Abstract.-If inbreeding depression is caused by deleterious recessive alleles, as suggested by the partial dominance hypothesis, a negative correlation between inbreeding and inbreeding depression is predicted. This hypothesis has been tested several times by comparisons of closely related species or comparisons of populations of the same species with different histories of inbreeding. However, if one is interested in whether this relationship contributes to matingsystem evolution, which occurs within populations, comparisons among families within a population are needed; that is, inbreeding depression among individuals with genetically based differences in their rate of selfing should be compared. In gynodioecious species with self-compatible hermaphrodites, hermaphrodites will have a greater history of potential inbreeding via both selfing and biparental inbreeding as compared to females and may therefore express a lower level of inbreeding depression. We estimated the inbreeding depression of female and hermaphrodite lineages in gynodioecious Lobelia siphilitica in a greenhouse experiment by comparing the performance of selfed and outcrossed progeny, as well as sibling crosses and crosses among subpopulations. We did not find support for lower inbreeding depression in hermaphrodite lineages. Multiplicative inbreeding depression (based on seed germination, juvenile survival, survival to flowering, and flower production in the first growing season) was not significantly different between hermaphrodite lineages (8 = 0.30 ± 0.08) and female lineages (8 = 0.15 ± 0.18), although the trend was for higher inbreeding depression in the hermaphrodite lineages. The population-level estimate of inbreeding depression was relatively low for a gynodioecious species (8 = 0.25) and there was no significant inbreeding depression following biparental inbreeding (8 = 0.01). All measured traits showed significant variation among families, and there was a significant interaction between family and pollination treatment for four traits (germination date, date of first flowering, number of flowers, and aboveground biomass). Our results suggest that the families responded differently to selfing and outcrossing: Some families exhibited lower fitness following selfing whereas others seemed to benefit from selfing as compared to outcrossing. Our results support recent simulation results in that prior inbreeding of the lineages did not determine the level of inbreeding depression. These results also emphasize the importance of determining familylevel estimates of inbreeding depression, relative to population-level estimates, for studies of mating-system evolution.Key words.-Biparental inbreeding, gynodioecy, inbreeding depression, Lobelia siphilitica. Received October 8, 1997. Accepted June 25, 1998 Inbreeding depression, the reduced fitness of progeny derived from inbreeding relative to outcrossing, has received much attention in recent decades. According to the "partial dominance" hypothesis, inbreeding depression is a consequence of in...