Sources
Every work this project has engaged with, filed by how far we actually got through it. One we read whole. Most we did not, and the label says so. A bibliography usually flatters its author. This one is here to bound what we are allowed to claim.
Not listed: software tools and encoding standards, which are credited in the log entries that use them, and works we know only as citations inside other works. Naming those here would claim an acquaintance we do not have.
The last three groups are the honest part. Five works sit within reach with not one line of them quoted yet, and clearing the unbought shelf would cost about €350 before tax plus $43, with four more items nobody would quote us a price for. That ceiling is what this project is allowed to claim about several of the hypotheses it discusses, and pretending otherwise would be the easiest lie on the site.
Read in full2
We went through the whole thing. Two entries qualify, and we would rather print that number than round it up.
The subject of entry #026. Read in full, all 29 pages: every load-bearing claim was pulled out with a page locator, the checkable parallels were verified against our locked copy of the corpus, and the damaged readings were compared with the original GORILA plates. Self-published, without the independent expert review a journal would arrange.
Ester Salgarella and Anna Judson, “Signs of the time? Testing the chronological significance of Linear A and B palaeography”, Ariadne Supplement 5 (2024), pp. 359–379
Read cover to cover to find out whether a per-sign palaeographic grading of the corpus exists. It does not: the study covers ten signs, and it argues against treating palaeography as chronological evidence. That answer closed a line of work rather than opening one.
Consulted22
Used for specific passages, tables or data, each verified in the source itself. We claim nothing about the pages we did not use.
Margalit Finkelberg, “Minoan Inscriptions on Libation Vessels”, Minos 25–26 (1990/91)
The first hypothesis through the testability audit, and the closest reading in the project short of a full one: her thirteen sine qua non traits are quoted verbatim with page locators, and all sixteen inscriptions she cites were checked against our locked copy of the corpus. We worked through the argument, not yet through every page, so it sits here and not above.
The backbone of entry #015. The sign and impression counts, the “legion” of decipherments, the “typewriting, not printing” correction and the unicity-distance argument are all quoted from these four pages, which we downloaded and worked through.
The forgery accusation quoted in entry #015, used with its own caveats on the page: not peer-reviewed (the author says so in the addendum) and published in the author’s own magazine. Also the source of the museum’s “non movable” reply and of the catalog of decipherment attempts.
The find circumstances (the foreman’s evening inspection), the 242 count, the recent dating argument and the chronological authenticity argument (comb-sign parallels excavated in 1955 and 1965) in entry #015 come from this signed research summary, read via the Internet Archive because the site blocks robots.
Heraklion Archaeological Museum, exhibit page “The Phaistos Disc”
The institutional record for the object entry #015 describes: find date of 3 July 1908, the stamped-before-firing technique, the museum’s dating label, and the careful official wording on the script’s relation to Linear A.
Michael Ventris and John Chadwick, Documents in Mycenaean Greek, Cambridge University Press, 1956
The primary text for our Ventris control. Blegen’s “too good to be true” letter is quoted from these pages and was re-read there, not lifted from a secondary account. At 494 pages, we mined it. We do not claim to have read it whole.
John Killen and Anna Morpurgo Davies, memoir of John Chadwick (1920–1998), Proceedings of the British Academy 115 (2002), p. 133 ff.
Our own records called this a memoir of Ventris until we opened the file and read the first line. It is a memoir of Chadwick, and it carries the locked-crate answer to Beattie’s accusation that Ventris had worked backwards from the tablets. That passage was verified here. We used those pages, not all thirty-six.
The paper we declared and replicated in phase 4. The replication forced us through its method and results line by line. Their solution counts did not reproduce under any reading of the method we could state, and we documented the divergence instead of tuning constraints until the numbers agreed. We make no claim about the sections the replication never touched.
Quoted verbatim for two facts we needed: how the database is actually structured, and that its transliteration values are Younger’s rather than an independent judgement. Both mattered because we were asking whether SigLA could serve as a palaeographic oracle. It cannot.
Michele Corazza, Fabio Tamburini, Miguel Valério and Silvia Ferrara, “Unsupervised deep learning supports reclassification of Bronze age Cypriot writing system”, PLoS ONE, 2022
The method template we costed for clustering sign images: their pipeline on Cypro-Minoan is the closest published precedent for what a phase-5 image study would do. Consulted for the method, not the data, which is neither Linear A nor redistributable.
Louis Dautais and Marc Gabolde, “‘Crossing the Wine-Dark Sea’ from Knossos to Memphis: Hellenized Cretans in Egypt in the Mid-Eighteenth Dynasty (BM EA 5647)”, Journal of Ancient Egyptian Interconnections 47 (2025), pp. 7–28
Source of the closed list of seven Cretan names in Egyptian transcription that our comparanda audit treats as usable. The chain runs through Linear B rather than straight from Egyptian to Linear A, and saying so is half the value of the entry.
