Velar versus uvular fricatives

[Chengjiang]

OK, so I've noticed some things about velar and uvular fricatives:

• As far as I can determine, considerably more languages distinguish velar and uvular stops (e.g. /k/ versus /q/) than distinguish velar and uvular fricatives (e.g. /x/ versus /χ/), and most of the languages that do the latter also do the former. I'm sure there are others, but the only example I can think of that distinguishes velar and uvular fricatives without distinguishing velar and uvular stops is Seri. I think the languages surveyed in WALS bear me out on this, but I'm not absolutely sure since although it says how many languages have uvular fricatives it doesn't say how many contrast them with velar fricatives. (It does note that almost all languages with uvular stops distinguish them from velar stops.)
• For many languages I've read phonological descriptions of that have velar fricatives (without contrasting uvular fricatives), it's noted that the velar fricatives have uvular fricatives as allophones, free variants, or dialectal variants. I've seen this for Spanish, Dutch (a prominent dialect difference IIRC), German's ach-Laut, Arabic, Armenian, and numerous others.
• Relatedly, I often see the same dorsal fricative described as velar or uvular in different descriptions of the same language. This almost never happens with stops or nasals, and rarely with approximants or trills. (Admittedly IIRC a true velar trill may not be possible.) Or I'll see a language's dorsal fricative(s) described as one and then hear speech samples that make it sound like the other.
• Fairly often when I hear people describe how to make a [x] their description seems more in line with a [χ], e.g. "deep back in the throat", "gargling", etc. This might just be an aspect of how [x] is perceived by people whose languages don't have it, but it often ends sounding like [x] and [χ] are being treated as interchangeable.
• [χ] can optionally be articulated with a "turbulent" or "rough" quality, approaching a trilled fricative. I personally find the "smooth" and "rough" versions of [χ] to be more readily distinguishable than the "smooth" [χ] is from [x]. Occasional turbulent "[x]" seems to have a uvular component.

Is it possible, then, that the distinction between velar and uvular is less stable for fricatives than it is for stops, and (non-palatal) dorsal fricatives are likely to freely oscillate between velar and uvular realizations as a language evolves?

[Zaarin]

[χ] can optionally be articulated with a "turbulent" or "rough" quality, approaching a trilled fricative.

Speaking for myself, I can't even produce a uvular fricative without trilling it. My /χ ʁ/ are typically [ʀ̥ ʀ].

[Nortaneous]

As far as I can determine, considerably more languages distinguish velar and uvular stops (e.g. /k/ versus /q/) than distinguish velar and uvular fricatives (e.g. /x/ versus /χ/), and most of the languages that do the latter also do the former. I'm sure there are others, but the only example I can think of that distinguishes velar and uvular fricatives without distinguishing velar and uvular stops is Seri. I think the languages surveyed in WALS bear me out on this, but I'm not absolutely sure since although it says how many languages have uvular fricatives it doesn't say how many contrast them with velar fricatives. (It does note that almost all languages with uvular stops distinguish them from velar stops.)

Tibetan dialects, maybe? Qiangic and Naish generally contrast both /k q/ and /x X/, but there might be a language somewhere in there that lost /q/.

[Vijay]

Is it possible, then, that the distinction between velar and uvular is less stable for fricatives than it is for stops, and (non-palatal) dorsal fricatives are likely to freely oscillate between velar and uvular realizations as a language evolves?

I honestly don't see anything wrong with this idea so far.

[Chengjiang]

I'd like to add a rather striking case: Tehrani Persian has a phoneme that represents a merger of Arabic /q/ and /ɣ~ʁ/. It has a stop allophone word-initially (and I think in codas; can't remember for sure) I've seen varyingly transcribed as [q] and [ɢ], and a lenited fricative allophone intervocalically I've seen transcribed as [ɣ] and [ʁ], more often the former. Transcriptions seem to agree on the POA of the stop and the voicing of the fricative, but not vice versa. And a lot of them have the phoneme apparently changing POA when lenited. I'm not sure if it's actually doing that or if [ʁ] is just being broadly transcribed as [ɣ], but it's an interesting contrast.

