We used capture–mark–recapture models to investigate the effects of both individual and parental heterozygosity, measured at microsatellite loci on the survival of Seychelles warblers (Acrocephalus sechellensis), an endemic island species which went through a severe population bottleneck in the middle of the last century. We found that an individual's survival was not correlated with multilocus heterozygosity, or with heterozygosity at any specific locus. However, maternal, but not paternal, multilocus heterozygosity was positively associated with offspring survival, but only in years with low survival probabilities. A nestling cross-fostering experiment showed that this was a direct maternal effect as there was an effect of the genetic mother's, but not of the social mother's, heterozygosity. Heterozygosity–fitness correlations at microsatellite markers were generally assumed to reflect genome-wide effects. Although this might be true in partially inbred populations, such correlations may also arise as a result of local effects with specific markers being closely linked to genes which determine fitness. However, heterozygosity at the individual microsatellite loci was not correlated and therefore does not seem to reflect genome-wide heterozygosity. This suggests that even in a small bottlenecked population, heterozygosity–fitness correlations may not be caused by genome-wide effects. Support for the local effects hypothesis was also equivocal; although three specific loci were associated with offspring survival, including all single-locus heterozygosities as independent predictors for the variation in survival was not supported by the data. Furthermore, in contrast to the local effects hypothesis, the loci which contributed most to the heterozygosity–survival relationship were not more polymorphic than the other loci. This study highlights the difficulties in distinguishing between the two hypotheses.