Sound Change Quickie Thread

[Nortaneous]

From what I can see, Proto-Germanic had consistent initial stress (though I’m very unfamiliar with the language, so please correct me if that is incorrect). This makes the syncope rule consistent — in disyllabic words, it is always the second vowel which is deleted. By contrast, my language has weight-sensitive stress, which usually ends up on the first syllable, but not always. Is posttonic vowel syncope still plausible with this stress rule, or does that make it implausible?

I don't see any reason why it would be implausible.

Tocharian A had movable stress and such extensive syncope that *ɨ became epenthetic - then again, syncope in Tocharian was totally independent of stress, and syncope had no problem deleting stressed vowels - IIRC even in TB where we know what the stress system was and can be pretty sure there was no independent stage with a different stress system, although I might be wrong about that.

Along related lines, what happens in multisyllabic words? e.g. would *tapakar with initial stress turn into tapkar, or tapakx, or what? What about long words with secondarily-stressed syllables?

Depends on how secondary stress is handled.

You list a bunch of ways in which final consonant clusters get reduced. This is of particular interest of me, as one of the biggest problems I’ve been having has been getting rid of final clusters in a nice way, without messing up the paradigms too much. (Or, more accurately, my whole goal with these sound changes to mess up some paradigms horribly, but keep the rest sane.) Do you happen to have any source going into more detail about exactly how the cluster reduction proceeded in the various Germanic languages? I feel this would help me get a better sense of what options I have.

It mostly didn't. I made that all up. Germanic tends to keep its 'postsyllables' separate from its main syllables, and not to have much interaction between the two - the main exception I know of is German -Pən -Tən -Kən > -pm̩ -tn̩ -kŋ̩, which I think can simplify into plain -m -n -ŋ. (You could probably have -pən -bən > -ˀm -m or something if you wanted.) In English, postsyllables are mostly separate, but grammatical words can have tighter binding: isn't [ɪdn̩] but business [bɪznɨs], of them [ʌb̪ɱ̩] but oven [ʌvən]. But given that [bɪdnɨs] is also attested, this is plausibly a result of dialect mixing.

But what interacts with what is basically a set of decisions that languages can make. (For an extreme example, consider the Paman languages where word-initial consonants were transphonologized onto second-syllable vowels.) In Sinitic, presyllables didn't interact with main syllables much and were mostly lost without compensation; in Tangut, presyllables were mostly transphonologized onto the vowel of the main syllable; and in some Mon-Khmer languages (like Nyaheun and Kriang) presyllables mostly interacted with initial consonants. I think the tricky part is that initials usually are more prominent (and carry more information) than codas - AmEng, with its (partially morphology-driven) unusually large codas and its preferential coda syllabification (e.g. lawyer), is extremely unusual here. (For extreme examples in the other, more usual direction, consider Georgian - does Georgian allow coda clusters? - or Wutung, which can tolerate [highly structured] initial clusters of up to four consonants but which doesn't allow coda consonants at all, except in three or so words where you can sort of get a word-internal syllable coda -n.)

What you're proposing is weird but doable, and it would probably have implications for the structure of the language in general.

[Zju]

(For an extreme example, consider the Paman languages where word-initial consonants were transphonologized onto second-syllable vowels.)

Wait, whaat? How did that work out, exactly?

[Nortaneous]

(For an extreme example, consider the Paman languages where word-initial consonants were transphonologized onto second-syllable vowels.)

Wait, whaat? How did that work out, exactly?

https://en.wikipedia.org/wiki/Mbabaram_ ... e#Vowels_2

[bradrn]

From what I can see, Proto-Germanic had consistent initial stress (though I’m very unfamiliar with the language, so please correct me if that is incorrect). This makes the syncope rule consistent — in disyllabic words, it is always the second vowel which is deleted. By contrast, my language has weight-sensitive stress, which usually ends up on the first syllable, but not always. Is posttonic vowel syncope still plausible with this stress rule, or does that make it implausible?

I don't see any reason why it would be implausible.

