Although computerized instructional materials for foreign language instruction have begun to become available, computer assisted instruction (CAI) in this field has not lived up to its promise for four reasons: poorly designed materials, incompatible machines, incompatibility of machines with textbooks, and equipment costs. Real CAI in classical language instruction will appear when the profession: (1) draws on colleges and universities for the best-designed materials, (2) uses innovative techniques toward sensible ends, (3) develops and tests materials on mainframe computers and then converts them to microcomputer form, and (4) does not ignore the centrality of the textbook in instruction. Two CAI projects illustrate the importance of these principles: the Montevidisco Project at Brigham Young University--a videodisc simulation of a visit to a Mexican town, and the University of Delaware's Latin Skills Project using the computer to inflect variable parts of speech in Latin. The latter not only incorporates all the desired development features but also provides a greatly increased number of exercises without corresponding additional memory need, gives more sophisticated judging of student answers and feedback on partially correct answers, adds the ability to review missed verb forms "invisibly," by presenting a similar form several items after the missed one, and enhances flexibility without loss of individualization. (MSE)