Extreme environments offer one of the best opportunities to study biotic adaptations because evolutionary responses to extremes are often clear and unambiguous. We studied two populations of mosquitofish ( Gambusia holbrooki; Poeciliidae) in South Carolina, one in a normothermic environment, and the other in a near-lethal, thermally elevated system heated by nuclear reactor effluents for 60-90 mosquitofish generations, which offered an opportunity to observe selection for increased thermal tolerance. We performed three experiments. First, we determined the critical thermal maximum (CTM) of each population and, as predicted, found the thermal population to have a higher CTM. We then placed fish from both populations in an acute thermal LD50 stress, and compared genetic diversity (heterozygosities) of fish that died and fish that survived the stress. Survivors had higher heterozygosities, indicating that genetic diversity may contribute to thermal tolerance. Finally, we used a half-sib/full-sib experimental design to estimate heritabilities for temperature at death of over 32%, indicating that selection has not depleted the population of genetic variation associated with thermal tolerance. Our results have implication for global warming studies because predicted increases in global temperatures must be met by the ability of populations to adapt to higher thermal regimes. At least in part, these adaptations come from selection on genetic variation for temperature tolerance within populations.