To the eye, most common metals and ceramics used in commercial products appear to be uniformly solid. But at the microscopic level, they are polycrystalline, made up of aggregates of grains that have different sizes, shapes, and crystal orientations. The grains are tied together by a network of grain boundaries that influence the material’s properties. Materials scientists manipulate the density of grain boundaries to control the properties.  For the last 70 years, the movement of grain boundaries (which determines their density) was predicted using a theory that says that the velocity at which a grain boundary moves through a heated material is correlated to the boundary’s curvature. Our observations show that this theory, formulated to describe the most ideal case, does not apply in real polycrystals.  The finding means that grain growth models must be updated to more accurately predict the effect of annealing on the ultimate properties of the material.