Temperature gradient metamorphism in snow results in the development of branch grains at the expense of other types of grains. At constant density, the development of branch grains decreases the strength, thermal conductivity, and crystal number density while increasing the mean crystal size. A theory is developed that uses the following major assumptions: (1) at growth sites the thermal gradient is a function of crystal separation and (2) vapor flux is a function of local thermal gradient. A computer model based on the theory is compared with stereological measurements of snow from a carefully controlled experiment. The model calculates grain size distributions within the accuracy of the measurements. The major conclusion is that the theory explains the main process in temperature gradient metamorphism. The inclusion of surface energy considerations and stereological refinements would probably improve the theory when techniques for stereological analysis become more accurate.