Erg-abs universals

[Dauyn]

...or at least commons. Does anyone have any links or lists of common features of ergative-absolutive languages?

[Radius Solis]

One is that hardly any languages are purely erg-abs. Usually there'll be some nom-acc behaviors as well. For instance take Basque, which has an ergative-absolutive case-marking system, but uses nominative-accusative word ordering. Exception: some of the Australian languages are reputed to be purely ergative.

Another universal is that ergative languages, by their nature, lack passive voice. Instead many of them use a special voice called the "antipassive" which works similarly but in the opposite direction (I can explain how it works if you like).

That's all I can think of but there's likely more.

[Dauyn]

Thanks Rad!

I do understand the antipassive - it demotes the object, rather than the subject (like the passive).

Mary eats pie.
> Passive The pie is eaten (by Mary)
> Antipassive Mary eats (from the pie)

I also know that the subject of the Antipassive moves from ERG marking to ABS.

What I'm not sure of is whether I understand the semantics of it...

[chris_notts]

For instance take Basque, which has an ergative-absolutive case-marking system, but uses nominative-accusative word ordering.

What? Basque doesn't have nom-acc word ordering... Basque work order is more determined by information salience and pragmatics than grammatical relations. What Basque does have is the subject as the syntactic pivot... but that's not the same as saying that the subject is differentiated by word order.

Another universal is that ergative languages, by their nature, lack passive voice. Instead many of them use a special voice called the "antipassive" which works similarly but in the opposite direction (I can explain how it works if you like).

This universal simply isn't true. Basque has a passive but no antipassive and so do many other ergative languages (possibly most, depending on what you count as a passive). This is because Ergative languages often allow the easy deletion of the erg, as in Basque for example:

ikusi dut
I saw him/her

ikusi da
he/she was seen

Here, the ergative can be deleted simply by giving the verb an intransitive auxilliary, rather than the expected transitive. Some ergative languages allow the deletion of the ergative but don't explicitly mark that the argument has been deleted... if you consider these passives then I'd go so far as to say that probably most Ergative languages have passives. If you require explicit marking to count it as passive then still quite a few do.
As for Australian syntactically ergative langs... yes, they do exist. Even they though have at least one area that exhibits the notion of subject: imperative formation. It makes little sense to order a patient, since they don't have control over the action, so imperatives usually order S+A, not S+P. A language that had:

run-imp
Run!

hit-imp
Be hit!

probably doesn't exist.

[Radius Solis]

Another universal is that ergative languages, by their nature, lack passive voice. Instead many of them use a special voice called the "antipassive" which works similarly but in the opposite direction (I can explain how it works if you like).

This universal simply isn't true. Basque has a passive but no antipassive and so do many other ergative languages (possibly most, depending on what you count as a passive). This is because Ergative languages often allow the easy deletion of the erg, as in Basque for example:

ikusi dut
I saw him/her

ikusi da
he/she was seen

Here, the ergative can be deleted simply by giving the verb an intransitive auxilliary, rather than the expected transitive. Some ergative languages allow the deletion of the ergative but don't explicitly mark that the argument has been deleted... if you consider these passives then I'd go so far as to say that probably most Ergative languages have passives. If you require explicit marking to count it as passive then still quite a few do.

No, I don't count erg-deletion as passive, not even if there's special morphology involved in doing so. And I never said ergative langauge universally have antipassive voices, note that I used the word "many". I said ergative languages do not have passive voices. A passive voice is one in which a transitive sentence is detransitived with the patient in subject position in opposition to the default method of detransivizing. The definition of "ergative" casemarking requires the default method of detransitivizing to be, essentially, erg-dropping. Morphological indicators of detransitivization may be involved, but that doesn't make it passivization. It can only be passivization if the default/unmarked method of detransitivizing would be to drop the patient instead.

[Miekko]

Another universal is that ergative languages, by their nature, lack passive voice. Instead many of them use a special voice called the "antipassive" which works similarly but in the opposite direction (I can explain how it works if you like).

