Re: Optimum number of symbols

Raymond Brown <ray.brown@...> wrote:

>At 6:28 pm -0400 22/5/02, Mike S. wrote:
>>My main statement comes in a moment, but I want to inject here that
>>I disagree with your assessment that alphabetic scripts have never
>>been phonemic until the creation of some last century.  IMO, written
>>ancient Greek and Latin both fall solidly into the category of
>>phonemic scripts, the occasional quirk or irregularity notwithstanding.
>
>If 'twas just an occasional quirk, then the pronunciation of ancient Greek
>would be clear and non-contoversal.  It ain't either.
Actually, I feel that it wouldn't.  Phonemic scripts are essentially
not intended for, designed for, or capable of representing that level
of phonetic information.  It is very rare that any standard script would
do that, and furthermore, unnecessary.  Phonemic scripts only need to
indicate the *contrastive* units of speech.

If, on the other hand, you are pointing out that phonemic scripts
can be compromised by dialectal and diachronic changes (as was rampant
in Greek and later in Latin), then I have to say I think you are
demanding more than can be expected from any script, at least a non-
morphemic one.  And while a morphemic script such as a logogram might
arguably handle dialectal and diachronic changes better, such scripts
carry well-known costs which cast their overall desirability into
serious doubt.

Alternately, if you are pointing out that phonemic contrastiveness
is lost because of, for example, lack of marked vowel length, then
you are right.  Insofar as a script fails to contrast phonemes it
is less than phonemic.  However, I do hasten to add that, as I am sure
you know, many, if not the vast majority, of vowel ambiguities in Greek
and Latin were in practice pretty easily solved by context.  I am sure
problems did arise from time to time, but not often, I would take it.



>It's true we have a better understanding of Latin, but it's still not
phonemic.
>
>At 8:28 pm -0400 22/5/02, Nik Taylor wrote:
>>
>>There were several contrasts not captured by the script.  For example,
>>the Romans rarely marked long and short vowels, even tho it had the
>>possibility, and Greek had no way of indicating certain vowel contrasts
>>like (IIRC) /a/ and /a:/
>
>The Romans did occasionally use things like writing the vowel twice, or
>puting an 'apex' (mark rather like an accute accent) over the vowel; but
>neither became generally adopted.  In normal Latin spelling the phonemic
>contrast in vowel length simply is not marked.
>
>To Nik at 10:38 pm -0400 22/5/02, Mike S. repiled:
>[snip]
>>
>>That was the one biggie.  In Latin, semivowels and <Q> if you want
>>to get picky.  Other than that, highly phonemic.
>
>I don't understand why the lack of distinction between long & short vowels
>is described as 'the one biggie' and the lack of semivowels is dismissed as
>getting picky.  In Classical Latin [j] and [w] simply were *not* allophones
>of /i/ and /u/, they had separate phonemic status.  It is true that, for
>the most part, it can be predicted - when you get used to it - when {i} =
>/j/.  But that is simply not true of {u}.   The Romans themselves felt this
>to be awkward, tho they never got around to sorting it out.  But it is
>significant, I think, that one Claudius' extra letters was an inverted F to
>represent /w/.
I agree this situation was not ideal.  The Roman alphabet has always
been remarkably resistant to new characters, even 2000 years ago.



>Just to add to the fun, {i} in some words represented _two_ phonemes /j/ +
>/i/, e.g.
>adicere /ad'jikere/
>conicere  /kon'jikere/
>etc.
>As for Q, that was an attempt to distinguish between /u/ and /w/ in a
>context where the confusion would otherwise have been very problematic,
>namely:
>cu = /ku/
>qu = /kw/
That's correct.  They had an extra character to use.  Clunky but a
minor improvement nonetheless.



>So, e.g. example we can distinguish _cui_ /kuj/ from _qui_ /kwi:/.  But
>it's mighty odd way of going about things, and it does nothing to help
>other instances of {u}.  How can one tell from the spelling that, e.g.
>_soluo_ is disyllabic /'solwo:/ while _solui_ is trisyllabic /'solui:/ ?
>One cannot.
>
>I must disagree; I cannot count as _phonemic_ an alphabet where one symbol
>can represent three different phonemes:
>{u} = /u/ _or_ /u:/ _or_ /w/
>
>and another symbol may represent one of three phonemes or a group of two
>phonemes, i.e.
>{i} = /i/ _or_ /i:/ _or_ /j/ _or_ /jj/ _or /ji/
You have taken Latin to task for its vowels and semivowels.  Let's now
examine the consonants.  Prior to fricativization of semivowels and
palatalization of velars, Latin had the following consonants:

p b t d c g f s m n r l h

There were a couple "allographs": as you mentioned, <c> was replaced
with <q> to mark <u> as /w/; <k> replaced <c>, but very rarely;
<x> stood for <ks>.  Other than that, these were extremely well-behaved
consonant graphemes.  Even the quirks are totally predictable when
rendering their pronunciation.  Consonants are generally recognized
as more salient than vowels;  in any script it's more important
to get consonants right than to get vowels right.  Latin pretty
much gets all the consonants right.

