There is a great deal of interest in studying the joint action between chemotherapeutic agents, particularly in screening compounds for therapeutic synergism. In this paper, we propose an approach to study therapeutic synergism using logit models when the data are binary. The key is to convert the definition of therapeutic synergism to its equivalence in terms of the model parameters. Thereby, evaluating therapeutic synergism amounts to examining parameter estimates in the most parsimonious model fit to the data. The proposed approach was further applied to an experimental situation where it was desirable to rank several test articles when administered jointly with a common compound in the treatment of shock.