New paper on ordinals

This blog post is now a paper, which came out unexpectedly soon: ‘Ordinal Numerals as a Criterion for Subclassification: The Case of Semitic’.

Abstract: This article explores how ordinal numerals (like first, second and third) can help classify languages, focusing on the Semitic language family. Ordinals are often formed according to productive derivational processes, but as a separate word class, they may retain archaic morphology that is otherwise lost from the language. Together with the high propensity of ‘first’ and, less frequently, ‘second’ to be formed through suppletion, this makes them highly valuable for diachronic linguistic analysis. The article identifies four main patterns of ordinal formation across different Semitic languages. Together with innovations in the lowest two ordinals, these can be correlated with more and less accepted subgroupings within Semitic as a whole. Concretely, they offer support for the widely accepted West Semitic, Northwest Semitic and Abyssinian (Ethio-Semitic) clades as well as the recently proposed Aramaeo-Canaanite clade and provide new evidence for the further subclassification of Abyssinian that matches other recent proposals. However, no evidence was found to support the debated Central Semitic or South Semitic groupings. Given the accurate identification of accepted subgroupings and high level of detail, this approach holds promise for the classification of other language families, especially where other linguistic data are scarce.

Enjoy!

#Akkadian #Amharic #AncientSouthArabian #Arabic #Aramaic #GeEz #Hebrew #linguistics #ModernSouthArabian #news #ProtoSemitic #Ugaritic

Reconstructing the Modern South Arabian dual endings

One of the many lovable features of the Modern South Arabian languages is their productive retention of the dual, in nouns, pronouns, and verbs (all persons). Some examples from Omani Mehri (Rubin 2018):

• ġáygi ṯrōh ‘two men’, contrast ġayg ‘man’ (in practice, dual nouns are nearly always followed by the numeral ‘two’)
• perfect bǝg(ǝ)dōh, bǝgǝdtōh ‘the two of them (m./f.) chased’, contrast singular bǝgūd, bǝg(ǝ)dūt and plural bǝgáwd, bǝgūd
• imperfect yǝbǝgdōh, tǝbǝgdōh ‘the two of them (m./f.) chase’, contrast singular yǝbūgǝd, tǝbūgǝd and plural yǝbǝ́gdǝm, tǝbǝ́gdǝn
• independent pronoun hay ‘the two of them’, contrast singular hē ‘he’, sē ‘she’, plural hēm, sēn ‘they (m./f.)’

Let’s look at some reconstructions.

The verb

The verbal dual ending can be reconstructed for Proto-Modern-South-Arabian as *-óh, but its deeper Semitic origin has not been explained. Dufour (2022: 77) writes:

The suffix for the dual in verbs is stressed in MSA (stable in Soqotri). It is unclear what etymon should be posited for it. Akkadian and Classical Arabic have -ā, but such an etymon would not fit the MSA forms since, as we have seen, final vowels drop in Proto-MSA (cf. the exactly identical *-ā suffix marking 3fp in the perfect: Ga *ḳadarā> OMh. ḳədū́r, J./ Ś. ḳɔdɔ́r). On the other hand, Epigraphic South Arabian attests verbal dual suffixes written with a /y/, though what this orthography stands for is unclear. Perhaps we should therefore posit *-ay or *-āy.

Both of these reconstructions run into problems, but I think we can solve those issues by combining both options and reconstructing *-ayā. Consider the following:

First, the final *-h is probably automatically added to a stressed final vowel, or at least to *-ó(h). Rubin discusses this in his § 2.2.4.

How do we get a stressed word-final vowel? Shouldn’t word-final vowels be lost, as Dufour states? Well, if we reconstruct *-ayā, then the first of those two vowels isn’t word-final. Modern South Arabian stress is super weird, but the main rule (for words containing a Proto-West-Semitic low vowel) is pretty much: stress the last, non-word-final *a or *ā. In a reconstructed form like 3m.du. perfect *bagad-ayā, that gives us *bagad-áyā, with the stress in the right part of the word: the suffix.

Next, we have to assume that *-áyā contracts to *-ó(h). *a turns to Proto-Modern-South-Arabian *o most of the time, so this just means loss of an intervocalic glide, something I don’t mind at all. In fact, there’s MSAL-internal evidence for almost exactly the same change. III-y verbs, like *bakay–a ‘he wept’, show up in MSAL as *bokóh (> OMehri bǝkōh). Before suffixes, these verbs retain their y, like OMehri tǝwōh ‘he ate’, tǝwyǝ́h ‘he ate it (m.)’. And, what do you know, so do the duals: “sǝbṭáys ‘they (two) hit her’; śǝnyáyǝh ‘they (two) saw him’”.

