Two central claims of the theory of articulatory phonology are that articulatory gestures are phonological primitives, and that the main advantage of such a choice for primitives is that it unifies the physical and phonological levels of description. I argue that both claims are problematic. With respect to the choice of gestures as phonological primitives I discuss three problems: First, Browman and Goldstein argue that gestures can be used to signal lexical differences, but they do not argue that gestures are the best units for the job and hence do not provide an argument for treating gestures as the basic or primitive phonological units. Second, they do not explain the relative merits of gestures over segments in terms of this functional role, and they do not show how gestures and segments can be part of a single hierarchical particulate structure. Finally, even if we ignore these first two criticisms, the theory still faces a dilemma insofar as phonological gestures are said to be "defined" as invariant, whether or not they are so in fact. With respect to the second claim I argue that a "unification" between phonological and physical descriptions of speech of the sort articulatory phonology provides is not needed in order to show that the two levels are systematically related.