I wore it to formal occasions through school and university. It’s how Lucy first noticed me. She came up to me at the Rowing Club ball, started a flirty conversation about the kilt, and we danced.

I wore Clanranald because I was a Murdoch, and Murdochs, I’d been told, were a sept of Clan Donald. The MacDonalds of Clanranald came from the western Highlands and the Hebrides. Highlanders. Hard people who survived on oats, rain, and grievance. The tartan said I was one of them.

In August 2013 we visited the Isle of Skye, and stopped at Armadale Castle, the home of Clan Donald. At the visitor centre I started a deeper search. I met with the clan archivist. My question was practical: which part of the MacDonalds did I come from? The clan map showed Clan Donald territories scattered down the west coast of Scotland: Islay and Kintyre to the south; Skye and the Uists, Moidart and the western Highlands to the north. This meant my ancestors had a truly smorgasbord-like selection of regional wars to choose from. Murdoch could have come from any of them.

Home of Clan Donald, Armadale Castle & Gardens, Isle of Skye, August 2013

Her answer was the tartan. Whatever your family handed down, she said, was likely a good indication. Wearing a tartan different from your clan was a modern thing, the last fifty years or so. My hand-me-downs were older than that.

MacDonald of Clanranald. Western Highlands and Hebrides. Confirmed.

I liked the idea of being a Highlander. I felt a visceral connection to the landscapes of the Isle of Skye and Glen Coe. Since emigrating to the US in the early 2000s, that had only grown.

Glen Coe, July 2022.

I came away with the source material too: photographs of Black’s ‘Surnames of Scotland’, of the Clan Donald sept guide, of the clan map. They’d matter later.

During COVID I built out a first pass of the family tree, sketching the lines, filling in names and dates. A lot of it leaned on existing trees, which notoriously overreach. Everyone finds a famous ancestor, like Cousin John. In 2025 I revisited it, this time forcing a ground-truth view: every claim tied to an original document, the digital scans from the Scottish and English government archives.

I worked back through my direct Murdoch paternal line first. Each generation had to have an original document: birth, marriage, or death. The line ran back through the Clyde at Old Kilpatrick, through Lanarkshire and the industrial belt around Glasgow, and then, before that, into Ayrshire. Kilmarnock. Ayr. Ochiltree. Irvine. Five generations in a row, the early 1700s back into the early 1600s, all of them in Ayrshire. The trail goes cold around 1618, with a Johne Murdoch whose birthplace isn’t recorded. Records become sporadic or absent before then.

Ayrshire is not the Highlands. Ayrshire is Burns country. The Lowlands.

I grew up near Loch Lomond, at the start of the Scottish Highlands. My Murdoch ancestors did not.

Luss Pier, with a snow-capped Ben Lomond in the background, Loch Lomond, March 2016.

Reading the records is its own challenge. The handwriting is cramped, spellings pre-standardised, abbreviations baked into Scots legal usage that a modern eye doesn’t catch. I worked through the hardest of them with ChatGPT.

One word kept appearing: “umql”, sometimes “umqle” or “wmqll”. It was the abbreviation for “umquhile”, a Scots word meaning “the late”, “deceased”. Standard shorthand in seventeenth-century marriage, baptism and testament records, and entirely opaque to a reader who didn’t know it. The AI knew it.

Marriage of William Murdoch and Agnes Murdoch, Ayr, 11 February 1698. Old Parish Registers, Marriages 578, p. 51. ScotlandsPeople.

“Apud Ayr, Feb: 11. 1698. Quo die William Murdoch student of philosophy at the College of Glasgow and Agnes Murdoch lawfull daughter to umqll John Murdoch Merchant in Clyde [had] their names to be proclaimed in order to their Marriage. Consigned a Ginney money. Were Married on March first 1698 at Dumbarton by Mr James Dickmore.”

This is William Murdoch’s marriage. Born in Ayr in 1678, three generations of Murdochs surround the font at his baptism, his father was a maltman in the town. By 1698 he was a student of philosophy at the College of Glasgow, marrying Agnes Murdoch, the lawful daughter of the late John Murdoch (a different Murdoch family), merchant on the Clyde. They put down a guinea as security against the banns, married three weeks later at Dumbarton, and in September retrieved the deposit.

A philosophy student. A merchant’s daughter. A maltman. These were the Murdochs in 1698. Highly civilised, gainfully employed, and completely devoid of Claymores. Lowlanders.

