Population genetic theory predicts that selectively driven changes of allele frequency for both beneficial and deleterious mutants reduce polymorphism at tightly linked sites. All else being equal, these reductions in polymorphism are expected to be greater when recombination rates are lower. Therefore, the empirical observation of a positive correlation between recombination rates and amounts of DNA polymorphism across the Drosophila melanogaster genome can be explained by two very different types of natural selection. Here, we evaluate alternative models of effects of selection on linked sites by comparison of X-linked and autosomal variation. We present polymorphism data from 40 genes distributed across chromosome arms X and 3R of Drosophila simulans, a sibling species of D. melanogaster. We find significantly less silent polymorphism in D. simulans on the X chromosome than on 3R, but no difference between arms for silent divergence between species. This pattern is incompatible with predictions from theoretical studies on the effect of negative selection on linked sites. We propose that some form of positive selection having greater effects on sex chromosomes than on autosomes is the better explanation for the D. simulans data.T he amount of nucleotide polymorphism in a given region of the Drosophila melanogaster genome is positively correlated with the regional recombination rate (1, 2), which varies several-fold from one part of the genome to another (e.g., ref.3). The absence of a detectable effect of varying recombination rates on amounts of nucleotide divergence between species (1, 2) rules out variation in neutral mutation rates (4) across the genome as the explanation for the polymorphism data from D. melanogaster. Two kinds of population genetic models, both of which invoke effects of natural selection on linked neutral sites, have dominated the debate as to the cause of these patterns. One model (5-7) invokes effects of selection for beneficial mutants on linked neutral sites (i.e., hitchhiking effects), whereas a second model (8-10) invokes effects of selection against deleterious mutants on linked neutral sites (i.e., background selection). Both models predict that selection reduces polymorphism at linked neutral sites and that the magnitude of this reduction is greater when recombination rates are lower. Finally, for both models there is no expectation that variation in recombination rates affects divergence between species at neutral sites (11). The qualitatively similar predictions of the two models have made it difficult to determine their relative importance. The background selection model can provide a reasonably good fit to the polymorphism data, given certain parameters of deleterious mutation rate and recombination rate in D. melanogaster (9, 10). Similarly, a simple hitchhiking model also fits the D. melanogaster polymorphism data quite well (7, 12). Attempts to distinguish hitchhiking effects from background selection based on the frequency spectrum of variation in regions of very low cr...