Sound Change Quickie Thread

[Nortaneous]

Would people say that /u/ fortitioning to /kw/ word-finally and before voiceless consonants is a good idea?

w > kv / _# happened in Faroese, apparently including semivowels ejected in the Great Faroese Vowel Shift - e.g. kúgv [kʰɪkf]

I'm not sure if a no-closing-diphthongs analysis would work for Faroese like it does for English, but the details of skerping make it look like one would - maybe *kuː > *kuw > *kiw > *kiwː > *kigʷː > kʰikv? (If orthographic ú is /iw/, that'd also explain [ʊi̯] for í -- instead of iː uː > əj əw as in English, iː uː > uj iw.)

I vaguely remember Nortaneous once mentioning an instance of Vj > Vʃ in one of his blogs though...

I don't, but I might have. Index Diachronica has j > s in Blackfoot.

How common is /ᵐb ⁿd ᵑɡ/ without corresponding /ᵐp ⁿt ᵑk/?

very

the voiceless lenis stops become aspirate stops

this is a little weird

Given the above, how might "fortis" and "lenis" be defined in the proto-language itself as opposed to in its daughters?

Long/short? Or just unlenited/lenited (Celtic), or write it off as some vague and practically undefinable thing that's nevertheless obviously there (Korean).

Further, if I decide this is a branch of an even earlier proto-language, how might the distinction develop in the first place with the result that fortis consonants are favored word-initially and lenis consonants are favored word-finally (but the other way around not being phonotactically prohibited)?

You can get 'favored but not prohibited' by vowel loss at word edges or consonant cluster simplification. Maybe lenition in most postvocalic environments, or cluster reduction + initial consonants arbitrarily become fortis just because.

[Kuchigakatai]

the voiceless lenis stops become aspirate stops

this is a little weird

What do you think of the Middle Chinese development of voiceless stops into breathy-voiced stops, and then into aspirate stops before the level (平) tone?

I felt like writing a quick summary of the thing for those who read us, focusing on stops with evolving inventories. It was only after I wrote this post that I noticed Wikipedia has a useful table with the same information now which I could've just linked to, but meh. (I'm aware of the various shenanigans between stops and fricatives across time, but I'm skipping them because they're not relevant for this post.)


Early Middle Chinese stops in the Qieyun (early 7th century)
/pʰ tʰ tsʰ tɕʰ ʈʂʰ ʈʰ kʰ/
/p t ts tɕ ʈʂ ʈ k/
/b d dz dʑ ɖʐ ɖ g/

Then, in terms of POA, the so-called palatals (i.e. the alveolo-palatal affricates /tɕʰ tɕ dʑ/) merge with the retroflex affricates. I'm heading into Mandarin so I'll write these merged palatal-retroflex affricates with retroflex symbols, although they do end up as alveolo-palatal affricates in 19th-century Cantonese. In terms of MOA, voiced consonants become breathy-voiced.

Late Middle Chinese stops in the Guangyun (early 11th century)
/pʰ tʰ tsʰ ʈʂʰ ʈʰ kʰ/
/p t ts ʈʂ ʈ k/
/bʱ dʱ dzʱ ɖʐʱ ɖʱ gʱ/

Then the retroflex plosives merge into the retroflex affricates. Meanwhile, the breathy-voiced stops merged with the other two rows, with a tendency to merge into their aspirated counterparts when preceding a level (平) tone, and into their unaspirated counterparts when preceding any other tone. The 14th-century Zhongyuan Yinyun reflects a complete merger of the breathy-voiced stops, but it seems there was some tension between more conservative and more innovative speakers, since in the 15th century the Korean scholar-official Sin Sukju would describe the Mandarin of Ming dynasty officials conservatively, without the merger.

Old Mandarin stops in the Zhongyuan Yinyun (early 14th century)
/pʰ tʰ tsʰ tʂʰ kʰ/
/p t ts tʂ k/

Then the dental affricates and velars develop alveolo-palatal allophones that are pronounced the same, in complementary distribution with both series and also the retroflex affricates, creating Mandarin's famous awkward phonemic situation.

Modern (Northern/Standard) Mandarin stops (20th century)
/pʰ tʰ tsʰ tɕʰ tʂʰ kʰ/
/p t ts tɕ tʂ k/

[Pabappa]

How sure are we about the pronunciation of Middle Chinese? I mean, I know I asked this exact question before, but it was for one specific word, so im not repeating myself .... this time I mean, are we sure that there were ever voiced aspirates in Chinese to begin with? Specifically with that unpaired /bʰ/ we see in some reconstructions, exactly as in PIE, ... I wonder if the IE-ists and the Sinologists were borrowing ideas from each other. I should mention Im skeptical of the reconstruction of PIE as well.

[Knit Tie]

Can nasal+voiced plosive clusters like /mb/ simplify to just plain nasals unconditionally after all the nasals before everything else in the language become voicing on the vowels? Talking about hypothetical future Englishes here.

