The reproductive assurance hypothesis posits that selection favors self‐pollination in flowering plants where mates and/or pollinators are scarce. A corollary is that self‐pollinating populations are expected to be superior colonizers of mate‐ and pollinator‐scarce environments. The California annual Clarkia xantiana includes outcrossing populations (ssp. xantiana) and autogamously self‐pollinating populations (ssp. parviflora). Outcrossing is ancestral, and the subspecies have parapatric distributions with a narrow contact zone. We tested aspects of the reproductive assurance hypothesis by examining geographic and subspecies variation in the densities of mates and pollinators (native bees) and the density dependence of pollinator visitation and pollen receipt. Plant and flower densities, pollinator density, and pollinator visitation rates were lowest in the region of exclusively self‐pollinating populations. Pollinator assemblages there lacked Clarkia‐associated pollinator taxa that were common elsewhere. Self‐pollinating populations in the contact zone generally had densities and visitation rates intermediate between allopatric self‐pollinating populations and outcrossing populations. Visitation rate and pollen receipt increased significantly with plant density. These findings suggest that selection for reproductive assurance influenced the origin of self‐pollination and/or that reproductive assurance influenced the geographic distribution of self‐pollination. Geographic variation in pollinator assemblages may have generated variation in the value of reproductive assurance.