It's really helpful to understand the theory behind the pronunciation, instead of just repeating words and attempt to make the same sounds as my coach. As in "this is what your tongue should be doing, but you're doing this other thing instead."
You can even use the chart (and maybe Wikipedia) to identify specific reasons as to why you're doing something wrong. For example, there's just one way to pronounce "S" in Finnish, whereas there are four ways to do that in English. It's just something that never crossed my mind before starting to work with a professional.
Week, weak and wick will be pronounced and will sound exactly the same to a non trained Hebrew speaker, and to much much more hilarity - sheet/shit, peace/piece/piss.
We have only seven vowel sounds. A i u and open / closed e o. But different parts of the country often use only one of the two e and o sounds.
In the beginning of the class, the professor introduced himself with "my name is X, but you can call me Y if that is too difficult". The two names were mostly the same, except that the first sound in X was a complete mumble, where I could not make out what he had said (and he repeated it several times).
Later in the first class he handed out a syllubus which included his name in IPA. I immidietly knew not only how to pronounce it, but was able to do so without any difficulty.
(For those curious, his real name started with /ʒ/, while his nickname started with /dʒ/)
(Going off of memory here, I probably butchered the pronunciation anyway).
/ʒ/ does not occur in word-initial position in English, and for whatever reason it was incredibly hard for me to perceive it there.
There was probably other more subtle differences between the two that did not make it into the transcription (e.g., his native name was pronounced with his native accent instead of his English accent)
If you compare [ta] and [da], you find that the only difference is the time between when you make the consonant, and when your vocal chords start vibrating (voice onset time). In theory, VOT is a contimum, with any value being possible. However, in English it forms a tri-modal distribution /tʰ/ /t/ and /d/. The experiment artificially edited a sound to vary between /t/ and /d/, including with VOTs between the two that do not occur in English. What they found is that people put all of the sounds in 2 boxes, and were unable to distinguish between sounds in the same box, even if their VOT varied considerably.
However, when test subjects were played the same sounds, but told they were listening to rain drops, this effect disapeared, and they were able to distinguish between sounds in the same box.
Seems to be exactly what you describe.
They also used an ABX methodology, which forces the subject to put sounds into boxes by essentially asking if X is more like A or B, not if X is different from both.
The one I am thinking of used an odd-one-out methodology, where the subject was presented with AAX, and asked to pick which sound was different from the others (where the others were genuinly the same sound).
The one I am thinking of also found a priming effect, which yours apparantly didn't.
Doesn't gendarme start with /ʒ/?
Adopted word from French, but almost all words have origins somewhere else.
I never noticed that until now. While it's obviously not an English word, plenty of English speakers have discussed French explorer Jacques Cousteau. I can't think of any other examples.
One of the languages I know allows for a word-initial /ŋ/, which English does not have at all.
Vietnamese, by chance? I'm guessing that because the only initial /ŋ/ word on my radar is the name Nguyen.
There are (at least) 3 different consonants in Hindi that just sound like "T" to me. So it took me 3 attempts to spell this Hindi word correctly. The kids all absolutely lost their minds with laughter at this - they were all yelling (what I heard as) "no not tuh, TUH" and just couldn't understand why I couldn't tell the difference.
If you get a tone wrong, sometimes people will understand the erroneous word due to surrounding context. But pretty often you'll just elicit a blank stare. This is especially the case for short phrases, e.g. when you're asking for directions.
My teachers have no problem telling apart my j and q, but I rarely hear the difference despite all the cosonants in my native language coming in palatalized/non-palatalized pairs.
j / zh
q / ch (the former is further forward in the mouth than 'ch' in English, and the latter further back, with a fully clenched mouth)
There's a book called "Learner English" by Michael Swan and Bernard Smith for identifying common problems faced by English learners coming from different language backgrounds. They don't have a section on Finnish, at least in my second edition.
What are those four ways?
I did a search in cmudict and found examples!
sample → /s/
treasure → /ʒ/
visible → /z/
unsure → /ʃ/
I could probably have thought of these eventually, but cmudict made it a whole lot quicker.
