1. Taxonomic analysis of pollen collected by bees can provide insights into their host plant use, providing information about the plant species selected for targeted conservation strategies. The two main identification approaches used are morphological analysis of pollen samples affixed to microscope slides (i.e., microscopic palynology) and molecular analysis of samples. Both methods are widely used for freshly collected materials and have been compared in multiple studies, yet their application to archived samples remains to be explored. Archived samples may be particularly useful for the study of rare or protected species, particularly when historical foraging patterns are uncertain.2. We used both methods to analyse pollen collected by the endangered rusty patched bumble bee species, Bombus affinis Cresson, applied to museum-archived specimens. Pollen samples were removed from the corbiculae of bumble bees originally collected in Michigan between 1914 and 1974. Samples included 24 rusty patched bumble bees each with large pollen loads on both corbiculae, allowing for pollen from the same bee to be analysed using both methods.3. DNA metabarcoding detected more plant taxa than light microscopy, and DNA barcoding also had higher taxonomic resolution when compared to taxa determined using light microscopy. In many instances, pollen could only be confidently identified to tribe or family with light microscopy. Discrepancy between methods decreased when taxa identified via DNA metabarcoding were binned into ecologically relevant groups corresponding to those identified using light microscopy. 4. Although binning demonstrated smaller within-method variance, there was still minimal correspondence between the two methods. Results indicate there are benefits and biases unique to each method and highlight the utility of binning taxonomic results to morphological or ecological groupings.