Seven small matches, BanyanDNA and Patrick’s sister Catherine

There is a particular kind of stuck that Irish research seems good at reproducing in vast numbers – and I have been sitting in it for a while.

My second great-grandfather Patrick Caughlin was born around 1820 in the Clonfinlough area of Clonmacnoise, King’s County, and emigrated to Wisconsin, where he died in 1913. My tree says his parents were James Coughlan and Ann Devery. Not proven but somewhere someone had these names in a tree so I had them as a hypothesis to work on.

How Catherine turned up

What eventually moved this wasn’t a document. It was an afternoon spent going through somebody else’s DNA match list.

The tester was an older member of the family, a generation closer to Patrick than I am. Every generation you can move up dramatically increases the DNA shared with everybody above, which drags distant relatives up over Ancestry’s reporting threshold and into view at all. If there is an older tester in your family, their match list is literal gold for research. And if you have enough matches over on Gedmatch or My Heritage that you can ‘triangulate’ matches, that is like winning the biggest lottery in the universe. (You can only share a triangulation with someone who inherited it from the same ancestor).

I went through those match lists, searching on the surname — and searching every spelling of it. Coughlan. Coughlin. Coghlan. Caughlin.

And the same woman kept coming back. Different testers, unconnected to one another (or so it seemed). They were all in Australia – in tree after tree was Catherine Coughlan — a woman who married Thomas Brennan in Clonmacnoise Parish and sailed for New South Wales in 1841. That recurrence is a strong hint. One tree is somebody’s claim; a name that keeps surfacing across matches is worth attention.

Trees are just a lead. Many people (including myself) have been guilty of copying trees. But what all these trees did was tell me to be curious about this one woman named Catherine.

Catherine Coughlan Brennan

Catherine married Thomas Brannin — that is the register’s spelling — at Clonmacnois on 1 May 1841, with Bernard and Maria Hart standing as witnesses. Four months later she was on the other side of the world.

She arrived in New South Wales on the John Renwick on 31 August 1841, and died at Taroom, Queensland, in 1897, aged 78, having raised seven children.

Two things about that marriage record:

The spelling caught me out a second time. When I went to RootsIreland for the full citation I searched Brennan — which is how the family is spelled in every Australian record they left behind — and got nothing. The entry is indexed under BRANNIN, the register’s own spelling, and RootsIreland’s surname-variant matching (which is usually excellent) did not pick it up. I only kept going because I already knew the marriage existed. I then checked a different Caughlan marriage from the same register page to confirm the parish really was indexed, and it was.

She married just in time to be found. The Clonmacnois marriage register begins on 2 February 1841. Catherine married on 1 May 1841. Had she married much earlier there would be no record of it whatsoever — which is precisely why there is no baptism for her, none for Patrick, and none for their brother James. The register does not go back that far. She got in under the wire.

And Catherine, unlike Patrick, left records that name her parents. THE WALL.

The first is her assisted immigrant record. Every assisted emigrant to New South Wales was interviewed, and a clerk wrote down where they came from and who their parents were — at the moment of departure, from the emigrant’s own mouth. That is about as close to the source as Irish research usually gets.

Read what it gives me:

NATIVE PLACE — Clunemackny (Clonmacnois) which is not far from Athlone
FATHER — COGHLAN James (farmer)
MOTHER — Ann (dead)

Her father was James Coghlan, a farmer, of the Clonmacnoise district. Her mother was an Ann who had already died by August 1841. Catherine was twenty, Roman Catholic, a house servant, and she could read.

It is a wonderful record, but it stops one word short. Ann who? The clerk wrote the given name and moved on. No maiden name — which, for tracing a woman in Ireland, is the word that does all the work.

Two things we learned here though: The clerk marked the mother as dead and did not mark the father, which means James was most probably still alive when Catherine left home in 1841.

The missing word we needed – Ann’s surname – arrived from the other end of Catherine’s life. Queensland death registrations ask for the parents of the deceased, including the mother’s maiden surname. In November 1897 her son-in-law Patrick Meagher stood in front of the registrar at Taroom and supplied them.

Father: James Coughlan, Farmer. Mother: Ann Devery.

Between the two records I now had Catherine’s parents in full — father named and occupied at emigration, mother’s maiden name at death, the two agreeing on everything they overlap on.

And sitting there with both in front of me, I could see the route through the wall. If Catherine was Patrick’s sister, then her parents are his parents.

But what if I couldn’t find paper, yet, to connect Patrick (or James) to Catherine? What else could I do?

Why “if” needed more than a shared cM chart

Seven of the matches in that older tester’s match list descend from Catherine, through four of Catherine’s children. I did not match them, so using my own test they would have remained out of reach.

They share between 12 and 49 cM with our tester.

The three at the top of the chart below are the ones that made Catherine visible in the first place, and 49 cM is a solid match. But look at the bottom of that list. Twelve centimorgans is barely a match at all, and at that size a single value tells you almost nothing on its own. The potential relationships for 49 cM overlap so heavily that no shared cM chart on The Shared Centimorgan Project is going to separate them.