What I learned only later, reading Arthur Herman’s ‘How the Scots Invented the Modern World’, is that the named clan tartans are themselves a modern invention. The system was orchestrated by Sir Walter Scott for the King’s Jaunt of 1822 — George IV’s visit to Edinburgh, the first by a reigning monarch in nearly two centuries. Scott stage-managed the pageantry: chiefs in tartan, bagpipes, the whole Highland tableau. Most of the named clan tartans worn for that visit were invented for it.

The archivist’s logic at Armadale (your tartan tells you your clan) rested on a system barely two centuries old.

My kilt was a hand-me-down. My father wore it before me. So did his. What the Ayrshire records show is that the Murdochs of Kilmarnock and Ayr in the 1600s would not have worn Clanranald. They probably wore no tartan at all.

What the records don’t show is what came before 1618, where the trail goes cold. The surname itself has Gaelic roots: Murdoch from Muireach or Murchadh, meaning roughly “sea warrior” or “great chief”. On the standard sept rolls, Murdoch is listed under Clan Donald, and sometimes under others. That association is real at some depth, but it operates below where the records reach. Perhaps an ancestor finally grew weary of the weekly calendar of mutual clan slaughter, realising that surviving past thirty on an Ayrshire croft beat dying face-down in a bog for Clan Donald’s honour. The deep origin is genuinely unknown.

The precision was always the reach in any case. Clan Donald spans Sleat, Glengarry, Keppoch, the Lords of the Isles, and Clanranald, among others. The archivist’s reasoning at Armadale ran two layers: that a hand-me-down tartan tells you your clan, and then which branch. Both layers rest on Scott’s invention. Even granting the first, the second is a guess.

But somewhere in the nineteenth century (after Scott’s pageant, in the Victorian Highland revival) a Murdoch chose this cloth. Another wore it next. Then another. Then me, aged seven. Then, one evening at a Rowing Club ball, Lucy noticed it.

The tartan is a modern invention. So was the claim it was meant to carry. But the wearing of it (generation after generation of Murdochs reaching into the same fabric) is its own tradition. Not the clan’s. The family’s.

Highlander or not, the tartan stays.

Postscript: I would be remiss, with this Substack’s focus on ground truth, to not point out a persnickety detail in the above photo. The tartan that younger me is wearing is actually the ‘Dress’ version of the main MacDonald tartan. There is no ‘Dress’ version of the MacDonald of Clanranald sett. This is the kilt my Murdoch grandparents had made for me aged 16, and I preferred this sett.

The No True Scotsman fallacy is the move you make to protect a stereotype from a counterexample. You hold a picture of a Scot in your head: dour, thrifty, suspicious of show, canny to the point of meanness, Presbyterian to the bone, litigious when his interests are crossed, slow to part with money and slower to forget being short-changed. Someone produces a generous, sunny, open-handed Scot who buys the next round without checking the time. You say: “No true Scotsman would.” The fallacy edits out the exception so the picture stays clean.

Antony Flew named it in the seventies, but the move is older than philosophy. You may know it better as moving the goalposts. You believe the stereotype, you meet the counterexample, you tighten the rule. The goalposts shift; the cliché walks on, undisturbed.

It is why politicians reach for it. A standard that moves every time it is tested cannot be met, and cannot be failed.

My great-great-great-grandfather destroys the fallacy. Alexander Murdoch is the stereotypical True Scotsman made flesh. I find this more familiar than is entirely comfortable.

Alexander Murdoch was born at Hallhill, in the parish of Old Monkland, on 17 December 1799. He was a tenant farmer at Hilton, in the parish of Cadder, north of Glasgow. He died there sixty-five years later, at twenty past three in the morning on 13 October 1865, of typhoid fever with jaundice. The son who signed as informant on the death certificate was the Reverend Alexander Murdoch, minister of Loch Ryan near Stranraer. He was the eldest son, named after his father, who had been named after his.

That is the foundation of Alexander’s story. The rest is in the papers.

Start with the pony. He found it on his property in October 1854, and there is no record of how it got there. The neighborly response would have been to ask around. The lazy response would have been to keep it.

Alexander Murdoch hired solicitors.

Notice of Petition, Alexander Murdoch v. Owners of a Brown Pony. Glasgow Herald, October 1854.

The firm was Gordon & Leitch, of 3 Royal Exchange Court, Glasgow. They drafted a petition to the Sheriff of Lanarkshire, and on 6 October 1854 the Sheriff pronounced an order: notice was to be advertised in the Glasgow Herald, and unless the pony’s owner came forward and paid the expenses incurred (the cost of process, of advertising, and of keep) within eight days of publication, a Warrant of Sale would be granted as craved.

