Phonemes which are found in <5 languages or so

[Qwynegold]

There was some interdental consonant transcribed as h̪͆ (or was it x̪͆?) in some now extinct (?) language, but I can't remember which. Does anyone recollect this?

There's the voiceless bidental consonant /h̪͆/, which is transcribed /x̪͆/ in the context of the only language it is known to exist in: the Black Sea dialect of Shapsug (which is a western/Adyghe variety of the Adyghe-Kabardian continuum). I don't know whether it is alive or extinct.

Oh right, it was that one. Thanks!

[Pole, the]

Malayalam has both /ɕ/ AND /ʃ/ (definitely NOT */ʂ/). How common is that?

Depends mainly on how you define /ʃ/ and /ʂ/. Polish /ʂ/ used to be transcribed /ʃ/. Mandarin /ʂ/ used to be transcribed /ʃ̺/. If we take the most conservative definition of “retroflex” into account (i.e. “pronounced against hard palate”), then both of those count. (And probably many other, too. In general, I wouldn't classify the distinction as “uncommon”, especially in fricative-heavy inventories.)

[Vijay]

I'm a little confused about what your point is. Are you just saying /ʃ/ has historically been used to transcribe retroflex fricatives, too? Well, Malayalam doesn't have any retroflex fricatives, so that's not what I meant by /ʃ/.

[Pole, the]

I'm a little confused about what your point is. Are you just saying /ʃ/ has historically been used to transcribe retroflex fricatives, too?

No, I'm saying /ʃ/ has been used to transcribe the fricatives of Polish and Mandarin.

If you're saying the Malayalam fricative commonly described as “retroflex” isn't strictly retroflex, the same could be said about the two other languages.

[Vijay]

The Malayalam one isn't retroflex at all. I pronounce it literally the exact same way I pronounce <sh> in English, and I've never heard it pronounced any other way, nor have I ever heard of any native speakers of English pronouncing <sh> differently from me (i.e. it's not simply that I'm making a retroflex stop without being aware of it). Even right next to a retroflex stop, it's still not retroflex. The linguistic literature on Malayalam tends to be fairly unreliable because the few people who ever bother to work on it can make up whatever bullshit they want and no one will challenge them on it because nobody cares to check their data. I doubt the wiki article on Malayalam claims it's retroflex for any reason other than a) Panini said it was retroflex in Sanskrit, b) it still is for at least some speakers of some other Indian languages, and c) it looks neat on the chart.

Also:

In general, I wouldn't classify the distinction as “uncommon”, especially in fricative-heavy inventories.

Malayalam only has four fricatives, and the only one that ever appears outside of loanwords is [ɕ] (and even then, there aren't all that many such native words, and it seems to be in alternation with [t͡ʃ] in almost all of them). [f] appears almost exclusively in loanwords from English.

[Pole, the]

Well, that pretty much sums up my feelings, too, at least for Polish.

[Vijay]

For Polish maybe (although I dare say there are far more people working on Polish than there are on Malayalam, so there I think it's a little more complicated than people making shit up just because they feel like it). For Mandarin...I mean, in Mandarin, my understanding is that [ɕ] and [ʂ] are in complementary distribution, in which case they're not even separate phonemes to begin with. There may be varieties where they contrast, though; I'm not sure.

[finlay]

θ̪

*snorts with stifled laughter*

But seriously yeah Phoible is fairly useless in several respects: another problem I find is how it sticks to the traditional IPA chart so rigidly that it leaves things like ejectives, aspirated consonants and affricaes out of the main table in a separate list which is just like a pain to read. Just take the Archi entry for example; that must be like 5/6ths or more of the inventory that isn't in the actual table.

what the fuck. what use is that, even?

(1 thing I don't like a lot is when people use e̞ and o̞ for true mid vowels, as the IPA guidelines will literally tell you to use e and o - and in this case it just makes the chart unreadable)

(looks like it has a lot of data, though, never knew about it before)

[Vijay]

1 thing I don't like a lot is when people use e̞ and o̞ for true mid vowels, as the IPA guidelines will literally tell you to use e and o

Same. In general, I think it's weird (though probably not all that surprising...) how some of the more obscure symbols/diacritics are incredibly popular among Internet nerds.

[Soap]

θ̪

It could be a way of noting a true /θ/ as opposed to the less common https://en.wikipedia.org/wiki/Voiceless ... _fricative , if a lanuggae contrasts both.

