Noting that many language arts teachers have rejected tutorial software because of its inability to interact effectively with students, this paper explores implications of artificial intelligence research for human/computer interaction in reading instruction. Limitations of "exact match" curriculum designs in contemporary attempts to provide computerized interaction are surveyed and critiqued, and the importance of drill and practice to the development of language skills is emphasized. It is proposed that open-ended curriculum designs, which are process oriented, may help counter the limitations in developing adequate interactive programs. Finally, research questions in the development of artificial intelligence natural language systems are presented as central to the problem of"intelligent" human/computer interaction. The concluding comments assert that computers must be taught to simulate the same cognitive processes as occur in humans if meaningful interaction is to take place. (JD)