"The effects of inbreeding may not be as noticeable in the first generation as the invigoration immediately apparent after crossing". This statement, published in 1919, has received little attention, and has apparently never been tested empirically, although the reduction of the genetic load of populations by inbreeding is well known in theoretical terms. Because inbreeding increases homozygosity, and hence the effectiveness of selection against recessive or partially recessive detrimental alleles, changes in levels of inbreeding can lead to a reduction in the frequencies of such mutant alleles. This results in equilibration at higher population mean fitness and is referred to as 'purging' populations of their genetic load. Severe inbreeding can also reduce genetic load due to overdominant alleles, provided selection coefficients are not symmetrical at all loci, because alleles giving lower fitness will be reduced in frequency at equilibrium. With either fitness model, however, reduction in genetic load takes time, and the initial effect of an increase in inbreeding is reduced fitness due to homozygosity. There are few data relating to the extent to which fitness is reduced during inbreeding in a set of lines and to how long the reduction lasts before increasing again to the initial level, or higher. Inbreeding experiments involving sib mating in mice and Drosophila subobscura, and successive bottlenecks in house flies have yielded some evidence consistent with the purging hypothesis. Here, we report results of an experiment demonstrating a prolonged time-course of recovery of mean fitness under self-fertilization of a naturally outcrossing plant, and also compare our results with expectations derived by computer calculations. Our results show that the genetic load present in an outcrossing population can be explained only with a high mutation rate to partially recessive deleterious alleles, and that inbreeding purges the population of mutant alleles.