Re: ergative? I don't know...

>        1. Agent (AGT): entity which performs an activity or brings about
>a change of state (e.g. the crowg-AGT applauded.)
>        2. Patient (PAT): entity which- is viewed as existing in a state
>or undergoing change; viewed a slocated or moving; viewed as affected or
>effected by an entity (e.g. the sky-PAT is blue, the cat-PAT is in the
>house, or the bird ate the worm-PAT)
>        3. Experiencer (EXP) entity which experiences as emotion or
>perception OR an Agent (fm above) which acts unitnentionally (e.g.
>They-EXP love music or He-EXP fell (on accident)).
Here's the standard analysis, as I understand it (and for some reason I had
a very hard time understanding it).

Most now divide noun phrases into _core_ and _non-core_ arguments. Non core
arguments are things like locations, destinations, purpose, etc that
represent "indirect objects" of the verb. They are almost always optional,
and their functions vary pretty widely across languages (though they are
not without common features).

There are 3 basic functions that are found in sentences of human languages:

A (most agent like argument, subject of an transitive verb), P (most
patient-like argument, Object of an intransitive verb), S (Subject of an
intransitive verb). S may be close to your "Experiencer", though that term
is also used for something related but diffrent.

These are often marked somehow by morpholoy or syntax, into cases:

i. nom/acc system:

Nominative  (A, S)
Accusative  (P)

ii. ergative/absolutive system:
Ergative    (A)
Absolutive  (P, S)

iii. active system
Agent       (A)
Patient     (P)
Subject     (S)

These are the most basic systems. Variations include using different case
systems for nouns in a sentence and Argument morphemes in the verb (for
instance nom/acc noun cases paired with ergative/absolutive clitic
pronouns), different case systems depending on the tense of the verb or the
presence of some auxiliary: e.g erg/abs in the future, but nom/acc in
non-future.

Some active (and ergative languages) may differentiate "agent-worthiness"
in the S argument, so that intransitive sentences like "John kills" and
"the door closes" use different cases to reflect the difference in
agenthood.

Some languages use one system (ergative or accusative) in one tense, and
another in another tense (e.g. past tenses erg/abs, non-past nom/acc), some
use one system on bound verbal pronouns, and the other on fully expressed
nouns. These are the "split" systems.

Many languages have a hierarchy of agenthood something like this (It's a
postulated universal, but my memory may fail me on some of the later items):

1p > 2p > 3p > animate (present) > animate (absent) > inimate (present) >
    > inanimate (absent)

This may affect the choice of case, based on some dividing line between
agentive and non-agentive arguments (pick a line on the above, and you've
got a possible system for deciding whether something is an agent).

A few languages don't use the syntactic functions at all and instead mark
whether the agent is higher or lower on the relvance hierarchy than the
patient. this is a rare type, but is found in some Algonkian languages (is
that the proper modern spelling?).

The list of cases applied to other semantic roles, like location,
experiencer,  destination, purpose, etc. is variable. Frequently such cases
my be used with special verbs, or in place of a main case to reflect an
agenthood difference. S functions and P functions seem to be more prone to
this from what I can tell. Other cases are not infrequently required in
special constructions of some form or other.

There's a good explanation of this stuff with examples in Payne's book, and
Jack had a copy of longer screed I wrote on this (with citations). I
suspect this post is clearer than that one was, though. I don't think it's
made the FAQ yet, but it's on findmail.

Then theres' the trigger system, where several cases are in use, but one
nound is always found in the "trigger" case: the verb is then inflected to
mark what role that trigger is playing (agent/patient/experiencer/location,
etc...) from some closed list of roles.

These all interact with passivization and causatives, of course, as they
change the number (and syntactic assignments) of core verb arguments, as
well as the implications of agenthood in the verb. For instance an
absolutive subject with a  transitive verb can indicate an anti-passive
(the book reads well). That's where it gets complicates (and interesting).

Anyway, that's my current take on it.

  -- David

_________________________________________
David Durand              dgd@cs.bu.edu  \  david@dynamicDiagrams.com
Boston University Computer Science        \  Sr. Analyst
http://www.cs.bu.edu/students/grads/dgd/   \  Dynamic Diagrams
--------------------------------------------\  http://www.dynamicDiagrams.com/
MAPA: mapping for the WWW                    \__________________________