The toxicity of anticoagulant rodenticides to non-target species is one of the root concerns over wide-scale use of these compounds. Compared with the numerous studies documenting secondary exposure in predators, there have been relatively few studies on primary exposure in non-targets. We consider why primary exposure of non-targets occurs, which species are most likely to be exposed, how and why exposure magnitude varies, and whether exposure results in ecologically significant effects. Species groups or trophic guilds most at risk of primary exposure include invertebrates, reptiles, birds and mammals. Relatively little is known about exposure and particularly effects in invertebrates and reptiles although recent studies suggest that anticoagulants may impact invertebrates, presumably through different toxic pathways to those that result in vertebrate toxicity. Amongst higher vertebrates, primary exposure occurs in some bird species but there is little information on extent and importance. There are more studies on non-target mammals and it is granivorous species that are most likely to feed on bait and accumulate residues, as might be predicted given their ecological and trophic similarities to target species. However, studies suggest a surprisingly high degree of exposure in shrews, although it is unclear the extent to which this is primary and/or secondary. Overall, arguably the most striking aspect of primary exposure in mammals is the large-scale variation both in the proportion of animals exposed and the magnitude of residues accumulated. We consider the multiple abiotic and biotic factors that may drive this,including the direct and indirect effects of resistance in target species. In terms of ecologically significant effects, primary exposure clearly does cause acute mortalities in non-target vertebrates and these have been associated with significant population impacts on intensively baited islands where there has been limited or no potential for immigration. Localised population impacts have also been documented in mainland small mammals but most non-targets are likely to be r-selected species. Population declines may therefore be expected to be relatively short-term, provided baiting is episodic, as population numbers can recover through high intrinsic rate of reproduction in survivors, reduced density-dependent mortality, and immigration. However, prolonged or permanent baiting may potentially result in long-term depletion of resident non-target populations that is ameliorated only by immigration; such areas may act as population sinks.