About this site
Pictures of Linear A, for somebody who has heard of it and wants to look. Not a research workbench, not a decipherment, not a theory. Pictures first, words second.
Who
Matt Navarre, a database developer in Rethymnon, Crete. Scholars have been at this since 1900. He started in 2025 and has not read a sentence either. Nobody has.
How it is made
A script reads the corpus files and writes every page. Every count of the corpus is measured from those files at build time and prints its base beside it. Some numbers are typed in: figures from published papers, and results of our own tests. Each of those names its source, and a test also names the day it ran.
Every picture of a sign is a tracing from a real object, cut by SigLA, with one exception. The numerals figure on the known page is a modern drawing, and its caption says so. The SigLA credit sits in the footer of every page.
AI agents wrote the build scripts and the ingest code, under Matt's direction, and every change they made is a dated commit in the project repository.
What was checked
On 18 September 2026 a review listed 245 statements of fact on this site, from the page text and 15 sample generated pages. It checked each one against the data or a named source. 199 held up as written. 45 did not: they were wrong, overstated, stale, or had no source on file. This release rewrites or cuts those it can. The rest wait on a decision, and the project notes list each one.
What it is not
It is not a font demo. No Linear A font character appears anywhere on the site, because font shapes are modern and the incised shapes are the evidence. It is not complete. Where an input was not ready at build time the page says "coming" instead of pretending.
How to read the numbers
Every count on this site says what it counts. We hold three layers: the word list, the sign-by-sign text, and the SigLA tracings. They come from different sources and cover different documents. Sign *301 is in 35 of 701 documents with a word list, and in 272 of 1,707 documents with a sign-by-sign text. Both numbers are right. They count different things (and neither one corrects the other).
The number after "of" is the base, and it comes from the same layer as the count. A sign missing from one layer is not missing from the object. The text may be untraced, unpublished, or filed under another record. The document page says which.
In the word list, SigLA marks some signs as unsure. A sign page counts those apart, as "unsure", beside the main count, and never adds them in. A sign SigLA could not identify counts nowhere.
A box headed "Sources disagree" means two sources read the same spot differently. We show both, each with its source, and we do not pick one. A count cannot settle what a sign means.
Contact
Write to matt@navarre.training. Corrections welcome, with the document id and the source.
The gates
The build refuses to publish when any of these fails. The output folder is deleted, so a bad page never ships.
- No Linear A codepoint and no private-use character in any file.
- No em dash in any file.
- The SigLA credit line on every page that shows a tracing.
- Every image resolves to a file, with alt, width and height.
- Every internal link resolves to a page.
- One h1 per page.
- Every statistic carries its denominator and names its layer.
- No template prints a count without a denominator.
- Every document with a conflict entry shows both readings and both sources.
- Every copy file passes site/scripts/copy_check.py.
- Images come only from the permitted folders.
- No GORILA plate scan in any file, and no comparator manifest that lists one.
- No page carries text from a copy-check ignore region.
Licences
This site uses two licences.
- Our own work: CC BY 4.0. The project's own text, code and generated data are released under CC BY 4.0, copyright 2026 Matt Navarre.
- Material from SigLA: CC BY-NC-SA 4.0. The tracings, the sign crops, the sign images, and the transcription fields that come from SigLA are released under CC BY-NC-SA 4.0, the licence of their source. This includes the per-document JSON downloads, which carry the same notice. The credit line: Sign drawings from SigLA: The Signs of Linear A, a palaeographical database, by Ester Salgarella and Simon Castellan (https://sigla.phis.me/), licensed under CC BY-NC-SA 4.0. Changes: crops are shown unchanged, resized for the screen.
Sources
- Corpus. The project's merged corpus, built from lineara.xyz (Robert Hogan) and SigLA (Salgarella and Castellan).
- Cited, not relicensed. These works are cited as sources. This site does not relicense them, and their own terms stay in force. 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. Del Freo, M., and Zurbach, J. (2024), Recueil des inscriptions en lineaire A, Supplement 1, Etudes Cretoises 21.6, Ecole francaise d'Athenes. John G. Younger, Linear A Texts and Inscriptions in phonetic transcription and Commentary, and Linear A Lexicon (2024). lineara.xyz (Robert Hogan).
- Photographs. Photographs of 35 of 1,883 documents in the corpus are on this site, one per document, in the comparator only. They are published with the permission of lineara.xyz. Image rights: Ecole francaise d'Athenes. No other photograph is on this site.
- Map. Outline: Natural Earth, 1:10m land, public domain. Relief: NASA Blue Marble: Next Generation (public domain, NASA Earth Observatory, R. Stockli). Hillshade from Copernicus DEM GLO-30, produced using Copernicus WorldDEM-30 (c) DLR e.V. 2010-2014 and (c) Airbus Defence and Space GmbH 2014-2018 provided under COPERNICUS by the European Union and ESA; all rights reserved.
- Search. Pagefind (MIT) and MiniSearch (MIT), both served from this site. No third-party request leaves the page.
The sign browser's own credits file lists: sigla, corpus, interpretations.
This build
Built 2026-10-17. 753 of 1,883 documents in the corpus have a whole tracing. 786 of 1,883 documents in the corpus have per-sign crops.