Tocharian A had movable stress and such extensive syncope that *ɨ became epenthetic - then again, syncope in Tocharian was totally independent of stress, and syncope had no problem deleting stressed vowels - IIRC even in TB where we know what the stress system was and can be pretty sure there was no independent stage with a different stress system, although I might be wrong about that.

Along related lines, what happens in multisyllabic words? e.g. would *tapakar with initial stress turn into tapkar, or tapakx, or what? What about long words with secondarily-stressed syllables?

Depends on how secondary stress is handled.

Interesting, thanks! I had not known that Tocharian A had epenthetic [ɨ] — I’m more used to seeing that in Papuan languages.

You list a bunch of ways in which final consonant clusters get reduced. This is of particular interest of me, as one of the biggest problems I’ve been having has been getting rid of final clusters in a nice way, without messing up the paradigms too much. (Or, more accurately, my whole goal with these sound changes to mess up some paradigms horribly, but keep the rest sane.) Do you happen to have any source going into more detail about exactly how the cluster reduction proceeded in the various Germanic languages? I feel this would help me get a better sense of what options I have.

It mostly didn't. I made that all up.

Hmm, OK. That would explain why I couldn’t find any information about this when I looked it up. (But then why did you decide to call those examples ‘Fake German’ etc.? And do the ‘Fake Sino-Tibetan’ examples correspond any more to reality?)

Germanic tends to keep its 'postsyllables' separate from its main syllables, and not to have much interaction between the two - the main exception I know of is German -Pən -Tən -Kən > -pm̩ -tn̩ -kŋ̩, which I think can simplify into plain -m -n -ŋ. (You could probably have -pən -bən > -ˀm -m or something if you wanted.) In English, postsyllables are mostly separate, but grammatical words can have tighter binding: isn't [ɪdn̩] but business [bɪznɨs], of them [ʌb̪ɱ̩] but oven [ʌvən]. But given that [bɪdnɨs] is also attested, this is plausibly a result of dialect mixing.

Do you know if this a consequence of some general rule that makes ‘postsyllables’ less amenable to reduction than ‘presyllables’, or is it purely accidental that postsyllables happen to remain separate in Germanic?

But what interacts with what is basically a set of decisions that languages can make. … I think the tricky part is that initials usually are more prominent (and carry more information) than codas - AmEng, with its (partially morphology-driven) unusually large codas and its preferential coda syllabification (e.g. lawyer), is extremely unusual here.

So, thinking about this with respect to diachronics, what sort of impact does (if any) this tendency have on which sound changes are preferred?

What you're proposing is weird but doable, and it would probably have implications for the structure of the language in general.

Good to know! And yes, it does have some extremely large implications on the structure — in particular, stem alternations become entirely opaque, with the later change then simplifying some other paradigms back down to being synchronically nice. (The whole thing is part of a change from being disyllabic+isolating to being monosyllabic+agglutinating, with a concurrent increase in the vowel system size. It’s effectively going from CEMP to Hup.)

[Nortaneous]

(But then why did you decide to call those examples ‘Fake German’ etc.?

Might as well call them something.

And do the ‘Fake Sino-Tibetan’ examples correspond any more to reality?)

Yes, those basically follow the patterns of development in ST, although I'm probably wrong about Sikkimese.

Do you know if this a consequence of some general rule that makes ‘postsyllables’ less amenable to reduction than ‘presyllables’, or is it purely accidental that postsyllables happen to remain separate in Germanic?

That's an open question. I think it's probably accidental. There just aren't many examples of postsyllables. When they develop in ST, they don't look anything like Germanic postsyllables or ST presyllables - the postsyllable inventory is small enough to be a morphological matter more than a phonological one, I think - but they tend to fuse with main syllables. Mandarin erhua is the best-known example but there are others in Tibetic.

So, thinking about this with respect to diachronics, what sort of impact does (if any) this tendency have on which sound changes are preferred?

