Search Ace Linguist

May 28, 2024

Good linguists trust their ears - IPA isn't perfect!

The IPA is a pretty good invention, allowing us to transcribe languages with as much precision as we feel is necessary. It's especially useful for language learners, as each IPA character's name is basically an instruction on how to pronounce it. However, you should be careful when it comes to the IPA, because a transcription is not always reality!

To start with a trivial example, consider that the General American English rhotic sound is represented as /r/ in most transcriptions, even though it is not an alveolar trill in any variant. This is for reasons of tradition, convenience ('r' is easier to type than 'ɻ'), and generalization ([ɻ] is not the only realization of the American rhotic, and 'r' is a suitable enough symbol for the rhotic). I doubt anyone has been misled to believe that American English uses an alveolar trill, but it serves as an example of this disconnect between map and territory.

All this to say, if you are trying to learn a new language, it is not enough to just read IPA transcriptions. Practice listening closely to the language.

For example, most IPA transcriptions of Russian /o/ are [o]. If you try to learn Russian pronunciation by referencing a table like this (as I know many autodidacts on the internet do), you may think that Russian has a monophthongal [o] sound, much like Spanish.

Vowel phonemes
Front Central Back
Close i (ɨ) u
Mid e o
Open a

(Table from Wikipedia page on Russian phonology)

It is true that Russian [o] may be monophthongal, but Russian [o] is also often diphthongized! I had noticed this for years in listening to Russian, but it was only by looking up 'Russian o diphthong' and talking to Russian-speaking linguists that I found any sources on it. You'll definitely not find anything on it in beginner textbooks on Russian.

The /o/ vowel is a diphthongoid, with a closer lip rounding at the beginning of the vowel that gets progressively weaker [ᶷo] or even [ᶷɔᶺ], particularly when occurring word-initially or word-finally under the stress, e.g. očen' [ˈᶷoˑt͡ʃjɪn̠ʲ] ‘very’, okna [ˈᶷɔᶺkn̪ə] ‘windows’, moloko [məɫ̪ʌˈkᶷɔᶺ] ‘milk’.
(2015) "Illustrations of the IPA", Yanushevskaya and Bunčić. h/t to prikaz_da

The Swedish sj-sound is a drastic case of IPA misleading. If you have ever heard that the sj-sound is a 'coarticulation of [ʃ] and [x]', you may be entitled to financial compensation! Or at least, linguistic compensation, because it is not actually a coarticulation of [ʃ] and [x]. It's often something much simpler - a voiceless 'wh' [ʍ] (like in Southern American English), a labialized [xʷ], a [ʃ] or [ɕ]. Lindblad even offers up a velarized and labialized labiodental [fˠʷ] as a more likely pronunciation. The video below demonstrates:

I was never able to 'coarticulate' [ʃ] and [x], but a Swedish native speaker once told me that I had a pretty good approximation of 'sj'-sound. My approximation was based not off reading IPA descriptions but off listening to Swedish. I ended up doing something like [çʷ] or [xʷ], because that was what I heard.

This does not mean you shouldn't ever use the IPA a guide. Don't take broad IPA transcriptions as the final word, especially if you are just seeing a table of phonemes. If you notice that you hear something that does not seem to be in the standard IPA transcription, trust your ears! Look for articles on the phonetics of that sound if you can. Many descriptions of languages are old, or put together by someone on Wikipedia, and while there are many great Wikipedia editors, they may make mistakes or omit information for brevity's sake.

I'll finish with a quote from one of my phonetics professors - "write what you heard, not what you think you heard." What are some examples of pronunciations you've noticed that don't match common descriptions of the language?

6 comments:

  1. This isn't a problem with the IPA, though. This is a problem of mistaking phonemic transcriptions for phonetic ones, and of mistaking quick-and-dirty or neat-and-tidy "broad phonetic" transcriptions for narrow phonetic ones.

    I was never able to 'coarticulate' [ʃ] and [x]

    I am... and I have it on pretty good authority that that's one of the pronunciations of the phoneme in question, just by far not the most common one. The ones you list are all real, and so are several others! I'll look for the link.

    What are some examples of pronunciations you've noticed that don't match common descriptions of the language?

    Oh, tons. I'll just mention one that really is a matter of the IPA not having a symbol for it: the northern/Standard Mandarin x is not [ɕ], it's the dorso-palatal sibilant, for which the IPA does not have a symbol or even an obvious way to compose it of a symbol and a diacritic or three. Hundreds of millions of people do use [ɕ], but that's a southern accent that goes with merging the retroflexes (sh, ch, zh) into the alveolars (s, c, z).

    ReplyDelete
    Replies
    1. David MarjanovićMay 28, 2024 at 2:20 PM

      Oh, tons.

      Here's one: the mess that is the back allophone range of German /x/.

      (No intention to be anonymous. The new interface just has that as the default, so I didn't notice.)

      Delete
    2. David MarjanovićMay 28, 2024 at 2:43 PM

      Ah, there we go.

      “Official IPA charts do attempt to define [ɧ] as a specific sound, the coarticulated fricative [ʃ͡x]. And yes, that is one possible realization of /ɧ/. Which would be wild enough on its own already, but nope: that is just the beginning.”

