I argue that the intellectualist account of knowledge-how, according to which agents have the knowledge-how to (upvarphi ) in virtue of standing in an appropriate relation to a proposition, is only half right. On the composition view defended here, knowledge-how at least typically requires both propositional knowledge and motor representations. Motor representations are not mere dispositions to behavior (so the older dispositionalist view isn’t even half right) because they have representational content, and they play a central role in realizing the intelligence in knowledge-how. But since motor representations are not propositional, propositional knowledge is not sufficient for knowledge-how.