Background: Compared to the majority population of a host country, migrants tend to have different health risks and health behaviour. We have hypothesised that these differences diminish with time passed since migration. We tested this hypothesis by examining smoking behaviour among Turkish migrants and their children born in Germany (second-generation migrants), stratified by educational level and, for the first generation, length of residence. Methods: We estimated the prevalence of smoking based on the representative 2005 Mikrozensus, an annual survey including 1% of Germany's households. The 2005 Microcensus was the first to provide information that makes it possible to differentiate between first- and second-generation Turkish migrants. In total, 12,288 Turkish migrants and 421,635 native-born Germans were included in our study. The unit non-response is generally low (about 6% in 2005) because participation in the Microcensus is obligatory. Results: We found the prevalence of smoking in second-generation male Turkish migrants to be lower than in the first-generation group: 47.0% of first-generation Turkish migrants with a high level of education were smokers, in contrast to only 37.6% in the second generation. Within the German reference population, 29.9% were smokers. The percentage of Turkish women in our sample who smoked was generally smaller, but was not significantly lower in the second generation. In fact, the prevalence of smoking was highest among Turkish women of the second generation with a low level of education (40.9%), similar to younger second-generation German women with the same level of education. Conclusion: We present the first representative data on changes in the prevalence of a risk factor for many chronic diseases among Turkish migrants in Germany. Male Turkish migrants (with a high level of education) showed a decrease over the generations while smoking prevalences of female Turkish migrants increased. In the second generation, prevalences partly converged with those of the German reference population or were even higher. Our hypothesis – that migration can be interpreted as a ''health transition'' – was thus partly confirmed.