Five-year survival data were obtained in 97 percent or 1105 of 1140 new patients with histologically confirmed colorectal adenocarcinoma during a 12-month period in 1981 and 1982, as part of a large comprehensive population-based study of colorectal cancer incidence, etiology, and survival, The Melbourne Colorectal Cancer Study. Fifteen percent of patients were Dukes' A stage, 32 percent were Dukes' B, 25 percent were Dukes' C, and 29 percent were Dukes' D. At five years after diagnosis, the observed survival rate was 36 percent and the adjusted rate was 42 percent. Dukes' staging was a highly discriminating factor in survival (P less than 0.001). Survival rates were better in women than in men and better for patients with colon cancer than for patients with rectal cancer. Survival by Dukes' staging was not affected by colon subsite or by the tumor being the first and single tumor, metachronous tumor, or synchronous tumor. The survival of younger patients was better for Dukes' stages A, B, and C, and worse for Dukes' D. Survival was worse in the presence of bowel perforation in Dukes' C and D stages. Within Dukes' D (incurable cases), survival was best in the absence of hepatic metastases, slightly worse when only hepatic metastases were present, and poorest in the presence of both hepatic and extrahepatic metastases. Statistical modeling of survival determinants other than staging indicated that cell differentiation had the largest effect (survival decreasing with poor cell differentiation), followed by site (survival worse for rectal cancer than colon cancer), then age (survival better for younger patients), while bowel perforation had the smallest effect on survival.