[Yng]

I'd like to add a rather striking case: Tehrani Persian has a phoneme that represents a merger of Arabic /q/ and /ɣ~ʁ/. It has a stop allophone word-initially (and I think in codas; can't remember for sure) I've seen varyingly transcribed as [q] and [ɢ], and a lenited fricative allophone intervocalically I've seen transcribed as [ɣ] and [ʁ], more often the former. Transcriptions seem to agree on the POA of the stop and the voicing of the fricative, but not vice versa. And a lot of them have the phoneme apparently changing POA when lenited. I'm not sure if it's actually doing that or if [ʁ] is just being broadly transcribed as [ɣ], but it's an interesting contrast.

The phoneme has a stop articulation word-initially and is optionally lenited, I think, everywhere else. Prescriptively the gh and q are separate phonemes (as they still are in Afghan dialects and parts of Iran) and speakers often claim to distinguish them (and maybe do in some words non-initially), but in practice this isn't usually the case in speech. My impression is that the stop articulation is [ɢ] and the fricative allophone is either [ʁ] or [χ] in some cases of voicing assimilation. It's possible though that it varies between [ʁ~ɣ].

Incidentally, the historic /q/ phoneme was not just used in Arabic borrowings - it also represents lots of Turkic borrowings with back allophones of /k/. The /ɣ/, whatever it was, existed in native Persian words but also stood in for Arabic /ɣ/ in loanwords.

[Vijay]

I thought the phonetic realization of /q/ in Persian was just subject to dialect variation; I didn't know/realize it had anything to do with allophonic variation.

My cousin's wife is Iranian American. I'm not sure how her dad's name is spelled in English (maybe Qasim? My dad always spells it "Ghassem," but I think that may reflect his own pronunciation of that name better than it does the actual official spelling), but in Persian, I'm pretty sure it's قاسم. Apparently everyone, including fellow Iranians, pronounces his name differently, so he doesn't really care how anybody pronounces it.

[Quantum]

I've always intuitively felt (or perhaps more just general observation) that guttural sounds (velar to glottal) were the most likely to have free variation. Perhaps due to the dexterity of the front of the tongue as compared to the back. I honestly wouldn't be surprised if there were a language out there with [x~χ~ħ~ʜ~h] or even [q~ʡ~ʔ] or some combination therein. Often times it doesn't even seem clear if [ʁ] represents a fricative or an approximant.

[Sumelic]

I've always intuitively felt (or perhaps more just general observation) that guttural sounds (velar to glottal) were the most likely to have free variation. Perhaps due to the dexterity of the front of the tongue as compared to the back. I honestly wouldn't be surprised if there were a language out there with [x~χ~ħ~ʜ~h] or even [q~ʡ~ʔ] or some combination therein. Often times it doesn't even seem clear if [ʁ] represents a fricative or an approximant.

Voiced fricatives are rarely distinguished from voiced approximants at other POAs as well: how many languages contrast /β/ and /β̞/, or /ð/ and /ð̞/, or /j/ and /ʝ/ (the last is probably the most common).

[Curlyjimsam]

Languages generally have fewer fricative contrasts than plosive ones, I think - perhaps because fricatives (especially non-sibilant ones) tend to be less acoustically distinct from one another (and perhaps pose greater articulatory problems as well). Note how many learners of English struggle to distinguish /f/ and /θ/, how labial and glottal fricatives are sometimes allophonous, etc. The failure to distinguish /x/ and /χ/ may just be another specific instantiation of this.

It is of course a common trend amongst conlangers to have lots and lots of fricatives à la Western European languages, without realising how typologically unusual this is.

