Zompist Bboard Again

bradrn wrote: ↑Fri Mar 13, 2020 6:49 am I didn’t know that! I had thought that /ɜ/ and /ɘ/ were completely separate vowels on the phonological level, so thanks for explaining!

Well, they're a little separate -- the umlaut processes are in some cases not recoverable. From Bender 1968:

Thus one can say concerning forms such as these that é is either a lowered i (when the stem vowel e is dropped from independent, unsuffixed forms) or a raised e (when the stem vowel i is dropped), but its double origin makes biuniqueness impossible; one could not know whether to transcribe [bʷʊŋʷ] (béŋʷ) 'night' as beŋʷi or biŋʷe (with all word-final vowels to be dropped by a general reading rule) without additional information -- the quality of its stem vowel when suffixed. To complicate matters further, some forms have undergone reanalysis: one can hear both beŋʷin yinney and biŋʷen yinney 'night of yesterday' 'night before last' (but always béŋʷ 'night'). There is every indication that all instances of é can be derived ultimately from i or e and so written in base forms, but for reasons of the sort just noted, I have retained the fourth vowel ... while recognizing that the underlying Marshallese system is a three-vowel system.

(It's an interesting paper -- in the introduction, he suggests that other Micronesian languages, specifically Kosraean and Chuukic, might have VVSs as well.)

I’m not sure I understand this. Where do you see /yiyy/? And where did you see that /yiyiy/ is prohibited?

"{yiy} is for a "dwelling upon" version of i that occurs at the beginning of certain words, now generally written ii in the "new" orthography, phonetically pronounced [iː] and existing on the phonemic level as /jijj/, effectively making it identical to {yiyy}."

I don't see any instances of word-initial /yiyiy/.

I don’t understand this either. What do you mean by ‘orthographic long i’?

It's written <ii>.

Nortaneous wrote: ↑Fri Mar 13, 2020 6:24 pm

I’m not sure I understand this. Where do you see /yiyy/? And where did you see that /yiyiy/ is prohibited?

"{yiy} is for a "dwelling upon" version of i that occurs at the beginning of certain words, now generally written ii in the "new" orthography, phonetically pronounced [iː] and existing on the phonemic level as /jijj/, effectively making it identical to {yiyy}."

I don't see any instances of word-initial /yiyiy/.

I don’t understand this either. What do you mean by ‘orthographic long i’?

It's written <ii>.

I understand what you mean now, so now I can try to answer your original question:

Nortaneous wrote: If /yiyy/ is pronounced [i:], how is it distinguished from /yiyiy/? … Does short i contrast with non-syllabic i word-initially? … if there's just a rule that word-initial prevocalic /yiy/ is realized as [j], then there wouldn't be a need for a true phonemic semivowel. But if, say, [jæ͡ɑ i.æ͡ɑ iː.æ͡ɑ] all contrast, you need /i̯/.

I think that the confusion here is due to the fact that Bender originally created this notation for language learners, rather than linguists. When Bender created it, he wasn’t particularly interested in transcribing every detail of whether something was phonetic or phonemic; he was rather simply interested in giving a romanisation which allows the learner to pronounce words reasonably well. And this was his motivation in using {yiy} vs {'yiy} vs {yi'y}:

Bender wrote: Most words beginning in yiy differ in pronunciation between the two dialects [Rālik and Ratak] … The Rälik dialect adds an extra yi, while the Ratak dialect shortens the existing yiy so that it sounds little longer that the y of English yes … There has been a great deal of dialect mixture … and a few [words] are now usually given the Rälik pronunciation in both dialects. Such words are marked … by an apostrophe before the word [i.e. as 'yiy] … A few others are usually given the Ratak pronunciation in both dialects … indicated by putting the apostrophe … before the second y [i.e. as yi'y]

So from this, it sounds like it is unknown whether this is a phonemic or phonetic detail; it is just a detail of the pronunciation which was considered important enough to include in a largely phonemic romanisation.

