Generalized pollinators visit multiple co-flowering plant species and may transfer heterospecific pollen grains. Recent studies have indicated that the effect of heterospecific pollen (HP) on reproduction success is variable and depends on the identity of donor and recipient species. However, few studies have documented variation in HP receipt and evaluated the reproductive effects of HP receipt across geographic locations under natural conditions. We investigated the spatial variation of pollen deposition across eight sites and how the pollen receipt related to seed set of Salvia przewalskii, a subalpine perennial herb in Hengduan Mountain in southwest China. We found that stigmatic pollen loads substantially varied among sites for several metrics, including quantities of conspecific and heterospecific pollen, the proportion of HP, and species composition of HP donors. Five different plant families were the most common HP source at one or two sites, and the proportion of HP ranged from 3.4–51.3% across sites. The association of conspecific pollen with seed set was positive and variable among sites, whereas the association of HP receipt and seed set was negative and not significantly different among sites. Our results demonstrate variation in the quantity and fitness effect of pollen receipt across sites, which is a precondition for evolution of local adaptation. Further study of variation in patterns and effects of HP receipt for the same recipient species across natural communities would allow for better understanding of the ecological and evolutionary consequences of HP receipt. 

Field collections occurred from July 25 to August 19, 2021, during peak flowering of Salvia przewalskii. We collected 2.9 (median, 3) stigmas per plant, with a range from 1 to 10 stigmas due to variation in plant size. Because the wilted corolla and pistil usually abscise, we secured the corolla to the corona with a single horizontal pin at the corona’s midpoint to keep it from falling off. For each site, we performed a given procedure (labeling, stigma collection or seed counts) for all flowers on the same day. After three or four days, when the corolla wilted, we removed the stigma with clean forceps and stored it in a microcentrifuge tube containing 70% ethanol. In most cases, the style abscised naturally, which indicated that the fertilization had already occurred, and germinated pollen exines were still affixed to the stigma. Ten days after stigma collection, we counted the developed and undeveloped seeds of each flower, which were readily distinguishable based on size. This approach allowed us to link pollen receipt and seed number in the same flower. In some cases, stigmas were damaged or flowers (plants) were missing or showed evidence of herbivore damage. In total, 878 stigmas from 335 individuals were retrieved. We then observed the stigmatic pollen grains in the laboratory at 400× magnification and captured a series of images using a digital camera linked to a microscope, with one image per species of HP pollen on each stigma, the pollen from each HP species being sufficiently aggregated to be captured in a single image. Pollen grains were identified to the species level on the basis of morphological features including size, shape, and exine ornamentation, with use of a pollen reference library of co-flowering species within sites. We excluded the observations for which there was no pollen deposition (10 flowers) or conspecific pollen quantities were smaller than the number of seeds (19 flowers). We assume that these observations were due to loss of stigmatic pollen in the field during fertilization and stigma collection.