To date, studies on the mechanism of toxicity of pesticides are not yet an integral part of the toxicological evaluation process. However, in recent years mechanistic studies have played an increasing role in the assessment of toxicological hazards to man, and in this paper we have described two examples where an understanding of mechanism has contributed positively to risk assessment or has provided a surer scientific basis for the judgement of whether a potential hazard will be expressed in man. In the first example, an evaluation of the scientific literature leads to the conclusion that hepatic peroxisome proliferation in rats and mice is directly, or indirectly, related to the development of hepatocellular tumours. A wide range of non‐mutagenic chemicals elicit peroxisome proliferation in mouse and rat liver, but not the guinea pig or marmoset liver. Using one of the diphenyl ether herbicides, fomesafen, we have shown that isolated hepatocytes from mice and rats, but not those from guinea pigs, marmosets and significantly man, undergo peroxisome proliferation. Therefore, it seems reasonable to conclude that although fomesafen causes peroxisome‐related tumours in the mouse, man is neither susceptible nor sensitive to this mechanism. Consequently, we can conclude that fomesafen will not cause liver tumours in humans exposed to this herbicide. The herbicide paraquat, although safe in normal agricultural use, has been responsible for numerous human fatalities, almost exclusively as a result of the intentional ingestion of the concentrated commercial product. In this, the second example, studies on the mechanism of paraquat toxicity have provided a rational scientific basis for the evaluation of putative treatments and provide a framework for the development of novel approaches to reduce its toxicity. Also, a study of its mechanism of toxicity revealed that paraquat was selectively accumulated into the lung (in comparison with other tissues) by a previously unknown polyamine transport process. It was possible to determine the apparent kinetic constants for this uptake and use these pharmacokinetic parameters to predict that spray operators exposed to paraquat would not be at risk of lung injury. This scientific argument reinforced the epidemiological evidence that spray operators were clinically indistinguishable from an appropriate control population. In conclusion, mechanistic studies can contribute to the weight of evidence that is required in order to make scientific judgements on the hazards and risks that an individual pesticide will pose to man.