The observation that offspring produced by the mating of close relatives are often less fit than those produced by matings between unrelated individuals (i.e., inbreeding depression) has commonly been explained in terms of the increased probability of expressing deleterious recessive alleles among inbred offspring (the partial dominance model). This model predicts that inbreeding depression should be limited in regularly inbreeding populations because the deleterious alleles that cause inbreeding depression (i.e., the genetic load) should be purged by regularly exposing these alleles to natural selection. We indirectly test the partial dominance model using four highly inbred populations of an androdioecious crustacean, the clam shrimp Eulimnadia texana. These shrimp are comprised of males and hermaphrodites, the latter capable of either self-fertilizing or mating with a male (i.e., outcrossing between hermaphrodites is impossible). Hermaphrodites are further subdivided into monogenics (produced via self-fertilization) and amphigenics (produced via self-fertilization or outcrossing). Electrophoretic evidence suggests significant differences in heterozygosity among populations, but that selfing rates were not statistically different (average s = 0.67). Additional electrophoretic analyses reveal that three previously described sex-linked loci (Fum, Idh-1, and Idh-2) are all tightly linked to each other, with crossing over on the order of 1% per generation. Although selfing rates are clearly high, we present evidence that early inbreeding depression (hatching rates, juvenile survival, and age at sexual maturity) exists in all four populations. For all of these factors, inbreeding depression was inferred by the positive correlation of multilocus heterozygosity and fitness. Cumulative inbreeding depression (8) is between 0.41 and 0.47 across all populations, which appears to be too low to limit the effects of purging via identity disequilibrium. Instead, we suggest that the maintenance of inbreeding depression in these populations is due to the observed linkage group, which we suggest contains a large number of genes including many related to fitness. Segregation of such a large linkage group would explain our observations of the predominance of amphigenic hermaphrodites in our field samples and of survival differences between monogenics and amphigenics within selfed clutches. We propose that a modified form of the overdominance model for inbreeding depression operating at the level of linkage groups maintains the observed levels of inbreeding depression in these populations even in the face of high rates of selfing.