العربية

The Arabic script descends from the Nabataean variety of Aramaic, which descends in turn from Imperial Aramaic and, behind that, from Phoenician. Theodor Nöldeke made the argument in 1865 and it has held; the derivation of the letterforms is now close to consensus, and where scholars still differ is over the cultural route rather than the graphic ancestry, with a Syriac-influence thesis alive in the background and the Lakhmid court at al-Hira as the usual candidate for the crossing point.

The Nabataeans themselves are the interesting part. They appear to have spoken Arabic and written Aramaic, which is a mismatch a modern reader recognises immediately, and from about the third century they began using their Aramaic letters to write their own language. The script does not so much begin as drift. Current epigraphy handles this by refusing a single starting date: the CNRS DiCoNab project distinguishes Nabataean proper, then Nabataeo-Arabic, transitional and generally no later than the mid-fifth century, then Palaeo-Arabic, already recognisably Arabic; the last two together are "Developing Arabic", spanning the late third century to the early seventh.

Popular accounts run these together and they are separate systems doing separate jobs. إعجام, i'jam, is the pointing that tells apart letters sharing a skeleton, so that one shape becomes ba, ta or tha. تشكيل, tashkil, the harakat, marks the short vowels, and later the sukun, the shadda, the madda and the hamza.

The tradition dates i'jam to a reform under al-Hajjaj ibn Yusuf, governor of Iraq, who died in 714. The inscriptions do not co-operate. The Raqush stone of 267 CE already carries dots on dal, shin and ra; the fourth-century Jabal Ramm text has them on jim, ya and nun; a Greek and Arabic papyrus of 642 to 643 is full of them. Beatrice Gruendler's conclusion is that the system had finished developing in the first half of the first Islamic century, applied by habit rather than by rule. What al-Hajjaj's administration plausibly did was mandate the practice in official and Qur'anic copies, which is a different thing from inventing it. There are, in any case, three mutually inconsistent versions of who he asked, and the existence of three is the argument against relying on any.

The vowel marks have a better story and a worse provenance. Abu al-Aswad al-Du'ali, who died in 688 or 689, is credited with a system of coloured dots, and the scheme reported in the Fihrist is charmingly physical: a dot above the letter for fatha, the open mouth; beside it for damma, the mouth rounded; below for kasra, the lips parted.

The manuscripts do carry coloured dots, and the colours vary by region. Al-Dani, writing in eleventh-century al-Andalus, records Medina and the Maghrib using red for the vowels and yellow specifically for hamza, and Iraq using red for everything. An Andalusi Qur'an of 842 adds green, for the initial alif al-wasl.

What cannot be established is the attribution. Alain George's assessment is that the sources naming Abu al-Aswad and the others are not reliable enough to be treated as historical information, and that the accounts show a process of amplification and harmonisation, growing backwards from the practice to a founder. The red dots are real and datable to the Umayyad period. The man is a story.

The marks in use today are credited to al-Khalil ibn Ahmad al-Farahidi, who died around 786, and here the attribution matters less than the design. He replaced the dots with miniature letters: a small waw above for damma, a slanted alif for fatha, a fragment of ya below for kasra, a small shin for the shadda. The gain is that dots stopped doing two jobs at once. Before al-Khalil, a mark on a page might be identifying a consonant or supplying a vowel, and which one it was could come down to the colour of the ink.

By the tenth century the script had its machinery and acquired a geometry. Ibn Muqla, three times vizier in Baghdad and eventually mutilated and imprisoned by the man who displaced him, is credited with الخط المنسوب, al-khatt al-mansub, proportioned writing, in which every letterform is specified as a ratio against three measures. The first is the nuqta, which is not a round dot but the rhombus left by pressing the obliquely cut reed nib once against the page, so the unit scales with the pen. The second is the height of the alif, counted in those rhombi. The third is a circle whose diameter is that height, against which the curved letters are set.

It is an elegant system, it is recoverable in outline, and not one authenticated line in Ibn Muqla's hand survives. Everything known about it comes through later writers. Sheila Blair's warning covers the whole first two-thirds of the Islamic period: there are no surviving examples of a script identified by a named sample. The proportioned script as usually illustrated is a reconstruction from texts.

The first verifiable anchor is a generation later. The Qur'an copied by Ibn al-Bawwab at Baghdad in 391 AH, 1000 or 1001 CE, is in the Chester Beatty Library in Dublin, and it exists. It is also, on the usual account, the earliest surviving Qur'an written on paper, which is not a coincidence; the proportioned cursive hands are a response to a smoother surface. There is a twelfth-century anecdote in which Ibn al-Bawwab forges Ibn Muqla's hand so exactly that nobody can tell, and the anecdote is more useful than it looks: it records the tradition itself admitting that "a hand by Ibn Muqla" was already an uncertain category within a century of his death.

