MyHeritage v2.5 · read 10 September 2026

What the DNA says

A living descendant tested. It is the first check this archive has ever had from outside the paper record — and the first thing it does is destroy the most attractive claim in the tree.

A living descendant of Kenneth Lindsay D'Arcy (1927–2010) and Audrey "Dell" Murdoch (1929–2022) has taken an autosomal DNA test. She is living, so she is not named here and her kit is not identified — but her result can be published without her, and it is worth publishing, because every other page on this site rests on paper. The rule →

What an ethnicity estimate is, in one sentence. It compares your genome against reference panels of present-day people from various regions and reports which panels it most resembles. It does not find your ancestors. It cannot name one, date one, or confirm a single link in a pedigree — and the further back a claim sits, the less it can say, until at about eight generations it can say nothing at all.

The comparison

What the tree predicts, against what the swab says

The tree's prediction is computed by walking her pedigree and attributing each line to the country of its deepest ancestor whose birthplace is recorded — with both rejected grafts cut, because this archive does not accept them. Both grafts →

Region as the estimate reports it	The estimate	The tree predicts	Note
Scottish and Welsh	41.9%	29.7%	The estimate reports Scotland and Wales as one region. The tree has 26.6% Scottish and 3.1% Welsh.
English	19.1%	29.3%	The tree predicts more English than the estimate reports, and more Irish still.
Irish	19.1%	34.4%	Cork, Limerick and one Matson “b. Ireland” on the mother's side.
Dutch, Danish and Germanic	14.1%	6.3%	One man: John Blum, born in Germany, the tester's great-great-grandfather.
Breton and French	5.8%	0.4%	The largest gap on this table, and the one least likely to mean anything — see below.
British Isles, all four together	80.1%	93.4%	This is the row that matters, and the two agree.

The headline is agreement, not contradiction. The tree says this family is overwhelmingly British and Irish with one German great-great-grandfather. The DNA says the same. The internal split between English, Irish and Scottish differs — the tree predicts more Irish and less Scottish than the estimate reports — but those three populations are the hardest in Europe to tell apart, they have been exchanging people for two thousand years, and no company's panel separates them cleanly. A disagreement inside the British Isles is close to no information. A disagreement about whether the family is British at all would have been a finding.

The test worth running

Cut both grafts and the DNA barely notices

This archive rejects two joins the family tree carries: the D'Arcys of Hornby Castle, unevidenced, and the Sneyds of Keele Hall, disproved. Between them they carry 98 people and about thirteen extra generations, including the Earls of Holderness, the house of Oldenburg and the kings of Denmark.

So the prediction was computed twice — once with the tree as the family holds it, once with both grafts cut. The difference is 0.39 of a percentage point.

Predicted origin Tree as held Grafts cut

Ireland 34.38% 34.38%
England 28.91% 29.30%
Scotland 26.59% 26.56%
Germany 6.59% 6.25%
Wales 3.12% 3.12%
France 0.39% 0.39%
Denmark 0.02% 0.00%

A hundred people, thirteen generations, four royal houses — and less than half of one per cent. That is not a quirk of the arithmetic; it is what deep pedigrees are. Halve a contribution every generation and by the twentieth an ancestor is worth one part in a million. The glamour at the top of a family tree is, genetically, almost exactly nothing. It is worth knowing whether it is true. It is not worth knowing because it would show up in a swab.

The trap

4.4% Danish is not the kings of Denmark

Somebody in this family will eventually put "Danish 4.4%" next to "descended from Frederick I of Denmark" and join them. This section exists so that when they do, the answer is already written down.

The tree's own descent predicts 0.02%. Frederick I sits around generation twenty. Twenty halvings is one part in a million. The Oldenburg descent, taken entirely at face value, cannot produce 4.4% of anything.

You do not inherit from most of your ancestors. DNA is passed in chunks, and beyond about eight generations the majority of your ancestors on paper contributed no measurable DNA to you at all. A twentieth-generation ancestor is, for almost every descendant, genetically absent.

Danish, Dutch and Germanic together come to 14.1% — and that North Sea signal is exactly what ordinary English ancestry produces in these panels. England spent centuries absorbing Angles, Saxons, Jutes and Danes. An English-descended person routinely scores several per cent "Danish" without a Dane in sight.

