Re: Agents and patients

>I'm developing a new language for an RPG setting
>(Wamen and G'amah were for this one too). I thought
>of having agent/patient, with a twist regarding
>perception and non-volitional verbs.
>
>The usual pattern is subject-AGT, object-PAT; but
>for these verbs one might have subject-PAT, object-ABL
>(ablative):
>
>    3p.PAT house.ABL see
>    'They see a house.' (*'To them from a house is seen.'?)
>
>I don't know if this is likely to happen in a natlang.
>I know Tokana has a lot of possible cases for subjects,
>for example, dative for perception verbs, which is somewhat
>common patten. But this is quite different... The case names
>are maybe misleading (also see below).
Since you only have four cases--and will presumably use
each case to encode a variety of semantic roles--then *any*
nomenclature you choose will be misleading, as I discovered
myself for Tokana (cf. the recent exchange over what to
call the Tokana case that encodes "for", "about", and
"towards").  So I wouldn't worry about that.

As for your case-marking pattern for perception verbs,
I like it.  I don't know of any natlangs that do things this
way, but your scheme seems perfectly reasonable.

The tricky thing about perception verbs is to figure
out which participant is more "patient-like", the
experiencer ("they") or the stimulus ("the house").
In Tokana, I chose to associate the stimulus with the
patient role, and to express events of perception
using a metaphor whereby the stimulus (absolutive
case) becomes located within the mind of the
experiencer (dative case).  You made the opposite
choice, to associate the *experiencer* with the
patient role, and hence you have a slightly different
metaphor:  The stimulus is the source (ablative case)
of a change of state in the experiencer (patient case).
Both metaphors seem reasonable to me.

>What I'm not sure about here is what to do with the
>verb -- agreement would be marked in this language, but
>I'm not sure with which part of the sentence...
That would depend on how case is marked in your
language.  In languages where only one noun phrase
per clause agrees with the verb, agreement is often
(although not always) with the noun phrase bearing the
unmarked case (typically nominative in nom/acc
languages and absolutive in abs/erg languages).  Of
course, in many languages the verb agrees with two,
even three arguments in the clause...

Matt.