The difficulty of extrapolating from inbred mouse strains used for the assessment of toxicological responses to the human population that may be exposed to the potentially toxic substance is discussed. Conventional risk estimation procedures assume little heterogeneity among members of the population in their response to a substance. This ignores the inherent genetic diversity present in human populations. Measurement of enzymic variation reveals that humans are more variable genetically than most mammalian species. Common genetic variants abound. Genetic variability within or between inbred strains of mice is much less. Evaluation of toxicological hazards should also consider responses to potentially toxic substances by individuals who carry rare genetic alleles. The effect of these rare variants on viability is currently unknown. The manner in which levels of genetic variability are affected by population structure is also discussed. In contemporary human populations, migration rates are not sufficiently restricted, nor are inbreeding and phenotypic assortative mating rates sufficiently large, to play important roles in limiting genetic diversity. Hidden genetic variability in human populations could be enormous, leading to much wider responsiveness to the toxic effects of environmental chemicals than presently appreciated. There exists a potential threat that drugs or food additives may be judged harmless when tested in animals less genetically diverse than humans, when they may, in fact, have serious toxic effects in. a nontrivial number of humans, because of the great genetic diversity present.