Spoken Language Systems will require integration of continuous speech recognition and natural language processing. This is a proposed architecture for an interface between a continuous speech recognizer (CSR) and a natural language processor (NLP) to form a spoken language system. Both components are integrated with a stack controller and contribute to the search control. The architecture also allows a "Top-N" mode in which a "first part" outputs a list of top N scored sentences for post-processing by a "second part". An additional use for this architecture might be NLP evaluation testing: a common simulated CSR could be interfaced to each site's NLP to provide identical testing environments.