Abstract: The Ramsey Test idea has it that one finds a conditional acceptable to the extent that one finds the consequent likely on imagining the antecedent. So why does “If you had cheated on the test, Paris would have been in France” sound so wrong?
This paper argues that the acceptability of counterfactuals demands not just that one finds the consequent likely on the counterfactual supposition of the antecedent, but also that there be some relevant connection between antecedent and consequent. Extant work concerning indicatives, due to Igor Douven, understands relevance as evidential support, in Bayesian terms. We argue that the counterpart for counterfactuals can be *causal-explanatory* support, and work out the logical details.