During a three-year project (1977-1980) on computerized adaptive achievement testing, item characteristic curve theory (ICC) and adaptive testing strategies designed almost exclusively for ability testing were applied to achievement testing. Adaptive techniques substantially reduced test length without reducing quality, when applied to three problem areas: the use of unidimensional item pods, the measurement of multiple content areas, and mastery testing. The use of unidimensional ICC in achievement testing was further supported by the lack of effect on dimensionality of immediate knowledge of results. Because some results were inconclusive, more research is needed on inter-subtest branching, dimensionality of achievement over time in instruction, and adaptive mastery testing. Finally achievement measurement lacks a coherent framework. Norm referenced and criterion referenced testing approaches have little in common; and they do not relate to the crucial problem of measuring individual achievement gains during or after instruction. To incorporate inter-subtest branching, inter-time branching, and mastery testing, an adaptive self-referenced testing model is proposed. (CP)