In the present experiment, a complex task was analyzed into subtasks and then programmed. The main variable studied was the degree of similarity of problems given for practice on the subtasks. Problems were either heterogeneous or homogeneous. The expectation was that subjects who worked heterogeneous problems would perform the complex task better than subjects who worked homogeneous problems. The major finding of the study was that the heterogeneous-problems-group performed significantly better on the addition task than either the homogeneous-problems-group or the control group. A reason for this finding was suggested by an analysis of the error responses made by the groups. The heterogeneous-problems-group made fewer stereotyped errors than the other groups. Such a finding suggests that heterogeneous problems do not readily permit the stereotyping of error responses. The effect of the different practice problems was also found to be independent of the intellectual aptitudes of the subjects.