What is known, what is not

Short version. We can count. We cannot read.

Linear A is not a literature. The tablets are lists: a word, a number, a word, a number, then a total. The rest is mostly smaller stuff: 929 of 1,883 documents in the corpus are nodules and sealings, lumps of clay with a seal and a sign or two. The numbers can be read. The words cannot, and this page says how far each side got.

What holds up

The numerals

A vertical stroke is one. A horizontal stroke or a dot is ten. A circle is a hundred. A circle with strokes is a thousand. Write them biggest first, add them up, done (Corazza and colleagues, 2021). 352 of 1,707 documents with a sign-by-sign text carry at least one number.

Take the last line of tablet HT 13. It reads: a circle, three bars, then a small fraction sign. One hundred, thirty, and a half. The 6 entries above it are 5, 56, 27, 18, 19, 5. Add them and you get 130. The scribe did too.

One quantity per line, almost without exception: 1,288 of 1,292 lines with a numeral in the sign-by-sign text carry exactly one.

Want the lot? The numbers has every ku-ro total, every fraction sign and the sums by site and by commodity, in tables you can sort and download.

The totals

The word read ku-ro is on 35 of 695 documents with a word list. A number comes after it. Where the entries above it survive, that number can be their sum.

We checked the sums, in our own audit of 13 August 2026. The audit ran on the sign-by-sign text, not on the word list. Its unit is one ku-ro total, not one document (a tablet can carry two). For each total it added the entries in the span, back to the heading or to the total before, and compared. 10 of 35 ku-ro totals tested in the sign-by-sign text balance exactly. The other 25 each fail for a reason you can point at:

• 15 have text missing from the span.
• 4 have more entries after the total.
• 3 have no entries left to add.
• 2 balance in whole numbers and miss by the fractions.
• 1 is off by one or two, with no fraction in sight.

Each document page shows its own check, and a tablet with two totals shows both. The numbers lines them all up in one table.

In Linear B, which can be read, the same test on the word for "total" passes on only 4 of 49 cases (our test, 14 August 2026). Clay breaks, so a real ledger fails this test a lot. Linear A fails it less often than Linear B does, not more.

The libation formula

What the words mean is not known. We tested whether any slot in the phrase names the place, and our test found none that does (smallest p 0.12, n = 16, 13 August 2026). That is a failed test, not proof that no slot does: for 4 of the 9 slots, no result could have been significant, because each word in them occurs once.

The fractions

The small marks after a number are fractions. GORILA lists 17 fraction signs. Corazza and colleagues (2021) count 480 attestations of them and solved 12 of the 17 from arithmetic alone. The commonest, J, is one half (137 attestations). Then E is a quarter (105), D a sixth, B a fifth, K a tenth. Combinations add: J plus E is three quarters.

Our own check reaches one of those values without their paper. On HT 104, two J make a whole, so J is one half. The rest we take from them, with the citation.

The Linear B connection

Linear B is the script of Mycenaean Greek. Ventris read it in 1952. Linear B took most of its signs from Linear A. Meissner and Steele (2024) put the Linear A signs with a Linear B counterpart at 70 percent or more. The catalogue marks a shared sign with the prefix AB, and 94 of 378 sign types on this site carry it. Scholars read the Linear A sign with the Linear B sound. That is how ku-ro gets its name.

It is an assumption, not a proof. A sign-specific argument for the transfer exists for about 20 of the 58 syllabic signs (our count, 14 August 2026). Convention assumes all 58. Steele and Meissner put it plainly: "we have very little check on the results".

The scripts are related. The languages are not shown to be. The 138 word types shared between the two word lists are what sign frequency alone predicts (p 0.955 against a frequency-matched null, 1,000 draws; our test, 14 August 2026).

What does not

The language

Nobody knows what language the scribes spoke (Davis, 2026). "Family unknown" is the accurate phrase. "Isolate" claims more than 7,000 signs can show.

How much text is enough

The published corpus is 7,574 signs (GORILA plus the 2024 supplement). In ours, 2,261 of 8,949 sign slots in the sign-by-sign text are damage, where the clay is broken or the sign cannot be made out. Linear B, for scale, is 47,271 signs in the corpus we hold.

We measured how much text a decipherment needs (our test, 14 August 2026). Take Linear B, scramble its 90 signs at random, and attack it with a model of the language. At 250 signs the attack recovers 11 percent of the sign values. At 1,000 signs, 60 percent. At 3,000 signs, 89 percent. The curve flattens there. Call 3,000 the unicity distance, the point where the text has enough repetition to pin the key.

