The application of nonlinear mixing to model the band contrasts in the spectra of intimate mixtures shows that not only the band contrasts but also their ratios are sensitive to the presence of a featureless end-member. The ratio of the contrast of the weaker band to that of the deeper one decreases with increasing content of neutral phase. The degree and the rate of decline depend on a number of factors. When a neutral phase is present, the use of relative band contrasts to estimate the relative abundance of end-members results in systematic errors. In multicomponent mixtures the presence of a spectrally neutral phase leads to the underestimation of the component with the weaker band. The larger the difference in contrasts between the bands, the larger is the systematic error. Additional factors increasing the error are an opacity of the neutral phase, a large difference in brightness between the neutral phase and the components with absorption features, a decreased grain size of the neutral component relative to those of the other constituents, and decreased grain sizes of all end-members. The use of calibrations for binary mineral mixtures to analyze the spectra of multicomponent surfaces should be subject to errors when opaque constituents are present. Some variations in relative band contrast of asteroid spectra due to variable contents and/or particle size of opaques may be misinterpreted as variations in silicate modal abundance. ¿ 1999 American Geophysical Union