The effect of surface roughness on osteoblast proliferation, differentiation, and protein synthesis was examined. Human osteoblast-like cells (MG63) were cultured on titanium (Ti) disks that had been prepared by one of five different treatment regimens. All disks were pretreated with hydrofluroic acid-nitric acid and washed (PT). PT disks were also: washed, and then electropolished (EP); fine sandblasted, etched with HCl and H2SO4, and washed (FA); coarse sandblasted, etched with HCl and H2SO4, and washed (CA); or Ti plasma-sprayed (TPS). Standard tissue culture plastic was used as a control. Surface topography and profile were evaluated by brightfield and darkfield microscopy, cold field emission scanning electron microscopy, and laser confocal microscopy, while chemical composition was mapped using energy dispersion X-ray analysis and elemental distribution determined using Auger electron spectroscopy. The effect of surface roughness on the cells was evaluated by measuring cell number, [3H]thymidine incorporation into DNA, alkaline phosphatase specific activity, [3H]uridine incorporation into RNA, [3H]proline incorporation into collagenase digestible protein (CDP) and noncollagenase-digestible protein (NCP), and [35S]sulfate incorporation into proteoglycan. Based on surface analysis, the five different Ti surfaces were ranked in order of smoothest to roughest: EP, PT, FA, CA, and TPS. A TiO2 layer was found on all surfaces that ranged in thickness from 100 A in the smoothest group to 300 A in the roughest. When compared to confluent cultures of cells on plastic, the number of cells was reduced on the TPS surfaces and increased on the EP surfaces, while the number of cells on the other surfaces was equivalent to plastic. [3H]Thymidine incorporation was inversely related to surface roughness. Alkaline phosphatase specific activity in isolated cells was found to decrease with increasing surface roughness, except for those cells cultured on CA. In contrast, enzyme activity in the cell layer was only decreased in cultures grown on FA- and TPS-treated surfaces. A direct correlation between surface roughness and RNA and CDP production was found. Surface roughness had no apparent effect on NCP production. Proteoglycan synthesis by the cells was inhibited on all the surfaces studied, with the largest inhibition observed in the CA and EP groups. These results demonstrate that surface roughness alters osteoblast proliferation, differentiation, and matrix production in vitro. The results also suggest that implant surface roughness may play a role in determining phenotypic expression of cells in vivo.