Talk:International Phonetic Alphabet
|Discussions on this page often lead to previous arguments being restated. Please read recent comments and look in the archives before commenting, and read through the list of highlighted discussions below before starting a new one:
|International Phonetic Alphabet has been listed as one of the Language and literature good articles under the good article criteria. If you can improve it further, please do so. If it no longer meets these criteria, you can reassess it.|
|International Phonetic Alphabet has been listed as a level-4 vital article in Language. If you can improve it, please do. This article has been rated as GA-Class.|
|This article is of interest to the following WikiProjects:|
|This article has been mentioned by a media organization:|
The link to "current IPA chart" is broken. 126.96.36.199 (talk) 18:18, 24 February 2015 (UTC) As of 07 December 2015, this link is *still* broken. — Preceding unsigned comment added by ORD89 (talk • contribs) 19:45, 7 December 2015 (UTC)
Criticisms and/or Limitations
Hi. I know very little about IPA but I have a question about this article. The claim is made that IPA is "a standardized representation of the sounds of oral language." (see lede). Well, first does this mean ALL of the sounds? If not, then the sentence is misleading, if so, then I think it overreaches. You can't represent all of the sounds of many "oral languages" with a finite alphabet, can you? Specifically, any two 'similar' sounds having a different representation will have a variable number of possible intermediate sounds between them, which by definition IPA can't capture. This may or may not be relevant - IF it has been shown that IPA is both necessary and complete for the pronunciation for ALL of a speaker's words and sentences. But I doubt this has been done for all of the world's languages. This is objection/question one. Problem two is the well known (if I know it, then it must be well known) difference between a word in isolation and a word in a context. Most words vary in their sound depending on context (leading and following phonemes? blend with the phoneme). This means that there is NOT a unique pronunciation of ANY word. Putting it another way: how well does IPA cover oral languages? Next issue: does IPA extend to any non-Homo language? Its obvious, that it can't cover ANY language with audio signals outside of human hearing, but we have the problem of the difference between vocalizations and hearing, I'm not sure the former is a subset of the latter or do both sets have disjoint elements? (ie. there are vocalizations which can't be heard and, more obviously, sounds which can't be vocalized). If it is true that some languages include sounds which are NOT represented by the IPA, then this should be noted up-front. Same with IPA+Extension. Anyway, I am under the impression that the electronically recorded signal of oral language is the only accurate "representation" of what was said. IPA is clearly NOT that. (Nor is it the perceived sound, but that is impossible (afaik) to determine.) My point is that this article lacks a discussion of the qualifications and limitations of the IPA, which are surely known to some extent (my guess is to a pretty good extent). Other questions involve its use in non-IndoEuropean cultures. Do Chinese speech pathologists use it in their publications, for example? Is "attack" part of IPA "intonation"? How well are duration, vocal fry, nasal, and other influences captured? Its too bad that an expert isn't available for editing, but surely the science/engineering design of electronic voice synthesis MUST have a great deal to say about this, and surely a lot of that has been written and is accessible.188.8.131.52 (talk) 22:13, 24 September 2015 (UTC)
- Hello 216! Let me try to answer your questions one by one:
- All sounds
- Yes, that's what it means in the sense of all sounds that make sense to distinguish. There are in fact diacritics that can be used to distinguish intermediate sounds, in the few cases when that is necessary. That said, there are some distinctions that are not standardized yet; this typically happens when a linguist researches a language and finds that a distinction is meaningful in a language that has not found to be meaningful before. That's why I wouldn't write "all" in that sentence. But, yes, it is possible to represent all sounds of oral languages with a finite alphabet.
- Yes, pronunciation of words varies depending on context. That can be covered by IPA without a problem, but it doesn't have to be, and is not necessary for most standard situations. Whether to do that or not is not part of the standard, any more than it is whether you write A♯ or B♭ in standard musical notation.
- Non-human languages
- IPA is not intended for those.
- Non-European languages
- If you read the article, you can see that those are covered with IPA. While IPA was not devised for speech pathologists, I see no reason to assume that IPA can't be used by Chinese speech pathologists.
- Electronically recorded
- No recording, whether electronic or analog, is completely accurate. Neither is any notation system, either. That holds for language just as for music. It is a general limitation that has nothing to do with IPA in particular.
- How well are ... captured
- Well enough for the purposes listed in the article. Also, please note that "representation" does not necessarily mean "capturing".
- Voice synthesis aspects
- IPA does not claim to be a complete system to contain all information needed for voice synthesis.
