David Hume (1748) defended the regularity view on the grounds that unless causation was understood in this way, the very concept of causation would have no basis in sense experience, and therefore be rendered utterly meaningless. But Hume’s view suffers from two serious problems. First, it fails to distinguish regularities which are accidental (or even non-accidental!) from real cases of causation. Second, we can observe causation in cases of human action when we try to push, pull, or resist objects around us. Unless these “tryings” are illusions which make no real difference to the physical world, we have good reason (on pain of skepticism) to believe that some examples of causation are indeed observable.
David Lewis (1973) defended the counterfactual view according to which C causes E if and only if (in the nearest possible world) had C not occurred, E would not have occurred. But Lewis’ theory had three main drawbacks. First, Lewis’ way of measuring similarities between possible worlds yielded the wrong truth-values for counterfactual statements (e.g. in worlds containing small miracles). Second, he could not account for the direction of causation in simple worlds governed by classical Newtonian mechanics. Third, there are counterexamples that involve causation without counterfactual dependence, as in cases of preemption and overdetermination.
Therefore, in reply to both Lewis and Hume, I concluded that regularities and counterfactual relations can be evidence of causation, but they do not constitute what causation is.
 Admittedly, counterexamples to Lewis’ view based on overdetermination are not as convincing as those involving (late) trumping preemption.