The recruitment of scleractinian corals to settlement tiles is widely used to infer relative rates of recruitment to natural reef surfaces. On tiles fixed approximately horizontally, the majority of corals settle to lower surfaces, and it is assumed that this distribution reflects the benefits of avoiding algal competition, grazers, and sedimentation on upper surfaces. Using settlement tiles with and without refuges (shallow depressions, 2–10 mm in width and depth), we tested the hypothesis that coral recruits are not found on the upper surfaces of smooth settlement tiles because these surfaces lack refuges suitable for small colonies. Tiles were deployed for 4 to 14 months over 2 years in near-horizontal orientations on shallow reefs (~5-m depth) of one Caribbean and six Pacific islands, and following collection, tiles were inspected for coral recruits. In eight of nine deployments and in both regions, densities of recruits in upward-facing refuges were indistinguishable from densities on lower surfaces, or were several-foldhigherin upward-facingrefuges. In the ninth deployment (lasting14 months in the Caribbean) refuges were occluded by sediment and calcareous algae, and on these tiles densities of coral recruits were higher on the lower surface. Together, these results expand from a local-scale (Nozawa, 2008) to a global-scale the conclusion that the highest area-normalized density of coral recruits occurs in upward-facing refuges. An important implication of this finding is that coral recruitment on upward-facing natural reef surfaces will be increased by refuges that can be occupied by small corals.