The recursion deserves a moment. He had found the animal, and he could not advertise it without incurring expense, and he could not recover the expense without advertising it, and he could not sell the pony without an order of court. So he went to court to obtain an order, to advertise, to recover the cost of going to court. The pony was the collateral.

There is no possible outcome of this proceeding in which Alexander loses money on the pony.

The pony was not an isolated case. The Herald carried at least two other notices placed by him, both about livestock, both filed under his name at Hilton. In both, Alexander is making sure he does not lose money on an animal.

In one, he is the aggrieved party. Twenty-two black-faced tup hoggs and one black wether Leicester hogg, with two crosses, had been stolen or strayed from his farm at Hilton, near Bishopbriggs. A reward was offered to whoever could provide information leading to recovery. Information was to be given to him directly, or to a Mr James Coubrough at Blairlummock, Campsie. He had thought through the logistics.

Stolen or Strayed notice, Alexander Murdoch, Hilton. Glasgow Herald, 9 March 1855.

In the other, he is the finder. About a score of sheep had appeared on Hilton Farm, near Cadder. The owner could have them back on proving his property and paying expenses. Apply to Alexander Murdoch, Hilton.

Sheep Found notice, Hilton Farm. Glasgow Herald, 14 August 1848.

These are not the same incident. In one he has lost twenty-three sheep and would like them returned. In the other he has acquired about twenty and would like to be paid for keeping them. The principle is the same. Alexander Murdoch was not going to feed another man’s sheep, and he was not going to lose his own.

When he died in October 1865, the inventory of his personal estate ran to four full categories. The first item is almost too on the nose for a parsimonious Scot. Cash in the house: £1. Not “about a pound,” not “a small sum on the premises.” A pound.

Someone had counted it.

The rest is what you would expect of a man who counted:

• Household furniture: £47, 8s 6d. Appraised by a licensed appraiser named John Slater, because of course it was.

• Farm stock and implements of husbandry: £782, 5s 6d. Appraised by the same.

• £50 owed by a grain merchant on Cambridge Street.

• £163 from the City of Glasgow Bank, with interest tidily entered to the date of death.

• £51 in half-years’ rent from property at 70, 72 and 74 Rutherglen Loan. He was a landlord as well as a farmer.

• £250 lent to the Union Railway Company, being the price of property at Saltmarket the deceased had sold to the company. He had sold them the land and lent them the purchase price.

The total personal estate was £880, 14s, of which exactly £1 was within reach.

The settlement itself had been executed by him on 12 September, three weeks before he died. The codicil was annexed by the executors on 14 October, the day after death, and the whole thing registered at the Sheriff Court at Hamilton on 27 October. A typhoid fever with jaundice that runs three weeks gives a man time to put his affairs in order, and he had used the time. He did not die intestate. He died, in short, with his books closed.

The other thing the documents tell you about Alexander Murdoch is that the Murdochs of Lanarkshire kept marrying Bowmans, and saw no reason to stop.

His father, also Alexander and also a farmer, had married Isobell Bowman at Old Monkland in 1792. She was from the Cambuslang Bowmans, the daughter of a James Bowman who farmed there. The marriage produced a working farm and a stable household, which were the things a marriage was for.

So in December 1827, four days short of his twenty-eighth birthday, Alexander married Martha Bowman at Barony. A different Bowman family (Martha was from Middlequarter, in the Calton, daughter of John Bowman and Martha Law), but a Bowman nonetheless. He had married, in essence, his mother’s surname. Whatever objection might have been raised does not appear in the record. The marriage produced thirteen children and forty-six years of tenancy at Hilton.

The system was working.

In October 1875, ten years after his own death, Alexander’s son Robert Graham Murdoch came back to Old Kilpatrick with a motherless daughter. I have written in a previous piece about his years on the Demerara, and the two wives he buried in those years. What he did next was what his father would have done, and what his father’s father had done before that: he married a Bowman. Janet Bowman of Old Kilpatrick, no relation to either of the previous two. A third Bowman, a third Murdoch generation, a third parish.

He had a method for getting his life back on track, and he used it. Both Alexanders certainly would have approved.

You could explain this as demographics. Lanarkshire farming families drew from a small pool; surnames recur. The Murdochs would not have explained it. They would have said the method worked.

Try the fallacy on Alexander Murdoch and watch what happens. Did he take a stray pony to the Sheriff Court? He did. Did he keep £1 in the house and £250 out on loan to a railway company? He did. Did he advertise lost sheep in the Herald and then again found sheep, to avoid incurring any financial loss? He did. Did he marry his mother’s surname, send his eldest son to the kirk, and die three weeks after settling his estate, at twenty past three in the morning of a typhoid fever he had clearly been outlining a response to? Yes.

