zompist bboard

Lately I've noticed something about languages that have phonemic retroflex consonants. ("Retroflex" is used here in a broad sense, encompassing sub-apical palatals, apical postalveolars, and laminal postalveolars with no palatal element.) In the overwhelming majority of cases, these contrast with another posterior coronal series with a palatal element, whether that be palatoalveolar, alveopalatal, or the more forward/coronal variant of "palatal". Also, maybe this is just me displaying ignorance here, but in every case I can think of where a pre-retroflex phase of the language's development is known the language developed the retroflex series (or converted some existing series with a different POA to retroflex) no earlier than, and generally after, the palatal series developed. (The two main sources appear to be shifting some other postalveolar series to retroflex due to pressure from similar consonants, as in Polish and Russian, and interactions between anterior coronals and rhotics or other singleton retracted consonants as in Old > Middle Chinese and Proto-Indo-Iranian > Proto-Indo-Aryan. The Proto-Indo-Aryan example also ends up falling into the first case since the ruki rule sibilant shifted to retroflex as palatalization diachronics introduced an alveopalatal sibilant.)

Could it be, then, that we can state a relatively strong cross-linguistic tendency that languages will only develop a series of contrastive retroflex consonants when some other postalveolar series already exists? In other words, retroflexes are a more marked postalveolar series than the palatal(ized) ones, in a way beyond just being less common cross-linguistically in general, and so will typically not develop unless these are also present?

Qualifications: I'm restricting this to phonemically retroflex consonants, not to consonants with retroflex allophones, e.g. most coronals in Norwegian. (Although the environment that conditions these does fall into the "interactions with rhotics" pattern.) I'm also aware that some languages have a lone retroflex in a rhotic, so there may be some cases where this occurs despite no other postalveolars being present. Finally, there could be cases where a series that would most properly be transcribed as retroflex is loosely transcribed as something else (e.g. palatoalveolar) because of the absence of other postalveolars to contrast with.

Having said all this, I would love to proven wrong! If you have evidence against my proposal here, tell me!

I think the notation factor is more significant than you seem to think. How do you know if some miscellaneous language with a consonant transcribed with [ʃ] doesn't realize it as a "retroflex"? That aside, it's hard to evaluate because for a language to lack any palatal or palatalized consonants seems unusual to me in itself. I assume you're not requiring the language to lack /j/, but /tʃ~tɕ/, /dʒ~dʑ~ɟ~ʝ/, /ɲ/ and /ʎ/ are also fairly common. Is it OK if the language has palatals/palatalized consonants, but doesn't contrast palatal and retroflex sibilants of the same manner of articulation and voicing? One possible counterexample that comes to my mind is Old Spanish, which had alveolar, apicoalveolar, and post-alveolar sibilants; the latter seem to have developed into the velar fricative /x/ through an intermediate "retroflex" kind of stage. But Spanish also developed palatal /tʃ~tɕ/ and /dʒ~dʑ~ɟ~ɟʝ~ʝ/. Vietnamese is currently transcribed with retroflex /tʂ/ /ʂ/ /ʐ~r/, dental/alveolar /s/ /z/, and the palatals /c/ /ɲ/ /j/; although in different dialects and past speech the dental/alveolars have been connected to palatal consonants.

I think Sard is a counter-example— it has /ɖ/ but its /t d/ are dental, and there are no post alveolar or palatal stops.

(Or maybe it isn't, because it has post-alveolar fricatives? I'm not sure I understand your thesis.)

zompist wrote:I think Sard is a counter-example— it has /ɖ/ but its /t d/ are dental, and there are no post alveolar or palatal stops.

(Or maybe it isn't, because it has post-alveolar fricatives? I'm not sure I understand your thesis.)

I was including fricatives. My thesis was basically that a language is unlikely to develop retroflex consonants (of any kind) unless there are other postalveolar coronal consonants (of any kind) for them to contrast with. I was mainly thinking of stops, fricatives, and affricates here, but it wouldn't have to be limited to those.

