Researching distant family relationships like my mother-in-law's possible third great-grandfather Matthias Ambrose may present multiple hazards upon the paper trail. There is a gap between the earliest point at which we found Elizabeth Flowers, Matthias' likely daughter, reportedly in Ohio by 1814 and the date of his will, back in Bedford County, Pennsylvania, in 1804.
Granted, we already knew that Elizabeth and her sister Susannah married two Flowers brothers and that both families had migrated to the same Ohio location, but finding the surname "Flower" mentioned for both his daughters in Matthias' will still wasn't enough to persuade me that the search for Elizabeth's father was over. I've seen stranger coincidences.
What if we can use another way to confirm what we've found on the paper trail? That was my question today, when I delved into results from my husband's DNA test at Ancestry.com.
Usually I examine DNA results using a different approach—searching all matches by specific surname—but for today, let's play with the tool at Ancestry DNA called ThruLines™.
As it turns out, there were multiple DNA matches under the tile for my husband's suggested fourth great-grandfather, Matthias Ambrose. Normally, when I use this tool (which isn't very often), I use the "relationship view" rather than the "List" option.
Because of the number of matches for all Ambrose lines—and also due to some complicating factors which we'll discuss below—this time, I chose to examine the match results by list form. This provided a more manageable readout, as there were several descendants of Matthias Ambrose, for each of which the DNA matches were sorted.
Six children of Matthias Ambrose were represented in that readout, causing me to revisit Matthias' will and the listing of all his children. Keep in mind that his will named three sons, possibly six married daughters, plus two yet unmarried as of 1804, so the six DNA groupings at Ancestry lack any connections to the other lines, at least up to this point. (Because the ThruLines™ process relies so heavily on subscriber-posted trees, there are any number of reasons why the other lines aren't represented in this tool's list of Ambrose matches.)
All told, as of this date, there are sixty four people who have taken a DNA test at Ancestry who match my husband's results, and who also have Ambrose in their posted tree there. Not surprisingly, the largest number of Ambrose descendants were clustered into the line descending from Elizabeth (thirty two), followed by Susannah (fifteen).
Considering that Matthias Ambrose represents a fourth great-grandfather to my husband, the likelihood that any fifth cousins (if extending both lines of descent equally through the generations) would share his genetic readout is fairly slim. Just taking a quick look at the interactive Shared centiMorgan Project at DNA Painter reveals a likely range of shared cMs from zero to 117, with an average hovering around twenty five. The key point here is, those who "share" zero will obviously not show up as matches at all. It may not be possible, at this genetic distance, to confirm relationship via DNA.
Then, too, the smaller numbers may border on the unreliable. Various DNA experts warn against reading anything genealogical into numbers below ten centiMorgans. Or twenty. Or fifty. In the earlier days of utilizing genetic genealogy, I've even heard some instructors assert it is not worth a researcher's time to consider any matches under one hundred cMs.
Just taking two possible Ambrose matches as examples, we can see the wide variance at this distance of relationship. Using one of the other likely Ambrose lines—I chose Susannah, the second largest group, as Elizabeth was my husband's direct ancestor—I sampled two matches to my husband who were each estimated by Ancestry to be his fourth cousin, once removed. One match had eight centiMorgans in common with my husband, contained in only one segment—a quite predicable scenario. On paper, that match turned out to be a fifth cousin, once removed.
A second example, also estimated to be a fourth cousin, once removed, shared 148 centiMorgans with my husband, a considerably larger count. That number was spread out over eleven discrete chromosomal segments. Yet again, on paper, that match turned out to be a fifth cousin, once removed.
And here's that other issue I mentioned earlier. The difficulty in relying on the two Flowers wives may be obvious: not only were their descendants all related through their Ambrose ancestor, but also through their Flowers ancestor, since two brothers married two sisters.
Furthermore, because I know the rest of this family's story—they all settled in Perry County, Ohio, and remained there for well over one hundred years in a small, tightly-knit community—the number of other intermarriages between the community's families cause estimating relationship levels merely on the basis of shared centiMorgans difficult. While a study of Perry County's descendants is not an example of endogamy, it certainly approaches something a bit beyond mere pedigree collapse. I like to call it "endogamy lite."
However, excluding the two sisters' lines, the centiMorgan range for the non-Flowers matches was between six (the bottom number Ancestry will show) and twenty one. That lower end barely slithers beyond unreliable, although the higher side of the range might be somewhat helpful, but only in combination with some carefully-executed traditional genealogical research methods.
The fact that the Flowers-related matches also begin at that same doubtful six centiMorgan level may be telling. The fact that the upper end of the Flowers match range reaches to 388 centiMorgans for what relationship, by necessity of genealogical distance, should be beyond fourth cousin, tells us something else is going on, as well. There are likely so many other genetic connections represented in these matches that it tells us one thing above all: when using DNA testing at a relationship distance like this, proceed with caution.