It seems to me that IPA is badly designed and not suitable well for many languages other than English.

Some problems are:

  • It uses different characters to denote the same sounds. For example, [ʍ] and [w̥], or [ʃʲ] and [ɕ].

  • It does not have characters for even widespread sounds that are used in different languages. For example, the sound, conducted by English "ch" exists in English, Czech, Italian, Russian and many other languages. Yet IPA uses two symbols for it: [tʃ]. The Russian [ч] needs three IPA symbols: [t͡ɕ]

  • It uses the same symbols for different sounds. For example the above-mentioned transcription [tʃ] is used both for Belarusian phoneme which is denoted by the letter ч and for the combination of two phonemes [тш]. Similarly, Russian отселить "to resettle away" has [тс] while оцелить "to give aim" has [ц], the both again will be [ts] in IPA.

  • It uses different methods to indicate the same features in different sounds. For example, palatal variant of [ɣ] is denoted [ʝ], palatal variant of [tʃ] is denoted [t͡ɕ] while in other cases a superscript [xʲ] is used to indicate palatalization.

  • It uses completely different symbols for voiced and voiceless, or palatal and plain variants of the same sound, which does not reflect the sound proximity.

Reading professional literature on comparative linguistics I never encounter use of IPA. The authors usually use ad-hoc transliterations for even languages which have no written form, and transliteration systems of different branches may grossly differ making comparison difficult.

That said I wonder whether IPA really obsolete and whether there is recognized need for a better alphabet which could be used in comparative studies?

  • 8
    Don't forget that the IPA is an "alphabet", not an "orthography". Just as the Latin alphabet is used very differently for many languages including English and Italian, IPA was also designed to be flexible enough to be used for many languages. There are often multiple "orthographies" for writing phonemically the same language, but there will be a lot less (still some) variation in how to write phonetically. I don't believe Americanist transcription or any other transcription avoids every one of the issues you list. Commented Nov 14, 2012 at 12:12
  • 4
    I think that, for the purpose it was designed for -- transcribing the sounds of human languages, as they are produced by living speakers of the living languages -- IPA is fine. For phonemic purposes, which are almost always the points at issue in comparative linguistics, ad hoc subsets of IPA, augmented by traditional orthographic symbols like č, are common and unremarkable. For other purposes, you are free to design your own; the nice thing about standards, as I'm fond of remarking, is that there are so many to choose from.
    – jlawler
    Commented Nov 14, 2012 at 15:22
  • 4
    From Professor Wells' blog: "IPA symbols have always had to be interpreted in accordance with conventions implicitly or explicitly defined by the transcriber who uses the symbols. [...] It is convenient (= practical and sensible) for us to use the same phonemic symbol t for the unaspirated dental plosive of French, the aspirated dental plosive of Swedish, the unaspirated alveolar plosive of Czech, and the aspirated alveolar plosive of English." phonetic-blog.blogspot.com/2012/09/false-alarm.html
    – Alex B.
    Commented Nov 14, 2012 at 16:17
  • 6
    The IPA does let you distinguish between an affricate langsci.ucl.ac.uk/ipa/symbols.html and a cluster [ts]. You know, there's Google and the IPA Handbook, too - there's even a chapter on Russian there.
    – Alex B.
    Commented Nov 14, 2012 at 22:18
  • 4
    So what? Or would you rather have a different phonetic notation system for each language? By the way, we usually treat [ts] in English as a cluster and not an affricate because of phonological considerations. As far as articulatory phonetics is concerned, [ts] in English should be an affricate - a stop followed by a homorganic fricative.
    – Alex B.
    Commented Nov 14, 2012 at 23:54

3 Answers 3


Your criticisms of the IPA seem to be mainly based on patent misapprehensions about how its notation is correctly used.

