laterality, sibilancy, and rhotacity: can we combine the three features?

[LinguList]

We have laterality for lateral affricates and all laterals. Since laterality is a "base" feature that defines a base symbol, like "l", as is "sibilancy" (e.g., "s"), I wonder if we can treat the "rhotic affricates" in the same way. If we accept them (which makes sense to me), this would mean that we have one more base feature that is not listed as the official base, since it is often undefined. On the other hand, there is no clash between rhotacity and laterality and sibilancy, right? So could we not say that we an additional feature, maybe even just calling it "airstream" and giving it three values, "lateral", "sibilant", and "rhotic"?

I am suggesting this, because it would help us to handle these three features at once, while by now, we have quite some trouble, also in ordering the data.

[LinguList]

One important question here is, however, how we treat the numerous flaps, trills, and approximants: do we consider part of them as "rhotic"? Or do we consider this as language-specific? I am also asking, because one finds numerous combinations of affricates with "rhotic" sounds in phoible and the like, where it is often not clear why linguists would choose exactly those sounds: was it to mark an affricate as some sound unit, or rather really to point to a cluster that has distinct articulation for the stop and the liquid element?

[cormacanderson]

I don't see any obvious objection to this from a practical perspective. The categories of sibilant, lateral, and rhotic are mutually exclusive as far as I can see. As for the rhotic affricates, I think we have to distinguish different cases. There are stops with trill release, discussed here https://github.com/cldf-clts/clts/issues/45, and there are non-sibilant affricates in which the second part is a fricative that is represented using a rhotic symbol, for which a different representation might be possible, e.g. https://en.wikipedia.org/wiki/Voiceless_postalveolar_affricate#Voiceless_postalveolar_non-sibilant_affricate, cf. https://github.com/cldf-clts/clts/issues/51 for fricatives.

[LinguList]

Okay with trill release is easy to handle as a feature and to hard-code into consonants.tsv.

If we think of cases like "tr", which we find in phoible, I'd then write them as "t+superscript-r" and call them "voiceless alveolar rhotic affricate consonant", right? The alternative would be to do it similar to sibilant affricates and model the articulation on both elements, that is: r should have the devoicing marker on top, right?

[cormacanderson]

I'd say yes to modelling the articulation on both elements. These are parallel to /kx/ "voiceless velar affricate", to /cç/ "voiceless palatal affricate", also /pɸ/, /pf/, /tθ/, /qχ/ etc. so we would have:

• [ ] /tθ̠/ "voiceless alveolar affricate"
• [ ] /tɹ̝̊/ "voiceless post-alveolar affricate"
• [ ] /ʈɻ̝̊/ "voiceless retroflex affricate"

As well as the voiced versions which I can write out if you want. We have been discussing the representation of the fricative portion of these here https://github.com/cldf-clts/clts/issues/51. As you can see, there are a lot of diacritics on the post-alveolar (which should also have a retracted diacritic that I didn't write) and the retroflex. This is why Maddieson's policy of writing /ʈθ̣/ for the retroflex has some merit. I don't have a good solution for the post alveolar. Maybe @tresoldi has some idea?

[tresoldi]

There are times I want to do a PR to Unicode and the IPA...

As for a solution for the post-alveolar, you mean for the post-alveolar rhotic affricate, right? d̠ɹ̠˔?

[cormacanderson]

Yes, that's the one.

[LinguList]

Okay, @tresoldi, can you please help me on this, and spell out, tab-separated, as I need it to include it into consonants.tsv, all major sounds here? I'd then directly introduce those to the PR I am about to make and also map in phoible. If you give me the name only, I'd have to shuffle things, but I am about to work on pyclts, to get this right as well, by merging lateral and sibilant into "airstream". I'd then not add any "rhotic" airstream, unless you guys give me clear examples, but rather treat the case of "tr" as t with trilled release, okay?

[tresoldi]

I was looking again to the list of combining diacritical marks in Unicode, and nothing really satisfies me -- in all cases we would proposing something a bit too new.

I would suggest to stick with d̠ɹ̠˔. If you really, really want an alternative, in that model of mine that I had started, where I had no plans to follow IPA if it did not suit me, I had the fricative as θ˞ (U+03B8U+02DE), but mostly because it was easier for me to type it...

[tresoldi]

@LinguList , I can try, I assume you mean "a list of all affricates, with name and grapheme". I am sorry but I am still finding some difficulty in many discussions: while I can follow your overall reasoning on how to encode the features, I am unsure about what are the criteria for deciding on something as a feature, a value, or just exclude it, particularly when clusters are involved. Having laterality, sibilancy and rhoticity as value of an "airstream" feature is an example, because -- unless I misunderstood -- feature values are supposed to be exclusive, so you could not have both. On the matter, we might decide on a different name, as "airstream" is usually taken as "airstream mechanism".

[LinguList]

It is a lateral airstream as opposed to a normal airstream that makes laterals lateral, as far as I understand. Please provide alternatives, or we can stop discussion. It does not help to just dislike something, we need to be constructive.

Note that I am changing a lot of files now at the same time, it does not help me if we go back and forth with these decisions, so we need to decide this, and so far, laterality and sibilancy do not occur together.

And if it is not clear what I mean, I'll then deal with it myself.

[cormacanderson]

The cases of /tr/ in PHOIBLE have trilled release. I'm not keen on θ˞ as it seems to me that the rhoticity of these is secondary or sometimes absent, rhoticity anyway being a very slippery thing. Look at the alternative representations here https://en.wikipedia.org/wiki/Voiceless_alveolar_fricative#Voiceless_alveolar_non-sibilant_fricative, where the "rhotic" and the fricative are considered synonymous. This particular one is all over my English (most common non-initial allophone of /t/) and there's really nothing rhotic about it. The primary consideration would be to deal with the non-sibilant fricatives, and by extension the non-sibilant affricates. The default representation here is really messy, because we need both the raising diacritic and the retracted one for the post-alveolar. I don't see an alternative though, so ɹ̠˔ probably best.

