Endoscopic endonasal reductions have been addressed in 63 patients with blowout fracture of the medial orbital wall since 1992. The operations were carried out under general anesthesia with a magnified operative space projected on a television monitor by a charge coupled device video camera attached to the endoscope. The middle nasal turbinate was fractured toward the nasal septum, the uncinate process was cut off, and the bulla was opened. The ethmoidal bony partition and the mucous membrane were removed; however, the fractured bone chips of the medial orbital wall were preserved. The herniated orbital contents were pressed back into the orbital cavity, and the medial wall was set with 2-mm-thick bent silicone plates placed in the ethmoidal sinus. The plates were removed in the outpatient clinic 2 months after the operation. The surgical results of 21 patients treated with endoscopic reduction were compared with those of four patients treated with transfacial reduction with an iliac bone graft. All of the patients had isolated medial wall fracture and became aware of diplopia within 15 degrees in any direction from the primary position (straight gaze) before the operation; the follow-up period covered 6 months. The patients were classified into two categories according to postoperative double vision: "good," indicating no double vision or diplopia of more than 45 degrees, and "poor," diplopia of less than 45 degrees. Improvement of diplopia was observed in all patients without any complication. Of the 21 patients who underwent endoscopic reductions, 17 were classified as "good" and four as "poor." On the other hand, of the four patients who underwent transfacial reductions, three were classified as "good" and one as "poor." Significant differences were not observed between the surgical results of our two methods. Endoscopic endonasal reduction showed greater aesthetic advantages and, moreover, required no grafting. This technique is suggested as one of the most reasonable treatments of medial orbital wall fractures.