Re: Character Identity and Font Selection

From: Stephan Stiller <>
Date: Fri, 10 Jun 2011 18:54:11 -0700

I agree: in my experience, most linguists casually use IPA in a phonemic
manner, exactly in the way Philippe Verdy describes. Italian phonetician
Luciano Canepari argues that the IPA is used phonemically (not
phonetically) by most linguists. Some people emphasize a distinction
between "broad" (essentially phonemic) and "narrow" (more phonetic)
transcriptions, but most transcriptions are broad. (Interestingly,
Canepari invented his canIPA as a derivative of IPA more suitable for
very narrow transcriptions than even IPA with all its diacritics.)

The thing is that, fwiw, often cultural conventions trump accuracy in
picking symbols for such broad transcriptions. For example, [ə] is used
for the default reduced vowel in many language. (On that note, Korean
linguists tend to transcribe ㅓ [ɔ] as [ə] for really no good reason at
all, as it's not actually used a reduced vowel in Korean.) Linguists
have argued that standard British practice of transcribing the vowel in
"bet" as [e] is really outdated ([ɛ] is more appropriate). Linguist Ian
Maddieson has argued (personal communication) that [ʌ] is not the best
choice for the vowel in "but". American linguists tend to omit
diphthongization and vowel length [:] for vowels. Most linguists pick
symbols from the "most commonly used" IPA symbols (such as [r] instead
of the many r-derived symbols), with the rarer used ones appearing
mainly in JIPA.


On 6/10/2011 10:49 AM, Philippe Verdy wrote:
> 2011/6/10 ejp10<> wrote:
>> The phonemic level means that not all phonetic detail is recorded. It is common practice in the U.S. to transcribe American [ɹ] as /r/ UNLESS it is critical to record the phonetic details of said /r/
> Very true: we have to make distinctions between phonetic
> transcriptions (that should be language neutral), and phonologic
> transcriptions, even if they both use the IPA symbol set. Each
> language actually does not use the full set of IPA symbols but
> transcribes in its phonology all allophones using the same symbol,
> which is normally the same used for one the possible phonetic
> realizations of the same phoneme.
> In other words, if you have to tag a phonologic transcription, you
> MUST use the language code before the "-fonipa" transcription.
> Otherwise, the notation makes no sense at all (and not for example
> when producing automated aural rendering from this transcription).
> This is not true for pure phonetic transcriptions, for which the
> indication of the language should NOT be needed for correct rendering
> by some automated aural converter.
> Subtle phonetic transformations that occur in actual languages are
> generally not noted, as they are implicit for a specific accent of
> that language. The same phonologic transcription of the same language,
> should be usable to produce the various regional/societal accents of
> that language, that have EACH one distinct phonetic transcriptions
> using a much more complete and complex transcription system (this is
> highly difficult to standardize phonetic transcriptions, even between
> native speakers of the same language in the same region, as this also
> depends if you're a child or adult, or your male/female, or between
> levels of speech and attention, or speed of speech and who you're
> speaking to and how you want to focus your auditory, or if you're
> singing, playing theater, or if you are doing jokes, or speaking loud
> or very light...).
> An automated aural processor will then have to consider the language,
> and some other parameters, in order to convert the simplified
> phonologic notation into an actual phonetic transcription (using some
> complex rules) in order to render the actual speech.
> It is then perfectly valid to use, say, /r/ in English to denote all
> possible phonetic realizations, even in US English.
> French has a common standard for using /ʁ/ in its phonologic
> transcription, biut this is still a conventional notation, and here
> also you may use /r/ as well. If you need to make a clear distinction
> within a specific accent, to designate the actual phonetic realization
> where phonemes may NOT be replaced, you need to place the notation
> between [square brackets] instead of /slashes/. And an automated aural
> processor (such as an aural screen reader) should be able to detect
> the bracket/slash distinction to avoid some generation/transformations
> of allophones, and generate the wanted accents (but even in that case,
> the processor will have to add some other transforms, notably when
> creating subtle transitions to get a smooth language that can be
> understood and that is reasonnably pronounced by native speakers,
> because not all combinations of phonetic realizations of allophones
> are even possible in every position of the pronounced sentences).
Received on Fri Jun 10 2011 - 20:56:46 CDT

This archive was generated by hypermail 2.2.0 : Fri Jun 10 2011 - 20:56:46 CDT