This universal simply isn't true. Basque has a passive but no antipassive and so do many other ergative languages (possibly most, depending on what you count as a passive). This is because Ergative languages often allow the easy deletion of the erg, as in Basque for example:

ikusi dut
I saw him/her

ikusi da
he/she was seen

Here, the ergative can be deleted simply by giving the verb an intransitive auxilliary, rather than the expected transitive. Some ergative languages allow the deletion of the ergative but don't explicitly mark that the argument has been deleted... if you consider these passives then I'd go so far as to say that probably most Ergative languages have passives. If you require explicit marking to count it as passive then still quite a few do.

No, I don't count erg-deletion as passive, not even if there's special morphology involved in doing so.

What if the erg-deletion+morpho makes the abs argument the subject (as testable by pivots and stuff)?

[Radius Solis]

No, I don't count erg-deletion as passive, not even if there's special morphology involved in doing so.

What if the erg-deletion+morpho makes the abs argument the subject (as testable by pivots and stuff)?[/quote]

What does that have to do with anything? Isn't it normal for the abs argument to be the (syntactic) subject? If it isn't, why on earth would we call it absolutive in the first place?

[Miekko]

No, I don't count erg-deletion as passive, not even if there's special morphology involved in doing so.

What if the erg-deletion+morpho makes the abs argument the subject (as testable by pivots and stuff)?

What does that have to do with anything? Isn't it normal for the abs argument to be the (syntactic) subject? If it isn't, why on earth would we call it absolutive in the first place?[/quote]
In that event, it's not just the question of the erg having been omitted (and left a ? trace or something), but of the abs actually having moved up from within the VP to the position previously occupied by the erg argument. Which requires more than omission of the erg, and thus is closer to an actual passive than to something only involving deletion of an argument.

[chris_notts]

In that event, it's not just the question of the erg having been omitted (and left a ? trace or something), but of the abs actually having moved up from within the VP to the position previously occupied by the erg argument. Which requires more than omission of the erg, and thus is closer to an actual passive than to something only involving deletion of an argument.

Which, as you pointed out, does occur if you delete the ergative in a syntactically accusative but morphologically ergative language (like Basque). If you remove the ergative by giving the verb an intransitive auxilliary then the abs becomes the syntactic pivot instead of the ergative, so still many ergative languages have passives, even if you require some kind of surface marking.
I have to say that most of the books on voice that I've read agree with me on this issue, since I've seen many examples cited of ergative languages with passives... if it's not possible, then why exactly are so many people giving examples?
I don't accept that passives must be "in opposition to the default method of detransitivizing"... I'd accept a definition that there must be overt marking for passives to differentiate them from their transitive counterparts (in a syntactically ergative language that allowed unmarked erg dropping you could argue that the dropping was simply because the erg was not a core argument of the verb, ie all verbs in the language were intransitive with only the abs as a core argument) but I see no reason at all to limit passives to non-default methods, nor do I see how you can even ascertain in many cases what is basic and what isn't. If the passive is marked by a change in morphology (say, a change to intransitive verb agreement/inflectional markers) and the antipassive is too, then how can you argue that one is more basic than the other?

EDIT: example of an ergative language which has a passive that's not equivalent to normal intransitive marking on the verb: Inuktitut.

[Ketumak]

David J. Peterson has a good discussion of ergativity at:

http://dedalvs.free.fr/notes.html#ergativity

He sees the anti-passive as typical of ergative languages. All I know is that it was knotty problems like this the made me abandon ergativity in Lhemburan. I'd advise anyone to steer clear of ergativity unless their very sure of their material.

[Radagast]

EDIT: example of an ergative language which has a passive that's not equivalent to normal intransitive marking on the verb: Inuktitut.

Elaborate?

Also:
Can we agree at least that antipassives are only found in ergative languages and never in nom/acc languages?

