Arika Okrent presents 11 Fun Facts About the International Phonetic Alphabet; OK, most of them will not surprise anyone who knows anything about the IPA, but I for one didn’t know about the IPA typewriters (“Models publicized in a 1912 supplement to Le Maître Phonétique would cost $1600 and $3200 today”) or the fact that until 1971 articles in the journal of the International Phonetic Association were published in the alphabet — it’s quite disorienting trying to read the one shown, in French!
Comments
Trackbacks
-
[…] Hat looks at the International Phonetic […]
From the early days of 20th-century linguistics there has been an interest in a universal array of possible speech sounds, as if this is a restricted set and tells us something about universals. But if you look for example at the whole IPA you find, as Geoffrey Pullum has pointed out online (http://chronicle.com/blogs/linguafranca/2015/03/05/the-international-phonetic-alphabet/)::
“Studying the full array of symbols… reveals that for almost any point in the mouth or throat where an obstruction or radical restriction of the airflow from the lungs can be made by lips or teeth or tongue or pharynx, such a restriction will be used to produce at least some consonants… And for every reasonable position in which you can hold the tongue and lips and cheeks while permitting unimpeded airflow, there is a vowel sound employing that oral posture.”
In other words, in phonetics all the possibilities are realized (though there is that “almost” in the Pullum quote). This means that there are no phonetic universals. In case the issue isn’t clear, the search for universals involves consideration of everything that is logically possible, then noting what things have not been attested, and pronouncing these forbidden by universal constraints.
Of course there is the problem that by far the great majority of languages ever spoken are extinct. Yet the quest for phonological & syntactic universals continues.
“7. It’s fine-grained enough to represent different accents.”
Not particularly well without diacritics. I can’t for the life of me understand how you could look at the vowel space and decide that this is the most logical set of symbols for it, with three of the world’s most common sounds – [ä], [e̞] and [o̞] – being unrepresented.
Wow, links to Lingorado, which was fascinating.
http://lingorado.com/ipa/
This has the building blocks for my dream program: an AI that can generate and read text in novel accents.
Convert English text to IPA ala Lingorado, choose a few IPA alphabet substitutions, then IPA to speech. Maybe make a few careful natural swaps, or maybe let the program pick a few substitutions at random for completely alien accents.
That could be endlessly entertaining. And possibly even useful for some research into historical speech. Definitely entertaining though.
The French text reveals how much the “Standard” pronunciation has changed since it was written, in particular the long vowels (mostly the nasals and the â in the ending -ation (or -assion). I have long nasal vowels when utterance-final, but not earlier, so in l’écriture phonétique simplifiée I don’t lengthen the nasal vowel of simplifiée, but I would in the simple of c’est simple (not used in the text).
About the vowel chart, the basic four levels of height represent the non-nasal vowels of French at the time, with others added for German, English and a few others. I don’t think it is meant as “the most logical” universal chart.
They’ve mentioned IPA typewriters, but missed my pet favourite: cursive IPA.
There is a wonderful story in a comment by one “haroldfs” at Ken Miner’s link illustrating the usefulness of IPA. To summarise:
The writer was working with the Roma community in Tacoma, WA, to help obtain funding for a bilingual school for their children. But they refused to divulge any stories from their culture as they were forbidden to share information about their language and cultural practices.
Faced with this, the writer happened to find a book by a Swedish researcher who recorded in IPA what he had elicited from a Swedish man who had run away to live with the Gypsies as a boy, including some stories. The writer decided to try it on the Tacoma Roma and practised reading one story, which he then read to the Roma man.
After he read it, the Roma man sat staring at him, then said “Do that again”, which the writer proceeded to do. After staring at the writer, he went into another room to get the Baro (the “big man”) and asked the writer to read it again. The two stared at him some more, then left the room and engaged in a loud argument. They called him in and asked “Who are you?” They considered that he could only be the reincarnation of a deceased Roma, because there was no other way he could replicate the story. Cooperation was much better after that.
The story always impressed the writer’s students, who could then understand how IPA could be useful.
That’s a great story!
disconcerting: It does not help that there’s an ‘s’ and a ‘k’ missing from the image at the start of line 2 and 3 — thick book photocopy syndrome, I think — and sequences such as k i j a a (qu’il y a à) remind me of Victor Mair’s complaints about pinyin with word spaces for each character instead of phonetical word.
