Causal explanation is tricky. I have in mind what I think is the Aristotelian distinction between Formal Cause and Efficient Cause. Of course I interpret these from the standpoint of a soggy philosopher/scientist in 2009, so Aristotle would probably not concur. Efficient Cause is closest to folk usage of the word. A causes B iff A precedes B, A necessarily gives rise to B, etc. Let us not get into necessary and sufficeint conditions, but suffice it to acknowledge that often, this kind of causation is what we mean by ’cause’. Formal Cause is completely different. A full dynamical description of, say, a pendulum, that captures essential properties of the system (typically position of the bob over time) is a formal causative account, but it does not contain any notion of efficient cause. With a full dynamical description, we have determinism, but, critically, you can run time backwards or forwards, and the lawfulness remains the same.

Now fans of efficient cause like to distinguish between explanation and mere description. Formal cause, as I interpret it here, is description. But I claim that it is as full a causal story as one can have. The notion of mere description sugests that description is theory neutral. It is not. In providing a description of the movement of the pendulum over time, I have selected some properties of the world as relevant (idealized bob position) and some as irrelevant (pendulum colour). Description is never theory neutral. Efficient cause fans seem to suggest that a causative account takes a theory neutral description and improves on it by adding causal relations within a theory. I would argue that one can do no better than describe, and that description is the building of a theory.

Most of the time, the desiderata of a full formal dynamical description are not available to us. System boundaries are unclear, knowledge is incomplete, the objects of our attention are subject to too many uncontrolled influences, etc. That is just the everyday messy world of observational science. Efficient causal stories try and limit this complexity by positing hypothetical idealized entities A and B, which stand in causal relation.

By preferring formal causal accounts, we are acknowledging that any selection of A and B is the imposition of a theory, and full description is the best we can do. Efficient causal stories are just that: stories. They need langauge for expression (which formal causal accounts do not: they are captured by equations). Language gives us causal stories. They are important in coming to grips with the world in a common sense way. But they are not privileged. Nor do they trump formal causal accounts.