As for whether we can conclude anything about whether the distinction between all of these is phonetic or phonemic, here are my thoughts on the matter:

• Firstly, it seems pretty clear to me that {'yiy} and {yi'y} are at the very least distinct on the phonetic level, given that they occur in at least one minimal pair.
• I also managed to find a contrast between {'yiy} and {y} ({yẹy} vs {'yiyẹy}), plus a near-minimal pair between {yi'y} and {y} ({yẹy} vs {yi'yey}), so {y} appears to be distinct from both.
• But remember that in the VVS analysis, all glides are removed at the beginning of a word. Thus there needs to be at least a phonemic /i̯/ to ensure that the semivowel is always present at the beginning of {yi'y} words.
• There’s also the issue of Wikipedia’s phonemic transcription of /jijj/ for {'yiy} and /jij/ for {yi'y}. I think that /jijj/ is probably too complex a transcription; I see no observable distinction between /jijj/ and /jij/. (This is the same problem as your observation about {kakitehteh} vs {kakitete}.) However, I do think that this is distinct from /jijij/: /jij~jijj/ is twice-long [iː], whereas /jijij/ would be thrice-long /iːː/. But I think /jij/ is perfectly reasonable: the onset and coda glides get removed at the phonetic level, and you’re left with [i̯] = [j].

bradrn wrote: ↑Fri Mar 13, 2020 8:05 pm There’s also the issue of Wikipedia’s phonemic transcription of /jijj/ for {'yiy} and /jij/ for {yi'y}. I think that /jijj/ is probably too complex a transcription; I see no observable distinction between /jijj/ and /jij/. (This is the same problem as your observation about {kakitehteh} vs {kakitete}.) However, I do think that this is distinct from /jijij/: /jij~jijj/ is twice-long [iː], whereas /jijij/ would be thrice-long /iːː/. But I think /jij/ is perfectly reasonable: the onset and coda glides get removed at the phonetic level, and you’re left with [i̯] = [j].
[/list]

Why would /jij/ be long? If you had a word, say, /jayiyan/, wouldn't this be pronounced [tʲæ.i.ænʲ]? There are some examples of words in the MED with /yiy/ <i>:
- jañai /jahgahyiy/ 'boom'
- jei /jẹyiy/ 'duty that a younger sibling is expected to give to older siblings'
- jekāiōōj /jekayiyehej/ 'notches cut into a tree for climbing'
- jekṃai /jekṃahyiy/ 'coconut syrup boiled down from sap'

And there are some examples of /yiyy/ <ii>:
- keiie /kẹyiyyẹy/ 'strong, useful, serviceable'
- keiiuiu /kẹyiyyiwyiw/ 'swamp spring'
- ṃweiie /ṃẹyiyyey/ 'rich'

There are a few examples of /wiww/ <uw>, but aside from iọuwọ /yiyawiwwaw/ 'beware', all the examples I've found are word-initial. (But there might be others.) I wonder what this contrast is phonetically. Maybe intervocalic /yiy/ is just [j]? Or /yiyV yiyyV/ [i.V i.jV]? That'd be a weird contrast.

Nortaneous wrote: ↑Sat Mar 14, 2020 12:48 pm

bradrn wrote: ↑Fri Mar 13, 2020 8:05 pm There’s also the issue of Wikipedia’s phonemic transcription of /jijj/ for {'yiy} and /jij/ for {yi'y}. I think that /jijj/ is probably too complex a transcription; I see no observable distinction between /jijj/ and /jij/. (This is the same problem as your observation about {kakitehteh} vs {kakitete}.) However, I do think that this is distinct from /jijij/: /jij~jijj/ is twice-long [iː], whereas /jijij/ would be thrice-long /iːː/. But I think /jij/ is perfectly reasonable: the onset and coda glides get removed at the phonetic level, and you’re left with [i̯] = [j].
[/list]

Why would /jij/ be long? If you had a word, say, /jayiyan/, wouldn't this be pronounced [tʲæ.i.ænʲ]? There are some examples of words in the MED with /yiy/ <i>:
- jañai /jahgahyiy/ 'boom'
- jei /jẹyiy/ 'duty that a younger sibling is expected to give to older siblings'
- jekāiōōj /jekayiyehej/ 'notches cut into a tree for climbing'
- jekṃai /jekṃahyiy/ 'coconut syrup boiled down from sap'

And there are some examples of /yiyy/ <ii>:
- keiie /kẹyiyyẹy/ 'strong, useful, serviceable'
- keiiuiu /kẹyiyyiwyiw/ 'swamp spring'
- ṃweiie /ṃẹyiyyey/ 'rich'

There are a few examples of /wiww/ <uw>, but aside from iọuwọ /yiyawiwwaw/ 'beware', all the examples I've found are word-initial. (But there might be others.) I wonder what this contrast is phonetically. Maybe intervocalic /yiy/ is just [j]? Or /yiyV yiyyV/ [i.V i.jV]? That'd be a weird contrast.

Thanks for pointing this out! You’re right — that was an error in my reasoning.

bradrn wrote: ↑Tue Mar 10, 2020 2:28 am This map of Yap has lots of names which sound like they’re from a bad fantasy novel: Thileer, Fongoch, Musgow, Gachpar, Feeching, Kanifay, Teanfaar, Qeng, Qawoch, Luweech, Bulwol…

What about that big area labeled Tamil? What language do they speak there?