[Miekko]

EDIT: example of an ergative language which has a passive that's not equivalent to normal intransitive marking on the verb: Inuktitut.

Elaborate?

Also:
Can we agree at least that antipassives are only found in ergative languages and never in nom/acc languages?

Hm.
The similar transition in nom/acc langs would be nom(trans)>nom(intrans), acc>obl. While I am totally sure that this happens all the time and in very many languages, from a syntactic p.o.v. it doesn't make much sense to give that operation any specific name. If there were a syntactically ergative, morphologically nominative lang, it could make sense, though ... (then you'd have nom(erg)>nom(abs), acc(abs)>obl, where (X) is the 'syntactic case')

[chris_notts]

Elaborate?

From the Tarramiutut Subdialect of Inuktitut:

Typical intransitive verb:

(1) Jaani itir -tuq
Jaani(ABS) enter-IND(3s)
Jaani entered.

Typical transitive verb:

(2) Jaani-up nanuq qukir ?ta -nga
Jaani-ERG bear(ABS) shoot-IND-3sA:3sU
Jaani shot the bear.

Anti-passive of (2):

(3) Jaani nanur-mik quki-i-juq
Jaani(ABS) bear-SEC shoot-AP-3s
Jaani shot a bear.

Passive of (2):

(4) nanuq qukir-ta -u ?laur -tuq Jaani-mut
bear(ABS) shoot-PASSPRT-be-PAST-IND(3s) John-DAT
?The bear was shot by John?

Note that the verb in (4), qukir-ta-u-laur-tuq, contains the morphemes -ta-u- which are not TAM but passive and missing from (1), the example of a normal intransitive verb. These examples are from:

http://cslipublications.stanford.edu/LF ... 3beach.doc

Also:
Can we agree at least that antipassives are only found in ergative languages and never in nom/acc languages?

As Miekko says, an antipassive would make sense for a morphologically nom/acc but syntactically ergative language, but since no such language has been found so far, I'd agree it's pretty unlikely (although I wouldn't go so far as to say it's completely impossible). But ergative languages are generally less consistent in their ergativity than accusative languages are in their accusativity, so the fact that ergative languages do have passives sometimes whereas accusative languages don't really have antipassives shouldn't be a surprise.

[Radagast]

Anti-passive of (2):

(3) Jaani nanur-mik quki-i-juq
Jaani(ABS) bear-SEC shoot-AP-3s
Jaani shot a bear.

Passive of (2):

(4) nanuq qukir-ta -u ?laur -tuq Jaani-mut
bear(ABS) shoot-PASSPRT-be-PAST-IND(3s) John-DAT
?The bear was shot by John?

To me at least the antipassive example is a bit stretched. The way I have been taught to interpret that example is as a simple substitution of a transitive verb with an intransitive one with the introduction of the former object as an instrument (-mik is normally called the instrumental case). Is that enough to call it an antipassive? Whether it is a detransitivisation of a transitive verb or verb substitution is really unimportant because all verb roots can take both transitive and intransitive affixes.

To me the example is equivalent to

John was painting the paper (paper is object of the transitive verb)
John was painting on the paper (paper is location/instrument/receiver of the intransitive verb)

Would you call the second one antipassive?

[chris_notts]

To me at least the antipassive example is a bit stretched. The way I have been taught to interpret that example is as a simple substitution of a transitive verb with an intransitive one with the introduction of the former object as an instrument (-mik is normally called the instrumental case). Is that enough to call it an antipassive? Whether it is a detransitivisation of a transitive verb or verb substitution is really unimportant because all verb roots can take both transitive and intransitive affixes.

