In Los Angeles County, the age-adjusted incidence rate of colon cancer in men is almost 30% higher than that in women; however, in the descending and sigmoid colon, age-specific incidence rates for women are higher than those for men before age 55. Since menstrual and/or reproductive factors may be involved in producing this crossover in age-specific rates, they were examined in a population-based case-control study involving 327 white women with adenocarcinoma of the colon and age-, race- and neighbourhood-matched controls. After adjustment for other factors associated with colon cancer in this study (family history of large bowel cancer, total fat intake, calcium, weight and activity level), ever having been pregnant was protective (RR = 0.56, 95% CI = 0.33-0.97). For one to two pregnancies, the RR was 0.76 (CI = 0.42-1.37); for three or more pregnancies, the RR was 0.45 (CI = 0.25-0.81). However, the relationship between the number of pregnancies and colon cancer risk was actually U-shaped, with risk decreasing with successive pregnancies up to four and then increasing with additional pregnancies. The U-shaped relationship was present for incomplete as well as for full-term pregnancies and was more striking for cancers occurring in the distal (descending and sigmoid) than proximal (caecum to splenic flexure) colon. Risk was not related to age at menarche or use of exogenous oestrogens, but delayed natural menopause was weakly protective in the proximal but not distal colon. The crossover in incidence rates in the distal colon can be completely accounted for by the pregnancy effect. The U-shape of the pregnancy curve suggests the possibility of competing factors, some protective, especially after one or several pregnancies, and others conferring increasing risk with successive pregnancies, regardless of the pregnancy outcome.