The Gilliland Mystery

Family legend holds that young John was hidden in a well when Indians attacked.  His parents died in the skirmish, and John was raised by an adoptive family.  Whether the Hamilton surname was that of his biological parents or his adoptive ones is unknown.  We don’t even know if the story is true.

John Hamilton was born sometime between 1860 and 1865, according to various sources.  He reported that he was born in Texas, but there are no contemporary documents to corroborate that.  The first confirmed record of John Hamilton is that of his marriage to Mary Rowan Smith in 1888 in Conway County, Arkansas.  He and Mary farmed and raised their family in nearby Yell County, Arkansas.  He died there in 1937.

John’s great grandson Gary has been trying to identify John’s biological parents for several years.


DNA Testing Strategy

Autosomal DNA (atDNA) represents all branches of your tree, but it gets “diluted” with each generation.  As a great grandchild, Gary inherited only about 12.5% of John Hamilton’s autosomal DNA.  His best strategy to identify John’s parents, therefore, is to test multiple descendants of John.  There are two reasons to do this.  First, known cousins can be used to eliminate shared matches on Gary’s maternal and grandmaternal lines so that we can focus on those who are related through John.

Second, those other descendants might represent portions of John’s genome that Gary didn’t inherit.  The best way to visualize this is using the Coverage Estimator at DNA Painter.

The nine DNA testers shown in black in the diagram represent roughly 52% of John’s DNA.  That is, these nine kits could conceivably identify about half of the people John would have matched had he tested himself.  In practice, we only have direct access to five of these kits.  All nine can be used to sort matches, but our access only represents about 41% of John’s genome.

Unlike autosomal DNA, the Y chromosome is passed undiluted from father to son.  Occasional changes (mutations) can be used to gauge how many generations separate men whose Y-DNA matches.  Gary and two of his first cousins, SLH and JLH, have done the Y-111 test at FamilyTreeDNA, and two of the three have done the Big Y-700 test.  They are all direct male-line descendants of John Hamilton and match one another as expected with both atDNA and Y-DNA.  However, they don’t match any other Hamiltons; the majority of their top Y-DNA matches have some variant of the surname Gilliland (e.g., Gililland, Gilleland, Gillan, and Gilland).

Other than one another, their top Gilliland matches at the 111-marker level are PG, MG (both exact matches), DG (3 differences), LG (4 differences) and CG (6 differences).  Only PG shares atDNA with any of the Hamilton descendants.



DNA and/or census evidence suggested six hypotheses.  Not all are mutually exclusive.

  • Hypothesis 1:  Our John Hamilton is the 17-year-old John Hamilton censused in Cooke County, Texas, in 1880.  He was single, a lodger, and working as a porter in a saloon.  This is the only John Hamilton in that census who was born in Texas, of the right age to be our John Hamilton, and not living with family.  No DNA evidence has been found to either support or refute this hypothesis.
  • Hypothesis 2:  Our John Hamilton was the same person as John Gilliland (born circa 1861), son of John R. Gilliland (c.1831–c.1865) and his second wife Charity Wilke, from Van Zandt County, Texas.  The elder John Gilliland died in 1864 or 1865, which partially aligns with the orphan story.  The son was censused with his mother in 1870 but not in 1880.  Could he be the John Hamilton of Hypothesis 1?  If this hypothesis were true, Gary would be a junior third cousin once removed to DG, who matches on 108 of 111 Y-DNA markers.  FamilyTreeDNA gives this match only a 44–60% chance of being within the last 4–5 generations, and DG does not share atDNA with the known Hamilton descendants, so this hypothesis seems unlikely a priori.

  • Hypothesis 3:  Our John Hamilton was a son of Joseph Gilliland (1816–1873) of Tennessee.  This hypothesis is supported by an exact Y-DNA match at 111 markers to MG.  If this hypothesis is true, MG would be a junior third cousin once removed to Gary.  FamilyTreeDNA estimates a 98% chance that the connection is within five generations.  However, MG does not share atDNA with the Hamilton descendants, and all of Joseph Gilliland’s known sons can be traced.  None was old enough to have been our John Hamilton’s father.
  • Hypothesis 4:  Our John Hamilton was the grandson of Thomas Gilliland (≈1782–1852) and his wife Mary Ann Galliher (b ≈1790) from South Carolina.  This hypothesis is supported by an exact Y-DNA match at 111 markers to PG.  If this hypothesis is true, Gary and PG are 4th cousins.  PG shares atDNA with Gary’s sister (36 cM in one segment) and his two first cousins (23 and 37 cM, respectively; one segment each) but not with Gary or his Hamilton second cousins. FamilyTreeDNA estimates a 98% chance that the connection is within five generations, and we expect about half of 4th cousins to not share atDNA, so the DNA evidence supports this hypothesis. However, none of Thomas Gilliland’s sons are known to have lived in Texas, where John said he was born, or Arkansas, where John lived later in life.
  • Hypothesis 5:  Our John Hamilton was the 14-year-old John Hamilton censused in the township of Texas in Craighead County, Arkansas, in 1880.  He is listed as the son of Charles Hamilton and his wife Jane.  They would have been only 17 and 16, respectively, when John was born, and their next oldest child was 8 years younger than him.  The same couple were censused as presumed newlyweds in 1870 with no children, although John would have been at least 5 years old at the time.  Is this our John with his adoptive parents in 1880?  When he said he was born in Texas, was he referring to the township and not the state?
  • Hypothesis 6:  Our John Hamilton was someone else entirely.