Pink Trombone - https://news.ycombinator.com/item?id=18912628 - Jan 2019 (75 comments)
A mouth simulator - https://news.ycombinator.com/item?id=13973261 - March 2017 (1 comment)
However, the IPA has one distinguishing feature: it’s standardised, by the International Phonetic Association. By contrast, Americanist notation (the main competitor) is mostly unstandardised; for the most part, it’s not so much a ‘phonetic alphabet’ as much as a set of conventions people follow. And this doesn’t matter if the author has been careful to define their terms correctly, but it can be a real pain if they’ve forgotten. (e.g. for a while, I believed that Kalam had [r] as an allophone of [d], until I realised that the author had used ⟨ř⟩ to mean [ɾ], not [r]…) Thus, the linguistics community has mostly standardised on IPA, with the exception of some subfields which still use Americanist notation.
I wasn’t really aware of the IPA, is this mostly used for teaching purposes when learning a new language?
As far as people learning English as a second language it seems like one of the biggest hurdles is learning all the ‘exceptions’. Especially for people coming from languages where all vowels only have one pronunciation.
English pronunciations are so flexible & varied that it must be frustrating. There are so many words that you just have to learn individually through repetition, since they don’t follow the ‘framework’ of the base pronunciation rules.
It’s mostly used for linguistic purposes — most often in grammatical descriptions of obscure languages, or for precisely specifying the pronunciation of a word (e.g. in my dialect ‘mutable’, say, is [mjʉːtˢə̆bu]). But I believe it gets used in language learning as well.
No, it's used everywhere in phonology, for every language.
I started a project like that (also to train myself using web components), except it listed the phonemes of a given language, selectable to the user. I should complete it one day...
Sure, of course, but I was thinking mostly of those grammars (usually Sino-Tibetan ones, for some reason) which use IPA throughout instead of a romanisation. (I suppose that by now I’m just so used to IPA being used for phonologies that I don’t even think about it.)
> I started a project like that (also to train myself using web components), except it listed the phonemes of a given language, selectable to the user. I should complete it one day...
So like PHOIBLE? (https://phoible.org/)
One of the developers of this technique of teaching English sounds is a friend of mine. There's also games, matching words to colors, that can be played with all ages.
The few times I've been able to participate in one of these sessions, I learn something...
The difference between "door", "orange", and "dog" -- in southeastern states, I hear a vowel shift from one towards the others. My Californian friends couldn't quite hear it.
The next level up would be "phonemes", which must contrast against other phonemes within a particular language. For instance, in English /b/ and /v/ are phonemes, because "bat" and "vat" are two different words with only that portion changing. (This is known as a minimal pair.) However, in Spanish, /b/ and /β/ are not separate phonemes; they are two phones in complementary distribution.
It would be nice to extend this, perhaps on a different page, to examples from specific languages. That would fill in some missing parts---long and short vowels, nasal vowels, and diphthong---and also show the vowels within a specific context. I have these for a few languages and will try to reach out to the author in case they are interested in collaborating.
That's almost as magical to me as people who look at a sonograph and casually read off what the speaker said.
- phoneme, ns:"US English letter combinations", 
- phoneme, ns:"schema.org/CreativeWorks which feature said phoneme", 
AFAIU, WordNet RDF doesn't have links to any IPA RDFS/OWL vocabulary/ontology yet.
Audio doesn't do the same job, really, since it relies on a trained ear and the ability to distinguish what each individual sound is. Experts have this. The people that most need the help, don't.
English writing is not a good phonetic system for representing speech even for the English language: IMO a 1-to-1 mapping between pronunciation and spelling should be the goal.
You might want to explain how to pronounce the word in a British or Australian accent or compare it to how a word was pronounced 200 years ago. In that case you can use a phonetic alphabet to show the difference.
The idea is to have a writing system which can be used to accurately record all possible sounds. It's helpful for learning languages with bad alphabets such as English (https://www.teachingenglish.org.uk/article/teaching-pronunci...), for describing accents and for comparing languages and linguistics work.
