Arnica montana is a rare and rapidly declining, self-incompatible plant species. In 26 populations in The Netherlands we investigated the relationship between population size and genetic variation using allozyme markers. Genetic variation was low in A. montana ( H e ϭ 0.088). There were positive correlations between population size and the proportion of polymorphic loci, the number of effective alleles, and expected heterozygosity, but not with observed heterozygosity. There was a significantly positive correlation between population size and the inbreeding coefficient. Generally, small populations showed heterozygote excess, which decreased with increasing population size. Possibly, the heterozygous individuals in small populations are survivors from the formerly larger populations with relatively high fitness. The F statistics showed a moderately high level of differentiation among populations ( F ST ϭ 0.140 Ϯ 0.02), implying a low level of gene flow. For three out of four allozyme loci, we found significant inbreeding ( F IS ϭ 0.104 Ϯ 0.03). Only 14 of 26 populations were in Hardy-Weinberg equilibrium at all four polymorphic loci. In a subset of 14 populations of various size, we investigated natural seed production and offspring fitness. Population size was positively correlated with seed set, seedling size, number of flowering stems and flowerheads, adult survival, and total relative fitness, but not with the number of florets per flowerhead, germination rate, or the proportion of germination. Offspring performance in the greenhouse was not associated with genetic diversity measured on their mothers in the field. We conclude that the fitness of small populations is significantly reduced, but that there is as yet no evidence that this was caused by inbreeding. Possibly, the self-incompatibility system of A . montana has been effective in reducing selfing rates and inbreeding depression.