Does gjâ-zym-byn violate linguistic universals?

I've sometimes described gjâ-zym-byn as a psychological experiment conlang; that's less apt a description in recent years, now that I'm using it routinely and rarely if ever tinkering with its phonology or its basic grammar (and even the changes to the lexicon are growing more conservative over time), but it describes my original design and early use of the language pretty well. Among the design criteria was an intention to try out unusual grammatical and semantic constructions and see if I was able to learn the language fluently without altering those features. The question then arises: does gjâ-zym-byn violate any linguistic universals, and to the extent that it does, have I learned those aspects of it to fluency?

Some violations of apparent universals in gzb were due to my not knowing much yet about certain areas of linguistics when I started creating gjâ-zym-byn. For instance, natural languages that have clicks have them in a two-dimensional point of articulation/manner of articulation system, like other languages' pulmonic consonants. gzb has a fairly one-dimensional system, with ony one point of articulation (bilabial) having two clicks in different manners of articulation; and it has exactly one ejective. It may also violate a phonological universal by making clicks and ejectives syllable nuclei, and not onset or coda consonants. But that's due to young conlanger naïveté, not a deliberate violation of universals. Result of experiment: a system like this may be unlikely or impossible to evolve naturally, but it's not hard to learn. I found the clicks to be easier to learn to pronounce consistently in the contexts they appear in than most of the other non-English phonemes in gzb.

A universal that's so obvious that nobody has explicitly formulated it, as far as I know, is that if a language restricts certain phonemes to use in a certain part of speech or distributional class of words, that class will be ideophones or interjections. gzb violates this with all of its vowels, and more blatantly still with respect to the clicks and ejective mentioned above — each such consonant occurs in exactly one monophonemic word, one of the personal pronouns. Each vowel is also restricted to occur only in morphemes of certain distributional categories — some in noun roots, some in suffixes, others in postpositions, clitics or conjunctions.

gzb marks thematic relations with a large number of case-postpositions, and (for instance) not only consistently distinguishes agent, patient and experiencer with separate markers in both transitive and intranstive sentences, but also patient (i.e., thing physically affected by the action of the verb) vs. creative object of result vs. physical object of result vs. unaffected object of sensation/attention vs. transient object of performance... When I've talked about this on the CONLANG mailing list, nobody has mentioned natural languages that do things the same way; fluid-S active/stative languages exist, though they're rare, but I don't know of any natlang that distinguishes so many thematic relationships with distinct morphology or syntax. Result of experiment: mixed. Most aspects of this system I've learned to use fluently, and I even coin new case-postpositions on the fly semi-consciously from time to time, as a fluent speaker of a language with a powerful derivational morphology coins new compound and derived words on the fly without conscious planning; but my disfluencies log suggests that using the wrong postposition accounts for a bit less than 10% of my mistakes using gzb — not the most common class of errors, but far from the rarest.

Related to that, gzb is a verb-drop language, where if the case-postpositions marking a noun tell enough (in a given context) about what's going on, the verb may be omitted (and indeed, for some pairings of postposition with their basic meanings, no verb is allowed; any possible verb would change the meaning). Results of experiment: this works fine. I don't know if it violates a universal, though.

Anna Wierzbicka claims, among her semantic primes and universals, that all natural languages have a basic word meaning "good", that covers a wide range of specific meanings but generally means that the speaker or some salient person likes or approves of the referent of the word being modified by this language's adjective or stative verb for "good" in some unspecified way. gzb lexical semantics forces the speaker to be more specific: there is no broad word corresponding to "good", "bon", "agathos", etc. in the natural languages I'm familiar with, but more specific terms for "holy", "morally approvable", "just", "healthy", "tasty", "beautiful", "useful", "suitable for one's purposes", "appropriate to a thing's nature", "inducing contentment", "time-tested", "polite"... And all but one are derived words; none are root except for that meaning "holy", and it's not the root of most of the others — they're derived from various roots for abstract standards or mental states plus an "according to" or causative suffix. The same is true with respect to Wierzbicka's semantic prime "bad". Results of the experiment: too soon to tell. Confusing two words in the same distributional category and semantic field is a common type of disfluency in any language one is learning (or even in one's native language), and I don't have enough data to say whether I'm making such mistakes more often with the various hyponyms of a nonexistent "good" hypernym than in other semantic fields.

