Japonic family

[linguistcat]

This chart is sourced from "The Japanese Language Through Time" by Samuel E. Martin, copyright 1987, and ranges from early 700 CE to 1500 CE.

Screenshot (51).png (135.07 KiB) Viewed 21419 times

I am going to have to interpret this a little, so I'm going to say that going rightward, when going from a more solid line to anything less solid, or when going from no line to any amount of line, I will count that as the sound change starting and I will list the changes based on those dates. I don't always have the exact environments in which these would have happened, but I'll do my best to give them where needed. <N> marks the moraic nasal and <T> marks the consonant lengthening mora. # is the usual word boundary marker and $ marks a mora boundary.

ETA1: Edited so that syllables with * preceding them are notated in the standard V1/V2 format and assumed values according to Martin's reconstruction are below the sound changes for reference.
ETA2: Added the "starting phonology" that Martin seems to be working before the vowel values.

mwo,mo/mo
Introduction of #b, #d, #g, #z, #r from Middle Chinese
*to1,*to2/to
fa,wa/wa: V_
*ko1,*ko2/ko
*fi1,*fi2/fi
#n(i/u)[f,t,s,k]/#[b,d,z,g] (sporadic, mostly in particles and common verbs)*
*ki1,*ki2/ki; *ke1,*ke2/kye; *fe1,*fe2/fye
-ki#/-i# (in modern -i adjectives); -ki/-i:_te/ta (in yodan verbs)
-mi,-ni,-bi/N: _te, _ta; -mi,-mu,-ni,-nu,-bi,-bu/N (sporadic)
-fi-,-ti-,-ri-/T: _t(e/a): -f(i/u),-t(i/u),-k(i/u)/T: _(f,t,s,k) (and Tf/p:)
-ku#/-u# (in continuative form of -i adjectives)
kuwi/kwi; kuwe/kwe (some instances also borrowed from Middle Chinese)
wo,o/o
kuwa/kwa (some instances also borrowed from Middle Chinese)
*e1,*e2/ye
-fe/-we; -fi/-wi; -fu/-u
-we/-ye; -wi/-i
-fo,-wo/-wo
$we,$ye/$ye; $wi,$i/$i**
kwi,ki/ki; kwe,kye/kye
tu/tsu; ti/tSi

m n - t k f s r j w (while not mentioned, p may have been a marginal phoneme, or might have been an allophone of /f/)

i1 = yi, i2 = iy, e1 = ye, e2 =ey, o1 = wo, o2= o

*This may have also happened with #m(i/u) but I can’t find or remember any examples
** There was about a hundred years between this and the next change, hence why I kept them separate

If there are any questions or I need to reformat something to make the sound change clearer, let me know. These are not exhaustive because I know a few "onbin" changes are not accounted in all of this. And yes, most of these changes were unconditional, and others I can't find conditions for but know they didn't occur everywhere.

[bradrn]

Honestly, looking just at that image, I’m not sure it’s even possible to say much about specific sound changes with any confidence. e.g. for the first row, all we can tell is that kiy and kyi merged into something… but what? Just looking at it, I’m tempted to say ki, but I can’t know if this is true or not.

But I managed to find a copy of the book, and it looks like there’s a fair amount of information there — it’s somewhat too dense for me to understand, not being familiar with Japanese at all, but to what extent does it help to disambiguate that chart?

[linguistcat]

Honestly, looking just at that image, I’m not sure it’s even possible to say much about specific sound changes with any confidence. e.g. for the first row, all we can tell is that kiy and kyi merged into something… but what? Just looking at it, I’m tempted to say ki, but I can’t know if this is true or not.

Ah yeah, I was using the chart more for the timing of changes (when certain sound changes happened), and not the changes themselves. It's generally known in Japonic circles that Old Japanese had certain vowels, often i1/i2, e1/e2, and o1/o2 in English (and sometimes a neutral value for word-initial position and after consonants that never distinguished the two variants), that merged to modern /i/, /e/, and /o/. Different linguists have assumed different values for these vowels. But I was more concerned about when these vowels merged after which consonants.

I also figured I'd use the reconstruction of vowels that Martin used, since it at least gives us some guess at what those vowels were, which might give a better idea what sound changes might make sense in a similar situation. I'll also look through again to see if I can find some of the sound changes I know weren't present in this chart.

[bradrn]

Honestly, looking just at that image, I’m not sure it’s even possible to say much about specific sound changes with any confidence. e.g. for the first row, all we can tell is that kiy and kyi merged into something… but what? Just looking at it, I’m tempted to say ki, but I can’t know if this is true or not.

