We describe two complementary studies of diamond growth by chemical vapor deposition. In the first, the early stages of growth of randomly distributed nuclei on silicon are studied by scanning tunneling microscopy. For growth times from 1 to 30 min nearly all crystallites are three dimensional, and increase in volume as t1.5. Although this result could be interpreted in terms of diffusion limited growth, the conditions for diamond CVD are more consistent with rate limited growth where the crystals are expected to gain volume as t3. This anomaly can be explained in terms of a two-species growth mechanism in which the rate constant for carbon addition is proportional to the diffusion limited flux of atomic hydrogen. Other mechanisms giving rise to the observed t1.5 dependence are also considered. The second study uses both scanning electron and tunneling microscopies to examine the morphology of a boron-doped film homoepitaxial to the {100} surface of natural type 2a diamond. In regions distant from gross defects, this film is very smooth. However, gross defects appear to initiate growth of new epitaxial layers at a rate much higher than in defect-free regions. This observation suggests that diamond growth is promoted by “enabling defects” and that without such defects nucleation of new layers is a slow process and permits layer-by-layer growth at a much lower rate.