Lewontin argued that there is more genetic variation within human populations than between them, so they can’t really be very different. Of course that’s bullshit: humans have the second-largest morphological variation of any mammal, behind only dogs. How do you get significant phenotypic differences between populations when there isn’t much Fst between them?
One way (at the extreme) is to have a single allele, one that does a lot, vary strongly between populations: at the limit, be fixed in one and nonexistent in the other. There are not a lot of alleles like this in humans, but it happens. EDAR 370A is almost fixed in Northeast Asia, almost nonexistent in Europe and Africa: it results in thicker scalp hair, more numerous sweat glands, smaller breasts, funny teeth, and changes in the shape of the ears and chin. Just one allele: it would not show up noticeably in Fst.
The way at the other extreme consists of small, systematic differences in the frequencies of many alleles that affect a particular trait. Suppose that 100 alleles influence height – there are plus and minus versions of those alleles, with the plus versions slightly increasing height while the minus versions slightly decreasing height. A systematic, smallish increase in the fraction of plus variants would make a population average taller: quite possibly a lot taller. But this only involves differences in a tiny fraction of the entire genome – the GWAS hits for height – so again, no noticeable change in Fst. You get these correlated changes from natural selection. For that matter, sweeps like EDAR 370A are also products of natural selection.
According to Lewontin’s argument, Pygmies couldn’t really be hugely shorter than most other human populations. Yet they are short (five or six stds shorter), while Lewontin’s Fst argument is simply wrong. In fact the existence of Pygmies automatically disproved Lewontin’s argument, whether we understood the exact genetic basis or not. Counterexamples do that.
Understanding the genetic basis of the phenotypic differences between populations can often tell us something about the causes of those differences – for example the time frame.. You’d expect that the main lactose-tolerance alleles aren’t terribly old, since domestication of milk-giving animals happened relatively recently (within the last 10,000 years or so) . And they’re not terribly old. On the other hand, adaptations to cold and high altitude might be very old, even going back to archaic humans – and some are. Of course the phenotypic differences exist (and sometimes matter) even if we don’t understand the past selective pressures that created them.