Safaitic abecedaries

Safaitic abecedaries are ancient inscriptions that are interpreted to represent an ordered inventory of the letters of the Safaitic script. Below we illustrate how close some of our output correspondences may seem and allow ourselves to reintroduce historical considerations hypotheses once again.

Maghrebian ABC and the Tay 19 order

If we take the Maghrebian ABC sequence:

ʾ   b   g   d   h   w   z   ḥ   ṭ   y   k
l   m   n   ṣ   ʿ   p   ḍ   q   r   s¹  t
ṯ   ḫ   ḏ   ẓ   ġ   š

swap ḫ and š and swap ḥ and z:

ʾ   b   g   d   h   w   ḥ   z   ṭ   y   k
l   m   n   ṣ   ʿ   p   ḍ   q   r   s¹  t
ṯ   š   ḏ   ẓ   ġ   ḫ

We can represent it in 3 column rows that we turn into a boustrophedon

ʾ   b   g           ʾ   b   g
d   h   w           w   h   d
ḥ   z   ṭ           ḥ   z   ṭ
y   k   l           l   k   y
m   n   ṣ           m   n   ṣ
ʿ   p   ḍ           ḍ   p   ʿ
q   r   s¹          q   r   s¹
t   ṯ   š           š   ṯ   t
ḏ   ẓ   ġ           ġ   ẓ   ḏ
ḫ                           ḫ

We then read the right most table TB+LR and get the following:

ʾ  w  ḥ  l  m  ḍ  q  š  ġ  b  h  z  k  n  p  r      ṯ           ẓ  g  d  ṭ  y  ṣ  ʿ  s¹  t  ḏ  ḫ

We can compare this to a known abecedary found at Tayma (Tay19) by first removing non shared characters from our sequence:

ʾ  w  ḥ  l  m     q  š  ġ  b  h  z  k  n  p  r      ṯ           ẓ  g  d  ṭ     ṣ  ʿ  s¹  t  ḏ  ḫ

Tay 19 reads:

ʾ     l  ḥ  m  q  w  š  r  b  t¹  s  k  n    ḫ  ṣ  s³  p  h  ʿ  ḍ  g  d  t² ġ

Note that the four letters ṣ, p, h and ʿ are all scratched out in the Tay 19 inscription.

Is.R 204 and Is.Mu 650 as alternative renditions of the Ugaritic alphabet?

The hypotheses discussed give us a theoretical supplementary transliteration table for Ugaritic (and possibly Safaitic) which may be used in specific circumstances.

old => new
ʿ      r
g      k
ġ      ʿ
d      ḏ
ḏ      d
r      ġ

Alternative reading of the Ugaritic alphabet in rows of 13

ʾ b k ḫ ḏ h w z ḥ ṭ y g š
l m d n ḍ s r p ṣ q ġ ṯ o
t i u s2

Is.Mu650

h ʾ l t s ʿ d ḫ ṭ m b ġ f ṯ s & ʿ n z š ṣ g q w y h ḍ k

Is.Mu650 as a boustrophedon and read from left-to-right and top-to-bottom

            k   ḍ h y w q g ṣ   š
h ʾ l t s ʿ d ḫ ṭ m b ġ f ṯ s & ʿ n z

Is.Mu650 revised reading idea to mirror altenative Ugaritic alphabetic example above

                            ḥ
            k n ḏ h y w z g ṣ š
  ʾ l t b m d ḫ ḍ s r ġ p ṯ ṭ ʿ
                            q

Which is similar to Is.R 204

l ʾ r t b ʿ d ḫ ṭ m ẓ f q ṯ s¹ g n z s² ṣ ḥ ġ w y h ḍ r k

As a final note for this illustrative section, the Beth shemesh HLḤM abecedary written in Ugaritic script also follows a directional curvature similar to that seen in Is.Mu 650 and Is.R 204 that is otherwise atypical for most attested Ugaritic script inscriptions. Further observation of OCIANA entries Is.Mu 650 and Is.R 204 indicate that they may in fact be one and the same inscription. See: https://ociana.osu.edu/inscriptions/5413 and https://ociana.osu.edu/inscriptions/5034

The Ugaritic alphabet rows of 13 theory may be further supported by the presence of a ruler of length 13 on inscription Is.Mu 650/Is.R 204 with no other geometric shapes present.

AWS 169

Our research indicates that removing the first two letters of Safaitic abecedary AWS 169 and reversing it results in something very close to Is.Mu 650/Is.R 204. This is especially true when considering that some similar looking letters may have been swapped through a process of letter change coming from writer knowledge or transmission or by the quality of the copying by the publisher of AWS 169 (https://ociana.osu.edu/inscriptions/40295).

