass are fragile. More generally, the generalizations involved in causation are of the following form: if something has a certain property, then if it comes into a certain relation with something else, it will result in another event, with a certain objective probability. Call this generalization [G].
This generalization looks very much like the analysis of causal powers offered by R.Harré and E.H. Madden (1975). They defined the causal power of a thing in terms of its disposition to behave in certain ways in certain circumstances because of its nature. They wrote: "'X has the power to A' means 'X will or can do A, in the appropriate conditions, in virtue of its intrinsic nature' (p.86).
What difference does talk of 'causal powers' make to the analysis? From what they say, it seems that two conditions apply to things with a causal power ( p.93). First, things which have such powers are things of which [G] is necessarily true. The necessity, they explain, is a consequence of their having the property which makes that generalization true, i.e. which gives it the power it has: 'they must behave in the specified way in the given circumstances, or not be the things that they are.' (p.91) The same argument is given to support the second condition, which is that given the cause, that is the conjunction of the object with its triggering conditions, then the consequence will necessarily follow. Let me quote their argument in full.
I do not find this defense of natural necessity very persuasive, for reasons which I shall shortly explain. The question I pose in this paper is whether there are any good reasons for believing in it. Humeans would certainly deny that there are, for according to Hume regularities in nature are not necessary but entirely contingent, matters of chance. After discussing Harré and Madden's arguments, I look at Simon Blackburn's recent vigorous defence of a Humean position. There is a serious problem with his view of nature which is best met, I suggest, by adopting a theory of natural necessity. Finally I argue that Lewis needs this assumption for his own theory of objective chance and causation, though he has not recognised the fact.
What should we say about Harré and Madden's defence of natural necessity? Notice that their argument can go two ways: the necessity they claim could be a priori or a posteriori. One could say, for instance, that our conventional concept of crystal glass is one which includes its fragility as an essential, i.e. a necessary, characteristic. If something is made of crystal glass, then it is necessarily fragile, in the sense that if it were not fragile it would not be made of crystal glass. The truth of this claim rests entirely upon our definition of ordinary glass.
Alternatively, one might say that, given the universal law that things with the molecular structure of crystal glass are fragile, which we know from scientific inquiry, it is an empirical fact that things made of such glass are fragile, so that if something were not fragile it could not be made of such glass.
Harré and Madden employ both of these arguments: "We follow the scientific tradition in identifying the real essence of a kind, material or individual, with its nature, which is progressively revealed a posteriori by empirical investigation' (p.102). Thus science reveals the properties of crystal glass, and these consequently become part of the definition of that substance. It is then both a logical and a scientific fact that is something were not fragile it could not be crystal glass.
The immediate response of a Humean to any suggestion of natural necessity is disbelief. Harré and Madden might say it is a contradiction to say something is made of crystal glass and not fragile, but they are mistaken. It is entirely possible, say in another world, that things with the molecular structure of crystal glass are not fragile, but malleable or indestructible. Consequently the fragility of crystal glass in this world should be considered a contingent, and not a necessary fact.
This argument succeeds, I think, in destroying any argument which suggests that the fragility of crystal glass is a logical necessity. One can certainly envisage a world in which crystal glass is not fragile. If there is any necessity in the relation between the two it must be natural, not logical, necessity. To say A is a naturally necessary property of B is to say that for B to exist without A would be inconsistent with the laws which govern nature. The laws themselves could have been other than they are, in a different world, and in that sense they are contingent. But to say that a generalisation expresses a law, for instance that crystal glass is fragile, means, in part, that instances of the things it describes are bound to be related as it states. It is for this reason that laws are so useful in predicting what will happen, and in explaining what already has. Laws of nature do not express mere contingencies, but have counterfactual force. To deny natural necessity, one would have to deny this feature of natural laws.
Now we see why Harré and Madden's argument is not entirely successful against the Humean, for basically it begs the question against him. The two kinds of necessity involved in causal powers are both dependent upon the assumption that laws of nature state physically necessary relations between types of things in the world. Yet that it precisely what the Humean denies.
