The Riddle of the Labyrinth

By Margalit Fox

12 min read

Why does it matter? Because the lone-genius story of every great discovery is almost always a lie of omission.

We tend to celebrate the person who crosses the finish line. The lone genius, the flash of inspiration, the name on the cover of Science. What we rarely ask is who surveyed the terrain, who cleared the brush, who built the road so precise and solid that someone else could simply run it. Alice Kober spent the last decade of her short life doing exactly that — cutting 180,000 index cards by hand, filing them in cigarette cartons, building a logical framework so rigorous it could only lead one direction — and then she died at forty-three, two years before anyone crossed. The Riddle of the Labyrinth is the story of a fifty-year war to decode Europe's oldest written records, but it's also something more damning: a case study in how intellectual history gets told, and who it quietly, systematically forgets.

A 43-Year-Old Brooklyn Professor Died Two Years Before the Breakthrough She Made Possible

In 1950, a forty-three-year-old classics professor named Alice Kober died of cancer in her Brooklyn apartment, surrounded by thousands of handwritten index cards she had spent the previous decade filling with meticulous observations about an undeciphered Bronze Age script. Two years later, a young English architect named Michael Ventris cracked the code she had spent her adult life trying to break — and received all the credit.

Margalit Fox's book exists to correct that ledger. The script in question, Linear B, first emerged from the soil of Crete in 1900, when the archaeologist Arthur Evans broke ground at Knossos and within a week unearthed clay tablets that turned out to be Europe's oldest written records. Dating to around 1450 B.C., seven centuries before the Greek alphabet, the tablets were covered in strange pictograms no one could read. Evans dug up more than a thousand of them in his first season alone. For fifty years, the world's sharpest minds couldn't even identify what language they encoded.

Kober came closer than anyone. Working nights at her dining table without computers — she dismissed such machines with contempt — she built the analytical scaffolding that made the eventual solution possible.

Fox frames her mission with a particular edge: when Ventris died in 1956, the Times missed the story entirely — one tired night editor, and a death goes unrecorded. She's here, sixty years late, to file the obituary she should have. But the deeper debt is to Kober, whose name you almost certainly don't know, and should.

Decoding an Unknown Script in an Unknown Language Is Not a Puzzle — It's a Locked Room with No Door

Imagine you're handed a document in a script you've never seen, encoding a language that no longer exists, with no dictionary, no grammar guide, and no bilingual parallel text to cross-reference. That's not a puzzle. That's a locked room with no door — and it describes exactly the situation facing anyone who tried to read Linear B in 1900.

Any written text falls into one of four relationships between language and script. At the easy end: a known language in a known script — English in Roman letters, the page you're reading now. Move one step harder and you get a known language in an unknown script, or a known script encoding an unknown language. Etruscan can be read aloud with perfect confidence because its alphabet is Greek-derived — and yet it means absolutely nothing, because no one understands the language. Linear B in 1900 occupied the worst cell: unknown script, unknown language. No foothold anywhere. 'The toughest of all,' Fox writes.

What makes this genuinely hard isn't the absence of information. It's the presence of the wrong kind of confidence. The decipherment of Egyptian hieroglyphics offers the cautionary model. Thomas Young, a physician who had mastered a dozen languages before adulthood, correctly identified that certain oval cartouches contained phonetic spellings of royal names. He was right. And then he stopped, because the glyphs looked like pictures of real things — owls, reeds, hands — and he couldn't shake the conviction that pictures must mean things, not sounds. Champollion succeeded precisely because he overcame that assumption: the hieroglyphs were a mixed system, logograms and phonetic signs together, and the language they encoded was Coptic. Iconicity — the belief that a picture must represent what it depicts — had blinded a brilliant man to the actual structure in front of him.

Arthur Evans never learned this lesson. He spent decades staring at a Linear B sign that recurred constantly across thousands of tablets, convinced it depicted a throne with a scepter — clear evidence, he concluded, of royal ownership. It was the vowel sound 'o.' A man brilliant enough to find a lost civilization inside a Cretan hillside spent half a century insisting a vowel was a throne. That's the cognitive snare: the hunger to see meaning before structure is proven.

