In this paper I shall give reasons for rejecting one type of analysis of the basic constituents of action, and reasons for preferring an alternative approach. I shall discuss the concept of basic action recently presented by Alvin Goldman, who gives an interesting version of the sort of analysis I wish to reject. Goldman agrees with Danto that bodily movements are basic actions, and his definition of ‘basic’ resembles Danto's fairly closely. What is new is a useful concept of level-generation between actions, which Goldman uses both in his recursive definition of action (45) and in his definition of a basic action (67, 72), as one whose performance does not depend on level-generational knowledge. In brief, an action is an event which is level-generated by or capable of level-generating another action, and a basic action is one which is not level-generated by any other action. I shall examine this concept of level-generation, and point out incoherences I think endemic to views of this sort. In the last part of the paper I shall indicate the direction in which a more satisfactory account of basic action is to be sought. The criterion of basicness I shall sketch will select as basic actions not bodily movements, but a more interesting class of actions, and one whose demarcation can help us see the relation between actions and intentions, and the differences between intentions and other states of mind.