When you go to Wikipedia – which is something I’m sure the vast majority of readers will at one point in their lives have done – and look at pages related to Semitic languages, you’ll run into the claim that two current languages spoken in Northern Yemen and Southwestern Saudi Arabia, namely Raziḥit and Faifi, rather than being varieties of Arabic, are in fact last living descendants of the Old South Arabian languages.
Just a quick reminder: the Semitic languages are a language family that are generally grouped into various branches and sub-branches: firstly, there’s East Semitic – represented by Akkadian – and West Semitic, which is pretty much everything else. The subclassification of West Semitic is a bit more controversial, but the majority of scholars working on Semitic agree that West Semitic splits up into two branches: Central Semitic and Ethiopic. The Modern South Arabian langauges, spoken in eastern Yemen and western Oman (as well as on the island of Soqotra) are considered to have split off from West Semitic directly, although much of their history is still unclear.
Now, within Central Semitic we have another branch, Northwest Semitic – represented most famously by Hebrew, as well as Aramaic and Ugaritic. Arabic constitutes a branch on its own, as do the Ancient (or Epigraphic) South Arabian languages. There are many good reasons to adhere to this classification (which I won’t go into right now) but the most important thing here is that (early) Arabic and the Ancient South Arabian languages were cousins, kind of like English and Swedish; or Russian and Polish.
Why is this all important? Well, most people who are familiar with this subject are of the opinion that the Ancient South Arabian languages, which were in use in South Arabia from about the end of the 2nd millennium BCE died out somewhere between the 7th and 10th centuries CE (fun fact, this is one of the subjects I discuss in my upcoming dissertation). There are, however, also some people who claim that a a few remnants of these languages can still be found, in the areas mentioned above: the ʿAsīr and Ǧabal Rāziḥ regions in Southwestern Saudi-Arabia and North Yemen.
The argument is as follows: these linguistic varieties display a number of features – phonological, morphological, grammatical – that are not derived from Arabic. Therefor, these linguistic varieties represent a kind of Semitic that is not Arabic, most likely Ancient South Arabian. This argument is phrased relatively lightly in Watson et al’s 2005 article “The language of Jabal Rāziḥ: Arabic or something else”, as they note: “[a]t this stage, we leave open the question as to whether Rāziḥī is a dialect of Arabic or better regarded as another language with common features derived from a common Semitic source and others adopted from Arabic”. A similar argument comes from Alfaife’s 2018 master’s thesis: “[b]ased on this information, it could be that Faifi is a surviving Sabaic dialect which is still spoken today”.
How does this claim hold up when looking at the evidence? Let’s take a look:
There are a number of phonological characteristics in both Rāziḥīt and Faifi that set them apart: first of all we have the realization of the sibilant s2 as a lateral fricative [ɬ]. Similarly, the pronunciation of *ḍ in Rāziḥīt – depending on the whether the word is inherited or borrowed – can either be a voiceless lateral affricate [t͡ɬ]. Moreover Behnstedt (1989) mentions the pronunciation of the so-called emphatic ṣ as an affricate /st/, possibly metathesized from an original */ts/. Finally, both Rāziḥīt also exhibits assimilation of the nasal /n/ to any following consonant, which occurs with both nouns (e.g. ssān, “man”; ssāna, “woman”) and verbs (e.g. past našar vs present yuššur, “to go out in the evening”). (Watson et al. 2006)
With the exception of the last, all of these features are pretty archaic: s2 is reconstructed for Proto-Semitic as a lateral and is pronounced as such in the Modern South Arabian languages, such as Mehri and Soqotri. Furthermore /ḍ/, which is a glottalized voiced dental stop in Classical Arabic, was originally pronounced as a lateral as well, as many early loanwords from Arabic in languages such as Spanish and comments from early grammarians indicate.
And it also seems that /ṣ/, which is pronounced as a glottalized voiceless fricative (/sˤ/) in most Arabic dialects, was originally an affricate /ts/, much like in Hebrew. The final feature, the assimilation of /n/ to all following consonants is something that is sporadically attested in some Arabic dialects, but only in the case of a few lexical items. It is also inconsistently attested in Safaitic, but it really appears all over the place in Late Sabaic. (Stein 2011, Al-Jallad & Jaworska 2019)
So can we rely on phonology to prove whether or not Rāziḥīt and Faifi descend from the Ancient South Arabian languages? Apart from the apparent isogloss of consistent assimilation of /n/ to following consonants – which only occurs in Rāziḥīt, and not in Faifi – there’s no immediate reason to an Ancient South Arabian antecedant, rather than a dialect of Arabic with typically archaic features. But perhaps there is more to be learned by looking at Rāziḥīt and Faifi morphology:
2. Morphology: the definite article
Both Rāziḥīt and Faifi exhibit some rather interesting morphological features. An example of this is the realization of the feminine ending in nouns and adjectives, which in Rāziḥīt is distributed as either -Vh or -īt, depending on a number of factors which I don’t want to get into now. In Faifi, such a system does not exist at all. However, as Watson (2014) pointed out that the realization of the feminine ending as –t is something that occurs throughout the South Arabian linguistic area, taking examples from Ethiopic (Amharic) and Modern South Arabian languages (Mehri and Shehri).