I like the “Bijou phonetic typewriter”. Very chic.
Lars: k i j a a (qu’il y a à)
The IPA transcription is hardly more difficult than the common spelling. Writing k i for “qu’i” prevents the reader from expecting the word “qui”.
For me, this particular transcription confirms my long-held conviction that il should be pronounced i (except before a vowel), so ija (as I do say it), not ilja, originally a spelling pronunciation.
Bathrobe, Hat: yes, great story. Something similar happened to me two years ago, actually:
I was sharing an office with other instructors, and one afternoon a student came to the office, looking for his instructor (who was teaching a course in business, I think). As he wasn’t there I asked if I could leave him a message (I knew he would be coming later that afternoon) and the student did: he wanted me to make sure that his instructor had received his assignment (he’d e-mailed it without receiving any acknowledgment of receipt). I asked the student for his name: he was a native speaker of Arabic and was quite surprised when after a few tries I reproduced his name fairly accurately, including the /q/ and /x/: I proceeded to write it down in IPA.
And when the instructor, also a native speaker of Arabic, arrived at the office later that afternoon I delivered the message…and when I read out the student’s name said instructor practically jumped out of his skin, actually stepped away from me, turning pale as a ghost, and for a few seconds I genuinely thought he was going to faint. He then just stared at me with a bizarre mixture of terror and fascination and asked in a halting voice how on earth I had just done that. I showed him the IPA transcription and explained that this is a tool linguists use…but I sense that at some level he didn’t believe me, that at some level he was certain that I had to be an intelligence operative who must have lived in the Middle East or something similar.
Of course, there are limits to such stunts. If the student had been Nuxalk rather than Arabic, the chances of getting it right so quickly, or even reproducing it on demand, would fall considerably. Try saying [xɬpʼχʷɬtʰɬpʰɬːskʷʰt͡sʼ] ‘then he had had in his possession a bunchberry plant’ a few times.
@m-l, of course I’d quickly learn to disregard the word spaces in these sequences, just as I’ve learned to ignore a large percentage of the letters in normal written French.
But this is a representation of spoken French, or at least of written French read out, and I find it slightly absurd to pretend that the pronunciation is in all cases separable into units corresponding to orthographic words. Even if it allows you to discriminate between identical pronunciations that happen to be spelled as different numbers of words.
I don’t know if it’s possible to test whether qu’il y a is actually composed of four morphemes for native speakers, or if it’s just a conventional spelling for a single morpheme or two — but phonetically it’s produced as /kija/, and if listeners can disambiguate that, so can readers.
Lars: I find it slightly absurd to pretend that the pronunciation is in all cases separable into units corresponding to orthographic words.
Reading a phonetic transcription of a language one is used to seeing written is quite different from reading the usual spelling.
I recently discovered something similar with at least one version of French shorthand. (I never learned shorthand myself). The blog Chinook Jargon run by CJ specialist David Douglas Robertson often reproduces phonetic transcriptions of texts originally written using the old Duployé shorthand system, which was used by a French missionary over 100 years ago to teach native people in Southern British Columbia to read and write CJ which was then used as a lingua franca by both natives and newcomers. The texts in question include some written all in French (shorthand), with words separated as in the IPA text (all the French schwas are omitted, as for example in “ce que je fais” s q j fè). (The blog is interesting and entertaining both linguistically and culturally – worth looking at if you are interested in Northwest history on a small scale).
Or indeed just /kja/ depending on the speed of speech.
David: but phonetically it’s produced as /kija/ — Or indeed just /kja/ depending on the speed of speech.
I would say that /kja/ is not a pronunciation of qu’il ya (= que il y a) but from the very colloquial qu’y a (= que y a, leaving out the il altogether).
This omission of the pronoun il is quite frequent in colloquial speech, with phrases where il does not have a referent, as in (il) faut ‘must’, (il) vaut mieux ‘(better, rather (to do …)’, and some others including (il) y a “there is”. For example, in the folk song (or rather rhyme) Y a un’ pie dans l’ poirier … ‘there is a magpie in the pear tree …”
For pet IPA peeves, my current favorite one is the lack of distinction between more fronted and more backed [k] and [x], except with diacritics. Speaking of which: David M., how would you narrowly transcribe German ach, Aachen, and hoch? Are the velars pronounced exactly the same?
