Dust particles larger than 10 <em>&mu;</em>m are increasingly recognized as important to the climate system, yet models tend to underestimate their atmospheric abundance. Although recent work has improved representations of dust emission processes, dry deposition remains comparatively unexplored. Here we derive a mass-conserving parameterization of the dust deposition velocity that is physically consistent from submicron through giant particle sizes. We calibrate the parameterization using laboratory measurements and compare it to commonly used formulations. We find that assuming spherical particles overestimates the deposition velocity for supermicron dust, while assuming Stokes-regime settling increasingly overestimates the deposition velocity for particles larger than 10 <em>&mu;</em>m. Furthermore, the commonly used electric analogue formulation produces substantially larger deposition velocities than the mass-conserving formulation, particularly for particles larger than 10 <em>&mu;</em>m, although the magnitude and, in some cases, sign of this discrepancy depend on atmospheric stability, friction velocity, reference height, and treatment of particle rebound. Taken together, commonly used assumptions can produce deposition velocities and fluxes up to a factor of 1.6 larger than more physically consistent formulations. A survey of dust schemes used in climate models participating in the Sixth Climate Model Intercomparison Project found that all used non-mass-conserving dry deposition formulations and Stokes theory for spherical-particle settling. These results suggest that commonly used model parameterizations tend to overestimate dust dry deposition, which may contribute to an underestimation of the abundance and lifetime of large dust particles.