Well, what interacts with what? If something is transphonologized, what's it transphonologized onto? In Tangut, preinitials are transphonologized onto main vowels - *s.CV *l.CV > CV̰, *r.CV > CV˞ , etc. - but you could just as well imagine something where they affect initials instead. Same for mergers - there are a lot of different sound changes affecting vowels in American English, but when they're conditioned there's a general tendency for them to be conditioned by /l r g ŋ m n/ (voiced consonants that are either nasal or dorsal? you need velarization in the feature specifications for /l r/ for that tho), and maybe in another language they're mostly conditioned by prevocalic semivowels or the POA of the initial or nasal initials or whatever. In Turkic and Armenian you get interactions between initial consonant voicing and vowel frontness (or something - I'm not actually clear on how it works in Turkic) that are totally weird everywhere else.

[bradrn]

(But then why did you decide to call those examples ‘Fake German’ etc.?

Might as well call them something.

True, that. (Though please point this out next time you do this!)

Do you know if this a consequence of some general rule that makes ‘postsyllables’ less amenable to reduction than ‘presyllables’, or is it purely accidental that postsyllables happen to remain separate in Germanic?

That's an open question. I think it's probably accidental. There just aren't many examples of postsyllables. When they develop in ST, they don't look anything like Germanic postsyllables or ST presyllables - the postsyllable inventory is small enough to be a morphological matter more than a phonological one, I think - but they tend to fuse with main syllables. Mandarin erhua is the best-known example but there are others in Tibetic.

Sounds interesting… I know about erhua, but could you give some examples from Tibetic?

So, thinking about this with respect to diachronics, what sort of impact does (if any) this tendency have on which sound changes are preferred?

Well, what interacts with what? If something is transphonologized, what's it transphonologized onto? In Tangut, preinitials are transphonologized onto main vowels - *s.CV *l.CV > CV̰, *r.CV > CV˞ , etc. - but you could just as well imagine something where they affect initials instead. Same for mergers - there are a lot of different sound changes affecting vowels in American English, but when they're conditioned there's a general tendency for them to be conditioned by /l r g ŋ m n/ (voiced consonants that are either nasal or dorsal? you need velarization in the feature specifications for /l r/ for that tho), and maybe in another language they're mostly conditioned by prevocalic semivowels or the POA of the initial or nasal initials or whatever. In Turkic and Armenian you get interactions between initial consonant voicing and vowel frontness (or something - I'm not actually clear on how it works in Turkic) that are totally weird everywhere else.

I must admit, I’m not quite sure what your point is here. Alright, Tangut/Turkic/Armenian seem like fairly straightforward examples of initials transphonologising onto vowels — which makes sense, given the aforementioned tendency for initials to be more prominent — but unless I’m missing something, the conditioning factor in American English seems to be final consonants rather than initial ones, which doesn’t quite seem to fit with your other examples.

(By the way, I’m think your hypothetical feature specifications might actually work quite well for my own English, in which /l/ at least is always velarised. /r/ is a bit trickier; I believe my usual realisation is [ɻʷ], but I’m pretty sure I’ve seen [ʋ] and even [ʕʷ~ʕ͡β]. At any rate, there’s usually some sort of dorsality there.)

[Qwynegold]

I've had trouble explaining so I'm just going to use examples. This is Standard Japanese and well attested as occurring in the Muromachi period, which is part of why I had difficulty. Another reason is that it seems to do two things that sound changes are said not to do, namely, act on specific parts of speech (verbs and verbal adjectives), and that one of the sound changes reverses itself with the exception of a few words.

Ah, I see. I would've said that that looks like a morphology change rather than a sound change, but Pabappa seemed to know more.

Now, if you want to copy the same for /t/, then you could do t > D > 0. But if your conlang is very similar to Japanese, you wouldn't have any [ti], only [ts\i]...

[linguistcat]

Now, if you want to copy the same for /t/, then you could do t > D > 0. But if your conlang is very similar to Japanese, you wouldn't have any [ti], only [ts\i]...

