Probabilistic Causation and the Pre-emption Problem

Abstract

Probabilistic theories of singular causation claim that singular causal relations can be analytically reduced to probabilistic relations.' Though they differ in detail, these theories depend on the guiding idea that a cause makes a difference to its effect by making it more probable than it would otherwise be in the circumstances.2 Appropriately formulated, this idea applies uniformly to both deterministic and probabilistic causes. I wish to argue that singular causation cannot be reduced to probabilistic relations. There is a problem-I call it the pre-emption problem-that shows that the guiding idea that a cause raises the probability of its effect does not accurately match our intuitions about singular causation. The pre-emption problem is illustrated by examples in which there are two processes leading to some effect, one of which goes to completion and brings about the effect, but in doing so cuts off or pre-empts the other process.3 The significant feature of these examples is that the event that is the potential, but not actual, cause raises the probability of the effect, while the event that is the actual cause does not raise the probability of the effect at all. In ? II describe one such example in the context of a very plausible way of spelling out the guiding idea of probabilistic theories. An initially appealing response to the examples illustrating the preemption problem is to think that they can be overcome somehow by com-