Recent literature in the areas of reliability and maintainability evaluation and acceptance sampling has suggested the potential of trichotomous classifications of lot quality and reliability/maintenance levels as opposed to the usual dichotomous classification. Bray, Lyon, and Burr [1] have suggested a three-category approach to acceptance sampling. The purpose of this paper is to adapt this three-category procedure to the area of maintainability testing. The chief advantages of using maintainability test plans based on trichotomous classifications are that such plans do not require the assumption that the repair time distribution is log-normal as do the parametric plans described in MIL-STD-471A [2]. Furthermore, they utilize the data more efficiently than the distribution-free procedures given in the same publication.