Settling velocities in water were measured for natural quartz sand grains in the sieve-size range -0.75 to 1.50phi (0.35-1.68 mm). Obtained from beach sand samples, the grains were selected so as to provide a wide range of roundness from very angular to very-well rounded (rho r = 0.6-5.1). In a first set of experiments only two axial diameters were measured under a microscope, the longest axial diameter, D 1 , and the intermediate diameter, D i . In a second data set the smallest axial diameter, D s , was measured as well. Settling velocities were measured in a 6-meter long settling tube. It is found that the intermediate grain diameter is on average equal to the nominal diameter computed as D n = (D s D i D 1 ) (super 1/3) . This permits a simpler analysis of grain settling through microscopic measurements of D i alone but does not allow for corrections of grain shape effects. The measured settling velocity, w m , is then compared with the settling velocity of a sphere, w s , calculated using D i as the sphere diameter, and the relationship w m = 0.977w (super 0.913) s is found to be consistent with the data. An alternative approach is provided where w s is calculated directly from the sieve diameter, and then corrected to the actual settling velocity w m for that mean grain size. The results thus provide graphs and empirical relationships which permit the evaluation of actual grain-settling rates (rather than settling velocities of spheres) from the sieve diameter or from microscopic measurements of D i for individual grains. These procedures provide a mean settling velocity that takes into consideration the average degree of grain irregularities found in natural sands. A more detailed analysis is then conducted to examine the effects of roundness and sphericity on the settling of individual grains. It is found that grain roundness has no measurable effect on the settling rate. The settling velocity is shown to be a function of grain sphericity. either expressed as the Corey Shape Factor or E Shape Factor. Grain asymmetries must also be important, accounting for much of the scatter in the results.