Before turning to Humean theory, a small fault in Harré and Madden's analysis of causal power should be noted and corrected. They state that, if X has the power to do A, in virtue of its nature, then it will do A in the appropriate circumstances, unless X is a person, in which case she might not do A, but can do so (pp.86-7). Nancy Cartwright (1983) and others have taught us how mistaken it is to suppose that the outcomes described by natural laws really will occur, since their occurrence is subject both to a range of facilitating conditions and the absence of interfering conditions, both of which are entirely contingent. The shattering of the glass when dropped above a hard floor, for example, depends upon normal gravitational force, and the absence of someone catching the glass before it hits the floor. Philosophers now say, not that the effect will occur, but that there is an objective chance, or probability, of the effect occurring. This chance is a theoretical measure of the probability of the effect's occurring in an ideal world in which all the facilitating conditions are present and no interfering ones impede it. I shall follow the convention of saying that causes produce, not an event, but the tendency for an event of a certain kind to occur, meaning that there is an objective chance of its occurring in the circumstances. Harré and Madden are aware of this complication (pp.99-100), but did not incorporate it in their definition of causal power.
One final ambiguity remains, namely in the meaning of 'tendency' in this context. To say that crystal glass has a tendency to smash when dropped is to speak of its potential tendency to do so. When it is dropped, then its tendency to break is an actual tendency. Causes are events which activate potential tendencies by triggering the dispositions which they are. Causes make potential tendencies actual. Actual tendencies are sometimes mediated by micro-processes, as when the glass gathers speed as it drops towards the floor, and then the molecules of glass come apart under the impact of contact with the floor. It may be that the interaction of fundamental particles produces the only tendencies for which there are no micro-processes, though I still like to think of the action of the moon on the tides as action at a distance.
What, then, is the Humean response to this talk of natural necessity? Simon Blackburn (1993) has defended a radical Humean view of natural laws recently. Laws of nature, he says, do not provide a 'straightjacket', constraining events which fall under them, because their truth is entirely contingent. 'The ongoing regularity and constancy even of a thick nexus between one kind of event and another is just as much a brute contingent regularity as the bare regular concatenation of events (p.99). Galen Strawson has argued that the regularities of nature require some explanation (Strawson,1989, p.224), but Blackburn says we have no evidence that the regularities are constrained by God or necessity. Indeed we cannot even imagine how they would be controlled. 'The question is whether we know what governing or bringing about would be when we have no example, and indeed no conception, of the kind of fact alleged to be doing it' (Blackburn,1993, p.101). Even if you allow causal connections between events on particular occasions, there is no way of explaining 'how the 'musts' present on one occasion throw their writ over others' (p.106). Furthermore, suppose you postulate God or fundamental forces as sustaining laws of nature, what reason is there for thinking these will remain constant (p.98)?
Any theory which posits natural necessity, he says, 'looks extravagant. It asks from us more than we need' (p.106). Hume, and Blackburn, allow that after we witness regularities in nature we imagine it to be governed by causal necessity, but that is only 'a distinct mental set' (p.105). There is no need to suppose that anything really does govern the regularities we observe.
Do we need to postulate natural necessity, and is the concept of events being governed by natural laws intelligible? Blackburn answers no to both questions. Let me consider them in turn. First, we need to postulate natural necessity, not just to explain observed regularities as Strawson said, but to make predictions and explanations based upon them rational. This last point has not been adequately considered by the philosophers. According to Hume, we have the habit of mind of supposing natural necessity, but no reason for holding that supposition true. Blackburn says we do not need to suppose it true in order to draw the inferences and carry out the scientific inquiries we do (pp.106-7). Perhaps not. My point is simply that the inferences by which we predict the future and explain the past are irrational if we insist that in fact, in reality, all sequences of events are contingent, and could be otherwise than they are. In that case, as Blackburn points out, there is no reason to suppose that unexamined cases will conform to observed regularities. Indeed, in that case it is absurd to suppose they will. It is as mad as believing one will win the lottery this time, because you have failed to do so over the last fifty tries. There is nothing to warrant such a belief, no matter how strongly it is held. It is of course possible that future events will conform to past regularities, just as it is possible that this time you will win the lottery. But it is mad to believe it. In both cases what happens is entirely a matter of chance. We must assume natural necessity for our predictions and explanations to be rational.
This argument is not an argument explaining regularities, such as Strawson proposed. We do not have to establish the existence of God or 'fundamental forces' (Strawson,1989, pp.90-1). Rather it is a transcendental argument: it says what we must assume to be the case if we are to regard as rational predictions and explanations which are generally accepted as rational. As is well known, they require that generalisations about nature be not just accidentally true but have counterfactual force, the force of natural necessity. We can go on making predictions and explanations without assuming rational necessity, but while we believe nature to be entirely contingent, such predictions and explanations will be irrational.
