This paper is to my mind a classic, though we have already discussed at length most of its central themes. In Part I of the paper Miss Anscombe attacks the notion that causality must involve necessity and argues to the contrary that the central element in the notion of causality is the derivativeness of the effect from the cause; any necessity or universality is a further element and may be entirely absent. In the second part of the paper she argues that it is not mere ignorance of full causes that warrants the separation of causality from necessity, universality, and determinism. I will just make a few comments.

Part I:

1. First reply to "causality is not observable" objection:

     Notice the twofold reply to the charge that we can't observe causal efficacy. The first reply is that we observe causal efficacy if we are able, as we are, to use a wide range of causal terms and related natural kind terms correctly in perceptual situations. The idea of causality is an abstraction from countless active and passive verbs and natural kind terms and could not be introduced if we did not already have a large number of such words in our active vocabulary. (This is not to say that we have infallible knowledge in any such cases, but that is surely not required.)

2. Second reply to "causality is not observable" objection:

     The second reply is that we cannot in any case formulate the universal generalizations which, according to Humeans, are supposed to supply an account of what could not be observed in singular instances. (We have already raised this concern in our discussion of Mackie.) Notice Miss Anscombe's suggestion that we take laws of nature to be propositions ascribing properties to substances. These are not generalizations which tell us what always happens; rather, they tell us what to expect in the absence of interfering causes--and even here we must adjust for indeterministic causes.

Part II:

1. Causality, Predetermination, and Universal Determinism

     The second part of the paper is a bit more recondite. Miss Anscombe first tries to help us see the difference between causality and predetermination. Her claim is that in the absence of a "system" that entails the predetermination of a given effect, we have no antecedent reason to believe that an effect is predetermined by its causes--but this, of course, is no reason to believe that the effect in question is not caused.

     Also, Anscombe points out that someone--say, a physicist--might believe in the predetermination of each effect while not believing that the whole state of the universe at a given time is predetermined by its state at some previous time. The reason is, as she puts it, that such a physicist "may not think that the idea of a total state of the universe at a time is one he can do anything with. He may even have no views on the uniqueness of possible results for whatever may be going on in any arbitrary volume of space" (p. 142). Indeed, his belief concerning the singular effect is not that nothing could prevent (or could have prevented) the effect, but rather that a good scientific theory "should be such that only this result was possible as the result of the experiment." Universal determinism is compatible with this but not entailed by it. However, the non-determinist she has in mind "restricts his demand for uniqueness of result to situations in which he has got certain processes going in isolation from inconstant external influences, or where they do not matter, as the weather on a planet does not matter for predicting its course around the sun."

     Still, even this mitigated demand for predetermination looks false and in any case is not crucial for the various effects actually being caused. (Here she thinks, like Cartwright after her, that the solar system's "obligingness" gave rise to a false ideal of scientific explanation by suggesting that it would be easy to find similar stable models with no perturbing causes.) And even if Newton's system were a true deterministic system, it still wouldn't follow that universal determinism is true. What's more, even if we bring in other forces to supplement the mechanics, universal determinism still doesn't follow.

2. Necessitating and non-necessitating causes:

     The distinction she draws here dovetails nicely with my own account of natural necessity:

     A necessitating cause C of a given kind of effect E is such that it is not possible (on the occasion) that C should occur and should not cause an E, given that there is nothing that prevents an E from occurring.

     A non-necessitating cause is one that can fail of its effect without the intervention of anything to frustrate it (e.g., Feynman's bomb connected to a Geiger counter).

3. Non-necessitating causes and free choice:

     One question is this: Are choices, as non-necessitated effects, thereby "accidental" or "random"? No, since indeterminism is only a necessary and not a sufficient condition for free choice, which involves acting "according to an idea." (Notice the aspersions cast on the claim that free choice and universal determinism are compatible with one another.)

     The second, and more pertinent question, is this: Does the will, in acting on physical bodies, have to violate the laws of nature governing those bodies in order to exercise free choice? In particular, the laws governing the physical correlates of free choice are such that they allow us to "predict statistics of events when situations are repeated," and so a free choice may make its physical correlate occur in the wrong proportion:

    "The other objection is, I think, more to the point. Certainly if we have a statistical law, but undetermined individual events, and then enough of these are supposed to be purshed by will in one direction to falsify the statistical law, we have again a supposition that puts will into conflict with natural laws. But it is not at all clear that the same train of minute physical events should have to be the regular correlate of the same action; in fact, that suggestion looks immensely implausible. It is, however, required by the objection" (p. 146).

     Miss Anscombe's answer amounts to this: An act of will of a given sort might be associated with widely different sets of physical correlates. (Connectionism is relevant here). So there is no antecedent reason to think that acts of will are incompatible with the statistical laws governing such physical correlates. This is the point of the Coca-Cola example--something with the same intentional content is correlated with widely varying physical states:

    "It is not at all clear that those statistical laws concerning the random motion of the particles and their formation of small unit patches of color would have to be supposed violated by the operation of a cause for this phenomenon which did not derive it from the statistical laws" (p. 146).

Look at second last paragraph on p. 147.