R-sounds / "rhotics" -- practical definition?

[Daniel]

As far as I know, when I looked into this a few years ago, there was no great answer. But maybe some discussion here would be interesting or reveal one.

What is an "R sound or a "rhotic"?

As far as I know, it just literally refers to any sound written with an R (or technically Greek rho).

But then why are certain sounds written with R?

There's the slightly broader class of "liquids", and it seems easy enough to separate out the laterals. But that leaves only a subset of rhotics, eliminating some of the fricatives for example.


So just extensionally, rhotics consist of:
--some approximants
--some trills
--some flaps/taps
--some fricatives
--some vowels [word-final in German for example]
maybe more?

Places of articulation are generally coronal but may be gutteral as well. No obvious limits there. Is a bilabial trill considered a rhotic ever?


One possibility is that R is just whatever is left over after the well-defined classes are eliminated-- most languages have one (in Europe anyway), so that residue is written with R and considered a rhotic.

But what's really odd is that more or less we tend hear all of these sounds as /r/ via categorical perception. That's not always true, such as Brazilian Portuguese <rr> which is more like [h], but in general it works.


Questions:
1. Is there any coherent intensional definition for rhotics?
2. Are there any limits? Is there any reason why a nasal or stop or other sound couldn't be considered an "R"?
3. How reliably are these sounds perceived as the same thing? Does categorical perception work in both directions?
4. Are there prototypical R sounds that are found in more languages and are perceived more reliably as rhotics?
5. What's the historical origin of this? Is this just a convention to consider these sounds the same thing, a historical accident? Or is there more going on?

[MalFet]

Ladefoged used to treat rhotics in terms of a "family resemblance" rather than an identity class.

The simplest shibboleth is probably just "non-lateral liquids plus, within specific languages, other sounds that pattern with them". The category is more relevant for phonotactics than phonetics.

[Daniel]

"Family resemblance" makes sense and is fairly similar to what I said, and I can't think of a better explanation.
But that theory is known to fail as a means of categorization-- it isn't deterministic, it doesn't (without very complicated weighting) even properly distinguish boundaries. [That is, surely [l] is more similar to /r/ than Brazilian Portuguese [χ](?), whatever metrics are used.]


One problem is that, from what I've observed, coronals tend to be liquids or trills, while gutturals tend to be fricatives. So family resemblance seems to fail in that there is no real resemblance between the two groups, though of course there may be some exceptions here and there.
Is there acoustic similarity between these groups? Maybe something about F3? That's already beyond my knowledge of acoustic phonetics, but maybe that's what I'm missing.

within specific languages, other sounds that pattern with them

That's an interesting point I'd like to question: it implies that the "weird R sounds" are exceptions, but I don't know if that's true. Are you aware of statistics anywhere? Do coronal trills make up 50% of rhotics cross-linguistically? Or is that just the biased perspective from well-known European languages?

[MalFet]

Like I said, it's primarily a category established and mobilized for its phonotactic properties. You're not really going go get anywhere expecting it to behave coherently in plain phonetics.

[Daniel]

It's interesting to me that no one has a better answer than that. What is the underlying reason that such a phonotactic category would exist if there is no phonetic basis? By that, I mean that we don't generally observe other phonotactic categories established so arbitrarily (right?), but this one seems salient and reliable for the most part.

1. Are there any other similar categories you can think of?
2. What exactly defines this phonotactically? It falls somewhere between normal vowels and normal consonants on the sonority hierarchy?

[freknu]

I don't want to make any claims, and I'm sure I don't know even half of it, but to me it would seem that all rhotics are patterned after trills.

A very broad and generalised transcription would be:

• labial (front) ʙ — bilabial and labiodental
• coronal (middle) r — dental, alveolar, post-alveolar, retroflex, alveolo-palatal, and palatal
• dorsal (back) ʀ — uvular, pharyngeal, epiglottal, and glottal

The basic pattern is then:

rhotic

• trill (continual cycle)
• tap/flap (singular cycle)
• approximant (null cycle) → spirant (fortition) → [?sibilant (fortition)]

The perceptual archetype of a rhotic would be the trill, but the rhotic pathway (similar to plosive lenition: stop → fricative → approximant) allows it to shift depending on the cycle.

I don't know, it's probably not that simple ... but it's the pattern that I'm noticing.

