ABSTRACT Stellar population synthesis models can now reproduce the photometry of old stellar systems (age 2 Gyr) in the near-infrared (NIR) bands at 3.4 and 4.6 μm (WISE W1 and W2 or IRAC 1 and 2). In this paper, we derive stellar mass-to-light ratios for these and optical bands, and confirm that the NIR M/L shows dramatically reduced sensitivity to both age and metallicity compared to optical bands, and further, that this behavior leads to significantly more robust stellar masses for quiescent galaxies with [Fe/H] −0.5 regardless of star-formation history (SFH). We then use realistic early-type galaxy SFHs and metallicity distributions from the EAGLE simulations of galaxy formation to investigate two methods to determine the appropriate M/L for a galaxy. (1) We show that the uncertainties introduced by an unknown SFH can be largely removed using a spectroscopically inferred luminosity-weighted age and metallicity for the population to select the appropriate single stellar population (SSP) equivalent M/L. Using this method, the maximum systematic error due to SFH on the M/L of an early-type galaxy is at 3.4 μm and typical uncertainties due to errors in the age and metallicity create a scatter of . The equivalent values for optical bands are more than two to three times greater, even before considering uncertainties associated with internal dust extinction. (2) We demonstrate that if the EAGLE SFHs and metallicities accurately reproduce the true properties of early-type galaxies, the use of an iterative approach to select a mass dependent M/L can provide even more accurate stellar masses for early-type galaxies, with typical uncertainties of .