Palatal Stops

[Hallow XIII]

I was noticing this when I was looking over Irish phonology, but it really applies to pretty much everyone: something is wrong in the way the palatal stops are parsed. [c ɟ] are really two different sounds each, once articulated with the tongue touching the alveolar ridge and once without. Irish has both varieties, with the phonemic analysis being /tʲ dʲ/ and/c ɟ/respectively, but in phonetic notation both of these sounds seem to be represented by [c ɟ], at least in other languages (case example: Albanian /c ɟ/ is pronounced exactly like Irish /tʲ dʲ/). I don't know if there is some sort of diacritic I've missed (if yes, it's buried very, very, deep), but it seems to me there's a serious deficiency here, because in a phonology, unless the language contrasts these two and one has to be parsed differently, you don't know which is meant.

This must be rectified.

[Mednij]

I always thought that they were the same thing.

[Nortaneous]

Yanyuwa contrasts both kinds of palatal.

They really should add ȶ ȡ ȵ ȴ to IPA.

[Click]

They really should add ȶ ȡ ȵ ȴ to IPA.

Yes, they should.

[Rory]

because in a phonology, unless the language contrasts these two and one has to be parsed differently, you don't know which is meant.

In a phonology it doesn't matter which one is meant, because they're not contrastive. Looking at a phonological description of German, I can't tell if the /S/-sound is more like the English postalveolar or the Mandarin retroflex, or like neither. And that's fine, since it shouldn't matter (at least, it shouldn't if we're adhering to a modular structuralist-generativist substance-free phonology).

For a decent phonetic description though, yes, it's absolutely vital, although palatal stops are by no means outliers. Many aspects of the segmental description of a language are not phonetically clear - is there lip protrusion on the rounded vowels, or just lip rounding? What is the tongue posture for the sibilants? How long are the trills? Is the allophonic nasalization gradient or categorical? And what exactly is meant by "stress on the penultimate syllable"?

Languages are complicated

[Hallow XIII]

highly intellectual post

The point is, assuming I don't have the time to read a detailed phonological description, I'd still like to be able to look at an IPA transcription of something and be able to pronounce it in such a manner that native speakers would not like at me like I was some sort of mentally retarded Martian. I'm sure there are some poor bastards out there who, never having been to Albania, pronounce kuq [kuc] instead of [kuȶ] or, having read the Wikipedia article on Irish, waltz around saying ['ge:lʲȡə] until someone informs them of their mistakes.

[clawgrip]

highly intellectual post

The point is, assuming I don't have the time to read a detailed phonological description, I'd still like to be able to look at an IPA transcription of something and be able to pronounce it in such a manner that native speakers would not like at me like I was some sort of mentally retarded Martian. I'm sure there are some poor bastards out there who, never having been to Albania, pronounce kuq [kuc] instead of [kuȶ] or, having read the Wikipedia article on Irish, waltz around saying ['ge:lʲȡə] until someone informs them of their mistakes.

I would guess the average language learner would not be able to make that distinction even if they tried. People regularly fail to replace similar sounds like this. For example, Japanese speakers regularly fail to replace Japanese [ɕ] with English [ʃ] when Speaking English, but I wouldn't call them poor bastards because of it, and it doesn't hinder communication.

I also doubt there is a significant number of people learning Albanian who have never been to Albania yet are able to distinguish [c] and [ȶ].

[Rory]

highly intellectual post

The point is, assuming I don't have the time to read a detailed phonological description, I'd still like to be able to look at an IPA transcription of something and be able to pronounce it in such a manner that native speakers would not like at me like I was some sort of mentally retarded Martian. I'm sure there are some poor bastards out there who, never having been to Albania, pronounce kuq [kuc] instead of [kuȶ] or, having read the Wikipedia article on Irish, waltz around saying ['ge:lʲȡə] until someone informs them of their mistakes.

This is why we learn languages from teachers and not from Wikipedia. It's simply impossible to get a truly native pronunciation from an IPA transcription; that's not even what the IPA was designed for in the first place.

[gach]

Different notations do things differently. The mark for palatal articulation in UPA is the acute accent. This means that you can very easily distinguish between alveo-palatals and prevelars by sticking the palatal mark over alveolar and velar symbols. The down side is that in principle the palatal mark doesn't make a distinction between palatalised and purely palatal consonants. So without further specification the UPA ń can mean either /ɲ/ or /nʲ/ in IPA. If you want or have to contrast the two, you have to resort to special supplementary marks. Some old works that aim for exact phonetic transcription can be a pain to read because it's hard to dig the basic letter forms beneath all the diacritical marks.

[Xephyr]

This is why we learn languages from teachers and not from Wikipedia. It's simply impossible to get a truly native pronunciation from an IPA transcription; that's not even what the IPA was designed for in the first place.

Not everybody has access to native speaker teachers. To those people (e.g. me), shit like the IPA is a godsend, and in cases when it does not adequately do its job (such as when describing the numerous kinds of ""palatal stops"", an area which the IPA has long been observed to be very, very unsatisfactory), then it ought to be revised.

[Nortaneous]

IPA sucks for most things except total standardization and obscure sounds that APN can't write well. What else is there that IPA is better than either APN or canIPA at?

[Xephyr]

True, but those alternatives are still much preferable to "just ask your teacher".

[Nortaneous]

True, but that doesn't mean it's not obvious from looking at the thing that the IPA was designed by a committee of old French bureaucrats.

