In this study, we modelled population responses to chronic external gamma radiation in 12 laboratory species (including aquatic and soil invertebrates, fish and terrestrial mammals). Our aim was to compare radiosensitivity between individual and population endpoints and to examine how internationally proposed benchmarks for environmental radioprotection protected species against various risks at the population level. To do so, we used population matrix models, combining life history and chronic radiotoxicity data (derived from laboratory experiments and described in the literature and the FRED ERICA database) to simulate changes in population endpoints (net reproductive rate R-0, asymptotic population growth rate lambda, equilibrium population size N-eq) for a range of dose rates. Elasticity analyses of models showed that population responses differed depending on the affected individual endpoint (juvenile or adult survival, delay in maturity or reduction in fecundity), the considered population endpoint (R-0, lambda or N-eq) and the life history of the studied species. Among population endpoints, net reproductive rate R-0 showed the lowest EDR10 (effective dose rate inducing 10% effect) in all species, with values ranging from 26 mu Gy h(-1) in the mouse Mus musculus to 38,000 mu Gy h(-1) in the fish Oryzias latipes. For several species, EDR10 for population endpoints were lower than the lowest EDR10 for individual endpoints. Various population level risks, differing in severity for the population, were investigated. Population extinction (predicted when radiation effects caused population growth rate lambda to decrease below 1, indicating that no population growth in the long term) was predicted for dose rates ranging from 2700 mu Gy h(-1) in fish to 12,000 mu Gy h(-1) in soil invertebrates. A milder risk, that population growth rate lambda will be reduced by 10% of the reduction causing extinction, was predicted for dose rates ranging from 24 mu Gy h(-1) in mammals to 1800 mu Gy h(-1) in soil invertebrates. These predictions suggested that proposed reference benchmarks from the literature for different taxonomic groups protected all simulated species against population extinction. A generic reference benchmark of 10 mu Gy h(-1) protected all simulated species against 10% of the effect causing population extinction. Finally, a risk of pseudo-extinction was predicted from 2.0 mu Gy h(-1) in mammals to 970 mu Gy h(-1) in soil invertebrates, representing a slight but statistically significant population decline, the importance of which remains to be evaluated in natural settings.