bradrn wrote: ↑Wed Nov 13, 2019 3:25 pm
almost no language contrasts /ɨ/ and /ɯ/
I was just looking at the inventories that PHOIBLE says contrast /ɨ/ and /ɯ/---madness!
My impression is that phonologists usually treat /ɨ/ as [-front], grouping it with back vowels; don't know how sensitive that is to sound-change patterns, though, and I'm not sure how that'd work in the languages that, apparently, have the four-way contrast /i ɨ ɯ u/.
One possible variable is that I daresay it's often not entirely clear whether the non-/i/ high unrounded vowel should be considered /ɨ/ or /ɯ/---the Turkish nonfront unrounded high vowel gets transcribed both ways, for example.
bradrn wrote: ↑Wed Nov 13, 2019 3:25 pm
almost no language contrasts /ɨ/ and /ɯ/
I was just looking at the inventories that PHOIBLE says contrast /ɨ/ and /ɯ/---madness!
In what way exactly is that madness? Are you referring to the general unreliability of PHOIBLE, or something else?
My impression is that phonologists usually treat /ɨ/ as [-front], grouping it with back vowels; don't know how sensitive that is to sound-change patterns, though, and I'm not sure how that'd work in the languages that, apparently, have the four-way contrast /i ɨ ɯ u/.
So far, I only found one language with both /ɨ/ and /ɯ/ (Bora), and it has a vowel inventory of /a ɛ i o ɯ ɨ/. Maybe the reason /i ɨ ɯ u/ is so rare is because /ɨ/ and /ɯ/ are both [-front].
akam chinjir wrote: ↑Wed Nov 13, 2019 3:36 pm
I was just looking at the inventories that PHOIBLE says contrast /ɨ/ and /ɯ/---madness!
In what way exactly is that madness? Are you referring to the general unreliability of PHOIBLE, or something else?
(...snip...)
So far, I only found one language with both /ɨ/ and /ɯ/ (Bora), and it has a vowel inventory of /a ɛ i o ɯ ɨ/. Maybe the reason /i ɨ ɯ u/ is so rare is because /ɨ/ and /ɯ/ are both [-front].
Mostly that distinction seems pretty subtle to me, but it's also that a bunch of them have the full /ɨ ɯ u/, which contrasts three [-front] vowels---the approaches to these things I'm most familiar with seem to imply that this is impossible. (More or less, they take vowel place to be coded as [±front] and [±somethingelse] (for the latter, maybe [±peripheral]), which allows only two [-front] places.) Of course those approaches might be wrong, or the PHOIBLE data might be wrong. (Here's the full list of languages with /ɨ/ and /ɯ/.)
akam chinjir wrote: ↑Wed Nov 13, 2019 3:36 pm
I was just looking at the inventories that PHOIBLE says contrast /ɨ/ and /ɯ/---madness!
In what way exactly is that madness? Are you referring to the general unreliability of PHOIBLE, or something else?
(...snip...)
So far, I only found one language with both /ɨ/ and /ɯ/ (Bora), and it has a vowel inventory of /a ɛ i o ɯ ɨ/. Maybe the reason /i ɨ ɯ u/ is so rare is because /ɨ/ and /ɯ/ are both [-front].
Mostly that distinction seems pretty subtle to me, but it's also that a bunch of them have the full /ɨ ɯ u/, which contrasts three [-front] vowels---the approaches to these things I'm most familiar with seem to imply that this is impossible. (More or less, they take vowel place to be coded as [±front] and [±somethingelse] (for the latter, maybe [±peripheral]), which allows only two [-front] places.)
I’m not familiar at all with these sort of featural analyses, but this sounds like a reasonable explanation.
That’s exactly the same search I did, on exactly the same program. But I cross-checked these inventories with the ones on Wikipedia, and most of the time PHOIBLE is wrong. In more detail:
According to a SIL grammar Kenyang does indeed have all of /i ɨ ɯ u/. Wikipedia lists Kodagu (a.k.a Kodava) as having this contrast as well.
Matses and Bora also have both /ɨ/ and /ɯ/, but only because they have /ɯ/ where most languages have /u/.
Miraña is merely a dialect of Bora.
I couldn’t find any other information on Sedik, Kilba (a.k.a. Huba), Mishmi or Tangkul Naga.
PHOIBLE quotes Nimboran as having no rounded vowels, but according to Wikipedia this is incorrect. (But according to another source Nimboran is a standard example of an /ɨ ɯ/ contrast… I’ll put this one down as ‘inconclusive’.)
PHOIBLE quotes Ute as having all of /i ɨ ɯ u/, but according to Wikipedia it only has /i ɯ u/.
PHOIBLE quotes Apatani as having /ɨ ɯ/, but according to a grammar I found it actually has /ɨ u/. This is also the case for Sema and Wayana.
So out of 15 languages listed as having both /ɨ ɯ/, four do actually have this, five may have this, four do not, and one is merely a dialect of another.
