The Development of the Hijazi Orthography

3.1 Loss of wawation

In the earliest phases of pre-Islamic inscriptions in the Nabataean script, triptotic names are as a rule followed by wawation, a word-final wāw at the end of a name.^[16] Many of the Nabataean inscriptions are written in Aramaic, so we only ever find Arabic names that show such wawation. But in the earliest Arabic language inscription, the En Avdat inscription (can be dated after 95 BCE and before 125 CE)^[17] such wawation (and indeed also yāʾ for the genitive and ʾalif for the accusative) represent living final case vowels in Nabataean Arabic. In the centuries that follow the case system starts to break down, thus in JSNab 17 (267 CE), both names and nouns have wawation, but they no longer inflect for case.^[18]

In the course of the 4th century CE, wawation appears to have been lost completely, at least in some varieties of Old Arabic, and it had become a completely orthographic device retained on triptotic names only, and irregularly even there.^[19] By the 6th century in Arabic inscriptions, such as the Zebed inscription (512 CE),^[20] some names are still found with wawation, while others that one would expect to have it, lack it completely. The first dated Arabic inscription that lacks any wawation on names is the Jabal Usays inscription (528 – 9 CE).^[21]

The orthographic situation that we find in the Jabal Usays inscription is eventually what we see feeding into the early Islamic period, where wawation is only consistently retained in the spelling of the triptotic name ʿamr ‏عمرو‎. Due to the lack of the name ʿAmr in the Quran, there are no examples at all in the Uthmanic text of this once widespread practice. In the corpus of late (but mostly undated) pre-Islamic inscriptions collected by Fariq al-Sahra we see a mixed situation. Some names lack wawation, while others preserve it, thus FaS 5b attests

3.2 Loss of hamzah

One of the most notable changes in orthography between the pre-Islamic inscriptions in the Nabataean/transitional Nabataean script and that of the Quranic and early Islamic writing, is that its orthography reflects a dialect that has lost the hamzah. This is surprising, as what comes to be accepted as Classical Arabic has a very conservative retention of the hamzah. Research into the Quranic rhyme and orthography, however, has made it highly plausible that – unlike most Quranic reading traditions – the language of composition of the Quranic consonantal text had indeed lost the hamzah in almost all positions.^[24] Its use of wāw and yāʾ as “seats of the hamzah” there is not just an orthographic device, but a genuine reflection of the dialect it was meant to represent. This loss of the hamzah has traditionally been identified as a typical feature of the dialects of the Hijaz.^[25]

The pre-Islamic inscriptions denoted the hamzah with the sign that is, historically, to be expected: the ʾalif. This makes it clear that the Nabataean Arabic had conservative hamzah realization.^[26] In the rare cases where a word carrying a hamzah shows up in palaeo-Arabic inscriptions, these still use the archaic ʾalif for hamzah rather than the innovative wāw or yāʾ as we see in the Islamic era orthography. For example, the month name al-muʾtamar in the PalAr1 (513 AD) as

The only apparent examples of a lost hamzah can be found in theophoric names that end with ʾĒl. Examples for this occur in the Harran inscription (568 CE)^[30] in the name

3.3 Use of ʾalif for word-internal ā

As a result of the loss of the hamzah, words that historically had the sequence aʾ shift this sequence to ā. E. g. ‏راس‎ ‘head’ or ‏تاكل‎ ‘you eat’ were originally spelled etymologically to represent raʾs and taʾkul, but came to be pronounced as rās and tākul. This phonetic shift allowed the ʾalif to be reinterpreted as a mater lectionis in word-internal position for ā rather than be a sign for the hamzah.^[35] This shift is complete in the early Islamic period, and ʾalif for ā is used frequently both in our earliest Quranic manuscripts, papyri, and inscriptions.