With this theory in mind, the "original" positions of ḥ, q, ḍ and ṭ are called into question and, judging on the schematic copy, the apparent letter combination TH B G between the two main lines may partly be duplicate letters.

The inscription is given by the publisher as:

l z k h ʿ w ṭ ḍ ṣ ẓ m b n ḏ g f s² ġ ʾ m ḫ r q d t s¹ ṯ ḥ

Which reversed results in:

ḥ ṯ s¹ t d q r ḫ m ʾ ġ s² f g ḏ n b m ẓ ṣ ḍ ṭ w ʿ h k z l

However, when trying to reproduce this reading from the provided schematic copy, I get something slightly different even when trying to imitate the reading as much as possible:

  ṯ s¹ ḫ l ḥ q r t r ? f s² ġ ʾ b g ḏ n b m ẓ ṣ ḍ ṭ w ʿ h k z l
                   s¹
                   ??
                   y

If we then compare this with what we have described so far, we can reconstruct the following transliteration which, in safaitic script looks similar to Is.Mu 650/Is.R 204. We may even hypothesize that the "z l" letters look similar to a ḍ.

  ʾ b t l ḏ d m ḫ ṭ ? f ẓ ġ ṯ b g ḥ n s¹ g s² ṣ z q w y h k ḍ
                  r
                  ??
                  ʿ

Which is similar to Is.R 204:

l ʾ r t b ʿ d ḫ ṭ m ẓ f q ṯ s¹ g n z s² ṣ ḥ ġ w y h ḍ r k

Is.Mu650:

h ʾ l t s ʿ d ḫ ṭ m b ġ f ṯ s & ʿ n z š ṣ g q w y h ḍ k

Ugaritic modified. COL13:

ʾ l t b m i k d u ḫ n s2 ḏ ḍ h s w r z p ḥ ṣ ṭ q y ġ g ṯ š o

Which we can again compare with our modified Ugaritic abecedary:

ʾ b k ḫ ḏ h w z ḥ ṭ y g š
l m d n ḍ s r p ṣ q ġ ṯ o
t i u s2

Is.H 381, a Greek abecedary?

Two differing publications exist for this abecedary inscription:

One by M.C.A. Macdonald:

ʾ b {g} d z t ṯ q y m {ʿ} s² w l r ṣ ṭ h f s¹ n ḥ ḏ ġ ḍ k ḫ ẓ

And one by Ahmad Al-Jallad (2026) in which he asks whether the order isn't linked to the one known to us in the Greek alphabet:

ʾ b {g} d z t y q l m n w r ṣ ṭ h f k ḥ ḏ ġ ḍ s¹ ḫ ẓ ṯ

With the following swaps, we get a familiar 2D layout combining Tay 19 and ABG.

z <-> t
r <-> k
z <-> l
y <-> m
š <-> ẓ
ġ <-> ḏ
ḍ <-> h
ʿ <-> r
r <-> y
k <-> ḥ

Although these are too many tranformations to be taken seriously, we show the 2D result nonethless, which is:

ʾ b g
l t d
m q z
w n r
ḥ ṣ ṭ

ʿ p ḍ

k ġ ḏ
ḫ s h
š ṯ y
    ẓ

Khirbet es-Samra L.17 (KnA)

Through a handful of transformations, we can see how abecedary KnA might fir the 3 column boustrophedon arrangement:

The original publication by Knauf is the following:

l b ṯ d h w z ḥ ṭ y k m n ṣ r o p q s¹ t ġ ḍ ḏ g a ẓ s² ḫ

Note that G is written as TH and vice-versa according to the article.

l  b  ṯ
w  h  d
z  ḥ  ṭ
m  k  y
n  ṣ  r

q  p  ʿ

s¹ t  ġ
g  ḏ  ḍ
a  ẓ  š
      ḫ

With just a few swaps we also get the 2D layout

ʾ  b  g
w  h  d
ḥ  z  ṭ
m  k  y
l  n  ṣ

s¹ p  ʿ

q  t  ḍ
ṯ  ḏ  ġ
r  ẓ  š
      ḫ

Comparing character frequency and Safaitic script morphological groups

Amorite and Ugaritic are languages that are considered close. Here we attempt a crude comparison of their character frequencies and how they compare to Safaitic script morphological principle when transliterated into Safaitic.

