buffer vowel, misc replies

[Logical Language Group <lojbab@ACCESS.DIGEX.NET>]

>From: John Cowan <cowan@LOCKE.CCIL.ORG>
>Reluctantly, I have to weigh in on And's side here, after much mental
>muddling while writing the phonology paper.  The phonology is broken in
>the area of buffer vowels.
>
>The principal trouble is that buffering is done for the speaker's
>benefit (unlike the pronunciation of "'" as [T] (theta) which is
>supposed to be for the listener's benefit), but if the listener has
>learned an unbuffered dialect, he will not have space in his Lojban
>vowel map for the speaker's buffer, and will stuff it into one of the
>other slots ("i" or "y" or whatever), causing mishearing.

I don't agree.  If this was the case then we as native English speakers
would not be able to adapt to hearing other dialects of English that
have somewhat different vowel mappings, and more relevant - we could not
adapt to foreign speakers of buffered dialects.  The fact that I can
interpret most British English correctly in spite of what they do to
"er" solidifies the argument for me.

As long as the speaker uses a consistent buffer, which is maximally
distinct from other vowel realizations that the speaker uses, then the
listener has a very good chance of noting and/or filtering out the
buffer, sometimes subconsciously (as I usually filter out my wife's
odd-to-me vowel distinctions).  If you add relative shortness as a
feature, this should become an "easy" task.

>Unfortunately, I fear that although the phonology is broken, it may be
>too late to fix it.  Either we have to define the buffer as a definite
>though optional vowel, or we have to revise the morphology to allow
>buffer-hyphen equivalence (meaning that "patfymamta" and "patfmamta"
>mean the same thing -- currently the latter is a dubious le'avla).  This
>is scary, although it might work if it turns out that all le'avla that
>are banned either fail the slinku'i test or involve nonstandard medial
>consonant triples like "tfm" which is "t/f/m".

Yuck to both.

And:
>While the vowels we produce do not cover the whole space, any vowel
>that we hear will be classed as one of the phonemes.

I am skeptical of this.  In English, when we hear a phone that actually
occurs in English as part of English words, we do not classify that
phone as an allophone/realization of the other phoneme.  If I hear a
truly odd sound, like say /eu/ diphthong, a voiceless vowel whistle,
etc., I do NOT map it to a familiar phoneme.

Similarly, I understand that Japanese, who make no l/r distinction, do
NOT have trouble distinguishing the two sounds when English speakers use
them.  I also CAN distinguish between my wife's vowels for "Mary",
"merry", and "marry" (not to mention the verbal and tin "can"s), even
though I cannot reliably produce the distinctions myself.  Japanese do
not easily produce the l/r distinction as English speakers would do it,
but they do produce two sounds that we do recognize as l and r and can
tell apart.  I had no trouble recognizing o and u umlauts as distinct
sounds from any English vowel when first I started German classes, EVEN
when the sounds were produced improperly by other students.  They were
still distinctly different from those students' other realizations of
English or pseudo-German vowel phonemes, and I was prepared mentally for
the existence of extra unfamiliar phonemes that would map to the umlaut
vowels, so thus I heard them.

The Lojbanist should be prepared for exactly one phoneme that is clearly
distinct from all other Lojban vowel phonemes, though it might be
realized in different ways.  Some identifiable properties include
relative shortness (even shorter than an unstressed slack vowel), and
usually some form of centrality, but contrast is also important.

>I use a front [a] for Lojban /a/, and never a back [A], so in principle
>that would leave me free to use [A] as buffer.  But you would hear my
>buffer as /a/.

In principle, you can do a LOT of things.  What you actually CHOOSE to
do should depend on your audience as well as your own speech production
flexibility.  I am not SURE that I couldn't tell which vowel you were
doing, provided that you were consistent as to doing one or the other in
the proper places.

Cowan notes that the use of a buffer is one where we are catering to the
needs of the speaker rather than the listener - not the typical case for
Lojban.  But this is one case where I think there is a demonstrable
ability for listeners to accomodate speakers, provided that the speakers
themselves are consistent in their usage.

>Where does this "for English speakers" come in?  What counts is what
>Lojban speakers do.  Which sounds will a Lojban speaker hear as the
>buffer vowel?  Put another way, which sounds may be used for the buffer
>vowel?

Speakers of dialect Y of language X have a certain set of phonemes.
When listening to another dialect of X, they will map the realizations
of various speakers' vowels to some of those phonemes.  Exactly what
sounds they map will depend on the speaker's dialect, and thus will
differ from speaker to speaker.

Few English speakers produce a short high central lax semi-rounded vowel
as a realization of any English or Lojban vowel.  Thus that sound is a
likely candidate for English speakers to use as a buffer.  Russians
might prefer something a little different, since they have a high
central vowel that occurs in alternation to /i/ - maybe the lax rounded
vowel of "look", which will likely to be strongly contrasted with their
realization of Lojban u which wil probably be higher and more tense as
is the Russian 'u'.

>What matters for our present purposes is principle, not practice.

I disagree, because the purpose of principle in making a prescription is
to affect practice.

>  In
>practice the rules are flouted left right and centre.

If your physician issues a prescription, s/he likewise intends that it
be followed.  And unless you are far better than I am, you will be
unlikely to take your pills precisely 6 hours apart and at least 1 hour
after a meal.  But the physician still writes the prescription, not
because of principle, but because the act of specifying the times that
you take your medicine will indeed constrain your behavior, if not
perfectly.

Chris then says:
>Why do we allow for buffering in the first place?  Certainly there are
>other places we allow Lojban to throw difficult things at people (using
>both L and R, J and DJ, C and TC and S, h and X...)  A language that
>really tried for universally easy phonology would be better off
>borrowing from Hawai'ian or something.

It is NOT a question of "allowing" buffering.  Buffering WILL happen,
more or less, whether or not we want it to.  Different people will
buffer to different degrees both because of vocal tract differences and
because some languages have more or less onset times for start and
termination of the sound.

The reason for DEFINING the buffer is that if we do NOT say anything,
then we can be accused of ignoring a real phenomenon in language - that
people don't always pronounce phoneme strings ideally, and that we had
not dealt with the ambiguity that would result if, say, we were to use a
schwa as a buffer vowel as well as for hyphens.  That emphatically means
that we do NOT want schwa used as a buffer.  So by specifying SOME rules
about the buffer, we are in effect prescribing the first line failure
mode for the phonology, so that it degrades gracefully under pressure.

lojbab