[Nortaneous]

Can nasal+voiced plosive clusters like /mb/ simplify to just plain nasals unconditionally after all the nasals before everything else in the language become voicing on the vowels?

what?

[Knit Tie]

Can nasal+voiced plosive clusters like /mb/ simplify to just plain nasals unconditionally after all the nasals before everything else in the language become voicing on the vowels?

what?

VNT clusters, where T is a voiceless plosive, become ṼT clusters. Then clusters like VNDV, where D is a voiced plosive, become VNV (ND clusters aren't found in word-final positions). And then the rest of the VNC clusters become ṼC.

[Nortaneous]

Can nasal+voiced plosive clusters like /mb/ simplify to just plain nasals unconditionally after all the nasals before everything else in the language become voicing on the vowels?

what?

VNT clusters, where T is a voiceless plosive, become ṼT clusters. Then clusters like VNDV, where D is a voiced plosive, become VNV (ND clusters aren't found in word-final positions). And then the rest of the VNC clusters become ṼC.

why not just have -ND- clusters simplify first? ND > N is plausible, VNC > ṼC also; if this rule happens not to have any ND clusters to apply to because they've already become N, that's fine

[Vijay]

Then there would be no nasals left in the language. That would be pretty bizarre for a language with nasalized vowels.

[Nortaneous]

Then there would be no nasals left in the language. That would be pretty bizarre for a language with nasalized vowels.

VNC > ṼC leaves nasals - it just creates a distribution restriction! Initial, intervocalic, and postconsonantal nasals aren't affected. (Word-final nasals technically aren't affected either, but if you have VNC > ṼC, you probably also have VN > Ṽ / _#.)

[dhok]

Then there would be no nasals left in the language. That would be pretty bizarre for a language with nasalized vowels.

VNC > ṼC leaves nasals - it just creates a distribution restriction! Initial, intervocalic, and postconsonantal nasals aren't affected. (Word-final nasals technically aren't affected either, but if you have VNC > ṼC, you probably also have VN > Ṽ / _#.)

Not in American English (where C is an obstruent).

[Knit Tie]

what?

VNT clusters, where T is a voiceless plosive, become ṼT clusters. Then clusters like VNDV, where D is a voiced plosive, become VNV (ND clusters aren't found in word-final positions). And then the rest of the VNC clusters become ṼC.

why not just have -ND- clusters simplify first? ND > N is plausible, VNC > ṼC also; if this rule happens not to have any ND clusters to apply to because they've already become N, that's fine

That works, thanks!

[Nortaneous]

Then there would be no nasals left in the language. That would be pretty bizarre for a language with nasalized vowels.

VNC > ṼC leaves nasals - it just creates a distribution restriction! Initial, intervocalic, and postconsonantal nasals aren't affected. (Word-final nasals technically aren't affected either, but if you have VNC > ṼC, you probably also have VN > Ṽ / _#.)

Not in American English (where C is an obstruent).

American English has anticipatory vowel nasalization, but doesn't have loss of the nasal consonant itself, except arguably in intervocalic nt nd > ɾ̃.

[Vijay]

Then there would be no nasals left in the language. That would be pretty bizarre for a language with nasalized vowels.

VNC > ṼC leaves nasals - it just creates a distribution restriction! Initial, intervocalic, and postconsonantal nasals aren't affected. (Word-final nasals technically aren't affected either, but if you have VNC > ṼC, you probably also have VN > Ṽ / _#.)

I think I just got confused. Never mind. (Apparently, "everything else" in Knit Tie's original question excluded vowels).

[axolotl]

w > kv / _# happened in Faroese, apparently including semivowels ejected in the Great Faroese Vowel Shift - e.g. kúgv [kʰɪkf]

I'm not sure if a no-closing-diphthongs analysis would work for Faroese like it does for English, but the details of skerping make it look like one would - maybe *kuː > *kuw > *kiw > *kiwː > *kigʷː > kʰikv? (If orthographic ú is /iw/, that'd also explain [ʊi̯] for í -- instead of iː uː > əj əw as in English, iː uː > uj iw.)

I'm not sure what you mean by a "no closing diphthongs" analysis for Faroese. Faroese has many more diphthongs than just those two, and all of them are closing. In fact, the only one that may not be closing is /ʊi̯/, which is sometimes pronounced with the first part reduced to a semivowel. In particular, the sequence /vʊi̯/ is often collapsed to [wi], especially after other consonants. For example, in this song the phrase "Hygg út og síggj hvíta kavan" is pronounced something like [hɪkː ʉut oː sʊtʃː kʰwitʰa ke̞a̯van].