In the early Islamic period, many cases of word-internal ā are still spelled defectively. In what eventually becomes Classical Arabic orthography, most of these cases of defective spelling eventually come to be spelled plene, with only a few high frequency words that retain a historical defective spelling (e. g. ‏هذا‎ for hāḏā and ‏إله‎ for ʾilāh). While there appears to have been some degree of optionality of defective or plene spelling in the early Islamic period, this does not mean it was a free-for-all. A clear set of rules, dependent on the shape of the word, determines whether ā is written plene or not. Only words that fall outside the general rules that either require plene or defective spelling show some amount of optionality. Some of these rules were described first by Diem.^[36] However, his description was based on the modern Cairo Edition and therefore contains a number of inaccuracies in comparison to our earliest manuscripts. I have elsewhere described rules that better represent the orthography of the ʾalif in early manuscripts.^[37]

Plene spelling of ā is as of yet unattested in the pre-Islamic record. This does not necessarily need to mean that such a spelling practice is an innovation of the Islamic period. The inscriptions that are attested indeed lack plene spellings where they could have had them (e. g. qitāl as

3.4 Invariable feminine ending

In the earliest layer of Nabataean writing, Aramaic words with the feminine ending follow the feminine spelling convention <‐h> in the absolute, <‐t> in construct and <‐tʾ> in the definite form. However, Arabic names show a different distribution. Here, we see that the feminine ending is invariably <‐t> even in what would be an absolute position. This suggests that in early Nabataean Arabic the feminine ending was pronounced /-at/, as was the case in the Arabic of the Safaitic inscriptions. Construct phrases are of course also spelled with <‐t>, as, for instance, in the Namara inscription <mdynt šmr> /madīnat šammar/ ‘the realm of Shammar’.

In later writings in the Arabic script, we increasingly see that the feminine ending has undergone a shift of -at to -ah and thus the feminine ending is with ‏ـه‎ (e. g. FaS 2

By the time we get to the Uthmanic text, the orthography has undergone an unusual development. Not only does it use the ‏ـه‎ for all the indefinite and definite forms where it would be pronounced as -ah, it even starts to be written as such in construct, an environment where it was presumably never pronounced -ah. This orthographic development is therefore decidedly non-phonetic. This shift to this new non-phonetic spelling is not yet complete, and a good number of construct feminine forms (about 22 % of the total) in the Quran are still written with a ‏ـت‎ in construct, while many phrases occur in both spellings.^[39]

Only once Classical Arabic orthography crystallizes does the regular use of ‏ـه‎ in all positions become the standard. The Uthmanic text thus serves as a nice example of an orthographic reform in progress, although the construct feminine form in the pre-Islamic record is not attested often enough in our corpus to be sure that such a change was not already underway. The absence of a construct feminine spelled with ‏ـه‎ may just be an accidental gap.

3.5 ʾAlif al-Wiqāyah

One of the more remarkable features of the Quranic orthography is its use of the ʾalif al-wiqāyah. This is the employment of a non-phonetic word-final ʾalif after an expected word-final wāw. In Classical Arabic orthography, this is a purely morphological sign, which marks plural endings -ū and -aw. But this is not the spelling practice found in pre-Classical orthography. In the Uthmanic text as well as in early-Islamic writing the ʾalif al-wiqāyah is used for any word-final -ū or -aw, regardless of whether it is a plural verbal ending or not. Other cases of word-final wāw such as a consonant followed by -w or word-final -āw or -uww are spelled without the ʾalif al-wiqāyah in early Islamic orthography.^[40]

This orthographic device is not present in our earliest Nabataean inscriptions, where word-final -ū is simply written with wāw. This is clear in the transparently Arabic word ʾadḫalū ‘they introduced’ which is spelled <ʾdḥlw> in UJadhNab 109, dated to 455 – 56 CE.^[41] However, a recent discovery of a palaeo-Arabic inscription has now confirmed that the orthographic innovation of adding ʾalif al-wiqāyah to word-final -ū predates the Islamic period. The inscription is dated to 548 – 49 CE and contains fa-ʾaǧābū ʾamra-hū ‘so they obeyed/answered his command’ spelled

