any language families with kh/S correspondence

[Radagast the Third]

Do you all know of any language families with members that have the following regular sound correspondences?

k_h (or just k) - S
k_h - ts
g - z
j - t_h

[2+3 clusivity]

Please expand upon your question. Your notation is non-standard, so I am having a bit of difficulty figuring it out. The XSAMPA is fine, just the other ... stuff?

[zompist]

I'm not sure what you mean by th, but East Bay (in Austronesian) has a correspondence th / y, as in Sakao iɛð vs. Shark Bay θar 'four'. (These are cognates, both from proto-Vanuatu *vati. The example comes from Jacques Guy.)

Also, you can get x / š in East vs. West Iranian; cf. the words for 'six'.

[Radagast the Third]

I've corrected my Sampa. Too long since I've used it. The kh and th are aspirated stops.and the y is of course IPA [j].

[hwhatting]

You can find g > z (and k > t_s) in Slavic as a result of the so-called second palatalization. Before /v/, the correspondence can be observed inside Slavic, as the second palatalization did not take place in this position in Western Slavic:
Russian zvezda, Polish gwiazda "star"
Russian cvet, Polish kwiat "flower"

[zompist]

For kʰ/ts, you could peruse Sino-Tibetan-- e.g. Sgaw khwi / Suzhou tsiöy 'nine'.

For kʰ/s the same family has e.g. Tibetan khyi vs. Phom shi 'dog'; also note Phom shü 'nine'.

Heck, while I'm on that page, there's Cantonese jɐt vs. Monpa t’i 'one'...

[dhok]

Unconditional *k(ʰ) to tʃ(ʰ) is attested in, I think, one of the Mayan branches, but with *q turning to /k/ to follow up.

*y to /t/, without contrasting aspiration, happened in Cheyenne.

[Hakaku]

These are the closest examples I could find in the Japonic languages:

k_h (or just k) - S
"today"
Japanese /kjoo/
Shodon /kʰjuu/
Yoron /ɕuu/
Shuri: /t͡ɕuu/
Itoman /kɯɯ/
Yonaguni /su/

In this scenario, you could match Shodon /kʰj/ or Itoman /k/ to Yoron /ɕ/ (which is similar to /ʃ/).

k_h - ts
In the previous example, there's a correspondence that almost fits: Shodon /kʰj/ (or Itoman /k/) and Shuri /t͡ɕ/.

g - z
There's a regular correspondence between Japanese /gi, gj, ig, gei/ and Shuri /z(i)/. Some examples:

J: /isigaki/ O: /isizaci/ "Ishigaki (place name)"
J: /ojogu/ O: /ʔwiizuN/ "to swim"
J: /geinoo/ O: /ziinuu/ "performing arts"

That said, the sound usually transcribed as /z/ is actually closer to [d͡ʑ ~ ʑ].

j - t_h
The closest correspondence in this language family is Japanese /j/ which corresponds to Yonaguni /d/.

J: /jama/ Y: /dama/ "mountain"
J: /ja/ Y: /da/ "house"
J: /jonaguni/ Y: [dunaŋ] "Yonaguni (place name)"

[8Deer]

For *k - ts, there is Halkomelem. Most other Coast Salish languages have the correspondence *k - tʃ: Proto-Salish */kalax/ "hand" > Sliammon /tʃajəʃ/, Halkomelem /tseləx/.

[Radagast the Third]

Thanks, guys. This is great. I had a feeling these were not odd sound changes, just couldnt name any off the top of my head. The change k>t_S is of course well known, and it makes sense that kh>t_S>t_s as it seems to have done in Salish.

[vokzhen]

Unconditional *k(ʰ) to tʃ(ʰ) is attested in, I think, one of the Mayan branches, but with *q turning to /k/ to follow up.

Off the top of my head, Eastern Mayan kept around /q/, Yucatec merged q>k, and Huastec and most Western Mayan "satemized" q>k>tS. The Chujean branch only has partial k>tS though. I'm not sure what evidence there is for k>tS pulling q>k versus q>k pushing k>tS, though.

[dhok]

Unconditional *k(ʰ) to tʃ(ʰ) is attested in, I think, one of the Mayan branches, but with *q turning to /k/ to follow up.

Off the top of my head, Eastern Mayan kept around /q/, Yucatec merged q>k, and Huastec and most Western Mayan "satemized" q>k>tS. The Chujean branch only has partial k>tS though. I'm not sure what evidence there is for k>tS pulling q>k versus q>k pushing k>tS, though.

Ah, right, that's what I was thinking of- the Huastec satemization.

[Travis B.]

I do not get what is special about this; it is a simple result of palatalization combined with affrication followed by, in cases, deaffrication, which are very common crosslinguistically.

[Radagast the Third]

Noone said it was special. In fact I was hoping and assuming it was not, because I want to know the paths.

