Five models are introduced for the factor analytic treatment of a set of measures obtained for the same sample of persons on two different occasions. The models differ in terms of the assumptions made regarding the constancy of the (1) factor (actually component) score and (2) factor pattern matrices from occasion 1 to 2. Least-squares procedures are developed for the estimation of the component scores and patterns under four of the models; canonical correlation procedures are developed for the fifth. Illustrative examples using these procedures are presented, and the research implications of the hypotheses and procedures underlying each model are discussed. (Author)