2005-07-01Semitic meets Sinaitic conlang
I was just halfway through a draft of this, when an inopportune keystroke closed my browser (reaching for Control-T, hit Control-W, easy mistake on Dvorak). Blast. I've not got all day, and so this won't be as good as what I was doing before.
Anyhow. What I'm thinking of is, first, to make consonants semantic. Pick a core vocabulary of concepts (minimally, the semantic primes, but as a matter of realism you'd want at least the vocabulary of Dublex). Assign them to consonant sequences; think about the Huffman coding used in Plan B, and assign the shortest codes to the most-used concepts. (Yes, we're dealing with concepts so far, not words.) Notice that those codes are ultimately going to imply a sort of sound symbolism for words built with them, so when you have a choice for how to assign them, be smart.
There is a possibility of compounding here, and Plan B's provision for tree-building comes in handy—but remember that thematic roles, and anything smacking of them such as parts of speech, belong to a different layer: If you construct a consonant sequence for 'the guitarists' swift destruction of the hotel room' it is the same sequence as for 'the guitarists swiftly destroy the hotel room' (example after Pinker's The Language Instict). Same tree structure, same consonants with the same few tree structure markers.
Now, make vowels derivational (and perhaps partly inflectional, covering number, gender, speaker's attitude towards, determinedness, tensedness, or expressive force, or what have you). Determine just what operations you need. Note that Kali-sise seems to do OK with an extremely minimal set, but the Semitic languages have a larger array of such things (and I can't tell you how many times I've looked at a conlang's causative constructions and wondered why they just didn't use a derived hiph'il). If you want to go all-out, there is a hefty collection of carefully-selected, orthogonal derivation operations available in Rick Morneau's translation interlingua (most recently known as Ladekwa, I think). Anyhow, devise a way of assigning vowels to the consonant strings in order to give them parts of speech. Logli has some useful inspiration here, as do the Semitic CVCVC structures. Logli's system, though, includes a complete self-segregating morphology of its own, and you might or might not choose to keep that or the morphological features of Plan B.
(Make sure you've got a way to handle things if, say, your patient-noun form has two vowels, but the actual patient is a one-consonant word. You can allow a dummy consonant, you can employ diphthongs, or you can operate phonologically on the consonant root to create a standard-form stem. Your choice. Likewise, what do you do if you have more cononants than vowels?)
At any rate, now we're to the point where we can make tones syntactic. This is, of course, a blunt rip-off of the idea behind Gua\spi. Just how to divide the labor between inflectional marking in the vowels and strict ordering within the tonal syntax is a question you can answer more than one way. I can indeed imagine that, with the vowels already specifying part of speech, we just need to assign tones as though they were cases, distinguishing the ergative from the absolutive from the benefactive from whatever. We can do topic/comment by using one tone for a topicalized term, and having a particle fill that term's usual syntactic place. We can use a particle to suspend less-important modification or specification of an item, shifting that elaboration to the end of the utterance where it is picked up with another special tone. We can slide from syntax to pragmatics before we know what hit us. We can retreat and see what tones could contribute to an otherwise simple, complete syntax such as that of Kon'ya.
For that matter, Kon'ya is a fine model for many types of derivational processes that we might want. And when proposing basic vocabularies, I neglected to mention Sona.
So there you have it. You get clean designs with redundancy, the ability to promote or demote by importance (including packing-into-single-words) flexible syntax, and if I may say so a big dose of clever.