The difference in stress in the suffixed forms here is interesting. The 3m.sg. form tǝwy-ǝ́h shows a stressed suffix, part of a paradigm that is only used on the 3m.sg. and 3f.pl. perfects. These are exactly the forms that are reconstructed as ending in a low vowel, *-a and *-ā, respectively (explaining why the vowel right before the suffix is stressed).

Does this show that the 3du. forms did not end in *ā? Maybe. But maybe not. In Jibbali, the stressed object suffixes only occur on the 3m.sg., not on the 3f.pl. (Rubin 2014). As vowel length doesn’t usually play a role in MSAL vowel changes, this is probably due to analogy, the 3f.pl. taking the unstressed suffixes that are used everywhere else in the verb. In the same way, these suffixes could have spread to the 3du., also in Mehri where they didn’t make it to the 3f.pl. So, we could reconstruct the 3m.du. perfect forms as follows:

The numeral

The number ‘two (m.)’ has what looks like the same ending, as we saw in Omani Mehri ṯrōh. Should we reconstruct this in the same way? I don’t think so; we can get there without the triphthong. Based, in fact, on evidence from Modern South Arabian in particular, this is one of the words where we should probably reconstruct a word-initial consonant cluster in Proto-Semitic: the ‘two’ stem was probably just *θn-, with no vowel. If we add the dual nominative ending, that gives us *θn-ā. While word-final vowels normally don’t receive the stress in Proto-MSAL, here, it’s the only vowel in the word. So without further ado, we can imagine the development as pre-Proto-MSAL *θn-ā́ > Proto-MSAL *θr-óh > OMehri ṯrōh, etc. It’s really striking that the MSAL form seems to go back to a reconstruction with just an *-ā, just like Akkadian šinā. šinā doesn’t inflect for case (as far as I know), which would explain why we don’t get the expected oblique dual ending *-ay(na) here in Modern South Arabian. Adding this to my list of eerie Akkadian-MSAL isoglosses.

The feminine looks a bit confusing but at first sight I would guess it reconstructs to Proto-MSAL *θrót (e.g. Jibbali ṯrut). This regularly goes back to pre-Proto-MSAL *θn-át-ā; here, the dual ending is in a polysyllable, hence unstressed, and therefore lost.

The noun

Nouns mark the dual with a suffixed -i (mostly lost in Jibbali). Unfortunately, dual nouns can’t take possessive suffixes, so we don’t have any allomorphs to work with. Looking at other languages, I think our best bet for the reconstructed morpheme here is the nominal dual oblique ending *-ay. No nunation or mimation seems to follow (maybe because the numeral ‘two’ is always right behind the noun?). I’m not sure if *-ay should yield Proto-MSAL *-i; it doesn’t seem to in the jussive of III-y verbs.

The pronoun

Here are the forms from Rubin’s grammars, independent and suffixed:

ǝkáy indeed. Mehri gives us some support here for the idea that unstressed *-ay > *-i. Pronouns are generally a pain to reconstruct, because they all influence each other so much. I’ll venture a reconstruction of 2du. as pre-Proto-MSAL *ʔantay, *-kay and 3du. as *say, *-say, wonder out loud what the hell is going on with *k in the first person, and leave it at that.

Summing up

It looks like we can account for the Modern South Arabian dual suffixes by deriving them from *-ay in the noun and probably the pronoun, *-ā in the numeral ‘two’, and *-ayā in the verb. The first two morphemes are pretty much expected as the nominal oblique and nominative endings. For the verbal ending, as Dufour says, usually we’d expect *-ā (or is that too Arabocentric?). That suggests that we’re really looking at a double marking, *-ay-ā, with the dual verb tacking on the nominal oblique (and pronominal) *-ay suffix before the true verbal one. Maybe Old Akkadian, Ancient South Arabian, and Eblaite have some more to add to the story, but for now, this seems double-plus-good to me.

#linguistics #ModernSouthArabian #ProtoSemitic

Update on Mehri goats

Earlier today, I wrote:

PS *ʕVnz- ‘she-goat’ > Mehri, Harsusi wōz, Jibbali oz, Soqotri o’oz (? but then where did the *ʕ go?)

It just struck me that this is one of the lexically determined words that take ḥ- as the definite article in Mehri and Harsusi, at least. Many of these words used to start with a *ʔ—like M. ḥa-ynīθ ‘women’!—but not all of them; Rubin (2018) mentions some kinship terms where it’s analogical, for instance.

The word for ‘goat’ also happens to have a suppletive plural; from memory, that’s ḥə-rawn. This is probably one of the words where the shape of the article is due to original presence of *ʔ-: Rubin compares Syriac arn-o ‘mountain goat’.

Suppose the singular is from *ʕVnᵈz– and it took the ḥ-article by analogy with the plural. That means we might expect something like *ḥ–ʕōz for ‘the she-goat’. With two pharyngeals in a row, this would be a great environment for the *ʕ to be lost, yielding the attested form, ḥ–ōz. The indefinite form, wōz, would then in turn have been formed by analogy with the definite form. IMHO, this shores up the derivation from *ʕVnᵈz– and supports loss of *n directly before another consonant in an ancestor of the MSAL (provided we can make it work for Jibbali and Soqotri as well).

#Aramaic #linguistics #ModernSouthArabian #ProtoSemitic

‘Woman’ in Modern South Arabian, Amorite, and Ugaritic

EDIT: Roey Schneider reminds me he probably suggested this idea to me back in 2023! No plagiarism intended.

Some Modern South Arabian languages have a weird-looking word for ‘woman’: Mehri tēθ, Harsusi and Jibbali teθ. The θ makes it look similar to Proto-Semitic *ʔanθ–at-, which underlies Ugaritic a͗θt, Hebrew ʔiššā, Syriac <ʔntt-ʔ> at-o, Akkadian aššat- ‘wife’, etc. The same root also gives Arabic ʔunθ-ay– ‘female’1. But what about that initial t-?

For years, I’ve kind of assumed the Modern South Arabian words also come from something like *ʔanθ–at-, with the first part being lost and *θ-et then metathesizing to *teθ. It’s weird, but it was my best guess. But here’s a new guess I like better.

In late 2022 (paywalled), Andrew George and Manfred Krebernik published what they aptly referred to as “two remarkable vocabularies”, containing what is probably the first known connected text in Amorite, a Northwest Semitic language of the early second millennium BCE. One of the many surprises these texts contain is the word for ‘woman’ (unambiguously written with a Sumerogram in the Akkadian translation), ta-aḫ-ni-šum. Based on comparisons to the Semitic words above and known Amorite/Akkadian spelling conventions, this looks like *taʔnīθ-um, yet another different noun formation from the *ʔ-n-θ root. As I learned from a recent handout byTania Notarius, Ugaritic also attests a form that looks related: ti͗nθt ‘women’, ‘females’, plausibly /tiʔnīθ-āt-u/.

Both of these forms show a t- prefix, part of a pattern that usually forms abstracts—although concrete nouns in this pattern also occur, like Hebrew < Aramaic talmīḏ– ‘student’. And the Amorite, at least, lacks a feminine suffix. So that’s starting to look like our MSAL *teθ. Could this be a full cognate, with *teθ coming from *taʔnīθ-?

That depends on whether we can get rid of the first two radicals, *ʔ and *n. As far as I know, Proto-Semitic *ʔ was regularly lost on the way to Modern South Arabian. So that’s fine. What about *n, is this one of the (surprisingly) many branches of Semitic where it assimilates to following consonants? Let’s check out some likely etyma with *n before a consonant:

• PS *ʔanta ‘you (m.sg.)’ > Mehri, Harsusi hēt, Jibbali hɛt (if this is the right etymon)
• PS *ʔantum ‘you (m.pl.)’ > Mehri ətēm, Harsusi etōm, Jibbali tum, Soqotri ten
• PS *ʕVnz- ‘she-goat’ > Mehri, Harsusi wōz, Jibbali oz, Soqotri o’oz (? but then where did the *ʕ go? [update])

That’s all I’ve got, for now. The plural pronoun looks good, though. Of course, in *taʔnīθ-, the *n isn’t directly before the θ, so why should it assimilate? After assigning the stress to the first *a—a strange, but reliable rule in pre-MSAL—we could imagine something like
*táʔnīθ > *táʔnəθ (vowel reduction) >
*táʔənθ (metathesis) >
*táʔəθθ (assimilation) >
*teθθ (loss of the glottal stop, vowel contraction, MSAL vowel weirdness)
*teθ (degemination—not entirely clear whether this is regular).

