Tuesday, February 19, 2019

The Grand Unified Theory of dependent clauses

In honor of Koa's impending 20th anniversary, I hereby announce a concerted effort to figure out, once and for all, what is going on with dependent clauses: or in Koa-specific terms, as I now realize, clauses used as predicates. This includes nominal (or content) clauses, adverbial clauses, relative clauses, any situation in which a clause is being integrated into syntax in other than a free-standing way.

Further: my instinct going into this is that it's going to turn out that all these various "types" as our traditional grammars consider them actually break down to pretty much identical Koa structures, in the same way that Koa collapses the difference between nouns, verbs, adjectives, adverbs, etc. as IE languages conceive them.

I think the current state of dependent clauses in Koa is represented here, and with much greater sophistication than I had realized before I reread it. This gets us pretty close to a model that can be expanded and adapted to serve all needs coherently and transparently, but there are one or two apparent possible IE holdovers that need to be reconsidered. A summary:

  • Clauses can be made non-finite by deleting the 3rd-person finiteness marker i.
  • Nominal clauses can be preceded by ko (non-finite), or by ve (finite).
  • Adjectival clauses can be internally-headed with the head marked by ke (finite), or optionally preceded by u and gapped (non-finite).
  • Adverbial clauses can be preceded by ko (non-finite), or by ve (finite), following whatever other marking indicates the role of the clause.
  • Verbal clauses — which is to say basically clauses used as stative verbs — are simply non-finite.

Here's the thing: what we're talking about here at base is clauses being used as predicates. In all other cases in Koa, there is an extremely solid system for the use of a predicate as a noun, verb, adjective, adverb, etc., and this differs from the approach discussed above! For example, most particularly, why are adjectival clauses marked differently than nominal clauses?

Clearly, in Indo-European languages, adjectival (i.e. "relative") clauses are treated entirely differently than nominal ones, and that has equally clearly affected my thinking about this topic from the very beginning. Is there in actuality any meaningful way, however, that a relative clause can be said to be "different" from a nominal clause beyond any other predicate when used as an adjective versus a noun?

After some serious brain-twisting thought, and some delightfully old-school conversation with Allison, I have to conclude that the answer is actually no. Despite the way that this screamingly defies my IE intuition, I can't muster a single argument for treating them differently in Koa. Turkish, incidentally, agrees:

Aygül-Ø ekmeğ-i ye-diğ-imiz-i söyle-di-Ø
Aygül-NOM bread-ACC eat-PART-1PL-ACC say-PAST-3P
"Aygül said that we ate the bread"

ye-diğ-imiz ekmek
eat-PART-1PL bread
"the bread we ate"

Translating literally into English, the dependent clause here — yediğimiz ekmek or ekmeği yediğimiz — has a feel sort of like "the our-eating bread": "Aygül said the our-eating bread," etc. Traditionally, and as described by the principles above, these would be translated into Koa in distinct ways:

Áikulu i sano ko/ve [ nu suo ka lepa ]
Aygül FIN say SPEC/COMP [ 1PL eat DEF bread ]
"Aygül said that we ate the bread"

ka lepa (u) [ nu suo ]
DEF bread (REL) [ 1PL eat ]
"the bread we ate"

For the life of me, though, I just cannot rationally motivate the presence of u above. Why not instead:

ka lepa (ko/ve) [ nu suo ]
DEF bread (SPEC/COMP) [ 1PL eat ]
"the bread we ate"

Depending on whether we're calling the dependent clause finite or not, the literal translation comes out sounding something like "the bread of our eating" or "the bread such that we ate it." Either way it looks like very solid Koa of the usual variety in which there is no difference between traditional lexical classes! And as such we can apply this equally well to simple predicates:

ka sivu vihe
DEF leaf green
"the green leaves"

ka sivu ko/ve vihe
DEF leaf SPEC/COMP green
"the leaves that are green"

It appears that u is out of a job.

If this is so, we can rearticulate the principles of dependent clauses as follows:

  • Clauses used as predicates have an optional marker ko when non-finite, a mandatory marker ve when finite.

And that, ladies and gentlemen, is the way we like it in Koa. Here are some examples of complex predicates being used in different roles...

