How Linear B was deciphered

1. The Bronze Age Aegean

About 3000 to 1100 BC

Between Greece and Turkey lies the Aegean Sea, scattered with islands. At its southern edge sits Crete, long and mountainous, about 250 kilometers from end to end. In the Bronze Age, when tools and weapons were made of bronze, this sea joined some of the earliest towns and palaces in Europe.

On Crete lived the people we call the Minoans. That isn’t their own name, which we don’t know. Arthur Evans named them after Minos, the king of Cretan legend. From about 1900 BC they built large “palaces” at Knossos, Phaistos, Malia and Zakros: great buildings around a central court that were homes for rulers, but also storehouses, workshops, shrines and record offices.

On the mainland lived the Mycenaeans, named after Mycenae, their most famous citadel. They spoke Greek. From about 1400 BC they ran palace states at Mycenae, Pylos, Thebes, Tiryns and elsewhere, and by about 1450 BC Greek speakers were in charge at Knossos on Crete as well. Around 1200 BC the mainland palaces were destroyed, and writing vanished from Greece for about four hundred years.

Aegean dates come from styles of pottery tied to years by radiocarbon and by links with Egypt. They can be off by decades, and some, like the eruption of Thera and the fall of Knossos, are still argued over.

2. Three scripts on Crete

About 2100 to 1200 BC

Crete had writing early. The first script, which scholars call Cretan hieroglyphs, was used mostly on seal stones. Then came Linear A, written on clay tablets and stone offering tables from about 1800 BC. When Greek speakers took over Knossos, scribes adapted Linear A’s signs to write Greek: that is Linear B. The name “linear” simply means the signs are drawn as lines, not as detailed pictures.

All three mix two kinds of sign. Most signs stand for a syllable, like ka or po. Others, called logograms, stand for a whole thing: a man, a horse, a tripod, a jar of oil. Neither is an alphabet. The Greek alphabet, with one letter for each sound, came centuries later, borrowed from the Phoenicians around 800 BC.

3. Arthur Evans digs at Knossos

1900

In March 1900 the English archaeologist Arthur Evans began digging at Knossos, near the modern city of Heraklion. Within weeks his workmen were finding clay tablets covered with signs. There were thousands of them. Evans sorted the scripts and gave them the names we still use: hieroglyphic, Linear A and Linear B.

Evans was sure that the people who wrote the tablets were Minoans with a language of their own, not Greeks. He was a great excavator, and his view carried enormous weight. It was also wrong about Linear B, and for half a century it steered people away from the answer. He published only a small part of the tablets in his lifetime, so for decades anyone who wanted to try had very little to work with.

In 1936 Evans spoke at an exhibition in London about his finds. In the audience was a fourteen-year-old schoolboy named Michael Ventris, who made up his mind that day to read the script.

4. An archive at Pylos

1939 and 1952

In 1939 the American archaeologist Carl Blegen began digging at a hill near Pylos, in the southwest of mainland Greece. On the first day, his trench cut into a room full of clay tablets: the archive of a Mycenaean palace. They were in Linear B, about six hundred of them.

This mattered twice over. Linear B was not only a Cretan script; the mainland palaces used it too, in the heart of what would later be Greek-speaking country. And there was now much more text to study. The war stopped the work, but photographs of the tablets were made, and in 1951 Emmett Bennett published the Pylos tablets with a careful list of the signs, each with a number. Scholars still name a sign by its number, like *08, and an unread sign is written that way to this day.

Why did the tablets survive at all? They were never meant to. They were notes in sun-dried clay, kept for a season and then reused. When the palaces burned, the fire baked them hard, and so the destruction of Pylos and Knossos preserved their paperwork.

5. Alice Kober’s patience

1930s to 1950

Alice Kober taught classics at Brooklyn College in New York. She spent years on Linear B alongside her teaching, and she had a rule: don’t guess. Many people had announced what the language was and then read whatever they liked into the signs. Kober wanted facts that would hold whatever the language turned out to be.

She copied every word onto index cards, about 180,000 of them, many cut from scrap paper during the war. She counted which signs came where. In papers from 1945 to 1948 she showed that Linear B words change their endings, as words do in Latin or Greek, and she found sets of related words that differed only at the end. In those sets, one sign seemed to “bridge” the stem and the ending, so it had to share a consonant with one sign and a vowel with another.

Here is the idea, with words we can read today (she could not):

All three begin ko-no-. The first ends -so, the others -si-jo and -si-ja. If the stem ends in a consonant, so and si must share it, while jo and ja share another. Kober could prove relations like that without knowing a single sound. She built a small grid of ten signs from them. She died in May 1950, at 43, before the answer came.

6. Michael Ventris’s grid

1950 to 1952

Michael Ventris had served as an RAF navigator in the war and trained as an architect, and he never forgot the script. He took Kober’s method much further. He arranged the signs in a grid: signs in the same row shared a consonant, signs in the same column shared a vowel. For months it held only relationships, “row 3” and “vowel 2,” not sounds. He shared his progress in notes he sent to other scholars.

