accusative and ergative languages

Larry Trask larryt at
Mon Aug 9 11:42:01 UTC 1999

On Tue, 3 Aug 1999, Patrick C. Ryan wrote:

Well, I have visions of the moderator reaching for the plug, but here

> It is you who are missing the point. The absolutive NP in an ergative
> language is the patient.

Sorry; not so.  In Basque, and in most ergative languages, it is
perfectly possible for an agent to stand in the absolutive case, and it
commonly does so when the sentence contains an agent but no patient --
that is, with intransitive verbs which take agent subjects, like
`speak', `eat dinner', `dance' and `go'.

An ergative language, by definition, does *not* distinguish agents from
patients.  Rather, it distinguishes transitive subjects on the one hand
from intransitive subjects and direct objects on the other.  It makes no
difference whether a given NP is an agent, a patient, or neither.

A language that systematically marks the difference between agents and
non-agents is an *active* language, not an ergative language, and active
languages exist.  Among the ones I have seen reported are Crow, Eastern
Pomo, 19th-century Batsbi (I'm told that contemporary Batsbi is
different), and, according to some accounts, Sumerian.

> Lacking an agent (which would be the ergative), there is no
> possibility of a reflexive or reciprocal. It takes an agent and a
> patient to tango the Reflexive or Reciprocal.

Also quite false.  In Basque, just as in English, both reflexives and
reciprocals can occur freely in sentences which lack an agent or a
patient or even both.  In Basque, reflexives and reciprocals can occur
freely in sentences lacking an ergative NP.  Basque has a set of
dedicated reflexive NPs, of which <bere burua> `himself/herself' is the
third-singular one, and it has a dedicated reciprocal NP <elkar> `each
other, one another', which has no other existence in the language.
But look at these examples:

	Jon bere buruari mintzatzen zaio.
	`John is talking to himself.'

	Jon eta Miren elkarrekin joan ziren.
	`John and Mary went together.'
	(literally, `with each other')

The verbs <mintzatu> `talk, speak' and <joan> `go' are strictly
intransitive, and neither can occur with an ergative NP.  Yet both can
happily take reflexive and reciprocal NPs.

> Sumerian regularly expresses reflexives: how? by putting the agent
> in the ergative and the patient (which is the same) in the
> absolutive.

Very interesting, and a strategy which I have not often seen.
But this is only one strategy among many, and no universals follow from

[on my examples of reflexives lacking agents or patients or both]

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

> Larry, this is patent nonsense.

> In the sentence: "Susie saw the room behind her in the mirror",
> "Susie" is certainly an agent and "room" is certainly a patient.

Certainly not.  Try reading the definitions of `agent' and `patient' in
my dictionary.  In general, the subject of `see' is not an agent, and
its object is not a patient: the referent of the subject NP isn't
instigating anything, and the referent of the object NP isn't undergoing

> Your sentence is exactly analogous. "herself" is the patient (a
> paraphrase for "Susie"), and "Susie" is the agent.

'Fraid not.

> In your second sentence, it is equivalent to "Susie[1] annoyed
> Susie[2]". "annoy" is construed in English as a two-element verb
> (usually), of which the first element is the agent ("Susie[1]") and
> the second element is the patient ("Susie[2]').

No.  First, the rather unnatural `Susie annoyed herself' does not at all
mean the same thing as `Susie is annoyed with herself'.  Second, it is
most unlikely that `Susie' is an agent even in `Susie annoyed herself'.
People do not commonly set out deliberately to annoy themselves, as
required here by the agentive reading.

> It is incredible to find anyone, let alone a linguist, denying that
> any true reflexive construction does not mandatorily contain an
> agent, which is coterminous with its patient.

Well, it may astound you, but it's the plain truth, or rather it would
be the plain truth if you replaced `denying' by `asserting', which I
presume is what you meant.

[on my further examples]

>> 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.

> Agent: Susie; patient: Mike; second agent: Mike; second agent: Susie.

Nope.  `Susie' and `Mike' would be agents only if they consciously and
deliberately instigated the action, which is hardly likely to be the
case in any of my examples.  People do not deliberately set out to
collide with each other, to fancy each other, or to resemble each other:
all of these are things that happen to us, not things that we do.

