A reanalysis was conducted of Saunier's research (1985) on sources of variation in the National Research Council (NRC) reputational ratings of university faculty. Saunier conducted a stepwise regression analysis using 12 predictor variables. Due to problems with multicollinearity and because of the atheoretical nature of stepwise regression, Saunier's conclusions were only speculative. By using LISREL, the present reanalysis demonstrated the value of regressing reputational ratings on three latent variables: (1) department size; (2) faculty research productivity; and (3) the quality of program graduates. The model was tested using NRC data for each of six disciplines: English, French, philosophy, geography, political science, and sociology. The relative magnitude of the contributions made by the three latent variables depended upon the discipline of interest, but generally size and research productivity were found to be more important than the quality of the program graduates. (Author/GDC)