Re: Disambiguation of argument reference

In a message dated 10/8/02 8:36:21 PM, butsuri@MYREALBOX.COM writes:

>I've been trying to reform (or perhaps more accurately "form", as it's
>never really approached completion) the syntax of my conlang LC-01.
>I'm considering making it VSO now, although there are still some
>attractive qualities to an SOV scheme.
>
>What I'm thinking about presently is basic argument structure, without
>any complications of adjectives, adverbials or relative clauses.  With
>full case marking, you might think this would be simple.  However
>it's in the nature of LC-01 morphology (as I conceive of it) that any
>noun can be seen as a nominalization off a verbal root, and can thus
>take an array of arguments of its own.  (It's because I'm fairly
>committed to a Head-Modifier NP structure that I'm thinking of VSO,
>because arguments to a noun really have to be considered modifiers.)
>If these're simply marked for case, ambiguity can arise as to to what
>head a noun is an argument.
>
>A number of possibilities occur to me: Inflect case tags for level of
>reference, either absolutely or relative to the previous word;  or use
>postpositions and place them only after the entire NP to which they
>refer is completed.  These seem rather unnatural, though. Certainly
>postpositions don't co-occur with modified-modifier NPs much
>naturally, and this scheme would leave the possibility of having to
>end a phrase with a string of them - I'm not sure it'd even be
>speakable in complex sentences.*
I wonder if that first possibility is so bad, really. A verb in a main clause
can be marked one way, and verbs in dependent clauses may be marked in
another way (e.g., with some type of participle or non-finite form).
Especially since in your language nouns are already so similar to verbs, it
might make sense to have a parallel system for nouns. So you might have
(using prepositions, which would certainly work better ... who wants center
embedding? [isn't it a universal that VSO langs don't have postpositions?]):
want-to-take I wo goer
             = I want to take the one who went
want-to-take I wo goer to conference
             = I want to take (the one who went) (to the conference)
want-to-take I wo goer to conference-DEP
             = I want to take (the one who went (to the conference))
Of course the dependent marker could to the end, beginning, or middle of
either the head or dependent of 'to conference', or fuse with a case marker,
etc. whatever you like. But all you need is one extra morpheme.

>How does English deal with this?  Such ambiguities are possible,
>things like "I want to take the one who went to the conference".  But
>they don't seem to crop up much.  We don't use arguments to
>nominalizations much, we relativize them instead, and make their core
>arguments oblique, often with prepositions that don't make any sense
>to the main verb.  Of course, if they weren't they wouldn't have any
>case marking at all.
>
>A lot can be done with mandatory marking of transitivity on heads, but
>this only really helps us with core cases (and required oblique roles,
>maybe**).  Likewise gender or person agreement.  Having a variety of
>specialized adpositions helps too, but there will always be the
>possibility of ambiguity - there are some oblique arguments (locatives
>etc) which can apply to any verb, and it's possible that one of the
>arguments to a verb will be a nominalization of that verb.
>
>Well, excuse me for babbling here.  I hope I haven't said anything
>hopelessly wrong-headed - just thinking out loud.  Anyone have any
>thoughts on the matter?
Eloshtan is similar in that any noun can be seen as a nominalization of a
verb, though the arguments it can take are basically limited to core ones.
The word order is SOV. I ran your sentences through it (simplifying
'want-to-take' to 'take', and borrowing the word 'conference') and came up
with:

halo (joc) kanforo klaf qejjimt
goer (ACC-3) conference to-4 took-I   = I took (the one who went) (to the
conference)

kanforo klac halo (jof) qejjimt
conference to-3 goer (ACC-4) took-I   = I took (the one who went (to the
conference))

Firstly, the differences between joc and jof (which are omissible), and klac
and klaf, reflect the order of the nouns; if they were in a context where one
of the nouns had been mentioned already, the markings could be reversed.
Secondly, there is no ambiguity, because of the strictness of the word order.
Modifiers must come before nouns, so the second sentence could not be
expressed any other way. On the other hand, direct objects must come before
indirect objects, so the first sentence is not expressable any other way.

Well, though, Eloshtan isn't perfect. The following sentence is ambiguous:

kyemme joc ora klaf remt.
book ACC-3 thief to-4 gave-I   = I gave a book to the thief OR I gave to the
book thief (with an omitted DO)

kyemme ora klaf remt.
book thief to-4 gave-I   = I gave a book to the thief

This last sentence is unambiguous, because the accusative postposition can
only be dropped if it is an argument of a verb.
One more thing though: The first sentence can be clarified by adding the
marker -l to 'ora'. This affix merely shows that the noun is modified. It is
mandatory when there is an adjective, and optional for other modifers.
However, the ambiguous form would be more common, since the more common way
to express "I gave a book to the thief" is with the other form, where the
postposition is omitted.

Anyway, that's a long way of saying that if you play around with word order
rules and omissions, you can disambiguate a lot of the time, if not all of
the time. If you had rules like Eloshtan's (though mirror image) for LC-01,
you could have the unambiguous:

want-to-take I conference to goer wo
             = I want to take (the one who went) (to the conference)
want-to-take I goer wo conference to
             = I want to take (the one who went (to the conference))

And remember of course, there's nothing wrong with SOME ambiguity (unless
your language is supposed to be a loglang....); context can do wonders. Isn't
there some language where subjects and objects aren't even distinguished,
neither by cases nor word order nor verbal affixes, so the speaker just has
to figure out who's doing what? (Anyone know what this language is? I think
I've read about it but not seen the name, so I hope it actually exists.)

You might want to search the conlang archives for "vso nightmare"; there was
a similar discussion almost exactly a year ago....

>*Something like this, with wo as an accusative postposition:
>want-to-take I goer wo
>             = I want to take the one who went
>want-to-take I goer wo conference to
>             = I want to take (the one who went) (to the conference)
>want-to-take I goer conference to wo
>             = I want to take (the one who went (to the conference))
>
>** Are required prepositional arguments in English, like the
>   destination of "put", considered to be core or oblique?
Probably core, but I don't know for sure.


Josh Roth
http://members.aol.com/fuscian/home.html