zompist bboard

Does anyone know of any theories on what motivated Proto-Semitic /ɡ/ to develop into /ɟ/ in Arabic? Is this sound change attested crosslinguistically?

I started wondering about this while musing on the centum/satem split in Indo-European. If Proto-Indo-European, like Classical Arabic, had /ɟ/ and /k/ but no /ɡ/ or /c/, then both centumisation and satemisation would simply be a levelling of the gaps in the sound system rather mergers. The not-so-numerous instances of /k/ in satem languages that are traditionally reconstructed as original *k rather than *kʷ would represent sporadic failure to front analogically rather than a separate underlying phoneme.

I believe I've heard that Proto-Indo-European *ǵ and *ǵʰ have palatal reflexes in Uralic, but I'm not sure if the same correspondence is attested for *ḱ.

Šọ̈́gala wrote:Does anyone know of any theories on what motivated Proto-Semitic /ɡ/ to develop into /ɟ/ in Arabic? Is this sound change attested crosslinguistically?

Palatalization?

More specifically, unconditioned palatalisation of the voiced velar stop while the voiceless equivalent remains velar.

Šọ̈́gala wrote:More specifically, unconditioned palatalisation of the voiced velar stop while the voiceless equivalent remains velar.

Yes, that's very common. Well no, it's not, it's rare, but it's a well-known, relatively common rarity.

In some varieties of Arabic, *k also has a palatalized reflex. The ones I know about only have this conditionally (e.g. near front vowels and stuff like that) but it's possible some have this as an unconditional change as well.

Also, it seems like the common change of *q to [g] in some varieties of Arabic would be related somehow. I don't know if it's better classified as a push or a pull, but apparently some modern varieties of Arabic use a voiced uvular stop [ɢ] for this phoneme, and some historical grammatical descriptions of Classical Arabic classified it with voiced sounds (however, maybe this just reflected a lack of aspiration compared to other plosives or something).

I follow the interpretation that palatovelars were actually plain and the plain velars were actually uvular.

mèþru wrote:I follow the interpretation that palatovelars were actually plain and the plain velars were actually uvular.

Is this in the wrong thread?

mèþru wrote:I follow the interpretation that palatovelars were actually plain and the plain velars were actually uvular.

I agree, that one's pretty good. And, indeed, not altogether incompatible with what I said above: could be there was originally /ɟ/ /k/ /ɡʷ/ /kʷ/ /ɢ/ /q/. Indeed there do seem to be more than zero solid etymologies for a separate series (traditionally velars) contrasting with traditional palatovelars. A chain shift ɢ > ɡ and ɡ > ɟ, along the lines of what sumelic said, is plausible.

If you start out with no palatovelars at all, then it might be hard to explain the palatal reflexes in Uralic (assuming I have that part right in the first place). Arguably also hard to explain the unconditioned fronting of /k/ in satem languages in this scenario, although I think it's worth noting that, since most vowels in PIE are *e, a conditioned fronting would naturally be very frequent and might be generalised by analogy.

Semitic languages are difficult to find analogies for in other language families because of the way their grammar restricts what types of sound changes can occur. Conditional sound changes of consonants are rare because the same consonant in a root can appear in almost any environment in different forms of the same word. Hebrew managed to pull off the begadkefat shift, but even in Hebrew most sound changes for consonants seem to be unconditional. It is possible that Arabic at some point originally palatalized the velar stop too, though why that one would snap back to [k] while the other didnt I cant say. There does not seem to have been a preexisting /č/-type sound that would have gotten in the way of the voiceless velar shift. Perhaps its merely that very few languages have no /k/, but missing /g/ is common. One could similarly muse why it is that /p/, but not /b/, became a fricative in all positions in Arabic.

Šọ̈́gala wrote:If you start out with no palatovelars at all, then it might be hard to explain the palatal reflexes in Uralic (assuming I have that part right in the first place).

Not really, afaik most Uralic loans come from satem languages that already would have had fronting of some kind by the time borrowing occurred.

Arguably also hard to explain the unconditioned fronting of /k/ in satem languages in this scenario

"Satemization" shifts of q>k>tS aren't too uncommon. Southern Athabascan and Western Mayan languages have had similar shifts, for example, and it crops up in Neo-Aramaic dialects. In fact, Western Mayan and Neo-Aramaic are evidence of "satemization" and "centumization" popping up in closely-related language varieties, some have q>k>tS while others only have a merger of q>k, and in Neo-Aramaic you also have distinction between the two, while Chujean, on the border between "satemizing" languages to the east, "centumizing" Yucatec to the north, and conservative /k q/ languages to the east, as far as I can tell only partially completed k>tS but fully completed q>k.

Šọ̈́gala wrote:I believe I've heard that Proto-Indo-European *ǵ and *ǵʰ have palatal reflexes in Uralic, but I'm not sure if the same correspondence is attested for *ḱ.

It is. For example, '100' is *śëta > sata in Finnish, száz in Hungarian, čuohti in Northern Sami. (Contrary to the traditional transcription, *ś may have been a palatal stop originally.) This, and probably all similar loans, is also obviously from Proto-Indo-Iranian rather than PIE proper, what with the *m̥ >> /a/.

