If, because of genetic erosion, the level of homozygosity in small populations is high, additional selfing will result in small reductions of fitness. In addition, in small populations with a long inbreeding history selection may have purged the population of its genetic load. Therefore, a positive relationship between population size (or level of genetic variation) and level of additional inbreeding depression, here referred to as inbreeding load, may be expected. In a previous study on the rare and threatened perennial Salvia pratensis, a positive correlation between population size and level of allozyme variation has been demonstrated. In the present study, the inbreeding load in six populations of varying size and allozyme variation was investigated. In the greenhouse, significant inbreeding load in mean seed weight, proportion of germination, plant size, regenerative capacity, and survival was demonstrated. In a field experiment with the two largest and the two smallest populations, survival of selfed progeny was 16% to 63% lower than survival of outcrossed progeny. In addition, survival of outcrossed progeny was, with the exception of the largest population, lower (16% to 37%) than of hybrid progeny, resulting from crosses between populations. Effects on plant size were qualitatively similar to the effects on survival, but these effects were variable in time because of differential survival of larger individuals. In all populations the total inbreeding load, that is, the effects on size and survival multiplicated, increased in time. It was demonstrated that inbreeding load in different characters may be independent. At no time and for no character was inbreeding load or the heterosis effect correlated to the mean number of alleles per locus, indicating that allozyme variation is not representative for variation at fitness loci in these populations. Combined with results of previous investigations, these results suggest that the small populations are in an early phase of the genetic erosion process. In this phase, allozyme variation, which is supposed to be (nearly) neutral, has been affected by genetic erosion but the selectively nonneutral variation is only slightly affected. These results stress the need for detailed information about the inbreeding history of small populations. The relative performance of selfed progeny was lowest in all populations, in the greenhouse as well as in the field, and inbreeding depression could still influence the extinction probabilities of the small populations.