Saturday, February 24, 2007

Genitive Relationships

I think it's time to finally figure this thing out. Supposing we want to express something like "my father's house," through the years we've usually been thinking of something like

ka talo o ni ato
the=house GEN=1SG=father
"my father's house"

The big question in here is the o. Is it really necessary? What all is it used for? While trying to imagine what I'd do without o, it occurred to me that a structure like this might work:

ka talo ni ato
the=house 1SG=father

This follows our rule of head-modifier word order, and all; but would it work acceptably with the rest of the grammar? Well, just now I was thinking about direct objects. Take this sentence:

ni tata i loha ko sihi
1SG=dad 3P=love NONINST=vegetable
"my dad loves vegetables"

If this is how we're treating direct objects, and I think it is, then we should also have the following in parallel with our usual verb/noun/adjective series:

ka loha ko sihi
DEF=love NONINST=vegetable
"the vegetable lover"

ka mehe loha ko sihi
DEF=man love NONINST=vegetable
"the vegetable-loving man"

Note that we're going to have to look pretty critically at this last example, as this is basically a relative clause and I'm not totally sure that the clauses are going to be sufficiently distinct from each other articulated like this. But for the moment, let's assume this is okay to continue the argument.

If ka loha ko sihi means "the lover of vegetables," then we've got our genitive phrase! Of course, this is what a Latin speaker would call an "objective genitive," not a possessive genitive. But I really can't come up with a situation where there could be ambiguity between the two.

Well...except for one, whose importance I'm not yet sure of. How does one express the possessor of an agent? In most cases it's not problematic: "my father's builders" would be pretty unlikely to mean "the builders who built my father." But. I apologize for the following choice of verb, but it's the only one I can come up with that exhibits true ambiguity. What about e.g. "the emperor's killers?" I think a phrase like this probably occurs in Dune somewhere. There would be no way to know whether the assassins being spoken of in fact disposed of the emperor, or are employed disposing of others for the emperor. But maybe there's some way to resolve this by means of additional morphology. For instance, if we put a suffix on "killers" that indicates that the verb is the individual's profession (like we did with "wine salesman"), the ambiguity would no longer be pragmatically troublesome.

I'd like to take this moment to give thanks for the fact that this is not Loglan and therefore I don't need to worry about being able to clearly and unambiguously express every logical possibility.

Anyway, I feel like I'm getting away from my original topic. The point of all this is that I believe genitive relationships can be expressed simply by SPEC-NOUN SPEC-NOUN.

At this decision, though, a disquieting realization arrives. If we take our sentences about vegetables from above and use a pronoun object instead of a generic noun, look what happens:

ka tata i loha ni
DEF=dad 3P=love 1SG
"dad loves me"

ka loha ni
DEF=love 1SG
"the one who loves me," "the lover of me," "my lover"

We've just used ka X ni = "my X," instead of ni X which is what we've been fervently believing in for years. Uh-oh.

Okay, so I did think of this possibility originally -- in fact, in Ea opi le Koa I wrote ka tata o ni for "my dad." But I liked the way that saying ni tata was (1) shorter, and (2) treated "my" as a kind of determiner, which it is in a way.

Can I have it both ways? I don't think I can avoid the fact that there's no reason ka tata ni shouldn't mean "my dad," so what would be the difference between that and ni tata? Purely pragmatic? Hm. Well, variety is the spice of life, sure, but this is supposed to be an IAL.

There is one other disadvantage to doing pronoun possession this way, though I don't think it's strong enough to warrant throwing out the idea. If a postposed monosyllabic pronoun indicates possession, that means I can't use anything pronominal as a formative for my derivational system. That is to say, if ka talo ni = "my house," then there cannot be a word ka taloni. It's not that I'm into all that monosignificance crap, but this is just far too ambiguous for a well-designed system.

So, then, it looks like, instead of the 45-47 CV derivational formatives I was hoping for, I'm going to have to deal with around 40. Not such a huge loss, I suppose, but in a language with such a small phonology every such loss is pretty important. Here's what I have to exclude so far:

ka, a, ko, ti, i, ni, se, hi, nu whatever I end up choosing for 2nd and 3rd plural, if I decide I need them. Damn and blast.

Switching gears completely, I do want to mention that I've been throwing around the idea of providing optional reduplicative pronouns in addition to the usual monosyllabic forms. Thus, "I" would be either ni or nini, "you" = se or sese, &c. I momentarily thought, "wait! then possession could use the long forms, thus avoiding the derivation problem!" But ka talo nini is just far, far too long for a simple possessive phrase, in my opinion.

