The nature and appropriate application of the technique of multivariate analysis are discussed. More specifically, the intent of the paper is to demystify and explain the use of multivariate analysis as well as provide guidelines for selection of the most effective statistics for use in specific situations. For the purpose of this paper, the term multivariate indicates that there is more than one dependent variable that one is interested in analyzing simultaneously. The term univariate is used to mean any technique in which there is one dependent variable that is being analyzed. The relationship between multivariate analyses and canonical analysis is discussed, with specific reference to canonical correlation, multiple discriminant analysis, multivariate analysis of variance, multivariate analysis of covariance, and chi square analysis. It is demonstrated that multivariate techniques are subsets of canonical analysis and multiple regression analysis. The discussion of multiple regression analysis covers analysis of variance, analysis of covariance, discriminant analysis, chi square statistics, and t-tests. It is concluded that what frequently appears to be a multivariate analysis question may in fact be a univariate analysis question when one focuses on determining why the research is being conducted and how the results are likely to be used. It is this focus that should lead to the selection of a statistical model or technique. It is argued that considerations of the researcher's intent and the use to be made of the results are much more important than concerns about the Type I error rate. Moreover, univariate analyses are easier to interpret than multivariate analyses and, therefore, less prone to misinterpretation. (TJH)