Theoretical analyses of inbreeding suggest that following an increased degree of inbreeding there may be a temporary recovery of fitness, because of selection either within or among inbred lineages. This is possible because selection can act more efficiently to remove deleterious alleles given the greater homozygosity of such populations. If common, recovery of fitness following inbreeding may be important for understanding some evolutionary processes and for management strategies of remnant populations, yet empirical evidence for such recovery in animals is scant. Here we describe the effects of single-pair population bottlenecks on a measure of fitness in Drosophila melanogaster. We compared a large number of families from each of 52 inbred lines with many families from the outbred population from which the inbred lineages were derived. Measures were made at the third and the 20th generations after the bottleneck. In both generations there was, on average, substantial inbreeding depression together with a highly significant variance among the inbred lines in the amount of fitness reduction. The average fitness of inbred lines was correlated across generations. Our data provide evidence for the possibility of recovery of fitness at two levels, because (i) the average fitness reduction in the F20 generation was significantly less than in the F3 generation, which implies that selection within lines has occurred, and (ii) the large variance in inbreeding depression among inbred lines implies that selection among them is possible. The high variance in inbreeding depression among replicate lines implies that modes of evolution which require a low level of inbreeding depression can function at least in a fraction of inbred populations within a species and that results from studies with low levels of replication should be treated with caution.