3.6 Morphologically spelled definite article

In the Islamic period, the Arabic definite article is spelled morphologically, that is, it is spelled with ‏الـ‎ regardless of whether the lām is actually pronounced or is assimilated to the following consonant. This could be explained in two ways: 1) this could be an indication that the definite article did not assimilate to the following coronal consonants in the register on which the orthography is based. This would find support in Greek transcriptions of Arabic that lack this assimilation, such as the Damascus Psalm fragment and the Petra Papyri.^[43] 2) it could be morphological spelling, which at least in the Quran appears to be the more likely explanation. There, the definite article is even written as ‏الـ‎ before words that start with l, a place where it is certainly a morphological spelling, as a word like ‏اللطيف‎ al-laṭīf or ‏الله‎ aḷḷāh would always be pronounced with a geminate and thus always have assimilation.

This is an orthographic innovation from the perspective of Nabataean Arabic, where the definite article is <ʾl-> before most coronals (presumably because the definite article did not assimilate in Nabataean Arabic) but assimilates to l, as can be seen primarily in instances where the divinities aḷḷāh and allāt are mentioned, spelled <ʾlh> and <ʾlt> with only a single lām respectively. In the Quran, such phonetic spellings occur in a couple of cases, most notably seen with the regular spelling of al-layla as ‏اليل‎ (an idiosyncrasy it shares with FaS 5b, if that word is read correctly) as well as all relative pronouns that are spelled with a single lām. But in the majority of cases, it is spelled morphologically with two lāms, e. g. ‏اللعنه‎ al-laʿnah ‘the curse’ and ‏الله‎ Aḷḷāh ‘God’.^[44]

While in a number of the late pre-Islamic inscriptions Aḷḷāh is still written with a single lām (FaS 4, FaS 7, FaS 8), the vocative Aḷḷāhumma ‏اللهم‎ with double lām has now been found in two inscriptions: FaS 4a and the ʿAbd Šams inscription^[45] found in the region of Tabuk. Moreover, the spelling ‏الله‎ is now found in two inscriptions near al-Ṭāʾif.^[46] Al-Jallad & Sidky argue that writing the definite article may be an orthographic innovation that specifically belongs to the Mecca/al-Ṭāʾif orthographic practice. This would then also be the way through which the Quran would have inherited this orthography.

3.7 Loss of BAR logogram

A typical feature of the pre-Islamic Arabic inscriptions, which betray the origins of the Nabataean script, which used to be employed primarily to write Aramaic, is the use of the logogram BAR. This logogram started off as the consonantal spelling of the Aramaic word <br> bar ‘son of’, to write the Arabic word ibn. This shape becomes more and more cursive, so that eventually the shape looks neither like Arabic ‏بن‎ nor like Arabic ‏بر‎, but it is transparently an evolution from the original Aramaic <br>. As shown by Sijpesteijn,^[47] this orthographic practice continues in the Islamic period in the administrative papyri and inscriptions, and as I point out elsewhere,^[48] also in a number of early Quranic manuscripts (e. g. DAM 01 – 29.1, Arabe 330 g; Codex Amrensis I).

Nevertheless, writing the word ibn phonetically with (ʾalif) bāʾ nūn becomes more and more dominant in the early Islamic period. In inscriptions, we probably see its first securely dated example in 43 AH / 663 – 64 CE.^[49] A little earlier than that, it probably also was part of the Uthmanic archetype. While several early manuscripts use the BR logogram, the majority of early manuscripts including, for example, the Codex Parisino-Petropolitanus (last quarter of the 7th century?) use ‏ابن‎. It is noteworthy that Quranic manuscripts that use the modern spelling always use ‏ابن‎ with the ʾalif al-waṣl, even in ‘Name son of Name’ constructions like ‏عيسى ابن مريم‎ ʿīsē ibn maryam, where strict classical Arabic orthography would require the spelling ‏بن‎.

3.8 Dating the innovations

The above seven changes are the most notable developments of the Arabic orthography that can be shown to have been more-or-less complete in the early Islamic period and are reflected in the orthography of the Uthmanic text. Some of these innovations have been attested now in pre-Islamic inscriptions whereas others are only first evidenced in the early Islamic period, so far. The table below lists these seven innovations along with their earliest attestations.