[MalFet]

It's interesting to me that no one has a better answer than that. What is the underlying reason that such a phonotactic category would exist if there is no phonetic basis? By that, I mean that we don't generally observe other phonotactic categories established so arbitrarily (right?), but this one seems salient and reliable for the most part.

1. Are there any other similar categories you can think of?

Absolutely. In fact, I'd say that most typological generalizations about phonological interaction/phonotactics work like this. There's not really a clear phonetic definition of sonority, either, but nevertheless there are strong cross-linguistic tendencies governing syllable shape. The same is true for tone, metricality, and so on.

2. What exactly defines this phonotactically? It falls somewhere between normal vowels and normal consonants on the sonority hierarchy?

You're not going to get an exact definition. That's the punchline to everything I've been saying.

The category "rhotics" exists because of the confluence of two facts:
1) There are strong, cross-linguistic tendencies for how non-lateral liquids pattern, particularly with regard to syllable structure.
2) Many languages treat other, non-liquid sounds as either equivalent to or of the same class as [r]. These distributions are frequently explained by familiar historical patterns (like freknu mentions), but sometimes not.

In other words, a rhotic is "r-like". Some sounds are more likely to show up in that category than others (sometimes by wide, wide margins), but to find a categorical account of the class you ultimately need to consider languages individually.

Linguistics is pretty focused on universals, but personally I find it much more interesting when things are true 95% of the time. If something is true 95% of the time, it's neither arbitrary nor inevitable. The field is still struggling to figure out how to talk about that third way.

[Daniel]

Absolutely. In fact, I'd say that most typological generalizations about phonological interaction/phonotactics work like this. There's not really a clear phonetic definition of sonority, either, but nevertheless there are strong cross-linguistic tendencies governing syllable shape. The same is true for tone, metricality, and so on.

Hm.... so this suggests that rhotics are more similar to prosodic categories than phonemic ones. Interesting, no?

You're not going to get an exact definition. That's the punchline to everything I've been saying.

You're probably right. But why isn't that just hand-waving then? It's this mysterious category that we can sometimes observe but not define or explain. Few other phenomena exist quite like that, at least as far as I've seen.

In other words, a rhotic is "r-like". Some sounds are more likely to show up in that category than others (sometimes by wide, wide margins), but to find a categorical account of the class you ultimately need to consider languages individually.

So it's just de facto?
Is there any theoretical reason why would couldn't have other similar categories, say t-like sounds or l-like sounds? I realize we don't really have those, but based on what you've said there's no explanation for why we don't-- we just don't.

Linguistics is pretty focused on universals, but personally I find it much more interesting when things are true 95% of the time. If something is true 95% of the time, it's neither arbitrary nor inevitable. The field is still struggling to figure out how to talk about that third way.

I agree!



Regarding diachronic development:
1. Are "rhotics" then just, primarily, the remnants of more coherent categories? That would explain some of the weirder cases like vocalized final R in German, the <rr> of Brazilian Portuguese, and the Czech ř.
2. Why, freknu, do you assume that trills are most normal? From a European perspective (and therefore my default L2/typological perspective) that's true, but is there some broader reason?

[freknu]

2. Why, freknu, do you assume that trills are most normal? From a European perspective (and therefore my default L2/typological perspective) that's true, but is there some broader reason?

Yes, it's just an assumption. Looking at reconstructions of many languages, and not just Indo-European, the trill creeps up in quite many places — so it's just my assumption that the trill is the default rhotic.

On the other hand, I think I've read somewhere that the trill is one of the more difficult sounds for a child to learn, or am I thinking of another sound?

[Daniel]

I'm not sure if it's really the most difficult, but it is challenging. (Actually a lot of rhotic sounds are challenging!)

That doesn't seem unreasonable to me, but I wonder:
1. Whether the data really supports it; and
2. Why it seems reasonable to me. Cross-linguistic categorical perception?

[Trompette]

From what I understand, you're saying that rhoticity does not really describe a phonological category, more just left-overs. And you're saying it's more about how people speaking different languages perceive those sounds. So one phoneme could be rhotic for some people and not for others, for example because they speak different languages? My mother tongue is French and for me sounds like [ʀ, ʁ, x, ɣ, χ] and [r/ɾ] (is that tapped/trilled 'r' as in Spanish pero / perro?) all sound like 'r's. I think that, upon hearing those in isolation, most French people would agree (it's just a feeling really), although some of those clearly don't sound native. But for a native Dutch speaker, [ɣ] is a 'g' sound, isn't it? So not a rhotic?