That is the whole difficulty. Small matches are tricky. Taken one at a time they are not often very helpful except as leads. Taken together — seven of them, descending through four different children, tested against one specific family structure — they might be decisive.

Which is how I ended up with an account at BanyanDNA. I am new to it. This was my first real piece of work in the tool, and some of what follows is me finding the light switches and figuring it out. I still have a lot to learn so if you know more than me, please email me and send me your tips!

What BanyanDNA actually does

Most DNA tools work one match at a time: here is a number, here are the relationships that produce that number. Banyan works the other way round. You give it your tree and the shared cM for every tested person in it (anonymously), and it runs inheritance simulations — thousands of simulated times — across the structure you have drawn. Recombination is random, so every trial comes out differently, and after enough trials you have a distribution: what each pair in your tree should share, if the tree is right.

Then it lays your real numbers over the simulated ones and asks whether they fit. (If you like statistics : Each match gets a standard-deviation score, and the tree as a whole gets a chi-square, where closer to 1 means the structure accounts for the data well.)

It works on the shape of the tree alone. So it cannot tell me who Catherine’s parents were — the records do that. What it can tell me is how far apart Catherine and Patrick sit, which was the one piece neither of Catherine’s records could tell me.

Three trees, one difference

So I built three trees. Same tree, same seven matches, same everything — with Catherine put in a different ‘position’ (relationship) within the tree each time.

1. Full sister — both Catherine and Patrick are children of James Coughlan and Ann Devery.
2. Half-sister — daughter of James Coughlan by a different wife than Patrick’s mother
3. First cousin — daughter of a hypothetical brother of James Coughlan.

Then I ran each one and compared the results.

Structure tested	Chi-square	Matches outside expected range
Full sister	0.99	0 of 7
Half-sister	0.91	4 of 7
First cousin	0.47	7 of 7

First cousin fits worst. All seven matches sit above what Banyan expects for a cousin’s descendants — B1 shares 49 cM against an expected range of 0–19. But there’s a catch I didn’t appreciate at first: I can only see matches that are big enough for Ancestry to show me, so every model with distant relationships is going to look a bit “too low” against real matches. That pushes towards the closer answer. So cousin is the least likely of the three, but I can’t call it impossible. I plan to add more matches as I find them.

Half-sister fits, but less well. Halving the expected sharing pushes the four largest matches out of range. New matches may add more detail to this differentiation from sibling.

Full sister fits everything. All seven inside range, none further than 0.8 standard deviations from the simulated mean.

One check that reassured me: the matches on my own side, whose descent is documented, all landed where Banyan expected them in every run. When the known part of the tree fits, I trust what it says about the unknown part more.

Notice what did the work there. It wasn’t the 49 cM match — on its own that could sit comfortably in more than one of these models. It was the small ones, all seven of them, all pointing the same way at once.

Because I was suspicious of a margin as thin as 0.99 against 0.91, I re-ran both models at ten times the trials. Both came back identical to two decimal places. That tells me the numbers are stable. It doesn’t tell me the gap between them is big — 0.99 and 0.91 are both good fits.

What this does not prove

Banyan’s own documentation says a good score “does not mean that the relationships are correct”. Other genealogies can fit the same numbers. Consistency is not proof, and I would rather say so myself than have somebody say it to me.

Three specific holes, in order of how much they bother me.

Half-sibling is not excluded — and the record makes it concrete. Catherine’s mother was already dead in 1841. A widowed farmer with children remarrying is not an exotic scenario; it is the ordinary one. If James remarried and Patrick came from the second marriage, then Catherine’s parents still are not Patrick’s parents. The DNA leans away from this. It does not shut the door. Reassuring : Catherine and Patrick’s estimated birth dates are very close together, making it more likely that they come from the same parents. (Often a widower will remarry immediately so this is no guarantee).

There is a date issue. Catherine was twenty in August 1841, which puts her birth between September 1820 and August 1821. Patrick was supposedly born in 1820. The likeliest answer is that her age was rounded — and/or his date of birth estimated – as ages on these records routinely are. Until pensions came into being, nobody needed to know their age. Irish research throws this up all the time so it is less a worry than the first concern.

Where it leaves me

The DNA fits Catherine being Patrick’s full sister best, and her parents’ names are the names in my tree for his. I think she’s his sister. But seven small matches can’t separate full sister from half-sister on their own, or completely rule out a cousin.

Patrick’s parentage is a reasonable hypothesis — and I want to be precise about why, because it is the interesting part. It still does not rest on a record. It rests on Catherine’s two records, plus a measurement of the distance between her and Patrick that came out of seven matches – no one match of which could have carried it alone.

What it needs now is one document naming Patrick’s parents. That usually comes off a register page – which may be in some basement somewhere, locked in a vault growing mould.

---

DNA matches are anonymised throughout. If you descend from Catherine Coughlan and Thomas Brennan of Taroom, or from any Coughlan family of Clonfinlough and Clonmacnoise, I would very much like to hear from you.