Writing it out like that, the non-gemination of the θ (also word-internally, as in the Mehri dual tēθ–i) may also be a problem for assuming a derivation from the *ʔ-n-θ root.2 Still, this is commonly assumed; supporting evidence comes from the plural forms, like Mehri yənīθ, where the n is visible. So, since the t- in *teθ really does look like a prefix, I think Amorite *taʔnīθ- is an exciting form to compare.

#Akkadian #Amorite #Arabic #Hebrew #linguistics #ModernSouthArabian #ProtoSemitic #Syriac #Ugaritic

Rethinking Proto-Semitic

This week, I was stoked to attend a workshop in Marburg, Germany, entitled “Rethinking Proto-Semitic” and organized by profs Stefan Weninger and Michael Waltisberg. Despite some cancellations, the workshop had an amazing lineup of speakers—and a terrific atmosphere. Here’s my summary of the talks.

Leonid Kogan, “What can we learn from Eblaite on Proto-Semitic morphology?” Ongoing study and decipherment of the 24th-century BCE East Semitic language from Ebla, Syria shows the following features that are interesting for reconstruction:

Maria Bulakh, “Intercalated *a as a plural marker in Soqotri and its implications for the reconstruction of Proto-Semitic”. While superficially hard to recognize (and Jorik and I didn’t attempt to in our paper on this subject), reconstruction of Modern South Arabian and especially Soqotri attest insertion of *-a- between the second and third radical of *CVCC- nouns in the plural. No external plural suffix though.

Me, “Rethinking the Proto-Semitic stative”. Slides here. Got some good suggestions for languages where I could go looking for a synchronic distinction between resultative *qatal-a and preterit *ya-qtul.

Ahmad Al-Jallad, “Revisiting the post-verbal morphemes *-u and *-n(V) in Semitic: a proposal for a unified theory”. The different verbal suffixes/enclitics shaped like -u and -n(V) in Akkadian, Central Semitic possibly Modern South Arabian, and Gurage (South Abyssinian) could all descend from the Proto-Semitic *=u(m) locative, which gained various subordinating and durative meanings. Central Semitic *ya-qtul-u instead of *ya-qattal-u for the imperfect could show a collapse in the distinction between *ya-qtul and *ya-qattal related to the rise of the West Semitic perfect *qatal-a.

Michael Waltisberg, “Issues of reconstructive methodology in Semitics”. Based on his review of Rebecca Hasselbach(-Andee)’s 2013 Case in Semitic, Waltisberg discussed some methodological questions like whether our reconstructed Proto-Semitic represents an actually spoken language or just maps correspondences between different languages and whether there is room for dialectal diversity and different chronological stages within a protolanguage. (Prof. Hasselbach-Andee sadly had to cancel her planned attendance.)

Lutz Edzard, “Linguistic divergence and convergence in Arabic and Semitic revisited”. As the most protolanguage-sceptic scholar at the workshop, Edzard reviewed some of his problems with the linear-descent-only family tree model where every language in a family descends from a kind of ancestral singularity with no internal diversity.

Vera Tsukanova, “What can modern Arabic dialects reveal about the etymology of the L-stem in Semitic?” The development of the L-stem (*qātal-) in historical Arabic suggests that it is more likely that this stem originally had a concrete meaning like applicative that was bleached in some languages than that it was originally vague and acquired its specific meaning in pre-Arabic.

Eran Cohen, “Semitic k-based similative particles—comparative and diachronic aspects”. Different Semitic particles starting with k- can be diachronically related to each other according to recognized historical pathways of development.

Na’ama Pat-El, “Homomorphs and reconstruction”. We are probably not dealing with one, syncretic morpheme but rather two homophonous ones in the cases of 1) prefix conjugation 2m.sg./3f.sg. *t-; (2) f.sg. abstract noun/m.pl. adjective suffix *-ūt-; (3) f.sg. noun or adjective/weak root verbal noun or infinitive suffix *-t-. In the latter, most controversial case, Pat-El invoked some evidence that the verbal nouns like Biblical Hebrew šéḇeṯ ‘sitting’ (from y-š-b) are syntactically masculine (e.g. Ps 133:1).

