The purpose of this study was to develop a taxonomy which would categorize high level physics problem-solving behaviors, and to examine the usefulness of such a classification system. This classification of learning objectives is based on complexity, a nonarbitrary measure which does not rely upon comparison between students but rather is based on the number of computational steps required. An experiment was conducted to compare the results of criterion referenced items based on the classification scheme with problem difficulty. Significant differences indicated that performance on more complex problems could not be predicted on the basis of zero or one step problems. Included are samples of zero-step, one step and multiple step criterion test items. The computer programs used to analyze the data are also included. The project was carried out by the New York Institute of Technology and was funded by the U. S. Office of Education. (TS)