Since we have posited no causes of natural necessity, the question of how these causes govern nature does not arise. All we have assumed is natural necessity, that is that in this world things with certain properties necessarily, that is always, have certain powers, and that when those powers are triggered, certain active tendencies will necessarily, that is always, result. It is impossible, in this world, for these relations not to exist. Notice I have dropped all reference to forces and powers, which seem to suggest mysterious agencies at work. Like the Humeans, I prefer just regularities. But in this case the regularities are not just accidental, but universal, so that they could not be otherwise. That is why we call them necessary.
Is this talk of natural necessity compatible with the fact that, even in ideal conditions, the probability of certain outcomes given certain causes is less than one? For instance, suppose an antibiotic drug cures streptococcal infections in 80% of cases. Is the tendency to cure 80% of infections a natural necessity? If this percentage represents the frequency of cures in the long run, we assume for the sake of inference that in this world the frequency could not be otherwise, i.e. that it is necessarily what it is. On the basis of such frequency statements we infer the objective chance of a particular instance of the cause being followed by a particular instance of the effect, in ideal conditions.
Note that the fact that our knowledge of natural necessities is imperfect does not affect these arguments at all. We must assume the generalisations we have arrived at to be necessary for the inferences we make from them to unexamined cases to be rational. Which general laws we assume to be necessary might change as scientific inquiry proceeds, as more cases are examined. The point here is simply that they must be regarded as necessary for rational inferences to be drawn from them.
Some of the most perceptive discussions of objective chance are by David Lewis. It is interesting that he does not acknowledge the importance of natural necessity in drawing rational inferences about the chances of future events. He likes to relate the chances of an event's occurring to what it is rational to believe are its chances, given the best scientific theory about the world and the history of the world to date (Lewis, 1986, 1994). He calls this theory the 'Principal Principle'. This suggests that estimates of chance are entirely epistemic, functions of our theories and beliefs. This would be a convenient position for a Humean to adopt, confining necessities to theories and not having them in the world. Indeed he proudly declares at one point that with his theory 'the broadly Humean doctrine is upheld' (Lewis,1986,.p.112). He explains it thus:
However, elsewhere he denies that the chance of an event's occurring is a function of belief: 'this is not epistemology!...Rather, I'm talking about how nature--the Humean arrangement of qualities--determines what's true about the laws and chances. Whether there are any believers living in the lawful and chancy world has nothing to do with it.' Lewis, 1994, pp.481-2). Then why define the chance of an event's occurring in terms of rational credence, according to the Principal Principle?
As far as the present discussion is concerned, it does not matter which theory of chance Lewis prefers, an epistemic one or an objective one. Both require the decidedly non-Humean assumption of natural necessity. Inferences about chances of events in the world are not rational if the laws upon which they are based might not be followed in future. They are only rational on the assumption of natural necessity, as was explained above. And the objective chance of events occurring is always fifty fifty if, as Humeans hold, all events are contingent. To suppose that future events will follow past patterns is, for Humeans, merely to indulge in one kind of gambler's fallacy. The chance of an event's occurring only conforms to the regularities of nature if those regularities always hold.
Blackburn, Simon, 1993. Essays in Quasi-Realism. Oxford University Press, Oxford and New York.
Cartwright, Nancy, 1983. How the Laws of Physics Lie. Clarendon, Oxford.
Harré, R. and Madden, E.H., 1975. Causal Powers. A Theory of Natural Necessity Blackwell, Oxford.
Lewis, David, 1986. 'Causation' in Philosophical Papers. volume 2, pp.159-213. Oxford University Press, Oxford and New York.
Lewis, David, 1994. 'Humean Supervenience Debugged' in 'Symposium: Chance and Credence' in Mind 103, pp.473-90.
Mellor, D.H., 1995. The Facts of Causation, Routledge, London.
Menzies, Peter, 1996. 'Probabilistic Causation and the Pre-emption Problem', Mind 105, pp.85-117.
Strawson, Galen, 1989. The Secret Connexion. Causation, Realism, and David Hume. Clarendon, Oxford.