The central question of this project is: How are different sources of information which are retrieved from memory or provided by sensory input unified with language into an interpretation (comprehension) or message (production) beyond the single word level? Which neural networks are recruited for these unification operations? To what degree are these shared between production and comprehension, and what is the nature of their dynamic interplay with memory components, such as the mental lexicon or episodic memory of the prior discourse?