A truly baroque elaboration of the IPA is found in Luciano Canepari’s canIPA.
@Y:
My peeve is that, perhaps because of the lack that you identify, many people transcribe fronted /k/ and /g/ with /c/ and /ɟ/, which is ambiguous since the “palatal plosive” found in languages such as Hungarian sounds quite different to me (Wikipedia identifies the Hungarian sound as alveolo-palatal, and distinguishes it from the post-palatal sound that is much more common as an allophone of /k/).
@Eli Nelson: Yes!! This has confused me to no end.
@Y: Yeah, I’m a fan of canIPA. I’ve seen it argued that since there’s an infinite possible number of speech sounds, any attempt to significantly augment the IPA is a fool’s errand; but I would argue that if two sounds are distinct enough that we can make generalizations about their use in different languages (e.g. Hungarian “[ɟ]” is not the same as French [ɟ]; Spanish “[e]” is intermediate between Italian [e] and [ɛ]), then it’s worth having different symbols for them.
That’s what the diacritics are for. If a distinction is phonemic anywhere it gets a new symbol, otherwise not. I think that’s quite elegant.
@marie-lucie: ‘“ce que je fais” s q j fè’
Quite off topic:
I’m practicing my French listening skills recently, and often wish I had a dictionary that would let me look up French words by their phonemic representation. I don’t know if one exists, but I’m thinking I could create one by re-indexing Wiktionary data. (Another one of those projects that I will work on when I win the lottery and quit my day job)
Anyway, as part of that thought process, I started wondering how much more concise French would be if it were written phonemically. And that in turn made me wonder whether /ə/ is really necessary as a phoneme in (let’s just say contemporary Parisian) French. Which is what ties in to what marie-lucie mentioned. I mean, can’t almost all [ə]s be reconstructed from phonetic rules? I’d be happy to be shown how wrong I am.
an AI that can generate and read text in novel accents.
But can it do the police in different voices?
dainichi: a dictionary that would let me look up French words by their phonemic representation.
You could try the TLFI (Trésor de la langue française informatisé), a free access online dictionary. You can look up a word even if you don’t know the spelling, by writing it more or less phonetically according to general French usage, for instance “anfan” for ‘enfant’.
(more later about the schwas)
But can it do the police in different voices?
Only with its mutual friends.
m-l, that may be impossible to tell. You’re of course right that il is routinely omitted in colloquial speech before y a.
For me, yes, except for length: [axː], [ˈaːxːŋ̩], [hoːx]. But I’m sure you’re referring to more northern accents, where the ach-Laut is the uvular [χ] in most environments, and quite possibly intermediate between velar and uvular in some others – which ones exactly varies, but the closer to get to [ɑ], the closer you get to [χ]. (Consonant length hasn’t been phonemic there in a long time.)
Strangely, among its many good points, it contains a misunderstanding of aspirated consonants as clusters with [h]. That’s not, phonetically, what they are!
It is quite different. Rule of thumb: when Wikipedia claims a language has palatal plosives, it’s usually wrong.
To my surprise, the Latvian ķ ģ really are [c ɟ], though.
That’s the theory. The practice is a bit different; e.g., the symbol [ɱ] was created for the English allophone long before it, or something much like it, was discovered to be phonemic in one language.
Besides, sometimes it’s actually useful to have a phonetic rather than a phonemic alphabet. 🙂
French [g˖]. And the Greek “[c]” is the exact same [kʲ] that you can find in Russian </rant>.
That’s what the diacritics are for. If a distinction is phonemic anywhere it gets a new symbol, otherwise not. I think that’s quite elegant.
I think this approach yields a spotty and haphazard result, depriving us of many symbols that could be very useful in interdialectal and interlinguistic comparison. (And although there may be some merit to this stance regarding consonants, though I would still differ, I think it makes no sense to apply it to the smooth continuum of the vowel space.) As DM – and Canepari – observe, the IPA should abandon the pretension to being a phonetic rather than a phonemic alphabet if this is its operating principle.
There’s the fact, too, that although the IPA allows for the extensive use of diacritics, they’re inevitably ignored by all but the most assiduous transcribers, thus leading some distinctions to be arbitrarily privileged over others. The lenition of /t/ to a non-sibilant fricative in Irish English is no less salient or important than its lenition to [ɾ] in American English, but the former is shunted away as some insignificant trifle because the IPA lacks a symbol for it, while everbody and their mother knows about the latter. (And God forbid you should ever need to use multiple diacritics: there’s nothing elegant about [e̞̠].)
