What the DNA says

The finding Inferred, strongly

Kurt Israel Luwinski is very probably not Lynn Derrick Luwinski's biological father. Two independent lines of evidence say so, and the paper record was already hinting at it before any test was taken.

Not documented, and not closed. What would close it is set out at the foot of this page.

Nor does de Witt obviously fit

The temptation is to move the fatherhood four months earlier and be done. But an Afrikaner Oelofse–van Straaten line should read Dutch, German and French Huguenot. The observed Dutch is 1.5% and Germanic 1.4%, against English 18.5% and a named genetic group of Southeastern and Eastern England. That is an English family, not a Cape one. The question is open, and this archive does not intend to close it by guessing.

And one match is a generation closer The lead

The 297.5 cM match is not part of that knot — to her, the Harrison line is only a parent's third cousin. She sits nearer to Derrick than any of them:

1st cousin's daughter — 39.7% — most recent common ancestors: Derrick's grandparents.

In plain terms: her grandparent was a sibling of Derrick's biological father. And the detail that turns arithmetic into a place — her father was born in Johannesburg. Her family surnames are Halsey, Parr and Fenton.

Chromosome 2, 181,681,048 – 216,191,586 34.5 Mb

All three matches share that exact stretch with Derrick.

And a recombination you can see

On chromosome 18 one match's segment ends at 55,491,965 and the other's begins at 55,493,868. They abut, 1,903 base pairs apart. Two relatives holding adjacent, non-overlapping stretches of the same ancestral chromosome is the signature of a break in a shared ancestor — the point where one grandparent's chromosome was cut and handed down two different branches. Between them, the two matches reconstruct nearly the whole of Derrick's paternal chromosome 18.

One line, read at two depths Inferred

Four of those people are first cousins and their children to the holder of the big tree — 480 to 801 cM with her. They are one tight family, and Derrick stands at second-cousin distance from all of them.

The closest match inverts the ratio. She is 297.5 cM to Derrick and only 60–82 cM to every member of that core. She is not part of it — she is nearer to Derrick than they are. That is not two families. It is one line, read at two depths:

• The closest match's common ancestors with Derrick are his own grandparents. She was born in 1995, two generations below him, and MyHeritage reads her as a first cousin's granddaughter.
• The big-tree holder's are one generation further back — his great-grandparents. Second cousin, 68%.
• Which predicts that those two should be third cousins to each other. They are, at 82.1 cM. The structure is consistent from three directions at once, and that is the strongest thing about it.

Two surnames, at their two depths The short list

At grandparent depth: Halsey. The closest match's own pedigree is Halsey · Halsey · Parr · Parr · Fenton — every one of them living, and named nowhere here. Her match card carries the line that matters: “Ancestral places common to derrick luwinski and [her] include South Africa.”

At great-grandparent depth: Brown, Castle and Webb — documented four generations up in a tree of 10,715 people, and all of these people are dead, so the archive can name them:

Second cousins have parents who were first cousins. So if that reading holds, Derrick's biological father was a first cousin of Geoffrey David Brown or of Hazel Webb — a man of that generation, carrying on the other evidence the surname Halsey, and in Johannesburg in 1948. And Derrick's own great-grandparents are one of four couples: Brown–Brister, Castle–Wallis, Webb–Hayden, or the unrecorded couple behind that Unknown grandfather.

That last one is worth staring at. In a documented tree of 10,715 people, one grandfather is recorded as “Unknown” — in exactly the generation, and exactly the position, where an undocumented connection would sit. Set aside the following day — see below.

This is a structure, not a name, and it is labelled Inferred. What it does is turn an open question into a short list: four couples, one surname, one city, one decade.

Halved again, and the “Unknown” grandfather drops out 15 September 2026

The short list ran to four great-grandparent couples, and the archive made a point of the fourth — a grandfather recorded only as Unknown in a tree of ten thousand people. That was the wrong one to stare at, and one more look at the match pages says so.

The two people who are first cousins to the big-tree holder — 801 cM and 714 cM with her, and Derrick's first and third closest British matches at 333.9 and 165.9 cM — both list Brown among their ancestral surnames. One has Brown, Harrison; the other has Arrow, Brown, Noye.

First cousins share one grandparent couple. If both of hers carry Brown, they are on her father's side — descended from Walter George Brown and Agnes Kate Castle. And if Derrick is a second cousin to all three of them, then his own great-grandparents are the parents of one of that couple:

So Derrick's biological grandparent was a sibling of Walter George Brown or of Agnes Kate Castle, and his biological father was that sibling's son — a man born around 1910 to 1928 who was in Johannesburg in 1948, carrying on the other evidence the surname Halsey.

Held as Inferred, and here are its weaknesses. Brown is among the commonest surnames in England, and both trees are tiny — three people and five. A relationship estimate of “second cousin” carries real width; the same 333.9 cM also reads as half first cousin at 21%. What makes it worth acting on is that two independent first cousins point the same way, and that it is falsifiable: one reply naming a Webb connection would overturn it in a sentence.

First, a confirmation nobody had to argue for

Among the children of Walter George Brown (b. Great Yarmouth 1889, d. Hampshire 1971) is a daughter, Decima Audrey Harrison, 1928–1997. A Brown daughter who married a Harrison. That is where the Harrison match comes from, and it settles by documentation what was inferred yesterday from centimorgans alone: he is on the Brown side.

And then the short list itself Inferred

If Derrick is a second cousin to the tree's keeper, his biological grandparent was a sibling of Walter George Brown or of Agnes Kate Castle. Both sets are now named.

Eight people. Derrick's biological father was a child of one of them, born roughly 1910 to 1928, and in Johannesburg in 1948. The family is Norfolk — Norwich, Great Yarmouth, Caister-on-Sea, and Alfred John Brown was born at a hamlet called California on the Norfolk coast.

