Re: Character Identity and Font Selection

From: Stephan Stiller <>
Date: Sun, 12 Jun 2011 22:49:09 -0700

Dear CE Whitehead,

Thanks for your contribution.

> I think most phonetic transcriptions with the IPA are focused on
> transcribing features such as aspiration, labialization, length,
> tone, whether or not continuant, etc. that can be seen as minimal
> features in some language.
For broad/phonemic transcriptions, yes: minimal distinctions are
transcribed, and sometimes more, if it is of interest to phonological
analysis (eg looking at how vowel length is influenced by the following
consonant, even if the resulting length contrast is not distinctive).
For narrow/phonetic transcriptions, people often consider it to be
valuable to be finer-grained. I'm not saying one needs to always be
phonetic/narrow - I do recognize that phonemic transcriptions are very
useful (and I find that something like that, with more detail only on a
case-by-case basis, is the best choice for lexicography).

> I don't think voice quality distinctions between male-female or
> adult-child are normally transcribed using the IPA, although these
> would show up in speech analysis graphs.
Gender distinctions are /never/ transcribed. Okay, I bet it's possible
to find an exotic counterexample with a strongly gender-dependent
language, but normally not. I'd transcribe age-dependent variation only
if it's essentially dialectal, to document diachronic language change.

> Fundamental frequency can differ on the basis of gender/age, or even
> on the basis of language* in some instances, and while pitch contour
> is of interest in IPA transcriptions, fundamental frequency is less
> so I think (except to the extent that it is related to vowel
> quality, but then what's recorded is the contrast -- that is whether
> a vowel is spoken farther to the back of the mouth than another
> vowel uttered by the same speaker -- and not the actual fundamental
> frequency; if fundamental frequency -- apart from contrast -- were
> to become more of interest in transcription, then because I like
> Firth personally, I would hope to see some super-segmental symbols;
> but sound spectograms may be the best solution for recording
> fundamental frequency for now).
I believe you meant to write "Formants can [better: /do/] differ ...".
Fundamental frequency obviously varies with age as well, but fundamental
frequency is only very tenuously related (at least for the languages I
know) to vowel quality. In fact, almost by definition it's not, unless
you have accidental correlation. You do have some interactions between
fundamental frequency and other factors, such as for tonal genesis (eg
if you want to investigate how the Sinitic languages became tonal), but
that's all I've ever heard of in that regard.

> Note for anyone interested: apparently vowel quality -- related to
> fundamental frequency -- distinguishes infant speech on the basis of
> language earlier than other features do. Different features of
> vowel quality may be transcribed phonetically/phonemically,
> depending on the purpose of the transcription; there's often more
> interest in indicating contrasts, but not always; you could for
> example indicate that all the vowels in a particular language were
> breathy if such were the case, and then breathiness would not be in
> contrast to anything else in the language. For those interested
> in fundamental frequency and early vowel quality
Same as above: vowel quality is most likely not related to fundamental
frequency. (Interestingly if you point out to some monolingual speakers
of General American English that [ɛ͡ɪ] ("bay") is a diphthong, some of
them may respond in confusion that it's actually pitch contour
variation, but that's never the defining factor for vowel identity in
this language - it's independently overlaid, in general.) I think you
meant to write "formants, which define vowel quality ..." (simplifying a
tad here) "... interestingly are age-dependent". What's interesting
about this is that it seems to overthrow the whole theory of the first
two formants F1 and F2 "defining" the vowel space. And I've never seen a
good, conclusive analysis on how exactly the vowel space is transformed
for children.

> In any case, I personally would not think that having a new system
> would suddenly change the features linguists transcribe. (The
> current system works fine IMO if you note what features you are
> transcribing and what not and make sure you define symbols as needed.)
I was gonna end the discussion on canIPA or take it offline, but I'd
like to respond to your statement: canIPA's main merits are /not/ that
it gives you somewhat finer-grained resolution for the vowel space or
other features one may want to transcribe. In fact, compared to IPA,
canIPA gives you (informally speaking) perhaps half an order of
magnitude more accuracy (or even a bit less) compared to IPA - and
that's /not/ a lot. (This is also why I don't buy an argument that
people wouldn't be able to use it in a productive and meaningful way,
because we know that people differ quite a bit in their auditory
perception and production capabilities.) And we know that canIPA isn't
just some new age funny-diagrams thing, because the dimensions on which
it is modeling language are all generally well-established phonetic
dimensions. Imho, canIPA's main merit is that it fixes a bunch of
historical IPA flaws (discussions in JIPA are generally insightful, if
you're interested). IPA doesn't fundamentally change these days, just
like any standard (like Unicode) aims to preserve meaningful backwards
compatibility, and for a good reason at that. I'm trying to say this
without value judgment. Still, Canepari fixed a lot of legacy issues all
in one go, and his treatment of phonetics is more systematic and
/extremely/ insightful (eg how he systematically makes very clear
distinctions between phonemic transcriptions (which he recognizes are
useful) and all the allophones they appear in, in his language
descriptions // I don't have his book with me, but his writings seemed
insightful in other areas as well, as he said some interesting stuff
about prosody for example) - it's a fresh start, and he's obviously a
smart guy. Because his observations match mine, I like the system (it's
not a secret, and it's obvious from what I'm writing), but note that I
have /not/ proposed that anybody encode it now, because I know that such
a proposal wouldn't be successful. I mentioned canIPA merely as the only
possibility I could ever see the IPA block in Unicode to grow. IPA's
core won't grow much.

Received on Mon Jun 13 2011 - 00:52:08 CDT

This archive was generated by hypermail 2.2.0 : Mon Jun 13 2011 - 00:52:09 CDT