I have my disagreements with Canepari, mostly regarding his descriptive work (his observations can be tremendously incisive and helpful, but they’re sometimes overly ambitious and miss the mark), but I think his extended IPA is leagues ahead of the official version and proves that a rigorously designed phonetic alphabet can be a practical reality. Having acclimated myself to it for personal use in learning and description, I couldn’t imagine turning back to regular IPA – it feels almost as limiting as that old-timey dictionary notation that we all turn our noses at. It’s a shame that there isn’t any published canIPA font, but I was able to find a couple decent-looking homemade ones by snooping around.
Dinka, for instance, shows how the IPA fails even in its stated approach: it’s one of several languages that contrast dentals and alveolars, but for some reason this distinction doesn’t warrant separate symbols. There’s a generation of English-speaking language enthusiasts who’ve been conditioned to think that any [t] or [d] sound can be described as “alveolar”, an example of how the fuzziness of the IPA can actually encourage misinformation.
~~ shrug ~~
Design is one thing, implementation another. Changing the ambiguous status (without diacritics) of [t] and [d] would cost far more than it is worth at this stage. Similar reasoning accounts for allowing the various warts that Unicode has to remain intact, such as the occasional misspelled name like BYZANTINE MUSICAL SYMBOL FHTORA SKLIRON CHROMA VASIS.
@Lazar: I must be really old. I learned them all (from books, not in class–maybe the books were old) as “dental” and only later learned about alveolars.
David: kija vs kja “m-l, that may be impossible to tell. ….”
Have you never heard anyone say “k∂ ja” ‘que y a’ (no il at all? (as in qu’est-ce que y a pour dîner? ‘What’s for dinner?’) Of course not everybody would say that. I would not, but some people do.
David: ( French [ɟ]) – French [g˖]
What is this about?
“Speaking of which: David M., how would you narrowly transcribe German ach, Aachen, and hoch? Are the velars pronounced exactly the same?”
I am not an expert in all of this. But from personal experience with my childrens’ numerous friends, they all seem to pronounce words like this with such variation that I can only understand about half of it.
Are there really enough symbols in the IPA to transcribe the differences between words as even spoken by children?
I assume that a computer program could simply have millions of variations recorded and be able to classify them, and then reproduce them. Does the alphabet encompass that ability?
This is common in Dravidian and Australian languages, and it’s likely the reason for why the Indo-Aryan languages almost down to the oldest documented kind of Sanskrit have apico-dentals instead of the expected lamino-alveolars.
Iä! Iä!
Impossible to tell – the Parisians have deleted so many schwas that there’s only [kja] left whether il is intended or not. I’ve encountered [kilja], [kija] and [kja], and nothing else.
The French /k g/ are noticeably fronted, at least before front vowels including /a/. Some people seem to believe this is what the symbols [c ɟ] mean; I’ve even once seen them explained this way, with soundfiles, on a French university website. So when Lazar mentioned “French [ɟ]” yesterday at 2:57 am, that’s what I thought he meant.
BTW, there are languages with a phonemic contrast between front velars and back velars (which “are in no way equivalent to uvulars”). Here’s one.
Not quite, but canIPA with diacritics probably gets there.
dainichi: a dictionary that would let me look up French words by their phonemic representation.
There is a vocabulary list at the end of Bonnard’s 1915 “An elementary grammar of colloquial French on phonetic basis”: https://archive.org/details/elementarygramma00bonnrich which lists words phonetically (using IPA).
Another useful book along the same principles is the 1919 Phonetic French Grammar by Henry Morse Wells https://archive.org/details/phoneticfrenchgr00well.
You can test it out for yourself at Lingorado, where you can make an interpretation of content into the IPA for British or American elocutions and listen to the outcomes. Look at the British (above) and American (underneath) forms of “Herbs and tomatoes for your vitamin research center.”
The vowels of the IPA aren’t only a gathering of images—they’re likewise masterminded into a diagram where the position of the image generally compares to the position of the tongue in the mouth while delivering that vowel. It’s a key, as well as an intriguing visual presentation.
http://www.ongpohlin.com/sbf-center-office-space.html