Abstract

One of the most powerful tools in science is the analytic method, whereby we seek to understand complex systems by studying simpler sub-systems from which the complex is composed. If this method is to be successful, something about the sub-systems must remain invariant as we move from the relatively isolated conditions in which we study them, to the complex conditions in which we want to put our knowledge to use. This paper asks what this invariant could be. The paper shows that the kinds of thing that a Humean might point to – behaviour, laws, and dispositions – cannot play the role required of the invariant in question. Nor, indeed, can non-Humean causal powers of the kind advocated by contemporary metaphysicians such as Ellis and Lierse. The paper suggests that the analytic method presupposes a kind of entity that does not appear in standard ontologies – a metaphysically substantial notion of causal influence. This notion of causal influence is one that Cartwright has also seen the need for, though she does not seem to take the notion as seriously as she should.