A consequence of the recent origin of the genetic differences between Icelandic subpopulations is that allele frequency differences follow the null distribution predicted by neutral drift. Thus, there is little risk of false positive associations due to population stratification in disease association studies, despite the fact that there are genuine differences between regions. The same conclusion may be expected for other populations whose structure has arisen from recent genetic drift [29]. On the other hand, such populations are not well-suited for the detection of regionally specific natural selection reflected in unusual differences between subpopulations. For that purpose, subtly structured populations whose structure is due to more ancient population divergence, with large population sizes minimizing subsequent genetic drift, offer the greatest promise. For example, European American subpopulations exhibit unusual differences at the LCT, HLA and OCA2 loci that lie outside the null distribution with genome-wide significance ([13] and A.L. Price, unpublished data). The distinction between population differences attributable to recent drift and those arising from more ancient divergence is also likely to be of interest in studies of other subtly structured populations [22],[28],[30].
