The quest for genes associated with diseases is widely recognized as an essential task in the effort to investigate the genetic basis of complex human disorders and traits. A basic stage in association studies is the careful choice of the model population, with preference to closed groups having little population substructure. Here, we show evidence for significant geographic substructure (P=0.017) of the maternal lineage represented by mitochondrial DNA variation in one of the most commonly studied populations, the Ashkenazi Jews. Most of the substructure effect stems from differential representation of haplogroups K and H. Our results underline the essentiality of adjusting data of population genetic variation for substructure during the design of association studies, even in apparently closed populations.