It's based somewhat on Old Japanese, which might not have had [ts\] or [ts] as allophones of /t/ yet : D so it's not a concern.

[Nortaneous]

Sounds interesting… I know about erhua, but could you give some examples from Tibetic?

From A phonological history of Amdo Tibetan rhymes (doi:10.1017/s0041977x16000070):

Code: Select all

  0 -ba -bo
a a  ɑ   u
i ə  ɛ   
u ə  ɔ   u 
e e  ɛ    
o o  ɔ   u

So the Written Tibetan "postsyllables" -ba and -bo merge with open main syllables to produce correspondence patterns that aren't seen with any initial + final combination. (There aren't enough reflexes of -i-ba -e-ba to tell what those become.) There was probably an intermediate step with -ba > -wa, so this can be seen as a process of 'smoothing' like the fire-far merger.

I must admit, I’m not quite sure what your point is here. Alright, Tangut/Turkic/Armenian seem like fairly straightforward examples of initials transphonologising onto vowels — which makes sense, given the aforementioned tendency for initials to be more prominent — but unless I’m missing something, the conditioning factor in American English seems to be final consonants rather than initial ones, which doesn’t quite seem to fit with your other examples.

Right - finals. In two cases this is just ejection of -j-: the Californian raising of KIT to FLEECE before /ŋ/ (iŋ > ijŋ) and the occasional AmEng raising of DRESS to FACE before /g ŋ/ (eg > ejg), as well as the strictly phonetic æ > æj / _g _ŋ (which demonstrates an intervocalic /nK ŋK/ contrast - "panko" and "Bernanke" can have [eə] rather than [æj]). But also pin-pen, poor-pour(? - I actually have no idea which words have which), NORTH-FORCE, marry-merry-Mary (caused by the development of the rule of preferential coda syllabification of glides), various lax-tense mergers before /l/, etc. Preferential coda syllabification can arguably be taken as a case of extreme 'binding' between vowels and following glides (and less so resonants more generally - /j r w/ seem more tightly bound than /l/, which seems more tightly bound than /m n/... /g/ is somewhere in the middle despite not being a resonant, but it is prone enough to lenition that maybe we could call it /ɰ/

[Travis B.]

It mostly didn't. I made that all up. Germanic tends to keep its 'postsyllables' separate from its main syllables, and not to have much interaction between the two - the main exception I know of is German -Pən -Tən -Kən > -pm̩ -tn̩ -kŋ̩, which I think can simplify into plain -m -n -ŋ. (You could probably have -pən -bən > -ˀm -m or something if you wanted.) In English, postsyllables are mostly separate, but grammatical words can have tighter binding: isn't [ɪdn̩] but business [bɪznɨs], of them [ʌb̪ɱ̩] but oven [ʌvən]. But given that [bɪdnɨs] is also attested, this is plausibly a result of dialect mixing.

The English here has postsyllable reduction into [mː] and [nː] in words like problem, sudden, couldn't, and so on; note that the preceding vowel remains unnasalized, indicating that these are underlyingly still /bm/ and /dn/.

[bradrn]

Sounds interesting… I know about erhua, but could you give some examples from Tibetic?

From A phonological history of Amdo Tibetan rhymes (doi:10.1017/s0041977x16000070):

Code: Select all

  0 -ba -bo
a a  ɑ   u
i ə  ɛ   
u ə  ɔ   u 
e e  ɛ    
o o  ɔ   u

So the Written Tibetan "postsyllables" -ba and -bo merge with open main syllables to produce correspondence patterns that aren't seen with any initial + final combination. (There aren't enough reflexes of -i-ba -e-ba to tell what those become.) There was probably an intermediate step with -ba > -wa, so this can be seen as a process of 'smoothing' like the fire-far merger.

Thanks for the link! It looks really interesting after quickly reading through it; I shall have to look through it more carefully later.

I must admit, I’m not quite sure what your point is here. Alright, Tangut/Turkic/Armenian seem like fairly straightforward examples of initials transphonologising onto vowels — which makes sense, given the aforementioned tendency for initials to be more prominent — but unless I’m missing something, the conditioning factor in American English seems to be final consonants rather than initial ones, which doesn’t quite seem to fit with your other examples.