Superseded on 15 September. Two things above are wrong. The Castles are Hertfordshire, not Norfolk — only the Browns are Norfolk — and the census puts Derrick's father a generation higher than “a child of one of them”. The four Brown siblings are no longer part of his descent at all. What replaced this →

Which makes one thing suddenly obvious The hypothesis to test

Three of those eight are daughters — Alice Brown, Ada Castle, Martha M. Castle. A daughter's sons do not carry her maiden name. They carry her husband's.

And the surname sitting on Derrick's single closest match, whose shared ancestral place with him is South Africa, is Halsey.

If Alice Brown, Ada Castle or Martha Castle married a Halsey, then her sons were Halseys, born in exactly the right years — and one of them was Derrick's biological father.

It also puts Mervin Halsey, the grandfather the closest match named, exactly where he needs to be: a brother of Derrick's father, another son of that same marriage.

This is checkable without anybody's permission — so it was checked, the same afternoon, in the England and Wales marriage index. The entry is not there. See below.

Tested the same afternoon, and it fails Negative

The Halsey hypothesis was published and then put to the England and Wales marriage index within the hour. It does not survive.

No daughter of these two families married a Halsey in England or Wales. The neat join between the tree on one side and the surname on the other is not in the index, and the archive says so rather than leaving a tidy guess standing.

What survives, and what it means. The eight names came from the tree, not from the hypothesis, and they stand. So does something the tree shows plainly: Alice Brown, Ada Castle and Martha M. Castle have no spouse and no children recorded at all — only parents and siblings. They are genuinely untraced, which is exactly where an unrecorded branch would sit. That is not evidence for the hypothesis; it is the absence of evidence against it.

Three ways it could still be true, and one way it could be wrong from the start. The marriage may have happened in South Africa, outside the English index entirely. The Halsey name may enter a generation lower. Or — and this is the assumption worth naming — Halsey is simply the closest match's father's surname. Her link to Derrick runs through one of her four grandparents, and it could as easily be the Parr side, or her grandmother Cynthia Weir. This archive picked Halsey because it was the name on the card. Open

The grandparents she named a year ago, found

In June 2025 the closest match wrote that her paternal grandparents were “Mervin and Cynthia Halsey — her maiden name was Weir … Cynthia's middle name was Yvonne.” Every particular of that checks out, down to the middle name; the register spells the maiden name Wear rather than Weir.

And the Johannesburg cemetery records carry them both to the end: Mervin Richard Halsey, born 15 June 1930, died 30 May 2005, buried in Johannesburg — indexed there under the given name Kevin, which is the same man and the same middle name and year, mis-read. Cynthia Yvonne Elizabeth Halsey, born 1932, died 4 April 2000, buried at Brixton, Johannesburg.

A supporting negative: there is no Mervin Halsey in the England and Wales birth index for 1900–1940. He was not an Englishman who emigrated. He was South African.

Why his parents are now the whole question Inferred

Mervin was born in 1930, a generation ahead of Derrick. MyHeritage reads the closest match as Derrick's first cousin's granddaughter, and she is Mervin's granddaughter — which puts Mervin himself as Derrick's first cousin.

If that holds, Mervin Richard Halsey's parent was a brother or sister of Derrick's biological father.

One generation, one document. Three ways to get it, in order of likelihood:

• The 1952 register itself, image 2149. The 1956 form has columns for the parents of both parties; the 1948 form has none. 1952 falls between them, and the frame is already located.
• His estate. He died in May 2005, so there should be a J294 naming his father and mother. It is not in the gazette corpus under that spelling — but the Master of the High Court can be asked.
• His birth, South Africa, about 1930.

And the question already put to her this morning — who were Mervin Halsey's parents? — turns out to have been exactly the right one, asked before any of this was known.

Two words in a column nobody was looking at

Derrick's cluster is English. Mervin was born in the Cape, and there is no Hertfordshire, Norfolk or anywhere-else-English in him. His wife was the one born in England — and the match descends from both of them equally. The archive had simply picked the man's surname because it was the one printed on the card.

Three things that were wrong on this page until today Corrected

• The Castles were never in Norfolk. This page said “Norwich and Caister-on-Sea”. That is the Brown side. The Castles are Baldock and Norton, Hertfordshire, and Caroline came from Weston, four miles away.
• Caroline Wallis is Caroline Walls — her widowed father Joseph Walls, born at Weston in 1827, is living in the house in 1911.
• The list of eight was half wrong. Brown enters this family only as the man Agnes Kate Castle married. The four Brown siblings drop out of Derrick's descent entirely.

And the arithmetic changes which generation he sits in Inferred

Two of Derrick's closest matches now descend from one couple, through two different daughters:

Five centimorgans apart, down two separate branches. That symmetry is the whole argument: Derrick is not hanging off one twig of this family, he is equidistant from two — which puts him inside the Baldock household rather than beside it.

And it moves him up a generation. If Derrick's biological grandparent were a Castle sibling, the closest match would be a first cousin three times removed — about 117 cM expected. The reading is 297.5. If his biological father was himself a son of that house, the match is a first cousin twice removed, expected about 212 cM — and that is also, word for word, the relationship MyHeritage printed on the card: first cousin's granddaughter.

Derrick's biological grandparents are, on this reading, named for the first time: George Castle, born Baldock 1863, and Caroline Walls, born Weston 1863, married 1883.

Walked back within the hour. The paragraph above rests on the 292.7 cM match being a grandchild of Agnes Kate Castle. Her tree was then opened and she is a great-grandchild — her father Geoffrey David Brown, born 1926, is Agnes Kate's son. That is one generation further out than the arithmetic above assumed, and it changes the answer from settled to open. The corrected version is immediately below.

Two readings, and the numbers do not cleanly separate them

A fits the centimorgans better. B fits the ages better. MyHeritage's own algorithm returns one of each. And something in the readings is odd on either account: the two great-grandchildren average 313 cM and the great-great-grandchild reads 297.5, when a whole generation should roughly halve it. Two matches a generation apart at almost the same number is not what a simple line of descent produces.

