What a multivariate contrast analysis should entail and how such an analysis should be conducted are discussed. The three questions encompassing a contrast analysis are: (1) whether the contrast effects are generalizable; (2) which variables contribute most to the contrast effects; and (3) what is the construct underlying the contrast effects. Three sets of results are needed for a contrast analysis using computer packages: test statistic and P value; outcome variable ordering; and underlying construct definition. Some helpful statistical package programs are discussed, including: BMDP, SAS, and SPSS-X-DISCRIMINANT. Two real data sets illustrate the use of multivariate contrast analyses. In one of these, a one-factor data set with three levels, group sizes were 35, 81, and 37. The other was a two-factor data set with four outcome variables and 18 observation vectors. Conducting a partial pairwise contrast analysis when the data do not satisfy the equal covariance matrices condition is discussed. Seven data tables are included. (SLD)