The question I want to solve is what development took place. You seem to think that the verlar stops can beest be considered the original state with a fronting/palatalization and then affrication/deaffrication. That makes sense. But I was wondering if for example it might make sense to posit a palatal stop which split to the velar stop and palatal sibilant. But you are probably right that it is better to reconstruct the velar as the original state.

[Travis B.]

To my knowledge, fronting of velar and uvular consonants is very, very common, while backing of palatal and alveolopalatal consonants is very, very rare.

[----]

Heck, while I'm on that page, there's Cantonese jɐt vs. Monpa t’i 'one'...

Is it certain that these are actually cognates?

[zompist]

Heck, while I'm on that page, there's Cantonese jɐt vs. Monpa t’i 'one'...

Is it certain that these are actually cognates?

Why would you think they wouldn't be? The Cantonese is quite typical for Sinitic, the Monpa is quite typical Bodic; we have reconstructions in these and other parts of the family that seem to bridge the gap; numbers are fairly resistant to relexification.

[----]

Why would you think they wouldn't be? The Cantonese is quite typical for Sinitic, the Monpa is quite typical Bodic; we have reconstructions in these and other parts of the family that seem to bridge the gap; numbers are fairly resistant to relexification.

Because the word for 'one' in Written Tibetan (WT), another Bodic language, is gcig, which is demonstrably not cognate to the Cantonese term--final g in WT corresponds to final k in Cantonese, cf. drug ~ [lùk], both meaning 'six'. Proto-Sino-Tibetan also has multiple reconstructions for terms that mean "one". It seems preferable to say that gcig and t'i are cognate since they are more phonologically similar, rather than saying Tibetan's word for one is not cognate to the Monpa one, but the Monpa term is cognate to the Cantonese one, since Monpa and Tibetan are in the same branch, while Cantonese is not inside this branch.

[Sumelic]

To my knowledge, fronting of velar and uvular consonants is very, very common, while backing of palatal and alveolopalatal consonants is very, very rare.

The only examples of backing of palatals/alveolopalatals that I am familiar with are all more-or-less /ʃ/ > /x/ (Spanish is the main example that comes to mind, but also Slavic /x/ from PIE /s/ after ruki, and I think some Iranian languages have this). I guess the resulting velar fricative could potentially be fortited to /k/ or /k_h/, but my impression is that this would be another uncommon sound change.

[Vijay]

Some Eastern Indo-Aryan languages have that change, too. I think also some Semitic languages or something might. (This is my very vague recollection of an informal talk given by a grad student working on Semitic languages several years ago).

[dhok]

There does seem to be more of an exception to that rule when the coronal moves into a vacuum. E.g., several Austronesian languages changed *t to k after the glottalization or loss of original *k- most famously Hawaiian, but there are other, less dubious examples (since Hawaiian /k/ has a very wide range of allophony). Then there's colloquial Samoan, which merged *n and *ŋ as /ŋ/, but I don't know the chronology on this- maybe they merged to /n/ first, like in Hawaiian.

Other examples that come to mind include the shift of interdentals to labiovelars in a couple branches of Athabaskan- no vacuum there.

As for backing of palatals specifically, I could have sworn there's a Native American branch somewhere that had that change. I'll try and find it tomorrow.

[Tropylium]

The only examples of backing of palatals/alveolopalatals that I am familiar with are all more-or-less /ʃ/ > /x/ (Spanish is the main example that comes to mind, but also Slavic /x/ from PIE /s/ after ruki, and I think some Iranian languages have this).

Which is interesting, since /tʃ/ is also pretty common, and yet there never seem to be parallel examples of *tʃ > *tʂ > *kx > k. Spanish and Eastern Iranian still keep /tʃ/ around (though we might ask if *š from RUKI was a more retroflexed sound than *č from Law of Palatals). Slavic definitely did not have a suitable affricate around at the time, though, and also in Finnic *č merged with *t already before *š > *h.

(A few Northeast Caucasian and Nguni langs have *tɬʼ > kxʼ (> kʼ), and Aramaic has *tɬʼ > *kxʼ > *ɣ > ʕ, but that's not quite the same thing.)

[Radagast the Third]

So the interesting thing is that we have the following set of correspondences:

Language A - Language B
ʃ - kh /_back vowel
ts - kh /_front vowel
g - z /_back vowel

And in language B /z/ is the voiced alternant of /ts/ in several paradigms.


So I am wondering if

Language A backed *z to /g/ after having palatalized /kh/ to /ts/ʃ/. Or if language Bonly palatalized the voiced velar stop to /z/ but kept the voiceless velar stop intact - wiyh language A doing the opposite (palatalizing the voiceless velars and keeping the voiced velar).

[Richard W]

... numbers are fairly resistant to relexification.

The word for 'one' is relatively unstable.