Arthur Evans Found the Tablets and Then Spent Forty Years Making Sure No One Else Could Read Them

That cognitive trap had a face, and it belonged to the man who found the tablets in the first place.

Arthur Evans was simultaneously the man who gave the world Linear B and the man who made sure the world couldn't have it. Both things are equally true, and the second one is the one history tends to skip.

The tablets survived because of catastrophe. When the Palace of Knossos burned around 1400 B.C., the destruction baked the scribes' sun-dried clay records into something closer to ceramic — permanent, legible, durable across three millennia. Evans grasped this irony early, when rain from a leaky roof turned a batch of his own recovered tablets back into mud overnight. Fire had done what careful archival practice never could.

Then Evans became the obstacle the fire never was. Of the two thousand or more tablets he eventually unearthed at Knossos, he published reproductions of fewer than two hundred during his lifetime. When a Finnish scholar named Johannes Sundwall managed to see thirty-eight tablets in a Cretan museum and published drawings of them, Evans responded with fury. The tablets were his. The mystery was his. The eventual solution, presumably, would also be his — except that it never came.

His failure wasn't only about possessiveness. Evans had constructed an intellectual prison around himself, too. Convinced that Minoan civilization was a superior culture entirely separate from the 'rude' mainland Greeks, he declared with the authority of a field's dominant figure that the language of Linear B simply could not be Greek. Every serious scholar working on Linear B accepted his verdict. For fifty years, that dogma foreclosed what turned out to be the answer.

He died in 1941 without translating a single sentence. The tablets waited.

The World's Leading Expert on Linear B Worked at a Dining Table with Cigarette Boxes and a Dime-Store Hole Punch

Every night after her classes were done, Alice Kober sat down at her dining table in Flatbush, Brooklyn, lit a cigarette, and picked up her hole punch. Around her, stacked in empty cigarette cartons, were tens of thousands of handmade index cards — two by three inches, cut by hand from church circulars, the backs of greeting cards, and the cardboard covers of examination booklets, because wartime paper rationing left her no choice. On each one, in handwriting so small it required a magnifying glass, she had recorded statistical observations about the characters of Linear B. By the time she was done, she had cut, annotated, and sorted more than 180,000 of them. Open one of those cartons today and you catch a faint whiff of midcentury tobacco. 'FLEETWOOD IMPERIALS: A CLEANER, FINER SMOKE,' her file cabinets say.

The cards were the instrument of a method so radical it looked like restraint. While every other investigator was racing to guess what language the tablets recorded — Etruscan, Basque, proto-Celtic, take your pick — Kober refused to guess anything at all. Her governing principle was to treat the symbols as objects of pure form, nothing more. No sound-values assigned. No language hypothesized. The script would speak for itself, or it would stay silent. She called it inhabiting a world of 'form without meaning,' and she was willing to live there for as long as it took.

What this produced was not paralysis but precision. Using her dime-store hole punch, Kober assigned each Linear B character a fixed location on her cards, so that stacking cards together instantly revealed which characters those words shared. She tracked every sign's frequency not just in general but by position — initial, second, middle, next-to-last, final — and mapped how signs combined with one another. She once calculated that comparing each of Linear B's seventy-eight signs against every other would take roughly fifteen hundred hours. She bought a slide rule to speed the arithmetic.

Arthur Evans saw a picture and named it a throne. Kober saw a pattern and refused to name it anything until the pattern demanded a name. Michael Ventris, who crossed the finish line two years after Kober died, said publicly that he had arrived at his solution using the methods she devised.

Kober's Breakthrough Had No Name Because She Refused to Publish What She Couldn't Prove

What does it actually mean to prove something in a field built on guesses? For most scholars working on Linear B, proof was loose — a plausible argument, an intriguing pattern, a hunch dressed in footnotes. Kober held herself to a different standard, and in 1948 she published the result: a paper containing a grid of just ten symbols, captioned with almost comic modesty, 'Beginning of a Tentative Phonetic Pattern.'

