I stumbled upon this graph a week ago while looking at the supplementary information for the 2002 Gabriel et al. paper on the hapmap. Supplementary info isn't printed with the paper and you have to go to the website to look at it.

The layman's summary: you don't really need to know what F_st is to appreciate the significance of these plots. These graphics show whether allele frequencies between populations are positively correlated or not.

For example, consider base pair 100 on chromosome 1. Suppose that 30% of self-identified Yorubans have a G there and 70% have a C. With high probability, self-identified African Americans will have close to a 30/70 split, while essentially *no conclusion is possible* about what the G/C split will be in self-identified Europeans at that locus.

You can see this from the graph, as the Yoruban and African-American sample frequencies are very tightly correlated (fourth panel). However, the African-American and European sample frequencies are all over the map. Furthermore, the frequency differences between African and non-African populations are moderately correlated (fifth panel).