The classical canon is the six pens, الأقلام الستة, al-aqlam al-sitta: naskh, thuluth, muhaqqaq, rayhani, tawqi' and riqa'. They are a system rather than a list, six hands differentiated by proportion and curvature on a shared metrical basis, which is why they read as a family.

Kufic, the angular hand of the early Qur'ans, is the one to be careful with. Kufa was founded in 638, and the script that later took its name is attested well before that; the name is a modern scholarly convenience applied backwards, and François Déroche built an alternative classification of early Qur'anic hands specifically to get out from under it. Its most striking descendant is square Kufic, developed in twelfth-century Iraq, where the brick and the tile become the unit and the wall is written at building scale.

Naskh, from the root meaning to copy, is the plainer hand that lost the prestige contest and won everything else. It was the everyday administrative script while Kufic held the Qur'anic page; it became the basis of Arabic printing; and it is the default for Arabic typefaces in operating systems, which means it is the hand most Arabic is read in today.

Two names get conflated and should not be. Riqa', one of the six pens, is a small classical chancery hand. Ruq'ah is a modern Ottoman handwriting style, regularised in the nineteenth century by Mumtaz Efendi and Mustafa Izzet Efendi out of forms already in daily use, made of short straight strokes and simple curves, and it is what most handwriting and most street signage actually look like.

Nasta'liq belongs to the Persian page rather than the Arabic one. The name is a contraction of naskh-e ta'liq, hanging naskh, and it became the standard hand for Persian literature and later for Urdu; Elaine Wright's work indicates it descends from naskh alone and not from a marriage of naskh and ta'liq, so the name misdescribes its parentage. It is not a standard hand for Arabic, and the reason usually given is structural: the definite article al- gives Arabic a verticality that nasta'liq's long horizontals fight.

Diwani, the Ottoman chancery hand of the sixteenth century, is worth a sentence for its logic rather than its shape. It is dense, tightly interlocked, and rises from right to left, and the standard explanation is that its difficulty served confidentiality and made forgery hard. That explanation is universally repeated and it does read like received wisdom, so take it as plausible rather than demonstrated.

Strip every dot and every vowel mark from a page of Arabic and what is left is the رسم, the rasm, the consonantal skeleton. It is the oldest layer of the script and the layer in which the Qur'anic text was fixed; the canonical readings differ from each other, very largely, within a single rasm.

The counting is the point. Twenty-eight letters resolve to eighteen skeletal shapes, excluding the lam-alif ligature. In isolated and final position all eighteen are distinct. Connected on the left, in initial and medial position, the count falls to fifteen, because nun and ya collapse into the ba, ta and tha shape and fa collapses into qaf.

Two mechanical consequences are worth knowing before touching the keyboard. Arabic is obligatorily cursive: it joins in print as well as by hand, which is unusual among the world's writing systems. And six letters, ا د ذ ر ز و, join only to the right, so they have two forms each and they break a word open. A written word in Arabic is a run of connected clusters with gaps inside it, and the gaps are not word boundaries. That is a real obstacle for a new reader, and no amount of practice with the keyboard removes it, because it is a fact about the script rather than about the typing.

Ferguson borrowed the term from William Marçais, who had written on la diglossie arabe in 1930, and gave it a definition and nine rubrics: function, prestige, literary heritage, acquisition, standardisation, stability, grammar, lexicon and phonology. His four cases were Arabic, Modern Greek, Swiss German and Haiti, and Arabic is the one that carries the argument, because it is the only one where the high variety is a codified older stage of the same language with a sacred text attached to it.

Two of his observations survive better than the model. The first is that speakers' reports do not match their behaviour: educated speakers told him they never used the low variety, and observation showed them using it constantly, and when he asked how they spoke to their children they conceded the children would not understand the other one. The second is that he did not, in fact, claim a clean binary. He describes relatively uncodified, unstable, intermediate forms and cites the Arabic term for them, اللغة الوسطى, al-lugha al-wusta, the middle language.

He also predicted that Arabic would break into several regional standards. Greek went the way he expected, and Dhimotiki was made official in 1976. Arabic has not.

The refinements have mostly been attempts to describe the middle. Alan Kaye's point in 1972 has the sharpest teeth: the colloquials are well-defined systems because they have native speakers whose intuitions settle what is grammatical, and Modern Standard Arabic is ill-defined because it has none. That is why questions like "is this sentence grammatical in MSA" are harder than they look.

El-Said Badawi's five levels for Egypt, published in 1973, run from heritage fusha through contemporary fusha, the colloquial of the cultured, the colloquial of the basically educated, and the colloquial of the illiterate. The detail usually lost is that he drew them as overlapping bands, not steps, with the proportion of standard features falling and vernacular features rising down the diagram, and loanwords peaking in the middle rather than at either end.

