Historical and recent studies of Boulder County, Colorado (USA) bees (Hymenoptera: Anthophila) illustrate the potential and the pitfalls of using comparative collection data to evaluate faunal composition and change over time. A compilation of bee records from Boulder County (Scott et al. 2011) is used as a basis for re-examining the comparison of an historical data set (Cockerell 1907) with a recent one (Kearns and Oliveras 2009a,b). Despite numerical comparability reported by Kearns and Oliveras, the taxonomic and behavioral composition of these data sets differ markedly from each other and, in different ways, from that of the subset of bee species common to both and from the total fauna documented from Boulder County. The rank order of species richness across bee families and across cohorts of bees with different social behaviors and feeding preferences do not covary among data sets: taxonomically, colletids, andrenids and megachilids are relatively under-represented in the more recent data set, in which halictids are better represented than any other family. Behaviorally, parasitic and oligolectic solitary species are less well represented in both datasets, the paucity of Andrena Fabricius spp. being especially conspicuous in the more recent of the two, which is dominated by polylectic social bees. Ensemble comparisons of raw species richness may mask differences in the readiness with which bees with different taxonomic affiliations, social behaviors, and degrees of host plant specificity lend themselves to being sampled, and possibly to their appearing to undergo faunal turnover. When these behaviors are taken into account, these comparisons imply either (1) impacts, in descending order of severity, on parasitic bees, oligolectic solitary bees, and polylectic solitary bees or (2) the relative ineffectiveness with which these classes of bees are sampled with generalized pan- or bowl-trapping techniques. Seasonally limited summer sampling protocols may favor polylectic social bees, as reflected in halictids' being disproportionately well-sampled relative to andrenids, which account for a majority of early season oligolectic solitary bees—ironically those of potentially greatest interest when evaluating the stability of pollinator faunal diversity. By virtue of their abundance and protracted flight seasons, in contrast, eusocial bees appear to be among the most readily sampled, though potentially less relevant to the purposes of evaluating the integrity of animal-pollinated plant communities.