The relationship between Comptutational Lexicography and Computational Linguistics is examined. Section one is a general discussion of publishers' machine readable dictionaries, which points out their failure to adapt to users' changing needs. Section two discusses and evaluates attempts to process such dictionaries to obtain information useful for natural language processing, drawing a pessimistic conclusion. The third section investigates the current state of knowledge among Computational Linguists about machine dictionaries, which is found wanting. Section four highlights some of the problems for machine dictionary work caused by the adoption of modern techniques of system design. The last section discusses possibilities for future interaction between the two disciplines, and proposes a framework for common research.