[Sumelic]

θ̪

*snorts with stifled laughter*

But seriously yeah Phoible is fairly useless in several respects: another problem I find is how it sticks to the traditional IPA chart so rigidly that it leaves things like ejectives, aspirated consonants and affricaes out of the main table in a separate list which is just like a pain to read. Just take the Archi entry for example; that must be like 5/6ths or more of the inventory that isn't in the actual table.

what the fuck. what use is that, even?

(1 thing I don't like a lot is when people use e̞ and o̞ for true mid vowels, as the IPA guidelines will literally tell you to use e and o - and in this case it just makes the chart unreadable)

(looks like it has a lot of data, though, never knew about it before)

While it's true that sticking diacritics all over the place leads to harder-to-read transcriptions, I don't think the IPA's policy of using the technically "close-mid" symbols as de facto mid vowel symbols, or the technically "open front" symbol as a de facto open central vowel symbol, is ideal either. In something like the Phoible table, this would lead to the symbols appearing in the table, but with misleading labels and positions. I think it would have been better if the IPA had either defined separate symbols for an additional level of height (as in CanIPA) or made it so that the technical values of <e>, <o> and <a> were just mid front, mid back, and open central.

[KathTheDragon]

The problem is that phonemes are not really comparable (as Zompist has said somwhere, though I can't recall where off the top of my head), even if they're labelled the same, because they have different ranges of allophones.

[Pole, the]

The problem is that phonemes are not really comparable (as Zompist has said somwhere, though I can't recall where off the top of my head), even if they're labelled the same, because they have different ranges of allophones.

This. The vowels in IPA are just symbolic representations of sections of a continuous vowel space and adding new symbols / reassigning existing ones will not fix anything.

[Sumelic]

The problem is that phonemes are not really comparable (as Zompist has said somwhere, though I can't recall where off the top of my head), even if they're labelled the same, because they have different ranges of allophones.

This. The vowels in IPA are just symbolic representations of sections of a continuous vowel space and adding new symbols / reassigning existing ones will not fix anything.

On a sufficiently abstract level, that's true, but on a more concrete level, it seems to me that using specific descriptions like "close-mid" to describe what are actually true mid vowels, or "open front" to describe what is actually an open central vowel, and showing the phonemes in the corresponding wrong poisitions on a vowel chart, in contrast with grayed-out "open-mid" and "open central" positions, is misleading, and more misleading than simply labeling <e>, <o> and <a> as "mid front", "mid back", and "open" respectively. Obviously the IPA is not responsible for all of Phoible's foibles, but I think it contributed to the issues with the way Phoible represents vowel inventories.

The IPA is supposed to have a phonetic basis; of course that's just an approximation or abstraction, but in the context of trans-linguistic comparisons (as in Phoible) it's not just meant to be used as a convenient set of symbols that one can draw from when desigining phonological transcription systems. If that was all it were, it would be acceptable to use things like <c> = /ts/, <č> = /tʃ/; or <y> = /j/, or <ł> = /ɬ/; but none of these are standard IPA phonetic transcriptions. Officially defining certain symbols in an overly specific way and then telling people to use them in more vague ways based on the absence of other phonemes in the nearby area of the vowel space seems like a more confusing way of doing things than officially defining the symbols in vaguer terms and then allowing them to be used more specifically in the presence of other phonemes in the nearby area of the vowel space (with diacritics as an option if you want to indicate the same level of specificity in other contexts).

[gach]

On a sufficiently abstract level, that's true, but on a more concrete level, it seems to me that using specific descriptions like "close-mid" to describe what are actually true mid vowels, or "open front" to describe what is actually an open central vowel, and showing the phonemes in the corresponding wrong poisitions on a vowel chart, in contrast with grayed-out "open-mid" and "open central" positions, is misleading, and more misleading than simply labeling <e>, <o> and <a> as "mid front", "mid back", and "open" respectively. Obviously the IPA is not responsible for all of Phoible's foibles, but I think it contributed to the issues with the way Phoible represents vowel inventories.

It's a discussion that pops up regularly, so why not add my general thought as well. Phonemic description is in itself necessarily a level of abstraction and while the labels you choose for the phonemes should have a good grounding in the phonetic reality of the language, they should also be practical in use. The more exact descriptions of the phonemes should always be given in the form of textual discussion, involving allophony and the relevant contrasts.