A counter, off the top of my head, is Mixe languages. They have a near-complete shift of s>ʂ. [ʃ] appears but, at least in the South Highland that I've looked at the most, only as palatalized /ʂ/ (all consonants can be palatalized). Looking up a few, apparently Proto-Zapotec had both retroflex and non-retroflex, but some modern dialects merged them as retroflex. Afar (Cushitic) has /ɖ/ with no other postalveolars of any kind, I presume from an emphatic *ɗ. In Formosan languages, Puyuma has /ʈ ɖ/, and while it palatalizes /s/ to [ʃ] before front vowels, that it's still allophonic points to, though isn't conclusive of, a more recent origin than the retroflexes. Rukai has /ɖ ɭ/ according to Wikipedia and no other postalveolars, though according to the grammar of Puyuma I consulted /ɖ/ at least is from Puyuma loans.

My own idiolect of English has very strongly retroflexed /ʃ/, /ʒ/, and /r/. And of course no palatals. It's just an idiolect but I don't see why it couldn't become widespread. Also, I don't have retroflexes in /tʃ/ or /dʒ/; it's just the bare fricative forms.

Also, the retroflex /r/, at least, is fairly widespread in English even if the retroflex fricatives aren't.

Does that contradict what Chengjiang was saying, though?

I would agree with Sumelic here. The retroflex series are usually used in the description of a language in order to distinguish them fron another marked coronal series, e.g. palato-alveolars. If it's not the case, they are transcribed as /ʃ ʒ/ instead.

For example, there is no audible difference between /ʃ/ in Lithuanian and /ʂ/ in Polish or Russian, but Lithuanian lacks /ɕ/, which is present in both Polish and Russian.

Chengjiang wrote:Qualifications: I'm restricting this to phonemically retroflex consonants, not to consonants with retroflex allophones, e.g. most coronals in Norwegian. (Although the environment that conditions these does fall into the "interactions with rhotics" pattern.)

For the record, it's problematic to label the retroflex series in Norwegian/Swedish merely allophonic. They can be considered allophonic in cases of rC[+coronal] > C[+retroflex] across morpheme boundaries since the rule applies productively, but when they occur internally in words like kart I think it's preferable to treat the segments as phonemic because the application of the rule isn't quite regular. Kristoffersen devotes a lot of time to this subject in The Phonology of Norwegian.

With that said, I think the Scandinavian retroflexes might be better described as simply "apical coronals", because at least IMD they can actually be realized as alveolars or even dentoalveolars. I'm not sure if that would be considered to fall under the "retroflex" umbrella.

Thanks for the replies, everyone.

Pole, the wrote:For example, there is no audible difference between /ʃ/ in Lithuanian and /ʂ/ in Polish or Russian, but Lithuanian lacks /ɕ/, which is present in both Polish and Russian.

I thought Lithuanian had contrastive palatalization on most consonants, though, including the postalveolars. Correct me if I'm wrong, but isn't /ʃʲ/ more or less [ɕ]?

Sumelic wrote:I think the notation factor is more significant than you seem to think. How do you know if some miscellaneous language with a consonant transcribed with [ʃ] doesn't realize it as a "retroflex"?

I do realize this is a potential problem. I've been looking up articulatory descriptions of the consonants in any language I look at that is listed as having palatoalveolar consonants if it's not familiar to me. I figure consonants transcribed as alveopalatals are pretty unlikely to actually be non-palatalized.

That aside, it's hard to evaluate because for a language to lack any palatal or palatalized consonants seems unusual to me in itself. I assume you're not requiring the language to lack /j/, but /tʃ~tɕ/, /dʒ~dʑ~ɟ~ʝ/, /ɲ/ and /ʎ/ are also fairly common.

I certainly wasn't expecting the language to lack /j/. I was talking about palatalized coronal consonants, so palatoalveolars, alveopalatals, and the more front versions of "/c ɟ ɲ ʎ/". (I.e. palatalized laminal postalveolars, which in Sinological circles would probably be written /ȶ ȡ ȵ ȴ/.)

Is it OK if the language has palatals/palatalized consonants, but doesn't contrast palatal and retroflex sibilants of the same manner of articulation and voicing?

By "Is it OK", do you mean "Is it a counterexample?" If that is what you mean, I'd say I don't think it's a very strong one.

