Four replicate lines of D. melanogaster, which had been selected for increased abdominal bristle number for 58 or 69 generations, were pedigreed for nine generations under selection with or without replacements for failed matings (SW and SO sublines) and under relaxed selection also with or without replacements (RW and RO sublines).Natural selection was effective in reducing mean bristle number in both R Wand RO sublines (except in one line), but its opposition to artificial selection in SW and SO sublines appeared to be only indirect. The relation between the two selective forces was inferred from their effects on effective population size, the comparison of selection responses in SW and SO sublines, and the difference between expected and realized selection differentials.Fertility was the most important factor affecting variance effective population size, while fecundity (with upper limits on the number of offspring scored) and artificial selection were in most sublines similar in their relative importance. Measured by reduction in effective population size, the inbreeding effect of artificial selection confined to the immediate generation was small (6·4%), but the cumulative effect estimated from the observed rate of inbreeding was quite large (16 %), tending to increase with more response.The spread of genes from initial families seems to have been influenced by both artificial and natural selection. Correlations between the mean score of an initial family and its genetic contribution indicate that additive genetic variance was still available at this stage of selection. The number of initial families represented in SW and SO sublines was generally large even after nine generations, and few families made unusually large contributions.