To: Mark S.
From: Geoffrey Klempner
Subject: David Hume's account of causation
Date: 19th November 2009 22:16
Thank you for your email of 1 November, with your essay for the University of London Modern Philosophy: Descartes et. al. module, in response to the question, 'What was Hume trying to do in his account of causation? How successful was he in attaining his objectives?'
Apologies for the delay. I 'replied' on 9 November but for some reason failed to send the email.
There are some pertinent criticisms in your essay which relate to the narrowness of Hume's radical version of empiricism. However, I think that an examiner might feel that you have missed the bigger question: namely, whether or not we should be 'Humean' about causation. A relatively contemporary exponent of the Humean approach would be Carl Hempel's 'deductive-nomological' account of explanation. G.E.M. Anscombe, in her article 'Causation' (in OUP 'Causation and Conditionals') represents a notable attack on the idea that causation can be analysed in terms of deduction from a universal generalization.
The problem stems from the assumption which you make at the beginning of your essay, that Hume's primary objective is to 'give a genetic account' of the causal relation. Hume offers a genetic account, because this is the only methodology which he will accept as being consistent with empiricism. (By the end of your essay, however, there are hints that you recognize this point.)
In other words, the genetic account is offered as a means to an end: to determine the semantics and epistemology of the causal relation. It is in relation to these questions, that we should be asking whether or not 'Hume succeeds'. (Of course, it is possible that we may be led to conclude that he fails *because* of his assumption that the correct methodology is to conduct a genetic analysis.)
The 'big' question about causation is whether it could possibly be true that, e.g. when a tap of one of these keys causes a letter to appear on the computer screen, *all* that it is true to say about this evident fact is that a certain universal law holds, and that the particular case can be derived from that universal law. There are two aspects to this: first, the negative idea that there is *nothing* at all to the 'causal fact' that I have described, apart from the truth of a causal law, together with the observation that this instance falls under that law; secondly, the idea that when I assert that the key-tap caused the letter to appear, I am implicitly invoking a law which relates to all places and times, now and in a potentially infinite future, where a 'relevantly similar' key-tap and letter-appearance events occur.
Of course, Hume doesn't talk in this way (about 'potential infinity') but it is implicit in what he says about the difference between non-lawlike conjunctions and lawlike conjunctions.
The big problem with the law theory, as you point out, concerns the notion of 'relevant similarity'. The term usually used for this is 'ceteris paribus clauses', or 'other things being equal'. Not every key-tap is followed by a letter-appearance. The problem is, however hard we try to specify the exact circumstances in which one will always follow the other, there will always be something left out. Hume seems to have been unaware of this problem.
You also invoke the possibility of a 'nativist' account of the concept of a cause. Suppose it could be proved, through some kind of 'poverty of stimulus' argument, that human beings are born with the more or less implicit notion of a cause, without which we would be unable to learn how to make sense of our experience? Actually, some would argue that this has been done very successfully -- by Kant in the first part of his 'Critique of Pure Reason'. Causation is an 'a priori' concept, and Kant 'proves' this by means of his transcendental deduction of the categories. Or, more briefly, experience would be impossible if we were not equipped from the start with the idea of causation.
As P.F. Strawson argues in 'The Bounds of Sense', we can detach Kant's claim that determinism is true a priori from the conclusion of the transcendental argument: that we necessarily have an idea of cause would still be true in a universe where for some reason the thesis of determinism did not hold.
Regarding the two alternative definitions which Hume offers, in terms of the truth of a universal generalization, and in terms of the psychology of the formation of causal beliefs, one could argue either that the second is redundant, or that they are both (as Hume claims) aspects of the same theory. All that is necessary to say about the semantics of the causal relation is the deductive-nomological account. However, that still leaves a question unanswered regarding how causal beliefs are formed. Here, it does seem that Kant saw the point that Hume missed. Hume imagines a subject having all sorts of experiences, then 'learning' to identify 'causes' and 'effects': this is an impossible scenario, for the same reason as the scenario of a subject experiencing 'impressions' who has not yet 'learned' to interpret these impressions as perceptions of objects in space. In both cases, we can say that the fault lies with Hume's radical empiricism.
I need to pick you up on one point you make, in relation to Quine's denial of the analytic/ synthetic distinction. Quine accepts that we can make stipulative definitions. 'Henceforth, whenever I describe a man as a 'bachelor' I am stating that he is unmarried.' However, your objection that it is an 'empirical question' how the term 'bachelor' is actually used in a linguistic community makes Quine's objection look trivial. As if the assertion would be falsified if some ignoramus thought that 'bachelor' meant a man who makes cricket bats. Contemporary arguments over the analysis of philosophically or scientifically problematic concepts illustrate Quine's point better: that the question isn't 'what we mean by X' but rather 'what concept of X do we want' where the question is to be settled by appeal to logic and experience.
All the best,