If natural language question-answering (NLQA) systems are to be truly effective and useful, they must respond to queries cooperatively, recognizing and accommodating in their replies a questioner's goals, plans, and needs. Transcripts of natural dialogue demonstrate that cooperative responses typically combine several communicative acts: a question may be answered, a misconception identified, an alternative course of action described and justified. This project concerns the design of cooperative response generation systems, NLQA systems that are able to provide integrated cooperative responses. Two questions must be answered before a cooperative NLQA system can be built. First, what are the reasoning mechanisms that underlie cooperative response generation? In partial reply, I argue that plan evaluation is an important step in the process of selecting a cooperative response, and describe several tests that may usefully be applied to inferred plans. The second question is this: what is an appropriate architecture for cooperative NLQA (CNLQA) systems? I propose a four-level decomposition of the cooperative response generation process and then present a suitable CNLQA system architecture based on the blackboard model of problem solving.