Some ridiculous phonologies:

• Nuosu (Northern Yi) is more famous for its huge syllabary, but it also has the craziest vowel system I’ve ever seen: /a̱ ɛ̱ ɔ̱ e̝ ɤ̝ o̝ z̩̱ v̩̱ʷ z̩ v̩ʷ/. Yes, that’s right — it has syllabic /z vʷ/ instead of high vowels /i u/. The underline represents ‘tight throat’, a special phonological distinction which Wikipedia describes as laryngealization + retracted tongue root. The syllabic consonants have a bunch of phonetic representations, most of which are either silly or weird: for instance, they can become [m͡l̩] after /m/, assimilate to a preceding lateral, or gives a trilled release for preceding labial or alveolar stops (e.g. the word for “to hit” is /ⁿdv̩ʷ˨˩/ [ⁿdʙv̩ʷ˨˩]).
• Japhug has prenasalised, voiceless and aspirated stops for all of labial, dental, alveolar, retroflex, alveolopalatal, palatal, velar and uvular POAs (although the alveolar, retroflex and alveolopalatal ones are affricated). This happens for voiced stops as well, except that there’s no voiced uvular stops. (So there’s /ᶰɢ/ but no /ɴ/ or /ɢ/.)
• Hmong (all varieties) has a set of bilabial consonants with lateral release: /mˡ pˡ pˡʰ ᵐbˡ ᵐpˡʰ/. Wikipedia notes that the phonemic status of these consonants is controversial, and for good reason: on the one hand, if they aren’t phonemic, then they would be the only consonant clusters in the language (prenasalised stops aside), but on the other hand, no other language is known to use these sounds phonemically!
• In Adzera, every stop has a prenasalised counterpart. And I really do mean every: it has all of /ᵐp ⁿt ⁿtʃ ᵑk ᵑʔ/. I have no idea how you would even say that last one! (I looked at the source, but it wasn’t really any help).

bradrn wrote: ↑Sat Apr 11, 2020 4:04 am [*] Nuosu (Northern Yi) is more famous for its huge syllabary, but it also has the craziest vowel system I’ve ever seen: /a̱ ɛ̱ ɔ̱ e̝ ɤ̝ o̝ z̩̱ v̩̱ʷ z̩ v̩ʷ/. Yes, that’s right — it has syllabic /z vʷ/ instead of high vowels /i u/. The underline represents ‘tight throat’, a special phonological distinction which Wikipedia describes as laryngealization + retracted tongue root. The syllabic consonants have a bunch of phonetic representations, most of which are either silly or weird: for instance, they can become [m͡l̩] after /m/, assimilate to a preceding lateral, or gives a trilled release for preceding labial or alveolar stops (e.g. the word for “to hit” is /ⁿdv̩ʷ˨˩/ [ⁿdʙv̩ʷ˨˩]).

Not really "instead of". All the published descriptions are somehow inadequate. o (/o̝/) is realized as either high-mid or high, but i ("/e̝/") always sounds high to me, as does e ("/ɤ̝/").

Nuosu just has an extension of the vowel triangle. It's possible for vowels to become so high that they become +consonantal -- that is, so high that they take on a consonant-like degree of oral constriction. For the raised consonantal u̝, a labiodental offglide is probably more common than a trilled one (cf. Angami, where p b m kʷ gʷ > pf bv ɱ kf gv before ə < *u̝), but trilling isn't unheard of. (Also Nias t d > tf dv ~ pf bv before the vowel transcribed /u/ but according to the grammar realized as [ʋ̩].)

[*] Hmong (all varieties) has a set of bilabial consonants with lateral release: /mˡ pˡ pˡʰ ᵐbˡ ᵐpˡʰ/. Wikipedia notes that the phonemic status of these consonants is controversial, and for good reason: on the one hand, if they aren’t phonemic, then they would be the only consonant clusters in the language (prenasalised stops aside), but on the other hand, no other language is known to use these sounds phonemically!

That's entirely an artifact of analysis, but Pumi has been claimed to have unit retroflexed labial and velar plosives by phonetic arguments.

In Adzera, every stop has a prenasalised counterpart. And I really do mean every: it has all of /ᵐp ⁿt ⁿtʃ ᵑk ᵑʔ/. I have no idea how you would even say that last one! (I looked at the source, but it wasn’t really any help).

