Cretan Hieroglyphic

The documents are of two kinds. The first is seals: small engraved stones, and a few of bone or metal, with one to four faces cut with signs, meant to be pressed into clay; about 200 of them survive, from residential quarters and graves as well as palaces⁠[2]⁠[13]. The second is clay: fewer than 150 inscribed pieces, mostly four-sided bars, round medallions with a suspension hole, crescent-shaped nodules and a few page-shaped tablets, together with sealings that carry the impressions of seals⁠[2]⁠[8]. A count made in 1989 gave 150 seals and sealings with 832 signs and 120 other clay documents with 723 signs, about 1,550 signs in all; finds since then have added a little⁠[9].

The clay comes chiefly from four deposits. Evans found the Hieroglyphic Deposit in the palace at Knossos in 1900. The French excavations at Malia produced documents from the palace and, in the 1960s and 1970s, the largest group, from the Middle Minoan II building complex called Quartier Mu⁠[8]⁠[2]. At Petras in eastern Crete, a Middle Minoan IIB archive of clay documents, seal impressions and seals was excavated from 1995 in a sealed destruction level and published in 2010; it is the only hieroglyphic archive dug with modern methods⁠[25]. Phaistos has sealings and a few inscribed pieces⁠[13]. Isolated finds at Vrysinas near Rethymno in 2011, at the Knossos cult centre in the 2010s and on Samothrace show the script reached beyond the central and eastern palaces⁠[26]⁠[15]⁠[13].

CHIC publishes the whole corpus with photographs, drawings and transcriptions, and numbers the documents H 1 to 122 for clay, I 123 to 179 for impressions and S 180 to 314 for seals, with a few miscellaneous objects after them⁠[8]⁠[9]. Most of the material is in the Heraklion Archaeological Museum, with the Petras finds in Sitia and a few seals abroad⁠[27]. There is no open digital edition. The Florence DBAS-CHS database indexes the seals by sign, motif, material and context for personal and scientific use; the INSCRIBE project at Bologna has made 3D models of seals, bars and nodules that can be viewed but not downloaded⁠[28]⁠[23].

The script is almost certainly a syllabary with logograms, that is, a system in which most signs stand for a syllable and some for a whole word such as a commodity. Evans already saw that 135 signs were too few for a purely word-based script, and the 2024 volume records broad agreement that the system rendered speech systematically⁠[13]⁠[2]. CHIC's inventory is 96 syllabograms, logograms, numerals, nine fraction signs and two punctuation marks⁠[8]⁠[9]. The numerals are decimal, with signs for units, tens, hundreds and thousands, and are read without difficulty⁠[9]. Reading direction is mostly left to right on clay; on seals it has to be worked out face by face⁠[9].

The inventory is less secure than CHIC's numbering suggests. Montecchi, Ferrara and Valério reviewed every sign in 2021 by three tests: whether two shapes share the features that matter, whether they never occur on the same kind of document, and whether they swap with each other inside the same sign group. They proposed eleven groups of mergers, removed a sign known only from a lost document, reassigned three to Linear A, and found that 33 signs attested once would fall to about 18; the list would then hold about 98 signs⁠[16]. They also dropped the division into syllabic and logographic classes, which CHIC had taken over from Linear B without evidence from the hieroglyphic texts themselves⁠[16].

Something is known of use. The clay documents are accounts: sign groups followed by logograms and numbers, like the Linear A tablets but shorter⁠[8]⁠[29]. The seals carry short sign groups, and the 2024 volume reports growing agreement that these name the seal's owner or office rather than the goods stamped⁠[30]⁠[21]. One group, the Archanes formula, recurs from the late Prepalatial seals of Archanes-Phourni to the Neopalatial seal from the Knossos cult centre, which also links it to fraction signs⁠[14]⁠[15]. Sound values are not known. Where a hieroglyphic sign has a Linear A look-alike, the Linear B value is sometimes written beside it, but the catalogue sites that do this warn that the values are hypotheses⁠[27]⁠[31].

The seals-as-writing question has been argued on the evidence of layout. Decorte's 2017 study treated every mark on a seal face, including the fillers and dots CHIC set aside, as part of one graphic syntax and concluded that the line between script and ornament was drawn by the editors, not the Minoans⁠[11]. Jasink had already restored more than thirty seal motifs to the status of candidate signs⁠[10]. Ferrara, Montecchi and Valério in 2023 asked which excluded motifs behave like signs, by their position, their combinations and their association with recognised signs, and argued that some of them are writing⁠[12]. The counter-position, that the seals were luxury ornaments whose script was displayed more than read, was put by Jasink and Weingarten in 2024; it does not deny that the signs are script⁠[30]⁠[21]. No test has yet shown which excluded motifs carry sound.

