Growth rates of neoplasms could be calculated only on the basis of mitotic and apoptotic indices (MI and AI, respectively), assessed on tissue sections, if the duration of mitosis and apoptosis (Tm and Ta, respectively) in vivo were known. For humans, this is practically never the case. What use then can be made of MI and AI to arrive at a relative, crude estimate of the state of growth? As a model system to study this problem, we chose diffusely growing stage I + II non-Hodgkin's lymphomas (dNHL, n = 94). Cluster analysis revealed the existence of 3 highly distinct groups of dNHL (clusters I, II and III) in the MI vs. AI per case plot, with a roughly linear relation between both parameters. Most nosologic entities defined by the REAL classification comprise cases that were represented in more than one cluster. We adopted the simple formula GI (growth index) = XMI - AI, where X (= Ta/Tm) remains to be evaluated. Based on the assumption that spontaneous regressions of dNHL are rare but do occur, we estimated that X = 2 or, possibly, 3 are best fits for the pooled dNHLs studied. With the assumption of X = 2, (i) 2MI - AI gave relatively lower values for dNHL than proliferative indices such as %Ki-67+ cells; (ii) values for 2MI/AI per cluster showed a pattern inverse to that for %bcl-2+ cells; and (iii) a plot of 2MI - AI vs. 2MI/AI per case allowed the recognition, especially among NHLs with a low cell turnover, of cases where accumulation of presumably longer-lived cells is an important factor in determining growth.