Re: lohe, lehe & ka

[Jorge Llambias <jorge@PHYAST.PITT.EDU>]

la lojbab cusku di'e

> ke'a would not work for the proposed marker because it already has an
> assigned meaning, and it is easy to envision conflict in that meaning
> (unless I am missing something).   What happens when the reference is
> inside a relative clause and is NOT the relativized pronoun.

Same thing that happens when there is one relative clause inside another,
we have to resort to subindices.

Fortunately, {ke'a} is only rarely needed, and such embedding is even
rarer, so the problem wouldn't arise much in practice. The solution
(indexing) is a bad one, but it is sufficient for a problem that doesn't
seem to appear in practice.

> Composing
> on the fly something like:
> lo nanmu poi ganse leka le rozgu cu se panci
> I don't even see a way to use ke'a in the abstraction, and if it did, it would
> refer to the man.

I don't see what ke'a would mean there either, which makes me wonder
why that place needs a property. Does the man sense the smell or the property
of the rose having a smell? Is it a property of the rose, that he senses?

In any case, there is no confusion in that example, since there is no place
to use ke'a.

> Most of the gismu that have abstractions seem to be such that the focused
> place in the property abstraction is an echo of x1.  If you have
> a relative clause such that the relativized pronoun is NOT that x1, then
> you have a conflict in use of ke'a.

The property and the relative clause are at different levels, so it can
be disambiguated with subindices, if ever it is needed.

Jorge