Dissociated cancer cells are exposed to antineoplastic drugs (5 X 10(4) viable cells/drug for 1 hr or continuously) and cultured for 4 to 6 days in liquid medium. Cells are then stained with Fast green dye, sedimented onto slides with a Cytospin centrifuge, and counterstained with a modified hematoxylin and eosin technique. Dead cells stain with Fast green, and living cells stain with hematoxylin and eosin. Cell kill is calculated as percentage of control based on the relative numbers of living tumor cells, living non-tumor cells, and dead cells. Drug sensitivity could be assayed in 125 of 162 specimens of human neoplasms obtained from malignant effusions (16 of 18), excisional biopsies (31 of 44), needle biopsies (34 of 47), endoscopic biopsies (18 of 23), peripheral blood samples (19 of 20), and bone marrow aspirates (five of seven). The assays were successful (median of ten drugs tested) in: 46 of 64 adenocarcinomas; four of 11 squamous cell carcinomas; five of seven lymphomas; six of seven melanomas; two of four sarcomas; 18 of 20 transitional-cell carcinomas; 14 of 15 small-cell carcinomas; seven of eight myelomas; 12 of 12 chronic lymphocytic leukemias; seven of nine acute leukemias, and four of five "undifferentiated" carcinomas. The assay results demonstrated a strong correlation between the in vitro chemosensitivity of different types of tumors and the known clinical response patterns of these tumors. This assay can be used to determine which specific cells are killed in a heterogeneous cell population. Further work is needed to determine if this assay may be useful in blind screening trials for antineoplastic agents or if it may be of clinical use in predicting response to agents which are not cycle specific.