I guess the point is that the verb takes an anti-passive affix -i-, or at least that's how it's glossed. In the English example you gave there is no marker of valence reduction, whereas if (as described) -i- truly marks the removal of the patient as a core argument, then the overt marking of valence reduction is probably what makes the argument that it's a voice operation. I don't think the fact that the removed core argument is demoted to an oblique case (the instrumental) is an argument against, since the same occurs for passives and other voice operations in English (demotion of agent to "by.." oblique) and other languages.
There's also the obvious difference that in the anti-passive construction the remaining core argument is mapped to a different grammatical relation (ie ergative -> absolutive), whereas with the English example no such mapping takes place... what was the subject remains the subject, all you've done is turned the less important core argument in english (the object) into an oblique, which doesn't require any explicit morphology or anything in English. But I don't believe this is a necessary requirement for voice though: As I've already said, I see little reason for not classing valence decreasing operations with explicit morphological changes (like passives in some ergative languages) as voice even though they don't map the remaining core argument to a new morphological or syntactic case or role.
Still, this doesn't affect the passive example. And since it's similar in structure to the English passive, if we argued that Inuktitut (a morphologically ergative language) had no passive, then I don't see how we could avoid concluding that English also lacks a passive, given the similarity in structure (both are be + past participle by the look of it).

[Radagast]

It's a matter of how to interpret the affixes which is a tricky business in eskimoan.

But seeing as all verbal roots can take transitive or intransitive endings I don't see how the affixes in between which are only semantic (at least the way I have learnt to interpret them) can remove or add arguments to the core. -juq and -tuq are intransitive endings and they are not dependent on any of the other affixes and could be switched with transitive endings.

Also I cannot interpret the passive example as be + past participle. Unless of course Tarramiutut is fundamentally divergent from Kallaallisut.

[Miekko]

Anti-passive of (2):

(3) Jaani nanur-mik quki-i-juq
Jaani(ABS) bear-SEC shoot-AP-3s
Jaani shot a bear.

Passive of (2):

(4) nanuq qukir-ta -u ?laur -tuq Jaani-mut
bear(ABS) shoot-PASSPRT-be-PAST-IND(3s) John-DAT
?The bear was shot by John?

To me at least the antipassive example is a bit stretched. The way I have been taught to interpret that example is as a simple substitution of a transitive verb with an intransitive one with the introduction of the former object as an instrument (-mik is normally called the instrumental case). Is that enough to call it an antipassive? Whether it is a detransitivisation of a transitive verb or verb substitution is really unimportant because all verb roots can take both transitive and intransitive affixes.

To me the example is equivalent to

John was painting the paper (paper is object of the transitive verb)
John was painting on the paper (paper is location/instrument/receiver of the intransitive verb)

Would you call the second one antipassive?

According to this long book on Ergativity (with various papers by various linguists) that I read some time ago, whether you'd select the antipassive or the active in Inuit has more to do with whether the paper, in this sample sentence, is new or given, than with whether it being the object or the location/instrument/receiver of hte intransitive verb, and therefore, interpreting this construction in general in Inuit to be an antipassive jives well with the general tendencies of antipassives.

[Radagast]

It is definitely true that when it comes to discourse strategy it is basically the same. It is away of demoting the patient and focusing on the agent. It can basically only be done if the patient introduced is new and peripheral to the discourse and the agent is also topic. But I would still just see it as choosing an intransitive ending instead of the transitive one.

[chris_notts]

As a further argument that voice processes need not reassign core grammatical roles, here are some examples from a language called Ixil in Grammatical Voice by Klaiman:

