In a message dated 1/18/03 5:53:58 AM, [log in to unmask] writes:

>On Fri, Jan 17, 2003 at 04:11:18PM -0500, Josh Roth wrote:
>[snip]
>> This doesn't seem quite right. I see trigger systems as not necessarily
>> an alternative to accusative, ergative, tripartite, etc. marking, but
>> merely a way of marking roles. An noun's role can be marked on the noun
>> itself, on the article (German???), on the verb (langs with trigger
>> systems), by word order, etc - this is separate from the issue of which
>> semantic/pragmatic roles are grouped together or separated.
>
>I'm not sure I agree with this description of trigger systems. You have
>to
>keep in mind that the noun chosen as trigger can be *any* of agent,
>patient, etc., only this chosen noun is marked differently from others.

But it's more complicated then that. You have to ask, how is this noun
treated, compared to how other nouns would be treated if they were the
trigger, and how are they treated when they're not the trigger? You can still
determine subjects and objects, and see if subjects of intransitive and
transitive verbs are treated alike, and how objects are treated. See
languages below.

>There is no "subject" role which always get picked as trigger; the point
>is that the chosen noun can be in *any* of these roles, yet it is treated
>differently from the others. So there must be some special property
>associated with that chosen noun---you can call it trigger, or focus, or
>what-have-you. But if we were to follow your analysis, we'd have to either
>mark the role of every noun in the sentence on the verb, or always mark
>nouns of the same role (say, agent) on the verb, which isn't what trigger
>systems do.

There is a special property of the trigger noun, but that property is on
another plane than the semantic and syntactic roles that ergative and
accusative classification are based on. According to my analysis, you don't
need every noun's role marked on the verb. See below....

>> Kar Marinam is trigger, but also split-S. I can also envision
>> fully-fledged ergative, accusative, etc. languages that happen to have
>> trigger systems.
>
>I'd be curious to find out how you can have an accusative trigger
>language.

I think this should illustrate what I mean. Below is a sentence from a
hypothetical language, based on Eloshtan:

pe-h       iri-g     kafagla
person-NOM paper-ACC touch

iri-h     isi
paper-NOM fall

You agree that the above is accusative? Now we'll make it into an accusative
trigger language. (NT=nominative trigger, AT=accusative trigger)

pe-t      iri-g     kafagla-hu
person-TR paper-ACC touch-NT

pe-h      iri-t     kafagla-gu
person-NOM paper-TR touch-AT

iri-t    isi-hu
paper-TR fall-NT

Here's an ergative language, for some contrast, first without triggers:

pe-h       iri-g     kafagla
person-ERG paper-ABS touch

iri-g     isi
paper-ABS fall

And now with (here, ET=ergative trigger, AT=absolutive trigger):

pe-t      iri-g     kafagla-hu
person-TR paper-ABS touch-ET

pe-h      iri-t     kafagla-gu
person-ERG paper-TR touch-AT

iri-t    isi-gu
paper-TR fall-AT

Compare that to the first language. The languages each have their own way of
grouping together S, A, and O. All the triggers do is displace the marking of
a choice argument.

>[snip]
>> My last point: you said that *all* non-trigger items are marked equally
>in
>> trigger languages. But in Tagalog, directions and beneficiaries are always
>> marked distinctly. Here are a couple of sentences from the book:
>[snip]
>
>I think he meant that non-trigger nouns have their roles marked on
>themselves, whereas the trigger noun has its role "moved" onto the verb.
>Hence, the non-trigger nouns are on the "same level" whereas the trigger
>is on a different level.

If that's what was meant, I agree of course.

>
>T
>
>--
>First Rule of History: History doesn't repeat itself -- historians merely
>repeat each other.

Josh Roth