We have developed a two dose/single-sample protocol for the in vivo/in vitro rat hepatocyte unscheduled DNA synthesis (UDS) assay. In order to enhance the effectiveness of grain detection by image analysis we found minor technical modifications to be beneficial. These included the use of 3% acetic acid in ethanol (or 4% aqueous paraformaldehyde) for hepatocyte fixation and 0.5% aqueous eosin for staining. Also, there was no correlation between cell viability (measured using the trypan blue method) and UDS response and, therefore, we do not reject animals from data analysis unless hepatocyte viability falls below 50%. Furthermore, we suggest that cell attachment is a more reliable indicator of toxicity than of cell viability. Therefore, our range-finding test has been modified to incorporate an extra animal per group so that hepatocyte cultures can be evaluated. Comparisons of a two-dose/single-sample protocol with the currently accepted single-dose/multiple-sample protocol demonstrated that the former was an acceptable alternative, in that responses with standard positive controls are very similar with both protocols. Furthermore, the two-dose protocol has clear advantages in that it uses fewer animals, resources and time and has better statistical discriminatory power. As a move away from the use of arbitrary values for determining the performance and outcome of assays, we use criteria based on confidence limits placed on historical data ranges. Where necessary, statistical analyses of concurrent data is performed using rank transformation followed by parametric analysis of the ranks; this combines the generality of a non-parametric methodology with the power and versatility of parametric analyses.