Models of spoken word recognition in the localist connectinist tradition appeal to different types of activation flow to characterize the lexical selectin process during wich a competitor set in activated and the target word is identified in this set. This study investigates the nature of activation flow by means of experiments and simulations with the objective of (1) distinguishing between the effects of bottom-up and lateral inhibition and (2) testing the effects of top-down word to phoneme activation flow upon word recognition. Our results support the concept of bottom-up inhibition and raise doubts about the utility of top-down feedback.