Through a review of the literature, this paper explores the viability of the rotation of canonical correlation analysis results. The similarities and dissimilarities between factor analysis and canonical correlation analysis are examined. The logic supporting a preference for the rotation of structure coefficients as opposed to function coefficients is presented, along with some examples. The primary advantage of rotation appears to be an improvement in the interpretability of the results. A second advantage of rotation is that it can be used as an invariance procedure to reveal the common structure between samples. However, these are two distinct applications. Canonical analysis focuses on extracting orthogonal solutions that maximize the relationships between the two sets of variables. A small hypothetical data set is used to illustrate the results of different types of rotation using the ORSIM2 computer program. Concrete examples are used to show the following: (1) the sum of the canonical correlations after rotation is the same as before rotation; (2) the sum of squared multiple correlations after rotation is identical to the sum of pre-rotation squared canonical correlations; and (3) the total predictive power of each variable is unchanged by rotation. Six data tables are included. (Author/TJH)