Recently on PBS's "African American Lives," host Henry Louis Gates had his DNA tested to learn about his ancestry. Gates' family suspected its paternal ancestry could be traced to a white slave owner. But DNA testing showed that his Y chromosome did not match the man's white descendants. A second, newer test gave Gates another result he didn't expect: His DNA showed that only half of Gates' ancestry was African. The rest were apparently European.
DNA testing for genealogy has become increasingly popular, as a Newsweek cover story in February attests. Especially attention-getting have been efforts to trace genetic relationships along the male lineage. In January the New York Times wrote up attempts to trace Irish genealogy through the male line to Niall of the Nine Hostages, a fifth-century Irish warlord. Other tests have also shown that as many as 14 million men may share the Y chromosome of Genghis Khan. But tests that seek a single, Y-chromosome male lineage are limited: They leave out the vast majority of ancestors. Newer tests can survey all the DNA that can be inherited from either parent, but at a cost of precision: They don't tell which ancestors lived where, and they can't detect traces of ancestry.
The newer "genetic admixture tests" examine DNA from genes inherited from all of a person's grandparents, great-grandparents, and so on. A few of these genes reflect the part of the world where those ancestors lived. Like postcards, they track the movement of people from the lands of their ancestors to their current address. Scientists studying these genetic variations now focus on sites that vary between people by one chemical letter. They're called "single nucleotide polymorphisms," or SNPs. Some of these SNPs are important: They may contribute to traits like skin color or resistance to regional diseases like malaria. Others vary among populations just because of chance.
For geneticists, finding the SNPs that mark populations is a challenge. For the most part, the same SNP might be found in Africans, Europeans, and people from every other part of the world. It's now possible to test quickly for hundreds of SNPs by using special microchips that bind to the distinctive DNA sequences. These tests examine hundreds of SNPs at once; if among these a person has many that are common in Africa, it is likely that she has some African ancestors.
Admixture testing works best in groups like African-Americans, whose ancestors in Africa and Europe lived far from each other. Most of the ancestry of today's African-Americans can be traced to West or Central Africa, with a minority from other parts of the continent. (Gates' family is a bit exceptional in terms of origin.)
But for other groups things can get a lot more complicated. Many amateur genealogists are interested in whether they might have a Cherokee ancestor, for example. And for some people, admixture tests can give a relatively accurate answer about Native-American ancestry. But other people, including Greeks and Ashkenazi Jews, may have "Native American affinity," according to the tests, even if they and their ancestors have never been to America. As far as anthropologists know, there were no lost tribes connecting Greeks, Jews, and ancient Americans. So, maybe this "Native American affinity" reflects the scattering of alleles by prehistoric Asian nomads to the ancestors of Greeks and Jews as well as to American Indians. Maybe the SNPs that they share gave these groups a leg up in fighting diseases.
All we know for sure is that such genetic similarities can make ancestry testing very confusing. Suppose a person of mostly German ancestry discovers that his DNA has 6 percent Native-American affinity. Does he have a Native-American ancestor, or a Greek or Jewish ancestor, or all three? There's only one way the 6 percenter can know for sure: He has to know most of his genealogy already.
From a practical point of view, that is the biggest problem with today's genetic genealogy tests. In many cases, they can't tell you what you don't already know. And unlike DNA fingerprinting tests with error rates of one in a billion or less, the chance of misidentifying ancestral groups in these genealogy tests may be 5 percent or higher. With this chance of error, the test won't be wrong about a full Native-American grandparent, but it might be wrong about a great-great grandparent. In addition, SNPs that separate central Africans from northern Europeans aren't nearly as good at separating Ethiopians from Arabs. So, in the test results of some African-Americans, European means Europe, while in others, it may mean East African, or Arab, or Indian. Depending on where his African ancestors came from, Gates' apparently European origins might lie somewhere else entirely.
A deeper problem with admixture testing is its claim to identify the "ancestral components" of different populations. For example, admixture testing considers people from India to be a mixture of "Indo-European" and "East Asian" ancestors. And indeed, Indians have some alleles otherwise common in Europe, and some otherwise common in China. But Indian populations have been on their subcontinent for tens of thousands of years, and they have many alleles that don't come from anywhere else. Anthropologists studying genetic variation have always found complexity rather than simple one-plus-one racial mixtures. SNP-testing companies don't seem to have gotten that news.
SNP-based tests can help you find out where your great-grandfather came from. But his distant ancestors ultimately came from other places. So, most of your genetic ancestry will still be a question mark, no matter how many tests you shell out for.