Re: First thoughts on Imperial

Ian Spackman wrote:

<<<<
Basic idea:
Different classes of noun have different inherent cases, and are marked
*only* when used in a different function.  So there is likely to be an
animate->agent class, an inanimate->patient class, a place->locative class
(and maybe even a container->inessive class), a tool->instrument
class.  Hm, any others?
<<<<
Yes, done that, but not with arguments per se, only with nouns used as
tagging arguments/PoS. Actually, allnoun conlangs do that: they ascribe a
role to each noun mimicking its semantic definition: for instance the sememe
"agent" is used as the grammeme "agent". (well it's a bit different from
your exact question but i think delving your issue leads to that point ;-).
My conlang does the same with an "expletive" preposition "ai". Example:
"mimitangunya u parisa ai kitari" means "be-written-up @ agent that_is man"
= "to be written BY the man". The preposition "u" is a dummy circumstancial
one (locative, temporal or notional) and the preposition "ai" shows that
"agent" becomes the agentive tag of its complement "man" as expected its
semantic definition. I've read plenty of natlangs and conlangs doing that
too.

<<<<
Question: does this occur in natlangs?  (I think the answer is yes, or at
least something similar occurs, when it comes to the first two
classes.)  And if not, does it at least seem plausible?
>>>>
It does in a way in English "I ate yesterday inside the house" where neither
the "(yester)day" or the "inside" are potential DOs. Indonesian does that
all the time: "sebab" means "cause" and "because", "untuk" is "goal" and
"for/in order to", etc. Japanese needs combine such words with extra clitics
"ni", "to", "de", "kara" but the idea is the same. That's how I came to
realize that there isn't ten or twenty "natural" cases/roles/whatever in a
language but a as many as there are words in that language, and many more,
although you can screen and reduce them into 1 to 60 main categories
according to the use you make of them.

<<<<<
It's often discussed that languages don't need such-and-such POS - there's
such a discussion going on now - but I don't think I've ever seen mention
of languages with more PsOS than English.
>>>>>
Semantic roles, PsOSF and tags are different things. From what i've seen in
the languages i know, the number of PsoS is the same everywhere because PsOS
are really as many as possible coreferences between the arguments, the
predicate and the clause combiend with each other into pairs of one entity
and one behaviour. What varies is the number of tags needed and the
categories of tags. For instance, plenty of natlangs have no class of
adjectives because adjectives are dealt with as a predicate put in
subordination. What changes is the number of tags (grammemes) used. Other
ex: "John can see the bird in the tree" vs. "John in the tree can see the
bird": "in the tree" is a complement (behaviour) of two different entities
but tagged here as only one PoS while some langs make a mandatory
distinction btw "John who is in the tree" (complement to argument or implied
subordinated copula/predicate) and "can see in the tree" (complement to
predicate) and even sometimes "John can see in the tree" (complement to the
whole clause).