Magb wrote:For the record, it's problematic to label the retroflex series in Norwegian/Swedish merely allophonic. They can be considered allophonic in cases of rC[+coronal] > C[+retroflex] across morpheme boundaries since the rule applies productively, but when they occur internally in words like kart I think it's preferable to treat the segments as phonemic because the application of the rule isn't quite regular. Kristoffersen devotes a lot of time to this subject in The Phonology of Norwegian.

With that said, I think the Scandinavian retroflexes might be better described as simply "apical coronals", because at least IMD they can actually be realized as alveolars or even dentoalveolars. I'm not sure if that would be considered to fall under the "retroflex" umbrella.

Interesting. I hadn't heard of speakers realizing them as apical anterior coronals. Do they still contrast with the primary (laminal?) coronal series in that case?

At any rate, I am acknowledging Norwegian as a possible counterexample to this hypothetical tendency, alongside Afar as a definite counterexample and other parts of vokzhen's post as at least partial counterexamples. I still suspect this is a fairly strong tendency, but it's clearly not a universal.

Chengjiang wrote:Interesting. I hadn't heard of speakers realizing them as apical anterior coronals. Do they still contrast with the primary (laminal?) coronal series in that case?

Yeah, they still contrast with the laminal alveolars. The difference is fairly subtle, but one helpful cue is that the preceding vowel is usually apicalized. I suspect I could distinguish between words like mat /mɑːt/ and malt /mɑːʈ/ even if the final consonant had been removed from the audio, because of the apicalization of the vowel.

By the way, only /ɳ ʈ ɖ/ can be alveolar for me. /ʂ/ is always laminal postalveolar and /ɭ/ can't be apical dental/alveolar because that's already what my regular /l/ is, so I think my /ɭ/ is usually postalveolar. (I actually find /l/ and /ɭ/ pretty difficult to distinguish out of context.) My /ɽ/ is more of a "true" retroflex, pronounced with the tongue initially curled pretty far back and "slapping" against the alveolar ridge. In other words, it's complicated.

I also have no idea to what extent this stuff applies to Norwegians in general, but I do know that referring to the Norwegian retroflex series as simply "apicals" is quite common. To me they sound very different from the retroflexes in e.g. Indian languages.

On a slightly different but related note, how many languages do we know of which allow initial retroflexes? Having read about the Australian languages they generally don't seem to allow initial retroflexs and I'm not sure the Dravidian languages do either (correct me if I'm wrong). Further to this, if initial retroflexes occur, how might they have come about? Not many languages have the Tibetic initial rC clusters from which such consonants might arise, so where might they come from if they don't have that?

Malayalam does have them but only in onomatopeia, certain proper nouns (actually I can only think of one at the moment...), and (mostly English) loanwords AFAICT. (Some?) Indo-Aryan languages have word-initial retroflexes, and I'm pretty sure some Dravidian languages (e.g. Kannada and Telugu) have them, too (in other contexts besides the ones I just mentioned).

Polish and Russian have retroflex fricatives and certainly allow them initially.

In Polish and Russian these are depalatalized postalveolars, so it shouldn't be surprising.

A bunch of Californian languages, including Chamariko, Esselen, Obispeño Chumash, the Pomo languages, Salinan, the Yok-Utian family, Yuki, the Yuman langauges, and Wappo contrast front (dental or interdental; occasionally alveolar, usually transcribed as t) and back (alveolar or postalveolar; usually transcribed as ṭ) coronal stops, the latter of which would, in most cases, qualify for Chengjiang's broad definition of "retroflex." Actually, as I understand it, even Hindustani retroflex stops are sometimes allophonically realized as alveolar, so we may be dealing with basically the same phone in all of these cases.

All of these languages (with the exception of the Tachi, Yawelmani, and Yowlumni varieties of Yokuts) and also have t͡ʃ, so they don't contradict Chengjiang's hypothesis. As for initial "retroflexes," I couldn't verify that all of these languages allow them, but for the most part they do. Perhaps also of note is that, in pretty much all of these langauges, ṭ is frequently affricated to some degree, and I don't think it ever contrasts with a retroflex affricate, except possibly in Yawelmani Yokuts (which, as noted before, lacks t͡ʃ).

