The number of cells within tumor colonies has not been determined accurately in prior reports because, in all but small clusters, cells grow too closely and stacked to allow direct counting of cells by inverted microscopy. Therefore, we stained colonies formed in agar from 38 tumor cell samples of diverse histological origin, removed them with a micropipet, and directly counted the number of cells. The number of cells within colonies increased geometrically with colony diameter and inversely with the size of the cells within the colonies. The relationship can be described using linear regression: [In(no. of cells/colony) = 0.87 - 2.80 In (colony cell diameter) + 2.38 In (colony diameter)] which gave an R2 of 0.92. Measurements of colonies in agar showed that they grew as oblate spheroids rather than spheroids. In an average sample, 60-micron-diameter colonies contained eight to 10 cells; the range for the 38 samples was 1.2 to 48.5. These results precisely define colonies in terms of cell number. They allow calculation of the total number of cells formed from clonogenic cells, a more complete estimate of proliferation than conventional cloning efficiencies which only measure initial proliferation. Furthermore, because of the dependence on the size of the colony cells, if colonies are defined only by a specific diameter then they do not contain similar numbers of cells. Calculations which assume spherical colonies, rather than the oblate spheroid shape we found, greatly overestimate the number of cells within colonies. Our data can increase the accuracy of quantitation in the clonogenic system and thus improve the interpretation of the proliferation of clonogenic tumor cells.