[Pabappa]

This reminds me a bit of something I wrote maybe about six years ago on the KneeQuickie:

http://www.kneequickie.com/kq/Europe

After I realized that languages that have /X/ without /q/ were mostly or entirely European. And that European languages also universally fail to distinguish /x/ from /X/. I think I wrote that at about the same time as there was a thread on the ZBB about "Standard Average European". Im still pretty European in the sense that I have a hard time even pronouncing a [q] but I have no problem distinguishing [x] from [X] and no problem using such a contrast in a conlang (though, my days of using postvelar sounds of any kind are probably over).

[WeepingElf]

The claim that the Armenian stop system is "arguably the one closest to the original Indo-European system that still survives" is at least highly doubtful, and under the standard theory simply wrong.

[Vijay]

Languages generally have fewer fricative contrasts than plosive ones, I think - perhaps because fricatives (especially non-sibilant ones) tend to be less acoustically distinct from one another (and perhaps pose greater articulatory problems as well).

Yes, they're definitely harder to produce. All you have to do produce a plosive is block airflow out of your mouth with your lips or tongue at some particular position, whereas to make a fricative, you have to block it just enough to produce turbulent airflow.

[jmcd]

The claim that the Armenian stop system is "arguably the one closest to the original Indo-European system that still survives" is at least highly doubtful, and under the standard theory simply wrong.

He would likely change this among other things six years on. Another potential change would be to mention the likes of Limburgish for an example of pure vowel lngth distinction outwith Finno-Ugric.

[Tropylium]

Another potential change would be to mention the likes of Limburgish for an example of pure vowel lngth distinction outwith Finno-Ugric.

It's also not even a pure length distinction in most of Hungarian (short /a e/ are [ɒ ɛ], long [iː yː uː] do not effectively contrast with short /i y u/ at all), and vowel length in Samic is heavily linked to prosody, so that would leave pretty much just (the more archaic varieties of) Finnic.

On the other hand, as far as I know at least Czech and Lithuanian have pretty regular binary length contrasts? Unless you're counting tone, I guess…

I'm sure there are others, but the only example I can think of that distinguishes velar and uvular fricatives without distinguishing velar and uvular stops is Seri.

Closer to the topic: there might be or might have been a dialect of Pashto that fits the bill — several have /x χ/ continuing Iranian *š *x (though e.g. Wikipedia for some reason transcribes them as /ç x/), and /q/ is normally only found in Arabic loanwords.

[Vijay]

Really? The Northern Pukhto thing? I thought that was just a merger.

[Tropylium]

Really? The Northern Pukhto thing? I thought that was just a merger.

They merge in the northeast, but not everywhere. The usual transcription I've seen (no idea if there's an Arabic script solution too) is x̌ for the reflex of *š, x for the reflex of *x.

[Vijay]

I thought it was merged there and elsewhere it was [ʂ] (I think for the reflex of *š - this is the consonant in the language name itself that we're talking about, right?) vs. (uvular) [x], but maybe that's just in one dialect or some dialects. The first of those is written ښ, and the second is خ. In the dialects where they're merged, AFAIK they're both written خ.

[Nortaneous]

It is of course a common trend amongst conlangers to have lots and lots of fricatives à la Western European languages, without realising how typologically unusual this is.

This isn't limited to Western Europe, of course -- Wakhi, one of the easternmost IE languages, has a fricative inventory of /f v θ ð s z ʂ ʐ ʃ ʒ x ɣ χ ʁ/, Shughni has /f v θ ð s z ʃ ʒ x ɣ χ ʁ h/, Bashkir (Turkic) has /f v θ ð s z ʃ ʒ ɕ χ ʁ h/, Skolt Sami (Uralic) has /f v ð s z ʃ ʒ ʝ x ɣ/, Nivkh (isolate) has /f v s z r̥ r x ɣ χ ʁ/, and Japhug rGyalrong (ST) has /w s z ɬ l ʂ ʐ ɕ ʑ x ɣ χ ʁ/, where /w/ has the allophones [f β].

The claim that the Armenian stop system is "arguably the one closest to the original Indo-European system that still survives" is at least highly doubtful, and under the standard theory simply wrong.