Ah yeah, I was using the chart more for the timing of changes (when certain sound changes happened), and not the changes themselves.

Ah, OK. (Though note that absolute dating is probably not of interest for this project; even relative dating isn’t critical. My Polynesian source barely gives any ordering at all for sound changes. Of course, the more information the better!)

It's generally known in Japonic circles that Old Japanese had certain vowels, often i1/i2, e1/e2, and o1/o2 in English (and sometimes a neutral value for word-initial position and after consonants that never distinguished the two variants), that merged to modern /i/, /e/, and /o/. Different linguists have assumed different values for these vowels. But I was more concerned about when these vowels merged after which consonants.

Yes, I’m aware of this. I imagine that we’ll eventually land up with several different sources, which broadly agree on the sound changes but disagree on the reconstruction.

I also figured I'd use the reconstruction of vowels that Martin used, since it at least gives us some guess at what those vowels were, which might give a better idea what sound changes might make sense in a similar situation.

What I’ve done in my Polynesian changes is to use the asterisked symbol throughout, and then give the reconstruction in a table associated with each source. So following this pattern, you could simply specify sound changes like e.g. *o1 → o (or whatever it might be) , then give *o1 = [wo] when citing Martin. Other sources can then have other reconstructions.

[linguistcat]

What I’ve done in my Polynesian changes is to use the asterisked symbol throughout, and then give the reconstruction in a table associated with each source. So following this pattern, you could simply specify sound changes like e.g. *o1 → o (or whatever it might be) , then give *o1 = [wo] when citing Martin. Other sources can then have other reconstructions.

I've edited the post to reflect this usage and am planning to start on collecting changes from the next reference.

[bradrn]

What I’ve done in my Polynesian changes is to use the asterisked symbol throughout, and then give the reconstruction in a table associated with each source. So following this pattern, you could simply specify sound changes like e.g. *o1 → o (or whatever it might be) , then give *o1 = [wo] when citing Martin. Other sources can then have other reconstructions.

I've edited the post to reflect this usage and am planning to start on collecting changes from the next reference.

Great!

I’d be interested to get your thoughts on how to treat multiple sources. So far I’ve been keeping sound changes from different sources entirely separate, on the basis that they might not be entirely compatible with each other (e.g. they might assume different reconstructions, or it might be impossible to order them sensibly). But on the other hand, we should expect different sources to agree on many sound changes, so keeping them entirely separate would result in lots of annoying duplicate changes. Do you have any suggestions as to how this should be handled?

[linguistcat]

I’d be interested to get your thoughts on how to treat multiple sources. ... Do you have any suggestions as to how this should be handled?

On one hand, keeping reconstructions and the sound changes together would be ideal in most cases, especially if there are large discrepancies. In the case of Japonic, the largest discrepancies in both reconstruction and sound changes are in Proto-Japonic to the Ryukyuan languages and Old Japanese. I'm saving all that until I get the basic sound changes from Old Japanese onward, but those should agree for the most part. Old Japanese to any other stage of the language is a bit more straight forward and we have some written records to help us, even if exact values for some phonemes are in question.

That said, some reconstructions are more like Old Japanese and generally accepted with only minor details differing, and it would feel redundant to include the reconstruction over and over again. But I have a feeling most of the languages people will work with are less agreed on.

[Man in Space]

OK. Busting out my edition of Thorpe (1983), Ryūkyūan Language History.

Going to set out the abbreviations here:

Proto-Ryūkyūan: PR (I differ from Thorpe here, but for sanity's sake I follow the remaining abbreviations as he posited them)
Hateruma: Hateruma (H)
Hatoma: Hatoma (Ha.)
Ieshima: Ieshima (I)
Ikema: Ikema (Ik.)
Irabu: Sarahama (Sa.)
Iriomote: Sonai (So.)
Ishigaki: Ishigaki (Is.), Kābira (K), Obama (Ob.)
Kakeroma: Shodon (Sh.)
Kikai: Aden (A), Shidōke (Shi.)
Kobama: Kobama (Kob.)
Kuroshima: Kuroshima (Ku.)
Miyako: Ōura (O), Uechi (U), Yonaha (Yo.)
Ōgami: Ōgami (Og.)
Okierabu: Hentona (He.), Oku (Ok.), Nago (Nag.), Naha (Na.), Sakimotobu (Sak.), Shuri (S), Sumuide (Su.), Yonamine (Yon.)
Ōshima: Koniya (Ko.), Naze (N), Yamatoma (Ya.), Yuwan (Yu.)
Taketomi: Taketomi (T)
Tarama: Nakasuji (Nak.)
Tokunoshima: Kametsu (Ka.), Inō (In.), Matsubara (M), San (San)
Yonaguni: Yonaguni (Y)

Reconstructed Phonological System of PR

*m *n
*p *b *t *tʰ¹ *d *k *kʰ¹ *g
*s *z²
*w *r *j
*Q³
*N⁴

1. ". . .the aspirated consonants. . .of early loanwords from Japanese".
2. Based on other stuff Thorpe says, *z may have phonetically been [*dz].
3. Thorpe calls this "the syllabic consonant"; when it precedes an obstruent, it assimilates in four ways: ± anterior, ± coronal, ± voice, ± continuant.
4. This he terms "the syllabic nasal", which assimilates in two ways to a following consonant (not just obstruents): ± anterior, ± coronal.

*a *e *i *o *u¹

1. Thorpe states that PR *u was what he calls *ï when after a coronal obstruent; based on his distinctive features, it looks like this was realized as [*ɯ].

Some other salient tidbits:
- *t *d "are affricated before high vowels", and these plus *s *z "are palatalized before" the high front vowel; it seems that this meant [*ts *dz] before *u, [*tʃ *dʒ *ʃ *(d)ʒ] before *i (I'm not clear on whether *z was actually [*dz]).
- I'm just going to copy him verbatim here: "A non-sonorant is voiced in a third or subsequent mora, provided it is not preceded directly by a de-voiced (high) vowel."
- Clusters were disallowed unless bracketed on both ends by [+ syll] segments (vowels or syllabic consonants).

I'll try to get more done in the next few days; there's a couple hundred pages of dissertation to go through and I've a busy week ahead of me.

[linguistcat]

Very nice. This reminds me that I need to go back to my original post and edit in the reconstruction for Old Japanese pronunciation.

[Ketsuban]

Frellesvig (2010) is mostly concerned with attested historical forms (i.e. Old Japanese and later) but does devote a couple of pages to what internal reconstruction can tell us about the ancestor of Old Japanese, which he calls Proto-Japanese, including an "approximate relative chronology of the changes that took place between [Proto-Japanese] and [Old Japanese]".

The segments reconstructed for Proto-Japanese are the consonants *p *t *k *s *m *n *r *w *y and vowels *i *e *a *o *u *ɨ *ə. The only symbol that deviates from the IPA is *y [*j].

Frellesvig writes the Old Japanese kō-otsu distinction like this.

Standard	Frellesvig
i₁	i
i₂	wi
e₁	ye
e₂	e
o₁	wo
o₂	o

1. Contraction of falling diphthongs
   • *ui > wi
   • *ɨi > wi
   • *əi > e
   • *ai > e
2. Merger of central vowels: {*ɨ, *ə} > *ə
3. Mid-vowel raising: *e > ye word-finally, otherwise *i
4. Contraction of rising diphthongs
   • *ia > *ya > -ye
   • *iə > -ye
5. Mid-vowel raising: *o > -wo word-finally, otherwise *u
6. Backing of central vowel: *ə > o
7. Contraction of rising diphthongs
   • *ua > *wa > -wo
   • *uo > -wo

[bradrn]

*Q³
*N⁴

3. Thorpe calls this "the syllabic consonant"; when it precedes an obstruent, it assimilates in four ways: ± anterior, ± coronal, ± voice, ± continuant.
4. This he terms "the syllabic nasal", which assimilates in two ways to a following consonant (not just obstruents): ± anterior, ± coronal.

This is actually a very interesting case which I hadn’t considered. As discussed in the other threads, I’ve taken the approach of mapping transcribed consonants to their IPA realisations — but of course that’s not possible here, since these don’t have an IPA realisation. I’m not sure what else to do, though. Perhaps we could simply call them ‘[+consonant]’ and ‘[+nasal]’?

I'll try to get more done in the next few days; there's a couple hundred pages of dissertation to go through and I've a busy week ahead of me.

Fair enough! I’m busy too, given that my thesis is due in three weeks…

[Man in Space]

*Q³
*N⁴

3. Thorpe calls this "the syllabic consonant"; when it precedes an obstruent, it assimilates in four ways: ± anterior, ± coronal, ± voice, ± continuant.
4. This he terms "the syllabic nasal", which assimilates in two ways to a following consonant (not just obstruents): ± anterior, ± coronal.

