Predictors of novice learning in simulation environments were investigated in the domain of statistics. The first objective was to clarify the relation between intellectual ability and working method (e.g., orientation and systematical orderliness), and to determine the effects on learning of working method, independently of intellectual ability. A second objective was to determine whether instructional aid by presenting students with a well-structured task, instead of unguided learning by discovery, might compensate for lack of ability or a poor working method. Twenty-seven first-year university students of relatively high intelligence (n=14) or low intelligence (n=13) worked in either a structured or an unstructured learning environment. Thinking-aloud protocols were analyzed in terms of the quality of working method. Results indicate that the working method of high-intelligence subjects was significantly better than that of subjects of low intelligence; working method was also a strong predictor of learning, independent of intellectual ability. No learning effects due to the level of structure of the learning environment could be detected. Five data tables and a 28-item list of references are included. An appendix contains three definitions from the study. (Author/SLD)