A course in macroeconomics was used to investigate the cost effectiveness of computer-assisted instruction (CAI). The experimental design used an experimental section in which the students had the opportunity to use the computer to the intensity of their choice, choosing from tutorial lessons, games, and simulations. The students in the control section received their instruction by the traditional lecture-textbook method. Students were not randomly assigned to individual sections, but rather they selected a given section primarily on the basis of scheduled class times. Each instructor's experimental section, however, was randomly assigned. Results showed that under virtually all the conditions studied, gains in educational output did not outweigh the additional cost of CAI. Only when the costs of development of CAI materials was ignored did the CAI format appear marginally efficient. It was suggested that costs of CAI may be reduced when authors gain experience in writing CAI programs and that advances in computer technology may have an impact on CAI costs. (JY)