Risk assessments for bovine spongiform encephalopathy (BSE) should be based on the group risk and not the median individual risk. The group risk is calculated from the arithmetic mean risk, which in the case of dorsal root ganglia, is a factor of 50-fold higher than the median. For environmental routes, the arithmetic mean exposure is sufficient for risk assessment, while for food-borne routes failure to accommodate the variation in exposures to individuals across the UK population could overestimate the group risk considerably. Ignoring prion destruction by cooking could overestimate the food-borne risks still further. The recent estimate for the arithmetic mean cow-to-man species barrier of 4000 does not take into accounts either of these factors and thus may be too high. Until evidence for a threshold dose is demonstrated, public health scientists should avoid assessing safety on the basis of a 'minimum infective dose'. The incubation period observed in cattle-feeding studies, when completed, would continue to increase with decreasing dose below the ID50 if there is a threshold or co-operative effect. The question is raised of whether fears over BSE in drinking water contributed to the spread of foot-and-mouth disease across the UK in 2001; a risk tradeoff.