John Younger, Linear A texts and inscriptions, sign grids (ABgrids)
The Linear A to Linear B sound correspondences our phase-3 tests run on, together with the confidence legend marking which values are secure. Pinned to a dated Wayback snapshot because the live host is gone.
Ester Salgarella and Simon Castellan, SigLA: The Signs of Linear A (database)
An independent transcription we checked our own corpus against, sign by sign, and the source of the per-instance sign drawings a future image study would need. Its licence is non-commercial with share-alike, which bounds what we can redistribute.
mwenge, lineara.xyz corpus, pinned at commit 083cb0c
Where our 1,721 inscriptions come from, and the only source on this page we have read exhaustively in the machine sense: every record parsed, counted and cross-checked. Its licence is undeclared and we are still clarifying it with the maintainer before republishing any transcription.
mwenge, linearb.xyz corpus, pinned at commit 84e0b00
Where our 5,832 Linear B inscriptions come from (entry #021): the same maintainer as our Linear A source, parsed with the same machinery, locked to an exact version so the counts can be redone. Its licence is also undeclared; we publish only derived counts and statistics, never the transcriptions themselves.
One of the two known-language yardsticks in entry #022: 394 Sumerian literary works, included so we could see what our measurements actually catch when the language behind the writing is already understood. We took only aggregate counts from it, each checked against the published totals before use. None of its text is reproduced on this site.
The other known-language yardstick in entry #022: a Finnish collection of 15,136 sentences, used the same way and for the same reason. Finnish is written in an ordinary alphabet, which is what makes it a clean point of comparison. We used only aggregate counts, each checked against the published totals.
The one real corpus among the three comparanda added in entry #023: 816 Urartian inscriptions, 661 edited word by word, which come to 13,860 words of running text. Urartian belongs to the Hurro-Urartian family and stands in for that family as a whole, as its best-documented member; it is written in cuneiform, a writing tradition entirely apart from the Aegean scripts. We took only aggregate counts from it, each checked against the published totals, and reproduced the arithmetic independently. None of its text is reproduced on this site.
A comparandum weighed in entry #023 and set aside as a dictionary, not a corpus: 1,602 Greek headwords that scholars trace to a substrate spoken in Greece before Greek. It is a word list, so our census cannot be run on it; we counted it and used no more. Its forms are hellenised reflexes, not the substrate itself. None of its text is reproduced here.
The other comparandum set aside in entry #023 as a dictionary: 3,799 lemmas of the minor ancient Anatolian languages, each entry already carrying an editor’s analysis into roots and endings. Measuring that segmentation would measure the editor, not the language, so we counted it and stopped there. None of its text is reproduced here. The licence requires the deed hyperlink, which is provided.
The twin project of entry #020, published pseudonymously; cited, never contacted. We consulted its README, preprint and hypotheses table, and used its nine published anchor words as a check declared in advance on our candidate list. Its name-position files and its selection rules were deliberately never opened, so that our criteria could not lean on theirs.
The standard corpus edition and the final authority for sign classification. Our AB numbering descends from it, but always through other people’s datasets, never from the volumes. All five volumes are scanned in full on the École française d’Athènes own digital library, free and without an account. For entry #026 we went to volume V, pages 18–19, the plates of IO Za 2, and read the damaged spots off the photographs to check the study’s account against the original edition. We still have not gone through the whole thing.
In hand, not yet read4
Available to us and not read. These are on disk with the text layer checked. Not one line of these is quoted anywhere in our work yet. This is the reading queue, listed so the queue cannot quietly pass for reading.
Giulio Facchetti, “Minoico ki-ro”, Studi Micenei ed Egeo-Anatolici 44/2 (2002), pp. 173–197
Downloaded as a candidate Tyrrhenian source. The same-day audit filed it as an accounting analysis of KI-RO that does not carry the Tyrrhenian argument, which lives in a book we have not bought. We have not read this one yet.
Gary Rendsburg, “On Jan Best’s ‘Decipherment’ of Minoan Linear A”, Journal of the Ancient Near Eastern Society 14 (1982), pp. 79–87
The published demolition of Best’s Semitic decipherment, on disk and unread. Its future role is limited by design: a hostile critic can tell us where Best’s claims are, never what they are.
Cyrus Gordon, “Linguistic Continuity from Minoan to Eteocretan”, Studi Micenei ed Egeo-Anatolici 3 (1967), pp. 89–92
Gordon’s own four-page synthesis of his Semitic continuity claim, queued as the primary text for that lane. The book it condenses is on the shelf below, unbought.
Gareth Owens, “A Linear Inscription from Olympia”, Talanta XXX–XXXI (1998/99), pp. 143–146
Queued as the primary text for the Greek and Indo-European lane. Four pages, not yet read.
Priced and not bought11
Identified as relevant and not bought. Where we found a price, it is shown. This shelf is the ceiling on what we are allowed to claim about the hypotheses sitting inside it.
