The interstellar reddening curve predicted theoretically for small graphite flakes is in remarkable agreement with the observed reddening law, suggesting that the interstellar grains may be graphite and not ice. This possibility is not in contradiction with the high albedos of reflection nebulae at photographic wave-lengths, provided the particles have sizes of order 10 −5 cm. The origin of graphite flakes at the surfaces of cool carbon stars is considered, about 10 4 N stars in the galaxy being sufficient to produce the required density of interstellar grains in a time of 3 × 10 9 years. Grains tend to be formed in the pulsation cycle of an N star at temperatures < 2700°K. The grains have an important effect on the photospheric opacity, causing the photospheric density to decrease very markedly as the temperature falls towards 2000°K. It is this fall of density that allows the grains to be repelled outwards by radiation pressure and to leave the star altogether in spite of the frictional resistance of the photospheric gases. The grains do not evaporate as they leave the atmosphere of the star. Much of the empirical data concerning graphite is uncertain. The optical constants used in deriving the reddening law are consistent, however, with the reflectivity of polished graphite surfaces. A low value of 10 −3 has been used for the accommodation coefficient, since this has been obtained in the most extensive series of experiments. With this value it is shown that the expelled grains can reach a size of ∼ 10 −5 cm. The velocities of expulsion from stars can exceed 1000 km sec −1 . The possibility that interstellar grains are carbon particles produced by stars has important cosmogonic consequences. Grains can be produced in galaxies that contain little or no gas. Grains can be expelled from galaxies. Graphite is highly refractory and would not evaporate in H II regions. Graphite chemisorbs hydrogen and is therefore an effective catalyst in the production of interstellar H 2 . Indeed, graphite grains would be highly efficient in the production of interstellar molecules in general, whereas ice is probably very inefficient. Graphite possesses anisotropic conductivity and this may be important in the phenomenon of interstellar polarization, as has been suggested by Schatzman and Cayrel, and also, more recently, by Wickramasinghe.