The mechanics of snow friction are central to competitive skiing, safe winter driving and efficient polar sleds. For nearly 80 years, prevailing theory has postulated that self-lubrication accounts for low kinetic friction on snow: dry-contact sliding warms snow grains to the melting point, and further sliding produces meltwater layers that lubricate the interface. We sought to verify that self-lubrication occurs at the grain scale and to quantify the evolution of real contact area to aid modeling. We used high-resolution (15 µm) infrared thermography to observe the warming of stationary snow under a rotating polyethylene slider. Surprisingly, we did not observe melting at contacting snow grains despite low friction values. In some cases, slider shear failed inter-granular bonds and produced widespread snow movement with no persistent contacts to melt (μ < 0.03). When the snow grains did not move and persistent contacts evolved, the slider abraded rather than melted the grains at low resistance (μ < 0.05). Optical microscopy revealed that the abraded particles deposited in air pockets between grains and thereby carried heat away from the interface, a process not included in current models. Overall, our results challenge whether self-lubrication is indeed the dominant mechanism underlying low snow kinetic friction.