Re: Self-Segregating Morphologies

Mike S:
> Hello group!
>
> This is actually my second post here.  I tried to post a message
> early yesterday but apparently it has to go through listserv, not yahoo.
>
> Anyway, I am designing a logical conlang, and I have been
> experimenting with various self-segregating morphologies.  Here is a
> brief outline of systems I have looked at:
>
> A. Lojban Approach - basically, a required consonant cluster at (or
> near) the front of a word marks the start of the word;  required
> penultimate stress marks the end.  Note though, Lojban's full
> implementation is *much* more complex than the description here
> suggests.
>
> B. Katanda Approach - Morphemes and submorphemes of form C(C)V(V)(V),
> belonging to different categories, are combined to form roots, and
> then words.  A bit complex at first, but quite regular and very
> productive once grasped. I believe this system is quite ideal for
> Morneau's powerful derivational approach to language design.
>
> C. Ceqli Approach - A single fricatives or stop, or a cluster of
> fricatives and stops, marks start of word; root extended freely by
> vowels, liquids and nasals.  Utterly simple.  This system lends
> itself ideally to languages of the isolating sort.
>
> D. Vorlin Approach - Roots take form CV[CV]C, where [CV] can be
> repeated zero or more times.  Not perfectly self-segregating as
> implemented, but a good start; could be made self-segregating with a
> little tweaking.
>
> Here are some questions to consider:
>
> Are there any systems out there greatly different from the four that
> I have mentioned?
Yes. Let me first mention my preferred varisnt on the above sorts of
system:
* C alternates with V or (@) = elidable schwa.
* word boundary # is present iff in the environment V#CV or VCV#.

As for a gretly different system, I offer that of Livagian. With
Livagian the goal was to have self-segregating words while
constraining the phonological shapes of words as little as possible
(so as the facilitate another goal, that of maximizing sound--meaning
iconicity). The rules are quite complex, because the phonology is
worked out in great detail, but briefly it works thus:
* the grammar prescribes permissible word-initial sequences of
low-tone syllables
* a syllable (or sequence of syllables) that could be initial in
a word-initial sequence of low-tone syllables cannot occur word-
finally

In less accurate but more perspicuous terms, the rules can be
restated thus:
* the grammar prescribes permissible word-initial consonant
clusters
* a consonant or consonant cluster that could be initial in
a word-initial consonant cluster cannot occur word-finally

> Is it important to self-segregate the morpheme level, or is word-
> level self-segregation sufficient?
It depends. If the morphology is compositional, productive and
regular, then self-segregating morphemes are necessary. Otherwise,
they're not.

But as I've argued before, I think that deviating from isolatingness
-- i.e. introducing a distinction between words and morphemes --
adds much complexity to the grammar, and this addition needs to be
justified in some way conceptually (e.g. by making all closed-class
morphemes bound affixes).

> Should self-segregation be based on semantics or on phonological
> shape alone? (e.g., should prefix morphemes need to be memorized, or
> should such semantics be identifiable from morpheme shape?)
I think this comes down to a difference of degree rather than of
kind: if you don't have homophonous morphemes, then it's just a
matter of how specific the phonological rules for self-segmentation
are.

> How much importance do conlangers in general place on self-
> segregation?  For those of you who have build self-segregating
> morphologies into your conlangs, what sort of appraoch did you take,
> and why?
The point of having self-segmentation is as a necessary condition
for true grammatical nonambiguity. It's a slightly quixotic goal,
though, in that the ambiguities that actually do beset real
language use hardly ever involve misparsed segmentation, though it
must be conceded that in speech prosody helps segmentation, while
in writing the presence of spaces does.

> What system do you think is best and why?
It very much depends on what the other goals of the conlang are,
because different schemes impact differently on concision, on
flexbility of shape, and so on. Of the schemes you listed, I
like Katanda's the most. If I were to redesign Loglan/Lojban
(which is what Ceqli is doing), I would use the V#CV, VCV#,
C(@)C scheme I outlined above.

--And.