Big-Y Time Tree

FamilyTreeDNA recently introduced their new Discover™ tools for the Big Y test.  My favorite is the Time Tree, a graphical representation of the time to most recent common ancestor (TMRCA) between the men in the tree.  And it’s really, really cool!

The Time Tree is based on what’s called a “molecular clock.”  In essence, it estimates how frequently SNP and STR changes have occurred per generation throughout history, then calculates how long it’s been since pairs of men in the tree shared an ancestor.  Because DNA mutation is random, there’s necessarily a lot of uncertainty in these estimates.  Fortunately, we can quantify that uncertainty and work with it.  (As an aside, I coauthored a book chapter using molecular clock methods for bees, ants, and wasps.)

This is part of Gary’s Time Tree:

The first branch point, at the upper left, is labeled with the SNP mutation (R-FGC64966) that all seven of the descendants share.  That node occurred sometime between about 600 and 1300 CE (the dashed horizontal line) with an average around the year 1000 CE.  Obviously, that’s too far back to be relevant to Gary’s project, so we can ignore it.

The other nodes are far more interesting.  The one in red (R-BY94024) represents Gary and his cousin SLH’s shared grandfather, who was born in 1897.  The Time Tree average for their grandfather’s birth is 1876, not far off from the truth.

If you go to the “scientific details,” you’ll see a probability curve for this estimate.  Time is on the X-axis and likelihood on the Y-axis.  The higher the curve, the more likely that date is for Grandpa Hamilton’s birth.  The black vertical bar marks the average at 1876.  What’s even more impressive is that the highest point of the curve, called the mode and marked with the red arrow, is around 1895.  This is remarkably accurate!

To address Gary’s question, we’re most interested in the node labeled R-FGC64969.  That’s when Gary shared an ancestor with PG and MG.  Time Tree gives an average birth year of 1643 for that man, with a 95% chance that he was born between 1462 and 1775.  The mode is right around 1650 CE.

That estimate is both disappointing and useful.  It’s disappointing because 1650 is far too long ago to help us identify John Hamilton’s father.  It’s useful because it rules out Hypotheses 3 and 4 (which were based on Y-DNA matches to MG and PG).

By logical extension, we can also rule out Hypothesis 2, which is based on the match to DG, even though DG hasn’t done the Big Y test and is not in the Time Tree.  I say that because MG and PG are exact STR matches to Gary at 111 markers and DG is not (they match on 108 of 111 markers).  If Gary’s shared ancestor with MG and PG was born in the 1600s, his shared ancestor with DG is almost certainly further back.

If the Time Tree is accurate (and I suspect it’s not far off), we’ve been chasing a red herring with those atDNA segments shared by PG and the Hamiltons.  Those segments may have been inherited through the Hamilton’s grandmother or great grandmother; they’re not likely through the Gilliland patriline.

Finally, we have to consider the possibility that John’s father’s surname was neither Hamilton nor Gilliland.  Assuming an average generation time of 35 years, there were six generations between John’s birth and that of the 1650 ancestor.  The surname could have disassociated with the Y chromosome in any one of them.  That is not a happy thought.


A Reevaluation

While this result is not what we’d hoped for, it helps us redirect our efforts.  For now, I’m looking closely at Hypothesis 5, that our John Hamilton was the boy censused in Texas Township, Arkansas.  This makes a lot more sense from a geographic perspective than him being the porter in Cooke County, Texas.

Consider this map.  The red points represent the two John Hamilton’s in the 1880 census.  The blue points are the known locations of events in our John Hamilton’s life.  The 1800s in the US were characterized by a steady migration westward.  If the young John in Cooke County, Texas, wanted to farm, he would gone west for cheaper land, not east.  The John in Texas Township, Arkansas, on the other hand, is likely to have moved westward in search of his own property.


We still don’t know who John Hamilton’s parents were.  One complication is that we’ve yet to find a cluster of atDNA matches that could represent them.  All of the atDNA matches examined so far group with known lineages of the Hamilton tree.  That in itself may be informative; it may be telling us that John Hamilton was related to his wife’s family somehow.  That would not be unusual in farming communities of that time.

Gary is excited to see if any new clues about John’s ancestry show up in the comments about this story.  Do you have Gilliland ancestors?  Please chime in!

26 thoughts on “The Gilliland Mystery”

    1. This new Time Tree makes it so much easier! And the best part is that the estimates will get better as more people test and as more link their cousins to their trees to help calibrate the molecular clock.