[ŋʔ], and that's not even the only example of a phonemically glottal nasal being realized as velar (because a true glottal nasal is articulatorily impossible)

Nortaneous wrote: ↑Sat Apr 11, 2020 2:13 pm

bradrn wrote: ↑Sat Apr 11, 2020 4:04 am [*] Nuosu (Northern Yi) is more famous for its huge syllabary, but it also has the craziest vowel system I’ve ever seen: /a̱ ɛ̱ ɔ̱ e̝ ɤ̝ o̝ z̩̱ v̩̱ʷ z̩ v̩ʷ/. Yes, that’s right — it has syllabic /z vʷ/ instead of high vowels /i u/. The underline represents ‘tight throat’, a special phonological distinction which Wikipedia describes as laryngealization + retracted tongue root. The syllabic consonants have a bunch of phonetic representations, most of which are either silly or weird: for instance, they can become [m͡l̩] after /m/, assimilate to a preceding lateral, or gives a trilled release for preceding labial or alveolar stops (e.g. the word for “to hit” is /ⁿdv̩ʷ˨˩/ [ⁿdʙv̩ʷ˨˩]).

Not really "instead of". All the published descriptions are somehow inadequate. o (/o̝/) is realized as either high-mid or high, but i ("/e̝/") always sounds high to me, as does e ("/ɤ̝/").

I think that /e/ and /i/ often sound very similar. (At least, they do to me.) It’s even worse with /o/ and /u/.

Nuosu just has an extension of the vowel triangle. It's possible for vowels to become so high that they become +consonantal -- that is, so high that they take on a consonant-like degree of oral constriction. For the raised consonantal u̝, a labiodental offglide is probably more common than a trilled one (cf. Angami, where p b m kʷ gʷ > pf bv ɱ kf gv before ə < *u̝), but trilling isn't unheard of. (Also Nias t d > tf dv ~ pf bv before the vowel transcribed /u/ but according to the grammar realized as [ʋ̩].)

Yes, I figured this was what had happened. But it’s still pretty unusual for this to happen to both high vowels.

[*] Hmong (all varieties) has a set of bilabial consonants with lateral release: /mˡ pˡ pˡʰ ᵐbˡ ᵐpˡʰ/. Wikipedia notes that the phonemic status of these consonants is controversial, and for good reason: on the one hand, if they aren’t phonemic, then they would be the only consonant clusters in the language (prenasalised stops aside), but on the other hand, no other language is known to use these sounds phonemically!

That's entirely an artifact of analysis

Yes, but it’s a pretty interesting artifact of analysis. (Just like Marshallese and Arrernte, which I believe we were talking about earlier on this thread.)

but Pumi has been claimed to have unit retroflexed labial and velar plosives by phonetic arguments.

This is even weirder than all the stuff I put in my post! How would you even say such sounds? And do you have a source I can read through?

In Adzera, every stop has a prenasalised counterpart. And I really do mean every: it has all of /ᵐp ⁿt ⁿtʃ ᵑk ᵑʔ/. I have no idea how you would even say that last one! (I looked at the source, but it wasn’t really any help).

[ŋʔ], and that's not even the only example of a phonemically glottal nasal being realized as velar (because a true glottal nasal is articulatorily impossible)

That [ŋʔ] realisation makes sense. But the rest of that sounds pretty interesting; could you give some examples of a ‘phonemically glottal nasal being realized as velar’?

bradrn wrote: ↑Sat Apr 11, 2020 7:59 pm

but Pumi has been claimed to have unit retroflexed labial and velar plosives by phonetic arguments.

This is even weirder than all the stuff I put in my post! How would you even say such sounds? And do you have a source I can read through?

Rhoticization as a Secondary Articulation in Stops: Evidence from Prinmi

That [ŋʔ] realisation makes sense. But the rest of that sounds pretty interesting; could you give some examples of a ‘phonemically glottal nasal being realized as velar’?

Sure, place neutralization processes. If you have total place neutralization, it's usually to glottals, but for nasals it's usually to velars, as in Spanish. In Kagoshima Japanese, for example, final vowels drop under circumstances unclear in my source, after which final non-nasals become glottal, and final nasals become velar:
/kɯtsɯ/ 'shoe' > [kɯʔ]
/midzɯ/ 'water' > [miʔ]
/hirɯ/ 'noon' > [hiʔ]
/kami/ 'hair' > [kaŋ]

And decomposition: sometimes if you have a creaky-voiced vowel, let's say [a̰], it can be pronounced [aˀ] with decomposition into a modal vowel and following glottal closure. In some Skou languages, Ṽ varies with Vŋ.

Anusvara is also commonly argued to be a glottal nasal. Paul de Lacy analyzes anusvara as a glottal continuant, phonemically /ɦ̃/, contrasting with a glottal nasal stop, but it is not entirely clear to me why.

