Theiling Online    Sitemap    Conlang Mailing List HQ   

Re: Optimum number of symbols

From:And Rosta <a-rosta@...>
Date:Tuesday, May 21, 2002, 23:56
Before I launch into the specifics of a reply, let me bring to the
analysis of scripts a distinction analogous to that between word
and morpheme. A single character, analogously to a word, is the
minimal independent unit -- the minimal unit that can stand alone
without combining with other elements. But just as words can
be monomorphemic or polymorphemic, so a character can either be
a single unanalysable whole -- 'monomorphemic' -- or it can be
broken down into separately meaningful elements ('scriptemes'?)
that combine together to yield the overall signification of the
character -- 'polymorphemic'. These 'scriptemes' would be
analogous to bound morphemes.

Directed to Mike S:
I have the sense that in responding to Ray's original question, I
have been focusing on characters, while you have been focusing on
scriptemes. (E.g. in a syllabary where each character is
composed of a C scripteme and a V scripteme, you say this script
is merely an alphabet in disguise.)

Mike S:
> From: "And Rosta" <a-rosta@...> > > Formerly a requirement of characters in the Livagian script was that > > they could be written in a single stroke from top left to bottom > > right (or top right to bottom left when switching direction in > > boustrophedon mode). There was no trouble in generating scores of > > characters as simple as the roman miniscules, though clearly there > > is a loss of redundancy/distinctiveness. As the character inventory > > grew and grew, though, I decided to admit diacritical marks placed > > above and below characters, for they multiply the inventory of > > characters without greatly adding to the average graphical or > > scriptoral complexity. > > Might a good practical test of efficiency be to compare a syllabary's > hand to Roman cursive?
Yes, though working out the specific criteria for the comparison would be difficult. Based on both common sense and my own experience of playing around with Livagian and its thousands of characters, I'd say that the care needed to write characters is proportional to the number of characters in the script. Up to a certain point, it is quicker to write one character drawn from a set of (say) 250 than to write two characters drawn from a set of (say) 25, but beyond that point, say when there are 2500 characters, the effort of writing a single character so that it is distinct from the other 2499 is greater than the effort of writing more characters that can be written more sloppily because they contrast with fewer other paradigmatically.
> At first it seemed clear to me that an alphabet, by containing > fewer characters, would naturally be more efficient as a writing > system, insofar as each character could be made maximally simple. > Mulling it over and actually conceiving an optimizing procedure > for creating a script that uses a syllabary, I now believe that > it is possible, if not likely, that the syllabary approach might > generally be expected to be more efficient strictly as a writing > system, notwithstanding any other disadvantages such a script > might present.
Based on my own unscientific experience, I'd say that so long as a low degree of redundancy is acceptable, the graphical devices sufficient to create (say) 25 characters can without elaboration generate a much larger number.
> Oddly enough, it just struck me now that my own loglang lends itself > well to the syllabary approach. My language is riddled with > consonant clusters and diphthongs alike; however, I have contrived > a "deep morphology" in which all syllables are CV; consonant clusters > and diphthongs occur when a schwa is elided. This is the theoretical > justification by which schwa-buffering is legal; in theory, schwa > buffering merely reverses an underlying elision. But it just > dawned on me that it would equally justify a syllabic script. > Implementation would require 18 x 6 = 108 characters. How strange > it seems to contemplate.
Ah: so when I get my arse in gear and send you the Livagian phonology document, it will have a deep ring of familiarity to you, since your description of your loglang could equally well apply to Livagian.
> > > If syllables are more intuitive to learn and use, then there should > > > be no temptation to encode any phoneme-level information into > > > a syllabic script. Explicitly encoding phoneme-level information > > > into a syllabic script is analogous to encoding phonetic data such > > > as +/- voice, +/- velar into a phonemic symbols. > > > > Indeed. And in what way does using a featural alphabet undermine > > its alphabethood? For example, if characters with ascenders were > > voiceless consonants, with descenders were voiced consonants, and > > without either were vowels, would this prove to you that alphabets > > were not the optimal system and that pure featural scripts were? > > I perceive two questions here. First, to what degree does the use > of features change a script's status? Second, what does that use > suggest about the efficacy of the script's nominal level of analysis? > > The short answer to these questions is that it depends on the > nature of the use. If features are employed by the script's inventer > merely as a device to achieve basic visual contrastiveness among > the characters, in the same manner that random markings would achieve > the same purpose, then the answers are clearly 'not at all', and > 'nothing'. On the other hand, insofar as the features are added > as an aid to sound out the characters, I think the answers are less > obvious. In other words, the issue is not to what degree features > are *used*, but rather *depended upon*. I don't think this is hair > splitting either. I frame the question as, are you reading the > characters, or are you reading the features? If you are consistently > reading the features, then how do you justify the efficacy of the > script's nominal level of analysis?
