zompist bboard

Would any of you guys know examples of non-IE vowel shifts? Moreover if you do, I would prefer them to be taken from languages outside of the geographic territory of Europe.

I know of a few non-IE shifts myself like the Savonian Finnish vowel shift, or the great Saami vowel shift, but that's about it. Aside from that it appears that there were extensive vowel chain shifts in the Mon-khmer languages strickingly similar to those seen in Germanic. However I don't know the details. So If any body has an idea what happened in Mon-Khmer, I'd be interested to hear it. As a matter of a fact I believe the Sino-Tibetan are also reputed to have a few such shifts.

The reason I'm asking is because I'd like to get a non-IE perspective of the different ways in which vowel shifts can operate. So far most of all the linguistic litterature (especially Labov) has focused on IE languages which makes it hard to get a broader picture on the whole phenomenon. I'll be honest, I'm mainly looking for conlanging ideas.

I hope you can answer me!

Cantonese has a vowel shift described here.

E.g. 你 (nǐ, MC nɨ) has become nei, and 王 (wáng, MC jwaŋ) has become wong.

zompist wrote:Cantonese has a vowel shift described here.

E.g. 你 (nǐ, MC nɨ) has become nei, and 王 (wáng, MC jwaŋ) has become wong.

Interesting article. I have noticed this myself, though not to the point of seeing the regularity.

In Ryukyuan, standard Japanese /e e: o o:/ correspond to /i i: u u:/, but /ai au/ correspond to Ryukyuan /e: o:/. This appears to be a chain shift, but I'm not familiar with the diachronics on the Ryukyuan side. (The /au o:/ - /o: u:/ correspondence is slightly obscured by the fact that standard Japanese shifted /au/ to /o:/ as well.)

Zhen Lin wrote:In Ryukyuan, standard Japanese /e e: o o:/ correspond to /i i: u u:/, but /ai au/ correspond to Ryukyuan /e: o:/. This appears to be a chain shift, but I'm not familiar with the diachronics on the Ryukyuan side. (The /au o:/ - /o: u:/ correspondence is slightly obscured by the fact that standard Japanese shifted /au/ to /o:/ as well.)

From what I've been reading over at Amritas recently, apparently Japanese short /e o/ actually come from *iə *ə. Original *e *o went to /i u/ in both, and can only be told apart by lack of palatalization/labialization in Ryukyuan, or internal reconstruction of distinct sandhi effects or something.

I've posted some more eastern Uralic shifts in the correspondence thread, but here's a few again. Main developments only, not specific secondary splits.
Mordvinic
Proto-Uralic had *i *ü *u *e *ë *o *ä *a, plus some *Və, *Vj and *Vw sequences.
1) Mergers
*ü i (initially, *wi)
*ë a
2) Laxing
*i *u *ɪ *ʊ
*Və V (probably via a long vowel)
3) Raising
*e *o i u
*ä ɛ
4) Delaxing: *ɪ *ʊ e o
Leaves an inventory of /i e ɛ a o u/.

Nganasan
Proto-Samoyedic had *i *ü *ï *u *ɪ *ʊ *e *ö *ë *o *ä *a *å, and some *Və and *Vj sequences.
Mergers:
*ü(ə) *ï i
*ö(ə) o
*ë(ə) a
*ɪ *ʊ e o / _j, ə / stressed, i u / unstressed
Some sequences
*iə *eə *äə ja
*uə *oə *åə wa
(there was no *ïə *ɪə *ʊə *aə)
Raising chain shift:
*u y
*e *o ɨ u (not *e i!)
*ä *å e o
Leaves an inventory of /i ɨ u e ə o ja a wa/ and some Vj sequences.

Nenets (also from PSamoyedic)
Mergers:
*Və V
*e i
*ö o
*å a
Quality shifts:
*ü ʉ
*ä e
Length development:
All non-close vowels long
*ɪ *ʊ a
*ej, äj iː
*aj æː
(there's also /uː/, not sure where from. Maybe from the rarer back-vocalic *Vj, on whose fate I'm not sure)
Leaves an inventory of [i iː ɨ ɨː ʉ ʉː u uː eː ɤː oː æe a aː], however due to rampant consonant palatalization, [ɨ ɨː ʉ ʉː ɤː] can be analyzed as allophones of /i iː u uː eː/ when next to an non-palatalized consonant.

The Kagoshima dialect of southern Japan exhibits a strong level of vowel coalescence that can more or less be summarized in the following table:



So a word like 赤い /akai/ "red" in standard Japanese would thus correspond to /ake/ in the Kagoshima dialect. 今日 /kjo:/ "today", which goes back to the form /ke(f)u/, becomes /kju/. 太い /hutoi/ "thick" becomes /hute/ "large" (or /huti/ in some southern areas). 赤く /aka(k)u/ "(to become) red" becomes /ako/. 寒い /samui/ "cold" becomes /sami/. 書いて /kaite/ "write (te form)" becomes /kete/, and so on.

