A total of 123 patients were followed between January 1983 and July 1998 with 140 tooth-implant connected prostheses. The age of the patients at prosthesis installation ranged from 20 to 79 years (mean 51.8). 339 (Brånemark(R) system) implants were connected to 313 teeth. The loading time ranged from 1.5 to 15 years (mean: 6.5). 123 patients were randomly selected as a control group with freestanding implant-supported prostheses only. The age of the patients at prosthesis installation ranged from 22 to 78 years (mean 52.3). The loading time for the 329 freestanding (Brånemark(R) system) implants ranged from 1.3 to 14.5 years (mean: 6.2). Evolution of the marginal bone stability around the implant in the tooth-implant connected as well as the freestanding group was studied with respect to the prognosis of the implants. Over the period from 0 to 15 years, there was significantly more marginal bone loss (0.7 mm) in tooth-implant connected versus freestanding prostheses. No significant difference in marginal bone loss was found between the non-rigid tooth-implant connected prostheses versus freestanding prostheses. However, there was a significant difference in marginal bone loss for rigid and multi-connected tooth-implant connected prostheses versus freestanding ones. The results of this study indicate that more bone is lost around implants which are rigidly connected to teeth. This suggests that bending load, which is increased in tooth-implant connected prostheses, might be responsible for this phenomenon. These observations favor the use of freestanding prostheses whenever possible. However, the clinical significance of greater bone loss in rigid versus non-rigid connections might outweigh the annoying phenomenon of tooth intrusion in the case of non-rigid tooth connection, when connection is considered.