I doubt that it could be said that one is consistently reading the features, for it is known than in alphebetic writing we don't even read the letters, except for unfamiliar words: we read by overall word-shape. So let us suppose that these features -- these scriptemes -- are aids to learning unfamiliar characters, but not necessary to the fluent reader.
> Wherever such featural dependence indeed occurs, it is proportional > to the productivity of the features over the entire script. > The more productive a feature is, the more likely you are dependent > on it. The more you are dependent on it, the more doubt is cast upon > the efficacy of the system's nominal level of analysis.
Fair enough. By these criteria, your judgments make more sense to me now. For my part, I had been judging the nominal level of analysis by the minimal independent unit criterion, as outlined at the start of this message.
> Now, it can be seen that the potential productivity of features in > the two types systems differ by an order of magnitude. In an alphabet > of 30 symbols, any particular phonological feature, such as voice, > is not likely to distinguish more than 20 symbols. But even 20 doesn't > matter because all letters will appear so frequently that their > memorization ease is improved little by such featural regularizations. > Finally, I doubt it would be fruitful to bother actually trying explain > phonetic distinctions such as voice, which are even more subtle than > overall phonemic values, to children. > > In contrast, in a syllabary, it's easy to conceive of a system > in which a hundred characters are distinguished by a feature, which > in practice if not in theory, is simply marking some phoneme > such as a final nasal. With scores of uncommon syllabic characters > lurking ready to appear at any given time, it would be very much > in a learner's advantage to understand the phonemic significance > of a character's features. So, justified or not, I have wonder > whether the syllable is truly the optimal level of analysis.
I will agree with you that the syllable is not the optimal level for what scriptemes signify. The optimal level for the scripteme is (IMO) the voice/place/manner 'feature'. But for reasons given in earlier messages (and also by Nik), I prefer the syllable, when practicable, as the significatum of the independent character.
> Lastly, let me note that on the subject of features I don't really > have a horse in this race. Using features, or even depending on > them, never throws alphabets into doubt in the same way it throws > syllabaries into doubt. This is because a phonetic script could > never be a viable substitute for an alphabet. In contrast, > a syllabary can always be closely approximated, if not outperformed, > by an alphabet, with an order of magnitude fewer characters to boot.
I can't dispute your general point, but there are ways in which a suitably ingenious featural script could eliminate, say, the redundancy of marking place distinctions in each constituent of a homorganic cluster. -- Indeed, some tengwar modes do just that. So there remains the issue of what the optimal amount of feature- signifying elements to augment a basic alphabet is.
> > > To do so is to demolish the entire argument that syllable-characters > > > are easier to teach. > > > > Hardly. The independent pronounceability of a character is an > > important aid to first learning it. (And note how children learning > > to spell in English go through a phase of treating letters as > > syllables -- e.g. <lat> for <later>.) Once some symbols were > > learnt, the learner could then recognize the featural/phonemic > > element to their construction, and use knowledge of this pattern > > as an aid to learning the rest of the symbols. > > The independent pronounceability of a character may well be > an important aid to *first* learning it, but from this it doesn't > necessarily follow that it is an important aid towards mastering > the whole system, which of course, is the important goal.
I don't understand the difference between first learning the system and mastering it, and why the latter is the important goal. Well, of course mastery is the important goal, but the greatest hurdle to mastering a writing system is first learning it. As all of us parents know, learning to read is for children the most difficult and demanding part of the process that eventually leads to mastery of the language. Anything that could be done to ease that task is of immense value.
> Since you have pointed out that featural scripts lend themselves > to phonemic analysis, I feel compelled to point out that alphabets > equally lend themselves to being learned as the combinations of > letters in syllables or words. Educators commonly debate the > pros and cons of both learning styles, with some siding with > the "phonetic" approach and some advocating the "whole word" > approach.