A-k'oni in ta'n uula
2nd.erg-shoot 1st.abs with sling

Uula A-k'oni-b'e in
sling 2nd.erg-shoot-index 1st.abs

Basically, there are affixes on the verb that indicated the role of a fronted focused participant in the action, which may be an oblique like an instrument. However, these affixes don't demote either of the core arguments erg or abs to an oblique role (ie they are not applicatives as the term is usually used)... as shown by the ergative prefix a- and the 1st person abs enclitic in, both the actor and patient remain in the same roles in both sentences, even when the role of the sling is indicated by verbal morphology rather than by with a preposition.
Klaiman calls such systems information salience voice systems, ie systems whose primary purpose is related to marking information salience rather than promoting, demoting, deleting existing core arguments or altering the syntactic pivot. Ixil even has a separate voice operation (the anti-passive) for deleting or demoting the abs that isn't marked in the same way as this focus fronting is marked.
My point here is that voice as conceived by many linguists is a much more varied phenomenon than the definitions people usually give for it. If these operations don't fall under your definition of voice then you're free to think of them as something else, but like many things in linguistics voice can mean different things to different people and no-one can be considered to be correct and all the others wrong. Basically, voice covers most operations that through verbal morphology (and in some cases periphrasic constructions, as is the case for the English passive) alter the argument structure or valence of the verb. The Ixil voice markers above actually increase the valence, since, I believe, you can't simply delete an argument you've marked the role of in the verb (ie, the number of required arguments is raised by one). This, incidentally, makes me wonder why the remaining valence modification operations aren't considered voice.. if systems such as the Ixil system and applicatives are voice, why isn't the causative a voice? And if it is, why do we use the term voice instead of talking directly about valence and argument structure modification, or simply define the term voice to refer to those processes?

[Radagast]

I agree Chris, you are absolutely correct. If we choose to widen the terminology of voice to all operations that have the function of changing valence of verbs or otherwise change the focus from one argument to another semantically then these are of course all voice operations. I wouldn't personally use that terminology in neither the case from Inuit nor from Ixil maya. But both certainly do fall under the wider description. But under the wider description then I do think that the difference illustrated in the example "john paints the paper" vs. "john paint on the paper" is also describable as a voice operation.

[chris_notts]

I agree Chris, you are absolutely correct. If we choose to widen the terminology of voice to all operations that have the function of changing valence of verbs or otherwise change the focus from one argument to another semantically then these are of course all voice operations. I wouldn't personally use that terminology in neither the case from Inuit nor from Ixil maya.

To be honest, the more I learn the less useful I see voice as a catch all term. I'd much rather see the operations labelled voice in terms of three or more separate functions that they generally perform:

1) removing or making peripheral a core argument
2) marking pragmatic factors like topic or focus
3) altering the syntactic pivot
4) otherwise altering the argument structure of the verb (including adding core arguments)

Most things described as voice perform at least one or two of these functions, but voice operations from language to language don't seem to be a well defined or comparable group, since a superficially similar operation (say passive) can be used in very different ways from language to language. It seems to me very dubious that even a very restricted part of voice like passives can truly be treated as in any sense equal across languages... it seems to me that nothing should be just described as say a passive or an anti-passive or any other kind of voice without a thorough description of what, exactly, it's used for in discourse.
But if we must talk about voice, I maintain that for all intents and purposes many ergative languages have passives.

But both certainly do fall under the wider description. But under the wider description then I do think that the difference illustrated in the example "john paints the paper" vs. "john paint on the paper" is also describable as a voice operation.

Functionally it is equivalent. But I was working under the assumed definition that a voice operation like a passive requires a non-zero marking of some form. As I've said though, the more I think about it the less interesting or useful this whole debate seems, since it's all a matter of minor differences in definition.

[Radagast]

Yep.

[Accard]

Does split ergative languages use both passive and antipassive? For example, say that we have a language ergative in the past tense, and that it then uses the antipassive voice as a valency-decreasing operation. Would the nominative-accusative part of this language (the present tense) then use the antipassive marker as a passive marker, or would it inroduce a new marker althogether?

What kind of valency-decreasing operation does a tripartite language use? It must essentially be the same as an antipassive.

[Zhen Lin]

What kind of valency-decreasing operation does a tripartite language use? It must essentially be the same as an antipassive.

It could just as well be either, couldn't it? You could either delete the agent or the patient - whatever is left takes the intransitive case... so it might have passives and antipassives.

[Accard]

Even if it is not about ergativity generally - it is about valency-changing operations though - does anyone noe how the Tamil voices 'subject voice' and 'object voice' works?