Recent computational work gives the continuum some hard edges. The ALDi project scores sentences on a scale from pure standard to fully dialectal, with a corpus of 127,835 annotated sentences. The most informative number in it is the inter-annotator agreement, a Fleiss' kappa of 0.44; even trained annotators do not agree crisply on where a sentence sits, which is what a continuum looks like from the inside.

One distinction to keep. Educated Spoken Arabic is the flexible register of the conference and the current-affairs interview, and Gunvor Mejdell's summary of it is that no stability exists in it even at the level of the individual speaker. It is a zone of negotiated mixing rather than a variety with a grammar. Middle Arabic is a different thing altogether, a philological term for the mixed written register of papyri, Christian and Jewish Arabic texts and popular literature; Kees Versteegh warns explicitly against reading it as a stage on the Old, Middle and Modern English model.

There is no agreed boundary between Classical Arabic and Modern Standard Arabic, and that is the first thing to say rather than the last. The morphology and syntax are essentially continuous. The substantial difference is lexical: MSA carries a modern vocabulary built by coining from existing roots, سيارة for a car from the root meaning to travel; by extending old words; and by adapting loanwords. Beyond that the differences are tendencies, not rules, mostly toward shorter sentences and a plainer list construction.

It emerged from the nahda, and specifically from journalism and translation rather than from any committee. Napoleon's press reached Egypt in 1798 and left with the French in 1801; the Bulaq press followed in Cairo; al-Waqa'i' al-Misriyya, the first Arabic printed newspaper, appeared in 1828; Rifa'a al-Tahtawi founded the school of languages in 1836; al-Ahram began in 1875. The academies came afterwards, to regulate what was already happening: Damascus in 1918, Cairo by royal decree at the end of 1932. Their coinages have frequently failed to displace the loanwords already in use, which is the usual fate of language academies.

The terminology is worth getting right, because English gets it wrong. Arabic speakers do not ordinarily treat Classical and Modern Standard as two languages; both are العربية الفصحى, al-'arabiyya al-fusha, and where the distinction is needed the terms are fusha al-'asr, contemporary fusha, and fusha al-turath, heritage fusha. Those are the same two labels sitting at the top of Badawi's five levels. "Modern Standard Arabic" is a coinage of Western Arabists with no exact everyday Arabic equivalent.

Ethnologue records roughly 335 million users of Standard Arabic and zero first-language speakers. The zero is not an error. A child acquires a regional variety at home and meets fusha at school, in the Qur'an class and on the news, and learns it the way a second language is learned.

The psycholinguistics supports this better than most claims in this area. Saiegh-Haddad and Haj tested 120 children from kindergarten to Year 6 on whether they could judge pronunciations correct across three kinds of word: identical in dialect and standard, partly overlapping, and standard-only. Phonological distance predicted accuracy at every grade, and there was no improvement at all between kindergarten and Year 1. Asadi and Kasperski found the components of reading swap places between the two varieties: in the spoken variety, listening comprehension contributes more than decoding; in the literary variety, decoding dominates. That is the shape you would expect if the standard were being processed as a learned second system.

What does not follow, and is asserted constantly, is that diglossia damages literacy. Lama Bergstrand Othman's 2025 review makes the comparison nobody makes and sets minimum-reading-proficiency figures side by side: Brazil, which is not diglossic, at 43.5 per cent in 2019, against Egypt at 44.57 and Jordan at 46.64 in 2021, and Qatar at 80.39 and the UAE at 80.94 in the same year. The spread among the diglossic countries is much larger than the gap between the lowest of them and the non-diglossic comparator, which points at schooling, funding and household conditions rather than at the language arrangement. Othman also notes that the early-grade effects shrink, and sometimes disappear, by late primary (figures as reported by Othman from UNESCO).

Classification here is conventional rather than principled. SIL treats about thirty varieties as separate languages; the Library of Congress treats them all as dialects of one. Neither position is neutral and neither is wrong.

Two dimensions organise the variation. The obvious one is geography, running west to east. The less obvious one is the split between Bedouin-type and sedentary-type varieties, which cuts across geography and is a historical artefact of which populations settled where and when. Bedouin-type varieties keep the interdentals, use g for qaf, and preserve the feminine plural; sedentary-type varieties lose or shift the interdentals, use a glottal stop or q, and innovate a new indicative marker with a preverbal particle. The classification is historical, though, and modern urban speech mixes freely across it.

This is where the gap between what is repeated and what is known is widest, so it is worth stating the gap first. Trentman and Shiri open their 2020 study by observing that despite a plethora of beliefs about Arabic dialect intelligibility, empirical research remains extremely limited. The percentage tables that circulate on language-learning sites are not traceable to any study.

