Re: Optimum number of symbols

Mike S:
> On Sat, 25 May 2002 15:42:41 +0100, And Rosta <a-rosta@...> wrote:
> >John Cowan:
> >> And Rosta scripsit:
> >>
> >> > I don't think an alphabet lends itself in a trivially easy way to
> >> > a language with lots of lexically contrastive suprasegmental features
> >> > such as tone, nasalization and voice quality. Well -- the result may
> >> > be trivially easy, but the number of characters needed is
> >> > unsatisfactorily large. (Cf. the numberless threads on this
> >> > list about romanizations of Chinese.)
> >>
> >> I think that results primarily from a prejudice felt by Latin alphabet
> >> users that going past the Big 26 (or 27 at most) is unacceptable.
> >> Cyrillic, as Ivan pointed out, is much more willing to accept novel
> >> characters as needed by newly written languages.
> >
> >I don't think so. Rather, if you have a very large set of putative
> >segmental phonemes that are systematically and transparently
> >derived from a combination of a smaller set of features, a strict
> >alphabetic approach obscures that underlying phonological system and
> >requires an unnecessarily large inventory of symbols.
>
> I am not quite sure what you are proposing here.  Are you suggesting
> that we build characters from place, manner, voice, etc.?
Where 'appropriate'. For many phoneme inventories it would not
be appropriate; the phonemes of English don't neatly decompose
into voice/place/manner, for instance, in the sense that not all
combinations of v/p/m yield phonemes, so a decompositional
approach would require an additional set of rules defining the
set of permitted combinations of features.

However, imagine a consonant inventory where for every place of
articulation you have voiceless fricative, voiced fricative,
nasal, breathy voiced stop, aspirated stop, unaspirated voiceless
stop, plain voiced stop, ejective, and each of these can be
palatalized. For each place of artic, that gives 8 * 2 = 16
phonemes. I would find a featural script far more satisfactory
a written representation of such a language. That is, a
script based on primitives for place, manner and palatalization
would be more satisfactory than one with primitives for each
phoneme, not because of sheer number of primitives but rather
because the one better captures the underlying system to the
phonology.

> Or are you
> proposing a more complex system in which a cluster like /mb/ is
> marked only once as +bilabial?  I believe you suggested that in
> another post.
That wasn't what I meant here.

--And.