How I solved my great-grandmother's NPE using DNA
My great-grandmother, Catherine Maud Isabelle (“Belle”) Reeves, was born 21 Jan 1888 in Corsicana, Navarro, Texas. Her father, Tillman H Reeves, had died the preceding November - so we were told - in Comanche County, Texas, where the family lived. Shortly afterwards, her mother, Martha Caroline, packed up her five kids and moved to Corsicana, some 200 miles to the east, to have her baby. My aunt, the senior genealogist in our family, always thought that was strange: why would a woman in the advanced stages of pregnancy leave family, friends and community at such a time? We didn’t know, but figured, she must have had her reasons.
NPE discovered after 125 years
After my aunt's unexpected death in March 2013, I contacted our 2nd cousin, Tony, who gave me startling news: years earlier, another cousin found Tillman Reeves’s probate file at the Comanche County courthouse, and it proved Tillman died in October 1884, three years before Belle was born. Martha Caroline, presumably, left her community to avoid the stigma of being an unmarried mother.
Now I had a mystery on my hands, but I was optimistic I could solve it. For starters, I had learned how to identify my kits' crossover points. These are the points where the DNA you get from both parents breaks down into alternating segments from your grandparents. You need a minimum of three siblings to do the mapping, and I had tested my mom, her brother and sister (hereinafter referred to as the "kits"). Mapping the crossover points doesn’t tell you which segments came from which grandparent, however – only where there was a crossover. To do that, you need 2nd or greater cousin matches on some of the segments. I was also fortunate to have four of my kits’ half second cousins (Alva, Gary, James and Sharon) – each a grandchild of a different half sibling of Belle’s – tested at Family Finder. As a result, I knew which parts of the Belle-contributed segments came from Belle’s mother. And I could infer that for at least some of those segments that the kits got through Belle where they did not match Alva, Gary, James or Sharon, they were matching through Belle’s unknown father.
A possible candidate
One match in particular intrigued me: Gwen. She shared a total of 90 cM each with my aunt and uncle, including a single segment of 69 cM with my uncle, and 44 cM with my mother with a longest segment of 18 cM. She was matching only on segments the kits got from Belle, but she did not share DNA with Alva, Gary, James or Sharon. In a phone call with Gwen, she read her ancestors' names from charts prepared by her mother years earlier. Gwen's great-grandmother was named Mary Lane, and Mary's great grandfather's name, Simon D F Lane, was the name of the man in whose household Martha Caroline and her brother, James, were living in the 1860 census. Martha Caroline’s mother had died in childbirth with her in 1852, and her father, remarried with children from both wives in his home, had sent Martha Caroline and James to live with his friend and neighbor, Simon D F Lane.
I quickly confirmed that Simon D F Lane, Gwen’s ancestor, was the father of Simon D F Lane in the census. The younger Simon Lane was the uncle of George D Lane, Mary Lane’s father. The Montgomerys and Lanes had long been neighbors; George and Martha Caroline’s fathers, Alfred G Lane and Nathan Montgomery, were listed in the same 1840 Walker County, Alabama tax district. Since Gwen was a predicted 2nd-4th cousin with my kits, I homed in on George D Lane as Belle’s father.
George D Lane was living in Lampasas County, Texas in the 1880 census, two counties south of Comanche. His wife and several of their children died in 1884-1886. George remarried in 1888 and had two more children. But from about 1885-1888, he was single. The conditions seemed ideal for a union, however brief, between George and Martha Caroline, doubtless old acquaintances, each recently widowed.
More Lane connections
But I needed more than the sole match with Gwen and circumstantial evidence to prove that George D Lane was my ancestor. Unfortunately, on the segments where my kits matched Gwen, there were no other comparably sized segments or known Lane descendants. However, I had been in contact with other matches with Lane ancestry, including Robert, a descendant of Alfred Van Buren Lane, George D Lane’s brother. Robert shared 85 cM with my uncle – including a longest segment of 31 cM, and two additional segments of 9 and 16 cM; 42 cM with my mother; and 68 cM with my aunt (the two longest segments 27 cM and 8.4 cM). Like Gwen, Robert matched the kits only on segments they inherited from Belle, but never on segments with Alva, Gary, James and Sharon.