- Not sure what you mean by "attack". Are you referring to the term used for ADSR envelope? Would you have an example for when that is a relevant distinction for spoken language that is not already inherent in the distinction between initial consonants?
- Your main point seems to be that you miss coverage of limitations in the article. Your examples cover limitations that are par for the course; they remind me of someone demanding that a motor vehicle article should cover the fact that a car can't (normally) fly. If you have any specific sourced information on limitations (especially for the "how well are ... captured" questions), you are cordially invited to add that. I'll be happy to help you with that. — Sebastian 17:36, 28 September 2015 (UTC)
- I think Sebastian answers the question fairly well, but I'd qualify the answer somewhat. It is true that the IPA uses a finite set of symbols to represent points on an infinite space. For instance, there are more possible points between a high-front vowel and a high-mid front vowel than can be represented using the tools of the IPA. But this doesn't matter, because you only need enough symbols to distinguish the sounds that the world's languages distinguish. There's no language, for example, that distinguishes ten different vowels between [i] and [e], so the IPA doesn't need a means of making that many distinctions. Of course this means that the same symbol may represent slightly different sounds in different contexts. The symbol [r] (even if in square brackets) should not be taken to represent precisely the same sound every time it's used. As Sebastian says, if you need to make a finer grained comparison of two given sounds, you have diacritics to help you out. In any case, I think what the anonymous poster is pointing out is that no IPA representation of speech, however detailed, could provide enough information to allow the original speech to be played off. But that's not what's meant by "a standardized representation of the sounds of oral language." What's meant is that you have all the tools you need to distinguish the sounds of the world's languages, and that's pretty accurate. Garik (talk) 02:11, 29 September 2015 (UTC)
- But also, the OP's reference to audio recordings completely misses the point. A transcription of a language is there to capture generalisations: so when a young girl says [ˈzdrastvʊjtʲə] (or /bɔ̃.ʒuʁ/) in a high shrill voice, speakers of Russian (or French) recognise this as "the same as" a gruff male growling [ˈzdrastvʊjtʲə] (or /bɔ̃.ʒuʁ/), even though the audio recordings would be utterly different. Imaginatorium (talk) 13:13, 29 September 2015 (UTC)
Thanks for your helpful and thoughtful responses! Am I wrong in thinking that in order to match an audio recording of a single "word" to a specific IPA expression that all three of the following requirements must be met?
- 1. The IPA representation must explicitly or implicitly be tagged with the language it is encoding.
- 2. The listener/reader must be proficient in speaking that language.
- 3. The listener/reader must be proficient in using IPA specifically FOR THAT LANGUAGE.
- 1. The IPA representation must explicitly or implicitly be tagged with the language it is encoding.
I've seen several sources that make the claim that a given IPA symbol string is generally NOT unique unless the language is specified. I've also seen several sources which claim that there are nuances which are NOT captured by IPA and which require a proficient speaker of the langugage to decode from the IPA symbol string into a word/sound. I think that I'm not alone coming to this article thinking that IPA is an unambiguous way to encode the oral languages of the world, and it is not. If I understand correctly, two IPA experts may disagree when given the same symbol string. (I assume they could also encode the "same" word sound with significantly different strings). If I'm correct, then I think the lede should state that. Specifically, that IPA is language dependent and it may not fully encode a given oral word. It may or may not be helpful to point out that Wikipedia (en) points to the English IPA table (rather than "THE" IPA table). Oh, it may also be worthwhile to mention that no general IPA to sound algorithm exists. (Which is what I came looking for, originally.) Thanks again!184.108.40.206 (talk) 18:49, 7 November 2015 (UTC)
- "The IPA" itself is completely untied to languages, so it makes no sense to say it's language-dependent. It can sometimes be used in language-specific fashions, such as when using it phonemically (and not phonetically). That the IPA may not fully encode a given oral word is not the same thing as it being language dependent. It can't fully encode a given oral word simply because it's not a recording: only a perfect sound recording can achieve that. The IPA only provides an abstract representation of sound. An IPA-to-sound algorithm could in theory certainly exist, and in fact, there are certain speech synthesizers that come close. It would not, however, sound exactly like a native speaker of any language.
- And to address your three points explicitly:
- No, in a phonetic IPA representation, there is no need to explicitly or implicitly specify a language; it could even be gibberish not in any human language, and yet it could be spoken from the IPA representation. This is not the case with a phonemic representation, though, which a language does need to be associated with.
