accusative and ergative languages

[Larry Trask]

Pat Ryran writes:

> Now, when Larry recently quoted Dixon about the nature of the
> ergative, he conveniently neglected to mention that Dixon
> acknowledged that there were currently practising linguists --- not
> amateur linguists like myself --- still defending the passive
> interpretation of ergative constructions.

No; this is not so.  Read p. 189 of Dixon's 1994 book.

> I asked Larry where he had "shredded" this interpretation, and to my
> knowledge, got no answer. If I missed the "shredding", perhaps you
> will be kind enough to rehearse his performance for us. I have seen
> nothing by Larry's vehemence and your allegiance to support the idea
> that the ergative should not be interpreted as a passive.

Mr. Ryan, I'm afraid I lack the time to reply in detail to all, or even
most, of the 37 or so postings from you that greet me each morning.

But briefly: if a transitive sentence in an ergative language were
"really" a passive, then its absolutive NP would be the subject and
would exhibit subject properties.  But this not the case in most of the
ergative languages I have heard about.  In spite of the ergative
case-marking and/or verbal agreement, it is usually the *ergative* NP
which exhibits subject properties, not the absolutive NP.

Take any ergative language you like.  Assume that the subject of an
intransitive sentence is the sole (absolutive) NP in it.  Examine the
syntactic properties of that NP, and tabulate them.

Then look at transitive sentences.  Tabulate the syntactic properties of
the absolutive and ergative NPs, and compare these with the preceding.

In the great majority of ergative languages, it is the ergative NP which
shares the subject properties of the intransitive subject.

A few examples of typical subject properties:

The subject controls reflexive and reciprocal NPs.

The subject cannot itself be reflexive or reciprocal.

The subject controls the empty NP in an empty-NP complement.

The empty NP in an empty-NP complement is itself a subject.

A subject can be coordinated with another subject.

And so on.  There are also various language-specific tests.  For
example, in some varieties of Basque, the object of a gerund (but not
the subject of a gerund) goes, exceptionally, into the genitive case.
Neither the absolutive subject of an intransitive gerund nor the
ergative subject of a transitive gerund can be genitivized, but the
absolutive object of a transitive gerund can be.

One more test from Basque.  In Basque, the subject of an intransitive
sentence cannot be reflexive or reciprocal.  So, Basques can say,
literally, `Susie and Mike [ABSOLUTIVE] were talking to each other', but
they cannot say *`Each other [ABSOLUTIVE] was talking to Susie and
Mike'.

Now, in a transitive sentence, Basques can say `Susie and Mike
[ERGATIVE] slapped each other [ABSOLUTIVE]', but they cannot say *`Each
other [ERGATIVE] slapped Susie and Mike [ABSOLUTIVE]' -- or, in terms of
the discredited passive theory, they cannot say *`Susie and Mike
[ABSOLUTIVE] were slapped by each other [ERGATIVE]'.

In an intransitive sentence, the absolutive subject cannot be reflexive
or reciprocal.  In a transitive sentence, the ergative subject cannot be
reflexive or reciprocal, but the absolutive object *can* be.

All tests for subjecthood in Basque give the same result: it is the
ergative NP in a transitive sentence which shares the subject properties
of the absolutive NP in an intransitive sentence, and therefore it is
the ergative NP, and not the absolutive NP, which is the subject of a
transitive sentence.

There are unusual languages, of course.  In Dyirbal, and in
Nass-Gitksan, the absolutive NP in a transitive sentence shows at least
some subject properties, while the ergative NP does not show the same
properties -- though it does show other subject properties.  This is the
phenomenon we call `syntactic ergativity'.  But I know of no language
which is syntactically ergative without exception, and I know of few
languages which are syntactically ergative at all.

The "passive" view of ergative languages in general is indefensible.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk