Proto-Sinaitic and Wadi el-Hol: the first alphabet

The corpus is small and scattered. At Serabit el-Khadim, the Egyptian turquoise-mining site in south-west Sinai, Flinders Petrie found in January 1905 a small sandstone sphinx and a kneeling statuette in the temple of Hathor, and about eight rock-cut tablets above Mine L, all carrying signs that were not Egyptian⁠[1]. Later expeditions, above all Kirsopp Lake's Harvard campaigns of 1927, 1930 and 1935, raised the total to roughly forty inscriptions and fragments, numbered Sinai 345 to 375 in the Egypt Exploration Society's catalogue and its continuations⁠[22]⁠[23]. Recent counts give about 47 texts from Sinai, many only a few signs long, with one more at Timna in the Arabah⁠[4]⁠[24]. The sphinx (Sinai 345) is in the British Museum; the statuette and most of the removable panels are in the Egyptian Museum in Cairo; the rest remain on the rock⁠[25].

In 1993 John and Deborah Darnell found two further inscriptions at Wadi el-Hol, a pass on the desert road between Thebes and Abydos, among Egyptian rock inscriptions of the Middle Kingdom. One runs horizontally and one vertically; together they hold about 28 signs⁠[5]. They are the only texts in the script found in Egypt proper.

The standard catalogue is Sass's 1988 corpus, with a palaeographic table for each letter⁠[11]. Hamilton's 2006 monograph is the reference for the sign forms and their Egyptian prototypes⁠[3]. The Wadi el-Hol texts were published with photographs and drawings by Darnell, Dobbs-Allsopp, Lundberg, McCarter and Zuckerman in 2005⁠[5]. No open digital corpus exists: the photographs of the West Semitic Research Project are for study only, and the IFAO supplementary catalogue of 2012 is in print⁠[21]⁠[26].

The script is an alphabet in the strict sense of a consonantal alphabet, or abjad: each sign stands for one consonant and vowels are not written. This much has been accepted since Gardiner's 1916 paper, which observed that the signs are pictures of things whose Semitic names begin with the consonant the sign writes. The ox head is 'alp, so it writes the glottal stop; the house is bayt, so it writes b. Gardiner read the sign group that recurs on four of Petrie's monuments as b-'-l-t, Semitic ba'alat, "Lady", a title of the goddess equated with Hathor, and the preceding sign as l, "to"⁠[2]. That reading has held for a century and is the foothold for everything else. The sphinx carries the Egyptian dedication "beloved of Hathor, mistress of turquoise" on one shoulder and a Proto-Sinaitic line with b'lt on the other, which makes it a near-bilingual, though the two texts are not translations of each other⁠[25].

The sign inventory is reconstructed at about 27 letters, more than the 22 of Phoenician, because the early script kept consonants that Phoenician later merged, such as the two kinds of h and the sound ghayin⁠[3]⁠[27]. The letters are the direct ancestors of Phoenician, Hebrew, Greek and Latin letters, which is how their values are known: each sign can be traced forward through the Canaanite inscriptions of the Late Bronze Age to the readable Phoenician alphabet of about 1000 BC⁠[3]⁠[14]. Direction is not fixed: texts run left to right, right to left and vertically, often in the same place⁠[6].

The language is Northwest Semitic, the family of Canaanite, Phoenician and Hebrew, as the letter names and the few secure words show⁠[2]⁠[4]. Beyond b'lt, the readable material is thin: a handful of personal names, 'l "god" and rb "chief" at Wadi el-Hol, and the formula "to the Lady"⁠[5]⁠[4]. The first full sentence in the descendant Canaanite script is the seventeen-letter wish against lice on the Lachish ivory comb, read in 2022⁠[10].

Three questions have been tested repeatedly: the date, the inventors, and the readings. On the date, Petrie's c. 1500 BC rested on a potsherd and the stone of the sphinx⁠[1]. Gardiner placed the texts under Amenemhat III and IV (1831-1777 BC); Albright moved them to 1525-1475 BC on palaeographic grounds⁠[7]⁠[4]. Sass argued for the Middle Kingdom in 1988 and then, in 2004-2005, for a date around 1300 BC⁠[11]⁠[12]. The Wadi el-Hol texts, dated by the Egyptian inscriptions around them to about 1850-1800 BC, and the Lachish sherd, radiocarbon-dated to the 15th century BC, make a 13th-century invention impossible, though they do not by themselves date the Serabit texts⁠[5]⁠[14]. Parker's re-examination of the find spots at Serabit (2022, 2026) argues for the reign of Senwosret III and finds the New Kingdom arguments weak⁠[13]⁠[28]. Haring, an Egyptologist, calls the Twelfth Dynasty case strong but not closed⁠[29].

On the inventors, Goldwasser argued in 2006 that Canaanite workers who could not read Egyptian took hieroglyphic pictures and named them in their own language, which explains why Egyptian sound values were ignored and why several look-alike hieroglyphs serve one Semitic letter⁠[15]⁠[30]. Rainey and Rollston answered that writing was an elite skill and that the inventors knew Egyptian script⁠[31]⁠[16]. Hamilton's catalogue finds hieratic as well as hieroglyphic prototypes, which would require some scribal knowledge⁠[3]. Sanders and Wilson-Wright side with invention at the margins, citing the Cherokee and Hmong scripts as cases of pre-literate inventors⁠[32]⁠[4]. Colless holds that the alphabet was simplified from the undeciphered Byblos syllabary⁠[33].

On readings, Albright's 1966 decipherment gave values for 23 letters and translations of most texts as votive formulae; the values are largely accepted, the translations largely not⁠[7]. Later whole-text readings by Puech, Wilson-Wright and others disagree with one another⁠[8]⁠[9]. No statistical or machine-learning study of the corpus has been published; the texts are too short and too few for frequency analysis to decide anything.

Four positions are live on what the texts say. The first, held by most epigraphers, is that the inscriptions are short Northwest Semitic votive texts and names: dedications "to the Lady" at the temple, and names or pleas of workers at the mines. It rests on b'lt, on the sphinx's Egyptian parallel, and on the find spots⁠[2]⁠[25]. Wilson-Wright's reading of Sinai 357 as a votive text to the Hurrian storm god Teššob is an example of this position pushed to a whole text⁠[9].

The second position is that most of the texts cannot be read at all with present evidence: beyond b'lt and a few words, every full reading depends on choices of direction, word division and damaged signs that another reader can make differently, which is why Albright's, Puech's and Wilson-Wright's readings of the same stones diverge⁠[7]⁠[8]⁠[29]. On this view the Sinai corpus will remain a sign list with a handful of words unless longer texts are found.

The third, Petrovich's claim that the texts are Hebrew and name biblical figures, has been rejected by specialists because it assigns values not supported by the descendant scripts and assumes a Hebrew distinct from Canaanite a millennium before it is attested⁠[34]⁠[4].

A fourth question sits beside these: whether the Umm el-Marra cylinders of about 2400 BC are alphabetic at all. Schwartz reads their twelve signs (eight distinct) as early alphabetic letters; Rollston agrees the shapes match; Steele objects that eight signs cannot show what kind of system they belong to, and that a 500-year gap separates them from the next evidence⁠[17]⁠[20]⁠[19]. Richey takes the cylinders and the alphabetic marks on Sealand-dynasty tablets as signs of a northern strand in the alphabet's history⁠[18].

The date of the Serabit texts would be settled by stratified finds: an inscription sealed under a datable Egyptian deposit, or ceramic and radiocarbon evidence from the debris at Mines L and M, which Parker's work shows has never been systematically examined⁠[13]. A new excavation permit for the plateau would be needed.

The readings would be settled by a longer text or a true bilingual. The sphinx is the closest thing to a bilingual and gives only one word. A text of thirty or more signs in a secure context, or an abecedary (a list of the letters in order) from the Middle Bronze Age, would fix the sign count and the order, as the TT99 ostracon and the Ugaritic abecedaries do for the later alphabet⁠[35]. Reflectance transformation imaging and photogrammetry of the rock panels, which have weathered since 1905, would at least fix the sign counts; the 2005 Wadi el-Hol edition shows what such documentation can do⁠[5].

The inventors' literacy would be tested by the prototypes: if every letter can be derived from a hieroglyphic picture as seen on monuments and none requires a hieratic (cursive) form, Goldwasser's case is strengthened; if some letters demand hieratic models, as Hamilton holds, the inventors had scribal contact⁠[30]⁠[3]. A sign-by-sign test against dated hieratic and hieroglyphic palaeography is possible now with published material.

Umm el-Marra would be settled by more inscribed objects from third-millennium Syria, preferably longer than four signs, and by a published radiocarbon series for Tomb 4⁠[20]⁠[19].