- No, someone can read a phonetic IPA string if they are proficient in IPA, although of course, they may not be able to read it as quickly and fluently as someone who knows that string as a word in a human language.
- No, same as above.
- LjL (talk) 19:01, 7 November 2015 (UTC)
- We all started out as beginners. That's no argument. It's not harder than learning a new language with different spelling rules. --220.127.116.11 (talk) 00:51, 23 July 2016 (UTC)
Mid-line tilde not accounted for
Nowhere in this article (nor at Tilde#International Phonetic Alphabet) do we account for use of the mid-line, non-diacritic tilde, Unicode 'tilde' or 'spacing tilde' U+007E,
~ (not to be confused with: the mathematical "similar to" symbol a.k.a. 'tilde operator' U+223C
∼; the diacritic tilde a.k.a. 'combining tilde' a.k.a. 'non-spacing tilde' U+0303,
̃; or the 'wave dash' U+301C,
〜 used in Japanese). This article uses this markup at least once, and it's used many times without explanation at International Phonetic Alphabet chart for English dialects (where it is not even listed in the "Other symbols used in transcription of English pronunciation" table) and in dialect articles like California English.
It is most often used in a way that appears to indicate an intergrading range, e.g.
ɑ~ɒ, but may be intended to represent a choice between two exclusive variations rather than a range, or perhaps some specific blending of the two. Sometimes, however, it's used between more than two sounds, as in
ɛʊ̯~œʊ̯~œʉ̯~œɤ̯̈~œː~ʌʊ̯, an extreme example from International Phonetic Alphabet chart for English dialects, so a specific intergrade cannot seem to be the [usual?] meaning.
I'm seeing it used inconsistently, with no delimiters, with /.../ delimiters, and with [...] delimiters.
Many of these articles are actually using the U+223C mathematical operator
∼, when they should (surely?) be using the non-maths version, U+007E
Disparity between wikipedia consonant chart and official chart
The official IPA chart is here: https://www.internationalphoneticassociation.org/content/full-ipa-chart . The official consonant chart is vastly different to the one presented in this Wikipedia article. What is the justification for having such a disparagingly different chart to that of the official one? — Preceding unsigned comment added by Blimpblam (talk • contribs) 23:54, 16 January 2016 (UTC)
New revision (2015)
https://www.internationalphoneticassociation.org/sites/default/files/IPA_Kiel_2015.pdf --Petreleon (talk) 12:51, 16 March 2016 (UTC)
- @Petreleon: Read this: "The 2015 chart makes minor changes to wording and layout, but otherwise reproduces the appearance of the 2005 chart." Love —LiliCharlie (talk) 14:24, 16 March 2016 (UTC)
On including IPA charts that are not the IPA chart
We've long included a chart that is the concoction of one single WP editor -- a clear NOR violation, in my opinion. As of 2015, the IPA chart is distributed on a public licence in fully accurate form. So really, now is the time for us to go with just that single chart, which definitionally is the IPA. (N.B.: we include extensive discussion of alternative symbols elsewhere on WP.) I hope the editor who created the concocted chart will understand. BrucePHayes (talk) 20:23, 16 March 2016 (UTC)
IPA for beginners?
How do I make a pronunciation guide with the IPA? No, I'm serious. I can't figure out some of the sounds, the sounds on the website are annoying, and they are confusing because it's never just one sound. Is there a way to make this page easy to understand for people who have very little clue what the IPA means or how to use it? NewbTopolis Rex (talk) 01:19, 29 March 2016 (UTC)
- I'm not sure what you want to achieve. Maybe you can consult Help:IPA for English first, then move on to Help:IPA. If you want to learn IPA more thoroughly you won't be able to do so by imitation alone, but will have to learn to control the movements of your speech organs (see Articulatory phonetics and Some low-tech means of observing your articulators). Ian Catford's A Practical Introduction to Phonetics is a very helpful reader with "experiments" for self study, though I recommend the help of a teacher. — As to your proposal to change this page please read WP:NOTHOWTO. Love —LiliCharlie (talk) 02:14, 29 March 2016 (UTC)
Hello fellow Wikipedians,
I have just modified one external link on International Phonetic Alphabet. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
- Corrected formatting/usage for http://www.langsci.ucl.ac.uk/ipa/ipachart.html
When you have finished reviewing my changes, please set the checked parameter below to true or failed to let others know (documentation at
Archived sources still need to be checked