A boy puts his hand up
In 1936 a school party from Stowe was taken to Burlington House in London for an exhibition marking fifty years of the British School at Athens. The main attraction was a person rather than an object: Sir Arthur Evans, then eighty-five, excavator of Knossos, displaying the inscribed clay tablets he had pulled out of Crete three and a half decades earlier.
Evans explained that nobody could read them. Nobody knew what language they recorded. Nobody even knew whether the marks stood for letters, syllables or whole words.
A fourteen-year-old in the group put his hand up and asked whether the tablets had ever been deciphered. Told they hadn't, he decided — with the flat certainty available only to the very young — that he would do it himself.
That is normally the sort of story that turns out to be invented. This one isn't. The boy was Michael Ventris. Four years later, aged eighteen and still at school, he published his first paper on the script in the American Journal of Archaeology, under the title "Introducing the Minoan Language." He did not mention his age to the editors.
Sixteen years after that he was proved right. Weeks after the book proving it went to press, he was dead — killed on 6 September 1956 when the car he was driving struck a parked lorry at Hatfield, a few miles from where he was born. He was thirty-four. The coroner recorded accidental death. That was seventy years ago yesterday.
This is the story of what he did, and of the woman who did the hardest part and didn't live to see any of it.
The problem with no key
Decipherment usually needs a cheat. The Rosetta Stone worked because the same decree appears in Egyptian and in Greek, and scholars could already read Greek. Give a code-breaker a bilingual and the rest is labour.
Linear B had no bilingual. Worse, it had two unknowns at once: an unknown script and an unknown language. Every guess about the sound of a sign depended on an assumption about the language, and every assumption about the language depended on the sounds. There is no obvious place to put the first foot.
Evans made it harder still. He named the script's world "Minoan" and was certain the language was not Greek — he preferred some indigenous Cretan or Anatolian tongue, on the reasoning that Crete's civilisation was older and grander than anything on the Greek mainland. He also owned the material, and published only a fraction of it in his lifetime, which kept the working corpus thin for decades.
The consensus followed him, and dissent carried a cost. The Cambridge archaeologist Alan Wace, who argued that mainland Mycenaean Greeks had taken over Knossos, was reportedly kept out of excavation in Greece for years.
Ventris agreed with the consensus, in print. Through his teens and twenties he believed Linear B recorded something related to Etruscan, and wrote that the idea of Greek-speaking Minoans ignored historical plausibility altogether. Hold on to that. The man who eventually proved it was Greek had published the argument against it.
The woman with the cigarette boxes
Alice Kober was born in New York on 23 December 1906 to Hungarian immigrant parents, and taught classics at Brooklyn College from 1930 — assistant professor by 1936. Heavy teaching load, no research money, no institutional interest in what she did at night.
Her preparation was faintly absurd. If you are trying to read a script whose language is unknown, you cannot afford to be surprised by any grammar that exists. So she went and got them. Fluent in Greek, Latin, French, German and Old English, she then worked through Sanskrit, Hittite, Old Persian, Tocharian, Old Irish, Akkadian, Sumerian, Chinese, Basque, Carian, Hattic, Hurrian, Lycian and Lydian.
She wasn't reading them for pleasure or content. She was compiling an inventory of the shapes human languages come in — the ways they can inflect, agree, and change at the edges of words — so that whatever Linear B turned out to be, she'd recognise its behaviour.
Then there were the cards. Wartime paper rationing meant she couldn't simply buy index cards, so she made them: around 180,000 of them, cut by hand from the backs of greeting cards, church circulars and library checkout slips. Her filing cabinets were empty cigarette cartons. She punched holes in the cards so that any set sharing a feature could be lined up and read against each other in one pass — an analogue database, built on a lecturer's salary in Brooklyn, by one person, by hand. The reconstruction of that working method is largely owed to Margalit Fox's 2013 book The Riddle of the Labyrinth.
The grid
Here is what she found, and it is the actual engine of the decipherment.
Scanning her cards, Kober noticed sets of words that began with the same signs and ended differently, in consistent patterns — the same three-way variation recurring across different stems. She called them triplets. Their existence proved something enormous without revealing a single sound: the language was inflected. Word endings changed to mark grammatical role, the way Latin amicus / amici / amico does.
The second-order insight is the one that cracks it open. Linear B is syllabic — each sign is roughly one consonant plus one vowel. So at the join between a stem and its ending, a single sign is doing double duty: its consonant belongs to the stem, its vowel to the ending. Kober called these bridging signs.
Now follow the logic. If the same word appears with ending A and ending B, the bridging signs in each version must share a consonant — the stem's — while differing in vowel. Do that across enough triplets and you can start asserting relationships between signs whose sounds you still do not know: this sign and that one share a consonant; these two share a vowel.
Kober laid the results out as a chart — signs sharing a consonant along a row, signs sharing a vowel down a column — with every cell's actual sound left blank. She refused to guess.
That refusal is the whole discipline. Every failed attempt before her, and there were many, started by deciding what the language was and then bending the signs to fit; such schemes are flexible enough to produce words in almost any language you like, which is precisely why they prove nothing. Kober's grid contains no guess at all. It is pure structure, and structure can be checked.
