In a previous experiment, it was demonstrated that a wide marginal defect around an implant can heal with a high degree of osseointegration. The present experiment was performed to evaluate the degree and quality of de novo bone formation and osseointegration in marginal defects adjacent to submerged titanium implants. All mandibular premolars and 1st molars were extracted in four Labrador dogs. Four experimental sites were identified in the right side of the mandible. In two sites, custom-made implants with a sandblasted, large grit, acid-etched (SLA) surface were installed without further ostectomy (control sites). In the two remaining sites (test sites), a specially designed step drill was used to widen the marginal 5 mm of the canal. A barrier membrane was used to cover the implants in the defect sites. All implants were submerged. One month later, an identical procedure, including site preparation and implant installation, was performed in the left side of the mandible. Two months following the first implant installation procedure, biopsies were collected and prepared for sectioning. Ostectomy and implant installation in the control location resulted in a series of bone tissue alterations which eventually allowed newly formed bone to establish contact with the SLA surface. The marginal defect lateral to the implant in the test locations gradually became filled with newly formed bone. De novo bone formation started within the walls of the surgically prepared defect. Bone-to-implant contact was first established in the apical portion of the gap. This new bone tissue was in the coronal direction continuous with a dense, non-mineralized 'implant attached' soft tissue which, over time, also became mineralized to increase the height of the zone of bone-to-implant contact. The results suggest that healing of a wide marginal defect around an implant is characterized by appositional bone growth from the lateral and apical bone walls of the defect.