My guess on the vowels and semivowels is that surely both the Greeks
and Romans noticed and perhaps even resented the occasional problem
that polyphonous characters cause, but like us with our English,
were never so affected by it as to actually feel the need to reform
the system.  Bear in mind, as I have mentioned, that for a native
Roman, context would have solved most of the ambiguity.



>I still maintain that early alphabets 'tended towards a phonemic
>represntation' or, if you prefer R.Y. Chao's words:
>"An _alphabetic writing system_ is one in which each symbol corresponds
>more or less closely to the phonemes of the language."
Okay.  Saying "tended towards a phonemic represntation" and "more
or less closely" are fair ways to look at it.

In your last post, however, you said "alphabetic scripts have never
been phonemic until the creation of some last century".  That
statement strikes me as an overstatement.  If that was not your
intention I stand corrected.

As for Latin at least, with minor trivial quirks mentioned, every Latin
consonant phoneme corresponds with a grapheme; and the vowel graphemes
fail to contrast for length; the semivowels and highvowels are not
contrasted.  That's about it.

It's not a perfectly phonemic, but it is, as Chao put it, "more
or less closely" phonemic.



>>The basic and universal utility of the phonemic principle has been
>>in effect--indeed proven IMO--at least since the development of the
>>Greek alphabet with vowel letters, even if the Romans and Greeks did
>>not have the same precise modern understanding of "phoneme" we have.
>
>They most certainly did not.  Indeed, in view of different definitions of
>'phoneme' given by, e.g. the Prague School, by the British phonetician
>Daniel Jones, and by American linguists of the 1940s, and remembering the
>questioning of the phonemic principle recently on this list and And's
>outright rejection of the phonemic theory (and he is not alone in this), I
>don't think we can say the modern understanding is 'precise'.
I do not believe it's necessary to place stock in the theoretical
validity of the notion of the phoneme in order to recognize the
universal practicality of the alphabet.  The phoneme is in dispute
in some quarters, and quibbles are made in others.  If you prefer,
you could interpret the alphabetic grapheme as representing the
contrastive segments in which phonotactic position is regarded
as a feature (and each such position has an associated set of
contrastive segments), as in And's perspective (assuming I grasp it).
Furthermore, the fact that the same grapheme is assigned to
contrastive segments of different sets without distinguishing that
feature is not a fault because that can always be regarded as marked
by the relative positions of the consonant and vowel graphemes.
If it is objected that "woolly criteria" have been applied to
establishing the graphical identity between members of the contastive
segment sets, then it could simply be maintained that no claim
of theoretical identity is being made, *even* while the principle
of mnemonic reuse of graphemes is put into practice.

In other words, in my view, you would have to go much, *much*
farther than merely questioning or rejecting the notion of phoneme
to cast the universal practicality of the alphabet into doubt.
You would have to disprove the very existence of contrastive
segments in at least one human language.  I can't remotely
conceive what such a language would sound like or how it would
be written, and I very strongly doubt such a language exists,
has existed, or ever will exist.



>I have not denied the utility of the alphabet, otherwise the Greeks
>would've carried on with their ancient syllabary.  The utility of the
>alphabet is seen with the widespread adoption of the Roman alphabet in the
>last couple of centuries.
>
>But that the utility of of the alphabetic principal is _universal_ is not
>one that I readily accept.  China has been exposed to the Roman alphabet
>since the early 17th century - indeed, Matteo Ricci's system of
>Romanization of 1605 was but the first of a whole series of attempts at
>alphabeticization.  Now, if the 'phonemic principal' were really of
>universal untilty, it seems strange to me that almost 400 years later the
>main system for writing Chinese is logographic.  One must conclude, it
>seems to me, that, despite it disadvantages, it is well suited to represent
>the Chinese language and that the advantages of an alphabet do not outweigh
>the perceived advantages of the traditional script.
I should mention that I have refined my position in the last two days,
so I hope you don't feel like you're having to shoot at a moving target.

What I mean precisely by the "universality" or "universal practicality"
and similar terms is that because all human speech can be analyzed
in contrastive segments, *all* human languages lend themselves to the
alphabetic system of writing in a *trivially easy manner*.  This is
the special attribute that alphabets have that no other system of
writing has.  For example, a language with a simple syllable structure
may be a good candidate for a syllabary, but a language with a complex
one is not.  Likewise, a language with an isolating syntax may be
a good candidate for a morphemic system (assuming easy learnability
is not too important), but a highly inflected language with alternating
stems probably would not be.  Again, the claim here is relative easy
for *all* languages.  Only the alphabet fits in this category.

