Background The prevalence rate of overweight and obese has been escalating over the past two decades in China. Even so, the association between obesity and stroke still remains unclear to some extent. Aims The aim of this study was to elucidate the association between body mass index and stroke in a large Chinese population cohort. Methods A cohort of 26607 Chinese people, aged over 35 years, was investigated in 1987. Baseline information of body weight and height was used to calculate BMI (weight in kilograms divided by height in meters squared, kg/m2). Cox proportional hazards model was fitted to estimate hazard ratios of stroke adjusted for age, educational level, smoking and alcohol consumption. Results The 11-year follow-up revealed (241149 person-years) a total of 1108 stroke events (614 ischemic, 451 hemorrhagic, and 44 undefined stroke). Body mass index30<bold></bold>0 was an independent risk factor for stroke both in men and women. Compared with normal weight, hazard ratios for total stroke were 0<bold></bold>74 in men underweight (95% confidence interval: 0<bold></bold>53 approximate to 1<bold></bold>03), 1<bold></bold>63 overweight (95% confidence interval: 1<bold></bold>35 approximate to 1<bold></bold>96), and 2<bold></bold>20 with obesity (95% confidence interval: 1<bold></bold>47 approximate to 3<bold></bold>30); and with ischemic stroke, hazard ratios were 0<bold></bold>52 in those underweight (95% confidence interval: 0<bold></bold>30 approximate to 0<bold></bold>89), 2<bold></bold>08 overweight (95% confidence interval: 1<bold></bold>65 approximate to 2<bold></bold>62), and 3<bold></bold>80 with obesity (95% confidence interval: 2<bold></bold>47 approximate to 5<bold></bold>86). In women, the corresponding hazard ratios for total stroke were 0<bold></bold>79 underweight (95% confidence interval: 0<bold></bold>58 approximate to 1<bold></bold>07), 1<bold></bold>42 overweight (95% confidence interval: 1<bold></bold>16 approximate to 1<bold></bold>73), and 1<bold></bold>57 with obesity (95% confidence interval: 1<bold></bold>06 approximate to 2<bold></bold>31); and for those with ischemic stroke, 0<bold></bold>92 underweight (95% confidence interval: 0<bold></bold>59 approximate to 1<bold></bold>43), 1<bold></bold>90 overweight (95% confidence interval: 1<bold></bold>44 approximate to 2<bold></bold>50), and 2<bold></bold>42 with obesity (95% confidence interval: 1<bold></bold>50 approximate to 3<bold></bold>93). There appeared an evident dose-response relationship between body mass index and the risk of developing stroke, which still appeared, however, adjusted low for hypertension, diabetes, and heart disease. Decreased risk for stroke in the leanest group was confined to men only. No association was found between body mass index and hemorrhagic stroke in both genders. Conclusions Our data suggest that body mass index was an independent risk factor for total and ischemic stroke but not for hemorrhagic stroke in both genders. Association between body mass index and stroke was extremely mediated by hypertension, diabetes, and heart disease. Decreased risk for the leanest group was confined to men.