Ugaritic character frequency is the following:

n m b l t r y d š ʿ k ṯ ʾ p w ʾi h ḥ q ḫ ṣ g s ʾu ġ z ṭ ḏ ẓ s2

Amorite character frequency is:

r n ʾ b m l k d t ś y z ḫ š p q g w ʿ ḥ ṭ ṣ s ḏ h ẓ

If we modify these slightly, we get a list of characters that, when read in Safaitic transliteration show some morphological grouping similar to that of the MISS.A 2 abecedary.

Ugaritic character frequency modified for Safaitic transliteration:

n m b l h r ḥ z s¹ ḫ k š ġ p w ʾi t y g ʿ ṣ q ṯ ʾu ʾ d ṭ ḏ ẓ s2

Amorite character frequency modified for Safaitic transliteration:

r n ʾ b m l k z h s ḥ ḏ ʿ š p g q w ḫ y ṭ ṣ ś d t ẓ

If the guiding principle of the MISS.A 2 abecedary was originally Ugaritic or Amorite character frequency then one might reconstruct the original order as:

b r m
l n ḫ h z k
ḥ s¹
s² ġ f
ʿ w g
t ṯ y ʾ ṣ
ḏ d ẓ q
ḍ ṭ

Instead of MISS.A 2 which is below:

t ṯ y ʾ ṣ
b r m
l n ḫ h z k
ḥ s¹
s² ġ f
ḍ ṭ
ʿ w g
ḏ d ẓ q

MaSA

An example usage of this grouping of letters and a reconstruction of a version of MISS.A 2 based on phoneme frequency of Amorite names reported by GELB is the interpretation of an abecedary designated as MaSA. The abecedary reads in the following manner.

l ṣ ʿ b n ẓ w g ṯ d r k s² f m h ġ q ḍ ṭ ( f ) s¹ ḥ ḏ ʾ ḫ t z y

However, the extra "f" is described by C.A. Macdonald as a form of reading aid rather than a character. Indeed the inscription is laid out in a peculiar way not uncommon for Safaitic inscriptions which is that the end of the line snakes above the first such that we get in 2D:

                           y z t ḫ ʾ ḏ ḥ   s¹ 
l ṣ ʿ b n ẓ w g ṯ d r k s² f m h ġ q ḍ ṭ ( f ) 

With just 4 character swaps and MISS.A 2 letter grouping considerations:

  • ṯ and ḏ
  • r and ġ
  • ḫ and ʿ
  • z and ṣ

And ẓ, ḏ, d and q the "extra letter" groups of MISS.A 2.

                           y ṣ t ʿ ʾ ṯ ḥ   s¹ 
l z ḫ b n ẓ w g ḏ d ġ k s² f m h r q ḍ ṭ ( f ) 

We notice that the end of the inscription shows us an array of letters "y ṣ t ʿ ʾ ṯ" similar to that of the first group of MISS.A 2 but also the extra letters "Y S. O S T DH" that appeared when we reconstructed the Tay 19 HLHM inscription from the Magrebian abjadi sequence. These letters may help explain the discrepancy between 22 and 28 letter writing systems. Is it possibly something to do with Akkadian cuneiform not having equivalent distinguishing signs for those phonemes?

Regardless, we can hypothesize that the "s¹ ḥ ṯ ʾ ʿ t ṣ y" section was intended for insertion after the initial "f" as hinted:

l z ḫ b n ẓ w g ḏ d ġ k s² f s¹ ḥ ṯ ʾ ʿ t ṣ y m h r q ḍ ṭ ( f ) 

By swapping sections "ẓ w g ḏ d" and "m h r" by help of the presence of extra characters and our MISS.A 2 reconstruction, we get:

l z ḫ b n m h r ġ k s² f s¹ ḥ ṯ ʾ ʿ t ṣ y ẓ w g ḏ d q ḍ ṭ ( f ) 

Which is comparable to our MISS.A 2 / Amorite reconstruction which reads:

b r m
l n ḫ h z k
ḥ s¹
s² ġ f
ʿ w g
t ṯ y ʾ ṣ
ḏ d ẓ q
ḍ ṭ

Where pairs of groups 3,4 and 5,6 are flipped.

