The DNA index, percentage of S-phase cells, proliferation fraction, and glutathione (GSH) content were determined at more than 1100 separate sites in 140 human tumors and 140 normal tissues. The study showed that the variability was so great from site to site within a tumor that there was only a 61% chance of identifying an aneuploid tumor clone (when present) if only a single site sample was analyzed for DNA content. Similar broad variability was observed in the percentage of S-phase cells, proliferation fraction, and glutathione content. Since these tumor characteristics are often used to predict the outcome of therapy and patient survival, the inaccuracy and underestimation of the test results may cause conflicting or erroneous predictions. The probability of finding an aneuploid clone or elevated percentage of S-phase cells proliferation fraction and GSH content increased dramatically as the number of sample sites studied per tumor was increased. Statistical analyses indicated that in order to achieve a 90% probability that the test results for these parameters were representative of the whole tumor: (a) all single site testing should be abandoned; (b) assays should be performed on samples taken from 3-7 different sites within each tumor; or (c) samples from each tumor should be pooled and the analyses run on a thoroughly mixed or homogenized aliquot of the multisite sample.