philosophy is for everyone
and not just philosophers

philosophers should know lots
of things besides philosophy

Philosophical Connections

Electronic Philosopher

Feature Articles

University of London BA

Philosophy Lovers Gallery

PhiloSophos Home

International Society for Philosophers

G.E.M. Anscombe on singular causation


To: Tony L.
From: Geoffrey Klempner
Subject: G.E.M. Anscombe on singular causation
Date: 28 August 2006 10:22

Dear Tony,

Thank you for your email of 21 August, with your University of London Introduction to Philosophy essay in response to the question, 'Anscombe takes singular causation seriously, whereas Hume does not.' Explain and discuss.

You start off by defining 'singular causation':
We mean that this specific pair of events, A and B, has in itself some property or properties in virtue of which we can say that A caused B without knowing whether A-type events are always (or often, or indeed ever, apart from this one instance) followed by B-type events.

However, there is an ambiguity in this definition, which also appears in your account of what Hume means by 'necessity:
He takes it that what “necessity” really amounts to here is the firm mental conviction that we form, after observing repeated instances of A-type events being followed by B-type events, that A-type events will always in future be followed by B-type events. That is, we expect what he calls “constant conjunction” of cause and effect.

The ambiguity comes to light when one considers the two components in Hume's account of causation:

1. Hume is concerned to explain the psychological process by which we form beliefs about causes and effects. The mechanism is explained in terms of the theory of ideas and impressions.

2. 'A caused B' is true only if a type B event always and everywhere follows from a type A event (the 'univeral generalization' requirement).

Whatever 'necessity' or compulsion we feel in believing that A caused B, the necessity that characterizes causation is the necessity of a universal generalization, nothing more or less. B always follows from A, at every time and every place. This is in itself a very powerful claim to make, and one which can never be conclusively verified in actual cases, but it is one which is fully consistent with - indeed necessitated by - Hume's rejection of any 'metaphysical' component in causation.

Now consider this claim in relation to the idea of singular causation, and in particular in learning situations where one grasps the meaning of 'cause'. Anscombe would say that no-one in an everyday situation where causes and effects are identified is thinking of making such a huge claim. That is what her examples are meant to show. We know that Fred caught measles from Violet, without needing any concept of the Humean covering law that accounts for the causal connection.

The question of what causation is, is distinct from the question of how we know that a particular case is one of causation. Hume is prepared to allow 'clear experiments' which lead to a strong and justified conviction that A caused B. What this means, however, is universal generalization. Anscombe, on the other hand, believes that given that we grasp the meaning of 'cause' through such clear experiments, the question whether a causal claim logically entails a universal generalization remains open. Regularity and generalization may play a role in our grasp of causation but there is no reason why this role should be held to be constitutive of the very notion of a 'cause' in the way that Hume claims.

In principle, Hume indeed has no problem at all with the idea of a type A event which is so unique that it only ever occurs once in the history of the universe.

You say:
Hume would surely insist that you couldn’t say, in such a case, that A caused B. You would have to say that the “cocktail” caused B. That’s because, for Hume, causality implies constant conjunction, so it would be simply wrong to use the term even for patchy conjunction (where A-type events are quite often, but not always, followed by B-type events), let alone for a single isolated instance such as we’ve envisaged.

But this isn't necessary. Assume that the 'history of the universe' is some finite length of time. Then Hume can say that if at any time the history of the universe is repeated up to the time when A occurs, then B follows.

We have seen that Hume can allow single cases where we form the reasonable conviction that A caused B, and he can also allow cases where A causes B only on one occasion in the history of the universe. Both of these, I have argued, are fully consistent with a 'universal generalization' component in the analysis of causation.

This looks like a defence of Hume against Anscombe, to the effect that Hume DOES 'take singular causation seriously'. However, as I have indicated, Anscombe is making a stronger claim, to the effect that the universal generalization component is not part of the meaning of a 'cause'. The examples she cites (such as the case of catching a contagious disease) are not sufficient to show this. On the other hand, all she has to do is raise a legitimate doubt about Hume's claim about universal generalization.

The discussion does not stop there, of course. The whole point of the universal generalization component is to find a substitute for the naive notion of 'coming from' that seems to be part of our understanding of causation. If one rejects the Humean 'solution', then we are right back where we started!

All the best,