PS: I don't know IPA so I'm probably wrong about everything.
It took me ~6 months to understand where a word started and ended. When I tried saying the most simple things I got blank stares. But when when they repeated it back to me it sounded exactly like what I said all along.
Try to say "Dry wood burns doesn't it?" in Thai and you get different variations of the word "Mai" which sounds like "Mai mai mai maii?" ... or "spicy duck with pepper" that is just "phet ped pet ped PET ..." etc
I eventually ended up marrying a Japanese girl (so heating with wood was never an issue neither was ordering spicy duck with pepper). And their language is all rather monotone (as you probably know from your own time there).
using this IPA would have gotten me more ridicule than ordering spicy duck from a wooden grill. And if I would have trusted HN (and this chart) I'd still be feckless today thank you very much :P (I'll take an Indian Pale Ale though without all the moaning)
You might be able to hear an unfamiliar word and try to reproduce it, but in a lot of cases you will only get, at best, a rough approximation unless you also know how to shape your tongue, where the sounds come from and what the accent is.
English is a very forgiving language, generally speaking, but highly tonal languages (like Vietnamese and Mandarin) won’t let you get away with simple mimicry, at least not without a lot of work to nail the pronunciation.
For individuals wishing to learn a foreign language, the IPA probably isn't super helpful. Your mind needs to learn to pick up on new sounds if you are going to be able to converse anyway and you should definitely learn your target language's orthography to be a proficient reader/writer.
That said, I enjoy learning languages in addition to studying (comparing, contrasting, etc) and have found the IPA helpful to give me a leg up on that. Therefore, I recommend the IPA for polyglots and linguists.
(They do sound somewhat different but that's because English ch is always pronounced with rounded lips.)
Also, seriously, couldn't they think of a better name than "palato-alveolar" vs "alveolo-palatal"?
I was trying to find the Swedish "u" sound. According to the Swedish Wikipedia page on the sound, it says it's written as "u" in IPA, but it doesn't sound right at all. The "ʉ" sound is a lot closer, but still not right.
What is the correct IPA representation of the Swedish u, as in "ful"?
IIRC, Swedish /u/ is somewhat fronted, but not quite /ʉ/. It's somewhere inbetween those two. There isn't an IPA symbol for every possible sound, just the most common among languages, but there are diacritics for special cases or more precision.
See the "Advanced" diacritic:
The diacritics are also used for _phonetic_ rather than _phonemic_ renderings as the surrounding sounds in a word, etc. affect the exact outcome of an individual sound. Phonemic renderings in the IPA are typically enclosed in slashes and phonetic in square brackets.
Hold your hand in front of your mouth and feel the difference in pressure when you pronounce the first sound of the word "tune" versus the word "dune". This will let you feel that puff of "aspiration." Next, try putting your fingers over the front of your throat, (on your layrnx or where the Adam's apple is on men), and pronounce the sounds slowly. You may be able to feel the difference when your vocal cords start to vibrate as you say the "d" sound. In native English speakers this happens shortly after the pressure is released from your tongue, while in the English "t" sound, the vibration doesn't start until the vowel does. The "d" sound in the example from the chart has the vocal cord vibrations start immediately as the pressure is being released.
The aspiration will be even more obvious if you try speaking into a microphone.
Native speakers have the advantage in spoken conversation. They quickly understand not just the word, but the context. No native speaker would ever confuse "Give that document to Ted" with "Give that document to dead" since the latter is nonsensical. Even if you actually said "dead", most people would interpret it as "Ted" without even noticing anything odd about your pronunciation. (In fact, in a context like that between two vowels, it's entirely possible that you would keep your vocal cords buzzing and produce a "d" even if you'd meant to say "t").
There's very little blame to be placed on liaison, since word boundaries are unclear in every language, including languages that e.g. don't allow syllable-final consonants. English speakers feel like they have a strong grasp of where word boundaries occur because of the orthographic space, but the space in the writing system does not coincide with word boundaries in the language; it is heavily conventionalized.
