This study derives age-sex specific standardized mortality ratios (SMR's) among cancer (ca) patients with respect to population suicide rates. Suicide rates were derived for all 144,530 ca patients in the Connecticut Tumor Registry from 1940-1973. We also derived rates by number of years since diagnosis (y-s-dx) and by decade of dx. We hypothesized beforehand that suicide rates will (1) be higher in ca patients than the population; (2) be higher soon after dx than in later y-s-dx; (3) vary directly with spread of disease at dx, highest for metastatic, lowest for local. We address hypotheses (1) and (2) here. After collapsing age groups and y-s-dx in the observed 192 ca patient suicides, SMR's derived for age, sex and y-s-dx were tested with an exact Poisson distribution. Overall, SMR for males was significantly elevated (230, P less than 0.005), but not for females. In males, a test for trend in SMR's with y-s-dx showed a significant decline, supporting hypothesis (2) for males (chi2 = 4.0, df = 1, P less than 0.05). No trend was found for decade of dx. On combining decades, male SMR's were significantly elevated in 10 of the 25 cells of age by y-s-dx, while none of the female cells was significant. We conclude that both hypotheses were supported for males, and neither for females.