Some Dilemmas for an Account of Neural Representation: A Reply to Poldrack