This is of course distinct from /h̃/, which straightforwardly exists in some Sino-Tibetan and probably African languages: in Lisu, for example, nasal vowels don't otherwise exist, but vowels are nasalized after the glottal fricative except in an imperative particle. So there's /h̃/ contrasting with a marginal /h/. In Laze, the only syllables in which nasal vowels occur are [hwã hɑ̃ hæ̃ hṽ̩ hũ hwɤ̃ hĩ hĩẽ æ̃ ṽ̩ ɔ̃], but it's unclear how to handle these -- [æ ɔ] never occur as full syllables, so we could make the case for /h̃/ in all but the last three and then /æ v̩ ɔ/ with spontaneous nasalization (taking [v̩] as /vv̩/), or say the last three start with /ɦ̃/, or (the solution taken in the phonology by Michaud and Jacques, and probably the worst option) say there are nasal vowels with restricted distributions.

Nortaneous wrote: ↑Sat Apr 11, 2020 10:29 pm

bradrn wrote: ↑Sat Apr 11, 2020 7:59 pm

but Pumi has been claimed to have unit retroflexed labial and velar plosives by phonetic arguments.

This is even weirder than all the stuff I put in my post! How would you even say such sounds? And do you have a source I can read through?

Rhoticization as a Secondary Articulation in Stops: Evidence from Prinmi

Thanks! So if I’m understanding correctly, it looks like those are just regular plosives coarticulated with a retroflex rhotic.

That [ŋʔ] realisation makes sense. But the rest of that sounds pretty interesting; could you give some examples of a ‘phonemically glottal nasal being realized as velar’?

Sure, place neutralization processes. If you have total place neutralization, it's usually to glottals, but for nasals it's usually to velars, as in Spanish. In Kagoshima Japanese, for example, final vowels drop under circumstances unclear in my source, after which final non-nasals become glottal, and final nasals become velar:
/kɯtsɯ/ 'shoe' > [kɯʔ]
/midzɯ/ 'water' > [miʔ]
/hirɯ/ 'noon' > [hiʔ]
/kami/ 'hair' > [kaŋ]

That makes sense. But I’m not sure I would agree that the velar nasal here is phonemically glottal.

And decomposition: sometimes if you have a creaky-voiced vowel, let's say [a̰], it can be pronounced [aˀ] with decomposition into a modal vowel and following glottal closure. In some Skou languages, Ṽ varies with Vŋ.

…

This is of course distinct from /h̃/, which straightforwardly exists in some Sino-Tibetan and probably African languages: in Lisu, for example, nasal vowels don't otherwise exist, but vowels are nasalized after the glottal fricative except in an imperative particle. So there's /h̃/ contrasting with a marginal /h/. In Laze, the only syllables in which nasal vowels occur are [hwã hɑ̃ hæ̃ hṽ̩ hũ hwɤ̃ hĩ hĩẽ æ̃ ṽ̩ ɔ̃], but it's unclear how to handle these -- [æ ɔ] never occur as full syllables, so we could make the case for /h̃/ in all but the last three and then /æ v̩ ɔ/ with spontaneous nasalization (taking [v̩] as /vv̩/), or say the last three start with /ɦ̃/, or (the solution taken in the phonology by Michaud and Jacques, and probably the worst option) say there are nasal vowels with restricted distributions.

Aren’t these just cases of rhinoglottophilia?

And an unrelated interesting feature: this Wikipedia article implies that the orthography of Aklanon writes /ɣ/ as ⟨e⟩. Understandable, but very weird. (Although admittedly I can’t find a source for this, so it could just be a typo.)

i think theyre confusing it with /ɤ/, the "ram's horn" vowel. wikipedia seems to be having issues right now but i found it on cache: https://webcache.googleusercontent.com/ ... clnk&gl=us

bradrn wrote: ↑Sun Apr 12, 2020 12:03 am Aren’t these just cases of rhinoglottophilia?

Diachronically, but not synchronically. Laze has the minimal pair [hæ hæ̃] (but this is the only minimal pair between a nasal and an oral vowel); Lisu probably has [hɑ̃] contrasting with [hɑ]. If you don't want to posit nasal vowels with extremely restricted distribution (and why would you want to?), you have to say there's /h̃/.

There's also the Austronesian language Seimat. There's a full inventory of nasal vowels with a similarly restricted distribution: aside from the common Austronesian process of rightward nasality spreading from /m n ŋ/, nasal vowels only occur after /h w/, and this contrast is maintained even in the context of nasality spreading; Blust says this is by differences in the degree of nasality.