So the honest statement is narrower than the one published an hour earlier, and it is this:

Derrick descends from George Castle and Caroline Walls of Baldock. Whether they are his grandparents or his great-grandparents is not settled, and centimorgans alone will not settle it.

Under A his father is George W. Castle b.1885 or Frederick Castle b.1890 — 63 and 58 in 1948. Ernest b. 28 November 1904 is out: the 1939 Register has him at 29 Norton Crescent, Baldock, a butcher (incapacitated), married, with children, which matches the “invalid” of 1911 and does not match a man in Johannesburg nine years later. Under B his father is one of the grandchildren of that house, born roughly 1910 to 1928 — which is the shape the archive had a week ago, before the surname changed.

What became of the Halseys, for the record NAAIRS, 54 documents

They are no longer Derrick's family, but they were chased for a week and the findings should not simply vanish. The South African national archives index returns, under Halsey:

• WLD 12617/1978 — divorce: Mervin Richard Halsey versus Cynthia Yvonne Elizabeth. They separated in 1978, twenty-six years after St Martin's-in-the-Veld.
• MHG 6348/64 — Erral David Halsey, the estate of the 25-year-old who died at Walvis Bay in July 1964, born at Spaarwater, Heidelberg, Transvaal.
• MHG 790/55 — Penrose Thomas Halsey, who married Helen Edith Stevens at Lower Umgeni, Durban, in July 1914.
• And a Kimberley family older than any of them: Walter 1899, William 1917, Sarah 1920.

A negative worth keeping: the South African death-registration form of that period names parents only where the deceased was under five. Erral David Halsey's 1964 form is complete, legible, and names neither.

And the match list itself now answers to the name Swept the same evening

The search box on Derrick’s DNA Matches page takes an ancestral surname and searches all 5,464. In August it was given Leibholz and Luwinski and returned zero for both. Tonight:

Stated carefully, because the last time this page got excited about a surname tag it was wrong. The 14 is not proof: the top hit is the Castle descendant already known, and thirteen matches at 11–38 cM is the range where a shared surname is suggestive and nothing more. What it does do is put Castle on the opposite side of the ledger from Leibholz and Luwinski, which have never returned a single person.

A trap walked up to and stepped around Checked, not published

Searching the same list for Wear returned eight, and the 292.7 cM match was top of that list too — which looked, for about a minute, like her tree independently containing the Hitchin marriage. It does not. Opening her tree and searching it shows the “Wear” in it is Weare: Henry Weare born 1591, Grace Clarke born 1611.

This is precisely the error this archive published a lesson about in the Figueiredo case — believing a surname tag without opening the tree behind it. The lesson held, one day later, on a finding it would have been very convenient to believe. Open the match before believing the label.

The lesson, which is the same lesson as last time

A week went into Halsey because it was the surname on the card. It was the wrong half of a married couple. The archive had even written down the objection — “Halsey is simply the closest match's father's surname … her link to Derrick could as easily be her grandmother” — and then carried on down the man's line anyway.

Write the objection down, then act on it. Two columns on a register settled in one glance what a week of surname-chasing could not.

Split the closest match's 297.5 cM in two

~44 cM of it lies in regions Derrick also shares with a Castle cousin — a single block on chromosome 2, from 173 to 218 Mb, which all four of them carry. That is the Castle inheritance, and it behaves exactly as it should.

~253 cM of it lies nowhere near any of them. Chromosomes 1, 6, 7, 8, 12, 15, 18, 20 and 22 — nine segments Derrick shares with her and with no descendant of Agnes Kate Castle.

If Derrick and the closest match had only the Castle couple in common, that 253 cM would have nowhere to come from. They share two ancestors, which is to say they share a married couple.

Her great-grandparents on that side are Henry Robert Wear and Martha Maud Castle, married at Hitchin in 1914. Derrick descends from both — so his biological father was their son, and a brother of Cynthia. Which is, word for word, MyHeritage's own most probable relationship for that pair: first cousin's daughter, 39.7%.

And it kills the alternative outright. Under the other reading, Derrick’s relationship to the closest match would be the same relationship the 292.7 cM match has to her. Hers reads 82.1 cM. His reads 297.5.

And a machine that knew none of this put the same people in one box 16 September

Everything above was reasoned out by hand — shared-match lists read one page at a time, segments copied off match pages, a couple inferred from the shape of an overlap. On 15 September MyHeritage produced an AutoClusters report for the same kit: 110 matches sorted into 20 clusters, 5,361 cM in all.

AutoClusters knows nothing about this archive’s reasoning. It groups matches purely by whether they match each other. It produced the same group.

Cluster 3 is the paternal line, and it dwarfs everything else — twelve people, 1,706 cM, nearly a third of all clustered sharing in the kit and almost three times the next cluster. All three anchor matches sit in it together.

But it does not separate Wear from Castle, and that is worth being clear about. The closest match descends from both halves of the couple; the two Castle cousins descend from one half. All three therefore match each other, and this algorithm is bound to put them in a single cluster. A married couple looks exactly like this. Cluster membership proves paternal, not Wear — which is what the twelve-region signature above is for.

The bifurcation is the other finding, and it is cleaner than anything argued here. Derrick’s mother was Portuguese, from Lourenço Marques. So every Portuguese cluster is maternal and every British one is paternal — and there are four separate Portuguese clusters, fifteen people, about 776 cM, against a British side that outweighs them several times over. For a man whose father of record was a German Jew from East Prussia, that is the 0.0% Ashkenazi result restated in a different currency.

The names are not here and will not be. These are living people. This archive publishes DNA match holders as role, country and cM only, and the export itself is held off the repository. Three of the four Portuguese clusters contain someone sharing more with Derrick than anyone this archive has ever written to about his mother — but writing to them is David’s to do, not this archive’s.

The Wear signature, as coordinates 16 September

The argument above has always been made by exclusion — these regions are the ones the Castle cousins do not share. That is hard to test a new match against, because it asks a reader to hold twelve absences in mind.

