Several previous attempts at selecting for increased life span with Drosophila have failed to obtain a response to selection, and postulate that life span is controlled by non-genetic maternal effects instead of genes. In other experiments, however, populations have responded. This study uses a set of true-breeding long- and short-lived stocks developed by in vitro selection to examine the effect on life span of developmental environment and outcrossing. The expression of genes for life span is shown here to vary as a result of a gene-environment interaction and is strongly affected by the environment during development. The longevity of F1 crosses between long- and short-lived parental lines is additive when the density of larvae is high and unfixed, showing that life span is controlled by genes and not nongenetic maternal effects. But, when numbers of larvae are held low and constant, as in studies where selection fails, life span of crosses and parental lines is greatly restricted. The failure of some previous attempts at selection, therefore, appears to have resulted from the introduction, through experimental design, of strong artifactual environmental effects that limit the phenotypic expression of genes for life span and the effectiveness of selection. Comparison of fecundity in parental and F1 lines shows that selection for increased life span antagonistically reduces early-life fecundity. Short-lived, parental lines, reproduced at an early age in life, lay 22-24 per cent more eggs in the same period than do long-lived lines selected for reproduction at a late age.