Tab. 2:

Seven innovations of the Arabic orthography and their earliest attestations

	Earliest Attestation	Example
Loss of wawation	Jabal Usays Inscription (528 CE)	ʾana ruqaym bin muʿarrif al-ʾawsī ‘I am Rqym son of Mʿrf the Awsite’
Loss of hamzah	PERF 558 (22 AH / 643 CE)	katāyiba-hū ‘squadrons’
ʾAlif for word- internal ā	PERF 558 (22 AH / 643 CE)	šāh ‘sheep’
Invariable feminine ending	PERF 558 (22 AH / 643 CE)	ḫalīfat tidrāq ‘the representative of Theodrakos’
ʾAlif al-Wiqāyah	Unpublished (548 – 49 CE)	fa-ʾaǧābū ‘so they obeyed’
Morphological definite article	ʿAbd Šams inscription (late 6th / early 7th century CE)	bi-smika ḷḷāhumma ‘in your name O God’
Loss of BAR logogram	Hisma 1 (43 AH / 663 – 64 CE) Uthmanic text (ca. 650 CE)	ʿAdiyy ibn Kaʿb ʿĪsē ibn Maryam

So far three of these appear in the 6th century:

1. Loss of wawation

2. ʾalif al-wiqāyah

3. Morphological definite article.

Three more can be seen to have become complete in our very earliest Islamic documents:

1. Loss of hamzah

2. Use of ʾalif for word internal ā

3. Invariable feminine ending.

One innovation is not yet complete in the early Islamic period (loss of BAR logogram). The first clear evidence of the replacement of the BAR logogram appears a couple of decades after the earliest Islamic attestations. But the use of the BAR logogram continues into the 2nd century AH.

The difference in attestation suggests that the Hijazi spelling reform that gives us the Islamic era orthography was not a single event in which the whole spelling system was abruptly changed. Although it should be noted that earliest attestation of an innovation of course does not mean that this represents the first employment of said innovation.^[50]

What we can exclude with certainty is that all of the typical orthographic developments that we see in the Islamic period were only innovated with the rise of Islam. But we may ask ourselves to what extent it is possible that the three innovations that we only see for the first time in the Islamic period did indeed originate at that time.

Robin believes that the use of ʾalif for ā was innovated in the Islamic period.^[51] Yet his conviction seems to be based on the mistaken notion that our earliest piece of evidence of the ʾalif used as a mater lectionis is first attested in 40 AH. But, in fact its first attestation is already in PERF 558 from 22 AH, as well as in the Uthmanic Archetype, which must have been canonized somewhere during the reign of ʿUṯmān b. ʿAffān (23 AH‒35 AH).^[52] The Uthmanic text as seen in the earliest Quranic manuscripts, and thus also the archetype, already displays a fairly complex and conventionalized set of rules of when to write and not to write the ʾalif to represent ā, which means that it already had had some time to develop. It is worth noting that the one preserved non-Uthmanic Quranic text, the Sanaa Palimpsest, likewise follows these scribal conventions.

I would therefore carefully propose that this innovation was already under way in the pre-Islamic period. This innovation was presumably localized to the Hijazi scribal school as it presupposes the typically Hijazi phonetic innovation of the loss of the hamzah. Given that this innovation is dependent on the loss of hamzah spelling, that innovation too must be placed in the pre-Islamic period.

As for the generalization of the feminine ending, I agree with Robin that this seems to be a continuation of an orthographic development that was already under way.^[53] The feminine endings that were actually pronounced as -ah, i. e., when the feminine ending was not in construct, already began to be spelled that way some centuries before Islam. The central innovation then is the rather unusual choice to apply this spelling non-phonetically to construct forms. Even in the early Islamic period, spellings with tāʾ are not at all uncommon. In the Quran, as many as 22 % of the construct feminine nouns are still spelled phonetically.^[54] It is possible that this spelling practice – perhaps a wish for a standardized spelling that enforced a single word form regardless of context^[55] – was indeed only introduced in the Islamic period, and the relative frequency of the older phonetic spelling in the early period is then explained as a not quite complete application of this reform. Hopefully, in the future more palaeo-Arabic inscriptions in the Hijaz will be discovered that will allow us to confirm or disconfirm the hypotheses about the date of the innovations of these innovations that up to now have not been attested in the pre-Islamic record.