The fusion extends to particles as well, so that 事は /koto/ + /wa/ → /koto a/ → /kota/ thing.TOP; 俺は /ore/ + /wa/ → /oi a/ → /oja/ 1.TOP. これを /kore o/ → /koi o/ → /kojo/ this.ACC. And 山に /jama/ + /ni/ → /jama i/ → /jame/ mountain.DAT/LOC.

There are some cases of /u/ and /i/ alternation: 人 /hito/ → /huto/ "person", 一つ /hitotu/ → /hutotu/ "one".

The other major shift associated with vowels is high vowel deletion, where syllables containing /u/ or /i/ will be reduced to either a moraic obstruent or nasal. Thus, 犬 /inu/ becomes /iN/, while 小路 /sjo:di/ "lane~alley" (from /seudi/) becomes /sjuQ/. And just because Japanese didn't have enough homophones already: /kubi/ "neck", /kugi/ "nail", and /kutu/ "shoes" all reduce to /kuQ/.

For further details: http://www.geocities.jp/rtymg285/ and http://en.wikipedia.org/wiki/User:Io_Ka ... _Processes

Zhen Lin wrote:In Ryukyuan, standard Japanese /e e: o o:/ correspond to /i i: u u:/, but /ai au/ correspond to Ryukyuan /e: o:/. This appears to be a chain shift, but I'm not familiar with the diachronics on the Ryukyuan side. (The /au o:/ - /o: u:/ correspondence is slightly obscured by the fact that standard Japanese shifted /au/ to /o:/ as well.)

I'd say the full Ryukyuan shift goes roughly like this:

North:
/e/ → /1/ → /i/
/u/ → /1/ → /i/ after the alveolars /s/ and /t/
/o/ → /u/

South:
/i/ → /1/
/e/ → /i/
/o/ → /u/
(/1/ → /M/)

Both:
/ai/, /ae/ → /e:/ (→ /i:/)
/au/, /ao/ → /o:/ (→ /u:/)
/eu/ → /u:/

I'm not sure about other sequences, but that's essentially the bulk of it. In the Northern languages, Amami retains /1/, which merges in Yoron and Okinawan with /i/. In the southern languages, some variants have /1/, others /M/. Yonaguni merges /1/ and /i/, and then raises pretty much all mid vowels.

Other particularities include an aspiration distinction in Amami which, iirc, stems from the palatal or non-palatal features of the vowels /e/ and /i/. Some Amami variants are reported to have more vowels, but I haven't seen any studies on them; chances are they just add some nasalization. Okinawan has word-initial vowel glottalization, where the high vowels /i/ and /u/ will be lost if they're immediately followed their approximant counterparts: /ija/ [?ija] → /?ja/ "you". It also features coalescence when the topic marker /ja/ follows certain long vowels: /i:/ + /ja/ → /e:/, /u:/ + /ja/ → /o:/. Miyako also has an additional s-like vowel characterized by the symbol /ɿ/, which seems to have, at least partly, historically arisen from vowel devoicing.

Thanks for all your replies!

Hakaku wrote:I'd say the full Ryukyuan shift goes roughly like this:

North:
/e/ → /1/ → /i/
/u/ → /1/ → /i/ after the alveolars /s/ and /t/
/o/ → /u/

South:
/i/ → /1/
/e/ → /i/
/o/ → /u/
(/1/ → /M/)

Both:
/ai/, /ae/ → /e:/ (→ /i:/)
/au/, /ao/ → /o:/ (→ /u:/)
/eu/ → /u:/

Do these shifts apply accross the board to all vowels in any environment or do they only occur in certain syllables? The reason I'm asking is because it's been suggested that vowel chain shifts only normally occur in languages that have somekind of a stress accent and especially in languages with very strong stress accent (such as the Germanic languages). Likewise when vowel chain shifts do take place, they only affect the stressed syllables. So I'm wondering if your Japanese shifts targetted certain (prominent?) syllables in particular. Also I do not believe Japanese has lexical stress accent at all although it has pitch. Does Japanese perhaps have somekind of phrasal stress accent then?

Tropylium wrote: I've posted some more eastern Uralic shifts in the correspondence thread, but here's a few again. Main developments only, not specific secondary splits.
Proto-Uralic had *i *ü *u *e *ë *o *ä *a, plus some *Və, *Vj and *Vw sequences.

I never seen this *Və sequence before (aside of course from reading your excellent work on Frathwiki). Is this *ə what older uralic literature used to label as *x which was conceived to probably be somekind of laryngeal sound?

Tropylium wrote:Some sequences
*iə *eə *äə> ja
*uə *oə *åə> wa

Do these occur only when the diphthongs are not preceded by a consonant?

Aside from that I'll post later (perhaps tonight) a summary of the Cantonese vowel shift taken from that article zompist suggested I read.

LoneWolf wrote:I never seen this *Və sequence before (aside of course from reading your excellent work on Frathwiki). Is this *ə what older uralic literature used to label as *x which was conceived to probably be somekind of laryngeal sound?