[Nortaneous]

I'm not sure what you mean by a "no closing diphthongs" analysis for Faroese. Faroese has many more diphthongs than just those two, and all of them are closing. In fact, the only one that may not be closing is /ʊi̯/, which is sometimes pronounced with the first part reduced to a semivowel. In particular, the sequence /vʊi̯/ is often collapsed to [wi], especially after other consonants. For example, in this song the phrase "Hygg út og síggj hvíta kavan" is pronounced something like [hɪkː ʉut oː sʊtʃː kʰwitʰa ke̞a̯van].

Closing as in rising in height. English is traditionally analyzed as having a vowel system of /æ ɑ ɛ ʌ ɔ eɪ̯ oʊ̯ ɪ ʊ i u aɪ̯ aʊ̯ oɪ̯ ɚ/, but for various reasons it's better to analyze at least East Coast AmE as having /æ ɑ e ʌ o i ɚ u/, and to reanalyze the standard /eɪ̯ oʊ̯ i u aɪ̯ aʊ̯ oɪ̯/ as /ej ʌw ij uw aj æw ɔj/. For example, this explains l-breaking -- it's hard to explain why Vl > Vəl after /aɪ̯ eɪ̯ i/ but not /ɛ ɪ/, but it's easy to explain why jl > jəl. (Some Americans also have l-breaking after /w/, so semivowel + l codas are disallowed. The same restriction applies to rl - when I was young I trained myself out of pronouncing "Carl" with two syllables, because it was spelled as if it only had one, so that must be right. And it's pretty common to hear it with two. Other -rl coda clusters only exist in surnames but IME always take epenthetic schwa.) This also explains fronting of "/aʊ̯ oʊ̯ u/" - there's no +/-front contrast before /w/ anyway, so the vowel centralizes phonetically.

It seems like Faroese could take a similar analysis. For skerping, you have ɔu̯ ʉu̯ > ɛ ɪ before <gv>, so maybe these should be analyzed as /ɛw ɪw/ - and then you have ɛi̯ ʊi̯ ai̯ ɔi̯ > ɛ ʊ a ɔ before <ggj>, so maybe these should be analyzed as /ej uj aj oj/. But I'm not sure if this is a workable analysis; there may be reasons against it, and even if it works, it doesn't have the same benefit that it does in English, where it makes the vowel system a little more crosslinguistically reasonable, eliminates long vowels, and almost simplifies the phonotactics. (Unfortunately, it ends up prohibiting stressed open syllables with vowels other than /ɑ o ɚ/, and that's not a natural class at all. If you call /o/ a low vowel, which it phonetically is, that explains the cot-caught merger and lets you say stressed open syllables can only contain low vowels or /ɚ/, which is a little more reasonable. For maximal simplicity we'd also like to be able to get rid of /ɚ/, but there's no way to do that; it has to be a monophthong because it doesn't condition l-breaking.)

The Faroese opening diphthongs can't be analyzed this way, but opening diphthongs in place of a fourth vowel height are also attested in Romanian and Chechen.

The VC co-occurrence restrictions this implies aren't a problem. There are languages where certain coda consonants can only occur with low vowels, but nobody says /ak/ is a diphthong.

[dɮ the phoneme]

In an inventory that already has /z ʒ ʑ/, how plausible does dʒ dʑ > ʒ ʑ without dz > z look?

[Vijay]

It looks pretty normal.

[Knit Tie]

So I've decided to include pervasive devoicing in my conlang, both word-finally and before voiceless obstruents. So far, the main results of this devoicing are:

w → ʍ → xʷ → x
j → j̥ → ç → ʃ
l → l̥ → ɬ
NC → N̥C → hC → Cʰ
NS → N̥S → hS → Sː

Later on, ɬC clusters simplify to just ɬ, while Cɬ clusters become the affricate tɬ.

How does the above look in terms of realism?

Also, I wonder what I should do with situations like wNC, where the above environments are together? Just combine them into xCʰ in the end? Perhaps make the voiceless resonants merge together?

[Nortaneous]

w → ʍ → xʷ → x
j → j̥ → ç → ʃ
l → l̥ → ɬ
NC → N̥C → hC → Cʰ
NS → N̥S → hS → Sː

reasonable

Later on, ɬC clusters simplify to just ɬ, while Cɬ clusters become the affricate tɬ.

I tried looking into Mizo initials once and it looked to me like *Pl > tl, so that's attested; non-plosives + ɬ would probably become something else, depending on the non-plosive. ɬC > ɬ is a little weird, but in some Qiangic languages there's /ɬ ɮ/ [ɬt ɮd], so ɬt > ɬ at least seems reasonable... again tho, I'd expect different outcomes for different clusters

Also, I wonder what I should do with situations like wNC, where the above environments are together? Just combine them into xCʰ in the end? Perhaps make the voiceless resonants merge together?

xCʰ seems like the simplest outcome, but further developments could occur from there

[Knit Tie]

What would you say I should do to get both ɬ and tɬ intervocalically, then?