  1. I read this article with great interest! I just noticed the surname Gilliland showing up in my MyHeritage matches’ trees yesterday. I did a search just now on the name Gilliland, and I have a match whose kit is managed by a Gary Hamilton, with a YDNA Gilliland 1845 listed as the tester’s 2nd great-grandfather. Now, I am wondering if that match is the same family as you’ve written about in this article.

    I do not know what assistance I can offer other than that this match is a maternal match for me. But, I will be happy to try to help however I can. Who knows, maybe this will somehow help me to breakthrough my own stubborn brick wall. 🙂

  2. I am the sister of MG that is the exact Y-DNA match to Gary. Let me know if I can provide any information. We are fairly certain we are descended from the Gilliland’s from Chatham County NC. My sister is a fairly close atDNA match to one of William Gilliland’s descendants.

  3. What do you know about Charles Hamilton’s wife, Jane? Is it possible that she was a Gilliland and John was her nephew? There was a lot of back-and-forth movement between East Texas, Western Arkansas, and Southern Missouri from the 1800s into the 1930s so assuming he would only have moved west may be limiting your research. Lastly, I have 2nd cousins who are descended from Edward Franklin Gilliland (1845-1933) of Alabama. Their branch ended up in Southeast Missouri; I haven’t researched any of the other sons. My cousins haven’t done a DNA test. Let me know if you need further info.

    1. I’ve seen two possible surnames for her, and neither is Gilliland. There are (at least) two unrelated Gilliland lineages in the US.

      1. Also, the Gilliland branch I referred to above has an inherited neurological disorder (similar to ALS); may be something to look for in death records.

  4. If John Hamilton was related to his wife’s family somehow, you should be able to quantify it to some extent, such as restricting how close they are.

    Also, NPEs?

  5. Wow, intriguing case! Just wanted to say that I’ve lived in Craighead County for 25 years and had never heard of Texas Township (or the apparent Little Texas Township!). Also, if you need anything local, let me know and I’ll try to help!

    1. Thank you for the offer! We’d love any insight into the area in the 1860s. Were any Indian attacks reported nearby? Civil War skirmishes that might have been misinterpreted by a young child as an Indian attack? Of course, we don’t even know if the Texas Township hypothesis is true, but it’s our best lead at the moment.

  6. Sorry for the long dissertation… I’m not as adept at the DNA research as Gary. I tend to have some luck with paper records so I’m thankful for this discussion thread that discusses the topic.

    My sister SG has the most autosomal matches to Josiah Gilliland’s descendants. You will find his oldest son William (b. 1813) in the 1840 Chatham Co NC census. Joseph, our line (1816) and younger brother John (1818) are found in a separate household with other siblings and both parents, Josiah and Sarah “Bray”. Both Joseph and John migrated to west Tennessee after that. They can be found in separate, but nearby households in the 1850 census of Madison Co TN. There are several Gilliland’s in Gibson Co TN that adjoins with Madison Co.

    SG has matches to older brother William’s line that remained in NC. This was found in Ancestry’s ThruLines…

    GE (4th cousin 1X removed) 67cm / 2 seg.
    DO (5th cousin) 51 cm / 4 seg.
    MI (5th cousin) 48cm / 2 seg.
    AO (5th cousin) 40cm / 4 seg.
    EP (5th cousin) 20cm / 2 seg.

    Also one with younger brother John who migrated with Joseph…

    AF (5th cousin 1X removed) 8cm / 1 seg.

    1. Those top matches to your sister are certainly promising. When we get into more distant relationships (4C and further), I like to see at least one segment decent sized segment, say at least 15 cM and preferably at least 20.

  7. Great post.
    Leah, can you please clarify something for me?
    The Time Tree has what look like error bars either side of each SNP mutation.
    So I would expect those to correspond to the range found in the two SNP mutation probability curves. But they don’t. The “error bars” show considerable overlap between the possible ages of mutations R-FGC64966 and R-BY94024. But the probability curves don’t. So what are those “error bars” ?

    1. Oh wow, that’s a really perceptive observation. The error bars are similar to the 95% confidence intervals, but not exactly the same. I have no idea why! Perhaps FTDNA updated the analysis at some point but didn’t update all of the associated web pages?

  8. I’m thinking you should be able to determine at least the minimum distance, such as not a sibling or maybe not first cousin. Or, eliminate his descent from certain of her family members.

    1. Yes! This has already been helpful in suggesting a testing strategy for getting closer to an answer. There are two prongs to it. First, MG is the only one in the group to have tested at Big Y-500. I suspect we’ll get better resolution of that long branch to him if he’s upgraded. Second, there’s a Y-37 tester on the same branch as PG. I’m hoping that upgrading him to Y-700 will break up that long branch to PG and tell us whether Gary and his cousin are closer to PG or to MG and perhaps refine the date estimates.

  9. Don’t know if I’m related but it was a interesting read nonetheless! Wish I did a DAN test to find out more about my family tree.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.