What? Isn't that uncontroversially true under the standard theory?

[Vijay]

It is of course a common trend amongst conlangers to have lots and lots of fricatives à la Western European languages, without realising how typologically unusual this is.

This isn't limited to Western Europe, of course -- Wakhi, one of the easternmost IE languages, has a fricative inventory of /f v θ ð s z ʂ ʐ ʃ ʒ x ɣ χ ʁ/, Shughni has /f v θ ð s z ʃ ʒ x ɣ χ ʁ h/, Bashkir (Turkic) has /f v θ ð s z ʃ ʒ ɕ χ ʁ h/, Skolt Sami (Uralic) has /f v ð s z ʃ ʒ ʝ x ɣ/, Nivkh (isolate) has /f v s z r̥ r x ɣ χ ʁ/, and Japhug rGyalrong (ST) has /w s z ɬ l ʂ ʐ ɕ ʑ x ɣ χ ʁ/, where /w/ has the allophones [f β].

Toda (Dravidian) has a bunch of fricatives, too! It may be a little weird cross-linguistically, but it's even weirder for Dravidian because usually Dravidian languages don't have many fricatives. In Malayalam, you could probably even argue that fricatives are limited to loanwords (though /t͡ʃ/ is native).

EDIT: To be fair, though, Malayalam has A SHIT TON of loanwords.

[Chengjiang]

I've always intuitively felt (or perhaps more just general observation) that guttural sounds (velar to glottal) were the most likely to have free variation. Perhaps due to the dexterity of the front of the tongue as compared to the back. I honestly wouldn't be surprised if there were a language out there with [x~χ~ħ~ʜ~h] or even [q~ʡ~ʔ] or some combination therein. Often times it doesn't even seem clear if [ʁ] represents a fricative or an approximant.

Voiced fricatives are rarely distinguished from voiced approximants at other POAs as well: how many languages contrast /β/ and /β̞/, or /ð/ and /ð̞/, or /j/ and /ʝ/ (the last is probably the most common).

Yes. I know Dutch traditionally distinguishes /v/ and /ʋ/, although some dialects have merged /v/ with /f/, with Afrikaans filling the gap by shifting /ʋ/ to /v/. Castilian and various other dialects of Spanish distinguish /j/ from /ʝ/; I don't know of any examples for the bilabial and dental continuants, although there could be some.

Come to think of it, does any language distinguish /ɣ/ and /ɰ/? I can't think of any, and honestly I see [ɣ] used to represent an approximant in loose transcription (or sometimes even [ɣ̞] in narrow transcription!) so often that it makes me wonder if there's much point in having separate characters for these sounds.

[Zaarin]

I've always intuitively felt (or perhaps more just general observation) that guttural sounds (velar to glottal) were the most likely to have free variation. Perhaps due to the dexterity of the front of the tongue as compared to the back. I honestly wouldn't be surprised if there were a language out there with [x~χ~ħ~ʜ~h] or even [q~ʡ~ʔ] or some combination therein. Often times it doesn't even seem clear if [ʁ] represents a fricative or an approximant.

Voiced fricatives are rarely distinguished from voiced approximants at other POAs as well: how many languages contrast /β/ and /β̞/, or /ð/ and /ð̞/, or /j/ and /ʝ/ (the last is probably the most common).

Yes. I know Dutch traditionally distinguishes /v/ and /ʋ/, although some dialects have merged /v/ with /f/, with Afrikaans filling the gap by shifting /ʋ/ to /v/. Castilian and various other dialects of Spanish distinguish /j/ from /ʝ/; I don't know of any examples for the bilabial and dental continuants, although there could be some.

Come to think of it, does any language distinguish /ɣ/ and /ɰ/? I can't think of any, and honestly I see [ɣ] used to represent an approximant in loose transcription (or sometimes even [ɣ̞] in narrow transcription!) so often that it makes me wonder if there's much point in having separate characters for these sounds.