4.1 Morphophonological spelling of suffixes

In Arabic (and even more generally Semitic) orthography, geminate consonants are not spelled distinctly from singleton consonants. Thus, nazala and nazzala are written homographically as ‏نزل‎ and can only be distinguished by context. The same is true when a verbal form ends in the same consonant as a subject pronoun suffix. Thus mit-tu^[57] ‘I died’ (Q19:23) is spelled ‏مت‎ with only a single tāʾ, and likewise ʾāman-nā ‘we believe’ (Q2:8) is spelled ‏امنا‎ with only a single nūn. This is different, however, when the pronominal suffix is a second- or third-person object suffix.^[58] If these occur right after an identical consonant, they are consistently spelled as two separate consonants. Thus, yudrik-kum ‘overtake you’ (Q4:78) is not spelled phonetically **‏يدركم‎ with a single geminate kāf but morphologically ‏يدرككم‎ with two kāfs. The same is true for yuwaǧǧih-hu ‏يوجهه‎ ‘he directs him’ (Q16:76) and yukrih-hunna ‏يكرههن‎ ‘compels them’ (Q24:33). To correctly spell words like these in such a manner requires a significant amount of morphological awareness on the side of the scribe and implies formal education in proper orthographic spelling.

It is worth noting that with the first-person pronominal suffixes -nī and nā, no such morphological spelling principle is observed. Thus, lā taftin-nī ‘don’t put me on trial’ (Q9:49) is spelled with only a single nūn ‏لا تفتنى‎. This same process is observed quite frequently even across vowels. Thus, we find examples like ‏مكنى‎ (Q18:95) which is recited makkan-nī ‘has established me’ for what in Classical Arabic would be makkana-nī, and ‏تامنا‎ (Q12:11) which is recited taʾmanʷnā, tāmanʷnā or tāmannā ‘trust us’ for what in Classical Arabic would be taʾmanu-nā.^[59] This kind of assimilation across short vowels is seemingly attested once in the pre-Islamic record for the second person plural bāraka-kum spelled ‏بركم‎ in the Rīʿ al-Zallālah inscription.^[60]

4.2 La-ʾ as ‏لاا‎

One of the orthographic practices that is rather unusual, as it creates ambiguity rather than resolving it, is the frequent appearance of the spelling of the asseverative particle la- followed by a word that starts with a hamzah with lām-ʾalif (resulting in a sequence lām-ʾalif-ʾalif). The most well-known example is no doubt the one found in modern print editions of the Quran, Q27:21 ‏لااذبحنه‎ la-ʾaḏbaḥannahū ‘I will certainly slaughter him’, but a decent number of other examples are found irregularly in early manuscripts, as well as a fair number where the readers disagree whether the ‏لا‎ should be read as la- or lā. Moreover, it is clear that this orthographic practice continued also in early Christian Arabic.^[61] It should be noted that this specific orthographic practice needs to likewise be learned by the scribe to do correctly. The obvious spelling would simply be with a single ʾalif, and doing so actually helps the reader as in such cases the la- cannot be mistaken for the negative particle.^[62]

Due to the frequent appearance in the corpus of late pre-Islamic inscriptions of the formula la-ʾūṣikum ‘let me certainly urge you’, likewise spelled several times with two ʾalifs, we now have clear evidence that this unusual orthographic practice had already developed in pre-Islamic times. This spelling occurs in three inscriptions: FaS 4, FaS 5b, FaS 8

