Although the rodent comet assay is gaining acceptance as a standard technique for evaluating DNA damage in vivo, there is no internationally accepted guideline for its conduct and several aspects of its experimental design have not been optimized. For example, no standard positive control is used, there is no agreement on how tissue toxicity should be measured and sources of experimental variability have not been considered in relation to experimental design. This study showed that methylnitrosourea is a good alternative positive control inducing DNA damage in all tissues examined (stomach, liver, blood and bone marrow) over a dose range of 25-100 mg/kg at both 3 and 24 h after treatment. At the highest dose, significant toxicity was seen in all tissues using the neutral diffusion assay and also by histopathological/haematological analysis, except in the liver where no change was seen even 7 days after dosing. Analyses using control data pooled from several studies showed that, as expected, the greatest variability was seen between tissue preparations from different animals and that different numbers of animals were required to detect the same fold increases in different tissues. Power analyses showed that, preparing three gels for each tissue and scoring 50 nuclei per gel, a group of six animals allows 2-fold increases over control in the liver, bone marrow and stomach and a 3-fold increase in blood to be detected with 80% probability. It is recommended that similar investigations of experimental variability should be performed to determine optimal experimental design in any laboratory using the rodent comet assay.