One test has been run properly. Čéplö and colleagues, in 2016, put Maltese, Libyan and Tunisian speakers through a three-part audio protocol, 160 words, 60 sentences and two comprehension passages, with 24 respondents at each of three sites and no prior exposure between them. Tunisian and Libyan listeners understood each other at roughly 73 to 80 per cent. Maltese listeners understood about 30 per cent of both. Tunisian and Libyan listeners understood about 40 per cent of Maltese, which makes the asymmetry itself a finding.

What that study does not do is the comparison everyone actually asks about, which is Moroccan against Gulf, or Moroccan against Egyptian. As far as I can establish, nobody has run it. The honest position is that Maghrebi and Mashriqi speakers reportedly struggle with each other, that this is close to universally reported, and that it has not been measured the way the central Maghreb has (a search that finds nothing is not proof of absence).

Trentman and Shiri's own result is the surprising one. Proficiency in a familiar dialect, or in fusha, did not predict comprehension of an unfamiliar one. The assumption that a strong standard base carries a listener across the map does not survive testing.

On the Egyptian question, the exposure is real: a century of Egyptian cinema, music and broadcasting, and Egyptian teachers staffing school systems across the Gulf and parts of North Africa. But the best evidence is behavioural rather than comprehension-based. Sonia S'hiri found that Tunisian speakers converge unilaterally toward Middle Eastern speech in mixed encounters, and argued that this reflects the relations between the communities rather than communicative necessity. That is worth sitting with, because one-way accommodation produces the asymmetry as much as it reflects it. If Tunisians always adjust and Egyptians never do, Egyptians never get the practice, and the incapacity becomes self-sustaining.

Arabizi is what happens when a script meets a system that cannot render it. Latin letters plus digits chosen for their resemblance to Arabic letters: 3 for 'ayn, 7 for ha, 2 for hamza, 5 for kha. It emerged with SMS and early web forums, and the spelling conventions split along the colonial language of literacy, so Morocco and Lebanon write on French lines with ch and ou, and Egypt and the Gulf on English lines with sh and oo.

The detail worth noticing is that Arabizi is written from the sound rather than from the etymology, so it records the dialect where the Arabic script conceals it. A Cairene writes 2alb for "heart" and a Gulf writer writes galb; the standard spelling, قلب, is the one that hides which of them is speaking.

The larger shift is not romanisation, though, but dialect written in Arabic script, which has expanded enormously since the mid-2000s. The measurable proxy is that identifying a writer's province from a tweet is now a tractable machine-learning task, because there is enough naturally written dialect to train on; one shared task covers a hundred provinces across twenty-one countries.

And it is not new. Muhammad Taymur was writing in Egyptian colloquial between 1918 and 1928; Naguib Mahfouz, Tawfiq al-Hakim and Yusuf Idris each negotiated the same problem, usually by narrating in fusha and letting the dialogue move toward the vernacular. In Morocco, دارجة, darija, now appears in newspapers, advertising, fiction and literary translation, including novels written entirely in it. The internet changed the volume and the domains; it did not invent the practice.

Whether this amounts to the arrangement breaking down is an open question and should be left as one. Lotfi Sayahi describes a gradual encroachment of the vernacular on standard domains, with electronic communication driving it. Against that, no Arab state has adopted a vernacular as its written standard, no vernacular has a codified orthography with institutional backing, and fusha is still the language of the law, the sermon, the news bulletin and the classroom. What has changed is that a large informal written register now exists which Ferguson's model had no slot for, because in 1959 writing was expensive and gatekept, and now it is neither.

Maltese is the test case for what would have to happen for a dialect to become a language, and it is instructive that the answer is not the one people expect.

It descends from the Arabic of Muslim Sicily, arrived in Malta with settlers in the early eleventh century, and was cut off by the Norman conquest of 1091 and the expulsions completed by 1249; it has developed on its own for roughly eight hundred years. It is written in Latin script, and its dictionary is about a third Arabic and half Sicilian and Italian, though that count understates the Arabic, because it is a dictionary count rather than a count of running text and the high-frequency core and the whole grammatical system are Semitic.

Neither the script nor the vocabulary is the argument, though. The argument is that Maltese stands in no diglossic relation to fusha. Every Arabic vernacular has Standard Arabic above it as its written form, its liturgical register and its prestige norm. Maltese has its own standard, its own orthography, formalised in 1924, its own literature and its own official status, and Standard Arabic plays no part in the speech community at all. That is what makes it a language and the vernaculars dialects, notwithstanding that Maltese is probably structurally closer to Tunisian than Tunisian is to Iraqi.

It is also the only standardised Semitic language written exclusively in Latin script, and the only Semitic official language of the European Union.