The Byblos syllabary

The corpus is a handful of small objects from one site. Dunand's ten inscriptions of 1945 are lettered a to j. Two are bronze tablets inscribed on both faces: tablet c, about 16 by 11 cm with about 225 signs, and tablet d, about 21 by 12 cm with about 457 signs, together carrying two thirds of all known text⁠[3]. Four are bronze spatulas, flat triangular blades about 5 by 9 cm with a stem handle, inscribed on one or both faces with 17 to 84 signs each; Dunand took them for labels attached to votive gifts⁠[1]. Four are fragments of stone stelae, of which fragment a, with about 120 signs in a monumental style with ruled lines, is the largest, and fragment g is written in vertical columns⁠[3]. In 1978 Dunand added four short stone fragments, k to n, with 10 to 45 signs each; four lines on fragment l have been read by James Hoch as an Egyptian hieratic date formula, the one externally readable passage in the corpus⁠[2]⁠[16]. A few Byblos signs also appear under later Phoenician inscriptions, on the Azarba'al spatula and other objects reused as palimpsests, and isolated signs occur on weapons and pottery; Colless and Garbini have assigned further objects from Megiddo, Egypt and elsewhere to the same script, assignments that Vita and Zamora call very dubious⁠[3]⁠[17]. The signs on the bronzes were punched with chisels rather than scratched. The find contexts were recorded by Dunand but none is stratigraphically secure, which is why the date is open⁠[6]. The objects belong to the Lebanese antiquities service and are kept in Beirut; some travelled to Leiden for the 2024 exhibition Byblos: A Legacy Unearthed⁠[18]. The only full edition is Dunand's, with photographs and his drawings; later analysts have mostly worked from the drawings, a dependence the GEAS assessors single out as a weakness⁠[1]⁠[11].

The size of the inventory is the strongest structural fact. Dunand counted 114 signs; Garbini reduced that to about 90 by treating damaged forms and variants as one; Vita and Zamora put the true figure at no fewer than 100, and the GEAS transcription works with about 100 graphemes⁠[4]⁠[3]⁠[11]. The two tablets alone use 74 distinct signs and the other texts add 30⁠[4]. A set of that size cannot be an alphabet and is unlikely to be purely logographic, so the script is taken to be a syllabary, probably of open consonant-vowel syllables, with some logograms; Garbini noted that a Semitic consonant set times three vowels would give 66 to 168 signs, which brackets the inventory⁠[3]. A statistical estimate of the phonetisation degree, about 69 percent, implies that roughly a third of the signs should be logograms⁠[11].

Direction is settled: right to left, with one spatula face running left to right and one stele in vertical columns⁠[4]. Short vertical strokes on several spatulas and stone fragments appear to divide words, though not consistently⁠[1]. Sobelman in 1961 and Mäder from 2019 have shown that recurring sign sequences, likely morphemes, and probable word boundaries can be extracted from the texts without knowing any sound values⁠[13]⁠[12].

The origin of the sign shapes is partly known. Dunand found about 50 signs similar to Egyptian hieroglyphic or hieratic forms and 25 apparently taken from the hieroglyphic repertoire; Hoch argued that the signs derive specifically from the Egyptian uniconsonantal signs, and a GEAS shape comparison found that eight of the twelve most frequent Byblos signs lie close to an Egyptian consonant sign⁠[3]⁠[19]⁠[20]. Egyptian inspiration is therefore the consensus, but a direct and exclusive derivation is not proven⁠[3]. Two sound values, me and pa, have been proposed from a cylinder seal that names Akhenaten's three daughters in Byblos signs beside their hieroglyphic names on the Berlin family stela; the paper is not yet published⁠[14]. Everything else about the sounds and the language is hypothesis.

The first attempts came within a year of publication. Bedřich Hrozný, the decipherer of Hittite, tried the stone stele in 1946 and Édouard Dhorme published a Phoenician reading of the whole corpus the same year, starting from a group of seven strokes on tablet c that he took for a year number in a date formula and reading the preceding signs as Phoenician b-šnt, "in the year"⁠[21]⁠[7]. Dhorme assigned values to about 75 signs, but the texts that resulted were not coherent Phoenician and the reading was not taken up⁠[4]. Malachi Martin re-examined the objects in 1961 and reclassified the signs into 27 palaeographic classes, correcting Dunand's readings, but never published the promised continuation⁠[22]⁠[23].