Obviously this is a spectrum; some transcriptions are so broad that they transcribe the english <r> as /r/ when the context is clear that we're talking about english, even though it should properly by /ɹ/. And in my narrow transcription, I didn't bother to notate vowel length, because it didn't matter for the given example.
(Note: I use <> for orthography (how it's written in the language), // for broad transcription, and, and  for narrow transcription)
Ah yeah, that's true. This is because most Spanish speakers pronounce /t/ as denti-alveolar (the tip of the tongue touches the back of the teeth) while most English speakers pronounce /t/ as purely alveolar (the tip of the tongue stops at the back of the alveolar ridge). I'm not sure why the IPA chart doesn't differentiate these specific phones, honestly. I imagine it's because although the sounds are very different cross-linguistically, maybe there aren't any languages where these phones are contrastive and thus they don't warrant separate glyphs in the IPA. Still, I'd love to see them!
That said, I think this is actually a relatively uncommon case in the IPA, because in addition to this simple chart the IPA also specifies a number of diacritics for more complex and precise transcriptions. These diacritics are often not used in broad transcriptions within a single language because most languages don't differentiate phones at this level in a contrastive manner. But in a narrow transcription, the diacritics are used as appropriate, generally to help the transcriber make a point about what is noteworthy in that transcription.
For instance, in English let's consider the "t" as you mention. We'll look at "tap" and "pat". The broad transcriptions for these words are /tæp/ and /pæt/, respectively. But the narrow transcriptions would be more like [tʰæp̚] and [pʰæt̚] (Note that broad transcriptions are phonemic and given between slashes, while narrow transcriptions are phonetic and given between brackets). What this shows is that, in English, syllable-initial voiceless stops are aspirated (indicated by the little "h"), while word-final voiceless stops have no audible release (indicated by the angle symbol, known as "corner"). So in addition to the main letter glyphs of IPA, there are tons of diacritics that help you write your transcriptions more precisely!
It's because sounds vary continuously, but the IPA is by necessity discrete. (A problem that is much worse for vowels, but still comes up for consonants.) You have to collapse variation somewhere.
The IPA notably doesn't have dedicated symbols for affricates. (Sounds that consist of a stop immediately followed by a fricative in the same location, like "z" in "pizza" or "ch" in "chill".) So English "ch" is represented /tʃ/, or if you want to be really explicit there's an affrication diacritic.
Recognizing that affricates consist physically of a colocated stop and fricative was felt to be a theoretical advance. But there's a funny story -- Peter Ladefoged went to document a language somewhere (the Americas?) and was proud to use the IPA to record its sounds. He insisted on it.
He insisted on it even after discovering that the language in question made a phonemic distinction between affricated tʃ, which he represented by /tʃ/, and /t/ followed by /ʃ/ without affrication, which he was forced to represent by /t.ʃ/. This would have been much easier to understand with a more bespoke system representing the two adjacent sounds as /tʃ/ and the affricate as /č/.
I get the continuous vs discrete aspect for sure, but you'd think we'd at least have different symbols for each place of articulation! /t/ crosses four different places (dental, denti-alveolar, alveolar, post-alveolar) and the best we can do is cut that down into two groups of two places with the dental diacritic which says "some teeth contact". I think if I were to design such a transcription system from scratch, I would try to make it at least possible to express each place/manner combination, but that's just me!
Looking at the example sidethread of Mandarin pinyin "q", it is a laminal palatal aspirated affricate. Its equivalent in English is a much more complex issue than the equivalent of a Spanish "t", which everyone agrees on. The "q" may be perceived as a "ts" (witness "Tsingtao beer"), which is the English sound combination most closely matching the positioning of the tongue, though not the positioning of the restriction in airflow, or it may be perceived as "ch", which is the English single sound most closely matching the place and manner of articulation. Different English speakers may even disagree on the interpretation while listening to the same speech, and the same English speaker will disagree on (or be confused about) the interpretation when listening to the same speaker produce the sound multiple times.
