Testicular cancer incidence is increasing rapidly in several countries. Environmental causes acting early in life are suspected but have not yet been identified. We conducted a cohort study to identify parental risk factors for testicular cancer among farmers' sons. Children born in 1952-1991 to parents who were farm holders at the time of the agricultural censuses in 1969-1989 were identified in the Central Population Register (Oslo, Norway). The resulting cohort of male offspring (n = 166,291) were followed up in the Cancer Registry of Norway (Oslo, Norway) for 1965-1991. Exposure indicators were derived from census information on activities on the farm. The cancer incidence was compared with that of the total rural population, and potential risk factors were analyzed by Poisson regression. In a follow-up of 2,924,663 person-years, 158 incident cases of testicular cancer were identified. The study population had a higher incidence of testicular cancer than did the total rural population, particularly at ages 15-19 years and in western Norway. Specific fertilizer regimens on the farm were associated with testicular cancer (rate ratio = 2.44; 95% confidence interval = 1.66-3.56), in particular nonseminoma (rate ratio = 4.21; 95% confidence interval = 2.13-8.32). The rate ratio estimates were highest for boys ages 15-19 years and for a subset of study subjects who were considered more likely to have grown up on a farm. Nondifferential misclassification and bias toward unity are likely because exposure information was available only at the farm level and only for census years. The fertilizer indicators were not available early in life for most subjects, and precise interpretations are difficult. A hypothesis worth considering is that excess nutrient run-off from agriculture constitutes a risk. However, inferences concerning the biological basis of our observations can scarcely be made.