I am looking at Help:IPA/Nguni and Help:IPA/Welsh, and wondering what the exact difference is between these sounds, and if there are any good audio recordings (or if you can make one!) showing how they are different:


It seems that is a "silent m", and both the aspirated versions of m are indistinguishable. What really is the difference between the two aspirated h's, and how does that differ from the so-called voiceless m?

  • 2
    What's the purpose of the Romanization scheme? Who's going to be using it? (Also, aren't Nguni and Welsh both already written in the Latin alphabet?) – Draconis Dec 27 '20 at 20:40
  • 3
    Why does a new phonetic alphabet help that, though? Won't they just need to learn your custom phonetic transcription instead, which will have about as much complexity as IPA (if you really want to support "every major language") and won't be compatible with published dictionaries and other resources? – Draconis Dec 27 '20 at 20:58
  • 4
    The general public doesn't care about language sounds not in English, and therefore they don't use systems for rendering foreign sounds. That's such a fundamental point that you ought to be able to grasp, and see that attempts to convey this information via letters is hopeless. To succeed, you have to use actual sound files. – user6726 Dec 28 '20 at 0:52
  • 5
    @LancePollard Personally, my issue is that you tend to immediately dismiss all previous work on problems and try to reinvent every wheel yourself. Which is perfectly fine if this is just for fun! There's nothing wrong with that, and I've come up with plenty of idiosyncratic transcription systems for my own use or just to play around with orthography. But if your goal is to make something for real-world use, you should also consider the different systems that have already been developed, and the issues they've run into and overcome, instead of immediately… – Draconis Dec 28 '20 at 1:30
  • 4
    …dismissing them as "failures" just because you personally dislike them. Given how widespread IPA is, clearly it's fulfilling some purpose for its users. There's no universal panacea that solves every problem; a design is successful or unsuccessful relative to who's using it, and for what purpose. – Draconis Dec 28 '20 at 1:33

"Exact differences" only exist between languages, and you absolutely cannot rely on a person's writing system as a resolution of how things are pronounced. First, most linguistic material is not created by highly-trained ear phoneticians with classical Edinburgh-type knowledge of IPA letters. Second, even looking at material created by such scholars, the IPA has only a certain granularity, so you cannot write "exact differences" between languages. You can perhaps get "close enough".

Any of [mʱ m̥ mʰ] are good enough for denoting "that phoneme", and all three are used in some work. For example, in Klamath [m̥] was used because it was easy (in the typewriter days); [mʱ] tends to be popular with people having certain ideas about phonetics; Gray uses [mʰ] for the Tanzanian language Kisi. Given your interest, the most important thing to realize is that written representations of languages are generally not reliable reproductions of how things sound, unless there is an accompanying detailed description of pronunciation. I might therefore describe for you how "mh" in Gogo sounds different from "mh" in Shona or Lhasa Tibetan, though I would prefer to refer you to sample recordings (sorry, I'm fresh out).

Given that IPA is inadequate for conveying the facts of pronunciation about languages, unless you invent a new system with 10 times the number of letters (and can provide reference values for all of those letters), you will still face the granularity problem, that any writing system can only get so close to actual pronunciation.

For your three choices, the first has a breathy voiced release; the second lacks vocal fold vibrations during its production; the third has "aspiration". If you find two differences being contrasted in Ladefoged and Maddieson's book, you can take that as an indication that there could be is a meaningful claim underlying a choice of spelling. The only language that I have encountered with actual voiceless nasals is Angas, where in the production of utterance-final nasals (and liquids), the vocal folds stop vibrating in the middle of the consonant (that's more a "have to see it" than "can hear it" feature). They do not sound at all like "m̥" in Tibetan; Tibetan "m̥" (romanized as "mh") also does not sound like Shona mh which is breathy throughout (thus [m̤], which you didn't ask about).

  • Thank you for saying IPA is inadequate, I was thinking that myself. Those charts like the vowel chart (x and y axis of mouth positions) is much more accurate. Trying to simulate pronunciation with a speech synthesizer is even more accurate than that, and finds IPA almost useless. You need a mathematical model which can model human speech production to be really accurate. No lettering system will be this accurate. So all I am looking for is a lettering system that is "close enough", as it is never going to be anything close to a speech simulator model. But that's not what I need, I need letters – Lance Pollard Dec 28 '20 at 0:41
  • 2
    Why do you need more that this: web.uvic.ca/ling/resources/ipa/charts/IPAlab/IPAlab.htm ? – user6726 Dec 28 '20 at 0:47
  • Niggle: IPA has symbols, not letters. There are no letters in the IPA. – Araucaria - him Dec 31 '20 at 0:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.