Now, djr33, you say,

But what's really odd is that more or less we tend hear all of these sounds as /r/ via categorical perception. That's not always true, such as Brazilian Portuguese <rr> which is more like [h], but in general it works.

On one Wikipedia article I've read recently -- and which I can't seem to find now, I will edit my post if I do -- they said that in some area of England, in some cases 'r's were pronounced [ʋ]. I did not check the reference they gave for that, but would you second that assertion? And if it were true, would you say that that sound is then rhotic for those speakers of English, though it probably sounds more like a 'w' or 'v' sound (to me at least)?

Now, I'm just recalling something a Scotish woman once told me. Her 'r's were tapped alveolars and she said that at school, they had tried to teach her to say them the 'proper' way (is it a retroflex palatal approximant in British English?). She said she couldn't because the English 'r' sounded too much like an 'l'. Then for her, perhaps the English 'r' which is rhotic for some of you wouldn't be rhotic?

[Daniel]

And you're saying it's more about how people speaking different languages perceive those sounds. So one phoneme could be rhotic for some people and not for others, for example because they speak different languages?

Actually, I'd think that in general "rhotic" is a relatively cross-linguistic category so that speakers of most languages would have some intuition about this category. At least it seems like we have some sense of "oh that's an R" regardless of which language we speak. It fills a certain place in the consonant chart or something.

But then your examples are good ones: it does vary by language/dialect. I don't know what to do about that-- there's no longer even a vague "rhotic" category except orthographically. Very strange.

My mother tongue is French and for me sounds like [ʀ, ʁ, x, ɣ, χ] and [r/ɾ] (is that tapped/trilled 'r' as in Spanish pero / perro?) all sound like 'r's. I think that, upon hearing those in isolation, most French people would agree (it's just a feeling really), although some of those clearly don't sound native.

If you can manage to find a truly isolated native speaker, would they agree with you?

One possibility, which would challenge this idea of "rhotics", is that Rs are actually due to language contact-- we are told that people "say R funny" in certain languages, and then we start to map the sounds that way. So this could truly be a European phenomenon. I wonder if, for example, speakers of various R articulations in Asia or Africa also perceive "rhotics", without the influence of European orthographies. (Might be hard to test that these days.)

But for a native Dutch speaker, [ɣ] is a 'g' sound, isn't it? So not a rhotic?

Hmm... good point. Is it mutually exclusive? Could /g/ sometimes be pronounced as a rhotic without eliminating the rhoticity of the sound?
But that is a compelling argument.

they said that in some area of England, in some cases 'r's were pronounced [ʋ].

I don't know, but I'd like to hear more about this
I was wondering if there were any labial examples, and apparently there are!

[MalFet]

Absolutely. In fact, I'd say that most typological generalizations about phonological interaction/phonotactics work like this. There's not really a clear phonetic definition of sonority, either, but nevertheless there are strong cross-linguistic tendencies governing syllable shape. The same is true for tone, metricality, and so on.

Hm.... so this suggests that rhotics are more similar to prosodic categories than phonemic ones. Interesting, no?

You're not going to get an exact definition. That's the punchline to everything I've been saying.

You're probably right. But why isn't that just hand-waving then? It's this mysterious category that we can sometimes observe but not define or explain. Few other phenomena exist quite like that, at least as far as I've seen.

Virtually all phonological phenomena exist precisely like this.

Typological categories are virtually without fail permeable and generalized. If you consider that to be handwaving, so be it, but you're assuming a kind of abstracted realism that hasn't been tenable in the discipline for four or five decades now. These categories do not exist in some kind of platonic ideal realm. They are descriptive generalizations that attempt to convey broad patterns within a vast dataset that is necessarily more complicated.

In other words, a rhotic is "r-like". Some sounds are more likely to show up in that category than others (sometimes by wide, wide margins), but to find a categorical account of the class you ultimately need to consider languages individually.