Moreover, both Rāziḥit and Faifi exhibit variations of the prefixed definite article, which is typical in Arabic but not present in Ancient South Arabian, where definiteness was marked on the noun as a suffix. Throughout Yemen the definite article occurs as Vm-, in the case of Faifi this is /im-/ (e.g. im-laḥža, “the window”); in Rāziḥīt the definite article is iC-, with full assimilation to the following consonant (e.g. ib-bēth). In the light of the presence of an unassimilating definite article /in-/ in the area directly east of Ǧabal Rāziḥ, Watson (2006) suggested an original article */in-/ for Rāziḥit as well. Very recently, Al-Jallad demonstrated that a North Arabian inscription predating the 4th century CE contains the /Vm/-article. On this basis he suggested that despite the /Vm/-article being one of the typical isoglosses of Yemeni Arabic, it originated in North Arabia (Al-Jallad fc.)
3. Morphology: 1st and 2nd person past tense
Another feature of Rāziḥīt is the paradigm of the 1st and 2nd person past tense ending. In most dialects of Arabic (as well as in Classical Arabic), the verbal paradigm of the 1st and 2nd person past tense is marked by a suffixed /-Vt/. However, many Yemeni Arabic dialects have a paradigm in which this marker is /-Vk/ instead. So in Rāziḥīt you’d get sarḥuk, “I went”; sarḥik, “you (m.) went” and sarḥič, “you (f.) went”.
Interestingly, Faifi is one of the dialects of the region that has a paradigm with /-Vt/ instead, which is typical for Arabic dialects of the Yemeni Tihāma (the plains between the Red Sea and the Sarāt mountains). Faifi does display some interesting conservative features when compared to Classical Arabic: in his MA thesis, Saleem Muhammed al-Faife gives an example of this paradigm (Al-Faife 2019). As you can see, the system is phonologically rather close to that of Classical Arabic:
|Classical Arabic||Faifi||English translation|
|ḍarab-tu-hā||ʿagar-t-ha||“I hit her”|
|ḍarab-ta-hā||ʿagar-tə-ha||“You (m.) hit her”|
|ḍarab-ti-hā||ʿagar-ti-ha||“You (f.) hit her”|
What is important to note here is that the t-paradigm is characteristic of Northwest Semitic and Arabic, not for Ancient South Arabian. In ASA, like in Rāziḥīt and many other Yemeni Arabic dialects, the 1st and 2nd person past tense verbal conjugation was marked by -/Vk/. This is something it shared with the Ethiopic languages as well as (historically) with the Modern South Arabian languages (Watson 2014).
4. Morphology: the relative pronoun
Finally I’d like to discuss one of my favorite features: the relative pronoun. In Classical Arabic, the relative pronoun is inflected for number and gender (and in the case of the dual also for case). While I remember writing in my MA thesis that there are no spoken dialects of Arabic that have a relative pronoun that is inflected, but again, there are a number of exceptions that occur in Southwest Arabic. Rāziḥīt distinguishes four relative pronouns: ḏa (masculine singular), ta (feminine singular), wulā (human plural) and ma (non-human plural). Faifi distinguishes three: ḏi (masculine), tī (feminine), and ʾuḏi (plural). This is noteworthy, as the vast majority of Arabic dialects have an uninflected relative pronoun ʾillī, with ʾəldi and ʾəddi attestd as well, which is most likely eventually related to Classical Arabic ʾallaḏī (Rabin 1951).
The situation in ASA is somewhat more complicated: evidence from Sabaic indicates the presence of a paradigm for the relative pronoun, which was asymmetrically inflected for gender, number, and case (only clearly attested in the plural), i.e. ḏ– (m.); ḏt- (f.); ʾlw (m.pl.nom.); ʾly (m.pl.acc/gen); ʾlt (f.pl). Apart from this, there is also evidence for an invariable relative pronoun ḏ-, occuring with any antecedent (Stein 2011)
Conclusion: are Faifi and Razihit surviving descendants of Ancient South Arabian?