The picture stays clean. There is nothing to edit out. Alexander Murdoch is the One True Scotsman.

He is also, I should note, my great-great-great-grandfather, and the line is not entirely broken. I once took Whirlpool Corporation through what I can only describe as a sustained correspondence over a leaky KitchenAid dishwasher, and came away with compensation, a formal written apology, and a shiny new KitchenAid stand mixer.

The reel was unlabeled. The yellow Kodak box it came in was labeled ‘Oslo’ and must belong to another reel entirely that I have yet to uncover. What I had was three minutes of 8mm Kodachrome, over 60 years old with the plastic spool holding the film breaking apart. I ran it through a Kodak digitizer myself.

The first thing it gave back was a man in dark trousers and a soft brimmed hat finishing a golf swing on a links course, gallery in coats behind him. I had no date and no place. I asked. The answer came back: Arnold Palmer, St Andrews, 1960.

Cine still. Royal Troon practice rounds, July 1962

That would have been a good answer. The 1960 Open at St Andrews was Palmer’s first, the Centenary, the tournament that brought American professionals back to the Championship. Kel Nagle beat him by a stroke. It is the Open everyone remembers from that decade.

It was also impossible. My father was in the United States that July. The diary, transcribed a few weeks earlier, has him in Detroit on the 9th, the day the Open ended, and aboard a Britannia aircraft out of Idlewild the night of the 23rd. He did not land in Britain until the day after the tournament was a fortnight in the past.

So the model had the player right and everything else wrong. The diary closed off 1960. That left the rest of Palmer’s appearances at Open Championships, of which there are many, and a separate complication. Through the 1960s Palmer played a series of exhibition matches at Scottish courses. Any of those was a candidate. 1960 St Andrews is the most-photographed Palmer event in Britain by an order of magnitude, and the model was defaulting accordingly.

The clothing and the context closed it. Dark trousers, a soft brimmed hat, an overcast sky. A gallery in suits and overcoats. Multiple balls visible on the turf in front of him and no playing partner in frame. Tournament dress and gallery turnout, but a practice round, not competition. A frame-grab matched against the standing record of Palmer in tournament play narrowed it to Royal Troon, July 1962, the practice rounds for that year’s Open. He was defending champion, having taken Birkdale the year before. He would win again at Troon the same week.

Cine still. Royal Troon practice rounds, July 1962

Palmer won that week, by six strokes from Kel Nagle, the man who had beaten him at St Andrews two years earlier. Nicklaus, twenty-two and just professional, played his first Open in the same field.

My father was twenty. He was between his second and third years at Cambridge, reading mechanical sciences, on his summer holidays. He had a cine camera. He had been filming since he was fourteen.

What came next on the reel was different. The colour gave way to black-and-white, and the frame was no longer outdoors. Ian had filmed his own television. A man in a suit, appearing to sing (there is no audio on 8mm film). A studio set behind him. A caption I could just read across the bottom of the screen: “FIRST TV TRANSMISSION FROM FRANCE”.

Cine still of BBC broadcast. 11 July 1962

The identification was straightforward once the caption was visible. Telstar 1 had been launched from Cape Canaveral the day before, on a Thor-Delta rocket. The job it was built to do had not been possible by any other means.

The transatlantic cables had carried voice and telegraph since the 1850s. Television was bandwidth on a different level that the cables of the day could not handle. Until Telstar, video crossed the Atlantic on aeroplanes: the BBC’s coverage of the Queen’s Coronation in 1953 reached American audiences by being developed and edited in the air, on a flight from London to New York.

Telstar was the first means of crossing in real time.

It worked at once. On the night of the 10th, a panning shot of the American flag at Andover, Maine, with The Star-Spangled Banner playing, was relayed to Pleumeur-Bodou in Brittany and to Goonhilly in Cornwall. The next night the French replied: a tape of Yves Montand singing La Chansonnette over images of Paris by night. The BBC carried the French side that week, with the caption Ian filmed.

Ian was in a sitting room in Glasgow. He understood the significance of what was on screen. The technology, the cultural impact, and what that would mean. He loaded the cine camera and pointed it at the television.

It was the same week as Palmer’s practice rounds at Troon.

The third element of the reel was sailing. A dinghy with a red sail, calm water, a pale building among trees on the far shore. The colour was good and the day was fair. I had no idea where it was.

