PETER MEDAWAR proposed that senescence arises from an age-related decline in the force of selection, which allows late-acting deleterious mutations to accumulate. Subsequent workers have suggested that mutation accumulation could produce an age-related increase in additive genetic variance (VA) for fitness traits, as recently found in Drosophila melanogaster. Here we report results from a genetic analysis of mortality in 65,134 D. melanogaster. Additive genetic variance for female mortality rates increases from 0.007 in the first week of life to 0.325 by the third week, and then declines to 0.002 by the seventh week. Males show a similar pattern, though total variance is lower than in females. In contrast to a predicted divergence in mortality curves, mortality curves of different genotypes are roughly parallel. Using a three-parameter model, we find significant VA for the slope and constant term of the curve describing age-specific mortality rates, and also for the rate at which mortality decelerates late in life. These results fail to support a prediction derived from MEDAWAR's "mutation accumulation" theory for the evolution of senescence. However, our results could be consistent with alternative interpretations of evolutionary models of aging.