As already pointed, vowel phonemes especially exist as diffuse regions of allophony within the vowel space. There might be a point in trying to pick the most appropriate IPA symbols for the vowels by locating the focal points of these regions and taking the nearest symbols. Unfortunately this will often lead to a distracting amount of diacritics and special letter forms. I'd argue that in practice it's much better to choose symbols that are practical in use and the least offensive for the audience. Choosing practical notation for the phonemes will most likely lead to a slight loss of information in the phonemic transcription but that's a price well worth paying for increased clarity. The choice of phoneme letters won't anyway reveal anything about the ranges of allophony on its own.

There's furthermore the matter of what phonemic contrasts are relevant to the phonology. If a language, for example, has only one front unrounded mid vowel, /e/ = [e ~ e̞ ~ ɛ], that only contrasts in height with high and low vowels, how much sense does it make to argue if it's a "true mid vowel" or something higher or lower than that?

[Travis B.]

Take English (I am using the English I am familiar with out of familiarity) - there are two series of obstruent phonemes, conventionally in IPA transcribed using voiceless and voiced symbols. The problem is the contrast between the two really is not one of voicing except between sonorants (and not even then necessarily, as /t d/, due to /t/ voicing, and /tʃ dʒ/, due to /dʒ/ remaining voiceless, typically do not contrast for voice intervocalically), and for fricatives, initially. Thus it is best to describe this contrast as a fortis-lenis contrast, one for which we are just using voiceless and voiced IPA symbols out of familiarity and convenience. Therefore, using IPA for phonemes is really just a matter of convention rather than IPA describing phonetic reality.

[gestaltist]

Nobody mentioned the whistled sibilants of Shona yet. They don't even have an agreed-upon IPA transliteration.

[Vijay]

Well, I think it's fair to say lots of sounds don't have an established transcription, and those sibilants are pretty common in Southern Bantu languages, IIRC.

[Nortaneous]

low semivowel somewhere in the weird part of Asia, I forget where, Sangtam maybe?

also http://phoiblesearch.herokuapp.com/

the PHOIBLE site itself is open-source and somewhere on Github, but I tried to mess with it once and couldn't make heads or tails of it

[Vijay]

Northeast India is the weird part of Asia?

[gach]

low semivowel somewhere in the weird part of Asia, I forget where, Sangtam maybe?

There's the low glide in Marphali and perhaps in many other languages of the area, if you follow the references. However, isn't this essentially an uvulo-pharyngeal approximant? In that case it doesn't sound different from the glide realisation of the "voiced pharyngeal continuant" aka the "voice pharyngeal fricative", which I think is a common realisation for it.

[Soap]

I read some of the PDF and I can see the argument for calling /ʕ/ a semivowel, and that if it is a semivowel, it's certainly a semivowel of a low vowel such as /ɑ/. And that analyzing it as a semivowel fits in perfectly in a language that already has semivowels /w j/ or at least one of those. I agree with all of that.

Only thing I doubt is that there is an example of /ʕ/ actually evolving from an /ɑ/ or similar low vowel the way /w j/ often evolve from /u i/. The paper seems to indicate it may have come from a rhotic consonant, which means it was never a vowel or even part of a diphthong. So while I think it's interesting that /ʕ/ can be analyzed as a semivowel I wouldnt agree with assigning /ɑ̯/ an existence separate from /ʕ/ and therefore considering it a rare phoneme.

A possible counterargument, i guess, would be if /ɑ̯/ always came with open lips and a depressed frontward part of the tongue. The paper talks about the two /j/'s of Spanish .... one assimilates the lip shape of the adjacent vowel, the other always looks like /i/ ... if the paper goes on to say that the /ɑ̯/ does indeed always have the mouth shape of the full vowel /ɑ/ then I take back what i said.

[Frislander]

I was going to propose the nasalised glottal fricative, but apparently rhinoglottophilia is much more common than I thought.

[Zaarin]

Some dialects of Haida have /ʔʷ/, which is pretty unusual if I'm not mistaken. Also Masset Haida has /ʡ͡ʜ ʜ/.

[Frislander]

Some dialects of Haida have /ʔʷ/, which is pretty unusual if I'm not mistaken.

I think I mentioned this before but this is pretty much universal in the Circassian dialect continuum (this is a result of a historical change of uvular ejectives (labialised or not) to glottal stops), I thought it was the only language which had it but thanks for this additional attestation!