The value in using the computer as a teaching machine is likely to be a function of how well it can adapt to individual differences in performance. The term "adaptive" is used to imply that the student is routed to material which follows different teaching strategies and that tasks are put together, on-line, to fit a particular student's competence. The components of an adaptive instructional system are, representations of tasks and student performances, and a set of teaching operations which are controlled by means-ends guidance rules.
Two examples which illustrate these components are discussed. The first, set in junior mathematics, uses generative material and adjusts task difficulty, type of feedback and remedial teaching to suit individual performances. Experiments which were necessary to establish the means-ends guidance rules are outlined. The second example, the planning of investigations in applied statistics and in physical chemistry, concerns the teaching of undergraduates. The material is pre-stored and the difficulties in designing adaptive decision rules in these contexts are indicated. However, experiments show the need for such rules. Finally, the integration of this teaching with more conventional methods is briefly outlined.