OBJECTIVE Since the early 1980s to relieve functional disturbances after rectal excision, we have been performing nerve-sparing surgery for rectal cancer. The aim of this study was to analyze patterns of recurrences, especially concerning causes of local ones. Furthermore, we would like to address the criteria we used in patient selection to effect successful nerve-sparing surgery. METHODS From 1982 to 1991, 306 patients underwent nerve-sparing operations, which may be categorized into three types: 1) total autonomic nerve preservation (125 cases), 2) complete pelvic nerve preservation (105 cases), and 3) partial pelvic nerve preservation with removal of parasympathetic nerve (79 cases). Single and multivariant regression analyses were conducted to investigate patterns of recurrence, especially causes of local ones. RESULTS Sixty-five patients (21 percent) developed recurrent tumors, 19 of which (6.2 percent) were local. Using Dukes terms, there were five patients with Dukes A 13 with Dukes B, and 47 (35 percent) with Dukes C stages. Rate of local recurrences was 13 percent in patients with Dukes C tumor. According to single-variant analysis of Dukes C patients, the following factors are thought to influence local recurrences: number of lymph nodes metastases, level of primary growth, and direction of lymphatic spread. Multivariate regression analysis suggested that lymph node metastasis was the most important and influencing factor on local regrowth (P < 0.002). CONCLUSIONS Compared with local recurrences is so-called extended surgery appeared to be lower. Our current policy is aggressive application of nerve-sparing surgery, even to patients with node-positive rectal cancer, taking into consideration the exact extent of cancer spread. From the viewpoint of neuroanatomy related to mesorectum, we discussed patient determination for our nerve-sparing surgery.