A Gallimaufry

Na'vi

I've managed to get myself on a panel for a local con this fall to chat about the science of Avatar. In theory, I'm there to talk about language. We'll see how many people in the audience are interested in that.

Tsrai

Tsrai now has nice set of postural verbs. I'm still thinking about the semantics of these, especially in verb chains, but I'm ridiculously pleased to be able to say this,

we	or-ta	blas	rabbë	dzwai
3SG	drink-PST	beer	get.horizontal	happen
He drank so much beer he ended up on the ground.

Inspiration

I recently got myself a copy of The Languages of Native North America by Marianne Mithun. What an astonishing diversity of languages this continent used to have. Language inventors will find so much inspiration in this book.

How to handle Tsrai compounds?

I tend to worry about lexical expansion early in the creation of a language. It's entirely possible for me to have several bits of derivational morphology tested and ready before I've even started seriously with, say, the verb system. I favor rather complex derivational systems (see the madness of Vaior's derivational system), but in Tsrai I'm trying to be more restrained. So, there will be much more compounding using root words and little or no bound morpheme use. I'm currently thinking I'll follow the Vietnamese practice, and write each element of the compound separately.

Since Tsrai so far has been head-initial for noun phrases, it seems best to use head-initial compounds (although the typology on this isn't so clear-cut). The big question now is — how to mark plurals? Since most noun plurals are marked with the suffix -ne, do I tack that onto the compound head, or the full phrase? What if the head noun reduplicates for plurals? Using syur person (pl. sisyur), véu study; inspect:

syur véu student (person-study)
?sisyur véu students?
?syur véune students?

Right now I'm leaning toward sisyur véu, but I'll worry about this for a few more days before deciding. Hopefully it won't take as long as Bixwá's relative clauses took — I worried about that for longer than a month before making a decision.

And I'm not even ready to worry about verb compounds. Do they keep the same headedness pattern? Right now I'm inclined to make N-V compounds follow that order, verb last, dli word, ba see:

?syur véu ba dli a student reads
?syur véu dli ba a student reads

And verb compounds will have the same puzzle for plurals. Again, for now I'm inclined to make the plural and tense marking go into the original V element, wherever it ends up in the compound.

Embracing Redundancy, Ambiguity and Nonorthogonality

Even though I have never created an auxlang (i.e., an auxiliary language, like Esperanto), there are still certain habits its easy for conlangers to fall into which seem to be more suited to auxlangs. The longer I create languages, though, the less I'm willing to tolerate some of these things.

The first habit is efficiency, the avoidance of seemingly redundant things, such as multiple forms of agreement. No human language is perfectly, or even partially, efficient. Indeed, redundancy in a spoken language is a positive benefit, since language happens in a noisy medium. Nonetheless, nearly all my languages avoid certain features. For example, I typically use either strict head marking or strict dependent marking. Rarely do I allow overlap, which is somewhat unusual.

At this point Bixwá is not going to chage in this regard, but I decided in Tsrai that I will mark the plurality of the subject in both the subject and the verb.

Gad kóis A man sleeps
Órói këskóis men sleep (reduplication for the verb, and the plural of man happens to be irregular, órói)

This double-marking of plurality is quite redundant — but also quite common in human languages.

The second habit is the tendency to avoid ambiguity. In beginning (or engineered) conlangs, this commonly presents itself in the form of perfectly identifiable word classes. Every derived adverb in Esperanto, for example, ends in the same letter, -e. In Láadan all the speech act morphemes begin in the letter b-. In one of my earlier languages, Mavod, all the tense markers begin with the sound sh-, and other word classes exhibit the same patterning. I've been getting away from this habit, which I now find rather dull, and in Bixwá I have evidentials I would once have felt slightly uncomfortable about and avoided. They are suffixed to the clause, usually landing on the verb:

-aazh	obvious, "of course, as we know"
-sh	direct perception
-xw	supposition, inference
-jin	report
-jíín	report from untrusted source

Only the two report evidentials show any family relationship. Further, there are plenty of verb stems that end in -sh and -xw, a situation I would once have avoided.

The third habit I call "orthogonality," which presents itself in multiple guises. I'm abusing the mathematical concept of orthogonality, but by this I mean the tendency to ensure maximal distinctiveness in whatever feature we happen to be working on at the moment. For example, we might mark all tenses using the same morphological pattern, or we might fill out all the possibilities of tense and aspect if we've decided a language needed that. Some natural languages do work this way, but plenty do not. Navajo, for example, is most preoccupied with marking aspect rather than tense. Nonetheless, it does have a future. For Tsrai, I've decided the verb has only past and non-past tense marking.