Are you sure you understand the meaning of the term `agent'?

> I suppose we could dance around the definitions of "grammatical" and
> "semantic" but the fact is that, e.g. "reflexive" means simply: "a
> verb having an identical subject and direct object".

No, not remotely true.

To begin with, a reflexive is not a verb at all.  An NP can be a
reflexive, and a clause can be reflexive.  But a verb can't be
reflexive, at least not in English.  (The Romance languages, of course,
have verb-forms which are commonly called `reflexive verbs', but the
term is used here in a rather special sense: the Romance reflexive verbs
are not, in general, strictly reflexive.)

Moreover, the two coreferential NPs need not be the subject and the
direct object.  See the examples above.

> The fact that you want to extend "reflexive" into areas in which it
> does not belong based on pseudo-reflexive constructions in some
> languages does not alter one iota what a true reflexive is.

I've no idea what you mean by a `true reflexive', but it doesn't appear
to bear much resemblance to what everybody else calls a reflexive.
And I certainly can't guess what you mean by `pseudo-reflexive'.

[on Pat's claim]

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

[and my response]

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

>> All intransitive, but all with multiple NPs.

> Well, it is my fault for leaving out the qualification "essential" or
> "core".

> Actually, I thought you might grasp that without the qualifiation.

> Do you think there is a difference between:

> "Joe is hitting" and

> "Joe is sleeping"?

> Or is it just a peculiarity of my personality that would make me ask:
> "Hitting whom?" but NOT ask "Sleeping with whom?" ?

Well, I am astonished.  In my experience, `sleeping with' is not
something you can do by yourself, and `sleeping' is not at all the same
activity as `sleeping with'.  Of course, things may be different in
Arkansas, but, given what I've been reading about Bill Clinton, I doubt
it. ;-)

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


>> Nope.

> Larry, I do not need nor want you to assume the prerogative of
> correcting my expressed impressionistic opinions.

Well, perhaps you'd like to lay out your reasons for seeing Basque as
highly unusual.  I've been studying the language for 28 years, and I
haven't noticed many peculiarities.

> It is bizarre, in my opinion, because of its weird phonology,

"Weird"?  Why "weird"?  Basque has an exceptionally simple segmental
phonology, comparable in many respects to that of Castilian Spanish.
It has a small set of phonemes, very simple phonotactics, and few
alternations.  Phonologically, it is *much* simpler than English, and
simpler than almost any other European language I can think of.  It
lacks the mutations of Celtic, the consonant gradations of Finnish, and
the elaborate stem-alternations of most IE languages.

> and the habit it has of borrowing (according to you and Michelena)
> vocabulary from everywhere for just about everything so that one is
> hard put to find originally Basque words for anything.

A grave overstatement.  Basque has indeed borrowed extensively, but in
this it is hardly alone, especially among minority languages.  And it is
trivial to list hundreds of Basque words and morphemes which are clearly
native and probably ancient.  Where are you getting this stuff from?

> Shibatni's endorsement is implied by his allowing Estival and Myhill
> to publish under his aegis.

No.  An editor does not, in general, endorse all the views expressed by
those contributing to the book he edits.  I have edited some books
myself, and am about to edit another.  Like any editor, I do not prevent
contributors from publishing views I disagree with.

> Are you not the one who is forever complaining about wrong-minded
> ideas being allowed to see the light of a printing press?

To be precise, I have often complained about transparently shoddy work
being accepted for publication, even in refereed books and journals.
It is an editor's responsibility to prevent crappy work from appearing
in his books and journals, but not to suppress conclusions he doesn't

> Would Estival and Myhill be published in a book under your
> editorship? Fat chance!

You do me an injustice.  When editing or refereeing, I have frequently
accepted or approved work with whose conclusions I do not agree.
I content myself with correcting obvious errors, pointing out flaws in
argumentation, and cleaning up the English.

You seem to have a jaundiced view of the academic world.

Larry Trask
University of Sussex
Brighton BN1 9QH

larryt at

More information about the Indo-european mailing list