Predictions for the evolution of mating systems and genetic load vary, depending on the genetic basis of inbreeding depression (dominance versus overdominance, epistasis and the relative frequencies of genes of large and small effect). A distinction between the dominance and overdominance hypotheses is that deleterious recessive mutations should be purged in inbreeding populations. Comparative studies of populations differing in their level of inbreeding and experimental approaches that allow selection among inbred lines support this prediction. More direct biometric approaches provide strong support for the importance of partly recessive deleterious alleles. Investigators using molecular markers to study quantitative trait loci (QTL) often find support for overdominance, though pseudo–overdominance (deleterious alleles linked in repulsion) may bias this perception. QTL and biometric studies of inbred lines often find evidence for epistasis, which may also contribute to the perception of overdominance, though this may be because of the divergent lines initially crossed in QTL studies. Studies of marker segregation distortion commonly uncover genes of major effect on viability, but these have only minor contributions to inbreeding depression. Although considerable progress has been made in understanding the genetic basis of inbreeding depression, we feel that all three aspects merit more study in natural plant populations.