So it's just de facto?
Is there any theoretical reason why would couldn't have other similar categories, say t-like sounds or l-like sounds? I realize we don't really have those, but based on what you've said there's no explanation for why we don't-- we just don't.

Based on what I said in what sense? There is a tremendous amount of very good research on exactly this question, and I'm not sure why anything I've said should suggest otherwise. In very brief summary: Approximants have some basic articulatory and perceptual characteristics that make them both flexible and salient. This opens the door wide to diachronic, sociolinguistic, and allophonic variation.

Being able to talk about these clusters is useful, so we have invented for ourselves the word "rhotic".

[Daniel]

Virtually all phonological phenomena exist precisely like this.

Typological categories are virtually without fail permeable and generalized. If you consider that to be handwaving, so be it, but you're assuming a kind of abstracted realism that hasn't been tenable in the discipline for four or five decades now. These categories do not exist in some kind of platonic ideal realm. They are descriptive generalizations that attempt to convey broad patterns within a vast dataset that is necessarily more complicated.

The borders are fuzzy. But there is no "natural class" that is anywhere near as fuzzy as rhotics, right?

So are you saying that's just a convenient example of how weird languages are in general? Or is there something still exceptional about rhotics?

If we did want to come up with some kind of universal deterministic system, I bet it would be easier without rhotics, even if still hard/impossible. At least it appears easier.

Are all classes so flexible? Is it conceivable that a language could have a T as a vowel?

Approximants have some basic articulatory and perceptual characteristics that make them both flexible and salient. This opens the door wide to diachronic, sociolinguistic, and allophonic variation.

Being able to talk about these clusters is useful, so we have invented for ourselves the word "rhotic".

Alright. Can you please analyze/contrast the terms "rhotic" and "approximant" for me? Are they equally arbitrary and (non-)deterministic? Or is there something more "real" about the approximant category compared to the (artificial) "rhotics"?

...so we have invented for ourselves the word "rhotic".

So is this entirely due to the analysis? Is "rhotic" not something that is really part of language, but just something linguists have interpreted for convenience?
(I'd argue that "vowel" seems to be a pretty reasonable part of languages inherently, even if a bit fuzzy; it's not just due to to he analyst.)

[MalFet]

Could a voiceless obstruent (like "t") function as a vowel? Hypothetically, sure. For a number of well-understood reasons, however, it's not very likely.

I'm having a very hard time following your interpretations of my statements (i.e., no, this is definitely not just a convenient example of how weird languages are in general), and likewise I'm at a bit of a loss for how to proceed. I suspect that we are operating under some very different understandings about what phonology is.

Specifically, the idea of a "universal deterministic system" is deeply antithetical to virtually all work done in phonology over the last century. That's an extremely important point. In terms of pure phonology, categories construct a closed system. There is no equivalence across languages. Asking what an English /r/ has in common with a Hindi /r/ is not meaningful except as shorthand for a very different question about perceptual typology. If we're talking about perceptual typology, of course the category is not deterministic. How on earth could it be?

What the term "rhotics" describes is a cross-linguistic pattern in the relationship between a cluster of phonetic natural classes (roughly, non-lateral approximants) and a bunch of language-specific phonological categories. What's interesting about rhotics is that in many languages the phonological category extends (sometimes significantly) beyond the boundaries of the phonetic natural class. That fact of extension is itself motivated by the articulatory and perceptual properties of approximants. The typological generalization is that non-lateral approximants and other sounds treated as equivalent to them by speakers of a particular language frequently have distinctive distributions within syllables.

What's "fuzzy", then, is not the border in any sense, but rather the degree of isomorphism between language-specific phonological categories, on the one hand, and natural phonetic classes, on the other. In those terms, hunting for the exact border that separates a rhotic from a non-rhotic is missing the point entirely. It's like arguing that the term "vowel" is just handwaving because we don't know for sure whether or not "l" counts as one. That's just not how it works.

In that regard, rhotics are pretty middle-of-the-pack conventional. They're more typologically irregular than some things, but less typologically irregular than others. Far from being unusual, this is what the bulk of phonology looks like once you get beyond the trivial stuff that shows up in problem sets. The notion of rhotics is no more and no less coherent than five dozen other different words we use in typological comparisons: "tone", "sonorance", "vowel", "syllable", "stress", etc. These things are all generalizations with significantly different organizations in different languages.