According to Wikipedia, Avatime contrasts /ɣʷ/ and /w/--and so do the reconstructions of Proto-Italic and Proto-Afro-Asiatic. Given that, I would presume some Afro-Asiatic languages may also have that contrast, but I'm not particularly familiar with Cushitic or Chadic languages.

[Valdeut]

Castilian and various other dialects of Spanish distinguish /j/ from /ʝ/;

Do they really contrast [j] and [ʝ] in the same environment? I don't speak Spanish so I'm actually curious to know, but reading on Wikipedia, I got the impression that /ʝ/ and /j/ have a different distribution. The only place where they overlap seem to be after /n/ and /l/ and here, /ʝ/ is realized as an affricate [ɟʝ]. Are there some dialectal differences, perhaps? Additionally, some speakers may apparently contrast them in words like abyecto ('abject') vs abierto ('opened'), and I have no idea how /ʝ/ is realized here.
(Wikipedia uses "non-syllabic /i/" rather than /j/, but whether and [j] are different phonemes is irrelevant here)

Apparently, there are some dialects with contrast "non-syllabic /u/ and "a rare consonantal /w̝/". This sounds like a contrast between /w/ and /ɣʷ/ (once again, it's not important here whether syllabic /u/ and non-syllabic /w/~/u/ should be analysed as one phoneme). Wikipedia gives the near-minimal pairs:
deshuesar [de̞zw̝e̞ˈsaɾ] ('to bone') vs. desuello [de̞ˈswe̞ʎo̞] ('skinning')

https://en.wikipedia.org/wiki/Spanish_p ... Consonants

[Ser]

Castilian and various other dialects of Spanish distinguish /j/ from /ʝ/;

Do they really contrast [j] and [ʝ] in the same environment? I don't speak Spanish so I'm actually curious to know, but reading on Wikipedia, I got the impression that /ʝ/ and /j/ have a different distribution. The only place where they overlap seem to be after /n/ and /l/ and here, /ʝ/ is realized as an affricate [ɟʝ]. Are there some dialectal differences, perhaps? Additionally, some speakers may apparently contrast them in words like abyecto ('abject') vs abierto ('opened'), and I have no idea how /ʝ/ is realized here.
(Wikipedia uses "non-syllabic /i/" rather than /j/, but whether and [j] are different phonemes is irrelevant here)

There are dialects in South America that distinguish /ʎ/ [j] vs. /ʝ/ [ɟʝ]~[ʝ].

[Sumelic]

Castilian and various other dialects of Spanish distinguish /j/ from /ʝ/;

Do they really contrast [j] and [ʝ] in the same environment? I don't speak Spanish so I'm actually curious to know, but reading on Wikipedia, I got the impression that /ʝ/ and /j/ have a different distribution. The only place where they overlap seem to be after /n/ and /l/ and here, /ʝ/ is realized as an affricate [ɟʝ]. Are there some dialectal differences, perhaps? Additionally, some speakers may apparently contrast them in words like abyecto ('abject') vs abierto ('opened'), and I have no idea how /ʝ/ is realized here.
(Wikipedia uses "non-syllabic /i/" rather than /j/, but whether and [j] are different phonemes is irrelevant here)

There are dialects in South America that distinguish /ʎ/ [j] vs. /ʝ/ [ɟʝ]~[ʝ].

Really? Interesting, I'd never heard of this. I don't speak Spanish, but all the material I've read indicated that de-lateralized /ʎ/ merges with /ʝ/, not with [j].

The spelling suggests, and the analysis on Wikipedia states, that [j] is not an independent phoneme, but an allophone of the vowel /i/.

Interestingly, it seems that word-initial "y" and "hi," which for most modern dialects are described as having identical pronunciations as the phoneme /ʝ/, are distinct in some dialects (something like [ʒ] and [ʝ] respectively), which suggests that word-initial "hi" maybe was once pronounced /j/. Or maybe it's an orthographically conditioned split. Also mentioned some here