Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site sphinx.UChicago.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxt!houxm!ihnp4!gargoyle!sphinx!mmar
From: mmar@sphinx.UChicago.UUCP (Mitchell Marks)
Newsgroups: net.nlang
Subject: Re: Query on ergative languages
Message-ID: <853@sphinx.UChicago.UUCP>
Date: Sat, 20-Jul-85 19:37:23 EDT
Article-I.D.: sphinx.853
Posted: Sat Jul 20 19:37:23 1985
Date-Received: Mon, 22-Jul-85 07:20:45 EDT
References: <6312@ucla-cs.ARPA>
Organization: U Chicago -- Linguistics Dept
Lines: 84


	There has been a lot of ugly disagreement on just how to describe
what's special about ergativity.  So I'll just say, not that I think
Rich Wales's way of describing it is wrong, but that there are other
descriptions which I find easier to use.  He wrote that ergative languages
use an agent/patient distinction while nominative-accusative languages use
a subject/object distinction.  From my perspective, the former is semantic
and the latter is syntactic, and all languages use both; one dimension of
difference among languages is how they handle the mapping from one set of
roles to the other.  (There is some justice to saying that classic
highly ergative languages must have a very different notion of subject from
classic nom-acc languages, and defining these syntactic roles is tricky
and should be done language by language.)
	Let's start off with a very unhelpful definition:

	An ERGATIVE LANGUAGE =df a language with a large number of
	ERGATIVE CONSTRUCTIONS or PROCESSES.

This entirely leaves open what that threshold level would be; the idea is
not so much that you can draw your own line for saying which languages are
or are not ergative as a binary feature, but rather to encourage talking about
languages as "highly ergative", "slightly ergative" etc in a scale.  Then
the real work has to be done in the explanation of "Ergative Construction."

	Let 'S' stand for "the single term (noun phrase) in an intransitive
sentence".  It mnemonically plays on the notion 'subject' but doesn't require
it as a more primitive notion.  
	Let 'A' stand for "the term referring to the agent in a
transitive sentence", and let 'O' stand for "the term referring to the patient
or goal in a transitive sentence".  These *do* depend on more basic notions.

	Then: An ERGATIVE CONSTRUCTION or PROCESS =df one which treats
	S and O similarly, and in contrast to A.
	A NOMINATIVE-ACCUSATIVE CONSTRUCTION or PROCESS =df one which treats
	S and  A similarly, and in contrast to O.

            ERGATIVE                                    ACCUSATIVE

               S                                             S
                \                                           /
                 \                                         /
                  \                                      /
        A           O                                   A         O

The most important area for ergative vs nom-acc constructions is in positional
or case marking of central terms in root sentences.  Originally, this was
all that was meant in the basic distinction of a language being ergative or
not, and is still what is usually meant by an unqualified statement that
such-and-such "is an ergative language".
	English has only remnants of a case-marking system, but a pretty
rigid positional system.  We say
			He ran.
			He hit him.
so English is a nominative-accusative language, to the extent that the
two-way distinction applies here at all.  It would be ergative if the
pattern were
			Him ran.  or better yet:  Ran him.
			He hit him.

	But, as I was urging in the general definition, there are
other areas of ergativity.  These can be found even in a language which
is mostly accusative.  For example, those who know French may want to
take a look at what happens in the faire-causative.  My French is not
very good, so there will be some mistakes in this example, but I hope
it will illustrate the point anyway.

	1.  Jean mange.   (john eats)
	2.  Jean mange le pain.   (john eats the bread)
	3.  Je lui fais manger (`a Jean).   (I make/have him [John] eat.)
	4.  Je le lui fais manger.  (I make/have him eat it.)

I actually need another couple of sentences, to show that in (4) the 'le'
pronominalizes 'Jean' and the 'lui' pronominalizes 'le pain' -- better
choice of the food, for gender or number, would have done it.  
	Now, note that 'Jean' is the S in (1).  'Jean' is the A and 'le
pain' is the O in (2).  Sentences (3) and (4) embed (1) and (2) respectively
in the causative construction (or they would, if the objects were spelled
out).  Notice that the S and the O went to 'lui' while the A went to 'le'.
Thus this qualifies as an ergative construction, though we would hardly want
to call French an ergative language in general.
-- 

            -- Mitch Marks @ UChicago 
               ...ihnp4!gargoyle!sphinx!mmar