The manner in which the conditional independence graph of a multiway contingency table affects the fitting and interpretation of the Goodman association model and of correspondence analysis is considered. Estimation of row and column scores is presented through a framework that includes both models. (SLD)