To: Christian M.
From: Geoffrey Klempner
Subject: David Hume's two definitions of 'cause'
Date: 2 October 2007 11:42

Dear Chris,

Thank you for your email of 23 September, with your University of London essay in response to the question, 'In both the Treatise and first Enquiry, Hume provides two definitions of "cause". What does the second definition add to the first, and why did Hume think it necessary to introduce it?'

This is not a bad essay. You have said enough to answer the question. The only point which made me reach for my red pencil was your all-too brief discussion of the exchange between Robertson and Richards. I would have liked to have known more about why Richards thinks that Hume wants to defend a distinction between 'natural' cases of causation and 'unnatural' ones.

However, there is a lot more to say in this topic. In what follows, I am going to try to fill in the gaps and suggest things to think about in relation to Hume's theory.

I would disagree that, strictly speaking, the first definition does not contain an explanation of necessity. What it leaves out is the naive/ unreflective idea of *necessary connection*. To assert that events of type A are always followed by events of type B is to make a very strong claim: at all times and at all places if A occurs, then B occurs afterwards. This is a universal generalization with unrestricted domain. No human being can ever survey all the possible instances of A 'causing' B according to this definition. In other words, there is necessity, but it is the necessity of unrestricted generalization, or 'lawlike' necessity.

What human beings naturally believe is another matter. According to Hume, when we observe A 'causing' B, we imagine something invisible connecting A and B, a 'causal influence' which passes from one to the other. This belief cannot be defended by observation or logical analysis. So, the task for Hume is to provide an error theory, an explanation of why we falsely believe in a necessary connection, in addition to his analysis of causal necessity. He is providing an explanation of a philosophical error. Once we recognize the error, we will still naturally expect A to follow B, but we will no longer be inclined to offer incoherent accounts of what it means to say that 'A caused B'.

You are right to suggest that the situation here is in some respects comparable to Hume's account of the rationally unjustifiable but natural belief that objects continue to exist when unperceived, and have a distinct existence from the act of perception. In both the case of causation and the case of spatio-temporal continuants (or 'substances') Kant was able to complete the train of thought that, arguably, Hume was grasping at: that in order to have experience at all we need to place our experiences in a spatio-temporal and causal framework. Hume's 'fictions' become Kant's a priori categories of substance and cause.

However, there is arguably an important difference between the two cases. Whereas everyone is agreed that Hume's difficulty with spatio-temporal continuants was the result of his over-exacting empiricism, there are many philosophers today who would defend a Humean account of causation in terms of lawlike connection. And indeed, Kant's does not attempt to resurrect the 'necessary connection' but merely argues for the necessity of causal determinism, as a precondition for the possibility of experience.

Another point I wanted to make is on the borderline of this essay question, but could occur in an exam - and is at least worth mentioning here. You say, '...the impression of the first forces the mind to associate the second one'.

Critics of Hume have pointed out that the banned notion of 'necessary connection' reappears in the idea that an impression 'causes' a corresponding idea. In Hume's defence, one would apply his own analysis of cause in terms of lawlike necessity. The Humean philosophical psychologist, inspecting the workings of the Human mind, makes a generalization about the association of ideas which is claimed to be true - as a lawlike necessity. That is what the 'forcing' or 'causing' consists in, and nothing more.

My last point concerns the claim that 'A is always followed by B'. Hume was fully aware that circumstances differ in each case of A, and so if one is attempting to formulate a causal law, one would have to add extra conditions. At one point in the Treatise, he goes into some detail about 'rules for judging causes and effects' which take this into consideration. Throwing a stone against a window does not always break it; but only if the stone is heavy enough, thrown with sufficient force, while the window is fragile enough etc. etc. A criticism of Hume (made by Elizabeth Anscombe) is that it is in fact impossible to ever give an example of a 'lawlike generalization' that applies to any particular example of cause and effect. The 'ceteris paribus' clauses ('other things being equal) are infinite. So it turns out that this 'law' is an ideal law that can never be formulated - rather a long way to come for a strict empiricist.

All the best,

Geoffrey