Two computer-assisted instruction programs were written in electronics and trigonometry to test the Wollmer Markov Model for optimizing hierarchial learning; calibration samples totalling 110 students completed these programs. Since the model postulated that transfer effects would be a function of the amount of practice, half of the students were required to complete one practice problem successfully before moving on to the next stage; the other half had to do two practice problems successfully. All students completed the courses successfully; students who had one success at each stage did about as well as those who had two successes. The Wollmer was thus not suitable for optimizing instruction, in terms of minimizing overall time, in the particular courses. Perhaps the main reason for this result was that, as the student works up to the top of the hierarchy, the sheer number of subskills involved in the final task becomes a major determinant of the practice time, and the number of practice trials has a relatively minor effect, unless a large number of practice trials are given. (EMH)