Abstract. -Previous attempts to model the joint action of selection and mutation in finite populations have treated population size as being independent of the mutation load. However, the accumulation of deleterious mutations is expected to cause a gradual reduction in population size. Consequently, in small populations random genetic drift will progressively overpower selection making it easier to fix future mutations. This synergistic interaction, which we refer to as a mutational melt-down, ultimately leads to population extinction. For many conditions, the coefficient of variation of extinction time is less than 0.1, and for species that reproduce by binary fission, the expected extinction time is quite insensitive to population carrying capacity. These results are consistent with observations that many cultures of ciliated protozoans and vertebrate fibroblasts have characteristic extinction times. The model also predicts that c10nallineages are unlikely to survive more than 10 4 to 10' generations, which is consistent with existing data on parthenogenetic animals. Contrary to the usual view that Muller's ratchet does more damage when selection is weak, we show that the mean extinction time declines as mutations become more deleterious. Although very small sexual populations, such as self-fertilized lines, are subject to mutational meltdowns, recombination effectively eliminates the process when the effective population size exceeds a dozen or so. The concept ofthe effective mutation load is developed, and several procedures for estimating it are described. It is shown that this load can be reduced substantially when mutational effects are highly variable.Received June 27, 1989. Accepted January 17, 1990.All populations are doomed to eventual in small populations. On the other hand, extinction. The reasons include temporal Kimura et al. (1963) argued that the load variation in the environment, demographic due to mildly deleterious mutations is instochasticity (random variation in survi-versely proportional to population size, apvorship, birth rate, and sex ratio), and ge-proaching the selection coefficient in the abnetic problems such as inbreeding depres-sence of back-mutation. sion, the loss of adaptive variation by These results for sexual populations in random drift, and recurrent deleterious mu-drift-mutation-selection balance may be tations (Soule, 1987;Lande, 1988). This pa-contrasted with Muller's (1964) idea that a per is concemed exclusively with the latter mutational ratchet operates in asexual popproblem. Deleterious mutations impose a ulations, continually driving them to lower load on populations through the reduction and lower fitness. The idea here is that, in in the mean survivorship and/or reproduc-the absence of sex, no genotype can ever tive rates of individuals (Haldane, 1937; produce an offspring with fewer mutations Muller, 1950;Wallace, 1987). IDtimately, than its own load. In a finite population, this load should be manifested in a higher there is always a possibility that the class ...