4.3 Consonantal dotting

Robin suggests that the consonantal dotting as we see it in the Islamic period is an innovation of the early Islamic Medinan polity.^[63] But there are certainly some caveats to be placed with this conclusion. While it is true that as yet, not a single pre-Islamic Arabic inscription has been found that has consonantal dotting, this can hardly be considered compelling evidence of its absence. Of the thousands of early Islamic inscriptions that have been documented, only an extremely small percentage uses any consonantal dotting at all – significantly less even than contemporary papyri. Considering the much smaller corpus of pre-Islamic inscriptions, the absence of consonantal dotting in the pre-Islamic period could be entirely coincidental. It is a bizarre stroke of luck that the very earliest Islamic Arabic inscription, the Zuhayr inscription (24 AH / 644‒45 CE), just so happens to have consonantal dots.

The evidence so far remains inconclusive, but I can think of some arguments in favor and against the idea that the consonantal dotting was introduced in the early Islamic period. In favor is the fact that the use of consonantal dotting was not yet entirely standardized, something we could perhaps have expected had the dotting been in use for a long time already. Nevertheless, the points of disagreement are quite sparse. Most words are distinguished with consonantal dotting as they are used to today, save for the qāf/fāʾ distinction, which are distinguished with a dot above and below. There is, however, some disagreement between different documents which of the two gets which placement. Some manuscripts follow the Maghrebi practice of putting a dot above for the qāf and a dot below for the fāʾ, while other (most notably manuscripts in the C.I style) have the reverse situation with a dot below the qāf and a dot above the fāʾ.^[64] Likewise, manuscripts in the LH/A style seem to systematically use an undotted denticle to denote yāʾ whereas bāʾ, tāʾ, ṯāʾ and nūn are consistently dotted, whereas the consistent dotting with two dots below only developed later.^[65] Both of these behaviors seem to suggest that dotting was not yet completely standardized. The lack of standardization of dotting might lead us to the conclusion that the system had not been in use for a long time yet.

What speaks in favor of the system being more archaic, however, is that its usage is highly idiosyncratic, and scribes almost appear to go out of their way in order to dot exactly the words that least need it. There is a pervasive pattern both in Quranic manuscripts as well as early papyri to dot word-final nūn, one of the few letters that is completely unambiguous in the Arabic script. Adam Bursi shows that there are a number of recurring practices of dotting that seem to point to a developed and often idiosyncratic scribal practice.^[66] Such idiosyncrasies give the impression of a system that has already had quite some time to develop. As it stands now, it is impossible to be sure whether the dots really were invented in the Islamic period, the absence of dots in pre-Islamic inscriptions can certainly not be used as compelling evidence in favor of that.

4.4 Other complexities

There remain several other orthographic complexities that show up in early Quranic manuscripts which are sufficiently specific and difficult to understand, that it seems unlikely that they are recent innovations. For example, we find that if a word starts with a hamzah and is followed by a yāʾ with an intervening short or long vowel, and such words are preceded by the prepositions bi- or li- the yāʾ is spelled twice. Thus, we find forms like ‏لاييتنا‎ for li-ʾāyātinā, ‏فبايى‎ for fa-bi-ayyi, ‏باييم‎ for bi-ʾayyāmi etc.^[67] Such orthographic practices have not yet been attested in the pre-Islamic record but are quite complex while not resolving any ambiguities. Such an idiosyncrasy suggests a developed scribal practice.

Another case in what appears to be a related practice is a mixed-historical spelling of words that historically contained a hamzah that was adjacent to i, ī or y.^[68] These are irregularly spelled with ʾalif followed by yāʾ word-internally. The most well-known case of this is ‏مايه‎ for miʾah ‘100’, which is an orthographic practice that continues into the Islamic period, but several other cases occur in early Quranic manuscripts. The ʾalif no doubt points to a historical memory of the presence of a hamzah, but what is surprising is that it always occurs in the order ʾalif-yāʾ, even if logically the yāʾ precedes the hamzah, such as in ǧīʾa ‘it was brought’ spelled ‏جاى‎,^[69]šayʾ ‘something’ spelled ‏شاى‎.