So it has been computed the other way round and written down. Subtracting every region either documented Castle-side cousin shares from the closest match’s twelve segments leaves 253.2 cM in 12 regions. These are the coordinates. Any further match who shares these is a candidate for the paternal side, and can be tested in seconds rather than argued about.

GRCh37/hg19, as MyHeritage reports positions. The subtraction is done on physical coordinates and the cM of each remainder is apportioned by length, which is an approximation — cM and base pairs do not scale together. It reproduces this archive’s previously published figures of 44.3 cM shared and 253.2 cM not shared to the decimal, which is the reason to trust it and also the reason not to over-read the last digit.

Because he was in Johannesburg Documented

He was thirty-one in June 1948, living in Johannesburg, and married to somebody else. The marriage ended in a divorce his wife brought in 1951.

And he explains a thing this archive could not: why his sister Cynthia was in Johannesburg at all. She married there in 1952, aged twenty-one, having been born at Hitchin. A young woman does not sail from Hertfordshire alone and by chance. Her brother had been there for at least twelve years.

What is proven, and what is a strong guess Read this part

Proven to the standard this archive uses for DNA: Derrick descends from Henry Robert Wear and Martha Maud Castle — from the couple, not from one of them. That is what 253 centimorgans outside every Castle region means, and nothing else explains it.

What was not proven, this morning: that the son in question is Leonard. That rested on circumstance — he was the only son the records happened to name, the right age, in the right city. The 1921 census cannot show children born after it. So the question was put properly.

Fourteen fields, and not one of them is a parent

Six registers, three decades, two form designs, and the answer is the same every time. The state asked more and more about the two people in front of it — their race, their identity numbers, their exact dates and places of birth — and never once asked whose children they were.

One caveat: the 1970 example is a South-West Africa entry from Gobabis. But it is the same B.M.D. 3 form used across the Republic, and the same fourteen fields as the 1956 Johannesburg sheet.

64 Greenfield Road, Greenside, Johannesburg

Cynthia is twenty-one and a spinster, and that is the address she gives. An unmarried woman of twenty-one in 1952 is, on any ordinary reading, living with her parents. This is very probably the household address of Henry Robert Wear and Martha Maud Castle — seventeen years after they left Letchworth, and seventeen before he died in this city.

It is the first address this archive has ever had for them. And unlike a grave it can be worked: the Deeds Office property index, Johannesburg street and telephone directories, the voters rolls. If Leonard is ever found at that address, the Berea coincidence stops being the only geography in this case.

The Wears were Anglican — which corrects a half-thought

Married by banns, at St Martin’s-in-the-Veld, and the denomination line reads Anglican Church.

Leonard’s own wedding in 1940 was performed by W. P. Whelan, who wrote Catholic Priest on that same line. The letter sent to the Catholic parish at Rosebank guessed that “the church followed the bride”. That guess now has evidence. His sister married Anglican; Phyllis Agnes Robertson’s mother was a Momple, a Mauritian name, and Mauritius is Catholic. And the archive should stop half-thinking the Wears were Catholic.

And a Catholic register would not have named the parents anyway. Rosebank replied that the 1940 register survives and contains the entry, and that it does not carry them — “Unfortunately not, as attached”. The letter produced the page; it did not produce the parents.

A fourth ante-nuptial contract New lead

The 1952 contract is Cynthia’s rather than Leonard’s. But an ante-nuptial contract is a notarial deed that recites each party’s particulars, and a deed for a Wear daughter executed in Johannesburg in 1952 may well name her father — which is the one fact this archive cannot otherwise document.

Duplicate Original Marriage Register, no. 13 of 1940 — Rosebank, Johannesburg

7 September 1940, solemnised by W. P. Whelan, who filled in the Denomination line in his own hand: “Catholic Priest”. Witnesses Mary Una Robertson and James Rogers.

What not to read into it Method

Army is what the clerk wrote. Not a force, not a unit, not a number. He could in principle have been a British soldier posted to the Union. This archive’s own finding that the whole Wear household had left England before 29 September 1939 makes the Union Defence Force much the likelier of the two — he was already in the country, and enlisted there. But the register does not say so, and neither will this page.

Two routes — one closed, one open Leads

1. The SANDF Documentation Centre, Pretoria. South African service records of the Second World War survive there, and an attestation form gives full name, date and place of birth, next of kin, and in most cases the names of the parents. That is the exact gap. This archive’s whole chain rests on Leonard being a son of Henry Robert Wear and Martha Maud Castle, and the only source that says so is a seven-person family tree built by his daughter. An attestation form would make it a government record.

2. The Catholic parish register at Rosebank. Closed 22 September 2026 The reasoning was that civil registers have no parents’ columns and Catholic sacramental registers do. The reasoning was sound and the premise was wrong for this parish. The Church of the Immaculate Conception answered in a week: the 1940 marriage register survives, it contains this entry, and — in the parish office’s own words — “Unfortunately not”, it does not name the parents. There is also no second book to ask for: “There are no separate records for the banns of Marriage — the Catholic Church keeps one Marriage Register, the banns are included in the attached.” The page reproduced above IS the parish register. They confirmed Fr Whelan was parish priest at Rosebank from 1935, so it is the right church, and the books are at the parish rather than the Archdiocese. This route is closed by the only person who could close it, and route 1 is now the whole of it.

The two weddings, side by side

Eleven years, and the second wedding is the quiet one. That is a shape, not a fact, and it is set down here as a shape.

Duplicate Original Marriage Register, Johannesburg, no. 1397 of 1951

Solemnised 30 October 1951 at the Magistrate’s office, Johannesburg. Witnesses J. D. Malzer and L. H. Beck. Signed L.A.R. Wear and M.E. Dunlop.

How much weight that carries Say the weakness with the fact

Not much, on its own. Berea was a dense inner suburb of tens of thousands of people. The two addresses are three and a half years apart. Nothing here places the two of them in a room together, and a shared suburb is not a document.

What it is: the first geographic overlap this archive has found between Ivone and the man the DNA points to. Every previous link ran through segments and indexes. This one is two street names. It belongs on the record, and it belongs on the record with this paragraph attached to it.

