To examine potential confounders and dose-response data for the association between smoking and menopause, we used both a cross-sectional and case-controlled approach. In total, 10,606 middle-aged women residing in eastern Massachusetts were surveyed about their age at menopause and smoking history; 344 women (cases) with natural menopause prior to age 47 and 344 age-matched women (controls) who were still menstruating or who had a menopause after age 46 were selected for further study. Risk for menopause was assessed by Kaplan-Meier, Cox proportional hazards, or logistic regression models. From cross-sectional data on 8657 women aged 45-54, the hazards odds ratio for a natural menopause among women who ever smoked compared to non-smokers was 1.31 (95% C.L. 1.21-1.42) and among women who had accumulated 30 or more pack-years was 1.87 (95% C.L. 1.67-2.04) after adjustment for parity and weight. An additional potential confounder from the case-controlled study was lower educational attainment, and after adjustment for this variable, significant trends persisted for risk of early menopause associated with age began smoking (P = 0.03), years of smoking (P = 0.01) and pack-years of smoking (P = 0.03). This study demonstrates an association between smoking and early menopause in both cross-sectional and case-controlled data that is not confounded by parity, weight, socio-economic status, or nutritional variables.