Polarization provides useful quantitative information about scattering surfaces. In hyperspectral remote sensing of natural surfaces composed of granular materials, there are relatively few studies of polarization. Most earlier remote sensing studies of polarization have been based on multi-spectral data, and the majority focused on the negative branch of polarization, which typically appears at phase angles less than 20 degrees, using models with limited accuracy. Models of the positive branch have also shown limitations, particularly at longer phase angles. We review these earlier studies by Hapke and Shkuratov and present the results of our laboratory study using hyperspectral polarization imagery of particulate surfaces. Although the linear polarization ratio is typically a nonlinear function of phase angle, our results show that in an approximately linear region of the polarization curve, there is a correlation between the slope of the linear polarization ratio and the average grain size.