Stefan Weninger, “The Semitic Urheimat question: a review of the proposals and some perspectives”. An overview of some proposed points of dispersal for the Semitic languages since the late 19th century, the main contenders being the Arabian peninsula and East and North Africa. In the Q&A, Kogan added his own suggestion, published in an Encyclopedia Aethiopica article: Canaan.

Walter Sommerfeld, “The concept of a common Semitic cultural area (‘Kish Civilization’) in the 3rd millennium”. Contemporary evidence shows that there is no basis for Ignace Gelb’s concept of a distinctly Semitic culture in Early Dynastic northern Babylonia.

Apart from these talks, we spent about half the time in unstructured panel discussions, on phonology, morphology, methodology, and classification/Urheimat questions. Each discussion was kicked off by a short, stimulating talk, mostly by attendees who did not present full papers: Martin Kümmel, Michaël Cysouw, and Aaron Rubin. This was an experimental feature of the workshop, and I’m on the fence about it; the discussions were certainly fun and a lot of interesting points were brought up (e.g. Kogan: linguistic paleontology shows that Proto-Semitic speakers did know hyraxes but did not know oryxes, and only Canaan is [+hyrax][-oryx]), but it felt like they yielded fewer concrete insights than regular talks would have. It was a nice way to get some more people involved, though, also from adjacent fields (Indo-European/Indo-Iranian and Caucasian/Germanic linguistics).

All in all, it was wonderful to be able to fully geek out about Proto-Semitic and its daughters for a couple of days. There’s plans to publish proceedings, so hopefully in a few years you’ll be able to read all about these topics in full detail. Stay tuned.

#Akkadian #Arabic #Berber #conference #EastCushitic #Eblaite #Egyptian #Gurage #Hebrew #linguistics #ModernSouthArabian #news #ProtoSemitic

Earlier this year, I had two fun conversations with the team of the then newly-founded Kedem YouTube channel, which popularizes scholarship on the Ancient Near East and the Hebrew Bible. The first video was published yesterday. We talk about the concept of a language family, what languages constitute the Semitic language family, where Semitic comes from geographically and linguistically, how we can reconstruct earlier ancestors of the attested languages, and a few things this kind of reconstruction tells us about Proto-Semitic.

Stay posted for my second video with this channel, to be released sometime next year, on the different modern and—especially—ancient pronunciations of Biblical Hebrew.

https://bnuyaminim.wordpress.com/2023/12/30/video-intro-to-the-semitic-language-family/

#Afroasiatic #Akkadian #Amharic #AncientSouthArabian #Arabic #Aramaic #Beja #Berber #Chadic #Cushitic #Egyptian #GeEz #Hebrew #linguistics #Moabite #ModernSouthArabian #news #Omotic #Phoenician #ProtoSemitic #Tigrinya #Ugaritic

Open access in the latest issue of Afrika und Übersee (great publication experience, would recommend): ‘Two more contexts for Ge‘ez *u > u and three for *a > ǝ’. It’s a pretty technical article but I think I have some interesting things to say about various numeral patterns in particular. Abstract:

“The main Ge‘ez (Classical Ethiopic) verbal adjective is characterized by an ǝ-u vowel melody. Based on cognate evidence, the most basic form of this adjective, 01-stem 1ǝ2u3, derives from a *1a2uː3- pattern and thus shows assimilation of *aCuː > ǝCu. This assimilation does not operate in a set of specialized numerals shaped like 1ä2u3, which should be reconstructed as *1a2u3- with short *u. Short *u also yields Ge‘ez u in the nonaccusative case of the masculine cardinal numerals, like *ɬalaːθtu > śälästu ‘three’; this ending goes back to the Proto-Semitic diptotic nominative. The assimilation of *aCuː > ǝCu, on the other hand, also affected the personal pronoun *huːʔa-tuː > wǝʾǝtu, the perfect of fientive verbs like *gabaruː > gäbru ‘they did’, and the jussive of stative verbs like *yitrapuː > yǝtrǝfu ‘may they remain’. Ə was leveled to other parts of these paradigms, solving several longstanding problems of Ge‘ez morphology.”

https://bnuyaminim.wordpress.com/2023/12/16/new-article-some-geez-sound-changes/

#Akkadian #Arabic #GeEz #Hebrew #linguistics #ModernSouthArabian #news #ProtoSemitic

While reviewing proofs for an article that should appear soon, it struck me that the shape ordinal numerals like ‘third’, ‘fourth’, ‘fifth’ take in Semitic provides some evidence for subgrouping that I don’t think I’ve seen before. Quick recap: most scholars today accept something like the following family tree for Semitic, as compellingly presented by Huehnergard & Rubin (2011).

