Retroflex approximants in AE dialects

While looking up the best way to describe the aboriginal pronunciation of Uluru (/uluɻu/), I stumbled across retroflex approximants. The linked Wikipedia page states:

The retroflex approximant occurs in American English, Mandarin Chinese, Pashto, a few Brazilian Portuguese dialects ...

It then proceeds to state that red is pronounced as /ɻʷɛd/ in certain American dialects. Going by the audio sample provided on the page, I find this a little hard to imagine. The article links to another wiki on English phonology which confirms that:

Depending on dialect, /r/ may be an alveolar approximant [ɹ], postalveolar approximant, retroflex approximant [ɻ], or labiodental approximant [ʋ].

So, who in the US pronounces red as /ɻʷɛd/? How did this come about?

Solution 1:

Apparently this is considered the normal, default pronunciation in General American. Look at, and more importantly, listen to red at the Sound Comparisons site. Notice that the General American pronunciation is given as [ɻɛd], not [ɹɛd].

On the other hand, apart from the up-talking teen-aged boy’s pronunciation provided in the “General American” version, all the other North American versions use the [ɹ] version instead of the [ɻ] version.

I must confess that I myself am hard-pressed to hear any difference between those. Probably that’s because it’s not a phonemic distinction, but possibly because it doesn’t actually exist.

Another place that is alleged to have the retroflex version is Lewis, which I presume they mean the Isle of Lewis in Scotland. They give [ɻɛ ̝ˑd ̥] as the pronunciation. I can’t hear that being a great deal different either, so the retroflex aspect isn’t obvious to me.

If you really want a different r, listen to the one from Buckie, Coldstream, or South Wales, where it is a true rolled [r]. You might also check out their respective pronunciations of rain, right, and ring, which all display the same sort of regional variation as red does. Oddly, the Lewis pronunciation of ring sounds like a rolled retroflex to my ear.

Solution 2:

There are two distinct realizations of the /r/ (as in red) consonant phoneme in American English. Though the articulations are completely different, they sound quite similar.

• There's an coronal realization [ɻ], often described as somewhat retroflexed, which is the one you read about. I think that this articulation is less retroflexed than the Mandarin /ɻ/, which sounds quite different. It could more broadly be described as postalveolar. For some reason, descriptions of American English almost invariably pretend that this is the only realization in use.

• There is also a dorsal realization, sometimes transcribed [ψ]. Its existence is well-documented, but it's not widely taught (presumably because it's harder to describe). This phone is called the molar or bunched r. It can be described roughly as a back-palatal or pre-velar approximant that's somewhat bunched up along the left-right axis, producing a grooved articulation.

Because the acoustic difference is small, and the [ψ] realization isn't widely discussed, both realizations are often transcribed ⟨ɻ⟩ in the literature, and it will likely be hard to find information about the distribution of the articulations, or about where they came from.

More information:

• How is the pronunciation of r before th? Specific case: "north"
• IPA transcription of the American English “bunched” /r/

Labialization

People don't actually say [ɻʷɛd] (or [ψʷɛd]); initial /r/ is labiodentalized ([ɻᶹɛd] or [ψᶹɛd]), not rounded. Presumably this is where the [ʋ] realization (occurring occasionally as a speech impediment, apart from some dialects) comes from, from correctly imitating the lip movement but not the tongue position.

The degree of labialization is not phonemic, and assimilates according to context. For example:

• ride /rajd/: usually labiodentalized
• fried /frajd/: definitely labiodentalized rather than rounded
• pride /prajd/: usually bilabialized (compressed)
• Uruguay /(j)ur.rʉ/-: usually rounded

I don't labialize syllabic or coda /r/ at all, unless adjacent to a labial consonant. This probably varies by speaker.

Regardless, you're unlikely to be misunderstood, since the difference in sound is small.

Solution 3:

Here's one of the seven variants listed in Dialect Blog's The Wild World of the English “R”...

The “American” R: /ɻ/ (Retroflex approximant) Similar to the “velar approximant” described above. It is pronounced the same way, except the tongue is curved back just behind the alveolar ridge. You hear this most commonly in American and some Irish accents.

...so I guess the answer will be "quite a few", given that it's actually identified there as "American".