The grid was the payoff from what Ventris would later call 'Kober's triplets.' Think of Latin, where the word for 'farmer' changes its ending depending on whether it's the subject, the object, or the possessive — agricola, agricolam, agricolae. A syllabic script has to represent each of those forms one character at a time, which means the character bridging stem and suffix has to absorb both the stem's final consonant and the suffix's opening vowel at once. Change the suffix, change the vowel, change the bridging character. By patiently sorting through Linear B noun endings across three grammatical cases, Kober proved that two different-looking signs could share an underlying consonant while differing only in their vowels — and she proved it without knowing what any of those consonants or vowels actually were. Her grid mapped which signs shared a consonant and which shared a vowel, as pure coordinates, like a street grid before anyone names the streets.

Fox compares it to Sudoku: once you correctly place a single number, the constraints ripple outward and the rest of the puzzle becomes logically compelled rather than guessed at. The grid was that moment before the first number goes in — a structure fully ready for one confirmed value to set off a chain reaction through the rest.

The cruelest detail is this: Kober's private files, recovered after her death, contained a phonetic grid with more than twenty correctly placed characters — more than double what she had published. She simply hadn't satisfied herself that the evidence was airtight enough to print. The standard of proof that made her published work immune to attack also kept her unpublished work off the record. Ventris, working from her published grid two years after she died, filled in the cascade she had already mapped in private.

She wasn't slow. She was rigorous in a field that rewarded confident wrongness. And that rigor, the very thing that makes her method worth admiring sixty years later, is also what left the headline to someone else.

The Man Who Cracked the Code Spent a Decade Insisting the Answer Was Impossible

In the autumn of 1936, a fourteen-year-old schoolboy found himself face-to-face with Arthur Evans at a Royal Academy exhibition in London. Evans, eighty-five and the most celebrated archaeologist alive, was giving an impromptu tour when the youngest member of the group asked a quiet, eager question: had Evans really said the tablets couldn't be deciphered? Evans had. The boy's name was Michael Ventris, and he would die at thirty-four having solved what Evans spent a lifetime failing to crack.

The obsession that took hold in that gallery would consume Ventris for the rest of his life — partly because he was extraordinarily equipped for it. He had a permanently open critical period for language acquisition: while most people lose the ability to absorb new languages effortlessly around the onset of adolescence, Ventris never did. He mastered Swedish as an adult in a matter of weeks. He taught himself Russian from scratch to write letters to a sculptor friend. He navigated wartime bombing raids over Germany, then cleared space on his navigator's table to work Linear B transcriptions while flak shook the aircraft.

But the gifts came wrapped in damage. Jung, who treated Ventris's parents in Switzerland, had instructed them never to touch their son — physical contact, he advised, would produce harmful psychological attachments. A family friend was warned before meeting the boy: 'I must on no account touch Michael.' By the time Ventris was eighteen, both parents were dead — his father from tuberculosis, his mother from a deliberate overdose of barbiturates. Linear B was buffer and ballast both.

For years, it was also a trap. Ventris had decided at eighteen that the language of the tablets was related to Etruscan, and he held that conviction with a tenacity that resisted his own evidence. When he began building phonetic grids in 1951, modeling them on the framework Kober had published, his first attempt assigned wrong values to nearly seventy percent of the signs — most of the errors traceable to forcing the data toward Etruscan. When the grid began suggesting Cretan place-names that looked unmistakably Greek, he dismissed the results as a mirage. In the spring of 1952, he described his own mounting findings as a potential 'hallucination' and speculated that he might need to abandon the whole line of inquiry. He was two weeks away from something he couldn't yet let himself believe.

What finally cracked it was Kober's grid. By working through her published phonetic coordinates — ten characters, no sound-values assigned, just relational structure — and adding the place-name insight she had never lived to publish, Ventris triggered the cascade she had already mapped. The lock she had spent a decade building, he opened. He knew it, too: he said so publicly. The woman who had done the work was two years in the ground.

A Photograph Swimming into Being: The Moment Greek Emerged from the Grid

One night in early June 1952, the journalist Prue Smith and her husband arrived at the Ventris flat for a dinner party and found their host missing. His wife kept apologizing, refilling the sherry, letting the food sit. Then Michael Ventris burst into the room, disheveled and electric with something he couldn't contain. 'I know it, I know it,' he said. 'I am certain of it.' Smith assumed he'd confirmed his long-held Etruscan theory. He had done the opposite: he had just proved that the language of the tablets was an archaic form of Greek, seven centuries older than Homer.

