Although theoretical models often assume that the basic organization of the nervous system involves separate systems for perception, cognition, and action, neural data often does not fit into any of these conceptual categories. Here, an alternative framework is described, which focuses on interactive behavior and treats it as a continuous competition between representations of currently available potential actions. This suggests a neural organization consisting of two parallel systems: a system for action specification, which uses sensory information to represent currently available potential actions, and a system for action selection, which involves attentional and decisional mechanisms which determine the action that will be performed. It is proposed that neural processing occurs through two waves of activation: an early wave which specifies several potential actions and a later wave of biasing influences which selects one action for execution. A computational model of decision making is described within the context of this proposal, and simulations of neural and behavioral phenomena are presented.