This paper reviews multivariate hypothesis testing, interval estimation, and discriminant analysis procedures for an aptitude-treatment interaction (ATI) analysis and illustrates the procedures with data from a large experimental study of science education at the secondary school level. The treatment variable of interest is time allowed for study with three controlled levels and the student aptitude is reading ability. The treatment and aptitude variables are shown to interact for four dependent variables representing achievement and attitude, and the interaction is described. The multivariate linear model was used for the analysis. Two advantages of the model are discussed. The inferential payoff is better control of Type I errors in hypothesis testing and confidence levels for interval estimation; and the potential for simplification of the ATI description is greater. (Author/CTM)