As part of a project to develop high throughput versions of the comet assay (single cell gel electrophoresis), with a consequent need for more efficient scoring, we have compared the performance of visual scoring, automated and semi-automated image analysis when assessing comets in the same set of gels from dose-response experiments with typical DNA-damaging agents. Human lymphoblastoid TK-6 cells were treated with concentrations of methylmethanesulphonate between 0.04 and 0.6 mM, and peripheral human lymphocytes were incubated, after embedding in agarose, with H(2)O(2) concentrations from 2.5 to 160 μM. All three scoring methods proved capable of detecting a significant level of damage at the lowest concentration of each agent. Visual scoring systematically overestimates low levels of damage compared with computerised image analysis; on the other hand, heavily damaged comets are less efficiently detected with image analysis. Overall, the degree of agreement between the scoring methods is within acceptable limits according to a Bland-Altman analysis.