She died on 16 May 1950, aged forty-three. No cause is recorded in her correspondence, her death certificate or her obituaries; the family believed stomach cancer. She had been corresponding since 1948 with Emmett Bennett Jr., who was assembling the definitive sign list, and left him her archive; her papers now sit at the University of Texas at Austin.
She died two years before the answer.
A frivolous digression
Ventris, by now working as an architect in London, took the grid and enlarged it.
His method of working was unusual and probably decisive: he did it in the open. In 1950 he posted a questionnaire to the scholars around the world working on the script, asking what each of them actually believed, then wrote up the replies alongside his own view and circulated the lot as a "Mid-Century Report." He intended it to be his farewell to the problem. It wasn't. Over 1951 and 1952 he sent out twenty numbered "Work Notes" — mimeographed, posted to a scattered handful of specialists — showing his reasoning while it was still wrong.
The grid filled out. He began testing sound values, using probable place names as probes: Amnisos, Knossos, Tulissos, all still on the map of Crete. Slot those in, propagate the constraints across rows and columns, and other signs acquire values whether you like them or not.
What came out the other side was Greek.
Work Note 20, dated 1 June 1952, is titled "Are the Knossos and Pylos tablets written in Greek?" It reads like a man arguing with himself. He hedged the whole thing as "a frivolous digression" — his words — and predicted it would hit an impasse or dissolve into absurdities.
He had spent his life arguing the opposite case. He published anyway, with the disclaimer attached, and let other people check it. On 1 July 1952 he described the result on BBC radio. Among the listeners was John Chadwick, a Cambridge classicist and wartime codebreaker, who wrote to him; the two became collaborators.
The tripod tablet
A decipherment can still be an elaborate self-deception. What the theory needed was a genuine prediction — a text nobody had seen, which the system could be right or wrong about.
It arrived in March 1953. Carl Blegen, excavating the Palace of Nestor at Pylos, had recovered a tablet in June 1952 that was only now cleaned: PY Ta 641. Blegen applied Ventris's values to it, cold.
Next to a drawing of a three-legged pot, the signs read ti-ri-po-de — "two tripods." Next to a vessel with four handles, qe-to-ro-we: "four-eared." One with three handles, ti-ri-o-we. One with none, a-no-we. The words matched the pictures, and the pictures matched the numerals beside them.
Nobody had fed that tablet into the system; it had been in the ground while the system was built. Blegen wrote to Ventris. The argument was effectively over.
What was actually inside
And here is the part that stings.
For half a century people had wondered what the Bronze Age Aegean would say when it finally spoke — the world behind the Iliad, five hundred years older than Homer, suddenly audible in its own words.
What it says is: sheep.
The tablets are administrative records, essentially without exception. Wool, grain, bronze, textiles, perfumed oil. The Knossos archives track something on the order of 100,000 sheep across Crete. Scribes logged chariot wheels by condition and material — serviceable or damaged, elm or willow. Ration lists. Land holdings. Personnel rosters.
No poetry. No law code. No chronicle, no letters, no story of any kind.
The reason is the best fact in the whole business. These documents were never meant to survive. They were unbaked clay — cheap, soft, deliberately temporary. A tablet carried one year's business, and at year's end the clay was to be broken up, wetted and remade into next year's blanks; scribes even trimmed tablets down to save material and shelf space.
They exist because the palaces burned. The fires that destroyed the archives baked them, converting a year's disposable paperwork into the only permanent record of a civilisation. Around 5,500 fragments at Knossos, roughly 1,026 tablets at Pylos, 368 at Thebes, 88 at Mycenae. Whatever the Mycenaeans wrote on anything they intended to keep — parchment, wood, whatever it was — is gone entirely. What survives is exactly and only the material nobody valued.
Then there are the names. Among the shepherds, smiths and chariot-holders on those lists sits a-ki-re-u — Achilles — attested at both Pylos and Knossos. Names matching Aias, Hector and Pandaros turn up too. Not heroes; not those people. Just ordinary Bronze Age men carrying names that a poet, centuries later, would hand to kings.
We can read the payroll of the world that produced the Iliad. We cannot read one line of its poetry.
Who cracked it
There is a live argument about credit, and it's the most interesting thing left in the story.
Chadwick's The Decipherment of Linear B (1958) is the account most people have absorbed, and it gives Kober real but bounded credit: useful observations, only partly correct, comprehensively superseded by Ventris.
Others push hard against that. Without the grid there is no decipherment — the instrument Ventris used was hers, enlarged. And there is one documented detail with a sting in it: in 1948 Ventris wrote Kober a six-page letter arguing that her alternating endings did not indicate inflection. He was wrong and she was right, and she died before she could say so with the answer in hand.
The counter-case is honest too. Kober herself thought a solution was still distant, declined to commit to a language, and made no attempt at the leap. Ventris made a jump she had explicitly refused to make, and jumps are not footnotes.
The fairest reading is probably that she built the instrument and he used it — and that we are still arguing partly because she died at forty-three and he at thirty-four, so neither of them got the decades a reputation normally needs to settle.
One coda. Linear A — the older script, the same island, the language of the Minoans themselves — remains unread seventy years on. No bilingual, and far too few texts. Kober's method hasn't cracked it, because the method runs on volume, and the volume isn't there.