Three-field lymph node dissection has been widely used to treat thoracic esophageal cancer, but is very invasive and can cause serious complications. Whether cervical lymph node dissection should be performed in all patients with thoracic esophageal cancer remains controversial. We pathologically examined the recurrent nerve lymph nodes during surgery in patients with thoracic esophageal cancer to determine the presence or absence of lymph node involvement. In patients without recurrent nerve nodal involvement, cervical lymph node dissection was not performed. Treatment outcomes were analyzed to evaluate whether intraoperative pathological investigation was a useful procedure. Among 71 patients with thoracic esophageal cancer who underwent 3-field lymph node dissection, the rate of cervical lymph node metastasis was 40.9% in patients with recurrent nerve nodal metastasis on intraoperative pathological investigation, as compared with 10.2% in patients without recurrent nerve nodal metastasis (p=0.007). Multiple logistic-regression analysis showed that recurrent nerve nodal metastasis was a strong predictor of cervical lymph node metastasis (odds ratio, 2.98; 95% confidence interval, 1.139-7.775; p=0.03). Among 41 patients who underwent intraoperative pathological investigation, 10 had recurrent nerve nodal metastasis and underwent cervical lymph node dissection. Two of these patients had histological evidence of cervical lymph node metastasis. The remaining 31 patients had no recurrent nerve nodal metastasis on intraoperative pathological examination and therefore did not receive cervical lymph node dissection. None of these patients had cervical lymph node recurrence on follow-up. We compared patients who underwent intraoperative pathological investigation with those who underwent conventional 3-field lymph node dissection (without performing intraoperative pathological investigation). The rates of cervical lymph node recurrence were similar among the groups (2.6% vs. 6.7%), but the 3-year survival rate was significantly higher in the patients who underwent intraoperative pathological dissection (83.3%) than in those who underwent 3-field dissection (57.2%; p<0.05). Although this was a retrospective study, our results suggest that outcomes of patients undergoing cervical lymph node dissection according to the results of intraoperative pathological investigation are at least as good as those in patients undergoing 3-field lymph node dissection. We conclude that intraoperative pathological investigation of recurrent nerve nodal metastasis is useful for determining whether cervical lymph node dissection should be performed in patients with thoracic esophageal cancer.