Right - finals. In two cases this is just ejection of -j-: the Californian raising of KIT to FLEECE before /ŋ/ (iŋ > ijŋ) and the occasional AmEng raising of DRESS to FACE before /g ŋ/ (eg > ejg), as well as the strictly phonetic æ > æj / _g _ŋ (which demonstrates an intervocalic /nK ŋK/ contrast - "panko" and "Bernanke" can have [eə] rather than [æj]). But also pin-pen, poor-pour(? - I actually have no idea which words have which), NORTH-FORCE, marry-merry-Mary (caused by the development of the rule of preferential coda syllabification of glides), various lax-tense mergers before /l/, etc. Preferential coda syllabification can arguably be taken as a case of extreme 'binding' between vowels and following glides (and less so resonants more generally - /j r w/ seem more tightly bound than /l/, which seems more tightly bound than /m n/... /g/ is somewhere in the middle despite not being a resonant, but it is prone enough to lenition that maybe we could call it /ɰ/

Hmm, OK, I think I may be starting to understand what you’re saying here. From what I can tell, it looks like that initials are usually more prominent than finals, so languages usually have vowel changes conditioned by initial choice, but finals are more prominent in American English, so English vowel changes tend to be conditioned by finals instead — is my understanding correct?

[Nortaneous]

Hmm, OK, I think I may be starting to understand what you’re saying here. From what I can tell, it looks like that initials are usually more prominent than finals, so languages usually have vowel changes conditioned by initial choice, but finals are more prominent in American English, so English vowel changes tend to be conditioned by finals instead — is my understanding correct?

Not exactly. I think finals affecting preceding vowels is more common (in languages that have consonant finals) than initials affecting following vowels. The unusual thing about American English is preferential coda syllabification: if you have a /CVCV/ sequence, in many circumstances this is syllabified as /CVC.V/ rather than the much more crosslinguistically common /CV.CV/. (I think this is always the case for the semivowels /j r w/?) This is the result of a conspiracy of sound changes to produce it, including:
- the orange shift: ɒ.rV > ɑr.V ~ or.V
- the lawyer shift: ɔ.jV > oj.V
- the mirror-nearer merger: ɪ.jV > ij.V
- the marry-merry-Mary merger: æ.rV ɛ.rV eɪ̯.rV > er.V

Conspiracies are a perfectly natural occurrence in diachronic phonology - the Slavic open syllable conspiracy is the best-known example, but as we've seen, there are others.

But the thing I was referencing is the 'decision' that English (not just AmEng) 'made' that certain coda consonants Are Things That Cause Mergers, regardless of the details of those mergers. Most conditioned vowel mergers in English dialects, I think, tend to involve /r l ŋ g m n/ (in roughly that order of prominence):
https://en.wikipedia.org/wiki/English-l ... storic_/l/
https://en.wikipedia.org/wiki/English-l ... storic_/r/
- the thank-think merger (some AAVE)
- KIT > FLEECE before ŋ (Californian)
- æ > æj before ŋ g (some AmEng)
- the extended [i.e. before /g ŋ/ - and apparently some /k/ for some people, although I don't have this] LOT-CLOTH split (AmEng)
- the egg-vague merger (AmEng - maybe only East Coast? and I can't think of an actual merged pair for this)
- the pen-pin merger
And probably others I've forgotten - but maybe /m n/ don't belong here, since I can't think of anything but pen-pin and GenAm æ-raising (which is really a simplification of TRAP-BATH, which was extended in the US to words like "ant").

At the same time, there are things that have happened elsewhere that would be basically unimaginable here, like the Paman transphonologization of initial consonants onto second-syllable vowels and Adjarian's Law.

The English here has postsyllable reduction into [mː] and [nː] in words like problem, sudden, couldn't, and so on; note that the preceding vowel remains unnasalized, indicating that these are underlyingly still [bm] and [dn].