A third ante-nuptial contract New lead

Column 10 reads with. So a notarial contract was executed for Wear–Dunlop, Johannesburg, October 1951 and registered in the Deeds Office, alongside the two already sought for Kurt (June 1948, February 1956). An ANC recites the parties’ full particulars and is not confined to an eleven-column form. It is now the best available document naming Leonard in his own right, and it goes on the list for the Pretoria researcher.

Which closes it Inferred, but no longer circumstantially

Six children. Two sons. In June 1948 one of them was thirty-one, and the other was eleven.

The DNA says Derrick’s biological father was a son of Henry Robert Wear and Martha Maud Castle. The index says there were exactly two — Leonard, born 1917, and Bruce, born 1936 — and in June 1948 Bruce was eleven. The naming of Leonard Aubrey Robert Wear no longer rests on him being the only son anybody happened to have found.

Two of these six may be living, and are named anyway, at the direction of the family this archive belongs to. They appear here exactly as they appear in the General Register Office index and nowhere else — a name, a quarter, a district — and nothing further about either of them is published. See the method page.

Limits, because they are real. The index covers England and Wales only — a child born abroad would not appear, though the births run continuously at Hitchin from 1917 to 1936, so the family was there throughout. The index does not give sex; if the 1936 child is not a son, the conclusion is only stronger. And it clears up a loose end: Eileen Mary Wear b.1914 and Laurence Anthony Wear b.1915, both registered at Hitchin, are not in this list — their mother was not a Castle. A different Wear family in the same district.

Two documents would still be better than an inference: WLD 2026/1951, the divorce, and WLD 790/1957, an opposed application brought against his first wife six years later — applications of that kind are usually about children. Both are in the National Archives at Pretoria and can be ordered by number.

And the Y cannot help yet Checked, negative, and the negative means little

If this is right, Derrick's Y chromosome is a Wear Y, from Henry Robert Wear of Bermondsey. So the surname projects were checked. Of the 7,748 public members of the R-U106 project, only 34 sit under R-S11477, and the surname projects with members there are Ballew (16), Stafford (8), Blakely (8), Byrd (5) and York (4). No Wear, Weir or Ware.

That is not evidence against anything. The branch is barely tested, and its common ancestor lived around 200 BCE — ninety generations before any surname existed. The test that would actually settle it is a living male-line Wear descendant taking a Y test, whose result should be Derrick's. That is cheaper and sharper than Big Y-700, though Big Y would at least put Derrick on a branch young enough for a surname to mean something.

A correction owed to the person it was sent to Retracted

This morning’s message to the 292.7 cM match told her that her tree was wrong and Caroline was born a Walls, not a Wallis. Her tree is right. It has Joseph Wallis, 1827–1912, which matches the 84-year-old widower in the 1911 house exactly. The spelling came from a FindMyPast transcription of somebody’s handwriting, and a transcription is not a document. What the census actually adds is that Caroline’s father was living with them — not a new surname.

Then the family found this Independent

A small MyHeritage tree — seven people, last touched in May 2018, kept by a South African family who know nothing of Baldock, Letchworth or this research:

“Muad” is Maud. Martha Maud Castle, of Baldock, Hertfordshire — the woman this archive named from a birth index three days ago, sitting in a stranger’s tree as Leonard’s mother, and put there in 2018.

Two lines of work that never touched each other arrive at the same woman. This is the strongest corroboration the Wear reading has had, and it did not come from the DNA at all.

Held at its proper weight. A user-submitted tree is not a document — it can be wrong, and this one disagrees with FamilySearch about when Phyllis died (1973 here, 20 July 2004 at Key West, Florida there). What it cannot plausibly be is a coincidence: there is no reason for an unrelated South African family to give a Leonard Wear born in England in 1917 a mother called Maud Castle unless she was.

And it names the half-sister If the reading holds

Leonard had a daughter, and the tree says who she is without meaning to. Its profile page labels him “Father (of site manager)” — so the <Private> daughter is the person who built the tree, and opening it with her as the main person shows the name unmasked: Gillian Nel. Which is, exactly, Gillian Agnes Louisa Nel, born Wear, 1941–2024, married Willem Adrian Nel in 1969 — born a year after Leonard married Phyllis Agnes Robertson, and carrying her mother’s name.

On the reading this archive is testing, she was Derrick’s half-sister. She died in 2024, so the one test that would have ended this in a single reading — half-siblings share about 1,750 cM and nothing else looks like it — is no longer possible.

The whole tree is seven people: Gillian, her parents, and her four grandparents — and it stops there. It records no children for her, which is not evidence she had none: a tree built to record where you came from does not contain your descendants. Her 2024 estate file names her heirs, and that is the document to get.

Two more things fall out of opening it. Her Robertson grandparents are entered as Edgar and Sabine Robertson, which independently corroborates FamilySearch’s Anatole Edgar Robertson and Sabine Marguerite Momple. And Leonard’s own father is blank in it — she knew her grandmother was a Castle and did not know, or did not write down, that her grandfather was Henry Robert Wear of Bermondsey, who is buried at West Park in the same city. That much this archive can give her family back.

On the line of record: none — with one exception that is not on that line at all

Kurt Israel Luwinski married three times — Frances Hostowsky, Ivone in 1948, and Pauline Rappoport in 1956 — and one child is recorded across all three. On the paternity of record, Derrick has no brother or sister.

The exception comes through his mother, and so it survives whichever way the father question falls: Henrique Figueiredo Pinto da Rocha, 11 March 1941 – 27 October 2022, Ivone’s son by her first marriage. He is Derrick’s half-brother on either reading, because the dispute has only ever been about the father. His estate, 12294/2023, is the file this archive first wrote away for.

On the DNA line: unknown, and one name that is only a candidate Open

Leonard was married from 1940 to 1951 and again from 1951 to 1962. WLD 790/1957 — an opposed application he brought against his first wife six years after she divorced him — is the kind of case that is usually about children. But no child of his has been documented.

