Compaction bands are narrow, roughly planar zones of localized deformation, in which the shear is less than or comparable to compaction. Although there are differences in their appearance in the field and in laboratory specimens, they have been observed in both for high-porosity (greater than about 15%) sandstones. Because the porosity in them is reduced and the tortuosity increased, they inhibit fluid flow perpendicular to their plane. Consequently, they can alter patterns of fluid movement in formations in which they occur and are relevant to applications involving fluid injection or withdrawal. Formation of compaction bands is predicted by a framework that treats localized deformation as a bifurcation from homogeneous deformation. This paper gives a brief overview of compaction localization but focuses on field and laboratory observations that constrain two parameters entering the bifurcation analysis: a friction coefficient μ and a dilatancy factor β. The inferred values suggest that normality (μ = β) is not satisfied, and compaction localization occurs on a transitional portion of the yield surface, where the local slope in a plot of Mises equivalent shear stress versus compressive mean normal stress changes from positive (μ > 0) to negative (μ < 0). These inferences are at odds with critical state and cap theories that typically assume normality and predict dilation on the portion of the surface where μ > 0. In addition, the values suggest that the critical state (μ = 0) does not necessarily correspond to zero volume change.