Three learning models (adaptive mastery, typical mastery, and traditional non-mastery learning models) which employed different criteria for terminating computer-based practice in order to determine mastery or non-mastery of arithmetic skills were compared. The efficiency of two different sequencing arrangements (mixed and clustered) of practice items was also examined. All treatments involved the teaching of basic arithmetic skills to seventh-grade students. The adaptive mastery learning model produced the same high level of performance on both the posttest and a delayed retention test as the other two models, but requires less time, fewer practice items, and minimized overpractice. No significant differences were found between the clustered and mixed item arrangements. (Author/SD)