allophones of zero

[Logical Language Group <cbmvax!uunet!GREBYN.COM!cuvmb.cc.columbia.edu!lojbab>]

By way of history, the choice of hyphen, and the development of hyphen/buffer
involved a fair amount of testing, allbeit only with English speakers.
The original favored approach involved using vocalic 'r' as the hyphen, but
I believe that a significant minority thought this ugly especially in print.
The 'r' approach WAS retained after CVV rafsi at beginnings, and I used it later
in le'avla systmeatization.  The vocalic 'r' to some people
is the closest vowel sound to [y], of course.

Instead people chose y=@ as preferred.  This caused other problems, including
the fact that JCB at that time planned to use @ as the buffer.  Basically, all
the people who wanted to speak the language overruled 'simple' design.  JCB
 tried
and failed to satisfy people using the @ for both hypen and buffer.  I was one
of the opponents, and in the only contribution to the Loglan that JCB acknowl-
ledges, I proposed that speakers of buffered dialects would use 'iy' (yuh)
for the hyphen, and speakers of unbuffered dialects would use 'uh'.  I believe
that remains TLI design to the present.

On the other hand, at the gathering where we first put together the phonology
for Lojban, and at LogFests before and after, someone, I believe it was Jack
Waugh, convinced me-and-others that there would be a lot of people that would
speak half-buffoered dialects:  they would use buffers between some small number
of cluster (in Jack's case ml and mr initial, and paired stops as medials)
but unbuffered otherwise.  It was also easily demonstarted that such hyphenation
was close to subliminal, but that the lax sound in such clusters was so short a
as to be unidentifiable, even though Jack thought he was saying an @ when he
concentrated.

Since at the same time, we also realized that lujvo with hyphenation was going
to be much more common than conscious buffering, we designed Lojban to
de-emphasize the ambiguity question adn recognize actual usage - hyphenation
was deemed more important than buffering and the buffer was decided to be
some sound that would be ignored by the listener.  This can therefore either
be a sound whose length is so short that it is not deemed phonemic (as Jack's
usually was) and hence the listener can't even if concentrating, tell what
sound it is, or a sound that the speaker explicitly used and chose to be out
of normal phoneme space.  Experiments with'real listeners' showed that
a short sound like the 'i' of "bit" was successful with English speakers, but
we did not want to make this mandatory, since some others might find it hard
to distinguish.  We considered [y], but found it too hard to say as English
speakers (not that we buffered that much), but on the other hand we figured that
we could use it well-enough if talking to a listener who couldn't tell what
we were saying using an 'I' buffer.

We haven't talked about it, but it is plausible that we could accept the
reverse of the JCB plan and use [iy] diphthong as a buffer.  H,mmm.  Actually,
I think we did - it sounds hilariously strange.  As strange as the guy who,
back when we defined the apostrophe as any non-Lojban unvoiced consonant, chose
'th' instead of 'h' making everything he said sound like a lisp.  The
experimentation with live conversation showed that the current approach was most
viable, and easiest to do instinctively, but that we could, faced with a
listener who could not distinguish our speech clearly, make alterations and
with conscious emphasis on the buffer sound, make it virtually any sound nbeeded
to make it disntict to the listener.  Hence the current design 5that it merely
be different from the base sounds.  I have continued to find that the barred I
is most teachable, and we have real conversation that does not hinge on noticing
buffers.  Only when we get a speaker with a real problem saying or listeneing
to Lojban speech with the current design, will we have a conclusive test.
There is no sentiment among the design team to even consider a chnage until
such a speaker shows up and is interacted with for a while.  The strong
sentiment in the Lojban community for stability in baselines makes any
discussion of phonology changes beyond stylistic ones in how we teach the
language academic.

lojbab