[Hallow XIII]

True, but that doesn't mean it's not obvious from looking at the thing that the IPA was designed by a committee of old French bureaucrats.

Attempts at standardization of things tend to be hopelessly unsatisfactory, but at least the IPA is a better example that other such attempts that shall not here be named. Although the fact that the IPA cardinal vowel points are modelled on those of French does seem a bit presumptuous to me.

Also, what the heck is an APN? American Phonetic Notation?

[Drydic]

Yep.

It's not as standardized as the IPA, obviously, but the majority of it doesn't vary too much. Vowels are the exception to that generalization. Blah blah all the points in the schwa etc thread, yes, but having a consistent set of vowel symbols is extremely useful (this is part of why I hate "standardized" symbols for phonemes, at that point you might as well just stop writing a symbol and go with the the BATH vowel/the NURSE vowel system).

[Radius Solis]

The IPA is trying to meet more than one goal at the same time, and sometimes this creates a conflict of interest. First, it is put forward as a phonetic alphabet, not a phonemic one, being a system for transcribing phones according to their articulation; but then, to keep it to a manageable size and keep it appropriate for what most people want it for, it also tries to limit this to phones that are phonologically useful to transcribe, namely those that contrast in at least some human languages.

This is a good thing. A system that fully met one of these goals at the expense of the other would be a weaker tool than the IPA is. But because those goals are not completely compatible, it comes at the price of having to put up with issues like this one with the palatals.

The mismatch between phonetics and phonology in the palatal region is not an artifact of the transcription system, it is a real behavior of human languages. [t] and [k] do not have nearly so much of an issue: languages strongly tend to have a /t/ and a /k/, and there is a quite strong tendency for certain very specific articulations to fill these roles. If you imagine POA as a matter of points along a front-back spectrum, languages cluster tightly at certain points for where they put their /t/ and their /k/. But that's not nearly so true of languages with a /c/ - it is quite common for languages to have POA role for "between /t/ and /k/", but there is a lot less unity between languages as to what phones they fill this role with. There is no single pretty cluster at some mythical point point on the spectrum we could call a "true palatal stop", but instead five or six smaller clusters... and the most common phones to fill that role are [tS)] and [ts)], which, sadly, are the ones least describable as "palatal stops" in articulatory terms.

So, trying to be both articulatorily accurate and phonologically useful at the same time gets us into trouble. You guys suggesting the adoption of new symbols are, ultimately, quibbling over exactly where the balance between the two goals should lie. And it is quibbling. It's futile, in the big picture, because for every purpose for which you reduce the need to patch up your palatal transcriptions with English description, you introduce that need for some other purpose.

[Qwynegold]

They really should add ȶ ȡ ȵ ȴ to IPA.

The alveolo-palatals? They removed those from IPA and they're now transcribed t̠ʲ d̠ʲ n̠ʲ l̠ʲ IIRC.

[Hallow XIII]

They really should add ȶ ȡ ȵ ȴ to IPA.

The alveolo-palatals? They removed those from IPA and they're now transcribed t̠ʲ d̠ʲ n̠ʲ l̠ʲ IIRC.

Which is a bullshit, as everybody leaves out the lower diacritic.

[Travis B.]

I just love it when someone decides to arbitrarily downgrade something like this for no good reason...

[Rory]

They really should add ȶ ȡ ȵ ȴ to IPA.

The alveolo-palatals? They removed those from IPA and they're now transcribed t̠ʲ d̠ʲ n̠ʲ l̠ʲ IIRC.

No, they were never in the IPA to begin with. I've never seen them used outside of Chinese dialectology (although my experience is limited).

[Nortaneous]

So, trying to be both articulatorily accurate and phonologically useful at the same time gets us into trouble.

But IPA's design goals are pretty clear: minimize diacritics by creating two separate symbols for every pair of sounds that a language contrasts -- except, in practice, affricates, and the dental/alveolar contrast, where it would be a pain in the ass to create that many new symbols even though there are plenty of languages that contrast them. Distinguishing coronal and dorsal palatals doesn't go against those design goals at all; they contrast in Yanyuwa. Besides, when using IPA for phonetic transcription, there's no good way to specify which: the symbols for palatals are ambiguous (some idiotspeople even use them for the English postalveolar affricates!) and the diacritics that would be used for distinguishing them are hard as hell to make out in inline transcription font size.

It's not like the distinction between apical and subapical consonants, where the already unimportant phonetic detail is generally clear from the language you're talking about... but even then, they contrast in Toda!

[Radius Solis]

(some idiotspeople even use them for the English postalveolar affricates!)

Hardly idiotic in phonemic transcription, considering that /p t c k/ is our stop system. Disagreement about the /c/ part boils down to tradition, and squeamishness about phoneme symbols not matching their associated phones, neither of which change the basic facts about our language.

[Nortaneous]

Hardly idiotic in phonemic transcription, considering that /p t c k/ is our stop system.

Have fun writing rules for realization of the plural morpheme with that analysis.

[kodé]

Hardly idiotic in phonemic transcription, considering that /p t c k/ is our stop system.

Have fun writing rules for realization of the plural morpheme with that analysis.

Easy. /c/ = [+strid], /p t k/ = [-strid]
/z/ -> [Iz] / [+strid]_

[Radius Solis]

Quite so; I did not mean to imply any discarding of /c/'s sibilance.