A model of human linguistic performance for computer implementation, limited to the input-output interactions and internal processes available on today's digital computers. The model is to understand factual information input to it, answer questions, and seek information by asking questions. It need only learn the morphological and syntactic rules, rules of conceptual inference, and semantic correspondences for whatever natural language subset and database the user trains it to handle. Its "education" is by input of relational graph-structure representations of conceptual data structures. Basic to the model is the view of language as a continuous learning process, with semantic and syntactic information treated similarly as different phases of processing, in contrast to traditional views. The assumed stages are those outlined in RM-5210-PR. 43 pp. Ref.