Results using the same drug in phase II studies of treatment in ovarian cancer vary widely. An analysis of five phase II studies with a total of 93 patients was carried out to determine whether factors other than the efficacy of the drug affect response. The drugs for the phase II studies were chosen on the basis of in vitro activity or previous activity in humans. Univariate analysis showed that several factors were of significance in predicting response. The most significant was interval from the end of previous treatment to entry into a phase II study. Others were the original presenting stage of the patient, the second line treatment given and the best previous response to therapy. In multivariate analysis, however, only two factors were shown to be of importance which were interval and the FIGO stage of the patient. Using these two variables the discriminant analysis predicted 89% of those who did not respond and 75% of those who did, with an overall correct prediction of 85%. The importance of interval is emphasised by the observation that the response rate for those patients who progressed on treatment or who relapsed within 3-6 months of primary therapy had a response rate of less than 10%. Future phase II studies should probably exclude patients in this category, since the chance of their responding is very low.