This retrospective study was conducted to evaluate the relationship between the mood-stabilizing agents lithium, carbamazepine, and divalproex sodium and time course of drug response as measured by length of hospital stay.Seventy-eight patients were evaluated. Discharge date and length of hospitalization were obtained by review of the dictated discharge summary by investigators who were blind to medication status. Drug choice, including blood levels, was then identified as well as any adjunctive treatment. Statistical evaluation (SAS software) of time to improvement was conducted using nonparametric survival models. Possible confounding variables were assessed by parametric survival regression models and proportional hazards models.The individual survival curves of four major antimanic treatment groups revealed a significant difference by the second week of hospitalization in the divalproex sodium and the lithium/carbamazepine combination groups as compared with the lithium or carbamazepine groups (chi 2 = 13.83, df = 3, p = .003). The mean +/- SD length of stay was approximately 40% shorter for the divalproex sodium (10.2 +/- 2.0 days) and the lithium/carbamazepine combination (11.7 +/- 2.1 days) groups compared with the lithium alone (17.6 +/- 1.0 days) and the carbamazepine alone (18.1 +/- 3.0 days) groups. Regression analysis of possible confounding variables including adjunctive treatment (benzodiazepines and neuroleptics), severity of illness (number of years ill and previous hospitalizations), and discharge date revealed no statistically significant effect.Shortcomings of our study include those inherent in the design of a non-randomized, small naturalistic study and retrospective chart review. Our study does, however, suggest different drug response rates for the treatment of classic and dysphoric mania and warrants further controlled investigation.