To sum up: "the X of the Y" is to be expressed as ka X ka Y. If Y is pronominal, the structure, at this point, can be either Y X or ka X Y. last thought before I give it a rest for tonight. If ka talo ni and ni talo are really equivalent, then either of the following would have to be acceptable for "my father loves me":

ka ato i loha ni
DEF=dad 3P=love 1SG

ka ato i ni loha
DEF=dad 3P=1SG=love

Um. Huh. This will require thought.

Friday, February 23, 2007

NP Modifiers

Though we have not yet RIGOROUSLY defined the semantics of this operation, we've determined that when two nouns stand in the relationship X Y, X is interpreted as the head and Y as the adjectival modifier. What if Y is not a noun, though?

Turkish, for instance, has a special marker that ties locative phrases to NPs. So

resim masa-da-Ø
picture table-LOC-3sg
"the picture is on the table"

But if we want to say "The picture on the table is pretty," we can't just say

*resim masa-da güzel-Ø
picture table-LOC pretty-3sg

...despite the fact that it makes perfect sense in direct translation. Instead, we see the following:

masa-da-ki resim güzel-Ø
table-LOC-ADJ picture pretty-3sg
"the picture on the table is pretty"

For this reason I was nervous about making any unresearched decisions about similar PPs in Koa. The sentence "the man in the house is bad" (and yes, I know I desperately need to expand my lexicon), for instance, in literal translation would run ka mehe ne ka talo i pua. But is

mehe ne ka talo
man LOC=DEF=house
"man in the house"

okay? Usually I asked this question and then pondered whether it ought to be mehe o ne ka talo, with this o showing that the preceding lexeme is about to be modified by something (I believe o is going to figure in my possessives -- maybe I'll explore this next time). I think, though, that I may be able finally to answer this one satisfactorily. Let's compare the following parallel structures:

As a predicate:

ka talo i iso
DEF=house 3P=big
"the house is large"

ka mehe i ne ka talo
DEF=man 3P=LOC=DEF=house
"the man is in the house"

As an instantiated adjectival:

ka iso
"the big one"

ka ne ka talo
"the one in the house"

Therefore, directly modifying a noun, and voilà:

ka talo iso
DEF=house big
"the big house"

ka mehe ne ka talo
DEF=man LOC=DEF=house
"the man in the house"

Considering that these phrases are parallel in the top two cases, I don't see why they would need to be differentiated in the final one. One less thing to worry about.

Wednesday, February 21, 2007

More Thoughts, Obtained Whilst Tossing & Turning

We need a complementizer. At least sometimes. It may need to be defined for each type of complement clause. But clauses of wishing, &c., definitely need it. For example:

ni halu = I want
ni mene = I go
ni halu [ni mene] = I want my goer

Note that this last result is not quite what we want when we nominalize that complement clause. Better would be

ni halu ko mene = I want [the idea/state/quality of] going OR
ni halu ko ni mene = I want [the idea/state/quality of] me going

Is it being too Indo-European to say that the pronoun can be omitted when the subject of the matrix and complement clauses are the same? Thus

ni halu ko mene = I want to go
ni halu ko se mene = I want you to go

What other kinds of complement clauses are there? I need to find a list. Reported speech and the like, now, that's an interesting one.

Le Catherine i loha le Heathcliff = Catherine loves Heathcliff
Le Edgar i si sano [le Catherine i loha le Heathcliff]
= Edgar said that Catherine loves Heathcliff

That doesn't strike me as being terribly bad, or anything. The only issue is that, in Koa, all NPs must have a specifier, and the nominalized clause le Catherine i loha le Heathcliff does not. Furthermore, isn't what Edgar said really an idea/state/quality, which is what I keep using ko for? In this case the sentence would need to run

Le Edgar i si sano [ko le Catherine i loha le Heathcliff] = Edgar said that Catherine loves Heathcliff

The jury's still out on this one for me. I need to think some more about it with other verbs -- Edgar believes, Edgar knows, Edgar reports, etc. One last thought: If Le Edgar i na halu [ko le Catherine i loha le Heathcliff] is the only possible translation of "Edgar doesn't want Catherine to love Heathcliff," wouldn't it be weird to have a non-parallel construction without ko for the reporting kind of thing discussed above? Hm.

Monday, February 19, 2007

Changes to Koa

A lot of changes have been brewing in my mind since I last wrote about Koa, and I'd like to get them documented before I move on again.

An asterisk (*) by a Koa word indicates that the word is being used for illustration purposes only and has not been officially decided upon to fill that semantic slot.

1) The 3rd-person marker i is required in all cases, including before aspect markers: le Keoni i si nae... . This is necessary because si nae can function not only verbally but also adjectivally and nominally. Ka mehe si nae = "the man who has seen"; ka si nae = "the one who has seen."