I’m generally skeptical about West Semitic as a group because I think everyone’s favourite West Semitic innovation, the *qatala perfect, may be a retention from Proto-Semitic. But among some other innovations (I particularly like relative/demonstrative *θū > *ðū), this subgroup is supported by the shape of the ordinals. Akkadian has a *CaCuC– pattern, as in:

• Old Babylonian šaluš– ‘third’, rebu– < *rabuʕ– ‘fourth’, ḫamuš– ‘fifth’
• Old Assyrian rabū-t-um ‘the fourth (f.)’, rabū-ni ‘our fourth witness’, ḫamuš-ni ‘our fifth witness’

In West Semitic, the normal ordinal has a different, *CāCiC- pattern, as in:

• Classical Arabic θāliθ-, rābiʕ-, ḫāmis-
• Ge’ez śaləs, rabəʕ, ḫaməs
• Mehri (Modern South Arabian) śōləθ, rōbaʕ, ḫōməs
• probably also Sabaic θlθ, rbʕ, ḫms; Ugaritic θlθ, rbʕ, ḫmš…

In the rest of Northwest Semitic, one trace of this pattern might be found if the consonantal spelling tltʔ in Daniel 5:16 (Biblical Aramaic) stands for *tālítā ‘as the third one’ (Suchard 2022: 224). Otherwise, Aramaic and Canaanite have a different pattern: *CaCīC– followed by the nisbe suffix, which has a special shape in Aramaic. Examples:

• Biblical Hebrew šlīšī, rḇīʕī, ḥămiššī (probably influenced by šiššī ‘sixth’, itself a new formation for expected **šḏīšī)
• Syriac tliṯoy, rbiʕoy, ḥmišoy

So, we have three patterns: *CaCuC-, *CāCiC-, and *CaCīC–īy/āy-. Which one is oldest and which ones are innovative?

Interestingly, Ge’ez and Modern South Arabian both have a special set of numerals that specifically refer to periods of time like days:

• Ge’ez śälus, räbuʕ, ḫämus
• Mehri śīləθ, rība, ḫayməh

In the article I’m proofreading, I argue these can all be reconstructed as *CaCuC-. This also matches Biblical Hebrew ʕāśōr ‘tenth (day)’ and may be related to dialectal Arabic names for the days of a the week like ʔaθ-θalūθ and ʔar-rabūʕ (borrowed from Sabaic???). This matches the Akkadian pattern for the normal numerals, which also happens to be attested with reference to a period of time in Old Assyrian ḫamuš-t-um. It’s more likely for an old formation to be preserved in a specialized use like referring to numbers of days than for something specific like that to be generalized for ordinals in all contexts. *CāCiC– also has an obvious origin, as this is the productive pattern for active participles and we can imagine a kind of shift from ‘being third’ as a participle to ‘third’ as an ordinal. So in terms of innovations, this looks like:

*CaCīC–īy/āy– is so restricted that it is most attractive to see this as a late innovation shared by Canaanite and Aramaic. If so, that would support Pat-El & Wilson-Wright’s (2018; paywalled?) argument on other grounds that these two families form a subgroup within Northwest Semitic.

An intermediate *CaCīC– pattern without the nisbe suffix added might be attested in Biblical Hebrew šālīš, which not only means ‘one-third (of some unknown measure)’ but is also a military rank that has traditionally been explained as the ‘third man’ on a chariot besides the primary warrior and the driver.

This pattern also forms fractions in Aramaic, as in Imperial Aramaic rbyʕ and Syriac rbiʕ-t-o ‘quarter’. So maybe we should see the pre-Aramaeo-Canaanite development as a shift from still very active-participle-y *CāCiC– to more productively adjectival *CaCīC-, with the extra adjectival nisbe suffix being added later for good measure. Maybe that last step took place after the ordinals had started to shift in meaning to fractions (which are nouns, not adjectives), giving something like *rabīʕ–īy– an original literal meaning like ‘quarter-y’.

In conclusion, an ordinals-based family tree ends up looking like this:

https://bnuyaminim.wordpress.com/2023/11/03/ordinal-numerals-as-shared-innovations-in-semitic/

#Akkadian #AncientSouthArabian #Arabic #Aramaic #GeEz #Hebrew #linguistics #ModernSouthArabian #ProtoSemitic #Ugaritic