BACKGROUND Anterior cantilever resin-bonded prostheses fail as a result of a labio-lingual peeling action, which creates a stress concentration within the adhesive layer. OBJECTIVE The purpose of this study was to identify the factors that determine the retention of an anterior resin-bonded prosthesis and to seek to eliminate the stress concentration within the adhesive layer by fundamentally altering the prosthesis design. METHODS The first experiment involved 40 Ni/Cr (Wiron 99) beams with a width of 5 mm, thickness of 0.5 mm, and lengths ranging from to 13 to 22 mm. The beams were cemented onto a block of the same material using an adhesive resin luting agent (Panavia 21). The length of the beam that was bonded ranged from 1 to 10 mm, resulting in a bonded area ranging from 5 to 50 mm(2). A load was applied onto the cantilevered portion of the beam 2 mm from the end, causing a peeling action. The force (N) required to debond these beams was measured using a pull-to-fracture test. Subsequently, a second experiment was undertaken, and 7 beams with an altered point of attachment (new design) were tested. The new design had the point of attachment of the cantilevered portion located centrally on the bonded area of the beam. Implementing this new design clinically would result in a cantilevered resin-bonded fixed partial denture that would have the connector arm attached more centrally on the retainer wing. The data were analyzed using a 1-way analysis of variance (alpha = .05), and a Tukey pairwise comparison test was used when the results was statistically significant. Two finite element analysis (FEA) models, one simulating the first experimental design and the other simulating the new design, were created. A load was then applied on the cantilevered portion of the beams similar to the experimental models, and the stress patterns were examined. The numerical values of these resultant stresses were plotted graphically. RESULTS The direction of load application, which may be transferred to a clinical setting as labio-lingual forces, was identified as the dominant force responsible for debonding. The new design, which addressed this problem, showed a significant increase (P < .001) in retention. The FEA models identified the stress concentrations within the adhesive layer of the traditional design, which were eliminated when the new design was tested. CONCLUSIONS For the in vitro model, loads that may be interpreted clinically as labio-lingual forces resulted in the lowest forces required to cause debonding, and these forces were independent of the surface area of bonding. Altering the point of attachment of the cantilevered portion onto the retainer caused a significant increase in the forces needed to cause debonding.