WH 3598

We briefly introduced abecedary WH 3598 in a modified form on our dedicated [MISS.A 2 abecedary] page. At first glance, it's a very particular alphabetical order and reads as:

{f} t {b} z ḫ d q ḏ h n ʾ ṣ m {ẓ} s¹ g ṭ w ʿ ḥ s² r l k ṯ y ḍ ġ 

However, the relationship between it and MISS.A 2 may simply be that WH 3598 is a 2D reading of a MISS.A 2 order:

{f} t {b} z  ḫ  d  q  ḏ  h  n
ʾ   ṣ  m {ẓ} s¹ g  ṭ  w  ʿ  ḥ
s²  r  l  k  ṯ  y  ḍ  ġ

When read in columns, we get:

{f} ʾ s² t ṣ
r {b} m
l z {ẓ} k ḫ
s¹
ṯ d g y
q ṭ ḍ ḏ
w ġ
h ʿ n ḥ

The last group can be considered a clue indicating to some form of Akkadian character frequency. Indeed, if we flip:

f, y
n, ẓ
s², ṯ
g and ġ

We get:

{y} ʾ ṯ t ṣ
r {b} m
l z {n} k ḫ
s¹
s² d ġ f
q ṭ ḍ ḏ
w g
h ʿ ẓ ḥ

Our result is comparable to MISS.A 2 but with extra characters h, ʿ, ẓ, ḥ not rendered in Akkadian cuneiform:

t ṯ y ʾ ṣ
b r m
l n ḫ h z k
ḥ s¹
s² ġ f
ḍ ṭ
ʿ w g
ḏ d ẓ q

BES20 592

This abecedary is 31 characters long and reads:

ṯ rh ḅn mʿgs¹kḥm{l}ḏr
tfġʾṣl{n}ḫ ys²wẓ{h}{d/y}ṭ{ḍ}

It does contain some duplicate letters: n, h, l, m, r. These duplicates may be present due to multiple factors of which include low quality of writing by the student and uncertain readings by modern scholars.

When reversed we get:

{ḍ} ṭ {d/y} {h} ẓ w s² y ḫ {n} l ṣ ʾ ġ f t r ḏ {l} m ḥ k s¹ g ʿ m n ḅ h r ṯ

We can then write this in two lines of approximately equal length:

{ḍ} ṭ {d/y} {h} ẓ w s² y ḫ {n} l ṣ ʾ ġ f t 
r ḏ {l} m ḥ k s¹ g ʿ m n ḅ h r ṯ

And style it such that letters line up:

{ḍ} ṭ {d/y}  {h} ẓ w  s² y ḫ {n} l ṣ ʾ ġ f t 
r   ḏ {l}   m ḥ  k s¹ g  ʿ m  n  ḅ h r ṯ

Which, when transliterated into Ugaritic script gives us a regular

{𐎝} 𐎉 {𐎄/𐎊}  {𐎅} 𐎑 𐎆 𐎌 𐎊 𐎃   {𐎐} 𐎍 𐎕 𐎀 𐎙 𐎔 𐎚
𐎗  𐎏 {𐎍}   𐎎 𐎈 𐎋 𐎒  𐎂  𐎓   𐎎 𐎐   𐎁 𐎅 𐎗 𐎘

If we apply the following changes:

  • Last r -> z
  • Switch ḏ and ḥ
  • Switch s² and ẓ
  • Consider g as q
  • Consider the first m as Ugaritic 𐎓 copied directly rather than into Safaitic.

We get the following list of Ugaritic letters:

{𐎝} 𐎉 {𐎄/𐎊}  {𐎅} 𐎌 𐎆 𐎑 𐎊 𐎃   {𐎐} 𐎍 𐎕 𐎀 𐎙 𐎔 𐎚
𐎗  𐎈 {𐎍}   𐎎 𐎏 𐎋 𐎒  𐎂  𐎓   𐎎 𐎐   𐎁 𐎅 𐎇 𐎘

This arrangement, when read from top-to-bottom and left-to-right, is an inventory of Ugaritic letters arranged by letter shape. Note that 𐎘 may appear as a star on your screen but is closer to a vertical bar with a dot in the middle.

Inscriptions flagged as harder to interpret

The following inscriptions are harder to interpret due to poorer documentation or the inscription itself being harder to read because it is worn or because the direction of reading isn't entirely linear:

  • AWS 169
  • BES20 592
  • Is.H 381
  • KMD 28–31
  • KRS 1685
  • RWQ 67
  • WH 3598

No known readable pictures exist for the following abecedaries:

  • K. Es-samra L17.
  • AWS 169
  • MaSA
  • RWQ 67

Project details

For more exploratory results on the alphabetical order, see the preprint and repository.

The preprint can be found at the following address:

https://zenodo.org/records/21959747

The project repository is on GitLab:

https://gitlab.com/hugo.deploys/abjads

Next page on reviewing or contributing to the project →