What I do *not* claim for the alphabet is both "universal optimality"
and "universal desirability".  These are totally different claims.
Because universal practicality is not the same as universal optimality
or universal desirability, the fact that the Chinese elected not to
adopt an alphabetic script does not have any direct bearing on what I
do claim.  Chinese scholars were happy with their system, so they kept
it.  That's not surprising; we've mentioned historical momentum, of
English orthography for example, in several posts.   The point is,
if the Chinese *had* elected to adopt an alphabet, it *would* have worked.



>On the other hand, Arabic numerals have been adopted universally in Europe
>and, indeed, now most of the world (either in their westernized or native
>Arabic form) because they were perceived to be universally advantageous.
Indeed they are advantageous.  Another great *advancement* in history.
Or do you maintain that Roman numerals are just as good as the Arabic?
After all, they *were* in use for *centuries* and, as you have insisted,
"nothing is superior to anything else".





>1. The utility of applying the phonemic principle to every language is not
>proven
>2. No alphabet, before some created in the 20th cent., were in any case
>truly phonemic
>3. Alphabetic systems tend towards phonemic representation; some are closer
>than others (for all sorts of reasons)
>
>Therefore, I contend that an alphabet is not necessarily the most suited
>way of representing every language; other methods may be equally valid or,
>indeed, better for a particular language.
My replies:

(1) I've already stated my position clearly on the alphabetic/
phonemic principle.  I refer you to my post to Christophe yesterday
for additional reference.

(2) True.  It's a matter of degree.  I find it difficult to
minimize the strong if not predominant phonemic characteristics
in Latin and Greek script the way you do.  I'm willing to accept
that that's your perception though.

(3) I agree.

The alphabet can be used to provide the expected "floor" of acceptable
performance though.  I agree that other systems can be optimal,
based on one's criteria.



>>It's true that these scripts, which might be called proto-alphabetic,
>>indeed marked consonant phonemes.  As I understand it--I might be
>>wrong--most of these languages (unlike Greek) had only a few vowel
>>phonemes, and thus could more easily afford not marking them.
>
>No - it's that in Semitic languages vowel patterns are largely predictable.
Okay.



>>I would
>>have to think though, that although these scripts were basically
>>adequate, problems would occasionally arise in any script with such
>>an array of polyphonous symbols.
>
>Apparently not.  It was only when certain texts were held to be sacred and,
>in the case of Hebrew, the language was moribund as a spoken language or,
>in the case of Arabic, the language starting developing different spoken
>dialects (like modern English is doing), that a need was felt to mark the
>vowels so that the sacred texts would be read correctly.  Even now, Arabic
>is normally written without vowel marks, the latter being used almost only
>in texts of the Koran. The revived modern Hebrew in Israel is not normally
>written with vowel marks.
>
>>Everything else being equal, wouldn't
>>you have to agree that such a script is improved by marking vowels?
>
>It depends on the language, I think.  It would be interesting to learn the
>observation of our Hebrew speakers on this list (and, indeed, Arabic
>speakers if we have any).
I suppose it's tautological to point out that a language that never
relies on vowel contrast (if true) to distinguish words does not
need vowel markings.  That does seem pretty unusual though.



>>It is clear to me that scripts arose through a centuries-long
>>evolutionary process, in which enhancements were gradually added.
>
>Development, I go along with, but 'evolution' rings alarm bells.
Okay.  I won't say "evolution".



>>The development of consonant-only syllabary (if I may call it that)
>>represented a major step in this process;
>
>Essentially an alphabet - and long anticipated in Egyptian writing.
We can call it that.  I'd prefer "proto-alphabet" though.



>>the Greek innovation
>>to add vowel letters to their alphabet represented another.
>
>..and the early Greeks didn't even realize they were doing it  :)
>It was lucky happenstance.
Really?  When they built the acropolis, did they not realize they
were doing that either?  Interesting people if so.



>>Both these cases, IMO, marked a definite objective improvement
>>over the older systems.  This is not to do not deny the fact
>>that the older systems were adequate;
>
>When systems like the Egyptian lasted about four millennia and the Chinese
>system is still going strong after three and a half millennia, I'd say
>describibg them as "adequate" is being a bit patronizing.
Ah yes, we wouldn't want to patronize ancient civilizations, would we.



>>I do claim however
>>it is possible to make relative utilitarian judgements
>>in certain areas.
>
>Yep - In theory, the Shavian alphabet should surely be more useful - it
>takes up less space for a start, is far more regular etc etc.  Indeed, Shaw
>himself was convinced that the uilitarian minded Brits would abandon the
>Roman alphabet in favor a phonemic alphabet on purely practical &
>utilitarian grounds.  But the phonemic Shavian alphabet has never caught
>on; we find our less-than-phonemic version of the Roman alphabet adequate.
I don't think the Shavian script is so great.  Its failure to catch
on is no wonder, and certainly bears nothing on our present discussion.


Always interesting.


Regards

Replies