Re: [bpfk] camxes and syllabification in zi'evla

On Sun, Oct 26, 2014 at 5:48 AM, And Rosta <and.rosta@gmail.com> wrote:

"Morphogical syllables" (maybe renamed to something a little less susceptible to confusion) would be fine. My questions would then be what the rules are and why. The norms of language don't constrain morphophonological rules much, so they can be as weird and wacky as necessary. The rule you give below, CVC*, seems pretty straightforward.

Here's the theory of morphological syllables in a nutshell. There are two kinds of syllables: vocalic and consonantal. Most syllables are vocalic. There are a total of 24.840 possible vocalic syllables, and 64 possible consonantal syllables.

The consonantal syllables are all of the form CR where C is any of the 17 consonants bcdfgjklmnprstvxz and R is any of lmnr different from C.

The vocalic syllables consist of all possible combinations of (morphological) onset-nucleus-coda.

There are ten valid (morphological) nuclei: a, e, i, o, u, ai, ei, oi, au, y

There are 138/139 valid (morphological) onsets. The dot/apostrophe (which are in complementary distribution, so they could be considered either one or two), the two glides i/u, the 34 controversial CG, the 17 single consonants C, the 48 permissible initials CC listed in CLL, and 36 permissible initials CCC based on the permissible CC. 80 of the 84 CC(C) onsets fall within the pattern [csjz][ptkfxbdgvmn][lr]. the remaining 4 are tc, ts, dj, dz.

There are 18 valid (morphological) codas: the 17 consonants and the empty coda.

That gives 138*10*18 = 24,840 or 139*10*18 = 25,020 possible vocalic syllables.

All words (except for cmevla) consist of a sequence of valid syllables.

There are also some constraints on which syllables can be adjacent: the final consonant of a syllable and the first consonant of the next syllable can't be the same, they can't have different voicedness, they can't both be sibilants, if one is x the other can't be c or k, if the first is m the other can't be z. Also, a syllable that ends with n can't be followed by one that starts with an affricate (tc, ts, dj, dz). Some of these constraints sound completely arbitrary, and they are. In addition, some combinations are disallowed only because they give the same result as some other combination, e.g. tav+la = ta+vla. It doesn't make any difference if we say tav+la is disallowed, or if we say it's equivalent to ta+vla.

Now, not every valid combination of valid syllables will result in a string of valid words. Here's where the rafsi madness comes into play. Syllables with a "y" nucleus in particular are very restricted in how they will combine, and most of them can never occur in any valid (non-cmevla) word. Consonantal syllables are also somewhat restricted in that they can't appear until a vocalic syllable has appeared (except again in cmevla). There are also sequences of valid syllables (called "slinku'i"), which cannot be a word or a sequence of words.

I think that's basically it, although I may be forgetting some detail or other.

Actually, that's not quite true. We do need to identify valid onsets
in order to determine words, but this discussion wasn't really about
onsets.

The question about onsets being whether CGV is a valid onset?

But "morphological onsets" are needed too, aren't they. E.g. /patrAma/ is two words /pa trAma/ whereas /partAma/ is one word, because of the rules for morpho-onsets.

Yes. By "this discussion" I meant the one that started this thread in particular, which was about a bug in the PEG morphology that allowed onset-less syllables after consonantal syllables, when my intention when writing the morphology PEG was that all syllables should have a non-empty onset. So we had words like mas-tl-a pos-tm-o.

Well, today's morphophonology is yesterday's phonology (e.g. the vowel alternation in _sane--sanity_), so it makes sense diachronically but not synchronically. But for Lojban you don't look for diachronic explanations. (In Lojban too the actual explanation is of course quasi-diachronic, in that the complex constraints on 'clusters' were likely invented before the buffer vowel.)

Yes, and not only that. Since lujvo came after cmavo and gismu, fu'ivla came after lujvo, and cmevla were probably there all along but in a parallel universe of their own, the rules encompassing them all constitute a complex patchwork which is hard to put together into a seamless whole.

Without a buffer vowel, it does make sense to limit the amount of
consonant clustering that can occur. If there was a buffer vowel,
the morphophonological syllable could still be onset-nucleus-coda as
now, but with the coda allowed to contain as many consonants as you
wanted. That's not how my dialect of lojban works though.

In what way is it not how your dialect of Lojban works? It would categorize as valid some words that you categorize as invalid? Or would it insert word-boundaries differently? The latter seems more significant an objection than the former.

Just the former. I would not want to categorize "poktpftcu" for example as a valid word.

So anyway, do you advocate abolishing the buffer vowel? An alternative would be to insist that every licit phonological string has both a CV syllabification (with buffer vowels) and a resyllabification without buffer vowels. That alternative strikes me as needlessly complex, but as still preferable to abolishing the buffer vowel.

I don't mind it appearing sporadically at the phonological level, I just don't want it at the phonemic level because I think it hinders more than helps.

mu'o mi'e xorxes