Research investigations on teacher training effects have focused almost exclusively on differences between group means. The present paper suggests that several interesting and important research questions might be answered by examining student variability both within and between classrooms. Student variability might be considered as an outcome to be studied when teacher behavior affects heterogeneity or as a nuisance variable that should be controlled. A hypothetical data set is presented to demonstrate valid statistical methods which could be used in investigations of teacher training effects. (Author)