Peto test procedures for the statistical evaluation of carcinogenicity studies require that each tumor in an animal that died intercurrently (or was sacrificed in extremis) be classified as either fatal, probably fatal, incidental, or probably incidental. There is considerable controversy as to whether or not the cause of death can be established with accuracy in rodent studies. In the present article, the causes of death or ill-being as found in 10 consecutive carcinogenicity studies--5 studies with 2400 OFA (Sprague-Dawley-derived) and Wistar rats and 5 studies with 2400 OF1 and NMRI mice--were re-examined. A cause of death or moribund state had been established in more than 80% of the cases in rats and in more than 70% in mice. These causes were, in rats, mainly pituitary tumors, chronic progressive nephropathy (males), mammary gland tumors (females), and subcutaneous tumors (males); in mice, mainly hemolymphoreticular tumors, lung tumors, liver tumors (males), and glomerulonephropathy. The criteria used for determining the tumorous or non-tumorous lesions as the cause of death were based on in-life and pathological findings. The validity of such procedures, the possibility of improving criteria in the future, and the usefulness of establishing causes of death in safety assessment are discussed.