Is that postsyllable reduction, or is it just reduction of a homorganic plosive + nasal sequence whether or not the nasal is syllabic? Grimes has [mː] for /bm/ in submit.

[bradrn]

Hmm, OK, I think I may be starting to understand what you’re saying here. From what I can tell, it looks like that initials are usually more prominent than finals, so languages usually have vowel changes conditioned by initial choice, but finals are more prominent in American English, so English vowel changes tend to be conditioned by finals instead — is my understanding correct?

Not exactly. I think finals affecting preceding vowels is more common (in languages that have consonant finals) than initials affecting following vowels. The unusual thing about American English is preferential coda syllabification: if you have a /CVCV/ sequence, in many circumstances this is syllabified as /CVC.V/ rather than the much more crosslinguistically common /CV.CV/. (I think this is always the case for the semivowels /j r w/?) This is the result of a conspiracy of sound changes to produce it, including:
- the orange shift: ɒ.rV > ɑr.V ~ or.V
- the lawyer shift: ɔ.jV > oj.V
- the mirror-nearer merger: ɪ.jV > ij.V
- the marry-merry-Mary merger: æ.rV ɛ.rV eɪ̯.rV > er.V

I’d heard about preferential coda syllabification before, but I hadn’t known about the sound changes that cause it. Do you know if this syllabification has been suggested for other English dialects or only for AmEng dialects in which these mergers have occurred?

Conspiracies are a perfectly natural occurrence in diachronic phonology - the Slavic open syllable conspiracy is the best-known example, but as we've seen, there are others.

Huh, I’ve never heard of the Slavic open syllable conspiracy, what is it? (The one I’m most familiar with is the tendency towards disyllabic roots in Austronesian.)

But the thing I was referencing is the 'decision' that English (not just AmEng) 'made' that certain coda consonants Are Things That Cause Mergers, regardless of the details of those mergers. Most conditioned vowel mergers in English dialects, I think, tend to involve /r l ŋ g m n/ (in roughly that order of prominence):
https://en.wikipedia.org/wiki/English-l ... storic_/l/
https://en.wikipedia.org/wiki/English-l ... storic_/r/
- the thank-think merger (some AAVE)
- KIT > FLEECE before ŋ (Californian)
- æ > æj before ŋ g (some AmEng)
- the extended [i.e. before /g ŋ/ - and apparently some /k/ for some people, although I don't have this] LOT-CLOTH split (AmEng)
- the egg-vague merger (AmEng - maybe only East Coast? and I can't think of an actual merged pair for this)
- the pen-pin merger
And probably others I've forgotten - but maybe /m n/ don't belong here, since I can't think of anything but pen-pin and GenAm æ-raising (which is really a simplification of TRAP-BATH, which was extended in the US to words like "ant").

OK, that makes sense now, thanks for clarifying! Indulging in a bit of speculation, it occurs to me that this tendency may have a simple explanation: large vowel systems are particularly prone to rearrangements, vowels are articulated in the same place as dorsals so interact most strongly with them, and as I noted earlier, /r l ŋ ɡ/ all have some sort of dorsal component to them, especially after a vowel (/m n/ are as you note less prominent in sound changes). Potentially this explanation may even have some predictive power… /j/ is the sole ‘missing’ voiced dorsal consonant from the above list, were there any sound changes conditioned by that one?

[Travis B.]

The English here has postsyllable reduction into [mː] and [nː] in words like problem, sudden, couldn't, and so on; note that the preceding vowel remains unnasalized, indicating that these are underlyingly still [bm] and [dn].

Is that postsyllable reduction, or is it just reduction of a homorganic plosive + nasal sequence whether or not the nasal is syllabic? Grimes has [mː] for /bm/ in submit.

I think it is just homorganic plosive + nasal sequence reduction whether or not the nasal is syllabic, actually, since I have it in both submit and good night outside of careful speech.