There is exactly one name in the Johannesburg records that fits the shape, and it is offered as nothing more than that: Edwina Ann Wear, born 1945, died 30 June 2014, cremated at Braamfontein. The right surname, the right city, born inside his first marriage. Her parents are not recorded in anything found so far, and until they are she is a coincidence with good timing.

Everything findable about her has now been searched, and it runs out. The gazette gives her birth date — 28 May 1945 — her identity number 4505280044086, an address at 17 Pollie Street, Homestead, Germiston, and no surviving spouse, which on that form means there was none. She was cremated, not buried: certificate 65676, ashes released on 9 July 2014 to the undertaker rather than to any relative, so there is no grave and no stone to read. There is no marriage, no married surname, no obituary, no memorial. Google returns nothing for her name at all.

One document names her parents and it is offline: the death notice inside estate 33157/2014 at the Master of the High Court, Johannesburg. Until it is read the honest answer to does Derrick have siblings on this side is: yes — but she is the daughter in the Nel tree, not necessarily this woman.

Ernest is out, on a document rather than an argument

The 1939 Register has him at 29 Norton Crescent, Baldock: Ernest Castle, born 28 November 1904 — “Butcher (Incapacitated)”, married, with Mabel (b. 1908) and children in the house. Incapacitated answers to the invalid of the 1911 census, thirty-eight years apart. A married, settled, disabled butcher in a Hertfordshire market town in September 1939 is not the man in Johannesburg in June 1948.

Frederick, and a caution this archive is publishing rather than papering over Unproven

A Frederick Castle married Ella Newman at Hitchin in the third quarter of 1922, and the 1939 Register puts them at 17 Orchard Road, Baldock — he a vertical driller on aircraft work, she at home, and with them one officially closed record: a child born after 1912. A son there would be exactly what the further reading needs.

And it may well be the wrong Frederick. The 1939 Register gives his birth as 10 April 1887. The 1901 household gives Frederick as eleven — born about 1890 — sitting beside a sister Ada aged fourteen. Two siblings born three months apart is not possible. And the Hitchin birth index contains no Frederick Castle born between 1889 and 1892 at all; what it contains is Frederick James Castle, born 1887.

So either a census age is wrong, or there are two Castle families in Baldock and this is the other one. The archive does not know which, and is not going to decide it by preference. It is written down here so that the next person does not inherit it as a fact.

“Kurt was twelve before his parents married. What if Jacob wasn't his father — could Kurt still be Derrick's?” Asked 15 September 2026

It is the best objection anyone has put to this page, because it attacks the right joint. Almost everything here assumes Kurt was Jacob's son. He was born on 11 May 1912 to an unmarried Auguste Luwinski, carried her surname, and his parents did not marry until 5 January 1924 — twelve years later. Jacob's paternity of Kurt was never documented at the birth. So: if Jacob were not Kurt's biological father, Kurt would carry no Ashkenazi ancestry, his Y-chromosome would be some other man's, and both of this page's main arguments would collapse.

It fails, and it fails on Auguste rather than on Jacob. Whoever fathered Kurt, Auguste was still his mother — that is the one thing the 1912 entry does record, and it is why he bore her name. So if Kurt were Derrick's father, Auguste would be Derrick's grandmother, and roughly a quarter of Derrick would be her: East Prussian, from Sauerbaum and Allenstein.

There is no room for a German grandmother. Every non-Portuguese percentage points at the Atlantic and the North Sea — English, French, Danish, Breton, Dutch — some 38% of him, against 1.4% Germanic. MyHeritage even names his additional genetic groups as “Southeastern and Eastern England” and “England”.

So the objection does not rescue Kurt. It would have had to survive two independent tests and it only addresses one: rewriting Kurt's father leaves Kurt's mother exactly where she was, and Auguste is not in Derrick's DNA either.

But the question underneath it is a real one, and it is already in the post Open

Was Jacob Leibholz actually Kurt's biological father? That is a separate question from Derrick's paternity, and this archive cannot answer it from DNA — there is no sample from Kurt and none from any Leibholz. But the paper is thinner than the family tree suggests, and one detail is genuinely odd.

Under German law a child born to an unmarried mother took her surname, and Legitimation durch nachfolgende Ehe — legitimation by the parents' subsequent marriage — would ordinarily have made him a Leibholz in 1924. He stayed a Luwinski, to his death in 1968, and registered his own son under that name. Meanwhile his brother Siegfried, born in August 1925 after the marriage, was a Leibholz from the start.

That asymmetry has an innocent explanation — a family that simply never bothered to change a grown boy's name — and a less innocent one. The archive does not know which, and the 1912 birth entry is the document that decides it: a marginal note of Vaterschaftsanerkennung or legitimation, or the absence of one.

It was asked for on 15 September, as the third question in the letter to the Landesarchiv Berlin — before this objection was raised, and for the same reason. If the margin is blank, then Kurt's descent from Jacob rests on nothing but family memory, and this archive will say so.

Every one of them had been written to in June 2025, and three answered

Opening MyHeritage's message window to send a carefully drafted first approach revealed that the matches had all been contacted fifteen months earlier, with one template — and that three of them had replied with names. Nothing in this archive had used any of it. It was found by accident, and it is the worst process failure of the project.

Two things follow at once. The non-paternity was already disclosed to these people in 2025, so the delicacy being written into fresh drafts was, to that extent, beside the point. And the original template asked the wrong question — it asked each of them to work out how they were related and to share a tree, which is labour, rather than telling them something they could react to.

What the three said

The second cousin at 333.9 cM, 18 June 2025 — and he offered everything. “You are very welcome to all the information I have — do you have access to Ancestry? Or I could draw it out. It seems likely Derek and I are related through Great Grandparents … My Paternal Great Grandparents are Harrison / Randall and Crisp and Bray.” He was never answered. It also explains why his MyHeritage tree holds three people: his real tree is on Ancestry. And because his paternal great-grandparents are Harrison, Randall, Crisp and Bray, the Brown in his surname list must be maternal — which is exactly what the Brown–Castle reading requires.

