Tailoring microstructure and composition are critical components for obtaining high-performance silicon nitride (Si3N4) ceramics. Anisotropic growth behavior of Si3N4 grains can be used to form very elongated grains, which serve to reinforce the matrix analogous to whisker-reinforcement of ceramics. The Si3N4 grain morphology is known to be very sensitive to the particular additive used, especially in the case of the oxides of the rare earths (RE) and Group III elements. However, the atomistic mechanisms by which this occurs has not been understood until now. A first-principles model, the differential binding energy, has been developed to characterize the competition between RE and Si for migrating to the beta-Si3N4 grain surfaces. The theory predicts that, of the RE, La should have the strongest and Lu the weakest preferential segregation to the grain surfaces. Additional calculations define the adsorption sites and their binding strengths for each of the REs on the prismatic plane of the Si3N4 grains. These predictions are confirmed by unique atomic-resolution images obtained by aberration-corrected Z-contrast scanning transmission electron microscopy (STEM). The combined theoretical and STEM studies reveal that the elements that induce the greatest observed grain anisotropy are those with the strongest preferential segregation plus high binding strength to the prismatic grain surface. Advances now allow one to use first principles calculations to determine the chemical affinity and bonding of the rare earths, as well as other elements, at the grain interfaces; atomistic factors that actually control the growth anisotropy as opposed to the commonly considered ion size, which at best, only provides a trend.