This isn't so unusual; something similar happens in Skouic. Most Skouic languages lost the oral/nasal vowel contrast after nasal consonants, but apparently phonemically nasal vowels are more strongly nasalized than allophonically nasal vowels: /ta na tã/ [ta nã tã̃]. (Hopefully the stacked tildes will display correctly!)

So in Seimat, there's the contrast /nahi/ 'to go' vs. /nahĩ/ 'work', presumably realized [nãh̃ĩ] vs. [nãh̃ĩ̃] with weak vs. strong nasalization. For /w/, there's the minimal pair /wat/ 'monitor lizard' vs. /wãt/ 'earthworm'. And nasal vowels only occur after /h w/, so we may as well reanalyze Seimat as having /h̃ w̃/ instead of nasal vowels.

For /w̃/, this is clearly supported by the diachronics - it's from *mw, whereas /w/ continues *w. /h̃/, on the other hand, is from *r, whereas non-nasalizing /h/ is from *p. Blust says this is probably due to rhinoglottophilia; an obvious-seeming alternative is *r > *ɳ, but Seimat also has *dr > k/x, so it seems reasonable to assume parallelism and postulate something like *r > *x > *h > *h̃. (Seimat appears to have generally collapsed the voiceless/prenasalized contrast, which would explain why *dr /ⁿr/ was devoiced.)

Pabappa wrote: ↑Sun Apr 12, 2020 12:15 am i think theyre confusing it with /ɤ/, the "ram's horn" vowel. wikipedia seems to be having issues right now but i found it on cache: https://webcache.googleusercontent.com/ ... clnk&gl=us

This seems possible, but that Wikipedia article definitely has some issues: it lists /ɤ/ in the second line, but not in the vowel inventory! Furthermore, A grammar of Aklan (Chai) lists the vowel inventory as /a i u/, without any mention of /ɤ/. Finally, that Wikipedia article gives the word for ‘house’ as ⟨baeay⟩; Chai lists the same word as ⟨baǥay⟩, where ⟨ǥ⟩ is his notation for the voiced velar fricative. So I think it’s pretty probable that Aklanon does actually have ⟨e⟩ for /ɣ/.

Nortaneous wrote: ↑Sun Apr 12, 2020 12:32 am

bradrn wrote: ↑Sun Apr 12, 2020 12:03 am Aren’t these just cases of rhinoglottophilia?

Diachronically, but not synchronically. Laze has the minimal pair [hæ hæ̃] (but this is the only minimal pair between a nasal and an oral vowel); Lisu probably has [hɑ̃] contrasting with [hɑ]. If you don't want to posit nasal vowels with extremely restricted distribution (and why would you want to?), you have to say there's /h̃/.

There's also the Austronesian language Seimat. There's a full inventory of nasal vowels with a similarly restricted distribution: aside from the common Austronesian process of rightward nasality spreading from /m n ŋ/, nasal vowels only occur after /h w/, and this contrast is maintained even in the context of nasality spreading; Blust says this is by differences in the degree of nasality.

This isn't so unusual; something similar happens in Skouic. Most Skouic languages lost the oral/nasal vowel contrast after nasal consonants, but apparently phonemically nasal vowels are more strongly nasalized than allophonically nasal vowels: /ta na tã/ [ta nã tã̃]. (Hopefully the stacked tildes will display correctly!)

So in Seimat, there's the contrast /nahi/ 'to go' vs. /nahĩ/ 'work', presumably realized [nãh̃ĩ] vs. [nãh̃ĩ̃] with weak vs. strong nasalization. For /w/, there's the minimal pair /wat/ 'monitor lizard' vs. /wãt/ 'earthworm'. And nasal vowels only occur after /h w/, so we may as well reanalyze Seimat as having /h̃ w̃/ instead of nasal vowels.

For /w̃/, this is clearly supported by the diachronics - it's from *mw, whereas /w/ continues *w. /h̃/, on the other hand, is from *r, whereas non-nasalizing /h/ is from *p. Blust says this is probably due to rhinoglottophilia; an obvious-seeming alternative is *r > *ɳ, but Seimat also has *dr > k/x, so it seems reasonable to assume parallelism and postulate something like *r > *x > *h > *h̃. (Seimat appears to have generally collapsed the voiceless/prenasalized contrast, which would explain why *dr /ⁿr/ was devoiced.)

That’s pretty convincing evidence. I think you’ve convinced me that some languages have at least phonemic /h̃/ (though I’m still not sure that there is a phonetic distinction between [h̃] and [h]).

bradrn wrote: ↑Sun Apr 12, 2020 1:24 am So I think it’s pretty probable that Aklanon does actually have ⟨e⟩ for /ɣ/.

