Technical Abstract: Bumble bees (Bombus Latrielle) are some of the most important native pollinators in North America, pollinating both wild flowers and agricultural crops often where other pollinators such as honey bees are absent. In the last decade at least six North American species of bumble bees (Bombus Latreille) have experienced significant range contraction or population declines, including one species (Bombus franklini Franklin) which may be extinct. The prevailing hypotheses attributing to the decline of North American bumble bee fauna include habitat fragmentation and degradation, as well as pathogen introduction from commercially reared congeners. With the exception of Bombus pensylvanicus (DeGeer) and Bombus sonorous Say, all of the at-risk species belong to a single subgenus (Bombus sensu stricto Latreille). In each case, population decline is documented, but the full extent of the range contraction and the underlying causes are not fully understood. One major limitation to studying range contractions is our limited understanding of actual historic range across the broad geography of the North America. Here we review the current body of knowledge of the North American representatives of the subgenus Bombus s. str. and present a method for creating species databases using specimens housed in entomological collections. These data can be used to generate probable historic range maps from specimens found across multiple entomological museum collections. We demonstrate the benefits of expanding data sets by focusing on one species, Bombus occidentalis, for which we show the power of the predictive models when the database used is increased in size from 973 specimens to 2928 specimens. The increase in specimens included material from nine institutions and several publications and these locations represented many different geographical areas within the species range. Use of the expanded data set increased the predicted historic range in geographic scope and likelihood of occurrence changed across the range indicating that specimens housed at one institution represent an incomplete data set. We provide preliminary maps of three other species in the subgenus, B. affinis, B. franklini and B. terricola, for which the databases are still incomplete. We compare these maps to the existing range limit maps used historically. While modeling has largely been used to predict species range changes in the future, these predictive range maps may be useful in initiating and assessing monitoring programs of at-risk species.