The weeks before that moment had felt, as Ventris described it, like watching a photograph swim up from a developer's tray. He'd used the Cypriot syllabary — an Iron Age script that scholars had long considered too distant from Linear B to be useful — to anchor a handful of sound-values in his phonetic grid. Once anchored, the grid's web of interdependencies did the rest. Reading one word as 'a-mi-ni-so' gave him two new characters. Those two gave him entire rows and columns. 'Ko-no-so' emerged — Knossos itself, spelled in syllables, confirming that Alice Kober's triplets marked the names of Cretan towns rather than grammatical cases. Then vocabulary started surfacing: 'to-so' and 'to-sa,' the Linear B spellings of Greek words meaning 'so much' in masculine and feminine. 'Ko-wo' and 'ko-wa' for boy and girl. Words that inflected for gender — the signature of Indo-European languages, entirely absent from Etruscan. The language he had spent a decade insisting couldn't be Greek kept insisting it was.

The structural twist is almost too neat. Kober had been exactly right about the structure of those triplets and exactly wrong about what they encoded. She had identified the ending patterns as grammatical inflection — the same noun in different cases. They were actually derivational: the name of a town, the name for its harbor, the term for a man from that town. She built the lock without knowing which door it opened. The wonder is in those syllables surfacing — 'ko-wo,' 'ko-wa,' boy and girl, hiding in plain sight for three thousand years — and the injustice is that she never saw them come up from the tray.

The Fire That Killed the Civilization Preserved Its Paperwork

What Ventris had proven could be read turned out to be, almost entirely, paperwork. The Mycenaean palace scribes worked on exactly that logic — clay tablets were temporary, pulped at year's end and remolded into next year's records. What preserved them was catastrophe. Around 1200 B.C., the fires that burned the palaces to their foundations accidentally baked three thousand years of immortality into what had been glorified scratch paper.

What that scratch paper contains is wonderfully, stubbornly ordinary. A king whose land plot was three times any subordinate's. A palace at Knossos tracking nearly a hundred thousand sheep. Gendered labor sorted into neat columns: women at the looms, men at the forge. Names ranging from the heroic — 'Who Kills in Battle' — to the contemptuous: 'Devourer of Excrements,' bestowed, the tablets suggest, on slaves and foreigners. The gods appear mid-transformation: Zeus and Poseidon already present, alongside pre-Greek deities who would vanish before the Classical Age — a female counterpart to Poseidon, a 'Mistress of the Labyrinth.' The tablets caught a civilization in the middle of becoming something else, then froze it there.

The people who unlocked those records didn't fare much better than the civilization they studied. Kober died at forty-three, her private grid — more than twenty correctly placed characters, never published — sitting in cigarette cartons in Brooklyn. Ventris died at thirty-four, four years after his announcement, in a crash that the people who knew him best declined to call an accident. He had already written a self-lacerating letter to the Architects' Journal renouncing his fellowship, declaring he'd lost faith in his own intelligence. The riddle outlasted them both. What remains is the mundane miracle: bureaucratic paperwork, accidentally permanent, deciphered because a woman in Brooklyn refused to guess until she was sure.

What the Cigarette Boxes Were Actually Storing

Picture two people who never knew each other existed. A Mycenaean scribe at Knossos, pressing a stylus into wet clay at the end of a long administrative day — sheep counts, ration allocations, the names of slave women and their children — fully expecting the tablet to be pulped by spring. And Alice Kober at her dining table in Flatbush, long after her students had gone home, cutting index cards from old church bulletins by lamplight and filing them into cigarette boxes. Neither was trying to last. The scribe's work survived because a palace burned down. Kober's survived because Margalit Fox decided it should. That asymmetry is worth sitting with. The fire was an accident. The book was a choice. And the tablets, when you finally read what's on them — slave women's rations, a king's acreage, a child called 'Devourer of Excrements' — turn out to be exactly as ordinary and irreplaceable as the woman who refused to guess her way through them.