In this paper it is shown how a generalized form of canonical analysis can be useful to reveal which parametric functions of a MANOVA model, for instance treatment contrasts or combinations of observed variables, are responsible for rejection of a general linear hypothesis on these functions....