  • In your first impungment of the IPA, you apparently are assuming two different notations of the same sound (in this case, a voiceless labio-velar approximant notated with [ʍ] and [w̥]) are both unanimously agreed upon transcriptions of one phoneme and are used interchangeably. In reality, they're no more than the result of differences in how the phoneme could be more accurately or most obviously denoted. Whichever notation is employed is more an indication of the preference of whomever it was who employed it rather than the IPA's notational ambiguity. It's a trivial objection to raise, nonetheless.

  • To your presumably Anglophone ears, [tʃ] and [t͡ɕ] most likely sound the same. In reality, they correspond to two disparate phonemes. When used by knowledgeable people, [tʃ] invariably denotes the voiceless palato-alveolar affricate, whereas the voiceless alveolo-palatal affricate is signified with [t͡ɕ]. It isn't just a case of one phoneme having different names and transcriptions.

  • There's a subtle articulatory distinction to be considered between a palatalized consonant and a palatal consonant which gives the two their different notations. Palatal consonants have a primary articulation with the tongue either toward or in contact with the mouth's hard palate. Some palatal consonants of import are: [ɲ], [c], and [ɟ]. A palatalized consonant (conventionally denoted with a superscript palatal approximant, as in [pʲ] to denote the palatalized voiceless bilabial plosive) are such that they possess a non-palatal primary articulation coupled with a palatal secondary articulation.

    Therefore, the palatalized phoneme [nʲ] is distinguishable from [ɲ] because of the latter's palatal primary place of articulation and the former's non-palatal primary place of articulation. There's only a soupçon of difference between the two, but it's still a difference that deserves to be accounted for.

  • Different symbols for different phonation isn't a legitimate criticism. Also, there are some phonemes which are given special diacritical marks to individuate their voiced and voiceless counterparts, like with voiceless implosives, for example. This is a complex and nuanced issue and I shan't waste any of my time explaining it here. If you're interested, a Google search will help.

  • 5
    2) My haste lead me to misconstrue your criticism as having the wrong ill-founded premise. The correct ill-founded premise seems to be that [tʃ] deserves one symbol and not two. On closer inspection, one finds that no affricate could possibly be transcribed with one symbol, because every affricate is necessarily–though not fundamentally–a combination of two separate phonemes. Invariably, [t]+[ʃ]=[tʃ]. In light of this, denoting [tʃ] with less than two phonemes cannot be done without a dramatic loss in precision. And the tiebar in [t͡ɕ] isn't a third symbol. It's used to prevent confusion.
    – Rajan S.
    Commented Nov 21, 2012 at 6:22
  • 1
    3) That's not true. The IPA makes perfectly clear the distinctions in notation and articulation of palatal and palatalized consonants. Widespread ignorance of these distinctions are the cause of these disparate types of phonemes' notations being used interchangeably. People's nescience is no fault of the IPA.
    – Rajan S.
    Commented Nov 21, 2012 at 6:32
  • 2
    1) That is what I mean. You can use different symbol or just add a diacritic. The alphabet is redunant. 2) You're wrong. [t] followed by [ʃ] is not an affricate consonant. As I already pointed out in Russian for instance there is clear destinction between affricate [ц] and sequence [тс]. If one wants to denote the affricate one has to write [t͡s]. That is three Unicode symbols for just one consonant. And that is in "phonetic alphabet", which is supposedly designed for better denoting sounds, not in just some weird real spelling. Highly suboptimal.
    – Anixx
    Commented Nov 21, 2012 at 7:00
  • 2
    1) I can only suppose that a phenomenon, wholly accounted for in the IPA, exists to individuate ц from тс. I think that отселить is really [от'селить], with т and с occupying separate syllables, and оцелить is a monosyllabic word. Another, perhaps more plausible, alternative is that the vowel "е" palatalizes the preceding paired consonants "тс", making it [ts­­J]. Because ц is a single (unpaired) grapheme, the palatalization is not conveyed. Therefore, the differentiation is [tsJ] and [ts]. This hypothesis is simply an extrapolation of a general Slavic grammatical rule I've read in a grammar.
    – Rajan S.
    Commented Nov 21, 2012 at 8:42
  • 2
    Sorry. I accidentally deleted the comment you responded to. I concur, actually. The IPA is obviously English-centric. But the assumption that the IPA does not account for other, non-English European languages is counterfactual. Most languages in Europe have the advantage of sharing a close genealogical or geographic propinquity with English, and are therefore readily accounted for within the IPA. Many languages autochthonous to places like Papua New Guinea, sub-Saharan Africa, the Caucacus, Latin America, western Canada, and Southeast Asia are the veritable sufferers of its notation's foibles.
    – Rajan S.
    Commented Nov 21, 2012 at 9:37

