It's certainly true that, as measured by one's family tree, the denominator in any fraction must be a power of two. But a fraction could come very close to resembling one-sixth when going back many generations. If a person's 32 great-great-great grandparents consisted of 5 people of a given descent, he or she would be 15.625 percent of that ancestry.

Here, one-sixth is a stand-in for the 16 percent figure that Kari Stefansson reported. But no one is claiming that exactly one-sixth of Watson's ancestors are of African descent in the last hundred-some generations, since that's not possible.

Particularly given Stefansson's doubts about the quality of the Watson genome sequence, it's possible that this figure is too high. Perhaps (for example) the actual figure is closer to 12.5 percent, which would suggest one great-grandparent of African descent.

The folks at 454 technologies, which sequenced Watson's genome, are preparing a paper on their analysis. It will be interesting to see what they come up with.