2) It's important to note that Esperanto/English-style nominalization must take into account the valence of the verb to be nominalized. ka suo does not mean "the food" but rather "the one eating." "The thing eaten" would need to be ka pa suo, as we're talking about the patient instead of the agent.

3) Verb modifiers in many cases do not need special markers, as I believe I had previously been employing mo. For example, if ka puhu koa is "the good speaker," then Le Keoni i puhu koa means "John is a good speaker" = "John speaks well." We will need to determine when this type of modification can be used and when a separate marker (mo, etc.) will be necessary.

4) I am feeling increasingly unhappy with my 3rd-person pronouns. Right now we've got hi = "he/she/it/they", ti = demonstrative. I'm feeling like hi is somewhat too neutral in phonetic structure to stand alone in phrases like ni nae hi. One possibility is to use ti or li for the 3rd-person pronoun and another morpheme for the demonstrative. Foremost in my mind is to for obvious reasons. To sa ni na si nae = "that I didn't see," etc.

5) As of the last described version, I was dealing with constructions like ni ka talo. These are no longer acceptable. ni talo is now the matrix; if necessary, it can be modified as ka ni talo, co ni talo, etc.

6) I was previously uncertain about whether an existential verb would be necessary. Furthermore, it was unclear to me what the selection criteria would be between that and the copula for e.g. locative constructions: ka keli i ne ka talo vs. ka keli i tai ne ka talo, etc. However, it is clearly important to be able to express concepts like "Does God exist?" If not by Ei ka teo* i tai?, then how? Maybe Ei ka teo* sa? Hm. Clearly further thought is necessary. And, though I certainly like ohi for the negative existential, I'm not certain whether (A) we need it and (B) it's appropriate given the fact that there are no other inherently negative verbs in the language.

7) I had been loosely referring to certain particles as "correlatives." I think this needs to stop because it confuses the issue. This being the case, how are co and hu actually used? And most specifically, what is really the difference between a and hu? If these are grammatically equivalent to ke, ti (or whatever its eventual form turns out to be), etc., we would seem to have

ka mehe = the man/men
a mehe = an unidentified man, some unidentified number of men
ke mehe = which man, what man, what kind of man
ti mehe = that man/men
hi mehe = her man
co mehe = all men
hu mehe = some men

I'm totally unclear at this point on the difference between a mehe, hu mehe. It could be that hu is talking about sets, whereas a is referring to definiteness. In this scheme, a mehe would mean "an unidentified man" or "some unidentified men," whereas hu mehe would mean "an unidentified set of men."

Okay, so that's fine. The question is, is that needlessly abstruse? And if we decide that it's not, we need to rigorously explore how it is to be used to render lots of English examples.

8) What's the negative "correlative?" If we start with hi na mehe = "he is not a man," then ka na mehe = "the one who is not a man." This does not seem to me to flow easily into what we may (possibly) have bene thinking before, viz. Na mehe i koa = "no men are good." Given, though, this could not be interpreted as "it is not a man that is good" which would be Na (a?) mehe sa koa or something. But if we do use na correlatively, then doesn't ka na mehe end up meaning...uh-oh, we've confused ourselves. Let's back the truck up.

i na mehe = is not a man
ko na mehe = the quality/state/idea of not being a man
ka na mehe = an instantiation of the quality/state/idea of not being a man

Okay. We're fine until we get to the correlative issue. If we do use it that way, then na mehe i koa would mean something like "there is no set of men such that the members of this set would be good." Then ka na mehe = "a definite instantiation of an empty set of men," which is clearly logical gobbledygook and not actual usable human language. So that's out.

Thinking about it, I believe co and hu perform rather different functions than the other specifiers. I think they need to be "outside the parentheses," as it were, the logical symbols A and E. So:

co mehe i koa = all men are good
hu mehe i koa = some men are good
na co mehe i koa = not all men are good (in other words, the same as hu mehe...)
na hu mehe i koa = it is not the case that some men are good = no men are good

There then seems to be a pragmatic difference between hu X and na co X which also needs to be rigorously defined. The only problem, then, is the fact that I'm not aesthetically impressed by na hu. Do I need to find a different morpheme for hu?

Fine. So how do I say "nothing?" Nahua? Oh my. "I don't want anything" = ni na halu hua or ni halu nahua? Okay, two thoughts. First, I don't like th efact that I have these two seemingly identical-meaning constructions, like in Esperanto. I really prefer the double-negative to this sort of thing, even though it's not "logical" -- so that would suggest ni na halu nahua. But I do worry about the logic of the thing. I think I really need to consider typology for this one: how many languages use a double negative in this kind of construction? I really want to avoid this working like Latin, in any event.

