The interrelationship of computer assisted instruction and individual differences was investigated. Two hypotheses were tested: (1) Bloom's hypothesis of decreasing individual differences under a mastery-learning strategy, and (2) the random walk interpretation of simplicial correlation matrices. Stanford's Computer-Assisted Instruction Math Program was run with 723 first to sixth graders in three States. Neither of the two hypotheses was supported by the data. For the first hypothesis a compensation effect of slow and fast learner variance components as well as an aptitude treatment interaction interpretation are discussed. The second hypothesis has been restated as a restricted random walk hypothesis. (Author)