George Mendenhall's 1985 book is the most complete claim. He posited 26 consonant signs and 63 consonant-vowel signs writing a West Semitic language older than the division between Canaanite and Arabic, dated the texts to about 2400 BC, and translated eight of them, tablet c as a marriage contract⁠[5]⁠[4]. Giovanni Garbini's review found the translations cryptic and the method unconstrained, and most Semitists set the reading aside; Colless is its one sustained defender⁠[24]⁠[10]. Jan Best's reading of 2008 to 2010, built on sign comparisons with Cretan Hieroglyphic and Linear A, produced a dedication to the sun god Shuraya in a Semitic language with Akkadian and Hurrian words; Fred Woudhuizen's 2007 variant added many logograms and was rejected by Best himself⁠[8]⁠[25]. In 2024 Elisabeth Schmutz and Michael Mäder tested the Best and Woudhuizen values against the GEAS methodology, the seal bilingual and a phonetisation statistic; the values failed the names on the seal, allographs had been posited only where translations needed them, and the proposal could not be falsified⁠[11]. Garbini's own 2009 reading, which links the spatulas to later Phoenician spatulas and takes tablet c as a remedy, is the one proposal the assessors judged methodologically admissible, and it awaits testing⁠[26]⁠[11]. Ihor Rassokha's 2017 reading of the texts as Sanskrit through Brahmi letter values assumes a script attested a thousand years after any proposed date for Byblos and has no following⁠[27].

The live positions divide along two axes, language and date. On language, the oldest position is that the script writes a Canaanite dialect close to Phoenician, as Dhorme held⁠[7]. Mendenhall and Colless hold instead for an archaic West Semitic language from before the Canaanite and Arabic branches separated, with Colless treating the syllabary as the parent of the proto-alphabet, so that 18 of the 22 Phoenician letters would descend from its signs⁠[5]⁠[10]. Best read a Northwest Semitic language with Akkadian and Hurrian loanwords, which would fit a Late Bronze Age cosmopolitan Byblos⁠[8]. Rassokha's Sanskrit stands alone⁠[27]. The default among epigraphers, stated by Vita and Zamora and by Daniels, is that the language is undetermined: a Semitic language is probable from the place, but nothing in the texts yet proves it, and the corpus is too small to force a reading⁠[3]⁠[28].

On date, Dunand's Middle Bronze Age placement, about 1900 to 1600 BC, remains the conventional one and is supported by the Egyptian look of the signs⁠[4]. Mendenhall and Hoch favour an earlier origin, in the Old Kingdom or Early Bronze Age, because the closest Egyptian parallels are old⁠[5]⁠[19]. Sass argues the opposite: the shared word dividers, framing and sign forms with ninth-century Byblian alphabetic texts, and spatulas inscribed in both scripts, place the pseudo-hieroglyphs around 900 to 830 BC, in which case some signs imitate cursive letters rather than foreshadow them⁠[6]. The two disputes interact. If Sass is right, the script is a short-lived prestige system of the Iron Age and Colless's genealogy of the alphabet is reversed; if Dunand or Hoch is right, the script is a Bronze Age Egyptian-inspired syllabary and a possible relative of the proto-alphabet, which itself is attested from about 1900 BC⁠[6]⁠[10].

A bilingual would settle the sounds, and one small one may already exist: the cylinder seal naming Akhenaten's daughters, if its attribution to the Byblos script holds, would fix several syllabic values at once and test every proposed syllabary against them⁠[14]⁠[11]. Any decipherment must now read those names correctly. A second check is internal. Mäder's Unicode transcription makes sign frequencies, positional statistics and recurring sequences reproducible; a proposed grammar must match the morpheme boundaries that emerge without sound values, and a proposed sign list must respect the variant distinctions in the data rather than merging them where a translation requires⁠[12]⁠[11].

The date could be settled without reading a word. The bronzes can be analysed for alloy composition and compared with dated Byblian bronzes of the Middle Bronze Age and the Iron Age; the palimpsest spatula, with Byblos signs under Phoenician letters, fixes only a lower bound⁠[6]. A re-examination of Dunand's excavation records, now returned from Geneva to Lebanon, might recover find contexts for individual objects⁠[6]. New photographs would help everyone: nearly all work since 1945 has relied on Dunand's drawings, and the GEAS assessors note that readings have been built on drawn forms that the photographs do not always support⁠[11]. Finally, new texts are the real need. A corpus of about 1,000 signs with no document longer than 460 signs cannot support a decipherment by internal analysis alone, and claims that the script occurs at Megiddo, in Egypt or in Italy should be tested object by object, since a confirmed find outside Byblos would both enlarge the corpus and date it⁠[3]⁠[17].