ADJECTIVAL: "a mommy-not-feeling-well night"

ivo (ko) [ ka mama na mai koa ]
night (SPEC) [ DEF mom NEG feel good ]

ivo ve [ ka mama i na mai koa ]
night COMP [ DEF mom FIN NEG feel good ]

NOMINAL: "she said that mommy's not feeling well"

ta sano (ko) [ ka mama na mai koa ]
3SG say (SPEC) [ DEF mom NEG feel good ]

ta sano ve [ ka mama i na mai koa ]
3SG say COMP [ DEF mom FIN NEG feel good ]

VERBAL: "this night is mommy-not-feeling-well-ish"

ti ivo i [ ka mama na mai koa ]
DEM night FIN [ DEF mom NEG feel good ]

ti ivo i ve [ ka mama i na mai koa ]
DEM night FIN COMP [ DEF mom FIN NEG feel good ]

ADVERBIAL: "while mommy's not feeling well"

he (ko) [ ka mama na mai koa ]
TIME (SPEC) [ DEF mom NEG feel good ]

he ve [ ka mama i na mai koa ]
TIME COMP [ DEF mom FIN NEG feel good ]

Adjectival clauses do still have a third, internally-headed option that I don't think is extensible to the other types:

ka talo (ko) [ le Iako si ete ]
DEF house (SPEC) [ QUOT Jack ANT make ]
"the house that Jack built"

ka talo ve [ le Iako i si ete ]
DEF house COMP [ QUOT Jack FIN ANT make ]
"the house that Jack built"

le Iako i si ete ke talo / ke talo sa le Iako i si ete
QUOT Jack FIN ANT make QU house / QU house FOC QUOT Jack FIN ANT make
"the house that Jack built"

Wait. Why does this exist, again? Just for fun? Greater trans-typological intuitiveness? Now that I'm looking at that, it doesn't seem to make any sense at all. I keep wanting to reduce it to ke talo le Iako si ete, as in ke talo le Iako si ete i sulu "the house that Jack built fell down." That agrees with my IE intuitions. But what in God's name is the difference between that and ka talo le Iako si ete i sulu as described earlier? And again, why should relative clauses be treated differently than other dependent clause types?

Okay, I reserve the right to bring this back should circumstances seem to require it at some point in the future, but for the moment I see absolutely no motivation for its continued existence.

A few final points:

1) For the type of clause that in English winds up headless, like "I want to know who said that," Koa syntax using the above principles gets a little muddy. It could be non-finite:

ni halu ilo (ko) keka sa sano toa
1SG want know (SPEC) who FOC say that

...which raises the heretofore never considered question of whether a clause with a focused constituent can be nominalized: an issue for another day. Sa could be left out of the above clause anyway without changing the semantics. In any event, that seems vaguely acceptable if perhaps a little IE-calquish; on the other hand the theoretically equally allowable finite version feels fairly bizarre:

ni halu ilo ve keka sa sano toa
1SG want know COMP who FOC say that

Something seems frankly wrong with this but I can't puzzle it out at the moment. Should ve be le? I don't know. HOWEVER, the point I want to make is that I think the most natively Koa way of expressing this actually avoids all of these structures entirely:

ni halu ilo ka sano toa
1SG want know DEF say that

Literally this is something like "I want to know the sayer of that," and it's a lesson that we really have to remember that Koa can do this and explore the scope and power of it more fully (or at all, really).

2) What's the pragmatic difference between finite and non-finite dependent clauses? Why choose one over the other? I think it's good to have both options, but at this point I don't have any sense of comparative uses.

3) I've said that the marker for non-finite dependent clauses is optional, but under what circumstances would/should it be omitted and retained? It's clearly going to have an impact on clarity and style. Part of me is wondering whether it should actually be optional at all. More later.

4) In the earlier post I referenced above, I was fretting a bit about te tai ko... types of clauses. Just to reassure my past self, I don't think there's any issue here, and we can phrase them either finitely or non-finitely:

te tai ko/ve ta lu lai he leo
ABL be SPEC/COMP 3SG IRR return TIME today
"maybe she'll come back today"

On the other hand, the way this is structured makes it look like a verbal structure like te tai is transitive, in that it accepts a complement! What's actually happening here syntactically? Apparently we can also say things like

koa ko/ve ta lu lai he leo
good SPEC/COMP 3SG IRR return TIME today
"it's good that she's coming back today"

Is this something other than fronting the predicate (as in, is the pre-transformation form really ko ta lu lai he leo i te tai)? Can we do that? When?

?iso to kunu
big DEM dog
"it's big, that dog"

?ma polo la koto ni
DUR run 1SG to home
"here I am running home..."

Should this be focus instead, like te tai sa (ko)/ve ta lu lai he leo? Or is something completely different going on? Um. Let me get back to you.

No comments: