A consecutive series of 123 colorectal cancers, prospectively followed for 3 years, was studied to determine if the prognostic significance of DNA ploidy related to: (a) thresholds used to define DNA aneuploidy, or (b) aneuploid sub-groups defined by DNA index (DI) and peak size. Aneuploidy was defined using 3 methods depending on the minimum proportion of nuclei considered to constitute an aneuploid peak; type 1, 5%; type 2, 10%; type 3, 10% if DI is 1.1-1.8, but 15% if DI is 1.9-2.1. DNA aneuploidy rates were type 1, 75%; type 2, 67%; type 3, 58%. The significance of clinical and pathological correlations varied with the use of different methods. All were associated with a significant DNA diploid survival advantage which was strongest for type 3 (p = 0.006). DI was unrelated to survival irrespective of the presence or absence of an associated S/G2; 33% with a DI of 1.1-1.8 and 35% with a DI of 1.9-2.1 survived. Prognosis was inversely proportional to aneuploid peak size, 48% with small peaks (less than 20%), 30% with intermediate peaks (greater than 20% less than 40%), but none with large peaks (greater than 40%) survived (p = 0.03). We conclude that: (a) thresholds used to define DNA aneuploidy affect the prognostic significance of DNA ploidy; (b) survival is independent of the DI of aneuploid peaks; and (c) measurement of aneuploid peak size refines the prognostic value of DNA ploidy.