Cruz and Zorc assert the rule and give highly credible example words, but I couldn't find any examples of the orthography - examples are transcribed. The Ram's horn definitely looks like a misreading.

Richard W wrote: ↑Sun Apr 12, 2020 4:56 am

bradrn wrote: ↑Sun Apr 12, 2020 1:24 am So I think it’s pretty probable that Aklanon does actually have ⟨e⟩ for /ɣ/.

Cruz and Zorc assert the rule and give highly credible example words, but I couldn't find any examples of the orthography - examples are transcribed. The Ram's horn definitely looks like a misreading.

The orthographical example I found was from the Wikipedia article, specifically from this table row in ‘Common phrases’:

Akeanon	Malaynon	English
Si-in dampit/dapit ing baeay?	Sa diin kampi mong balay?	Where is your house located?

After a bit of digging I eventually managed to find that balay means ‘house’ in the Cebuano language (source: https://en.wikipedia.org/wiki/Malaybalay#Etymology states that Malaybalay means ‘house of the Malays’), which is closely related to Aklanon. Thus baeay must be the word for ‘house’ in Aklanon. And it just so happens that Choi (1971) gives a phonetic transcription of the Aklanon word for house as /baǥáy/, where /ǥ/ is stated to be the velar fricative. (This is consistent with Cruz and Zorc, who state that Aklanon /ɣ/ corresponds to other Visayan languages’ /l/.)

Erica Hoffmann’s thesis A Preliminary Phonology of Bana describes all sorts of weird and wonderful vowel systems (although as the title suggests it focuses on Bana):

• Word-finally, Higi has an unusual four-vowel system: /i e ɛ a/, with /e/ being very rare. But only three vowels are allowed medially: high /i~ɪ~ɨ~ʊ~u/, mid /e~o/, low /ɛ~ʌ~ɔ~a/. And this tiny system is reduced in several other weird ways as well: for instance, /i/ and /ɛ/ are neutralised utterance-initially, despite the fact that those vowels aren’t contiguous on the vowel chart! Similarly, /e/ and /a/ are neutralised in certain other circumstances, even though they are not contiguous either. And apparently the four vowels get reduced to only two in ‘certain grammatical constructions’.
• Miya has a three-vowel system /ɨ a aː/ (although Hoffmann states that it ‘can be stretched into eleven phonetic vowel realizations’), but /ɨ/ has a predictable distribution and is essentially epenthetic, with the complication that it is inserted before affixes are applied rather than afterwards.
• Guɗe has another unusual four-vowel system: /ɨ ɨː a aː/. These have many allophones, as expected, mostly depending on the ‘colour’ (labialisation and palatalisation) of surrounding consonants. But the rules for this allophony are fairly unusual: most interestingly, ‘two contiguous consonants of different colours cancel out each other’s effect on the vowel’.
• Bana has a similarly weird vowel system: /ɨ ɛ a/, with /ɛ/ being very rare outside plurals. It also has a palatalisation as a ‘syllable-level autosegmental feature’.

bradrn wrote: ↑Sun Apr 12, 2020 12:03 am

Nortaneous wrote: ↑Sat Apr 11, 2020 10:29 pmSure, place neutralization processes. If you have total place neutralization, it's usually to glottals, but for nasals it's usually to velars, as in Spanish. In Kagoshima Japanese, for example, final vowels drop under circumstances unclear in my source, after which final non-nasals become glottal, and final nasals become velar:
/kɯtsɯ/ 'shoe' > [kɯʔ]
/midzɯ/ 'water' > [miʔ]
/hirɯ/ 'noon' > [hiʔ]
/kami/ 'hair' > [kaŋ]

That makes sense. But I’m not sure I would agree that the velar nasal here is phonemically glottal.

In Cantonese, [ʔ] and [ŋ] are allophones of the "zero-initial", or default initial if you will of the C(w)V(m,n,ŋ,p,t,k) syllable structure. Middle Chinese distinguished /ʔ/ and /ŋ/ as different initials, but these merged in Cantonese, with [ʔ] tending to appear before the phonemes /e~ɛ:, ø~œ:, ɐ, a:, o~ɔ:, ʊ~u:/ in high-register tones or the mid-level tone, and [ŋ] before those same vowel phonemes in low-register tones. (Before /ɪ~i:, y:/, they have further merged with /j/ and now /j/ is typically used: [jɪp ɥy:n], although these may sometimes be [ʔɪp ʔy:n] among some people.) However, /ŋ/ still had some faintly marginal existence as a phoneme in colloquial words of probably non-Sinitic origins (e.g. 啱啱 [ŋa:m˥ŋa:m˥] 'just now, just a moment ago', also [ʔa:m˥ʔa:m˥]) in the mid-20th century, and at least in more formal registers people would also use it before vowels with mid-level tone.

