The Language & Its Script

Tamazight is not a dialect of Arabic. The two belong to different branches of the same very old language family, which puts them roughly where English and Hindi sit: related far enough back that neither descends from the other, and not mutually intelligible in either direction.

That is not a contested position. It is the ordinary classification used by everyone who works on these languages, and it has been since the nineteenth century, when the family was still called Hamito-Semitic. The sections below set out what the claim rests on, what the strongest objection to it is, and why that objection does not survive contact with the evidence.

A primary branch of Afroasiatic

Berber, or Tamazight, is one of the principal branches of the Afroasiatic language family, alongside Semitic, Ancient Egyptian, Cushitic and Chadic (Greenberg 1963). Arabic sits inside the Semitic branch. Tamazight sits in a different branch of the same family.

This is a genetic classification in the linguistic sense, which means it rests on systematic correspondences in grammar rather than on geography, religion or politics (Kossmann 2012). The things that establish it are the unglamorous ones: the shape of the pronoun system, the way verbs are inflected for person, the basic vocabulary that languages almost never borrow.

Two features make the point well, because neither has an Arabic equivalent and neither is the kind of thing a language picks up from a neighbour.

The first is the way Berber builds its verbs. The system is organised around aspect rather than tense: the same root appears in distinct aorist, perfective and imperfective stems, formed by internal changes to the stem itself and by prefixed and suffixed person markers. Semitic languages also inflect verbs by internal change, which is a family resemblance, but the actual Berber system is not the Arabic one and cannot be derived from it (Kossmann 2012).

The second is the state distinction on nouns. Berber nouns appear in two forms, conventionally called the free state and the annexed state, and which one is used depends on the noun’s grammatical position, for example after certain prepositions or as the subject following its verb. Arabic has nothing of the kind. A learner coming from Arabic does not find a familiar system with unfamiliar words; they find a different machine (Kossmann 2012).

How far back the common Afroasiatic ancestor lies is a separate question, and this site will not put a number on it. Proposed dates differ by several thousand years depending on the method used, and none of that work is cited here yet. What the classification establishes is the relationship, not its date.

The strongest objection: “but it’s full of Arabic words”

This is the real argument, and it deserves a real answer rather than a denial. Northern Berber varieties do contain a great deal of Arabic: not only vocabulary but borrowed phonemes, some borrowed morphology, and contact-induced changes in syntax. Anyone who has heard Tarifit spoken in a market has heard it. Centuries of contact in a shared society do that.

The answer is that borrowing is not descent, and the two leave different fingerprints.

English took an enormous amount of vocabulary from French after 1066, to the point where much of its formal register is French in origin. English is not a Romance language. It is a Germanic language with a very large French vocabulary, and the way you can tell is that its grammatical core, the pronouns and the irregular verbs and the way sentences are built, stayed Germanic. The same test applied to Tamazight gives the same kind of answer: the pronouns, the verb system and the noun states are Berber, whatever proportion of the dictionary arrived from Arabic.

There is a scholarly literature on exactly this, and its framing settles the question before you read a page of it. Kossmann’s book-length study is titled The Arabic Influence on Northern Berber, and it works through what Arabic has done to the lexicon, the phonology, the morphology and the syntax, variety by variety (Kossmann 2013). That study is only coherent if there are two languages for influence to pass between. You cannot write four hundred pages about what one language borrowed from another and conclude they were the same language all along.

Living varieties, and how many people speak them

In Morocco the language lives as three large, mutually influencing varieties: Tarifit in the Rif, Central Atlas Tamazight in the Middle Atlas, and Tashelhit in the Souss. Modern descriptive work treats these as members of one coherent Berber branch, not as dialects of Arabic (Kossmann 2012).

Speaker numbers are harder to state honestly than they look. Morocco’s 2014 census asked which local languages people use, which is a different question from which is their mother tongue, and reported that roughly 26 to 27% of Moroccans use an Amazigh variety, with Tashelhit around 14%, Central Atlas Tamazight around 7.6 to 7.9%, and Tarifit around 4% (HCP 2014).

Those figures are disputed, and it would be dishonest to quote them without saying so. Academic estimates run considerably higher, the methodology has been publicly questioned by the head of Morocco’s own royal institute for Amazigh culture, and census self-reporting on language is sensitive to how the question is asked and to what respondents think the answer is for. The honest statement is that the number is in the millions and that the precise share is contested. It does not need to be precise to make the point, because the antiquity of a language does not depend on its current speaker count.

A script attested for two millennia, and encoded for two decades

The Amazigh alphabet, Tifinagh, descends from the Libyco-Berber consonantal script. Its use is fixed to a firm calendar date by the bilingual Punic and Libyan dedication from Dougga (Thugga), which dates itself to the tenth year of the reign of Micipsa, 139 BCE (Dougga 139 BCE). The script is attested on undated inscriptions that are very likely older; the Dougga stone is simply the one that carries its own date.

The other end of that history is just as checkable. Tifinagh was encoded in Unicode 4.1.0, as one of eight complete scripts added in that release, occupying the block U+2D30 to U+2D7F (Unicode 4.1.0). That is a mundane technical fact and a useful one: it means the script has a documented character repertoire, stable code points, and text that renders on any modern device. A writing system with a dated inscription from 139 BCE at one end and a Unicode block at the other is not the profile of an “unwritten dialect”.

Official in two states

Whatever its deep history, the present legal status of the language is a matter of public record in both countries where most of its speakers live.

Article 5 of Morocco’s 2011 Constitution declares Amazigh an official language of the State alongside Arabic, with an organic law governing how that status is implemented in teaching and public life (Morocco 2011).

Algeria followed in 2016. Article 4 of the revised constitution states that “Tamazight shall also be a national and an official language”, and commits the State to developing it “in all its linguistic varieties in use throughout the national territory”, while Article 3 keeps Arabic as the official language of the State (Algeria 2016).

Official status is not evidence of antiquity and is not offered as any. It is here because the question “is this a real language or a dialect” is often really a question about recognition, and on that narrow point the answer is written into two constitutions.

Where to go next

The dated inscription behind the script claim, and the wider material record it sits in, are in Stone, Script & Settlement. For why the population history is a separate question from the language, and what genetics can and cannot establish, see Blood & Continuity.

---

Every citation on this page was checked against the original source, with one exception that is labelled as such: the 2014 census figures come from reporting of the official release, because the primary publication could not be reached. What was checked, and what was not, is recorded in each entry in the bibliography.