[tresoldi]

Ok, for all other matters, just please provide me an example with column names and one or two rows of what is needed. I will do my best to do it and refrain from any theoretical discussion.

[LinguList]

What I want is very simple, and I'd also ask @cormacanderson to provide data in the future in exactly this form, as it is easier if you do it directly. /1

[LinguList]

If you want to provide a direct sound to CLTS, you need to determine its sound class. For example, if it is a consonant, you need to modify the consonants.tsv table. /2

[LinguList]

This table has a header of the following form:

GRAPHEME	PHONATION	PLACE	MANNER	ALIAS	EXTRA	NOTE
pt	voiceless	bilabial-and-alveolar	stop

/3

[LinguList]

If you provide additional articulations, these are added to the field EXTRA, in the form feature:value. Consider how we treat the l:

GRAPHEME	PHONATION	PLACE	MANNER	ALIAS	EXTRA	NOTE
l	voiced	alveolar	approximant		laterality:lateral

/4

[LinguList]

So if I ask you to provide me with the data in this form, I ask you to provide the data in tab-separated form in such a way that I only need to insert it into that file.

@cormacanderson's

/tθ̠/ "voiceless alveolar affricate"

Then becomes:

tθ̠ voiceless   alveolar    affricate           

/5

[LinguList]

Then I can copy-paste it. But beware: if you give me an affricate fricative, you need to also add the "extra" features, and this is something you have not done so far, you just pasted me the "with_friction" or similar, and I have to figure out what feature it is, because I am only given a feature value.

What you have to understand is that our table is constructed like this: we provide BASE features phonation, place, manner, and all other features using the feature:value construct in the column EXTRA. The reason is that we would have too many features otherwise, and could not flexibly modify them. But it means, when adding features, one needs to think of both the feature value and the feature name itself, and this is what makes it difficult so far, since I receive often requests to add things that I have to translate to this system. /6

[LinguList]

So (and this is my last remark): I was just asking for you, @tresoldi, or @cormacanderson, to help me a bit by providing the new sounds in this format, which you can paste as code in github, by putting it into the `` marks. /end

[tresoldi]

I think these are the sounds. No special diacritic on the plosive for alveolars and post-alveolars, as I undestood. @cormacanderson , could you confirm?

GRAPHEME    PHONATION   PLACE   MANNER  ALIAS   EXTRA   NOTE
tθ̠ voiceless   alveolar    affricate           
tɹ̠̊˔   voiceless   post-alveolar   affricate       
ʈɻ̝̊    voiceless   retroflex   affricate       
dð̠ voiced  alveolar    affricate
dɹ̠˔    voiced  post-alveolar   affricate
ɖɻ̝ voiced  retroflex   affricate           

[cormacanderson]

@tresoldi yes, I think so. This is how we deal with other affricates, e.g. ts̪.

We should also take the opportunity here to add the corresponding fricatives, so if you could also add these please @LinguList

GRAPHEME    PHONATION   PLACE   MANNER  ALIAS   EXTRA   NOTE
θ̠  voiceless   alveolar    fricative   θ͇      
ɹ̠̊˔    voiceless   post-alveolar   fricative       
ɻ̝̊ voiceless   retroflex   fricative       
ð̠  voiced  alveolar    fricative           
ð͇  voiced  alveolar    fricative   +       
ɹ̠˔ voiced  post-alveolar   fricative
ɻ̝  voiced  retroflex   fricative           

This will also resolve https://github.com/cldf-clts/clts/issues/51.

[LinguList]

Nice, this improves the mapping more. Question, @cormacanderson, what do we do with tɾ? Is this still a trilled release? Or should we add a "tapped release"? But then, I don't know the symbol.

[cormacanderson]

I have amended the table above to add frequent aliases for the alveolar fricatives, as these are also used in the literature and are not otherwise likely to cause confusion.

[LinguList]

Alias syntax is different: you write an extra line with the same sound, copying it, but you add a + in the ALIAS field.

[cormacanderson]

For tɾ in PHOIBLE if I remember right one case was trilled release (Arigna) and I think I questioned the data with Mbembe.

[cormacanderson]

Aye aye captain

GRAPHEME    PHONATION   PLACE   MANNER  ALIAS   EXTRA   NOTE
θ̠  voiceless   alveolar    fricative   +
θ͇  voiceless   alveolar    fricative       
ɹ̠̊˔    voiceless   post-alveolar   fricative       
ɻ̝̊ voiceless   retroflex   fricative       
ð̠  voiced  alveolar    fricative   +
ð͇  voiced  alveolar    fricative
ɹ̠˔ voiced  post-alveolar   fricative
ɻ̝  voiced  retroflex   fricative           

Like this?

[LinguList]

And ɖɽ is our alveolar affricate, right?

[cormacanderson]

That would be a voiced retroflex with trilled release

[LinguList]

voiced retroflex stop with trilled release?

[LinguList]

ɖʳ

[cormacanderson]

Exactly. Also two aliases in the affricates

GRAPHEME    PHONATION   PLACE   MANNER  ALIAS   EXTRA   NOTE
tθ̠ voiceless   alveolar    affricate   +
tθ͇ voiceless   alveolar    affricate       
tɹ̠̊˔   voiceless   post-alveolar   affricate       
ʈɻ̝̊    voiceless   retroflex   affricate       
dð̠ voiced  alveolar    affricate   +
dð͇ voiced  alveolar    affricate
dɹ̠˔    voiced  post-alveolar   affricate
ɖɻ̝ voiced  retroflex   affricate