Recently, however, genealogists became aware of a new tool that could provide evidence for such hypotheses: DNA testing. When I learned of this new technology, I realized that such a test could show whether the early Rockwell settlers in the American colonies were related. Besides William and John in Windsor, John of Stamford, and Josiah of New London and Norwich, there was a Robert Rockhold who settled in Virginia and Maryland. Some early records call him Robert Rockwell, and some of his descendants took the Rockwell surname.
In 2001, I proposed that the Rockwell Family Foundation sponsor a DNA test. The Board of Directors approved spending $6,000 to cover about 35 samples, and we recruited several representatives of the three Connecticut lines, the Rockhold family and descendants of particular individuals of interest, whose ancestries were either unknown or in dispute.
The science behind the Project
DNA is well known as the conduit of our genetic natures. It resides within the nucleus of every cell in your body, carrying the genes passed down from both parents, determining everything from eye color to susceptibility to certain diseases. So it is remarkable that most of the DNA is �non-coding��it apparently does nothing at all. It�s �junk DNA.� This is particularly true of the DNA in the Y chromosome. While the X chromosome is packed with genetic code, the Y chromosome codes for maleness�and not much else. Furthermore, while the �junk DNA� in the other chromosome pairs mixes into new combinations with each generation, the junk DNA in Y shows little mixing at all with its counterpart, X. As a result, it is passed virtually unchanged from father to son. This is what makes the Y chromosome especially useful to genealogists.
Just what is being looked at in a DNA surname test? Researchers have noted that in numerous locations on the Y chromosome, a string of amino-acid base pairs may repeat itself several times. This string is called a �short tandem repeat,� or STR. The exact number of repeats may differ between individuals, usually over a specific range, for example 9 to 14. Each possible count is called an �allele.� If the allele at a given position (locus) tends to stay the same from generation to generation, the STR is considered a useful �marker� for testing relationships between individuals, for it will get passed down from fathers to sons over the generations, and cousins on different branches of a family will show the same allele. While two men may have the same alleles on some of these markers, a whole ensemble of matches on different markers points to a close relationship. Usually, a father and son will have the exact same alleles in each marker.
I say �usually� because now and then, during the process of copying and recomination at the time of conception, an imperfect copy of an STR is made and the marker picks up an extra repeat or two�or loses one or more. Then the son will differ from the father at that one locus, though the rest are the same. Such �mutations� happen only occasionally. The rate of mutation on different markers may vary�it�s still early in the study of such things�though an average of one mutation every ten generations has been suggested. But these mutations occur at random, so in a given paternal line, the ensemble of alleles (called a �haplotype�) may remain the same for many more generations than ten, or a mutation may occur on two different markers within a few generations. But due to the relative stability of the haplotype, genealogists are able to learn whether two men of the same surname are paternally descended from a relatively recent common ancestor.
Just as the STRs at some loci change so frequently that they are not useful as markers, others mutate so slowly that a given set of alleles at these markers tend to indicate a general sector of the population: a large group of anciently related peoples, called a �haplogroup.� Population geneticists have studied genetic clues of various types to reconstruct the ancient history of humanity and their migrations across the globe, and have detected several waves of migration into Europe. Correlation between these findings and certain Y chromosome alleles allows the DNA test results to connect one�s paternal line with one of these ancient populations.
The results were quite successful, as they established that the three old Connecticut families were indeed closely related. In recent years, as the number of markers tested has risen to as many as 43, there have been samples from different lines (William and John) who matched on every marker. But some samples show mutations from the "normal" values found, and these may help to distinguish different branches. One mutation consistently shows up among descendants of Jonathan Rockwell of Ridgefield, son of John of Stamford, and not among descendants of his brother Joseph. This may indicate that the mutation occurred in Jonathan himself. Meanwhile, the Rockhold family (some branches of which have the Rockwell surname) shows a very different haplotype. This will help descendants of Rockwells whose lineage is incomplete to determine whether they should be looking toward old Connecticut or the Pennsylvania/Maryland base of the Rockholds.
On August 1, 2007, Relative Genetics announced some changes to some of the markers used in the test. The alleles for a given locus were adjusted to reflect new standards in the field. The upshot for us is that the alleles determined during past tests have changed by a set amount:
Another nomenclature change, reported Aug. 24, 2007, results in the increase by 1 to alleles at DYS-441. Thus, the standard value for the Rockwell haplotype has been changed from 13 to 14. Here's the announcement.
Soon after the above announcements, Relative Genetics was sold to Ancestry.com, which began offering DNA testing. Our project thus migrated to dna.ancestry.com, but not all of the original samples were loaded at the new page there. Reportedly, Ancestry tried to reach all participants through available e-mail addresses, and only loaded results for those donors who responded. Several of our participants did not have e-mail or had changed addresses.
May 12, 2010: After the Project moved to Ancestry, some of the above changes in marker definition seem to have been overruled! In particular, Y-GATA-H4 is back, not "H4.1" anymore, and its current values are back to 11 or 12. This is vexing, since I upgraded the values for the 2nd and 3rd editions. Please subtract 10 from the totals you see in the table on page 421 (2nd ed.) or page 463 (3rd ed.)!
A recent upgrade of the sample for a descendant of Uriah Rockwell (column D in the table accompanying the DNA essay) shows the standard alleles for all values; it matches columns A and R. This has implications for the question of John of Butternuts (column K). If a 43-marker test were done on descendants of John and Simmons, and a mutation appeared in both, this would be evidence that Simmons was his father, while lack of a mutation in John that appeared in Simmons's haplotype would indicate that Uriah was his father. Any qualified volunteers?
We can also use volunteers from other branches of the family. An outstanding question, mentioned in the book, is: at what point did the DYS-239 mutation arise. The early results suggested it was confined to the descendants of Jonathan, son of John of Stamford, but then it turned up in our first (and as of April 2010 only) sample from the Thomas branch, via the Jabez Rockwell clan in Butternuts, N.Y. Since descendants from Joseph, the third son of John, consistently lack this mutation, we have two alternative theories:
And on the off-chance that descendants of the William and Josiah lines are reading this, we'll be happy to recruit new donors from these as well! And you don't have to go through Ancestry.com, since the various testing companies use many of the same markers, including DYS-439.
Here is an outline of the early generations in various William-line descents, showing how many donors from each lineage known to have male descendants have participated in the project, and where there are few or no samples so far. Thanks to Mark Rockwell for this summary.
Our sample from a descendant of William (m. Dorinda Conklin) of Lake County, Indiana (U13), is a case in point. His early residences suggest a John-line descent, but lines descending from Jonathan don't indicate a good candidate for the birth family of this man. My hypothesis that he represent a descent from Simeon (R11), son of Joseph, may still hold, with the mutation having occurred in a recent generation, as with the first Thomas sample. Another sample from a different descendant of this William would be interesting, as it may lack the mutation.