The relation to Linear A has been tested by comparison of sign shapes and of sign behaviour. Owens argued in 1996 from shared shapes and shared groups for a common Prepalatial source⁠[5]. Ferrara, Valério and Montecchi compared the two scripts sign by sign in 2022⁠[6]. Davis in 2024 applied the method he had used on the Phaistos Disc, comparing how the two scripts combine their shared signs within words, to ask whether they write one language; Donnelly's review found the result hard to assess and objected that excluding the cat-face sign AB 80 from the data was over-cautious⁠[20]⁠[32]⁠[21].

Decipherments have been few and none has been accepted. Best and Woudhuizen derived individual signs from Egyptian hieroglyphs and read texts as Luwian; the derivations are disputed sign by sign and the readings have not been taken up by Anatolian or Aegean specialists⁠[17]⁠[21]. Kenanidis and Papakitsos treat the script as an ornamental form of a reconstructed syllabary with Sumerian sign names, and Revesz places it in a script family read through Uralic; both rest on resemblance of shapes and neither has been tested against the corpus by others⁠[18]⁠[19]. There has been no published statistical or machine-learning decipherment attempt specific to Cretan Hieroglyphic, which reflects the lack of a machine-readable corpus.

On the language, the only position with a specialist following is that Cretan Hieroglyphic writes the same Minoan language as Linear A. It rests on the scripts' coexistence in the same palaces, their shared signs and the same kinds of document, and on Davis's syllabotactic comparison⁠[4]⁠[5]⁠[20]. It is not established: the two scripts were kept apart by the same administrations, which has been read as a sign of two languages or two communities, and the shared signs need not share values⁠[4]⁠[6]. The Luwian reading of Best and Woudhuizen, the Sumerian-named proto-syllabary of Kenanidis and Papakitsos and the Uralic family of Revesz are the published alternatives⁠[17]⁠[18]⁠[19]. The default among epigraphers is that the language cannot be identified from this corpus, whatever it is⁠[13].

On the seals, three positions stand. CHIC's is that some seal graphs are writing and the rest decoration, the test being attestation on clay⁠[8]⁠[2]. Decorte's, shared in substance by Jasink and by the INSCRIBE team, is that the whole seal face is script or belongs to its syntax⁠[11]⁠[10]⁠[12]. Jasink and Weingarten's is that the seals were ornaments carrying names in a script meant to be seen, which is compatible with either of the others⁠[30].

On origins, Evans's Egyptian stimulus is no longer accepted as the source of the signs, though Egyptian glyptic and local seal-carving both shaped the medium⁠[33]⁠[21]. Whether the Archanes group is the first stage of this script, an independent script, or early Linear A is unresolved; the 2024 volume rejects the last of these⁠[14]⁠[30]⁠[13]. Whether the Arkalochori axe and the Malia altar stone belong to the script at all is likewise open: CHIC includes them with reservations, and the axe's signs are closer to the Phaistos Disc⁠[9]⁠[7].

A bilingual would settle the sound values, and none exists; the nearest thing is the set of sign groups shared with Linear A, which can be tested but not read⁠[5]⁠[6]. Short of that, three things would move the problem.

The first is an open, sign-level digital corpus: every document in CHIC and later finds transcribed in a stable sign numbering with object type, site, date and the position of each sign, as SigLA has done for Linear A⁠[34]. With it, the 2021 sign mergers could be checked by anyone, the seals-as-writing question could be put as a statistical one (do the excluded motifs occupy the slots that signs occupy, more often than chance), and the Davis comparison could be re-run with different sign sets⁠[16]⁠[12]⁠[20]. Encoding the script in Unicode, which has not been proposed, would make such a corpus durable⁠[22].

The second is new texts from controlled excavation. Petras showed what a sealed deposit adds: documents, impressions and the seals that made them in one context, dated by pottery⁠[25]. A longer clay text, or more seals paired with their impressions, would fix which seal graphs were pressed into administrative clay⁠[8].

The third is a test of the Minoan-language hypothesis that does not depend on shared shapes alone: whether the sign groups that recur in both scripts, such as the Archanes formula and the Linear A libation formula, behave the same way in the same positions, and whether the Linear A fraction values found in 2021 also make the hieroglyphic totals add up⁠[15]⁠[35]⁠[16].