Several empirical models for deriving FeO concentration from multispectral imaging of the Moon are tested using model spectra of lunar soil. The spectra are computed from a radiative transfer model based on the work of Bruce Hapke and cover the system plagioclase-orthopyroxene-clinopyroxene-olivine-ilmenite, and vary the Mg number and maturity over lunar relevant ranges. The high degree of correlation between measured and spectrally estimated FeO when applied to lunar soils returned by Apollo provides constraints on the abundance of certain rock types, and these constraints are consistent with geochemical mixing models. Good FeO estimates are yielded for mixtures of plagioclase, clinopyroxene, and ilmenite by all the algorithms. It is found that all spectral algorithms show large errors in the presence of significant olivine and significant errors in the presence of ferroan orthopyroxene. The best performance is shown by the LeMouelic et al. (2002) algorithm that is also insensitive to topographic shading in the absence of ilmenite. Previously noted differences between spectrally derived FeO and that derived by other techniques are all consistent with reported mineralogical differences.