[Nortaneous]

I’d heard about preferential coda syllabification before, but I hadn’t known about the sound changes that cause it. Do you know if this syllabification has been suggested for other English dialects or only for AmEng dialects in which these mergers have occurred?

I think it's been suggested for RP to some extent, but it's clearly strongest in AmEng.

Huh, I’ve never heard of the Slavic open syllable conspiracy, what is it? (The one I’m most familiar with is the tendency towards disyllabic roots in Austronesian.)

here

Indulging in a bit of speculation, it occurs to me that this tendency may have a simple explanation: large vowel systems are particularly prone to rearrangements, vowels are articulated in the same place as dorsals so interact most strongly with them, and as I noted earlier, /r l ŋ ɡ/ all have some sort of dorsal component to them, especially after a vowel (/m n/ are as you note less prominent in sound changes).

Well, do similar things happen in Austroasiatic?

Potentially this explanation may even have some predictive power… /j/ is the sole ‘missing’ voiced dorsal consonant from the above list, were there any sound changes conditioned by that one?

/j w/ are both missing, but they can only occur in coda position due to resyllabification or the Great Vowel Shift, which in GenAm output vowel + semivowel sequences. Possibly this is because x > j~w / V_C created /ij uw/ sequences which merged with /i: u:/, but most of Germanic underwent some kind of raising shift in long vowels, didn't it?

[Richard W]

Indulging in a bit of speculation, it occurs to me that this tendency may have a simple explanation: large vowel systems are particularly prone to rearrangements, vowels are articulated in the same place as dorsals so interact most strongly with them, and as I noted earlier, /r l ŋ ɡ/ all have some sort of dorsal component to them, especially after a vowel (/m n/ are as you note less prominent in sound changes).

Well, do similar things happen in Austroasiatic?

The distribution of final palatals and velars is often complementary after some vowels, which evidences mutual influence. In Khmer, the reflexes of central low vowels depend on whether the final consonant is final or not. Whether final velar stops contrast with the a glottal stop also depends on the vowel in Khmer.

[Travis B.]

/j w/ are both missing, but they can only occur in coda position due to resyllabification or the Great Vowel Shift, which in GenAm output vowel + semivowel sequences. Possibly this is because x > j~w / V_C created /ij uw/ sequences which merged with /i: u:/, but most of Germanic underwent some kind of raising shift in long vowels, didn't it?

The English around here certainly doesn't have vowel + semivowel sequences corresponding to historical long vowels as you posit, aside from /aɪ/~/əɪ/, /aʊ/, and final and intervocalic [w] descended from historical /oʊ/ in final position or before another vowel respectively, and it appears that the English around is descended from a dialect pretty close to the dialect from which GA is descended, generations ago. (Remember that video that someone posted a link to with someone from Green Bay, WI who spoke pretty close to GA in the 1950's. Of course, the dialects here in southeastern Wisconsin could be further from GA than that, as both of my parents were born in the 1950's, and their speech clearly is not GA at all, having the NCVS and many consonant elisions and extra palatalization and all.) Of course, the dialect here has innovated new vowel + semivowel sequences thanks to consonant elisions and like; e.g. dialect for maybe is [meːj], contrasting with may and May [me(ː)], and pretty is [pʰʁɘj], contrasting with pre- [pʰʁi(ː)].