Brent Davis, “The Phaistos Disk: A New Way of Viewing the Language Behind the Script”, Oxford Journal of Archaeology 37.4 (2018), pp. 373–410
The 2018 statistical comparison entry #015 mentions, arguing the Disc encodes the same language as Linear A. We have read only the author’s own talk summary, not the article: the claim is reported at that strength and no more.
Giulio Facchetti and Mario Negri, Creta minoica, Olschki, 2003 (ISBN 9788822252913)
This is the Tyrrhenian hypothesis, not a summary of it, and there is no free substitute. Until we buy it, anything we say about that family is bounded by what its critics report.
Peter van Soesbergen, The Decipherment of Minoan Linear A, volume I, parts I–II
The Hurrian proposal, and the most expensive item on this page. It is also the clearest illustration of the ceiling: one hypothesis family costs more to read than this project has ever spent.
Cyrus Gordon, Evidence for the Minoan Language, Ventnor Publishers, 1966 (ISBN 9780911566062)
The book-length Semitic argument, second-hand print only. We have the four-page synthesis instead and say so, rather than letting the synthesis stand in for the book.
Cyrus Gordon, “Notes on Minoan Linear A”, Antiquity 31/123 (September 1957), pp. 124–130
Where Gordon’s Semitic case begins, and the reason our log dates it to 1957. We have the DOI and the catalogue record. We do not have the article.
Cyrus Gordon, “Akkadian Tablets in Minoan Dress”, Antiquity 31/124 (December 1957), pp. 237–240
Three months after the first paper Gordon moves from West Semitic to Akkadian, which is why “the Semitic case” is really two. We read the opening extract, “From that moment on I approached Linear A with Akkadian in mind”, and nothing past it.
Ester Salgarella, Aegean Linear Script(s), Cambridge University Press, 2020
Holds the finer palaeographic judgement phase 5 wanted, in print and in an image-only appendix. We could not confirm it contains a per-sign graded table rather than prose and plates, so that question stays open instead of being settled by assumption.
Cook, The Decipherment of Linear A, Pen & Sword, 2025 (ISBN 9781036134655)
The Egyptian proposal. Trade press rather than peer reviewed, which lowers what buying it would settle, though not its price.
L. R. Palmer, “Luvian and Linear A”, Transactions of the Philological Society, 1958
The other name on the Anatolian lane beside Finkelberg, and the older one. Behind a publisher paywall that answered our request with a payment-required response.
Jan Best and Fred Woudhuizen, Ancient Scripts from Crete and Cyprus, Brill, 1988
The expanded statement of the Semitic reading that Rendsburg attacks. We hold the attack and not the position, which is the wrong way round and worth admitting. The 1988 paperback is out of print; only the publisher ebook is left, and it is the second most expensive line on this page.
Michael Ventris and John Chadwick, “Evidence for Greek Dialect in the Mycenaean Archives”, Journal of Hellenic Studies 73 (1953), pp. 84–103
The middle link in the chain our Ventris control depends on, from Work Note 20 in June 1952 to this announcement to the 1956 book. We verified the two ends and paid for neither the middle.
Never reached7
Searched for and not reached: out of print, not digitised, or blocked at every door we tried. An entry here means we do not know how to get it, not that it does not exist.
Vladimir Georgiev, in Linguistique balkanique 7 (1963), pp. 1–104
Cited by others as a foundational statement of one hypothesis family. Every catalogue we could query either returned nothing or refused the query, so we can name the gap and not close it.
Vladimir Georgiev, Le déchiffrement des inscriptions crétoises en linéaire A, Sofia, 1957
The companion to the entry above and equally out of reach: a library catalogue answered with an access error and no second-hand copy surfaced anywhere. Not proof it is unobtainable, only that we do not know how.
Jan Best, in Talanta XIV (1982)
The primary text Rendsburg demolishes. Two free Talanta PDFs exist, but their pagination does not match what Rendsburg reviews, so we cannot confirm they are the same text. Holding a file is not holding the source.
Gareth Owens, “The Structure of the Minoan Language”, Journal of Indo-European Studies 27 (1999)
The fuller statement behind the four-page Olympia note. The journal’s online archive begins seven volumes later, and the one reseller link we found redirected to a block page.
Gareth Owens, Kritika Daidalika (ISBN 9789025610968)
Collected papers that would consolidate the Greek and Indo-European lane. The bookseller record answers, verbatim, that this edition is not available: zero new and zero used copies.
Michael Ventris, Work Notes on Minoan Language Research, ed. Anna Sacconi, 1988
Work Note 20 of 1 June 1952 is where the Greek hypothesis is first written down, and the whole force of our Ventris control rests on its date. We take that date from the memoir rather than from the notes themselves, and the entry is here so the difference is visible.
Margalit Finkelberg, “The Language of Linear A: Greek, Semitic, or Anatolian?”, in R. Drews (ed.), Greater Anatolia and the Indo-Hittite Language Family, JIES Monograph 38
The later restatement of the hypothesis we audited first. Our own two records disagree about whether the volume can be borrowed or only viewed, which is its own small lesson: we have not been back to settle it.