The conditional analysis of ability statements has many versions. In this paper I will deal with the version which claims that ‘x can do y’ is equivalent to ‘if x were to choose to do y, then x would do y.’ However, my comments should be equally applicable to any analysis of ability statements that can properly be called a version of the conditional analysis. The intuition behind the conditional analysis is that what it is for one to be able to do something is for one's choice to be effective. To have an ability to do y is for it to be true that one's choosing to do y would be effective - one's choosing to do y would result in one's doing y . But this intuition is not captured by the conditional analysis in its standard form, and a restriction is needed to mend this defect. This restriction is based on a distinction among counterfactual conditionals.