Groups of middle-aged (30–50 years) and elderly (65–81 years) men and women were compared on three problem-solving tasks, including written problems, the 20-Questions procedure, and problems administered on a Heuristic Evaluation Problem Programmer. By including education, occupation, and nonverbal intelligence scores in a regression analysis, along with age, sex, and their interaction, it was determined that education and nonverbal intelligence explained the age and sex effects which had been found when an analysis of variance was performed on the data. The implications of the results are that multiple regression is preferable to analysis of variance for analyzing several correlated subject variables and that previously obtained age and sex differences may have resulted from the effect of correlated subject variables, such as nonverbal intelligence and education, rather than from the effect of age and sex directly.