Predictive drug testing on human tumor cells in order to define the appropriate chemotherapy will remain imperative as long as the anticancer agents available are few in number and show only limited activity. The advantages of an effective test would lie in obviating the need for testing antineoplastic agents on large cohorts of patients for assessment of drug activity (phase II studies) and in allowing determination of optimal use of anticancer agents (phase III trials). Such an in vitro test could help to better define dose and schedule of drugs preclinically. The additive value of individual drugs could be determined on tumor cells in vitro in order to define the best combination chemotherapy in vivo. Test-directed therapy would avoid unnecessary drug-related morbidity in patients with refractory tumors. Chemotherapy treatment would be more than justified even with side effects if palliation or even prolonged survival could be anticipated as a result. The benefits of predictive drug testing on human tumor cells would extend beyond improvement of individual patient treatment if the testing helped to identify new active agents. This spectrum of benefits to the entire field of oncology pro­ vides tremendous motivation for the development of such testing. Although a number of chemosensitivity tests have been proposed since the advent of modern anticancer chemotherapy, interest has been renewed by the possibility of cloning human tumor cells on agar plates, with a view to testing drug activity on cells with high pro­ liferation capacity.