Variation in mutation rate, attributed to differences in both generation time and in metabolic rate, has been invoked under the neutral theory of molecular evolution to account for differences in substitution rate among mammalian lineages. We show that substitution rates at fourfold-degenerate sites and at sites in noncoding regions do not vary between the primate and rodent lineages, implying mutation-rate uniformity. In contrast, the substitution rates at nondegenerate sites vary both within and between lineages. This difference in substitution-rate pattern between the two types of site is incompatible with neutral theory but may result from substitutions occurring by fixation of slightly deleterious mutations. Variation in the rate of protein evolution among mammalian lineages appears to be due more to differences in population fixation rates than to biochemical or physiological differences affecting mutation rates.