We thank Hemminki and coworkers for their interest in our recent findings concerning parental transmission of type 2 diabetes . Based on their and our data [1, 2], they concluded that paternal and maternal history of diabetes simply add rather than act multiplicatively to increase risk for diabetes in offspring. They do so because in the Swedish model, a fully multiplicative model was rejected on statistical grounds. We agree that based on these data, it is very unlikely that paternal and maternal transmission of diabetes act in an entirely multiplicative manner. However, both in their and in our study, the relative risk predicted by an additive model (3.7 for their study and 3.4 for our study) is lower than the actually observed risk (4.3 for their study and 3.9 for our study). Earlier findings were also consistent with these results, with relative risks of 3.6 in Pima Indians and 5.9 Framingham Offspring predicted by an additive model while actually observed risks were higher, with values of 3.9 and 6.1 respectively [3, 4]. Thus, data of all four studies to date are in the same direction, with a higher actually observed risk in subjects with two affected parents than predicted by an additive model. Therefore, given that a multiplicative model does not fit the data, acceptance of an additive model as sole valid alternative seems an oversimplification, with the truth lying in between both of these extremes. It is important to acknowledge the possibility of nonadditive genetic effects (epistasis or gene–gene interactions) of individual loci playing a role in transmission of risk of type 2 diabetes, because this is considered a source of missing heritability in complex diseases, including type 2 diabetes . As many low-penetrance loci may contribute to genetic susceptibility for type 2 diabetes , such a scenario does not seem unrealistic.