We have used the effect of therapeutic agents on clonogenic growth in agar to discriminate between active and inactive agents for malignant melanoma. We report a prospective study of single-agent chemotherapy for metastatic melanoma. Forty-five separate in vitro/in vivo correlative trials were conducted in 34 patients. A number of agents were used in these evaluations, including actinomycin D, Amsacrine, bisantrene, mitoxantrone, BCNU, vinblastine, vindesine, 5-fluorouracil, MGBG, etoposide, interferon, tamoxifen, and 13-cis-retinoic acid. At the "cut-off" concentration, a colony survival less than 30% was designated as "sensitivity" and greater than 30% as "resistance." Clinical sensitivity was designated to include complete, partial, and mixed responses and was predicted in eight of 18 trials (44%). Clinical resistance (nonresponse) was predicted correctly in 24 of 27 cases (89%). Using Fisher's exact test the association of in vitro and in vivo results was significant (P = .05). These results offer further support for the concept that clonogenic assays may help select useful agents for clinical trials in metastatic melanoma.