Purpose To explore the appropriate method of mediastinal lymph node dissection for selected clinical stage IA (cIA) non-small cell lung cancer (NSCLC). Methods From 1998 through 2002, the curative-intent surgery was performed to 105 patients with cIA NSCLC who had been postoperatively identified as pathologic-stage T1. According to the method of intraoperative medistinal lymph node dissection, they were divided into radical systematic mediastinal lymphadenectomy (LA) group (n = 42) and mediastinal lymph-node sampling (LS) group (n = 63). The effects of LS and LA on morbidity, N staging, overall survival (OS) and disease-free survival (DFS) were investigated. Also, associations between clinicopathological parameters and survival were analyzed. Results The mean numbers of dissected lymph nodes per patient in the LA group was significantly greater than that in the LS group (15.59 +/- 3.08 vs. 6.46 +/- 2.21, P < 0.001), and the postoperative overall morbidity rate was higher in the LA group than that in the LS group (26.2 vs. 11.1%, P = 0.045). There were no significant difference in migration of N staging, OS and DFS between two groups. However, for patients with lesions between 2 and 3 cm, the 5-year OS in LA group was significantly higher than that in LS group (81.6 vs. 55.8%, P = 0.041), and the 5-year DFS was also higher (77.9 vs. 52.5%, P = 0.038). For patients with lesions of 2 cm or less, 5-year OS and DFS were similar in both groups. Multivariate analysis showed that lymph node metastasis was the unique unfavorable prognostic factor (P < 0.001). Conclusions After being intraoperatively identified as stage T1, patients with lesions between 2 and 3 cm in cIA NSCLC should be performed with LA to get a potentially better survival, and patients with lesions of 2 cm or less should be performed with LS to decrease invasion.