Background: Faecal incontinence is difficult to treat. A variety of reconstructive procedures has been described, but none is entirely satisfactory. This study evaluated the feasibility of cross-innervating a skeletal muscle neosphincter with the pudendal nerve in a canine model. Methods: Thirty dogs were rendered surgically incontinent (the pudendal nerve was cut and the external sphincter was partially excised). A neosphincter was then created using the semitendinosus muscle. In ten dogs pudendal nerve transposition (PNT) to the nerve to the semitendinosus muscle was performed. Ten dogs were given a dynamic neosphincter by inserting a pulse generator aT 6 weeks. The remaining ten dogs served as controls with passive semitendinosus wraps. Anal manometry was performed before operation and monthly for 5 months. Muscle biopsies, performed at the initial operation and at 5 months, were stained for slow- and fast-twitch fibres, and were examined histologically. Results: At 1 month, mean sphincter function was 32 per cent of the preoperative value in the control animals, 34 per cent in the PNT group and 27 per cent in the electrostimulation group; all dogs were incontinent. At 5 months the mean recovery of sphincter function was 42 per cent of the preoperative value in controls, 100 per cent in dogs with PNT (P<0·001) and 63 per cent in dogs having electrostimulation (stimulator on) (P = 0·02). Six dogs with PNT had squeeze pressures equal to or greater than preoperative levels. At 5 months the ratio of slow to fast fibres was significantly greater in all dogs (control P = 0·01, PNT P < 0·005, electrostimulation P < 0·001). Conclusion: Use of the pudendal nerve to innervate a canine skeletal muscle anal wrap produced a functional anal sphincter that was superior to electrically stimulated and passive wraps.