accusative and ergative languages

Larry Trask larryt at
Sun Aug 1 12:25:27 UTC 1999

On Mon, 19 Jul 1999, Patrick C. Ryan wrote:

[on my list of typical subject properties in Basque]

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

I've chosen Basque because it is the ergative language I know the best,
and I can speak with some authority here.  But most other ergative
languages I've read about do not appear to be significantly different --
though a few certainly are.

> And how were these properties selected?

Empirically.  See Ed Keenan's famous article in C. N. Li (ed.), Subject
and Topic.

[on my examples of reflexives and reciprocals]

> Since '*Each other was talking to Susie and Mike' is equally
> ridiculous, I fail to see any valuable point made.

I'm afraid you have failed to understand the point of my examples.

If transitive sentences in ergative languages were "really" passives,
then the absolutive NP should be the subject, and it should exhibit
typical subject properties, such as an inability to be reflexive or
reciprocal.  In Basque, however, absolutive NPs in transitive sentences
can freely be reflexive or reciprocal, whereas ergative NPs cannot.
It is only in intransitive sentences that absolutive NPs cannot be
reflexive or reciprocal.  So, ergative NPs in transitive sentences --
but not absolutive NPs in transitive sentences -- share the subject
properties of absolutive NPs in intransitive sentences.  This is exactly
the opposite of what is predicted by the "passive" theory, and it
constitutes a nail in the coffin of that theory.

> Frankly, I am amazed. A reflexive requires an agent and a patient,

Sorry; not so.  Consider these examples:

	Susie saw herself in the mirror.
	Susie is annoyed with herself.
	Susie excelled herself.

The first two certainly, and the third arguably, contain no agent, and
the third one certainly, and the first two arguably, contain no patient.
Yet all are overtly reflexive.

> and a reciprocal requires two agents and two patients.

No; same problem:

	Susie and Mike collided with each other.
	Susie and Mike fancy each other.
	Susie and Mike resemble each other.

Not an agent in sight, and not many patients, either.

Reflexives and reciprocals are *grammatical* constructions, not semantic
states of affairs.

> An intransitive verb, by definition, has only one NP element

Also wrong:

	Susie is sleeping with Mike.
	Susie smiled at Mike.
	Susie got ready for Mike.

All intransitive, but all with multiple NPs.

> so any two(or four)-NP-element construction obviously is a
> contradiction in terms.

Not remotely true, I'm afraid.

[in the same vein]

> More nonsense! 'Each other' cannot function as an ergative or
> nominative subject. So what? That is semantic not grammatical.

Completely false.  Suppose Susie slapped Mike, and Mike slapped Susie.
The semantics is straightforward:

	Slap(Susie, Mike) And Slap(Mike, Susie)

This semantics remains invariant however a language may choose to
express it grammatically.  Grammatically, a language is free in
principle to express this as `Susie and Mike slapped each other', or as
`Each other slapped Susie and Mike', or in any other way it chooses.
In Dyirbal, for example, a transitive verb is made reciprocal by adding
to it an affix which renders it strictly intransitive, and there is no
overt object. (And reflexives are made in the same way, with a different
affix.)  Turkish is rather similar.  The semantics is constant, but the
grammatical expression varies widely.

> That it can be used in some languages in an oblique case (as in
> English above) is to be expected.

Possibly so, but the crucial observation is that, in the vast majority
of known languages, a reflexive or a reciprocal cannot stand in subject
position.  The possible placement of reflexive and reciprocal NPs
therefore provides us with valuable evidence in identifying subjects --
and, as it happens, with powerful evidence against the "passive" view of
ergative languages, which makes all the wrong predictions.

> Of all the languages I have ever seen, Basque is, by a mile, far the
> most "unusual" language.

Nope.  Except that its morphological ergativity is unusually
thoroughgoing, Basque is not unusual in any respect I can think of.
Basque is syntactically unremarkable and morphologically highly regular.

Basque only becomes "unusual" if you insist on applying to it
demonstrably fallacious views like the "passive" theory of ergatives.
Only then does it start to appear bizarre.


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

> 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.

Estival and Myhill, in Dixon's account of them, emphatically do *not*
embrace the passive theory of ergatives: they only endorse the view that
ergatives invariably *originate* from passives.  To my knowledge,
Shibatani has never offered the slightest endorsement of the passive

As for the linguists of the past, well, they didn't know much syntax,
and they got it badly wrong -- as we now realize.

The linguists of the past frequently believed in all sorts of crazy
things which we have long since laid to rest: stadialism, linguistic
Darwinism, primitive languages, all sorts of things.  Even the great
Otto Jespersen, to cite just one name, believed in primitive languages
and in absolute progress in grammatical systems.  But nobody believes in
this stuff today.

Linguistics has moved on in the last few decades, and in our
understanding of syntax as much as in any other area.

Larry Trask
University of Sussex
Brighton BN1 9QH

larryt at

More information about the Indo-european mailing list