Michael Cooley's Genetic Genealogy Blog GEN • GEN
11 March 2018

Three Y-DNA Duncan Haplotypes

I'll begin with a caveat. While I've been off working on the Cooley DNA Project, getting my BA in history and an MA in English (creative nonfiction writing — I graduated December 2017 at the age of 67), I've largely ignored my Duncan research, have had little correspondence with Duncans in recent years, and have become largely unfamiliar with the work they're doing. I need also point out that I'm not directly connected to the Clan Donnachaidh DNA Project. I do not represent it nor am I familiar with its ongoing research. But as a genealogist of forty-plus years, the Duncans represent one of my most enduring mysteries, and as an admin for nine DNA projects, I've learned quite a lot about genetic genealogy. So here goes.

Everyone who spends any time researching the Duncans of Fauquier County, Virginia, and Rowan/Surry/Stokes County, North Carolina comes away confused. Part of the problem is simply that there are too many genealogical clues that don't quite add up. But although we still a long way to go, genetics has sorted out several Duncan conundrums.

My primary interest is Charles Duncan, born in Fauquier County on 19 Jun 1761, and his wife Margaret Kirk. Charles's family is well-documented in their family bible as recorded in Charles's Revolutionary War pension record. His birth is mentioned but his parents are not.

Here are three Y-DNA Duncan haplotypes, to the first 12 DNA markers:

These are so different that the families could not have possibly been related, patrilineally anyway — that is, through the Y chromosome. In fact, 12-marker test results are so low-resolution that people of dozens, even hundreds of surnames could match one another. We'd have to go back thousands of years to find a common ancestor in the male line — and he was certainly not named Duncan.

Charles's haplotype matches that of #3 — except for the first allele, which has a value of 12 instead of 13 (but more on that in a bit). The first thing we discover is that these markers disproves the speculation that Charles was of family #2, which allows me to definitively state that Charles Duncan (1761-1838) was not the son of John Duncan and Elizabeth Holtzclaw, as is often stated. The DNA renders any dispute to that claim meaningless. In fact, Charles was of haplotype #3 and was related, in some way or another, to John and Dinah Duncan. For this study's purpose, haplotype #2 is entirely out of the picture.

Nor do I believe that Charles was a son of John and Dinah (Bradford) Duncan. Although their Charles might have been of about the right age, consider the following:

Family tradition states (and this has never been disputed) that Rev Landon and and his brother Rev Charles Duncan were sons of Chloe (aka Clorinda) Duncan, a daughter of John and Dinah, and an unknown man. This explains two points: why Landon's Y-DNA does not match any other Duncan family, and why Alamander was Landon's uncle.

So, if all of this is true and we dot all the I's and cross all the T's accordingly, the inevitable conclusion is this:

However, because Charles's relationship is described only to Landon, and because Charles (the younger) was about the same age as Alamander and Chloe (and it's pretty clear from the described relationships that he was not their brother), it occurs to me the following could also be true:

Charles would be Landon's first cousin once removed, which matches nearly anyone's definition of cousin. It also solves some problems with supposed birth years. This is now, then, my favored theory. But either solution aligns well with the principle points of my thesis: that Charles was a son of Charles, and that the pair belongs to the family of haplotype #3. Let's move on.

Duncans and Kirks

Before delving into the Duncan DNA evidence I'd like to provide a sense about this complicated Duncan/Kirk genealogy, further complicated by the fact that the immigrants Kirks — as far as I know — have no better proofs than the Duncans. Consider the following marriages:

Untangle that spaghetti! If anyone can be provide evidence, including DNA evidence, that confirms or contradicts any of this, please feel free to contact me.

Group B of the Clan Donnachaidh DNA Project, to which Charles belongs, is quite large. The project's admin, Tim Duncan, has determined through DNA testing that these Duncans likely came from the Isle of Bute, part of a group of islands in the Firth of Clyde, Scotland. But in scanning over the data, a small anomaly immediately presents itself. The two descendants of Charles, listed at the very top of group B, have a value of 12 at the first marker, DYS393. This is a deletion from the obvious ancestral value of 13. That's no problem. In fact, it might help some future researchers.

It's not often that a family can be parsed down using Short Tandem Repeats (STRs) — with any dependability, anyway — in this case, the 12 repeats of AGAT at DYS393. The reason is that STRs are somewhat flaky. For example, it's possible (if not likely) that Charles was born with 12 repeats, his son Braxton with 13, and his son Thomas again with 12. On the other had, the deletion of a repeat could have happened independently among the descendants of brothers Dalton and William. In the previous article, "The Ashenhurst DNA Project," I cite a perfect example of such collateral mutations at DYS391. This is the reason we can't take the unusual 12 repeats in our results awfully seriously. What matters with STRs are the trends. Any one repeat value doesn't really mean much (unless it can be proved otherwise). So it remains to be seen whether this marker represents a trend among Charles's descendants. Test results for a descendant of one of Braxton's brothers would be useful to that end.

Isle of Bute
STR Trends

But let's pull back and look at other possible trends among Group B. I've copied all the results to a page simply titled Duncan/GroupB and rearranged the entries as best as I could regarding name, date, place and Y-STR marker (the mismatches highlighted in yellow). Conventionally speaking, names, dates, and places are immutable, but mutations, by definition are not. For that very reason, it's impossible to perfectly line up everything, but we can observe possible STR trends.

I've put the at the very top the five testers who claim their Earliest Known Ancestor (EKA) to be Marshall Duncan the Elder and have selected the fourth entry with which to compare. Now look down the first column to find our Charles testers' 12 repeats. That's the context by which I'll study their results.

The first thing to notice is that the upper portion, from Charles and up, is largely free of mutations. Does that mean that the topmost testers are descended are all from Marshall? Perhaps, but look at the entry for John Duncan Sr just below the Marshalls. He was born 15 years earlier. Is this the John Sr mentioned above, the father of John Jr? Well, as I said, my Duncan genealogy is rusty, so we'll just look at the results.

Next look at position #59 (the loci positions are noted in the column at the bottom row of the table you've just brought up). The three Marshall descendants who have tested that far out have a value of 19 whereas all others have 20. We have enough samples here to consider this as a likely trend. Outliers will certainly show up in future tests, but we can say for now, considering the available data, that Marshall Duncans descendants will probably have a value of 19 while the others will have 20. (Remember, we're talking STRs so there will likely be exceptions.) Given that, one of our Charles testers has a value of 20 repeats in that position.

Also note that at locus 34 and 35, the vast majority of Group B testers have 38-40. Whoever the Most Recent Common Ancestor (MRCA) for Group B Duncans was, he almost certainly had those values, known as ancestral values (as opposed to derived values). The rest of it can be chalked up to those flaky STRs.

There are other such trends that seem to pop up. For example, we can be nearly certain that Benjamin Duncan born c1764 had 15 repeats at loci 3 and 18 repeats at loci 32. Looking at these results, though, I wouldn't hazard a guess as to which line he belonged.

So here we have perfect proof that STRs can be very helpful, but can absolutely not be counted on.


Y-STRs are wanderers; they come and go, sometimes in disguises. But there's a kind of mutation that is a definite homebody. Once it finds its place, it wants to stay put. These are called Single Nucleotide Polymorphisms (SNPs). Y-STRs may generally point toward a haplogroup, a designation for a population, but Y-SNPs literally define haplogroups.

SNPs occur when one of the four genetic letters that make up our genomes is altered from one value to another. For example, the SNP known as FGC33214 changed from its ancestral value of C to a T at position 8222189 of the Y chromosome. Such changes in and off themselves mean little when they happen outside the coding regions of our DNA. They're not genes and play no part in our health and are not influential in natural selection nor, therefore, in the evolution of our species. Y-SNPs are silent witnesses as they pass from father to son for generations. I'm unaware of an age estimate for FGC33214 but judging from FTDNA's SNP tree it's of relatively recent origins, to the point, anyway, that it defines at least some — if not all — of the Group B Duncans.

The following is an alphabetical arrangement of the Duncans who have had advanced Y-SNP testing. Next to it is the SNP descent for the "Bute Duncans" over a very rough timeline of more than 10,000 years. The highlighted SNP, DF13, is about where the Group F Duncans, the second haplotype listed above and the group that John L Duncan belongs to, separated from the tree. And it appears that Alex Williamson at The Big Tree has found a new terminal SNP for kit #167920 (John Duncan of Rothesay, Bute, Scotland). In other words, for now, SNPs FGC19842 and A1147 are "brothers" and constitute subclades R-FGC19842 and R-A1147, with FGC33214 being downstream of A1147.

Drum Roll, Please

Considering all we've looked at (and there's much I've not described), I'm going out on a limb to suggest this partial tree for the Group B Duncans.

I know of no genealogical record that connects John Sr, Marshall, and Joseph. But the DNA suggests they were reasonably closely related, very possibly brothers. And it's certainly possible that Charles Duncan (1761-1838) is yet another degree of cousinship from Landon, but he doesn't seem to fit elsewhere.

Other early Duncans shows up in the Group B results — William Duncan born 1690 (kit #66454) and William Duncan born 1692 (kit #22694). This/they could be another "brother" but I know nothing of that lineage. A tester for Thomas Duncan, born in 1700 and married to Elizabeth Outten, is positive for A1147. And there's quite a bit online about a Robert Duncan, born 1692, who is said to have married Ann Gallop, but I don't recognize him among the Duncan results. Might we have this family of Duncans?

In summary, my analysis is based on the following points:

  1. Charles Duncan belongs to Group B (per DNA)
  2. He had a brother William (clear citation is needed), who was not John's son, William
  3. He was a son of another Charles Duncan (1787 tax list)
  4. He was a cousin, not an uncle, of Landon Duncan (Landon's letters)

The latter point convinces me that Charles couldn't have been the son of John and Dinah, as is commonly noted on the web. Besides that, family tradition tells us — to the extent that we can trust it — that John's son Charles died unmarried. Of course, that doesn't mean he didn't have children — and twins Charles and William could have been illegitimate one-offs — but if the DAR is correct and John was born about 1726, it seems unlikely he would have been a grandfather at 35!

Caution, however! I've seen no genealogical evidence that John Duncan Sr had a son named Charles, nor that John Jr had a brother by that name. And it's possible John Sr had another "brother" who produced another slew of Charleses. In that case, though, the degree of cousinship between Charles and Landon would begin to get rather remote. There might yet be another creation story out there, but I'm having trouble coming up with one.

I'm not privy to the Big Y results themselves and I know the identity of only two Duncan testers at FTDNA. It's my suggestion, however, that at least one person who might be descended from John Sr. commit to the test. There's certainly no guarantee, but it's possible a distinguishing SNP has emerged somewhere in that line. If such a person has Big Y-tested, I'd love to take a gander. Contact me below.

1 I don't know whether the record is extant. We're relying on a report that states, "what Charley Duncan copied from family Bibles when in Indiana in 1910." Nevertheless, there's no reason to doubt it.

2 Described as son of John Sr's in the 1766 deed, John Darnell and wife Elizabeth to John Duncan Jr, son of John, and the same John Jr is clearly defined in the 1768 deed, John Duncan Jr and wife Dinah to Thomas Pope same land purchased from John and Jane Darnell. (Hmm. Who was John Darnell's wife?

3 "SNPs are people too."

4 Dates for the sons are uncertain. I'd love to see and would be happy to record the proofs.

5 I've not yet seen it, but a document apparently exists that suggests Charles and William were twin brothers. Certainly, there's nothing contradictory in the record.

6 In a letter to Hiram from his cousin Jane H Shreves, another Duncan/Bradford grandchild, she acknowledges Hiram's interest in the Bradfords and asks about "Aunt Cloah" and her (Jane's) cousins "Landon & Charles." She also states that her mother, Leanna Duncan White, "wishes to be remberd [sic] to her brother [Alamander, Hiram's father]."