Salmoneus wrote:

Šọ̈́gala wrote:More specifically, unconditioned palatalisation of the voiced velar stop while the voiceless equivalent remains velar.

Yes, that's very common. Well no, it's not, it's rare, but it's a well-known, relatively common rarity.

Are you sure you're not confusing this with unconditional lenition of *g? I don't think I've heard of any languages with the same thing as Arabic; and in fact I know of several that have a kind of opposite fronting asymmetry (e.g. /tʃ ɟ/ in Spanish, /ts dʒ/ in Gallo-Romance).

Now unconditional ɣ > j is attested though, and unconditional j > ɟ is perfectly normal, so that would still predict that unconditional g > ɣ > j > ɟ should be possible (if rare). Doesn't help us a whole lot in the Arabic case of course, since the language retains /ɣ/ and /j/ all the way since Proto-Semitic. Unless we assumed that *ɣ was actually /ʁ/ at the time, and that the fronting was through /ʝ/?

---

As long as we're on Arabic and POA chainshifts BTW, here's one speculation to ponder on: Proto-Semitic might actually have had a full series of postalveolars! The *ṯ *ṱ *ḏ series, traditionally reconstructed as interdentals *(t)θ *(t)θʼ *(d)ð, is actually only based on the Arabic reflexes /θ ðˤ ð/: just about everywhere else they merge with other things (usually the plain sibilant series, but in Aramaic the dental stops). Afro-Asiatic reconstructions however suggest that these come at some time depth from postalveolars *č *čʼ *dž: e.g. *ṯawr 'bull' ~ Cushitic *čawr 'id.'. But if so, there's no particular reason for the fronting to be pre-Proto-Semitic! It could well be a late shift particular to Arabic.

This would give a bit more leeway on how to route the later developments such as *ṯ > š. We would not need to assume that this was earlier than the retraction of *s to /ʃ/ + the deaffrication of *ts to /s/ (which seem to be pretty widespread), and it could have instead been simply */tʃ/ > /ʃ/. One other possibly relevant thing is that Ugaritic allegedly later on recycles the letter for ḏ as a letter for /ʒ/ (where that comes from though, I don't know; it's not actually from older †ḏ, which rather merges with /d/).

Proto-Semitic's one of those wondefully complicated areas of research that I'd love to learn more about some time.

Šọ̈́gala wrote:More specifically, unconditioned palatalisation of the voiced velar stop while the voiceless equivalent remains velar.

This isn't an explanation so much as a more detailed description of what happened. What caused this palatalisation? Why was it unconditional? Why did whatever caused it not affect the voiceless velar?

(These are questions of a sort which often do not have easy answers.)

Curlyjimsam wrote:

Šọ̈́gala wrote:More specifically, unconditioned palatalisation of the voiced velar stop while the voiceless equivalent remains velar.

This isn't an explanation so much as a more detailed description of what happened. What caused this palatalisation? Why was it unconditional? Why did whatever caused it not affect the voiceless velar?

(These are questions of a sort which often do not have easy answers.)

I've always suspected that the voiced velar fronted due to pressure from the uvular stop, which some sources reconstruct as voiced in early Arabic. In this model, /k/ didn't have a correspondingly voiceless uvular to trigger fronting.

Also, I do think there's something fishy about the three Proto-Semitic sounds traditionally reconstructed as interdentals, since the only descendant in which we're sure they had those values is Arabic. Myself, I've kind of leaned toward Proto-Semitic contrasting two columns of sibilants (possibly with a marginal third in plain voiceless fricatives only) for a while.

Proto-Semitic relies too heavily on Arabic in exactly the same way that Proto-Indo-European relies too heavily on Greek and Sanskrit.

KathTheDragon wrote:Proto-Semitic relies too heavily on Arabic in exactly the same way that Proto-Indo-European relies too heavily on Greek and Sanskrit.

Absolutely.

The only thing that even makes me hesitate in saying that the "interdentals" were a sibilant series is Aramaic reflecting them as stops; I'm not aware of many instances of sibilant affricates or fricatives unconditionally hardening to stops. (Spirants hardening, yes.) The only one I can think of off the top of my head is apparent [s] > [t] in Vietnamese.

ts > t : Proto Austronesian to Proto Malayo Polynesian, Proto Austronesian to Kavalan, Amis, Bunun, Siraya ; Mpi, Bola (Sino-Tibetan)
s > t : Proto-Oceanic to a variety of languages including groups in the Southeast Solomons, New Caledonia, Micronesia, etc. (these are independent developments); many examples in Kamarupan (Sino-Tibetan)
tʃ > t : Blackfoot, Bola (ST)
ʃ > t : also Blackfoot, Ugong (sort of, it's dental)(ST)

P.S.: s > t in Vietnamese is more than just apparent, it's the accepted theory by scholars

I feel like there should be more emphasis on Hebrew and South Semitic languages (especially the latter).

To my knowledge, the PS "interdentals" are written as such as a matter of tradition (like Akkadian sibilants), and most Semiticists these days reconstruct them as something like /ʦ ʦʼ ʣ/ or /ʧ ʧʼ ʤ/, which not only makes more sense in terms of Semitic but also broader Afro-Asiatic reconstruction.