Age-specific mortality rates were studied at two adult density levels in four inbred lines of Drosophila melanogaster. In experimental populations, adult densities were maintained at constant levels throughout the experiment by replacing dead flies with live, marked mutants. In control populations, densities declined naturally as the cohorts aged. For all experimental populations the best mortality model is the two-stage Gompertz model, with slower mortality acceleration at older ages. Flies in the experimental populations generally lived longer than flies in control populations, regardless of sex, genotype, or initial density level. The data demonstrate that deceleration of age-specific mortality rates at older ages is not caused by declining cohort densities. Mortality deceleration is a real phenomenon that raises serious questions about the evolution of senescence.