In Bixwá I have deliberately mixed aspectual marking among different sorts of morphology. For examples, the aspect "begin to" is marked with the aspect prefix koo-, but if you want to talk about a stative verb (what English uses adjectives for) entering a state, one uses a mix of the perfective prefix ho- and the outward direction prefix, cho-. For example, jed is be cold. To say "it got cold," you use chohojed (cho-ho-jed out.away-PRF-be.cold).

I have also mixed up certain matters of syntax and pragmatics in different places in the language. For example, Bixwá does have some conjunctions that work as in English, but it also has some preverbs that do jobs much like conjunctions (preverbs are mostly adverbial particles that must occur in a fixed relationship to the verb, and which don't participate in other morphology). For example, the preverb wil indicates continuation, "then, and then, next,"

Né wil kora-n ho-deezh.
1SG PREVERB:then dog-ACC PFV-see.
And then I saw the dog.

I have also grabbed the preverb sa', thus, there, at that time, for a discourse function — it asserts narrative integrity, that the statement is related to the current discourse. This is useful when something surprising or unexpected is being said, but I find I use it a lot in any narrative. I've also grabbed the preverb jééx, which carries "emerging, up and out" ideas, to be yet another way to indicate an inceptive or inchoative sense for transitive verbs (distinct from the stative verb cho-ho- pairing).

Some years ago Jesse Bangs posted to the conlang mailing list The Conlanger's Rant, which suggested among other things conlanging schools and conlang criticism, himself being a devotee of the "naturalistic" school. Along with a lot of conlangers, I was rather turned off by this idea. Among other things, I've never aimed at perfect naturalism in a language. I certainly had no desire to belong to a Naturalistic School of Language Creation, and still do not. But my tastes in how a language I'll find interesting ought to be constructed have changed, with the surprising result that I've been pulled more in the naturalistic direction, not from a desire for naturalness, but from boredom.

The Birth of Tsrai

The World Atlas of Language Structures is just a wonderful way to spend hours. One thing I realized while reading some articles is that in my years of conlanging I have systematically avoided using certain features of language that are very common across the world. I tallied up a list in my mind of things I've avoided, and sure enough, the outlines of a new language started to appear — Tsrai.

I have only rarely used reduplication, a process that is ubiquitous in natural languages. For Tsrai, I decided to use reduplication to indicate number, as a marginal process for nouns but the most common way for verbs.

I strongly favor very simple sound systems for my languages. Even if I have a large inventory of sounds, I keep the syllable structure quite simple and open, at most allowing resonant codas. So for Tsrai I've decided to use a moderately complex system, with a few more complex onset types allowed. This means my decision to use reduplication has resulted in some hefty tables of behavior. Of course, a recent exposure to Squamish may also have something to do with that. The big question was how to reduplicate syllables with complex onsets. I decided that since the second element of any complex onset is either an approximant or a resonant, to impose sound changes similar to Ancient Greek for such reduced syllables. For example tyar reduplicated is tityar (< *tytyar).

Other tendencies of the sound system are inspired by the Nobiin language.

I tend to favor VSO or SOV languages, so Tsrai is solidly SVO. I am also very fond of case marking, but for Tsrai I've gone isolating, using word order for syntax. I've taken inspiration from Yoruba and Vietnames (also SVO languages) and used certain particles to mark focus for fronting behavior, as in —

Lë ba gad dai I see this man.
Gad dai fë lë ba I see this man. (fë is the focus particle)

I decided to step away from my aspect obsession. Verbs are marked only for tense — a past vs. non-past distinction only in verb morphology — letting adverbs and verb auxiliaries take up the slack.

I briefly considered using some sort of ablaut change in verbs to make ergativity a lexical category, a la Classical Chinese. But that got too messy for other plans for the language, so I tossed it. The idea may reappear for transitivity matters.

I do want to include verb chaining, but this presents some interesting design questions. At the moment a verb's form may be changed in two ways. First, reduplication for plural subjects, as in këskóis from kóis sleep. Second, it may take the suffix -ta to indicate past tense. The suffix is prone to assimilation, so that the verb varag choose, select may appear as vëvarag (pl.), varakka (past) or vëvarakka (past pl.). The syntax questions right now for verb chaining are (1) do all verbs need to be marked for number and tense and (2) if not, would the first or the last verb set the number and tense for everyone else in the chain.