And one remark that does not fit. He checked Salkow, de Witt, Luwinski and Rocha against his own matches, and reported that “Rocha was the only name which appeared.” Rocha is Ivone's side — the maternal, Portuguese one. Either he has an unrelated Portuguese connection, or the tidy split of this archive's two clusters is wrong somewhere. It is the one loose thread in the whole DNA picture, and it should be put to him directly. Open

The closest match of all, 17 June 2025 — and she named four grandparents. Her maternal grandparents were John and Elsie Parr; her paternal grandparents Mervin and Cynthia Halsey, Cynthia born Weir, her mother possibly Mord. Since her common ancestors with Derrick are his own grandparents, one of those couples ought to be the join — most plausibly the Halseys. Four new surnames to work: Halsey, Parr, Weir, Mord.

And the sentence worth staring at, from the match at 165.9 cM, also 17 June 2025: his great-grandfather was “Walley Brown, who was married twice (possibly at the same time)” — and this was on his mother's side. A Brown great-grandfather, maternal, married twice and perhaps bigamously, is precisely the shape of thing that leaves a branch nobody has traced. And Walley may well be Walter — the big tree's grandfather is Walter George Brown, 1889–1971. He added that his sister keeps the family tree and he would pass the message on. Nobody followed up with her either.

What it changes Method

• No cold letters. Everyone in the British cluster already has an open thread. The follow-up sent today went into the existing one, because the drafted version opened “I hope you don't mind a message out of the blue” — which would have been false.
• Answer the man who offered his tree, first. Fifteen months late is better than never, and the offer may still stand.
• Ask for the sister who keeps the other family's tree.
• Put the Rocha question directly. It is the only fact on the table that the two-cluster model cannot absorb.
• The drafted letter to the closest match is obsolete — she has already answered the question it asks.

Read the inbox before writing to it. A day went into composing careful first approaches to people who had already been approached, three of whom had already replied with names.

R-S11477 — R1b, and specifically the North Sea branch

MyHeritage's own page puts it plainly: “R-S11477 is a human paternal lineage nested within the broader R-Z381 haplogroup. Every man who carries it descends from a single ancestor who lived around 200 BCE.” The haplogroup-family badge on the same page reads R1b / Z381, and the node above it on the time tree is R-FGC13328, about 600 BCE.

The chain is R1b → M269 → P311 → U106 → Z381 → FGC13328 → S11477, and the branch that matters is U106. U106 is the North Sea, Germanic branch of R1b: it peaks in the Netherlands at about 37%, and runs at roughly 21% in Germany and Austria and 18% in England and Denmark.

Ashkenazi paternal lines do not look like this. Haplogroup J alone accounts for about 38% of the Ashkenazi paternal gene pool; the characteristic lineages are J1, J2, E1b1b, R1a and G. R1b does occur among Ashkenazi men — but the literature describes R1b-M269 there as repeated punctuated introgression of European Y chromosomes, and the two Ashkenazi R1b clusters that are identifiably Near Eastern belong to R1b-Z2103. A different branch of R1b entirely. Not U106.

What it proves — and what it does not Read this before quoting it

This is not a comparison with Jacob Leibholz's Y-chromosome. The archive does not have Jacob's, and it almost certainly never will, because his male line is exhausted:

So the Y-DNA works by character, not by comparison. It says that Derrick's unbroken father-to-father line belongs to the branch of R1b that peaks in the Netherlands and England, and not to any lineage characteristic of Ashkenazi Jewish paternal descent.

On its own that would not be decisive. A Jewish family can carry a European Y-chromosome — that is exactly what “European introgression” means, and it is why nobody should argue from a haplogroup alone.

But it is not on its own. It is the second independent system to return the same answer. The autosomal estimate says 0.0% Ashkenazi where a quarter is required. The Y-chromosome says the paternal line is North Sea R1b. Different molecules, different inheritance, no contradiction between them, and no evidence anywhere pointing the other way.

Leibholz: nothing. Luwinski: nothing. Exhaustive

Not one of the 5,464 people who share measurable DNA with Derrick carries either name or lists it among their ancestors. This page has said from the beginning that there is no German or Jewish cluster near the top of the list. That was a statement about six matches opened in full. It is now a completed search.

The Portuguese half of that sweep was overstated, and is withdrawn Self-corrected

For about an hour this page said the same search had found Ivone's family — 27 matches for Figueiredo, 13 for Mota — and called it the best lead on her parentage the archive had ever had. That was written before a single one of those matches was opened. Opening them undoes it.

“Shared ancestral surname” is a string match. On the top Figueiredo match — 29.0 cM, fifth or fourth cousin — the two people MyHeritage is pairing are Ivone Mota Figueiredo, born 1922 at Lourenço Marques, and Águeda Maria de Figueiredo, the match's direct ancestor six generations back, born before 12 April 1713. Two centuries and an ocean apart, joined on a string, in a country where Figueiredo is one of the commonest names there is. The shared place is no better: it resolves to “Portugal”, and beside the match's own parish the panel reads “No ancestors in derrick's tree.”

And the Mota matches share no surname with him at all. MyHeritage says so on the 50.1 cM match in plain words. The reason is in Derrick's own indexed surname list — Figueiredo, Leibholz, Luwinski, Schmul, Thimm — which does not contain Mota, because the source tree carries it as a middle name. The thirteen hits were other families' Motas.

The archive already held much better Portuguese matches, worked weeks ago and still on Ivone's page: 108.7 cM at second cousin's son, and 133.2 cM. A 29 cM surname coincidence was never going to beat them. The lesson is cheap and worth keeping: open the match before believing the label.

What the sweep did establish Documented

The two zeroes stand, and they are stronger than they looked. MyHeritage indexes Leibholz and Luwinski as two of Derrick's own five ancestral surnames. So the search was not firing blanks at terms the system had never heard of — it had real, indexed names to match, and across 5,464 people it returned nothing for either.