The tree already explains the continental share better. One man — John Blum, born in Germany, the tester's great-great-grandfather — accounts for a predicted 6.25% on his own. No royalty required.

The same applies, in reverse, to Breton 4.2%. The D'Arcy name is Norman — it is a place in France. The name → But a surname arriving with one man in the eleventh century leaves no autosomal trace whatever, and 4.2% Breton is not evidence of it. A name is not a chromosome.

What the numbers cannot do

Four limits, stated before anyone quotes the percentages

Version This is v2.5. The provider's own FAQ opens by explaining how it differs from v0.95. The same saliva has produced different numbers before and will again — the genome did not change, the reference panel did.

Confidence The slider sits middle of the slider — showing 2 of 3 genetic groups. The list you are shown is a setting, not a result. Move it and regions appear and vanish.

Small numbers Everything under about 5% is at or near the noise floor. French 1.6% is not a French ancestor; it is a number that could be zero.

Individuals No percentage on that page names, dates or confirms a single person. Not one join in this archive is settled by it, and none is overturned by it either.

Where the trail actually stops

One side of this pedigree is half the size of the other

Her ancestry divides evenly between two parents. This archive's coverage of them does not.

Side Ancestor slots filled Reaches

Kenneth Lindsay D'Arcy — D'Arcy and Sneyd 157 of 304 · 52% generation 21, though only through the grafts

Audrey "Dell" Murdoch — Murdoch and Atwell 25 of 52 · 48% generation 5, and then nothing

The names where her mother's side runs out are worth reading aloud: Claud Murdoch. Janet Walker. William Watt. Marion Fleming. Agnes Isabel Scott. Five Scottish surnames, four to five generations up, and not one recorded birthplace between them.

This is the research directive, and it did not come from the percentages. It came from counting. The Ayrshire Murdochs are a quarter of this person and the archive has twenty-five ancestor slots filled out of fifty-two — against three hundred and four on the other side. The Murdochs → · The queue →

What would actually be evidence

Two things on the same results page are worth more than every percentage on it, and one of them is the strongest test available to this archive by any means.

Y-DNA, from a male D'Arcy. The tester has no Y chromosome. A male-line D'Arcy descendant does, and it descends father to son essentially unchanged. It would place the D'Arcy paternal line against every other tested D'Arcy in the world — which is the most direct attack anyone can make on Robert D'Arcy's parentage, and it beats every archive visit on this site's list. not yet done

The match list. Shared matches carrying their own trees can confirm or break a specific join — the Saniger line and the Hornby graft above all. Unlike an ethnicity estimate, a match is evidence about individuals. not yet examined

The genetic group, "England #13". Groups are far more geographically specific than the headline regions and often name a county. Worth reading before anything above it. not yet read

And one correction, recorded where it happened. The first version of this analysis put the tree's predicted Scottish share at 0.02% and announced that the DNA was contradicting the tree. It was not. The fault was in the method: it attributed each line to the deepest ancestor on record rather than the deepest ancestor with a known birthplace, which threw away St Quivox in Ayrshire because the men above it are names without places. The corrected figure is 26.6%. With the rest of them →

Predicted origin	Tree as held	Grafts cut
Ireland	34.38%	34.38%
England	28.91%	29.30%
Scotland	26.59%	26.56%
Germany	6.59%	6.25%
Wales	3.12%	3.12%
France	0.39%	0.39%
Denmark	0.02%	0.00%

Version	This is v2.5. The provider's own FAQ opens by explaining how it differs from v0.95. The same saliva has produced different numbers before and will again — the genome did not change, the reference panel did.
Confidence	The slider sits middle of the slider — showing 2 of 3 genetic groups. The list you are shown is a setting, not a result. Move it and regions appear and vanish.
Small numbers	Everything under about 5% is at or near the noise floor. French 1.6% is not a French ancestor; it is a number that could be zero.
Individuals	No percentage on that page names, dates or confirms a single person. Not one join in this archive is settled by it, and none is overturned by it either.

Side	Ancestor slots filled	Reaches
Kenneth Lindsay D'Arcy — D'Arcy and Sneyd	157 of 304 · 52%	generation 21, though only through the grafts
Audrey "Dell" Murdoch — Murdoch and Atwell	25 of 52 · 48%	generation 5, and then nothing