Re: Character Identity and Font Selection

From: Stephan Stiller <>
Date: Sat, 11 Jun 2011 14:03:48 -0700

> Yeah, right. Just like wine tasters. The accuracy is largely
> self-delusion - controlled experiments show that wine professionals
> can't reliably identify wines, or even reliably distinguish red from
> white.
Just like piano touch, which has only ever been shown to be
distinguishable if the finger-to-key attack noises are included. But
that's not a surprise because the dimensions on which to distinguish
have never been specified or are obviously culturally different (for
piano touch or dimensions of wine taste) and commonly described with
creative poetic adjectives "gone mad" ...

> All my LabPhon colleagues think that detailed transcription is bogus;
> I've never seen any controlled experiments demonstrating reliable
> transcription at fine levels. If you have any, I'd love to see them.
... but you know that this isn't how people describe the phonetic space.
But we know (kinda) what the dimensions of the vowel space are. We may
not know what exactly ATR is, we don't think formants are /exactly/ like
vowel height and frontness, and we also don't know how the vowel space
is actually transformed for children - but we know a lot about phonetic
dimensions. And it depends on what you mean by "detailed". (1) Note that
some IPA symbols are of limited meaning. I hear they have both
"mid-centralized" and "centralized" just because the committee couldn't
agree on which to pick in (?)1990. (2) If you include all combinations
of IPA diacritics, then - yes - I'll agree with you. (3) If you have
reference recordings and (say) an applet to aid you in transcription,
then, conceptually, it of course ought to be possible. canIPA is his own
reference system for doing things, and - I can see you "really wouldn't
like it" - but a paper transcription has the advantage over a recording
that it's aiming to be speaker-independent. To distinguish (say) all
"cardinal" canIPA vowels in isolation with long-enough reference
recordings is possible, because we know that the physical dimensions are
real. For example if I want to fine-tune my pronunciation, I may want to
transcribe native speakers on a fine-grained level. His density of
distinctions in the vowel space quite closely correlates with what I can
perceive. And you know that some people are natively better at this sort
of thing than others.

>>> Even with years of ear training, phonetic transcribers often don't agree at
>>> the level of detail the IPA alone offers.
>> But part of the reason is the fact that reference recordings of a few
>> notable phoneticians have historically differed.
> True. Cardinal [a] being particularly problematic!
Yeah, they messed up with the a-like symbols down there for the low
vowels ...

>> Okay, seems like we have a factual disagreement. The following article
>> should address this: Holger Schmitt, JIPA 37:3 (2007). I doubt everyone
>> on this list in interested in the details, but, briefly, Schmitt
>> described the current (somewhat messy) state of affairs and proposes to
>> unify to the open e symbol.
> Yes. What he says is that there is no standard practice - in his
> survey of works, he finds a slight majority for<e>, but as he doesn't
> list the works consulted, one can't see how he sampled. My personal
> impression is for a majority of<ɛ> in phonetically oriented works,
> but that's just an impression.
Sure, let's weaken both of our statements a bit, and it seems pointless
to go further here.

But the larger point was that some conventions are culturally arbitrary
and aren't questioned by people (esp. if they can't make the necessary
distinctions). Why do many British transcriptions use vowel length
indications, while American ones rarely do? (And if the British do, they
don't normally apply it to [æ] - why?) If I hear "tapped" r in American
English as closer to a [d] than to a Spanish r-tap (and noticeably
distinct from it), why shouldn't we use a d-derived symbol or symbol
cluster then? I have seen phonologists mess up their analyses based on
taking broad transcriptions too literally. I guess it's about what you
use IPA for. That a detailed system can't be learned I can't
conceptually agree with. If you are saying that it's inaccessible to
many and that the effort may not be worth it (given how easy it is to
make a recording), that's a different statement.

>>> This is, in any case, influenced by your view on whether quality or quantity or both is the primary
>>> differentiator between tense and lax vowels in English.
>> Both. And the tense one is a diphthong (or "diphthongized", if you
>> like). I guess it's a good example of how perception can differ ...
> The tense are not always diphthongized. In much of northern England,
> the FACE vowel is a pure [e:], and GOAT is [o:].
Yes, yes. I was referring to the American practice of transcribing them
as monophthong, even though they're clearly perceived as diphthongized
by any non-native speaker, while in the UK people generally use diphthongs.

> Since you mentioned Maddieson, I have an ambition of contributing to
> Unicode: Maddieson was intrigued (at least at 3am after a night's
> drinking) by my /r/ sound, and as far as I can see, there's no real
> symbol for it, though I don't actually think it's an extremely unusual
> British variant. Maybe one day I will propose a symbol for it!
I take this as a joke - but either way, you'd have to prove that it's
either useful for a handful of language, or perhaps popularize it in a
system that people take up.

Received on Sat Jun 11 2011 - 16:08:33 CDT

This archive was generated by hypermail 2.2.0 : Sat Jun 11 2011 - 16:08:35 CDT