Abstract
Lupinus microcarpus is a self‐compatible annual plant that forms a species complex of morphologically variable but indeterminate varieties. In order to examine the hypothesis that varieties of L. microcarpus comprise genetically differentiated and reproductively isolated species, populations of L. microcarpus var. horizontalis and var. densiflorus were sampled from an area of sympatry in central California and genotyped using six microsatellite loci. Bayesian clustering divided the total sample into two groups corresponding to the named varieties with extremely low levels of inferred coancestry. Similarly, maximum likelihood and distance methods for genetic assignment placed individuals in two nonoverlapping groups. Evidence for isolation by distance (IBD) within each variety was found at shorter distance classes, but varieties remained differentiated in sympatry. Furthermore, coalescent estimates of divergence time indicate separation within the past 950–5050 generations, with minimal gene flow after divergence. A four‐level hierarchical analysis of molecular variance ( amova ) found significant levels of genetic differentiation among varieties (θ P = 0.292), populations within varieties (θ S = 0.449), subpopulations within populations (θ SS = 0.623), and individuals within subpopulations ( f = 0.421); but the greatest degree of differentiation was at the subpopulation level. Although it is sometimes assumed that the magnitude of genetic differences (e.g. F ST ) should be greater between species than among populations or subpopulations of the same species, shared ancestral polymorphism may lead to relatively low levels of differentiation at the species level, even as the stochastic effects of genetic drift generate higher levels of differentiation at lower hierarchical levels. These results suggest that L. microcarpus var. horizontalis and var. densiflorus are recently diverged yet reproductively isolated species, with high levels of inbreeding resulting from the combined effects of limited gene flow, demographic bottlenecks, and partial selfing in finite, geographically structured populations.