Cine still. Gareloch, July 1962

The model offered the Lake District, and a building to match. I looked: a different building altogether. The water was right for the region, the wooded shoreline plausible. But the same reel had Palmer at Troon and the BBC’s Telstar broadcast off a television. Three minutes of Kodachrome was unlikely to span half of Britain. The location had to be Scottish, and reachable from Glasgow in a morning. That still covered a lot of water.

The next guess was Ardenvohr House on the Gareloch, which the model said had become the yacht club. The building in the frame was white. The yacht club was not. I grew up by the Gareloch. The white building wasn’t immediately obvious to me. I went to Google Maps and worked along the shoreline in Street View. The match came on the Ardencaple Hotel at Rhu. Plenty of houses on the loch are white, but the Ardencaple has a particular shape, with two smaller buildings attached at the sides, and a set of windows that arrange themselves across the front in a pattern that doesn’t repeat. All of it was in the frame.

Cine still. Gareloch, July 1962

Ardencaple Hotel, Rhu. Google Street View, 2026

There was a reason the loch felt unfamiliar. Construction of the Polaris submarine base at Faslane began in 1963.

At Nassau, the previous December, Kennedy and Macmillan had agreed Britain would have Polaris. Faslane was named as the base the following April. The ground had not yet been broken. The Gareloch in Ian’s footage was the one before any of it.

The Gareloch I grew up on came later.

That my father sailed the Gareloch and that I later did is not much of a coincidence. Glasgow sailing families went to the Gareloch. Most of them did. The Ardencaple was the village landmark for both of us, in his time and in mine.

None of this was Oslo. That reel I have yet to find.

What the reel returned, in the end, was not a coincidence of place. It was a week. Troon practice rounds, Monday and Tuesday the 9th and 10th of July. The Yves Montand transmission from Pleumeur-Bodou, Wednesday night the 11th into Thursday the 12th. The Gareloch, a fair sky and a red sail, the weekend that followed. Three minutes of Kodachrome that until recently sat unwatched, now located in time and place.

One week, recovered.

Subscribe now

Milton dictates to his daughters, by Eugène Delacroix, (Public Domain Image)

The claim came down via my grandmother, who’d had it from her grandparents, and she passed it on the way you’d pass on a slightly suspicious heirloom; with a raised eyebrow and the implicit understanding that you were free to make of it what you wanted. She was acerbic and funny and would have enjoyed what follows.

The route ran through her mother’s line, the Woolleys, back through the Andrews and the Hawes and eventually to a Sarah Milton, born in Harefield in 1813, who married into the line at St Marylebone in 1835. From Sarah you climb backward through three more generations of Miltons in Ickenham and Ruislip (William, William, James) until you arrive at a John Milton baptised at St Botolph without Aldersgate, in the City of London, on 30 March 1705. That John is the hinge. Everything before him is the question.

The Ancestry tree, when I first got stuck into it, was confident. It linked our John to a father, also John Milton, born in Richmond, Surrey in 1680. From there it leapt three more generations and landed squarely on Sir Christopher Milton, the poet’s younger brother, the Royalist lawyer, Baron of the Exchequer under James II. From Sir Christopher to his brother John is one short step. The chain looked, on first inspection, like a chain.

Then you start pulling on the chain. Or rather, you ask an AI to help you pull on it, and then you spend a weekend checking what it finds, because AI is exactly as confident as Ancestry and exactly as wrong, until you make it look at the actual documents.

St Botolph without Aldersgate, parish register, 30 March 1705. London Metropolitan Archives

The 1705 baptism entry, when you find the actual image, is four words long. “John, son of John Milton.” No mother named, no occupation, no honorific. Sir Christopher had been knighted nineteen years earlier; any grandson of his baptized in a London parish would have been entered as “Mr.” or “Gent.” at minimum. Plain entries belong to plain people. The clerk who wrote those four words was registering a tradesman’s child, not a baron’s.

The Richmond link fares no better. There is a John Milton baptised at Richmond in October 1680, exactly as the Ancestry tree claimed. But when you read the actual parish register (there’s a free transcription on the Internet Archive, edited by a Mr Smith for the Surrey Parish Register Society in 1903), the entry reads “John, son of William Milton.” Not John. William. Ancestry had the date right and the parentage wrong, which is a particular kind of wrong: confident, plausible, and undetectable unless you bother to look. Richmond turns out to have had its own perfectly ordinary Milton family (Robert, William, Alexander, fathering children across four decades) with no apparent connection to the poet’s people at all.

