A study of 130 undergraduate students enrolled in a course on auiovisual techniques sought to determine whether heuristic or algorithmic computer-based problem solving training would be differentially effective for students varying in cognitive complexity as measured by the Educational Set Scale (ESS). The interaction was investigated between one trait variable (cognitive simplicity-complexity), two treatment variables (heuristic or algorithmic training and numerical list or multiple choice response mode) in a math problem task. Dependent measures of interest were method of solution, total time in problem, and, for the list response mode group, total number of steps in the problem. Results indicated that compatibility in degree of structure between problem solving instructions and response mode led to performance that was stable across the continuum of conceptual preference. (CMV)