I understand what you mean, though I believe that what is known as the 'whole word' approach works on the principle that the word is learnt as a visual gestalt, and there is accordingly no effort to stick to words with 'phonetic' spelling.
> It is only the essential flexibility of the alphabet > --or the careful encoding of features into a syllabary--which > would allow this debate to even occur.
OK.
> I notice that you attempt no defense of nonfeatural scripts. > Given that featural syllabics lend themselves to phonemic analysis, > and letters lend themselves to being learned in combinations, > perhaps we could come to the agreement that neither featural > scripts nor alphabets have a clear advantage over the other. > > I must continue to maintain however, that except perhaps in the > rarest of cases, that nonfeatural scripts are simply not as easy > to learn as alphabets. While alphabets have every functional > advantage that nonfeatural syllabaries have, syllabaries have > one clear disadvantage that alphabets don't: the number of characters. > A properly designed alphabet will contain C + V characters while > a best-case minimum for syllabaries is C x V. In order to justify > this handicap, there would have to be a much more obvious advantage > to adopting a nonfeatural syllabary than I have seen this far.
I mildly agree with you.
> I have to say I find it puzzling that syllabaries in general-- > not referring to the featural sorts you have described--have > been so consistently vaunted for their learnability, particularly > by westerners.
Is it correct to say these are not featural -- that the shape of a character gives no indication of its value? (I think it's true of the Cherokee syllabary, but that never struck me as admirable.) Many others on the list have far more expertise than me in answering that question, so I won't attempt to answer it myself.
> If I want to learn a new language, I'd much rather > have it use an alphabet, or a script that has phonemelike features. > And I would suspect most informed people, regardless of their L1, > would agree. Any system that lends itself to phonemic analysis, > even a foreign one, will let any adult who already grasps the > phonemic concept read the language in a very short time, regardless > of the phonology; the same simply can't be true to the same degree > about a nonfeatural syllabary, especially one tailored to a language > with a nontrivial syllable structure.
I agree.
> > > If understanding phonemic distinctions are needed, or > > > at least helpful, in learning or using a syllabic script, then I > > > can't see any reason not to use a phonemic script in the first place. > > > > Learnability and economy, as already stated. > > > > --And. > > I do find syllabaries attractive in three areas. First, is > the combinatorial freedom you mentioned in an earlier post. > This appeals to me not so much for any immediately obvious > boost to efficiency or some other practicality but rather for what > it offers in the way of a thought experiment. I am not at all > sure what fruit such a project might yield, but implementation > would certainly justify adopting a syllabic script if undertaken.
I see this combinatorial freedom as having merits that are primarily conceptual rather than utilitarian. The combinatoric criterion for defining minimal units is far more robust and determinate than other criteria that yield smaller (or larger) minimal units.
> Second is the possibility of a more efficient writing system. > Having very briefly looked at the question from the standpoint > of communication theory, I quite inadvertantly convinced myself > in this area; however, I have a very strong feeling that > obtaining the full benefits would require a strong commitment > to a non featural script. Probably features could be maintained > in the less common characters. I really think writing efficiency > trumps learnability anyway; you only have to learn once, > but you will write countless times. If you're attracted > to the syllabary idea, why not exploit this to the hilt?
Hmm. I think that it needs a really really hefty gain in writing efficiency to justify imposing extra torments upon our children, or denying them the power of reading for more time than is necessary. Other points to consider from a utilitarian perspective are: * readability, since we spend more time reading that writing, * writing by keyboard (or similar) -- most of us (who are here at least) write much more on a keyboard than by hand And lastly, for me, there are more considerations to writing and to language in general than purely utilitarian ones. For example, it is conceptually desirable that the writing system mirror the phonological system.
> Third is an area I have heard none of the syllabary advocates > mention so I'll mention it myself: reading. This is not to > be confused with learnability, where I continue to feel that > alphabets have the unassailable highground. What I have not > seen explored in these posts is whether a script might be > designed in order to maximally benefit reading speed and > comprehension. I don't know what the answer is, nor do > I know how this might interfere with writing efficiency. > But I do think these latter two areas, reading and writing, > are ultimately the only places to look for a true claim > of the efficacy of syllabaries.
Reading and perhaps, for similar reasons, handwriting, would be helped by having characters that maximally differentiate the overall shapes of words and/or morphemes. But this does not necessarily bear on the syllabary/alphabet debate. --And.

Reply

John Cowan <jcowan@...>