Well, Wikipedia may be wrong. Of course, errors usually don't survive long, as anybody who notices one can easily correct it, but how many people do actually know about these obscure languages and can spot and correct errors? Hence, I'd not invest more trust in Wikipedia than in PHOIBLE here.
WeepingElf wrote: ↑Thu Nov 14, 2019 9:19 am
Well, Wikipedia may be wrong. Of course, errors usually don't survive long, as anybody who notices one can easily correct it, but how many people do actually know about these obscure languages and can spot and correct errors? Hence, I'd not invest more trust in Wikipedia than in PHOIBLE here.
This is true. But PHOIBLE regularly gives me crazy inventories, and Wikipedia gives me sane inventories for the same language, so I’d say that Wikipedia is generally more accurate than PHOIBLE.
Nila_MadhaVa wrote: ↑Mon Nov 11, 2019 6:26 am A question that occurs to me in regards to begadkefat is why the lenition hasn’t spread to the rest of the paradigm by analogy?
Well, that would give us a system entirely without singleton stops, if I read you right. We'd have the geminates, but they would only occur between vowels, so that would be typologically unusual. I think if something like that were to happen it would likely affect only a subset of the stop inventory, not the whole thing. e.g. Arabic did /p/ > /f/ without /b/ > /v/, and did /g/ > /dž/ without /k/ > /tš/.
Ontena Gadsup had begadkefat-style lenition with CV(n ʔ) syllable structure, so the consonant inventory is /ʔ ɸ s ɾ x m n j w/ and phonetic plosives only appear in the sequences /ʔɸ ʔs ʔɾ ʔx ns/ [ˀp ˀt ˀd ˀk nt]. It's typologically unusual, but it can happen.
Nila_MadhaVa wrote: ↑Mon Nov 11, 2019 6:26 am A question that occurs to me in regards to begadkefat is why the lenition hasn’t spread to the rest of the paradigm by analogy?
Well, that would give us a system entirely without singleton stops, if I read you right. We'd have the geminates, but they would only occur between vowels, so that would be typologically unusual. I think if something like that were to happen it would likely affect only a subset of the stop inventory, not the whole thing. e.g. Arabic did /p/ > /f/ without /b/ > /v/, and did /g/ > /dž/ without /k/ > /tš/.
Ontena Gadsup had begadkefat-style lenition with CV(n ʔ) syllable structure, so the consonant inventory is /ʔ ɸ s ɾ x m n j w/ and phonetic plosives only appear in the sequences /ʔɸ ʔs ʔɾ ʔx ns/ [ˀp ˀt ˀd ˀk nt]. It's typologically unusual, but it can happen.
Interesting, so even a very unusual change could affect the whole paradigm. That's good to know, thanks for the info.
Does anyone have any advice re my question upthread about which forms would be likely to trigger changes spreading to the entire paradigm? In my lang roots take CVCCV (and sometimes (V)CCVC) forms as the most basic/commonly used stems, but stems of a CVCVC form are the statistical majority (about 3/4 stems are CVCVC). Any help figuring out which forms would be most likely to act as the model for analogy would be very welcome.
Would it be possible for a language to lose /l/ twice in its diachronic history? First everything vocalises, and then /l/ gets back into the phoneme inventory through Arabic and Africal loanwords and promptly gets either devoiced in clusters with voiceless consonants and word-finally and meeged with either /n/ before nasal vowels or /r/ everywhere else?
So the current state of the language is that /l/ is present only in recent loans and viewed as a very much unusual and marked phoneme because of that?
Of course it's possible. Speakers don't know what sound changes happened in the language or what its phonemic inventory was before they were born, and even if they did somehow know this, why would that stop repeated losses of the same phoneme?
WeepingElf wrote: ↑Thu Nov 14, 2019 9:19 am
Well, Wikipedia may be wrong. Of course, errors usually don't survive long, as anybody who notices one can easily correct it, but how many people do actually know about these obscure languages and can spot and correct errors? Hence, I'd not invest more trust in Wikipedia than in PHOIBLE here.
This is true. But PHOIBLE regularly gives me crazy inventories, and Wikipedia gives me sane inventories for the same language, so I’d say that Wikipedia is generally more accurate than PHOIBLE.
My advice would be to treat both PHOIBLE and Wikipedia as monstrously unreliable -- though Wikipedia probably is slightly more accurate -- and (as you seem to do when possible!) go by any available published grammars or peer-reviewed journal articles you can find that someone else may well have misinterpreted.
Whimemsz wrote: ↑Sat Nov 16, 2019 6:48 pm
Of course it's possible. Speakers don't know what sound changes happened in the language or what its phonemic inventory was before they were born, and even if they did somehow know this, why would that stop repeated losses of the same phoneme?
Certainly. Though my instinct in this situation would be to skip the middle stage and just have the language borrow foreign /l/ as [n], [r], etc from the get-go.