However, in recent decades, [ʔ] has taken over most contexts of [ŋ] in colloquial levels of the language, so that now [ŋ] is mostly just found before /ɐ, a:/ in low-register tones (and even there [ʔ] can sometimes be heard too). This has also meant that initial [ŋ] in its previous contexts has become associated with formal Cantonese in the cities, less urban ways of speaking, old folks' way of speaking, classic pop or rock songs, and the Cantonese of people overseas. I have come across YouTube videos of young Cantonese speakers born overseas where people from Hong Kong make comments expressing their shock to hear such young people use plenty of initial [ŋ] in words like 我 [ŋɔ:˩˧] 'I, me' (instead of [ʔɔ:˩˧]), a habit naturally acquired from their parents and a lack of contact with mainstream Cantonese culture in southern China / Hong Kong.

To me, this whole thing looks like [ŋ] being treated as equally glottal as [ʔ] phonologically speaking, seeing how it is added as a default initial and treated as an allophonic variant of [ʔ] itself.

Ser wrote: ↑Sun Apr 12, 2020 4:17 pm

bradrn wrote: ↑Sun Apr 12, 2020 12:03 am

Nortaneous wrote: ↑Sat Apr 11, 2020 10:29 pmSure, place neutralization processes. If you have total place neutralization, it's usually to glottals, but for nasals it's usually to velars, as in Spanish. In Kagoshima Japanese, for example, final vowels drop under circumstances unclear in my source, after which final non-nasals become glottal, and final nasals become velar:
/kɯtsɯ/ 'shoe' > [kɯʔ]
/midzɯ/ 'water' > [miʔ]
/hirɯ/ 'noon' > [hiʔ]
/kami/ 'hair' > [kaŋ]

That makes sense. But I’m not sure I would agree that the velar nasal here is phonemically glottal.

In Cantonese, [ʔ] and [ŋ] are allophones of the "zero-initial", or default initial if you will of the C(w)V(m,n,ŋ,p,t,k) syllable structure. Middle Chinese distinguished /ʔ/ and /ŋ/ as different initials, but these merged in Cantonese, with [ʔ] tending to appear before the phonemes /e~ɛ:, ø~œ:, ɐ, a:, o~ɔ:, ʊ~u:/ in high-register tones or the mid-level tone, and [ŋ] before those same vowel phonemes in low-register tones. (Before /ɪ~i:, y:/, they have further merged with /j/ and now /j/ is typically used: [jɪp ɥy:n], although these may sometimes be [ʔɪp ʔy:n] among some people.) However, /ŋ/ still had some faintly marginal existence as a phoneme in colloquial words of probably non-Sinitic origins (e.g. 啱啱 [ŋa:m˥ŋa:m˥] 'just now, just a moment ago', also [ʔa:m˥ʔa:m˥]) in the mid-20th century, and at least in more formal registers people would also use it before vowels with mid-level tone.

However, in recent decades, [ʔ] has taken over most contexts of [ŋ] in colloquial levels of the language, so that now [ŋ] is mostly just found before /ɐ, a:/ in low-register tones (and even there [ʔ] can sometimes be heard too). This has also meant that initial [ŋ] in its previous contexts has become associated with formal Cantonese in the cities, less urban ways of speaking, old folks' way of speaking, classic pop or rock songs, and the Cantonese of people overseas. I have come across YouTube videos of young Cantonese speakers born overseas where people from Hong Kong make comments expressing their shock to hear such young people use plenty of initial [ŋ] in words like 我 [ŋɔ:˩˧] 'I, me' (instead of [ʔɔ:˩˧]), a habit naturally acquired from their parents and a lack of contact with mainstream Cantonese culture in southern China / Hong Kong.

I had forgotten about Cantonese /ŋ~ʔ/ — thanks for reminding me!

To me, this whole thing looks like [ŋ] being treated as equally glottal as [ʔ] phonologically speaking, seeing how it is added as a default initial and treated as an allophonic variant of [ʔ] itself.

But I’m not sure how this follows. A phoneme can easily have two different allophones at two places of articulation, but that doesn’t necessarily mean that you have to treat those allophones as having the same place of articulation in some vaguely undefined way…

Regarding Aklanon, see https://en.wikipedia.org/wiki/Aklanon_l ... #Phonology where they mention the orthography.