Statistical analysis and multiple regression were used to determine and quantify the significant hygrothermomechanical (HGTM) variables which influence the tensile durability/life (cyclic loading, fatigue) of boron-fiber/epoxy-matrix (B/E) and high-modulus-fiber/epoxy-matrix (HMS/E) composites. The use of the multiple regression analysis reduced the variables from 15, assumed initially, to 6 or less with a probability of greater than 0.999. The reduced variables were used to derive predictive models for compression and intralaminar shear durability/life of B/E and HMS/E composites assuming isoparametric fatigue behavior. The predictive models were subsequently “generalized” to predict the durability/life of graphite-fiber/resin-matrix composites. The “generalized” model is of simple form, predicts conservative values compared with measured data, and should be adequate for use in preliminary designs.