Early in the crystallization of many tholeiitic basaltic magmas, plagioclase crystals cluster together into a 3-D cellular framework, which forms a passive marker capable of recording the deformation that accompanies compaction of crystal mush. Although irregular in detail, the overall structure is initially isotropic and becomes anisotropic as a result of compaction. We have developed three independent methods using digitized information from several (at least three) non-parallel thin sections to determine the 3-D anisotropism of the plagioclase network, from which the amount and direction of compaction can be calculated. One method focuses on the angular variation of the mean intercept along parallel traverses through the network; another examines the orientation and size distribution of individual links; and the third considers the average shape of interstitial regions outlined by the plagioclase network. A rigorous test of the hypothesis that the network is statistically anisotropic and estimates for the standard errors of the derived compaction parameters are deduced from the statistics of elliptically contoured non-normal multivariate distributions. We illustrate the methods using four non-parallel thin-sections from a sample of diabase 146 m above the base of the 300-m thick Palisades sill of New Jersey. Compaction of crystal mush in this sill has previously been postulated on the basis of chemical evidence. The textural anisotropy determined by the three different stereological approaches, can be shown to be statistically significant and to suggest compaction on the order of 7.2% in a direction roughly parallel to the intersection of the columnar joints at the sample site. Least squares estimates for the principal axes of the textural anisotropism have normalized dimensions of 1.035, 1.016 and 0.951 oriented at 123/13, 031/11 and 081/73 respectively.