The same was true of Eldon and Harold. Eldon is the great-grandson of Sarah Lane, Alfred and George Lane’s sister; and Harold is the great-great-grandson of Joseph Lane, these siblings’ uncle. But despite the strong matches with Eldon, Gwen, Harold and Robert, none of them were matching in the same places. This meant I could not triangulate the match (triangulation is the process of identifying three or more matches on the same segment and attributing their shared DNA to an identified, common ancestor). But this was not surprising. The odds are against two or more 3rd cousins matching on the same segment of DNA. Third cousins share the same 2nd great-grandparents. We each get approximately 6.25% of our DNA from our 2nd great-grandparents. The chances of three 3rd cousins getting the same exact, or overlapping 6% of their randomly passed DNA from one of their common ancestors is quite small.
Despite the odds, I was still hoping for a triangulated segment. Gwen, the closest match, was a half 2nd cousin once removed, so triangulation seemed to be at least a possibility. I knew that George had another child by his first wife whose descendant had tested, at Ancestry – Linda; and I had contacted Homer, a great-grandson of Hattie Lane, George D Lane’s half-sister. Both Homer and Linda were also the kits’ half 2nd cousins once removed. Happily, Linda agreed to transfer her kit to FTDNA, and Homer agreed to test. When their results came back, I got what I was looking for: triangulation.
The screenshot below is of my uncle's chromosome browser and shows a long overlapping segment on chromosome 10 with Homer and Linda. The segment spans the centromere, but still is a high match - 86 cM with Homer and 30 cM with Linda.
In addition, the small orange and blue overlap is one shared by all the kits with Homer and Robert on chromosome 9.
Although Gwen, Homer and Linda all had the same relationship to the kits, their common ancestors were different: in Homer's case, it was Alfred G Lane; in Gwen's and Linda's, it was George D Lane. The kits' and their matches' relationships are shown below.
The kits all shared the most DNA with their closest matches – Gwen, Homer and Linda. (Actually, Dewey shared a substantially larger amount with Homer and comparable amounts with Robert, Gwen and Linda.) Sherry's cM amounts were lower but she had more segments with Gwen (two) and Linda (four) than with her other matches. The number of segments – not just size – is also a factor in assessing relationships. It is more common than not that 3rd cousins will only share one segment of DNA IBD (although Robert, a 3rd cousin, shared 2 and 3 segments with Joan and Dewey, respectively).
When I compared the kits' matches through Belle's father with those through her mother, I could see that the kits shared slightly more DNA with the maternal cousins – which was not surprising since the matches through Martha Caroline were the kits' 1/2 second cousins – not (as in the case of Gwen, Homer and Linda) half 2nd cousins once removed.
I was fairly confident that I had identified the correct candidate as Belle's father. Gwen, Homer and Linda shared comparable DNA with the kits, but with the common ancestor in successive generations. An alternative scenario, in which Belle's father was a different son of Alfred G Lane's, would have meant that Gwen and Linda were the kits' half 3rd cousins once removed or, at best, half 3rd cousins – relationships which did not seem to be supported by the size or number of segments they shared. I based this on 17 3rd cousin once removed segments I recorded in my database of known cousin matches, with an average total cM of 77.08 cM and longest segment of 29.68 cM. I had no known 1/2 third cousins once removed recorded but figured the DNA would be appreciably below these amounts.
I also had to look at the circumstantial evidence. Alfred G Lane's sons, Alfred Van Buren Lane, Simon B Lane, Albert Galiton Lane, Jefferson Davis Lane, and Robert Lee Lane, were all living in either San Saba or Lampasas Counties, Texas in the 1880 and 1900 censuses. Robert Lee Lane was born in 1867 and so was 17 years Martha Caroline's junior, and, I thought, probably not likely Belle's father. But any of the others were possible - though each was married at the time of Belle's birth and raising a family. This did not rule them out, but it somewhat favored George D Lane, who was in similar circumstances as Martha Caroline, as each was recently widowed.
Because the Kits had the same relationship to Gwen and Linda as they did to Homer, but at a generational remove, and because they shared the same approximate amount of DNA, and amounts consistent with their relationships to their other Lane cousins based on the presumed connection, I believe the matching scenarios and circumstantial evidence prove, if not beyond all doubt, then with near certainty that George D Lane was Belle's father. Thanks to DNA, a 128 year old NPE was solved almost as soon as it was discovered. What were the odds?