This morning in the graduate seminar we’ll be discussing Niall Ferugson’s Virtual History, which (per Andrew Gelman here and here) seems to me an agreeably rigorous thought experiment on the nature of causation in history. A cause is that x without which no y; to establish the causality of x one ought to be able to show that without it, no y, which means a comprehension of the counterfactual.
It strikes me as odd that some historians remain unhappy with this concept. Tristram Hunt suggested awhile back that this was because counterfactuals are inherently conservative; I think this is true only if you are wedded to a radical or progressive concept of history in which the end is foreordained. (See responses to Hunt here.) It is of course also possible that counterfactuals discomfit because, past a certain nearby point, there can be no evidence for them. Or perhaps for the same reason Gelman likes them—they are kin to a quantitative modeler’s mindset.