Smoking is well known as a risk factor for esophageal cancer, but controversial as a prognostic factor. Moreover, evidence is scarce that a dose–response relationship exists. We conducted a retrospective study on the effect and dose–response relationship of prediagnostic smoking on the postoperative disease-specific survival of patients with lymph node-negative esophageal squamous cell carcinoma (ESCC). We enrolled 643 patients with lymph node-negative ESCC who had undergone esophagectomy between 1990 and 2005 at the Department of Thoracic Surgery, Sun Yat-sen University Cancer Center, Guangzhou, China. The patients' demographic, pathological, preoperative and cancer outcome data were obtained from medical records. These data were reviewed and analyzed using life table, Kaplan–Meier analysis and multivariate Cox regression. A significant reduction in 3- and 5-year survival rates was observed in smokers with lymph node-negative ESCC compared with those in non-smokers. The 3- and 5-year survival rates were 54% and 46% for smokers, and 67% and 64% for non-drinkers, respectively (P < 0.05). Multivariate Cox analysis revealed that smoking was an independent prognostic factor (P = 0.008, hazard ratio = 1.404). Both log-rank test (P = 0.065) and multivariate analysis (P = 0.091) showed no significant difference between the survival rates of light and heavy smokers. Prediagnostic smoking is an independent prognostic factor for patients with lymph node-negative ESCC, but the dose–response relationship needs further investigation.