Recent modeling efforts in San Francisco Bay suggest that more sites are suitable for eelgrass (Zostera marina) than those supporting extant populations; however, restoration attempts with transplants of whole shoots have had limited success. Flowering rates are relatively high in this system and there is at least one annual bed that relies on recruitment from seed each year. Here, we describe a mesocosm experiment to assess the potential to restore eelgrass to San Francisco Bay using seed. We measured seedling recruitment from one annual and two perennial donor populations in mesocosms seeded using a modification of the buoy-deployed seeding technique. The experiment included inoculation of half of the tanks with donor-site sediments, and microsatellite analysis of tissue from donor plants and recruited seedlings to compare genetic diversity in donor versus restored populations. Seeds from all three donor beds produced seedlings, with similar densities, growth rates, and maximum leaf lengths among donors. Inoculation with donor bed sediments led to a significant increase in seedling recruitment that was consistent across donors. Seedlings from the annual donor population did not develop flowering shoots within the first year in the mesocosms, implicating an environmental basis for the annual expression in the field. One of the perennial seed sources resulted in several flowering shoots by early winter, suggesting that environmental cues triggering flowering vary among donor beds. Microsatellite analysis is underway on both donor and mesocosm recruit tissues. Applicability of these results is currently being field-tested at three restoration sites within San Francisco Bay.