There are languages that distinguish retroflex stops from alveolar stops, and IPA obliges those languages with different symbols for the two places of articulation. Do you know of a language that distinguishes alveolar stops from dental stops?
> There are languages that distinguish retroflex stops from alveolar stops, and IPA obliges those languages with different symbols for the two places of articulation. Do you know of a language that distinguishes alveolar stops from dental stops?
I actually mentioned this in the top comment:
> I'm not sure why the IPA chart doesn't differentiate these specific phones, honestly. I imagine it's because although the sounds are very different cross-linguistically, maybe there aren't any languages where these phones are contrastive and thus they don't warrant separate glyphs in the IPA.
What I meant in my most recent comment was that, if I were going to create a new IPA from scratch, I would probably try to have some diacritic or distinct symbol for each place of articulation, such that any possible place/manner combination could be expressed uniquely. As it stands, the presence of distinct symbols in the IPA seems governed by whether there are examples of real-world languages in which phones are considered contrastive. This is perfectly pragmatic, but it leaves me feeling disappointed that I can't easily express something that is right there on the chart! Having <t> correspond to 3-4 places of articulation with only a diacritic to distinguish 1-2 of those is disappointing, even if it's technically sufficient for any transcription within a given language.
I think this is a pretty spot-on observation. IPA does offer a bunch of diacritics for making more narrow transcriptions. For instance, Spanish's dental <t> is [t̪].
However, the narrower the transcription, the less applicable it is to a language in general. Things like dialectal and even individual differences start to come into play. So in general transcription will only be as narrow as necessary to achieve some point.
That is not what "phoneme" means. They are different sounds, but they aren't different phonemes in either language; to be different phonemes, the same language would have to consider them different. Neither does -- in both cases, one is "/t/" and the other is "weird /t/".
However, since there is no alveolar [t] in Spanish, it will generally just be transcribed as /t/ rather than having to put the diacritic on all the time.
What I was talking about was how the Spanish /t/ has the tip of the tongue just a bit further forward, so that it touches the back of your teeth. English /t/ is, for most people, purely alveolar, so there's no teeth contact.
It’s even more crucial with d. An English native speaker who is a beginner in Spanish will pronounce tened in a way that will be interpreted by Spanish speakers as tener.
> “Mr Vimes," said Mrs Winkings, "ve cannot help but notice that you still haf not employed any of our members in the Vatch..."
> Say 'Watch', why don't you? Vimes thought. I know you can. Let the twenty-third letter of the alphabet enter your life.”
I am not a native English speaker, yet I decifer everyone execept gargoyles:
— e cuns uk ere um-imes an awks oo ugg
(Oh, and don’t worry; I’m a native English speaker, and I can’t understand Pratchett’s gargoyles either.)
Language is unconscious and you won't understand how it works until you let go of a lot of false culturally-programmed notions, and then actually study it by engaging with the work of linguists.
For Hindi there are additional phenomena that will alter the sounds of spoken language in certain contexts.
In French, Italian, or German, though, they already do. Those languages are quite orthographically regular.
I agree! I’ve heard that IPA was optimised for movable type, a medium in which it was easy to reverse and flip letters — thus we get e.g. ʌ,ɐ,ə,ɘ,ɔ,ɟ etc. (This, incidentally, is why you can write text upside-down using Unicode.) This can be contrasted with e.g. Americanist notation, which was optimised for handwriting: š, ȼ, ƛ etc. are easier to write than the IPA equivalents ʃ, t͡s, t͡ɬ. I’m not sure what a phonetic alphabet invented today would look like though — probably it would look something like X-SAMPA , restricted to mostly ASCII characters and making heavy use of uppercase, punctuation and digraphs. (A_i j}:z It r\e:li:, ba_"t It lUks lA_ik DIs.)
Edit: er, I meant high-pass filter, I guess.
I spent months brewing IPA's using the same basic recipe with tweaks in yeast, mash temperatures, and ingredients and have now moved on to do the same with a saison (just kegged the first today), but am very happy to see this chart pop up.