I think it may also violate a semantic universal (not according to Wierzbicka but to Brent Berlin and Paul Kay) with respect to the colors which are lexicalized as basic words; it has words for white plus the light primaries red, green and blue — not white, red, green and yellow, as in most or all natural languages that have four basic color terms. Results of the experiment: too soon to tell. I'm not an especially visual person, and I find, looking at my most recent analysis of the part of my corpus I have in electronic form (the bulk of it is only on paper), that the root {lě'ku} "white" occurs only twelve times (mostly in an opposite-derivation for "black", and in a single text), {ðru} "red" occurs only twice (once compouned with {lě'ku} for "pink"), and the roots {vrĭm} "green" and {hyrŋ} "blue" don't occur in this corpus at all. My disfluencies log doesn't show any mistakes with respect to color terms, but it's not very extensive.

In general, semantics is one of the areas I have been most experimental — there are other areas of gzb semantics where there is a missing hypernym relative to the natlangs I'm most familiar with, as with "good/bad" and color terms. gzb has words for "chair, sofa, bed, etc." and "desk, table, shelf, etc." and no hypernym for "furniture"; terms for "fuzzy border region" and "sharp boundary" and no general term that covers both; basic terms for "rapid transformation", "gradual change", and "decay" and a derived, less-frequent general term for "change", and so forth. But in these areas I don't know of any claimed universals, and I'd be shocked if "furniture" in particular is a universal even among natural languages spoken in higher-tech societies.

gzb has a default word order that's very rare in natural languages, Object-Verb-Subject; and it may violate a statistical universal with respect to those few natural languages, in having its complement postpositional phrases (time, place, etc.) come normally at the beginning of a sentence, rather than at the end, as in Hixkaryana, the only OVS natural language for which I've found good documentation. However, gzb is also a free phrase order language; other orders of the main sentence constituents also occur, though less commonly, and it's not terribly uncommon for a locative or temporal complement to go at the end of a sentence. Results of experiment: uncertain. Given gzb's syntax and pragmatics, it's hard to say how often my use of a word order other than (L)OVS is due to disfluency and how often it's due to a subconscious choice to emphasize a constituent displaced from its default location; especially since I don't have good statistics on how often I've used different word orders. It should be possible, given gzb's morphophonology and syntax, to extract that information from my electronic corpus, but I haven't done it yet.

Another word order universal is that OV languages very strongly tend to have the auxiliary verb after the main verb. In gzb, the auxiliary precedes the main verb (and can have the object of the main verb come between it and the main verb). However, most of the OV languages on which this universal is based are SOV, not OVS; there are too few OVS languages, and most of them are too closely related genetically, for correlations to be strongly predictive about other unrelated OVS languages. Result of experiment: this part of gzb hasn't been hard for me to learn, but that's not surprising because English and the other languages I know well all have Aux-V order.

Earlier versions of gzb violated more universals; for instance the first version of gzb had no root words for composite numbers, and no number base; composites were expressed by adding or multiplying words for prime numbers. I added root words for some powers of ten and sixteen after a few years, finding that this system was fun but (as I'd suspected when I started the experiment) not very practical. Even now, though, the only numbers less than 32 for which I use those root words for ten and sixteen are ten, sixteen, twenty and thirty; all others, plus some larger numbers are expressed as sums or products of primes (and 1, of course, as it strictly doesn't count as prime). Results of the experiment: universal confirmed! (Well, no, not really. But definitely not busted.)

Up until 2008, gzb violated the implicational universal that languages with object-verb order have standard-marker-adjective order in comparatives (i.e., if in English one said "Dog man bites" instead of "Dog bites man", one would also say something like "Dog man than smaller is" (or "man than er-small"?). I found this system harder to learn than other aspects of the language, so eventually I changed it to use a separate postpositional phrase for the standard + marker constituent, which normally comes before the comparative adjective or stative verb. The old system where the comparee and standard are linked within the subject postpositional phrase by a dedicated comparative conjunction is still used for comparative of equality ("is as ... as ..."), however. Results of the experiment: universal confirmed, sort of.

Main {gjâ-zym-byn} index
Derivational morphology
My conlang page
My home page

Last updated March 2014