When we look at what we know about the phonology and morphology of Ancient South Arabian – more specifically, Sabaic (and even more specifically, Late Sabaic), there is very little concrete evidence that either Rāziḥīt or Faifi can be considered descendants of the Ancient South Arabian languages. In my opinion, it is much more likely that both Rāziḥīt and Faifi are local variants of Arabic that are on the one hand extremely archaic while also having underwent some substratal influence of local pre-Islamic languages. This is clear, amongst under things, by the amount of Sabaic lexical items in Yemeni Arabic in general.
The phonological evidence is very inconclusive: much has been written about the fate of the sibilants in Late Sabaic, but there is definitely something weird going on when looking at Sabaic-Arabic language contact in this area (which is something I’d like to revisit in a later blog post). Phonology can also change considerably over short amounts of time, and is something that spreads relatively easily within a single linguistic area.
The morphological features discussed here are certainly of some interest: the definite article in both Rāziḥīt and Faifi are typical for the Southwest Arabia as a whole, but as the epigraphic evidence indicates, is more likely to be inherited from Early Arabic. Their different realizations in Rāziḥīt and Faifi is more likely to be a natural development rather than substrate influence.
In the case of the verbal paradigm of the 1st and 2nd person, Faifi displays an archaic system – but one that is wholly Arabic. In the case of Rāziḥīt, the case for substrate influence can be made much more strongly, as even the borrowing of an entire verbal paradigm is not unprecedented. 
Lastly, although there are some superficial similarities between the relative pronouns of Sabaic and Faifi, it is hard to see how Faifi ʾuḏi could derive from a Sabaic source. Again, it seems more plausible that this paradigm in Faifi goes back to Early Arabic. The system of Rāziḥīt on the other hand seems more likely to be related to that of Sabaic, although the non-human plural ma appears to be Arabic in origin.
I do not doubt that there has been a good deal of contact between speakers of Ancient South Arabian languages and the varieties of Arabic that were present in Yemen.  However, purely based on the phonology and morphology of Faifi and Rāziḥīt it does not seem plausible to suggest that they descend from the Ancient South Arabian languages.
Al-Jallad, Ahmad, and Karolina Jaworska. 2019. A Dictionary of the Safaitic Inscriptions. Leiden ; Boston: Brill.
Al-Jallad, A. forthcoming. “The History of the Am- Definite Article – South Arabian or Arabic.”
Alfaife, Saleem Mohammed. 2018. “A Grammar of Faifi.” M.A. Thesis, California State University, California.
Asiri, Yahya. 2008. “Relative Clauses in the Dialect of Rijal Alma’ (South-West Saudi Arabia).” Proceedings of the Seminar for Arabian Studies 38:71–74.
Behnstedt, P. 2016. Dialect Atlas of North Yemen. Vol. 114. Leiden: Brill.
Rabin, C. 1951. Ancient West Arabian. London: Taylor’s Foreign Press.
Stein, P. 2011. “Ancient South Arabian.” Pp. 1042–73 in The Semitic Languages – An International Handbook, edited by Weniger, S. Berlin/Boston: De Gruyter.
Watson, Janet C. E., Bonnie Glover Stalls, Khalid Al-razihi, and Shelagh Weir. 2006. “The Language of Jabal Rāziḥ: Arabic or Something Else?” Proceedings of the Seminar for Arabian Studies 36:35–41.
Watson, Janet C. E. 2014. “Southern Semitic and Arabic Dialects of the South-Western Arabian Peninsula.” Proceedings of the Seminar for Arabian Studies 44:147–53.
Watson, J. C. E., B. Glover Stalls, K. al-Razihi, and S. Weir. 2006. “Two Texts from Jabal Razih, North-West Yemen.” Pp. 40–63 in Current issues in the analysis of Semitic grammar and lexicon II: Oslo-Göteborg cooperation 4th-5th November 2005, edited by L. Edzard and J. Retsö. Wiesbaden: Harrassowitz.
 For example, Michin has French nominal morphology and Cree verbal morphology (Bradley, personal communication). Lamunkhin Even (Northern Tungusic) borrowed two full verbal paradigms from Sakha (Turkic) (
 One of the arguments I will propose in my doctoral dissertation is that there were considerable amounts of Arabic speakers in South Arabia before Islam.
2 thoughts on “Are Faifi and/or Razihit descendants of Ancient South Arabian??? (spoiler alert: no.)”
All I can add is that by “glottalized” you consistently mean “pharyngealized”; you do consistently use the right diacritic (ˤ) for “pharyngealized”.
Oh, and, “Michin” is Michif, on which there’s a lot of information available online, starting with Wikipedia.
Hello, my name is Ömer Aras. I’m a Ph.D. student in Turkey in the field of Islamic History. I work on Arab genealogies. I would like to ask you some questions about Yemen and genealogy, if you give permission. I tried to write on Twitter but couldn’t send a message. Best regards.