And the poet’s people, it transpires, run out. The standard accounts (David Masson's Victorian biography of 1880, updated by John T. Shawcross in 2004) tell the same story. The poet had one son, who died as an infant in 1651. His brother Sir Christopher had ten children. The sons either died young, died unmarried, or had only daughters. The last possible male-line grandson, a Charles born around 1684, vanishes from the record without a documented marriage. The youngest son, Richard, was a Catholic lawyer who fled to Ireland after 1688 and died there a bachelor in 1713, his probate noting tidily that he left no heirs. The line ends.

So the Ancestry chain was wishful thinking attached to a real baptism. The poet’s line dies with the seventeenth century. The Richmond Miltons are someone else’s family. The Ickenham Miltons appear to be a third unrelated cluster. Our John of 1705 was almost certainly a London tradesman whose father was another London tradesman, and the name Milton was, in the City of London in 1705, just a name.

I should have stopped there. The honest conclusion was on the table: the family story was wrong, and the neatness of the direct line was the giveaway. But I’d already spent a weekend on it, so I kept going, and I started reading about something called pedigree collapse.

Here is what pedigree collapse (a.k.a. implex) says. You have two parents, four grandparents, eight great-grandparents, and so on, doubling each generation. By the time you reach the early 1500s (roughly the era of Richard Milton of Stanton St John, the poet’s grandfather and the deepest Milton our chain even gestures at), you have, in theory, about thirty-three thousand ancestor-slots in that single generation {for the pedantic mathematicians amongst us, yes, ~n=2^15}. The trouble is that there were not thirty-three thousand distinct people available to fill them. There were villages of a few hundred. Cousins married cousins. The tree folds back on itself, again and again, until what looked like a branching structure becomes something more like a braided rope.

The mathematician Joseph Chang worked out the consequences in 1999. For any population that’s been mixing for long enough, you don’t have to go back very far to hit a person who is an ancestor of everyone alive in that population today. For Europeans, the figure is around six hundred years. Anyone living in 1400 with descendants at all is an ancestor of essentially every European alive now. Richard Milton of Stanton St John was born in roughly 1535. He is comfortably inside that horizon.

Which means that if Richard Milton has any living descendants (and he does) he has them in the tens of millions. Anyone of English descent is almost certainly descended from him. So is anyone whose ancestors married into English stock at any point in the last four hundred years, which is most of the Anglophone world. The poet himself, John Milton of Paradise Lost, has no surviving male-line descendants; but his grandfather is an ancestor of, conservatively, half the people I went to school with.

The family story, in other words, is true. Just not in the way it was meant.

What my grandmother passed on, with her raised eyebrow, was not really a claim about descent. It was a claim about membership; membership in a very large and entirely uninteresting club that contains most of England and large portions of the New World. The lore feels specific because it names a specific person. The mathematics is the opposite of specific. It is a way of saying: you, and roughly fifty million others.

This is, I think, the secret embarrassment of amateur genealogy (of which I am most certainly a participant). The further back you push, the less your discoveries belong to you. The Norman ancestor everyone is so proud of is also somebody else’s Norman ancestor, and somebody else’s, and somebody else’s, down the pub and across the road and over the Atlantic. The thinness of the line is the whole point. A tree with thirty-three thousand slots at the top is not really a tree. It is a delivery mechanism for the comforting illusion that you, specifically, came from somewhere.

My grandmother would have enjoyed that.

So I did not descend from John Milton. I am, however, in all probability, descended from his grandfather, along with most of the people reading this. Cousin John, it turns out, is as much your cousin, dear reader, as he is mine.

My wife, meanwhile, is descended from Oliver Cromwell, apparently.

Subscribe now

The diary in question is my father’s. He kept it on a five-month journey across North America in 1960, when he was eighteen, a trip sponsored by Royal-Globe Insurance (a Queen Mary crossing, work at the Goodyear plant in Akron, then a grand tour: Washington, Houston, Dallas, Phoenix, the Grand Canyon, Los Angeles, San Francisco, the Pacific Northwest, Chicago, Detroit, Toronto, Montreal, Quebec, the Catskills, Cape Cod) before flying home from Idlewild that July. I only found the physical diary in the last month. I had known vaguely that the trip happened but hardly any detail. I couldn’t ask him about it as he died when I was very young.

He also took a cine camera. On day one, watching the New York skyline come into view from the Queen Mary’s deck, he writes: “The Statue of Liberty was smaller than I had expected — maybe because I was aboard the Q.M. It was also much closer to the N.J. mainland and I was unable to get a cine shot without also taking in N.J. gasometers.” Two records of the same journey, in two media. The diary I could read with effort. The cine film was proving more challenging to digitize.

The obvious next step was to feed the diary into the latest AI models.