The IPA is useful for what it is: an alphabetic writing system biased towards "major European languages" for performing certain analyses/comparisons/descriptions based essentially on the typographical requirements and linguistic models and theory of 100 years ago. As such:

  • it allows you to analyse and denote certain phonetic/phonological phenomena that it is often useful to examine, e.g. cross-linguistic categorisation of sounds, comparison of broad features of the sound inventories of languages, description of broad/high-level pronunciation phenomena for the purpose of language description and teaching...;
  • it operates more or less as any "alphabet with diacritics" will operate, so you may as well use the alphabet + diacritics that are familiar to many people rather than inventing a new one with essentially similar problems (though there are other similar phonetic transcription systems: the "US" transcription scheme is systemically similar to IPA, but not actually identical in the specifics of which symbols are used where);
  • it's become "common currency" in dictionaries and teaching materials in some countries for some languages;
  • it is "common currency" for denoting pronunciations in various linguistic journals and in linguistic publications in some countries;
  • it is a tool to be adapted to specific languages/purposes and extended where necessary-- much of its effectiveness lies in how effectively the linguist is able to apply it to their needs;
  • using IPA with modern typography tools (read: "typing IPA on a computer keyboard") is a complete pain in the arse most of the time.

So, it's useful, but not the be-all and end-all of the universe:

  • many speech technology applications that require transcriptions don't actually use IPA, but rather some other scheme that is easy to type and more closely adapted to the particular language/analysis needs in question;
  • many phonetic and phonological studies nowadays aren't quite so preoccupied with the "armchair segmental analysis" for which IPA is most useful.

So, if IPA isn't useful for what you're doing, nobody's saying you can't use or invent something else that is. On the other hand, if you are in the business of performing the kind of descriptions/analysis that phonetic transcription is useful for and it is for general readership rather than an internal database of an IT application, you may as well use IPA as not: it is still "common currency".


The IPA clearly plays an important role: we need to be able to write pronunciation in an unambiguous way. But maybe we can improve on it.

First of all, IPA looks like a math formula, with all its diacritics, digraphs, and decorations. That makes it intimidating for learners and limits its value in pedagogy and even in public use (like dictionaries or Wikipedia).

Second, reusing the familiar Roman alphabet was an obvious design choice, but in practice it leads to some confusion. IPA letters like c and a are often misread. John Wells tells a story of including /reply/ in a read-aloud exercise, and having most of the students say "reply" (the y is rounded) [Sounds Interesting 3.1, p82].

Among the original design principles of the IPA were to have a letter for each sound, and to avoid diacritics [Handbook of the IPA, p.196]. Now, a century and change later, those principles have been abandoned. There are a LOT of sounds in the world, and the Roman alphabet simply doesn't have enough letters. New letters have been created, but they're not in normal fonts or on normal keyboards. And broad classes of sounds are written with diacritics or digraphs: aspirates, affricates, ejectives, labial-velars, and breathy/murmured sounds.

The vowel model for IPA is an abstract multiplication table of 3 backnesses × 4 heights × ± rounding [ibid 2.6]. One result is a cluster of hard-to-distinguish sounds in the center. Very qualified phoneticists can legitimately disagree on a transcription. Maybe a model based on the acoustics rather then the articulation would have been a better idea.

If we were starting over from scratch, there are some other flaws I'd correct: better symbol names, more consistent use of features, easier letters to write by hand, better integration with tone and intonation. Some day...

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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