The test put Linear A at 3,895 to 4,316 sound signs, counted on the word list as it stood that day. That clears 3,000 by a third, not by a mile, and a quarter of the slots are damaged. At exactly that size, the attack with the right language model recovers 91 percent of sign values. With frequencies alone, 9 percent. With the wrong language, 0.3 percent. So size is not what stops the reading. The missing input is a sample of the language, and no excavation has supplied one.

The main proposals, and where each stopped

Each proposal below is not demonstrated. None is disproved either, and the difference matters.

Greek. Duccio Chiapello (2010s onward) reads an early Greek dialect with Doric and Laconian features. Each reading is offered as a clue, and none has an independent check. The standard objection is structure: affixes turn up in about 59 percent of Linear A words against about 12 percent of Linear B words (Duhoux, 1978 and 1998). Not demonstrated.

Semitic. Cyrus Gordon (1957 onward) found about fifty lookalike words (about, because he kept revising the list), first Akkadian, then West Semitic. Best, Woudhuizen and Aartun followed. Vocabulary, never grammar, and the matches point at three branches at once. Davis (2026) allows that ku-ro may be a Semitic loanword. Not demonstrated.

Luwian. Leonard Palmer proposed an Anatolian language. Facchetti's survey files it with Gordon's work as formal coincidence. Not demonstrated.

Etruscan-related. Giulio Facchetti proposed it, with Negri, and ranks it himself as a suggestion that this much material can neither verify nor refute. Not demonstrated.

Hurrian. Peter van Soesbergen (1987, revised 2016 to 2017) reads ku-ro as Hurrian "again". Facchetti shows the same morpheme matches fit an Indo-European reading equally well. Not demonstrated.

Family unknown. Davis (2026), Duhoux, Salgarella and Younger transliterate and do not translate. That is a position, not a finding.

The proposals share one problem. With 60 to 90 open syllables, words of 2 to 4 signs, and spelling rules that let you add or drop consonants, lookalike hits are what chance produces. Our test (14 August 2026): meaningless strings shaped like Linear A words get an English gloss 93 percent of the time from a made-up dictionary. So this site offers no gloss of its own beyond ku-ro, the numerals and the commodity signs. Where a page quotes a scholar's gloss, it names the scholar.

Our own tests that failed

Our own ideas, with the number that stopped each one.

Scribal hands from 3D scans. We tried to tell scribes apart from stroke geometry on 3D-scanned tablets (17 faces, 928 strokes). Hand attribution scored AUC 0.452, chance is 0.5, and the kill line was set at 0.70 before we looked. Our test failed, so we dropped it. The same features tell the tablets apart (AUC 0.882), which answers a different question.

The listing search. In the SigLA pages of 14 August 2026, eight sign types appear in the traced drawings and on none of the 802 document pages. We thought that proved a bug in how SigLA draws its pages. Our test shows the gap follows the sign type. It cannot show the cause. Half a failure.

Sign *301 as a variant of na. A 2026 proposal that *301 is an allograph of the sign read na fails a permutation test, p below 0.0001, on the 22 word-internal *301 tokens in the word list. Its neighbours are not na's neighbours.

Sign *301 is this project's emblem because nobody has shown what it says. It appears on 272 of 1,707 documents with a sign-by-sign text, and on 6 of 34 documents with *301 in the word list it is stamped alone or nearly so on a small clay sealing. No value proposed for it is demonstrated. One of the proposals was ours, and we withdrew it.

Sources

• Corazza, M., Ferrara, S., Montecchi, B., Tamburini, F., and Valerio, M. (2021). The mathematical values of fraction signs in the Linear A script.
• Davis (2026). The Undeciphered Aegean Scripts.
• Del Freo, M., and Zurbach, J. (2024). Recueil des inscriptions en lineaire A. Supplement 1. Etudes Cretoises 21.6. Ecole francaise d'Athenes.
• Godart, L., and Olivier, J.-P. (1976 to 1985). Recueil des inscriptions en lineaire A (GORILA), vols. 1 to 5. Etudes Cretoises 21. Ecole francaise d'Athenes.
• Salgarella, E. (2020). Aegean Linear Script(s): Rethinking the Relationship between Linear A and Linear B. Cambridge University Press.
• Salgarella, E., and Castellan, S. SigLA: The Signs of Linear A, a palaeographical database. sigla.phis.me.
• Steele and Meissner. From Linear B to Linear A: The problem of the backward projection of sound values.
• Younger, J. G. (2024). Linear A Texts and Inscriptions in phonetic transcription and Commentary, and Linear A Lexicon.

Every count of our corpus on this page is measured from the corpus files each time the site is built, and names its base. The other numbers are typed in: figures from the papers above, and results of our own tests, each with the date the test ran. The unicity test, the ku-ro audit, the formula test and the 3D pilot are each written up with their data. Write to the address on the about page for a copy.