It's not just older Uralic literature; I'm pretty sure it's still the most commonly accepted hypothesis. It's backed up by the fact that, intervocalically, the sound has developed into (*checks Wikipedia*) *k in Samic, *j in Mordvinic and *ɣ in Ugric. Of course, you could argue that *x was found only in intervocalic position, and that the sound found before consonants was something else (such as *ə) - but for that, you need to reconstruct an extra phoneme (such as *ə) which would then have been found only before consonants. Which is certainly possible, but do we have enough evidence for this reconstruction to be preferable to the simpler one?

Xonen wrote:

LoneWolf wrote:I never seen this *Və sequence before (aside of course from reading your excellent work on Frathwiki). Is this *ə what older uralic literature used to label as *x which was conceived to probably be somekind of laryngeal sound?

It's not just older Uralic literature; I'm pretty sure it's still the most commonly accepted hypothesis. It's backed up by the fact that, intervocalically, the sound has developed into (*checks Wikipedia*) *k in Samic, *j in Mordvinic and *ɣ in Ugric. Of course, you could argue that *x was found only in intervocalic position, and that the sound found before consonants was something else (such as *ə) - but for that, you need to reconstruct an extra phoneme (such as *ə) which would then have been found only before consonants. Which is certainly possible, but do we have enough evidence for this reconstruction to be preferable to the simpler one?

I'm describing the vowel shifts phonetically here. [ə] can be simply treated as an allophone of /x/ when in the environment V_C (you may recall the IE laryngeals also have their vocalic allophones). That said, AFAIK there's no explicit evidence that intervocalic *x and this preconsonantal *x were the same segment; they're merely in complementary distribution.

LoneWolf wrote:

Tropylium wrote:Some sequences
*iə *eə *äə ja
*uə *oə *åə wa

Do these occur only when the diphthongs are not preceded by a consonant?

No, these are fully general. [j w] occur in Nganasan only in these difthongs (original *j *w > dʲ b). Coronal + j merges into a palatal consonant, tho.

LoneWolf wrote:Do these shifts apply accross the board to all vowels in any environment or do they only occur in certain syllables? The reason I'm asking is because it's been suggested that vowel chain shifts only normally occur in languages that have somekind of a stress accent and especially in languages with very strong stress accent (such as the Germanic languages). Likewise when vowel chain shifts do take place, they only affect the stressed syllables. So I'm wondering if your Japanese shifts targetted certain (prominent?) syllables in particular. Also I do not believe Japanese has lexical stress accent at all although it has pitch. Does Japanese perhaps have somekind of phrasal stress accent then?

In terms of Ryukyuan, the changes seem to be relatively consistent across the board. There are, however, a few words here and there that fall exception to the general rules, but it's unclear if pitch could be a driving factor at all. I know that in reconstructions attempting to understand the origin and similitude of the existential verbs 'oru' and 'iru' (Ryukyuan (b)uru and (b)iru), pitch is often suggested as the source of these variations. But the lack of such variability in other words brings doubt to the theory.

There aren't enough detailed studies on the Ryukyuan languages to rule out the possibility. As pitch accent is extremely variable across Japan, acting more like a register system or even tonal-like system in some southern areas, while being completely accentless in other regions. It may be possible for pitch to be a driving factor concerning the presence or lack of aspiration in Amami's set of plosive consonants, as well as those of Yonaguni, but I couldn't say for sure.

As promised here is a summary of the vowel shifts discussed in the article zompist linked. If anyone notices any mistakes or has suggestions please tell me. I'll add it to the correspondence library.

The Middle Chinese to modern Cantonese vowel shift

Source: Newman, J. 1983. Cantonese vowel shift. Cahiers de Linguistique Asie Orientale XII.1: 65-79.
Forms with asterisks are Middle Chinese (MC)

Prior to the Cantonese vowel shift the following sound changes took place:

1. Back assimilation: a>ɔ/__ ŋ/k

2. Vowel coloring:

uən/t>ɔn/t
yən/t>œn/t
iəŋ/k>iŋ/k

3. Diphthongization:

i>əi/{labio-velat}__
u>əu

e.g.: *liu ʽwillow’>liau>lau
*kui ʽexpensive’>kuai>kʷai

4. Open syllable lengthening: V>V:/__# e.g.: *ka ʽfamily’>ka:


The Cantonese vowel shift (also refered to as the inner-outer flip):

In the push chain account proposed by Newman, the Cantonese vowel shift was triggered by the change *ə>a which in turn caused an original *a to either lengthen or raise to ə when after an onglide (it isn’t quite clear to me in what environments precisely *a lengthened or raised). The push chain had the following consequences on the diphthong system:

Diphthongs having arose by step 3 above:

əi>ai
əu>au
iəu>au (the glide deletes by effect of the rule below)

Original MC diphthongs:

*ai>a:i
*au>a:u
*iau>iəu>iu (happens after the ə deletes, see rule below)

Other unrelated changes which have an effect on the overal outcome of the shifted forms:

ə>∅/glide__ e.g.: *pian ʽwhip’>piən>pin, c.f. *piən ʽguest’>pian>pan

glide>∅/C__ e.g.: *liəm ʽforest’>liam>lam