And there is no computed path to anybody. Filtering his matches for Has Theory of Family Relativity returns “There are no DNA Matches that fit your criteria.” MyHeritage has built no relationship path to a single one of the 5,464. Whatever is found here will be found by hand. Negative, and clean

One loose thread, recorded as loose. The two Portuguese matches that were opened both trace to the Viseu district — Treixedo in Santa Comba Dão, and Tondela and Castro Daire. The three are not one place: Treixedo and Tondela are nine kilometres apart, Castro Daire is forty-five kilometres from Tondela, and fifty-four separate the two furthest. Worth noting; not worth believing. Both are distant, one has no documented link to Ivone at all, and the closest Portuguese match by a wide margin records no place more specific than “Portugal”. If Viseu were the family's district, the strongest match ought to show it. Open

A document has since walked into the same concelho from the other direction. Ivone’s own birth registration, supplied by the family in September 2026, gives her father’s parish as VIMIEIRO and her mother’s as NEGOZELA. Both are freguesias of Santa Comba Dão — and Nagozela and Treixedo were merged into a single parish in 2013, which is to say they adjoin. Her mother came from the parish this match’s parish was merged with.

Every caution above survives intact: the matches are distant, one has no documented link to Ivone, and the closest Portuguese match still says only “Portugal”. What changed is the size of the coincidence, not its status. It was a district — Viseu has twenty-four municipalities and half a million people. It is now one concelho of eleven thousand and two adjoining parishes inside it, arrived at from a register that knows nothing about DNA. That is worth more than a match list agreeing with itself, and it is still not a relative. Recorded as what it is: a coincidence that got smaller.

Yes — but not from this result, and the reason is arithmetic Tested

FamilyTreeDNA's Discover pages are public, and they list the projects whose members carry R-S11477: Ballew (16 members), Stafford (8), Blakely (8), Byrd (5), York (4), plus a Norman England group and a German-language group. Real surnames, English and Anglo-American, on Derrick's own paternal branch.

They cannot name his father, because surnames are about 800 years old and this branch is about 2,200. R-S11477's common ancestor lived around 200 BCE. Every Ballew and Stafford on it is a cousin at roughly ninety generations, descended from a man who lived before any of those names existed.

That is the theory. It was then tested against Derrick's own match list:

Every one lands between 10 and 28 cM, at the noise floor, and not one is near the close British cluster at 297–334 cM. The prediction was that these surnames would be genealogically meaningless. The match list agrees.

The arithmetic, done properly at last New, 11 October 2026

Those counts had been copied from Discover for three weeks and nobody had put a base rate under them. The base rate is the master gateway project: 21 of the 23,730 public members of “R1b ALL Subclades” sit on R-S11477 — 0.0885 per cent. That is what “a project contains this branch” looks like when nothing is going on.

The two geographic projects carry no signal at all. The branch sits in a large English project and a large German project at chance. A note here had told the next session to go and read their member lists; that instruction is withdrawn, because it would cost an afternoon and return nine men whose presence means nothing. R1b-P89.2’s 592× is circular — it is a project defined by a SNP on this very branch.

And the five surname projects, ranked below them, are enriched fifty-three to two hundred and ninety-two times. Which corrects a sentence this archive has printed: “no surname has surfaced and none is predicted to” is true in its meaning — nothing in the public Y record names Derrick’s biological father’s line, and that is not weakened — and misleading in its words, because five surnames do sit on this branch far above chance. The honest form: five surnames sit on this branch at fifty-three to two hundred and ninety-two times chance, and not one of them is Wear, Weir or Ware.

It does not make him a Ballew. A surname project’s enrichment says that one documented paternal line of that name carries this branch; it says nothing about which of the branch’s other members descend from it. What it does is name the places a paper trail would have to come from if one ever comes — Ballew, Blakely, Stafford, Byrd, York, in that order of strength, and nowhere else in the thirteen.

The countries flatten the same way. Raw counts read Anglo-German — United States 30, England 18, Germany 7 — but against each country’s own pool in the same database the United States is 2.3× base, England 1.5× and Germany 0.95×, which is exactly chance. Estonia reads 7.5× and is one man; it is written down only so nobody later reads it off a chart. And 167 of 242 branch members state no origin at all, so every rate above is computed on under a third of the branch. The denominator is a sample of men who joined a project, not a population, and that is the honest limit of the whole table.

What would work: a Big Y-700, which would place Derrick on one of the 42 branches below R-S11477 — terminal branches in a well-tested lineage are often only three to six centuries old, and at that depth a shared branch usually does mean a shared surname. It would also open FamilyTreeDNA's Y-DNA match list, which is the literal answer to the question and which MyHeritage does not provide at all: a list of men sharing the paternal line, each with their earliest known ancestor's name, year and birthplace.

But the Y is not the best tool available here, and it is worth saying so. The Halsey match is Derrick's first cousin's daughter at 297.5 cM — which means their common ancestors are Derrick's biological grandparents. The surname being looked for is that grandfather's, and it is in her family tree, not in anybody's chromosome. The Harrison match is a second cousin at 333.9 cM, in the United Kingdom. Two messages could produce a name; a Big Y produces a branch number and a probability. Both are worth doing, in that order.

The label does not change, and that is the point Inferred

The method page has said since this archive was built that the paternity finding stays Inferred until a Y-DNA result moves it. The result has been run. It did not overturn the finding — it corroborated it from an independent direction — and there is no further test available, because Jacob's male line has ended.

So the tag stays Inferred. There is still no document and still no direct comparison, and a finding does not become Documented because it survived a test. What has changed is that the test nominated to break it was run, and it held. The archive can now say so without hedging.

Two things would sharpen it further. Big Y-700 — MyHeritage calls its own result a low-resolution estimate; a full sequence would pinpoint the terminal branch and place Derrick in a surname project, and the nearest surname in the British cluster is Halsey. And one message: if any of the Halsey, Harrison or Browne matches is a man descending father-to-father from that family, his Y is directly comparable to Derrick's, and that costs nothing at all.