Purpose: This prospective clinical trial examined the effect on teeth and implants when rigidly or nonrigidly connected in a cross-arch model. Materials and Methods: Thirty patients received 2 implants, 1 on each side of the mandible, and were restored with 3-unit fixed partial dentures connected either rigidly or non-rigidly to an abutment tooth. Patients were followed for at least 5 years post-restoration. Results: Repeated-measures analysis revealed no significant difference in crestal bone loss at implants (rigid versus non-rigid methods). An overall significant difference (P .001) was found comparing methods for teeth. Paired t tests revealed no significant differences in crestal bone levels for implants or teeth at the 5-year recall. Kaplan-Meier methods and the Cox proportional hazards model showed no differences between attachment methods with regard to success based on survival and bone loss criteria. During the 5-year recall period, 1 implant (rigid side) was removed. Four implants developed bone loss greater than 2 mm during the course of this trial. One tooth on the rigid side and 2 teeth on the non-rigid side had greater than 2 mm of crestal bone loss and were removed secondary to fractures. In all, 5 abutment teeth were removed, all of which had been treated with root canal therapy and fractured at the interface of the post within the tooth. There was no clear relationship of tooth fracture to attachment. Repeated-measures analysis of mobility values revealed no significant changes over the time course of this study, and paired t tests revealed no statistically significant differences between implants for mobility. Repeated-measures analysis and paired t tests for probing depth revealed no significant changes over the time course of this study. There were no significant differences in soft tissue indices for either attachment method. The percentage of patients who had measurable intrusion was 66% for the non-rigid group, and 44% for the rigid group; 25% of the non-rigid teeth had greater than 0.5 mm intrusion, compared with 12.5% for the rigid group. For the 2 time periods evaluated, there was no significant increase in intrusion over time. The non-rigid-side implant required more nonscheduled visits to treat problems than the rigid implant and the teeth. Discussion: Most patients were treated successfully with rigid or non-rigid attachment of implants to teeth. Conclusion: The high incidence of intrusion and non-scheduled patient visits suggest that alternative treatments without connecting implants to teeth may be indicated.