Causation and Reductive Explanation

Abstract

One of the most powerful tools in science is the method of reductive explanation, where we explain the behaviour of a complex system in terms of the behaviour of the parts from which it is composed. In order to employ this method we observe the behaviour of the parts in isolation and use this information to tell us how the parts will contribute the behaviour of the whole. Clearly then, the method assumes that something in what we learn when we observe the parts in isolation remains true when the parts are part of a larger whole. In particular, since all but the most trivial complex systems will involve causal interactions between their component parts, we must assume that something about these causal relations remains constant from one situation to another. Nancy Cartwright points out that we certainly donÇt assume that it is the behaviour of the parts that remains constant from one situation to another. Not only would such an assumption be false, it would trivialise the notion of reduction. Standard analyses of science try to avoid this kind of problem by turning their attention from actual behaviour to laws of nature, dispositions, and counterfactual truths. But we will see that these approaches cannot make sense of reductive explanation either: the facts they point to will either not remain constant from one situation to another, or else they will be useless in predicting the behaviour of complex systems. In response to this kind of problem Cartwright introduces the notion of a causal capacity and suggests that it is a systemÇs capacities that remain constant from one situation to another. However, I will argue that although Cartwright is on the right track, when it comes to understanding reductive explanation, her notion of capacities is no better off than the standard analyses it was set to replace. I will argue that what is assumed to be constant in reductive explanation are component causal influences. These influences sit somewhere between CartwrightÇs capacities on the one hand and the dispositions and counterfactuals that appear in standard analyses on the other. I finish by considering whether standard counterfactual, interventionist, or agency analyses of causation have the resources available to make sense of these component causal influences. If they do not, then it would seem that they cannot capture a notion of causation that is fundamental to scientific practice.