Pabappa wrote: ↑Sat Nov 16, 2019 7:52 pm
Certainly. Though my instinct in this situation would be to skip the middle stage and just have the language borrow foreign /l/ as [n], [r], etc from the get-go.
Can it still borrow /l/, though? There are a fair few languages with certain phonemes found only in borrowings, especially if the influence of other languages is considerable.
Plenty of Indo-European languages (Romance, Greek) lost /h/ twice, first (probably) as a laryngeal and then secondly from some other source (*s, *gʰ). But yes, sound change has no memory. I think there's also at least one Austronesian language that underwent the *t > k shift twice.
Pabappa wrote: ↑Sat Nov 16, 2019 7:52 pm
Certainly. Though my instinct in this situation would be to skip the middle stage and just have the language borrow foreign /l/ as [n], [r], etc from the get-go.
Can it still borrow /l/, though? There are a fair few languages with certain phonemes found only in borrowings, especially if the influence of other languages is considerable.
Yes, your original scenario is certainly still plausible.
In the evolution of Latin to Spanish, /ʎ/ [ʎ] was created and lost twice.
First, Latin sequences of liV [lj] and post-vocalic cl gl [kl gl] became [ʎ] in Proto-Western-Romance, e.g. fīlia > *[ˈfi:ʎa], auricula > *[auˈre:ʎa], coāgulāre > *[koaˈʎa:re]. This older [ʎ] is still found in Portuguese (filha, orelha, coalhar) and was common in France as recently as the 19th century (fille [ˈfiʎ], oreille [ɔˈʁɛʎ], cailler [kaˈʎe], now [fij ɔˈʁɛj kaˈje]). However, in Old Spanish it shows up having undergone fortition to [ʒ]: fiia [ˈhiʒa], oreia [oˈɾeʒa], cuaiar [kwaˈʒaɾ], which then become modern [ˈixa oˈɾexa kwaˈxaɾ].
Second, Latin geminate ll and word-initial pl cl fl stay as such in Proto-Western-Romance, and furthermore French conserves them well (with gemination removed for ll): Latin vallem > French val [val], plēnum > plein [plæ̃], clāvem > clé [kle], flamma > flamme [flam]. In Spanish, these ll and pl cl fl do develop into a newer [ʎ] a few centuries after the [ʎ] > [ʒ] change above, showing up in Old Spanish thus: valle [ˈβaʎe], lleno [ˈʎeno], llave [ˈʎaβe], llama [ˈʎama]. However, since the early modern period, Spanish has been undergoing a greater and greater loss of /ʎ/ by merging it into /ʝ/ [ɟʝ ʝ], so that nowadays it only survives in a few pockets in northern Spain and the Andes (in these areas, it might help that Basque, Quechua and Aymara have /ʎ/!).
dhok wrote: ↑Sun Nov 17, 2019 5:42 am
Plenty of Indo-European languages (Romance, Greek) lost /h/ twice, first (probably) as a laryngeal and then secondly from some other source (*s, *gʰ). But yes, sound change has no memory. I think there's also at least one Austronesian language that underwent the *t > k shift twice.
Spanish has arguably lost and regained it many more times. If you trace it all the way back, you have loss of laryngeals, loss of Latin /h/ (and the /h/ that Spanish gained by aspiration of/f/), then in many dialects creation of a new /h/ from /x/, and of another new /h/ from /s/, which has since been lost in some places (particularly in the Caribbean).
My conlang's stress rules are that stress always falls on the last heaviest syllable. Pretonic short vowels syncopate except in CVCVC syllables.
So, this creates syllables (a and ā represent short and long vowels, repsectively) like CaC.Ca(C) and CCā.Ca(C) and Cā.Cā.Ca(C) but I don't know what I can do so that CāC.Ca(C) and CCa.Ca(C) Cā.Ca.Ca(C) are also valid syllable shapes.
Additional information:
It has a triconsonantal root system, not sure if that makes a huge difference for this question
Inventory:
/a e i u /<a e i u>
/aː eː iː uː/<ā ē ī ū>
/aj əj aw əw/<ay ey aw ew>
/aːj əːj aːw əːw/<āy ēy āw ēw>
/m n ŋ/
/p pʼ b t tʼ d k kʼ g ʔ/
/s sʼ z ɬ ɬʼ ɮ x xʼ ɣ h/
/r j w/
Pharyngeal sound usually either lowers the vowel, lengthen the vowel, and (paradoxially) fronten the vowel. Then it's usually deleted or turned into glottals.
Note that pharyngeal sound doesn't necessarily pattern with phyaryngealized consonant. In Arabic, /a/ next to pharyngeal consonant is /æ/, but next to pharyngealized consonant is /ɑ/
Last edited by Xwtek on Fri Nov 22, 2019 9:21 pm, edited 1 time in total.
