The lithospheric architecture of the Colorado Plateau--Rio Grande rift transition zone reflects a narrow corridor of active tectonism and magmatism between the tectonically stable Colorado Plateau and the stable interior of North America. In this paper we analyze Bouguer gravity data, crustal thickness variations, xenolith data, and seismic wave speed structure to argue that the Rio Grande rift and southeastern Colorado Plateau is underlain by a low-density upper mantle province that does not trend along upper crustal tectonic boundaries but, rather, is correlated with regions of middle to late Tertiary magmatism. We model uppermost mantle density contrasts using Bouguer gravity along five parallel profiles across the Colorado Plateau and validate our models by converting modeled density contrasts to seismic velocity variations for comparison with regional seismic data, as well as comparing modeled densities to calculated densities from xenoliths. We find that upper crustal features and crustal thickness variations do not explain long-wavelength (>200 km) variations in Bouguer gravity across the Colorado Plateau--Rio Grande rift. A large 200--440 km wide region of low uppermost mantle density is modeled beneath the eastern Colorado Plateau margin, Jemez Lineament, Rio Grande rift, and parts of the westernmost Great Plains. This region has a density contrast of -90 to -30 kg/m3 relative to the adjacent Colorado Plateau and Great Plains provinces and coincides with previously imaged regions of low upper mantle seismic velocities. In situ densities inferred from xenolith data in the Colorado Plateau support the modeled upper mantle density variations and support the idea that the eastern and western margins of the plateau have been significantly modified by metasomatism. We suggest that the region of low upper mantle density likely reflects some combination of temperature variations, partial melt contents, and compositional variations related to Tertiary magmatic activity and/or thinned lithosphere. The relationship of this zone of modified upper mantle to extension in the Rio Grande rift remains enigmatic, however, particularly given the lack of spatial correlation between the trends of upper mantle structures and surficial extensional structures.