Tuesday, August 27, 2013

Hume and Anscombe on causation

Dear Max,

Thank you for your email of 26 September, with your essay for the University of London Introduction to Philosophy module, in response to the question, ''Hume holds that causal claims are always general claims, whereas Anscombe denies this.' Explain and discuss.'

As you stated in your email, you have ranged quite widely in attempting to answer this question. However, I don't feel that you have really succeeded in hitting the nail on the head.

The main problem with your approach arises from a confusion between what is stated when we make a causal claim -- in modern terminology, the truth conditions of a causal claim -- and the evidence that we may have on a particular occasion for making a causal claim, or the conditions for its verification.

'Hume holds that causal claims are always general claims.' What this means is that, according to Hume, when we make a causal claim we *always* commit ourselves to a general claim, even when this claim is not explicitly stated. What we commit ourselves to, when making any statement, is what follows if that statement is true. Of course, it may not be true. We don't know for certain. And indeed in any case of a general claim, we can never know because we are talking about *any* time and *any* place, not just our immediate environment.

So, for example, if I assert that throwing the stone at the window caused it to break, then according to Hume there is a statement, couched in purely general terms, whose truth follows from my assertion, assuming that my assertion is true (i.e. not only was the stone thrown and the window broken, but the former event caused the latter).

Anscombe (following Russell) sees a major difficulty here, in the very idea that such a general statement can be formulated, even in principle. The stone must be heavy enough (not a piece of gravel) must be thrown hard enough, the window must be sufficiently fragile etc. etc. This is not Anscombe's only objection. You mention some others. The point, however, is that this is precisely what she denies. According to her, it is not *always* the case that a causal claim commits one to the truth of a universal generalization. There are at least some causal claims which do not involve such a commitment.

That's one way in which generality is involved in Hume's account of causation. The other way, which you emphasize, arises in the context of our grounds for making causal statements. We witness a 'constant conjunction' between events of type A and events of type B and form the expectation that, when next presented with an event of type A there will be an event of type B. But Hume surely would not deny that there are all sorts of occasions when we make a causal claim in the absence of such prior evidence. That's because the instance in question falls under more general cases of things we have observed in the past. You can categorize causes into knocks, pushes, breakages, switches etc. When presented with a novel example of causation our minds readily find a suitable classification.

Hume has ample resources for explaining why his theory does not commit him to the truth of the statement, 'night causes day.' To state that a causal statement implies the truth of some generalization, which it would be possible to state in principle, does not entail that every generalization implies the truth of the corresponding causal statement. That's all Hume has to say in principle. After that, one gets down to cases (using Hume's 'rules for judging causes and effects' in the 'Treatise').

Insofar as Anscombe's argument seems to hinge on the fact that we are able to make causal claims without observing constant conjunctions, then the response should be that she is missing the point. 'You gave me the flu', I complain, based on no evidence other than that you had the flu two days ago and now I've got it. However, our world is not the world of primitive man, we know that the flu is a virus and that viruses can be transmitted in various ways. You don't have to probe very deeply to find a general law about infectious transmission. This is more or less the point you make.

However, I don't think that Anscombe is saying this, at least not exactly. One way to reformulate her argument against the claim that causation logically implies generality would be to imagine a possible world which is not too unlike our actual world, except for the fact that there are no scientific 'laws' as such. Imagine a meddlesome deity who is never content with just leaving things alone. You can 'do' science up to a point, but the efforts of scientific researchers are always thwarted in the end by inexplicable failures to reproduce experiments.

In this imaginary world, there are no general laws as such, that is to say, no possibility of making a true lawlike statement of the form, 'All As are B'. So, on Hume's account, there would be no causation. And yet in such an 'a-nomic' world, we get on perfectly well with our everyday, and ultimately unanalysable, notion of a cause.

What Anscombe is after, in other words is the *concept* of a cause. Russell thinks that this is ultimately dispensable, because science has found a better way, using the notion of a law. This arguably ignores the fact that our human world is in some sense prior to the world of science. Our social relationships, our ability to explain our actions to others or call others to account presupposes an understanding of causation which is prior to science.

All the best,

Geoffrey