The apparent activation energy for existing grain boundary diffusion data shows a tendency to decrease with decreasing temperature. Such a decrease may conceivably be due to at least two causes: (1) The grain boundary diffusion must involve several different types of atomic jumps in the core region, each of which is characterized by a different activation energy. The resulting diffusion is then a weighted average over these jumps. As the temperature is lowered, the jump paths with higher activation energies become progressively ''frozen out.'' If the paths are connected in a parallel arrangement (in analogy to an electrical circuit) the averaged effective activation energy decreases as the temperature is reduced. (2) Experimental results were obtained using polycrystalline samples containing a spectrum of different types of grain boundaries. The overall observed diffusion is, therefore, a weighted average over these boundaries. As the temperature is lowered, the boundaries with the higher activation energies become progressively frozen out and the averaged effective activation energy decreases. (GHT)