Interpretations of visible/near-infrared reflectance spectra of Mars are often complicated by the effects of dust coatings that obscure the underlying materials of interest. The ability to separate the spectral reflectance signatures of coatings and substrates requires an understanding of how their individual and combined reflectance properties vary with phase angle. Toward this end, laboratory multispectral observations of rocks coated with different amounts of Mars analog dust were acquired under variable illumination and viewing geometries using the Bloomsburg University Goniometer (BUG). These bidirectional reflectance distribution function (BRDF) data were fit with a two-layer radiative transfer model, which replicated BUG observations of dust-coated basaltic andesite substrates relatively well. Derived single scattering albedo and phase function parameters for the dust were useful in testing the model's ability to derive the spectrum of a "blind" substrate (unknown to the modeler) coated with dust. Subsequent tests were run using subsets of the BUG data restricted by goniometric or coating thickness coverage. Using the entire data set provided the best constraints on model parameters, although some reductions in goniometric coverage could be tolerated without substantial degradation. Predictably, the most thinly coated samples provided the best information on the substrate, whereas the thickest coatings best replicated the dust. Dust zenith optical thickness values ???0.6-0.8 best constrain the substrate and coating simultaneously, particularly for large ranges of incidence or emission angles. The lack of sufficient angles can be offset by having a greater number and range of coatings thicknesses. Given few angles and thicknesses, few constraints can be placed concurrently on the spectral properties of the coating and substrate. ?? 2004 Elsevier Inc. All rights reserved.