About raising shifts in long vowels, well, High German and Low Franconian first broke many long mid vowels into centralizing falling high diphthongs, and then monophthongized these diphthongs into high vowels (except in Upper German; note that it created new long mid vowels in a variety of ways) while also turning old high long vowels into closing falling opening diphthongs (except in Alemannic; note that the degree of opening varies, e.g. traditionally Standard Dutch only has open-mid diphthongs but much of contemporary Dutch actually has fully open diphthongs, like Standard German, while conversely there are High German dialects that did not fully open these diphthongs), Low Saxon turned long mid vowels into closing falling mid diphthongs (but created new long mid vowels as well; note that this is often hidden by conventional transcriptions of Low Saxon, which will write both /eɪ/ and /eː/ as ⟨ee⟩ and both /oʊ/ and /oː/ as ⟨oo⟩), Swedish and Norwegian first fronted their high back vowels and then raised their mid back vowels and raised and rounded their low long back vowels (I am not certain as to what extent this affected just long vowels or both long vowels and short vowels; note that Standard Swedish and Bokmål had all kind of vowel lengthening and shortening nonsense to confuse all this, which is not always reflected in the orthography). And of course we all know about what English did, but what is not always obvious to people is that Standard English was rather inconsistent in its raising due to irregular vowel shortenings at different stages or just failures to raise vowels freezing pronunciations, which is not always consistent with other Anglic varieties, e.g. Scots heid /hid/ for Standard English head /hɛd/ actually reflecting the regular outcome (aside from the Scots loss of phonemic vowel length).

[bradrn]

Huh, I’ve never heard of the Slavic open syllable conspiracy, what is it? (The one I’m most familiar with is the tendency towards disyllabic roots in Austronesian.)

here

Thank you! I shall have to add that to my list of books to read.

Indulging in a bit of speculation, it occurs to me that this tendency may have a simple explanation: large vowel systems are particularly prone to rearrangements, vowels are articulated in the same place as dorsals so interact most strongly with them, and as I noted earlier, /r l ŋ ɡ/ all have some sort of dorsal component to them, especially after a vowel (/m n/ are as you note less prominent in sound changes).

Well, do similar things happen in Austroasiatic?

I wouldn’t know; I know very little about Austroasiatic. (And why is Austroasiatic in particular relevant here?) But I do recall you saying something a while back about odd interactions between vowels and codas initials in Khmer… ah, here it is, from the old board:

Khmer can be analyzed as having an underlying vowel system of /a ɛ ɔ e ə o i ɨ u/, but with an additional +/-high feature. This is actually what happened diachronically; voiceless initials conditioned -high, and voiced initials conditioned +high, although with some phonation shit in between that doesn't make that any less bizarre.

Code: Select all

       a  ɛ  ɔ  e  ə  o  i  ɨ  u
-high: a  ae ɑ  e  aə ao əi ə  o
+high: iə ɛ  ɔ  ɪ  ə  ʊ  i  ɨ  u

Potentially this explanation may even have some predictive power… /j/ is the sole ‘missing’ voiced dorsal consonant from the above list, were there any sound changes conditioned by that one?

/j w/ are both missing, but they can only occur in coda position due to resyllabification or the Great Vowel Shift, which in GenAm output vowel + semivowel sequences. Possibly this is because x > j~w / V_C created /ij uw/ sequences which merged with /i: u:/, but most of Germanic underwent some kind of raising shift in long vowels, didn't it?

True, I didn’t notice that /w/ is also missing. But are you sure /j w/ can only appear in codas due to the Great Vowel Shift or resyllabification? Etymonline, for instance, lists an apparent counterexample in the form of may < OE mæg.

[Richard W]

I wouldn’t know; I know very little about Austroasiatic. (And why is Austroasiatic in particular relevant here?) But I do recall you saying something a while back about odd interactions between vowels and codas in Khmer… ah, here it is, from the old board:

Khmer can be analyzed as having an underlying vowel system of /a ɛ ɔ e ə o i ɨ u/, but with an additional +/-high feature. This is actually what happened diachronically; voiceless initials conditioned -high, and voiced initials conditioned +high, although with some phonation shit in between that doesn't make that any less bizarre.

Code: Select all

       a  ɛ  ɔ  e  ə  o  i  ɨ  u
-high: a  ae ɑ  e  aə ao əi ə  o
+high: iə ɛ  ɔ  ɪ  ə  ʊ  i  ɨ  u

Austroasiatic languages tend to have large vowel systems, so there is plenty of scope for minor interactions to have phonemic effect.

Nortaneous's example is an effect of initial consonants, not coda consonants.

[bradrn]

Nortaneous's example is an effect of initial consonants, not coda consonants.

Whoops, so it is! Fixed now.