Secondly, yes, nahua is not doing it for me. I will definitely need to find a new "some" morpheme.

9) On the model of Yoruba, I've been thinking that it may be okay not to have overt marking to disambiguate direct and indirect objects in ditransitive constructions. In other words, "the wine man told me that this wine is good" would be either

Ka sahima* i si sano ni ti sahi i koa or
Ka sahima* i si sano ti sahi i koa la ni

So, we've got


Incidentally, this is another reason why the 3p marker is necessary: otherwise a sentence like hi si sano ni ka sahi i koa would be ambiguous with "he said my wine is good" and "he told me the wine is good."

10) Regarding the above: complement clauses. I didn't use a complementizer up there; is that what I want to do? Typology again will be necessary.

11) I had previously been working with

me = accompaniment
mo = adverbial
pe = instrumental

I now think that some combination of me and mo may be able to accomplish this task perfectly adequately without a specific "instrumental" case. For example, in my old sentence Le Atami i si kani* a sono* late* pe hi kita* = "Adam played a pretty song with his guitar," couldn't we really say that "with his guitar" is modifying the verb, and therefore is adverbial? Why not mo hi kita? "His-guitar-ly?"

If so, mo seems to be operating something like an applicative in that one language in Describing Morphosyntax, in which there's a marker for "with respect to." It would also very much resemble the currently woefully underexplained use of the adverbial suffix in Esperanto.

What's needed now is, once again, a rigorous discussion of the difference between me and mo. I say this because before I had the idea that mo could function as described in the above sentence, I had been thinking that it was probably fine to use me -- not "logical," per se, but probably typologically acceptable.

I have also been thinking that me might be useful as the coordinating conjunction between NPs, as opposed to e for VPs and sentences. This parallels many thoughts I've had in the past, and happily is reinforced by usage in Swahili, in which na means both "with" and "and." So:

Le Mia me le Iúli i si tule ne ka talo e (si?) suo a sihi = "Julie and Mia came home and ate some vegetables." I sort of like that.

(or would that be ...suo hu sihi? Jesus.)

12) That age-old problem with root-worthiness rears its head again. I have sahi in my lexicon as "wine." Does "wine" really need a root of its very own? Can I not have sahi mean "alcoholic beverage" or "booze in general" and then modify it either with suffixes or conventionalized adjectives? So "booze-wheat" would be "beer," "booze-grape" would be "wine," etc.? There's no immediate answer to this question, but it does raise the fact that we need to get back to work on the derivation system so issues like this can be more satisfactorily addressed.

13) To which I say, I was remarking to Adam not too long ago that I thought the only way an oligosynthetic language could really be spoken by humans would be if the juxtaposition of morphemes was more mnemonic than semantically logical. So we remember that "carrot" is "fire-vegetable," not because it makes sense, but precisely because it doesn't.

I've always had trouble imagining how I was going to use my mere 50 available CVs to effect every possible necessary lexical derivation. It occurred to me that I could just define all 50 of them with random but highly memorable semantics. Well, some of them would be more traditional, like diminutive and augmentative, and probably a few more. Disparaging, etc. But the others could be elements, like "earth," "fire," "water," "air"; colors, like "white," "black," "red," etc. Perhaps a thought. I'd like to discuss it with a few people before I make a final decision. Really, the question is whether, if I go down that road, I'm utterly casting away any illusion that this could work as an IAL.

14) Directionality with verbs of motion. How does this work. I'm conditioned by the Slavic languages to really like directional prefixes. But we don't have a whole hell of a lot of CV morphemes remaining, and I sort of begrudge their use in this case. One thought I had was that we could use serial verbs for this sort of thing, reminiscent of the ol' Spanish "entró en la casa corriendo" type. So

Hi si entra* colo ne ka talo or hi si entra* mo colo ne ka talo or hi si entra* ne ka talo mo colo

If I were really trying to serialize my verbs à la Yoruba, Thai, etc., it would look something like hi si entra* ne ka talo colo, but this is too easy to confuse with talo colo = "?house of running?" I would need to mark that morpheme as a verb, but if I do so aren't I breaking typological rules about serial verbs? I'll have to give this more thought, but it's looking like mo may be my best bet.

Be aware, though: is it okay that mo kita* = "guitarly," mo colo = "runnerly"? "He entered the house in the manner of a runner?" Is there going to be a separate word that means "as/like?" And wait: mo kita = "in the manner of a guitar?" Er. This is why we need that rigorous definition I keep talking about.

Before I go to bed, which I really need to do right this minute, there could conceivably be a difference between mo colo and mo a colo, as mo is a particle and not a specifier and can therefore be followed by articles.

Okay. All for tonight.