We compared the efficacy of artificial and natural selection processes in purging the genetic load of perpetually small populations. We subjected replicate lines of the housefly (Musca domestica L.), recently derived from the wild, to artificial selection for increased mating propensity (i.e., the proportion of male–female pairs initiating copulation within 30 min) in efforts to cull out the inbreeding depression effects of long-term small population size (as determined by a selection protocol for increased assortative mating). We also maintained parallel non-selection lines for assessing the spontaneous purge of genetic load due to inbreeding alone. We thus evaluated the fitness of artificially and ‘naturally’ purging populations held at census sizes of 40 individuals over the course of 18 generations. We found that the artificially selected lines had significant increases in mating propensity (up to 46% higher from the beginning of the protocol) followed by reversed selection responses back to the initial levels, resulting in non-significant heritabilities. Nevertheless, the ‘naturally’ selected lines had significantly lower fitness overall (a 28% reduction from the beginning of the protocol), although lower effective population sizes could have contributed to this effect. We conclude that artificial selection bolstered fitness, but only in the short-term, because the inadvertent fixation of extant genetic load later resulted in pleiotropic fitness declines. Still, the short-term advantage of the selection protocol likely contributed to the success of the speciation experiment since our recently-derived housefly populations are particularly vulnerable to inbreeding depression effects on mating behavior.