To compare the rates of pollen depletion among C. virginica populations, we assessed the amount of pollen available in individual flowers (an estimate of the pollen standing crop) during the period of insect activity, from mid-morning until early or late afternoon. Each hour, we arbitrarily chose eight flowers scattered throughout the local plant population and collected the anthers from each flower into a microcentrifuge tube containing 1 mL of 70% ethanol. We chose flowers regardless of the number of dehisced anthers. We began anther collection each morning when the flowers’ anthers began to dehisce, usually between 8:45 and 10:00, but occasionally as late as 13:15 because of cool or wet weather. We concluded anther collection when insect visitation ceased or flowers began to close, between 13:00 and 15:00. We counted the number of pollen grains available in each flower sample using a Coulter Multisizer 3 particle counter (Beckman Coulter Inc., Brea, California). Prior to counting, we suspended pollen in 20 mL 0.9% saline. To obtain a representative count of the number of pollen grains in each flower, we took the average of four 1mL subsamples and multiplied the mean by the total volume of the sample. These counts included the number of pollen grains in undehisced anthers because undehisced anthers opened and emptied during the counting process.

To estimate the pollen presentation schedule in each population, we tracked anther dehiscence in ten male-phase flowers on each day of data collection. Prior to first anther dehiscence, we arbitrarily chose ten fresh male-phase flowers and marked them with dental floss and a label. Every 30 minutes, we recorded the temperature and then returned to these flowers and recorded the number of locules dehisced (two locules per anther for a total of ten locules). We concluded tracking each flower when all ten locules had dehisced or, if the ten locules did not dehisce completely, when insect activity ceased.

We estimated total pollen production during each day of data collection. We covered a sample of male-phase flowers with fine mesh cages prior to anther dehiscence to prevent insect visitation. After insect visitation ended for the day, we collected 5-15 samples of the unvisited flowers to estimate the total number of pollen grains. Because we collected these flowers at the end of each day, these counts exclude pollen that may have passively fallen from anthers.

To compare the visitation rates of flower visitors to C. virginica, we identified visitors to sets of flowers in sets of six 5-minute observation periods at each site. Observations were spread across the site, and when possible, we conducted 90 minutes of observations throughout the period of insect activity, from mid-morning until early or late afternoon. Prior to each observation period, we selected a new set of flowers that we could observe simultaneously, marked them using a hula hoop or embroidery hoop (depending on floral density), and recorded the number of male- and female-phase flowers. During the observation period, we identified each visitor and counted the number of visits by that visitor by to male- and female-phase flowers. We used knowledge from multiple years of collecting (Williams et al. 2013) to identify A. erigeniae to species and sex, Bombylius major to species, generalist bees in the genera Lasioglossum, Ceratina, and Hylaeus to a group we called “small generalist bees”, cleptoparasitic bees in the genus Nomada to genus, and other visitors to species, genus, or morphological group to the best of our abilities.