Enactivists often claim that since perception is one with action, it does not involve representations, hence perception is direct. Here we argue that empirical evidence on neural activity in the ventral premotor cortex confirms the enactivist intuitions about the unity of action and perception. But this very unity requires the detection of the action possibilities offered by the objects in the environment, which in turn involves certain representational processes at the neural level. Hence, the enactivist claim that perception is direct is wrong, or at least ambiguous and potentially misleading: in one important sense perception involves representations.