Heterozygotes for pericentric inversions are expected to be semisterile because recombination in the inverted region produces aneuploid gametes. Newly arising pericentric inversions should therefore be quickly eliminated from populations by natural selection. The occasional polymorphism for such inversions and their fixation among closely related species have supported the idea that genetic drift in very small populations can overcome natural selection in the wild. We studied the effect of 7 second-chromosome and 30 third-chromosome pericentric inversions on the fertility of heterokaryotypic Drosophila melanogaster females. Surprisingly, fertility was not significantly reduced in many cases, even when the inversion was quite large. This lack of underdominance is almost certainly due to suppressed recombination in inversion heterozygotes, a phenomenon previously observed in Drosophila. In the large sample of third-chromosome inversions, the degree of underdominance depends far more on the position of breakpoints than on the inversion's length. Analysis of these positions shows that this chromosome has a pair of "sensitive sites" near cytological divisions 68 and 92: these sites appear to reduce recombination in a heterozygous inversion whose breakpoints are nearby. There may also be "sensitive sites" near divisions 31 and 49 on the second chromosome. Such sites may be important in initiating synapsis. Because many pericentric inversions do not reduce the fertility of heterozygotes, we conclude that the observed fixation or polymorphism of such rearrangements in nature does not imply genetic drift in very small populations.