And now, just for fun, some examples I looked up to verify the presence of initial ṭ:

Chimariko: ṭamma "salmon meal"
Kashaya (Pomoan): ṭheqhále "elderberry"
Mutsun (Utian): ṭi﮲nuy "to wring out"
Salinan: ṭam "house"
Southern Sierra Miwok (Utian): ṭuhyu﮲yu "daddy longlegs spider"
Tulamni Yokuts: ṭiñak "nose"
Wappo: ṭ'ohtaʔ "catch (past tense)"

Vijay wrote:Malayalam does have them but only in onomatopeia, certain proper nouns (actually I can only think of one at the moment...), and (mostly English) loanwords AFAICT. (Some?) Indo-Aryan languages have word-initial retroflexes, and I'm pretty sure some Dravidian languages (e.g. Kannada and Telugu) have them, too (in other contexts besides the ones I just mentioned).

A consonant found only as the onset of a non-initial syllable can generally be pronounced word initially, but I suspect there is a tendency to assume one has misheard if one hears it in that place. For example, word-initial (and word-final) /ʒ/ is pretty unstable in Enɡlish, and the absence is probably similar to the near-absence of word-initial /r/ in many Eurasian languages. One can then find that a massive influx of foreign words can stabilise such sounds in a language, such as /v/ (Romance loanwords) and /z/ (Greek? loanwords) in English.

Sanskrit and Pali do have word-initial retroflexes, but they're very rare.

CatDoom wrote:Actually, as I understand it, even Hindustani retroflex stops are sometimes allophonically realized as alveolar, so we may be dealing with basically the same phone in all of these cases.

My understanding of retroflex stops in Hindi/Urdu is a bit different, namely that they're produced more like alveolar stops (with the tip, but apparently never the blade, of the tongue against the alveolar ridge) than retroflex stops in Dravidian languages are.

CatDoom wrote:interdental

Interdental stops exist? Cool. I thought of using them in a conlang but I was afraid they'd be too weird.

Southeast Asia and the Transhimalayan region have plenty of initial retroflexes. Chinese, Tibetan, Vietnamese, Qiang, Naxi, Yi/Nuosu/Loloish, Kra, and Hmong, for the most part, are based around monosyllables and have retroflexes. Several of those (Tibetan, Chinese, Vietnamese) formed them from Cr clusters, not rC (initial rC clusters in Tibetan actually just dropped the r-, not formed retroflexes).

EDIT: Some other, non-CV-monosyllable ones in the general region include a couple of the Formosan languages, rGyalrong, some Kiranti languages, and Burushaski that all have initial retroflexes.

Zaarin wrote:Interdental stops exist? Cool. I thought of using them in a conlang but I was afraid they'd be too weird.

Apparently the dental consonants in some Australian languages are pronounced interdentally as well. Some dialects of Mapuche do as well. As far as I know, no language contrasts interdental and dental consonants, so the difference isn't all that relevant. It appears that there may be a correlation between languages that contrast dental and alveolar stops and those that produce dental stops interdentally; the latter could hypothetically be an adaptation to maximize the articulatory difference between the two places of articulation.

If we're counting the Scandinavian retroflexes, Elfdalian apparently has initial /ɽ/. I don't know how it developed or how it's distributed though. From Yair Sapir: Elfdalian, the Vernacular of Övdaln - an article with an outline of Elfdalian (history, background, linguistic features, present:

Yair Sapir wrote: Old short l is often realized as /l/. Dalecarlian vernaculars have the peculiarity of having /ɽ/ even word-initially, as in luv /ɽʏːv/ ‘permission’.

It's the only Scandinavian dialect with any initial retroflexes that I'm aware of.

Chengjiang wrote:
I thought Lithuanian had contrastive palatalization on most consonants, though, including the postalveolars. Correct me if I'm wrong, but isn't /ʃʲ/ more or less [ɕ]?

The newest known for me work on Lithuanian (Abrazas et al. by Wikipedia) says that /ʃ/ and /ʃʲ/ are actually something like [ʃˠ] or [ʃʷ] (*[ʃˠʷ]) and [ɕ], so I think you are right, but we should have heard present standard Lithuanian to be able to confirm or deny it.