Sound changes before r

I know that Twii (southern Ghana) has [Ir] but I've never heard it pronounced. The book I have indicates that it's very difficult for non native speakers to differentiate [Ir] and [@r].

When I'm trying to pronounce the sounds u're wondering about they seem possible.

Those sounds are possible, but they're difficult to pronounce. That's why the often change. Some work better with tap/trill, while others work better with palatized/retroflexed

It's difficult to make those sounds because the "retroflex r" you're talking about is strictly an approximant (/r\'/ in X-Sampa) and in general these tend towards being nearly vowels themselves at times, making something like /i:r\'/ into practically a dipthong. In English this is often an inherent schwa sound which is our usual end all for a default unstressed vowel sound. This happens because in the production of this r sound, the air passageway is never fully constricted, so you really are saying /i:/ and then altering the vowel quality in the transition to /r\'/. I think it may be impossible, actually, to pronounce them the way you think you want it to be because of this physiological fact. This is the reason English variations like a good new york accent or in R.P., the final /r\'/ (or more often, /r\/) is barely audible or completely muted.

If you still wanted retroflex quality, but not the ickiness of vowel complications, you might consider the retroflex r flap thing prevalent in Indic/Dravidian languages which is what I assumed you meant originally until I kept reading your post. In this case, you actually bring the tip of the tongue in contact with the upper palate. This constricts the passageway, closing the syllable, and the only vowel heard is /i:/, as seemingly desired. It seems to me this can vary into approximant status other times, should your dialect desire, considering the small difference there is physiologically between the two. This doesn't have to be a full flap really, but as long as your tongue rises all the way up to close the air passageway, you won't have the vowel quality changed.

Also, you might consider the theoretical implications, which is interesting but I guess not important for an a priori set of constructed dialects. The flap and the trill, if I know what you're talking about, are related sounds physically, the trill being the more exaggerated mechanism. There is a tendency, laughably, to call all these sounds often represented by the latin character "r" some sort of class of "rhotic sounds," but this is obviously due to our heavy orthographic bias. The approximant "r" in the English word "ear," which I think is the consonant your referring to, is a largely unrelated sound, notably in terms of place of articulation and mechanism. Now suppose there is a word in your dialects that is variably /i:r/ (the trill), /i:4/ (the flap), or /i:r\'/ (the retroflex approximant), and let us also suppose your dialects have not been in contact with other non-related languages, but have diverged simply due to geographic separation or whatever else. Now I can't know this because it's your system. It should be theoretically improbable for your dialects to phonemically differentiate that radically across types of sounds (the flap and trill are very similar, the retroflex approximant very dissimilar), unless all three, or at least both sorts of consonants were present in the parent language and used for different things, and at some point folks A and B preferentially chose to replace all /r\'/ containing words with /r/ or /4/ and C folks preferentially chose to replace all /r/ or /4/ containing words with /r\'/. Alternatively, /r\'/ could have been introduced into group C by interaction with a different system. Thus you might want to flesh out the historical understanding of your dialect diversity.

Ahh, thanks for clearing me up on all that. I thought "retroflex" referred to the American English r, but that's not something I've been able to learn much about.

you might consider the retroflex r flap thing prevalent in Indic/Dravidian languages

What is that, exactly? Is that same sound as a single 'r' in Spanish? (And the 'd' sound in "middle" too) That's what I meant by the "flap" I mentioned before.

My system of dialects is pretty much along those lines, actually. I wasn't sure about the sounds being related, but I wasn't unsure enough to change it, and in any case, nothing is really permanently laid out yet. The phonology has [l] and [d] as related sounds, I suppose, but you're right, it probably makes more sense drop one. The idea was that the proto-lang had only the flap, and then it became a trill in some places, and less stressed in other places, and still remained the same somewhere else. As for interaction with other languages that have /r\'/ in their phonologies, that probably wouldn't have happened, at least not at the point I'm imagining the split took place. The only issue with that is that at the moment the 'r' phoneme is phonologically equivalent to [l] in terms of where it can occur, and I wonder if it will throw off the phonology if the 'r' isn't anything like the 'l', and never was. Mostly I just don't want it to evolve into a non-existant New York 'r' or anything like that, though.

It's the same idea as the spanish intervocalic -r- in terms of mechanism except that the tongue is curled into retroflex position. It actually is originally native to Dravidian languages and was borrowed into Indic. There is no Devanagari character for it in Sanskrit, but a modified one exists in Hindi.

There's a wiki that has a sound sample, although when I hear in Hindi, the tongue tip placement sounds pulled further back.

The term "retroflex" simply refers to place of articulation: here the tongue tip pulled behind the alveolar ridge to varying degrees. It says nothing about actually mechanism of sound production though.

Yeah with the theory stuff, you can do what you want. But in natlangs, you cannot introduce new phonemes into daughter languages by simple sound change mechanisms. I think introduction of a trill is considered a "new" phoneme, but I'm not sure because trills and flaps are so similar in concept. In terms of other "new" phonemes,for example, Hindi and Sanskrit have a fair amount of retroflex consonants. However, retroflex stop consonants are not endemic to the Indo-european family and came by way of contact with Dravidian languages in India.

Yeah, that's kind of what I had in mind.

One thing I thought of doing was just having 'r' have allophones of both of the sounds and then I don't have to pull anything out of the air. I'm not sure what the distribution for those would have been though, so I've still got some work to do. It's not a natlang, and I'm purposefully tramping on some universals for fun, but I don't want it to seem too contrived, and I want it work "naturally" in my artificial conworld. Heh. But didn't you say the trill was closely related to the flap? It's easy enough to imagine the trill devolving into the flap, and any rate.

They seem quite pronounceable to me. Rhotics do have a lowering effect on vowels, but I think the specific cases you mention are just features of the American English dialect(s), where /ir/ and /Ir/ have fallen together. Other dialects distinguish them as normal with the postalveolar /r/, as in 'serious' v. 'Sirius'. I don't know if there are any non-American dialects with a retroflex /r/, and how they'd treat these pairs if so.

the American English dialect(s), where /ir/ and /Ir/ have fallen together.

Yes, that's mine, unfortunately. I can't make [Ir] without pronouncing it [ir] or [3r] without pronouncing it [er]. I pronounce "serious" and "Sirius" exactly the same way...

Canadian English distinguishes (although I don't know how regularly) /or/ from /Ar/, whereas in American English they've merged.

Example: "sorry" CAN: /sori/ AMER: /sAri/

Amerenglish also has /Vr/ -- "or" /Vr/ vs. "are" /Ar/

I am American, and I say /sorri/. I also pronounce "tomorrow" differently than most.

I've never heard "or" pronounced /Vr\/. Usually it's /Or\/, sometimes /3r\/.