In this paper, we consider articulatory processes in a connectionist model of reading aloud to account for effects of manner of articulation of the initial segment in a variety of tasks. We first describe experimental results showing how flexibility in articulation can completely eliminate the a priori acoustic latency difference between plosives and non-plosives in some tasks, and exaggerate this difference in other tasks. We then simulate an expanded version of the Connectionist Incremental Articulation Model that incorporates Stevens’ (1998) 3 phases involved in articulating a speech segment.