This paper describes an analytical solution to a common problem in aptitude-treatment interaction (ATI). The researcher is often interested in the comparison between treatments for a large set of attributes. If this set is large, the chance of non-orthogonality is correspondingly large. In these circumstances the analysis and interpretation becomes equivocal. Using data from a study of contrasting teaching approaches in high school biology and thirteen attributes, a regression approach to the analysis is demonstrated. The redefinition of a new set of orthogonal variables is made on rational grounds rather than allowed to happen by default. (Author/BW)