accusative and ergative languages

Ralf-Stefan Georg Georg at home.ivm.de
Thu Jul 29 10:47:45 UTC 1999


>Pat responds:

>I suggest you read it. "But we do still encounter scholars who insist that
>there is a necessary diachronic connection, e.g. Estival and Myhill
>(1988:445): 'we propose here the hypothesis that in fact all ergative
>constructions have developed from passives'."  Are you suggesting that
>Estival and Myhill are not "linguists", or that Shibatani, in whose book
>this essay appeared, is not a "linguist"?  The quotation above is from p.
>189 of Dixon's 1994 book. What are you playing at?

Linguists can do wrong.

>Pat responds (on L.T's catalogue of subject properties):

>Why do you not give us an example of these so-called subject properties in
>some ergative language besides Basque? And how were these properties
>selected?

Empirically. What Larry gave was a *selection*, as you correctly name it.
It is so easy, you can do it yourself. S.b.

>Pat responds:

>If you are referring to the "tests" above, you have proved nothing.

This is much in line with what I said in  today's other post of mine. You
see, you read, you understand (or don't), but you reject nevertheless.
Always.
The subject-property-test *is* sufficient to discriminate between ergative
and passive constructions. Again:
Take an intransitive sentence in an exclusively accusative language. S-V
S is, by convention, called "subject" (just a convention). It's semantic
role can be one of the following: Agent, Experiencer aso. (some people
insist on Agent being restricted to agents of transitive constructions, bog
s nimi).
Now, take a transitive one. A-V-O
What accusative languages do is to treat S *and* A alike under all
circumstances (there are such languages, i.e. split-free accusative-lgs.).
This leads to the notion that these languages know a macro-category
comprising S and A, which most conventionally is called "subject" also, in
compliance with western grammatical tradition.
Now subjects have properties of various sorts, part of which have been
enumerated by Larry Trask. The real list is longer, and you can expand it
yourself. It is nothing more than a list of things which are generally true
for "subjects" in these languages.
Now, part of the things a passive formation does to a sentence is that the
*subject* is now what was semantically the undergoer/patient (choose your
favourite term) of the corresponding active sentence, an example:

active: Stefan proves Pat wrong (active, A-V-O)
           Pat is proven wrong by Stefan (passive, S-V[pass]-Agent-periphrasis)
"Pat", in this passive sentence, has been moved to a position (not only in
terms of word order), where it assumes all the functions commonly found
with S. It has become the "subject" of the sentence, according to
traditional terminolgy. The syntactic properties (partly enumerated by
Larry, but the list is longer) found with "Stefan" in the active sentence
are now found with "Pat" in its passive equivalent.

Now let's look at ergative constructions:

Again, we find intransitive sentences there: S-V (gosh, I'm starting from
scratch, but I think I have to)
And transitive sentences as well: A-V-O. What makes the construction
ergative is, of course, nothing but the fact that S and O are treated alike
in this language (or in some subsystems of this language).
Now, in a morphologically ergative language, such a sentence may look like:

A-erg V O-abs
a passive sentence, we recall, looks like (word order irrelevant): S V
Agent-periphrasis (P. is proven wrong by S., to take a random example with
positive truth-value, however;-).
Of course, ergative constructions and passives bear some resemblance: in
both the (semantic !) Agent is morphologically marked, in both the semantic
patient may be morphologically unmarked (this does not work when there is
an overt accusative marker in the language, of course).  However, this is
where the resemblance ends. In an ergative sentence like:
Stefan-Erg proves Pat-Abs wrong we should expect, under the
ergative-as-passive scenario, that subject properties remain with Pat (as
in the "true" passive above); however, they rest with "Stefan".
This is the general picture only, and the test has to be carried out with
real "ergative languages" of course, but this is what you will find. It is
nothing less that the watershed betweed ergative constructions and passive
constructions. It does not matter whether you accept the full list of
"subject properties" enumerated. You may accept some and dislike others.
With those you do accept, though, you will find this picture. The more
subject properties you can bring yourself to accept, the more instances of
the difference between ergative and passive constructions you will find.
It's that easy. But you do have to look at complex constructions, not only
at the minimal sentence.
This does *not* mean that ergativity in some given language may not have
come about via a passive transformation. It may. In order to claim the
right of being accepted as a "real" ergative construction (and not the
passive it may have been in an earlier life), the shift of subject
properties to the Agent-phrase (the ERG-NP) from the Patient-phrase is
crucial.
For a succinct demonstration that, what looked to some people like an
ergative construction, but is rather to interpreted as passive, see my (and
A.P. Volodin's) "Die itelmenische Sprache", Wiesbaden 1999, though I admit
that we could have been a bit clearer on the issue. If anything, this
thread has prompted me to write a clearer exposition of the relevant
chapter in this book.

>Pat responds:

>Obviously, I do not think so. And Estival and Myhill (and probably
>Shibatani) do not either --- not to mention the majority of linguists of the
>past.

Estivall/Myhill/Shibatani should speak up for themselves. As for the
linguists of the past, well, they are linguists of the past, if they had
got everything straight already in 1890, what the hell are we doing here (I
assume, that's what you ask anyway reading all this ;-).

Stefan

Stefan Georg
Heerstrasse 7
D-53111 Bonn
FRG
+49-228-69-13-32



More information about the Indo-european mailing list