Aims/hypothesis We investigated whether BMI predicts type 2 diabetes in twins and to what extent that is explained by common genetic factors. Methods This was a population-based twin cohort study. Monozygotic (n=4,076) and dizygotic (n=9,109) nondiabetic twin pairs born before 1958 answered a questionnaire in 1975, from which BMI was obtained. Information on incident cases of diabetes was obtained by linkage to nationwide registers until 2005. Results Altogether, 1,332 twins (6.3% of men, 5.1% of women) developed type 2 diabetes. The HR for type 2 diabetes increased monotonically with a mean of 1.22 (95% CI 1.20-1.24) per BMI unit and of 1.97 (95% CI 1.87-2.08) per SD of BMI. The HRs for lean, overweight, obese and morbidly obese participants were 0.59, 2.96, 6.80 and 13.64 as compared with normal weight participants. Model heritability estimates for bivariate variance due to an additive genetic component and non-shared environmental component were 75% (men) and 71% (women) for BMI, and 73% and 64%, respectively for type 2 diabetes. The correlations between genetic variance components (r g ) indicated that one fifth of the covariance of BMI and type 2 diabetes was due to shared genetic influences. Although the mean monozygotic concordance for type 2 diabetes was approximately twice the dizygotic one, age of onset of diabetes within twin pair members varied greatly, irrespective of zygosity. Conclusions/interpretation A 28-year follow-up of adult Finnish twins showed that despite high trait heritability estimates, only a fraction of covariation in BMI and incident type 2 diabetes was of genetic origin.