Anecdotal reports of hereditary colorectal cancer suggest that genetic anticipation (earlier appearance in successive generations) occurs, but ascertainment bias and cohort effects confound this interpretation. Using approaches that correct for such biases, we examined the age at diagnosis of colorectal cancer from family history questionnaires completed by subjects in the Johns Hopkins Hereditary Colorectal Cancer Registry; 475 parent-offspring pairs in 308 pedigrees were studied. We observed the expected cohort effect among offspring, in that the mean ages at diagnosis of those born before 1921, between 1921 and 1930, and after 1930 were 63 +/- 13 (SD), 57 +/- 10, and 42 +/- 10 years, respectively, while their parents' mean ages were 65 +/- 14, 66 +/- 14, and 58 +/- 15 years, respectively. In the cohort born before 1921, in which observation periods for both parents and offspring were comparable, there was no difference in age at diagnosis by pairwise comparison or life table analysis (P = 0.15 and 0.23, respectively; r = 0.32). Subgroup analysis of 67 pairs from 38 families that met the International Collaborative Group (ICG) criteria for hereditary nonpolyposis colorectal cancer (HNPCC) and of 14 pairs from 7 families with known germline mutations of DNA mismatch repair genes also showed no significant differences (mean age at diagnosis: 56 +/- 14 years for parents and 57 +/- 16 years for offspring from ICG families, and 45 +/- 10 years for parents and 44 +/- 12 years for offspring in families with known mutations). We also found no evidence for effect of parental gender on age at diagnosis in offspring of either gender, nor a secular trend toward younger onset colorectal cancer in this sample. In conclusion, there is no evidence for genetic anticipation or genomic imprinting of age at diagnosis in this sample of colorectal cancer families. Apparent anticipation appears to reflect a birth cohort bias of ascertainment.