This is actually a very interesting case which I hadn’t considered. As discussed in the other threads, I’ve taken the approach of mapping transcribed consonants to their IPA realisations — but of course that’s not possible here, since these don’t have an IPA realisation. I’m not sure what else to do, though. Perhaps we could simply call them ‘[+consonant]’ and ‘[+nasal]’?

One option is to star out features that don’t apply—phonix operates like this, giving things like [+cons *vc *ro *son].

[bradrn]

*Q³
*N⁴

3. Thorpe calls this "the syllabic consonant"; when it precedes an obstruent, it assimilates in four ways: ± anterior, ± coronal, ± voice, ± continuant.
4. This he terms "the syllabic nasal", which assimilates in two ways to a following consonant (not just obstruents): ± anterior, ± coronal.

This is actually a very interesting case which I hadn’t considered. As discussed in the other threads, I’ve taken the approach of mapping transcribed consonants to their IPA realisations — but of course that’s not possible here, since these don’t have an IPA realisation. I’m not sure what else to do, though. Perhaps we could simply call them ‘[+consonant]’ and ‘[+nasal]’?

One option is to star out features that don’t apply—phonix operates like this, giving things like [+cons *vc *ro *son].

Hmm, interesting approach. (Though given that we’re not limited to ASCII, I’d use ± rather than *.)

[Man in Space]

This is actually a very interesting case which I hadn’t considered. As discussed in the other threads, I’ve taken the approach of mapping transcribed consonants to their IPA realisations — but of course that’s not possible here, since these don’t have an IPA realisation. I’m not sure what else to do, though. Perhaps we could simply call them ‘[+consonant]’ and ‘[+nasal]’?

One option is to star out features that don’t apply—phonix operates like this, giving things like [+cons *vc *ro *son].

Hmm, interesting approach. (Though given that we’re not limited to ASCII, I’d use ± rather than *.)

In phonix the star specifically means “unspecified for”. The plus/minus indicates that the feature is specified. Put another way, the plus/minus is yes-or-no and the star is mu.

[Man in Space]

So the rules might look like this:

Q: [+cons +syll *ant *cor *vc *cont] > [±ant ±cor ±vc ±cont] / _[±ant ±cor ±vc ±cont]
N: [+cons +syll +nas *ant *cor] > [±ant ±cor] / _[±ant ±cor]

[bradrn]

So the rules might look like this:

Q: [+cons +syll *ant *cor *vc *cont] > [±ant ±cor ±vc ±cont] / _[±ant ±cor ±vc ±cont]
N: [+cons +syll +nas *ant *cor] > [±ant ±cor] / _[±ant ±cor]

That’s not what I was thinking, though. What I was getting at is that /Q/ and /N/ are phonemes, even ones preserved in modern Japonic languages. So, since they’re not IPA, we’d need a way to record their realisations.

[Man in Space]

So the rules might look like this:

Q: [+cons +syll *ant *cor *vc *cont] > [±ant ±cor ±vc ±cont] / _[±ant ±cor ±vc ±cont]
N: [+cons +syll +nas *ant *cor] > [±ant ±cor] / _[±ant ±cor]

That’s not what I was thinking, though. What I was getting at is that /Q/ and /N/ are phonemes, even ones preserved in modern Japonic languages. So, since they’re not IPA, we’d need a way to record their realisations.

OK, thanks for clearing that up. In that case the plus/minus might be a good solution.

[Man in Space]

Actually…maybe a question mark?

[bradrn]

Actually…maybe a question mark?

Another good idea! I’m still leaning towards ±, but it’s hard to know.

[BGMan]

Standard	Frellesvig
i₁	i
i₂	wi
e₁	ye
e₂	e
o₁	wo
o₂	o

I've wondered for some time if Old Japanese had a more Korean-like vowel system?

I do see cases like this video here, for example https://www.youtube.com/watch?v=4mWzhIcbQls where Old Japanese is interpreted as having a more Russian-like vowel system, with i, ï, e, ë, o more or less equivalent to Russian и, ы, е, э, о respectively... although they pronounce ö with a strange pinched pronounciation (similar to /ɤ/?) which IMO sounds horrible. (Personally, I would've expected something more like Korean eo ㅓ /ʌ/ for ö, since it's so common and a more relaxed pronounciation would be expected for such a common vowel.