Linear A
UnsolvedSummary
Linear A is the script of Minoan Crete. Its successor, Linear B, was deciphered by Michael Ventris in 1952 and turned out to record an early form of Greek. Because the two scripts share most of their signs, Linear A can be sounded out. Scholars can pronounce it.
They simply have no idea what they are saying.
What is documented
The script. Linear A appears on clay tablets, on vessels, and on cult objects, at sites across Crete and the wider Aegean. It was in use for roughly three and a half centuries, and it goes out of use around 1450 BCE, at the point when Mycenaean Greeks appear to take control of the island.
Linear B, and why it matters. Linear B replaces Linear A and is plainly derived from it: it uses most of the same signs. In 1952 Michael Ventris, working with John Chadwick, established that Linear B recorded Greek. That was one of the great decipherments, and it did something unusual: it gave scholars a set of sound values.
The consequence. Apply the Linear B sound values to the shared Linear A signs and you get syllables. Strings of them. They can be read aloud with reasonable confidence. What comes out is not Greek, and it is not any language anybody has been able to identify.
What can be understood, and it is not much. The corpus is largely administrative: inventories, lists, accounts. From position and context, a handful of items are broadly agreed. The sequence read as ku-ro appears at the ends of lists, in the place where a total would go, and is generally taken to mean "total". A sequence read as ki-ro appears in a way consistent with a deficit or an amount owed.
This is the level of comprehension: we can identify the word for "total" because of where it sits on the page, in the same way that you could work out the word for "total" on a foreign receipt without knowing a word of the language.
The corpus is too small. Around 1,400 inscriptions are known, and they are overwhelmingly short. The entire surviving body of Linear A text amounts to a few thousand signs. Decipherment normally requires either a bilingual inscription, of which there is none, or a large corpus in a language that can be guessed at. Linear A has neither.
The language. Unidentified. It is not Greek. It has not been convincingly connected to any known family. Proposals have included Anatolian languages, Semitic languages, and a relationship to Etruscan, and none has been accepted.
Leading explanations
The obstacle is not the script. It is the language.
Why Ventris succeeded and nobody has succeeded here. Ventris had a testable hypothesis. He guessed that Linear B might be Greek, applied it, and Greek words fell out. The hypothesis could be checked, because Greek is known: if the guess had been wrong, the output would have been noise, and it was not.
Linear A offers nothing to guess. There is no candidate language whose vocabulary can be tested against the output. You can produce syllable strings all day, and there is no dictionary anywhere on Earth to check them against. Every proposed language yields, in the hands of its proposer, some words that look promising and a great many that do not, and there is no independent way to referee.
What would break it. A bilingual inscription of any length. That is essentially the whole answer. Failing that, a substantial increase in the corpus, which would require new excavation and a good deal of luck. Failing both, the honest expectation is that Linear A stays unread.
What the popular version gets wrong
"Nobody can read Linear A." Everybody can read Linear A, in the sense of producing the sounds. What nobody can do is understand it. The distinction is not pedantic; it is the entire structure of the problem. This is not a code that resists being cracked. It is a language that nobody has ever heard.
"It's a cipher." It is not enciphered. It is an ordinary syllabary, written openly, by clerks, to keep accounts. There is no secret and nothing was hidden. The Minoans were not concealing anything from us; they simply died, and took their language with them.
"Someone will crack it with a computer." Machine methods have been applied and they run into the same wall. A statistical model can find structure. It cannot tell you what the structure means without something to anchor it to, and there is nothing to anchor it to.
"Linear A has been deciphered." Claims appear periodically. None has been accepted, and the test is simple and brutal: a real decipherment lets other people read new inscriptions. None of them does.
Current status
Unsolved. The script is understood. The corpus is catalogued. The sound values are largely known. The language is gone.
The registry keeps this record because of the particular kind of loss it represents, which is different from the others in this category. The Voynich may be meaningless. Rongorongo may not be writing. Linear A is unquestionably a real language, written down by real administrators to keep track of real jars of oil, and we are standing in front of it, sounding out the syllables, and hearing nothing.
Sources
- Ventris, M. and Chadwick, J. (1953 onward). Decipherment of Linear B, and Documents in Mycenaean Greek (1956).
- Godart, L. and Olivier, J.-P. Recueil des inscriptions en Linéaire A (the standard corpus).
- Standard scholarship on the readings of ku-ro and ki-ro from their positions in administrative lists.
- Evans, A. Original identification and naming of Linear A and Linear B at Knossos.
- Published assessments of proposed decipherments and of computational approaches to the corpus.
Last reviewed: July 2026. Records are provisional. Where the evidence changes, the entry changes. Found an error? Tell us.