The results of a combined analytical and numerical study of two locus selection models with partial selfing indicate that several commonly held opinions about the effects of partial self-fertilization do not hold in general. For example, the heterozygosity of a population may actually increase as the selfing rate is increased. Similarly, selection strong enough to guarantee a two locus polymorphism with complete selfing does not necessarily guarantee a two locus polymorphism with intermediate amounts of self-fertilization. The results presented here and a brief review of previously existing results indicate that the predictions of population genetic models based on the assumption of random mating will not be greatly altered by a small amount of self-fertilization, unless the loci involved are tightly linked. On the other hand, the results presented indicate that a very small amount of outcrossing may lead to marked differences from the expectation based on complete self-fertilization.