Try it the way he did. You know no sounds, only which signs share a consonant or a vowel:

Then came the leap. Some words turned up only on the Knossos tablets, never at Pylos. Might they be Cretan place names? Ventris tried the names of towns near Knossos, using the grid to spread each guess along its row and column:

ko-no-so for Knossos, a-mi-ni-so for Amnisos, its harbor, and tu-ri-so for Tylissos. The values fit. And when he put the new sounds into other words, they began to look like Greek. He had expected something like Etruscan, and at first he didn’t trust it.

7. “It is Greek”

1 July 1952

On 1 July 1952 Ventris announced on BBC radio that Linear B was Greek: a very old and awkwardly spelled Greek, but Greek. Among the listeners was John Chadwick, a scholar of early Greek at Cambridge. He wrote to Ventris, and they began to work together. In 1953 they published their case in the Journal of Hellenic Studies: “Evidence for Greek Dialect in the Mycenaean Archives.”

Why had it been so hard to see? Because Linear B spells Greek badly. Every sign ends in a vowel, so the script can’t write a consonant at the end of a word or most clusters of consonants. Scribes dropped some sounds and added empty vowels to others. One sign served for both r and l, and one for p, b and ph:

ka-ko is khalkos, bronze: the l and the s are missing. ku-ru-so is khrusos, gold: the first u is an empty vowel. pa-te is pater, father, without its final r. A reader who knew the language filled in the gaps, the way you can read “txt msgs.”

8. The tripod tablet

1953

Not everyone believed it. The test came from Pylos. In June 1952, before the broadcast, Blegen’s team had found a small palm-leaf tablet that wasn’t cleaned and read until the next spring. It lists vessels, each with a little picture. In May 1953 Blegen wrote to Ventris that the words, read with Ventris’s values, described exactly what the pictures showed. “All this seems too good to be true,” he wrote. “Is coincidence excluded?”

Beside the tripod picture stands ti-ri-po-de, “two tripods.” Beside cups with four handles, qe-to-ro-we, “four-eared”; with three handles, ti-ri-o-we-e; with none, a-no-we. Nobody could have arranged that match in advance. Most specialists were convinced within a few years. Ventris and Chadwick’s book Documents in Mycenaean Greek came out in 1956; that September Ventris was killed in a road accident, at 34.

Tap any word, logogram or number on the tablet to see what it is. The button steps through the reading one layer at a time.

9. What the tablets tell

About 1400 to 1200 BC

Anyone who hoped for poems or history was disappointed. The tablets are accounts: flocks and wool, oil and perfume, rations for workers, bronze for smiths, chariots and armor, land, and offerings to the gods. But accounts tell a great deal. The state was headed by the wa-na-ka, the wanax or king; next came the ra-wa-ke-ta, the lawagetas. A qa-si-re-u was a minor local official; his title later became basileus, the Greek word for king.

The gods are there by name: po-se-da-o (Poseidon), di-we (to Zeus), e-ra (Hera), e-ma-a2 (Hermes), even di-wo-nu-so (Dionysus), once thought to be a later arrival. One Pylos tablet sends gold vessels to Zeus, Hera and Drimios, son of Zeus, each with a man or a woman. Some scholars read those people as sacrifices; others think they were given to serve the gods. The tablet does not settle it.

What the tablets don’t tell is just as important. They name no Agamemnon, no Trojan War, no history at all. They are one season’s paperwork, saved by fire.

10. Linear A: still unread

About 1800 to 1450 BC, and today

If Linear B came from Linear A, why can’t we read Linear A too? We can, in a way, say its sounds. Many Linear A signs look like Linear B signs, so scholars read them with the Linear B values. That is a working convention, a likely guess, not something proved. And it gives words that match no language anyone knows.

A few things do come through. Some place names read the same in both scripts, which suggests many sounds carried over:

And arithmetic can show what a word does without telling what language it belongs to. On many lists, the word read ku-ro stands before a number that equals the sum of the lines above it, so it is widely accepted to mean “total.”

Stone offering tables from sanctuaries repeat a set of words, the “libation formula,” with forms like a-sa-sa-ra-me and ja-sa-sa-ra-me. We can see that they recur, and in what order. What they mean, no one knows.

11. How to judge a decipherment

Claims to have read Linear A appear every few years, often with a lot of attention. None has been accepted by specialists. That isn’t stubbornness. A real decipherment passes tests that Ventris’s did: it gives each sign one value and reads the whole corpus, not chosen words; it makes predictions that new texts can check; it turns known place names into the places we know; and specialists can test it and find that it holds. Try the tests on some made-up claims:

12. Where to go next

Learn the signs a row at a time, read the tablets for yourself, and try your own name in the scribes’ spelling.