Artificial selection for increased life span in experimental populations of Drosophila melanogaster sometimes produces long-lived populations that exhibit greater fecundity than unselected controls. The absence of a trade-off between survival and reproduction in these cases might be an artefact of the rich diet of typical lab culture; if nutritional resources are not limiting then there may be no need to trade off. Here I test the rich diet hypothesis by estimating genetic correlations between survival and age-specific fecundity in three nutritional environments. Experimental material consists of 58 recombinant inbred lines derived from an artificial selection experiment. Reducing the yeast content of medium causes substantial reductions in fecundity but does not alter patterns of genetic correlation. The correlation between life span and early fecundity is non-significant in all environments, while the life span correlations with mid-life fecundity are positive and statistically significant in all environments. The rich diet hypothesis is rejected. Qualitative features of fecundity trajectories are conserved across environments, with long-lived lines exhibiting a secondary peak of oviposition in mid-life. The micro-evolution of extended life span is not a monolithic process and does not necessarily involve direct trade-offs between survival and reproduction.