Figure 4. A Preference for different Clarkia pollen based on samples taken from foraging bees. Preference was calculated as the difference between frequency of a Clarkia species blooming in the community and frequency it was represented in bees’ pollen balls. Mean preference is plotted with a 95% confidence interval, and was estimated using quantitative amplicon sequencing. Estimates of preference did not significantly differ between quantitation methods (not shown). B Bar plots showing pollinator constancy. The far-left bar shows what we would expect if bees are inconstant: for example, the 34 bees caught in communities with four co-floweringClarkia species (black section of the far-left bar) should have four Clarkia species in their pollen balls. We estimated the number of Clarkia species found on bees in four ways: quantitative amplicon sequencing (qAMPseq), and relative read abundance (RRA) at different sample proportion cutoffs. Bees carried fewer species in their pollen balls than were flowering in their correspondingClarkia communities. The far-right bar is relative read abundance with no sample proportion cutoff; with this quantitation method, all samples contained all four species.