The literature on computer reliability is replete with very convincing arguments for the need and the use of self-repair techniques, as a viable approach to significantly enhancing the reliability of both maintainable and nonmaintainable computers. However, it would seem that no comprehensive and coherent program for the development and optimal employment of such techniques exists. This means that no method exists in the open literature for deciding the following: 1) what self-repair techniques, taken singularly or in combination, provide the greatest improvement in reliability; 2) what methods are optimum for initiating fault diagnosis and self-repair by redundancy and replacement; 3) what constitutes a closed set of self-repair techniques and what theory can be formulated to demonstrate the set's completeness; and 4) what is the effect of self-repair on the total system relative to design, maintenance, availability, and so forth.