      …and the clocks were striking thirteen.

      Delete
    3. This isn't a problem with the IPA, though. This is a problem of mistaking phonemic transcriptions for phonetic ones, and of mistaking quick-and-dirty or neat-and-tidy "broad phonetic" transcriptions for narrow phonetic ones.

      Fair point - it's a bit of clickbait title. :) It would be more accurate to say "good linguists trust their ears - don't rely on the pronunciation you're taught in the beginning of a textbook and/or skimming the phonology page of a language on Wikipedia."

      I agree that you can avoid the problem listed by using the IPA to make narrow transcriptions. The phonological descriptions are not faulty because they don't help language learners; that was never their goal! And even when discussing phonetics, it's hardly necessary or always relevant to include the narrowest transcriptions.

      That being said, my major point is that at some point, you must rely on your ears to guide you when learning the sounds of a new language. This is sort of an article I'd like to aim at myself in my younger days when I would hear something unusual and then think "I must be mistaken; I haven't found an IPA transcription of that yet." I thankfully grew out of that mentality, but I know some language learners who read a table once and don't really update their mental model afterwards.

      There are other reasons to not take existing phonetic descriptions of languages as 'the Bible', even if existing IPA transcriptions are narrow and accurate . New developments may arise in the language that haven't yet been described in journals. There may also come a time when the shade of difference is so minute that it may not be enough to be accurately captured by IPA. For example, "why is my pronunciation of Japanese /e/ not quite right?" is a question that has been plaguing me recently. Japanese /e/ is routinely described as being similar to Spanish [e̞], yet when I pronounce Japanese words with my Spanish [e̞], it doesn't sound Japanese to me, but like Japanese with a Spanish accent. I can tell the [e̞] needs to be moved down a little, but not to the extent of English [ɛ]. The correct degree of adjustment is something I will have to find out through trial and error. (I could even make a spectogram of a recording to get more hints as to where I'm going wrong if I'm really nuts.)

      I am... and I have it on pretty good authority that that's one of the pronunciations of the phoneme in question, just by far not the most common one. The ones you list are all real, and so are several others! I'll look for the link.

      Yes, please share the link if you can find it! And if you can find a video of a speaker using it in context, fantastic. The other pronunciations I've listed seem more common to me, but I admittedly struggle with the sound. Do you disagree with Ladefogen and Maddieson's doubt about the existence of the doubly coarticulated sound?:

      The most well-known case [of a possible multiply articulated fricative] is the Swedish segment that has been described as a doubly articulated voiceless palato-alveolo-velar fricative, i.e., ʃ͡x. The IPA even goes so far as to provide a separate symbol for this sound on its chart, namely ɧ. The sound in question is one variant of the pronunciation of the phonological element ʃ, which is highly variable in Swedish dialects, receiving pronunciations ranging from a palatalized bilabial sound to a velarized palato-alveolar one to a fully velar one. [I]t is not clear that any of the variants is actually a doubly articulated fricative.

      Footnote: One can nitpick and say that my Japanese /e/ issue comes from improper use of IPA. Perhaps the Spanish /e/ is being incorrectly transcribed as [e̞], for starters. All the more reason to practice closely listening!

      Delete
    4. Replying - I just noticed you replied to yourself with the link. Thank you, and I'll make sure to read it.

      Also remembered a very common but outdated transcription: Russian щ is often transcribed as something like [ɕtɕ] or "welsh cheese." Thankfully modern transcriptions correct this to [ɕ], and I haven't come across a Russian beginner's text that uses the 'welsh cheese' pronunciation in a while. My first few Russian textbooks did. It's not a bad transcription because at one point, it appears the sound really was pronounced like that (though I haven't come across an example of it in the wild). It would just be very misleading to a Russian-language student to learn that pronunciation!

      Delete
    5. "good linguists trust their ears - don't rely on the pronunciation you're taught in the beginning of a textbook and/or skimming the phonology page of a language on Wikipedia."

      Ah. A very different point – and one I wholeheartedly agree with!

      Perhaps the Spanish /e/ is being incorrectly transcribed as [e̞], for starters.

      The kinds of Spanish I've heard – from Spain and Chile, not the Caribbean! – have a straightforward [ɛ] as far as I've noticed (and likewise [ɔ]). But I've also heard a song that is partially in a Mayan language, and it has outright [e] and [o], which is actually what you'd expect from their Latin origins. I bet everything in between exists as well.

      English isn't uniform there either. I've noticed plenty of Americans replacing their DRESS with TRAP – [æ] – at least when /s/ or /k/ follow. At the other extreme, in South Africa, Australia and especially consistently in New Zealand, DRESS can be [e], as it was in early RP.

      What I've heard of Japanese has straightforward [ɛ ɔ], I'd say, but I haven't heard much.

      Actual mid vowels are characteristic of Israeli Hebrew – and I find them quite striking. I have no idea why they aren't more common worldwide.

      Russian щ

      [ɕtɕ] still exists, but it's rare by now. [ɕː] is pretty common, despite the marginality of length contrasts elsewhere in the language; [ɕ] may well be even more common nowadays.

      This whole process seems to be shared with Ukrainian, BTW.

      Delete