Specialist handwritten-text recognition (Transkribus) gave me a complete pass with errors I could see. “The fraud langan” for “the Grand Canyon”. “Awesane chasur” for “awesome chasm”. Sometimes its output drifted into German mid-page (”Efter welche das Recht, wenn der Vormundschaft des Mannes…”), which I took at first for the recognizer losing itself in a foreign training corpus, until I realized it was trying to read honestly what was on the page. Ian wrote in fountain pen on both sides; the ink bled through. Transkribus was looking at the back of one page through the front of it, and reporting what it saw. The seam was always visible. You knew where the trouble was.

Then I tried Gemini. Gemini produced something else: smooth, plausible, fluent prose. No visible errors. It read like a diary.

It was, increasingly, fictional.

What made me look closely at the manuscript was the Pacific Northwest. The Gemini version of the leg felt thin: the prose had the rhythm of summary rather than diary. I went to the handwriting.

What Ian wrote about his last day in Seattle, 28 June 1960, was four short sentences:

“Last day in Seattle. Ruth drove me out to the skiing area and Thunderbird hotel which we climbed up to. We took the ski lift down. The falls we visited on the way back.”

The Gemini version of those four sentences:

“Wednesday the 28th was the last day. I wrote to Seattle. Ruth took me to the sing and Thunderhead Hotel which we linked up to. We visited the falls on the way back.”

The date is wrong. The 28th was a Tuesday. “The skiing area” has become “the sing”. The hotel has become “Thunderhead Hotel”, which is a real brand of American motel; “Thunderbird” is not. And the sentence “We took the ski lift down” is missing entirely. It is the sentence that lets the place be located.

Sixty-five years later, the Thunderbird Lodge of 1960 is no longer there. It sat at the Snoqualmie Pass ski area, an hour east of Seattle. The modern interstate was still under construction; the road off the pass took you down past the town of Snoqualmie, and immediately past Snoqualmie Falls. “The falls we visited on the way back” names what Ian had no need to name. You may recognize the falls as those by the hotel in David Lynch’s Twin Peaks.

Had I been working from Gemini, I may have searched for “Thunderhead Hotel” and found nothing of use. The confident wrong noun would have closed the route, and the missing ski-lift sentence would have left the visit floating in geography. The model’s failure was small; the route it would have closed was the one that mattered.

In 1996, aged twenty-two, on my own American grand tour after graduating university, I visited Snoqualmie Falls. In 2004 I moved my family from London to Seattle and we lived, first, in the town of Snoqualmie itself.

Three Murdoch arrivals at the same place across forty-four years, and the leg of the trip that put my father on that ground was the one Gemini had silently mangled. If the model had been fluent enough that I had not noticed, I would never have known he had been there at all.

What worked was harder. It took weeks, several models in different roles, and a discipline against the failure mode that destroyed the Gemini run. Transkribus stayed on as the spine; a second LLM did reconciliation against the scans, applied one entry at a time, with a forced pause every three or four entries against silent drift. I built a primer for the LLM of period vocabulary and letterform pitfalls, a Names Gazetteer, a Places Gazetteer that grew with each pass. Then I did a proper review.

The failure was visible at scale, before any line-by-line check. Three artifacts survive from the work. Transkribus came in at 27,910 words; Gemini at 21,662; the reconciled text (what my father actually wrote) at 24,262. The first overshot, fed by bleed-through and garbled non-words. The second undershot by about eleven per cent, the loss not evenly spread but concentrated where the trip moved the fastest (and mattered most to me). The third is the actual scanned handwritten diary.

The cine film reels exist. They cover the same trip, the same Queen Mary deck, the same Akron streets, the same Grand Canyon, and maybe, just maybe, of my dad at the falls. I have a photo of me there in ‘96.

I have started viewing the cine film. The reels are 8mm Kodachrome, sixty-five years old, with splices that come apart in your hands as you work them through the digitizer. The first thing the recovery returned was the Statue of Liberty, viewed from the deck of the Queen Mary, with a figure leaning at the rail. It is the shot Ian writes about on day one — the one he could not get without also taking in the New Jersey gasometers. Sixty-five years later, the gasometers are gone. The shot is here.

The slides are still in shoeboxes I have yet to open. Whether the falls are in there, or in the reels still to come, I do not yet know.

What comes next for the cine film is the work the diary needed, in another medium. Reading motion. Recognizing a face at eighteen in a frame his children have yet to see. Identifying a place from a few seconds of footage. Restoring the color and cadence of Kodachrome that has gone mauve at the edges. The physical restoration and digitization is doable today by film recovery professionals. But deep contextual recognition of the film and compelling digital restoration is currently a complex, expensive workflow across multiple software tools, requiring extensive skill. In the next few years, no doubt the AI will integrate these tools, initiated by simple prompt. But I fear they will need the same discipline the diary transcription did: a stack, a refusal to trust fluent hallucination, evidence as ground truth.

The optimistic case for AI is not that it is magic. The optimistic case is the one I have just described. Without these tools, the diary stays half-readable, the names stay garbled, the cine films stay unwatched, and a generation passes. With them, and with the discipline to use them honestly, what comes back is real. Words my father wrote that I can read in his voice, a hotel name spelled correctly, a ski lift, the falls he saw on the way down, and the Statue of Liberty as he saw it on his first day in America.

We are the generation that can bridge back. The bridge is not free, and it is not automatic, and it is easy to do badly. But it is real. And the rest of the cine film is spooling.

Subscribe now

We flew into Timehri International, four years before it was renamed Cheddi Jagan, and drove north along the East Bank Public Road. The airport sits south of Georgetown on the east bank of the Demerara River, and the road threads the old sugar belt: villages, drainage ditches, cane fields stretching back from the road on the inland side, the river hidden behind a strip of houses to the west. It was less developed than it is now. I remember disembarking the plane and being hit by an intense humidity that my Scottish heritage was ill-suited for.

Timehri International Airport, Guyana, 1993. End of the expedition.

What I did not know, and would not learn for thirty years, was that one of those villages had once been a sugar estate called Providence, and that in 1867 a Scotsman of twenty-five had taken up his post there as manager.

The path to it ran through ancestry research. I’d been working through my family lines properly; first-party documents, parish registers reaching back to the early 1600s, the statutory records Scotland has kept since 1855; names and dates cross-checked against the source. When I got to my great-great-grandfather, the record showed three marriages in close succession. I assumed I’d conflated two men with the same name and went back to verify.

Three marriages, one man.

Then I looked at where the second one had taken place. Georgetown, British Guiana. That had to be a mistake too. I went back again.

Three marriages, one man, and a stretch of years on the Demerara coast.

His name was Robert Graham Murdoch. He was born in Cadder, Lanarkshire, in 1842, a few miles from the Old Kilpatrick parish where Murdoch and Bowman families would intermarry across at least three generations. He had grown up on the doorstep of Glasgow, in a Scotland whose shipyards were rising on the Clyde and whose young people were going out to staff the Empire. He had married Elizabeth Noble Houstoun earlier that year at Plantation Lusignan, a different estate further along the coast. By 1869 he was running Providence.

The records of 1869 are sparse but they are not silent. In May, Elizabeth died at twenty-eight; the death is registered at Lamaha Street in Georgetown. That same year, in July, the shovel gang at Plantation Leonora on the West Coast struck over withheld wages, the first of a wave of disturbances across the Demerara estates among the indentured Indian laborers who cut the colony’s cane. In November, Robert remarried, to Mary Uranie Hounslow, in Georgetown. They had a daughter in 1871. Mary returned to the UK and died in London that December at twenty-one, two and a half years after Elizabeth. The daughter, Martha, survived. Robert took her home to Old Kilpatrick and four years later married my great-great-grandmother, Janet Bowman.

That is what the records say. They do not say what he carried with him on the boat home from Demerara, and I do not propose to imagine it.

Placing myself on his ground took some work. I went to the maps to align my 1993 route with his: Lusignan, Providence, Georgetown, the East Bank Public Road. I dug out the slides I’d taken on the expedition, scanned years ago, and matched what I’d photographed against the coast as it is now. Gemini filled in the plantation’s own history.

What I can say with certainty is the geography. Providence sat on the East Bank of the Demerara, on the road that connects the airport to Georgetown. The estate is gone; the village name remains, and the cane lands have become other things. But the line on the map is the same. In 1993 I drove through what had once been his ground. I had no idea. I went on to Lethem, and to Georgetown, and to the Berbice. He had stayed where I was passing through until two wives were dead and a daughter was old enough to travel.

The interval was a hundred and twenty-six years. The road was the same road.

I don’t want to make more of this than it is. It isn’t reincarnation or design or a hand on the tiller. It is a coincidence. But coincidences have shapes, and the shape of this one is a stretch of tarmac on the East Bank of the Demerara. The same line of road, two Murdoch arrivals, a hundred and twenty-six years between them. He came for sugar. I came to build a school. Neither of us knew about the other when we landed.

I know about him now.

This is the kind of thing I keep finding. The records of the analog century are still in reach, the people who knew the people are still living, and the tools have caught up with the records. We are the generation that can bridge back. There are more echoes than I had expected.

Subscribe now