Question re: anti-passive voice

[2+3 clusivity]

As I understand it, anti-passive voice raises A to S and either (1) demotes O/P to an oblique case; (2) omits the erstwhile O/P; or (3) suppresses the existence of an O/P.

Regarding situation two and the basic assumption of A -> S, do all languages with an anti-passive demote O/P to an non-core oblique case, or can the O/P reemerge as an ergative? I.e. A-> O/P.absolutive and O/P -> A.ergative without a change in valence.

And, I suppose paralleling that, do any languages have a similar "non-valence adjusting passive voice" which rearrange A and P?

[vec]

I have never heard of such behavior... These are just my musings but I suppose its because a language's core arguments/grammatic relations are always to some degree tied to theta role. The subject always tends towards agent-like things or patient-like things, depending on weather the language is nom-acc or erg-abs, and the object always tends towards the other. A valency change operation such as voice is already encoding quite a "violent" change to the default syntax: the object gets promoted to the subject slot. It thereby kicks the previous subject out of its slot, which in turn doesn't simply move down the ladder to the object slot, it falls to the ground, so to speak. It then has to be "picked up" again as an oblique argument. To literally reverse the core arguments' case encoding seems like an entirely different beast.

[jal]

To literally reverse the core arguments' case encoding seems like an entirely different beast.

I recall a discussion about this years ago, when I discussed my conlang Kotanian, which has subject/object swapping in passive constructs. It was derided for being non-naturalistic (to which I simply replied it's spoken by aliens, so there goes your naturalistic argument :)).


JAL

[Cedh]

I wouldn't totally rule out the possibility that a natlang with something similar to what you describe can exist, but it would probably not be described as such:
- The case of the demoted O/P argument probably wouldn't be called "ergative", because it's part of the definition that an "ergative" case is used only for agent-like arguments.
- The most similar attested system that I can think of is usually called "direct-inverse alignment" (where case is typically unmarked and participant roles are inferred from things like person and/or animacy, with the "inverse voice" shifting around the default direction of action).
- "Philippine alignment" (aka "trigger") might also be considered to be similar, but of course this is only true if the agentive and patientive case are not clearly differentiated morphologically on the noun itself.

[Salmoneus]

Of course non-demoting passives are naturalistic - that's vanilla west-Austronesian. And indeed, these symmetrical languages tend to have the object-voice as primary iirc, so you could call them erg-abs languages with non-demoting antipassives - although I'm not sure it's helpful.

[Western austronesian languages are traditionally split functionally between Phillipine and Indonesian types - P-type systems tend to have multiple voices, whereas I-type languages have only the two symmetrical voices, but also use applicatives. Apparently this dichotomy is not perfect.]

[2+3 clusivity]

Hunh! Excellent food for thought. Thanks everyone.

[Western austronesian languages are traditionally split functionally between Phillipine and Indonesian types - P-type systems tend to have multiple voices, whereas I-type languages have only the two symmetrical voices, but also use applicatives. Apparently this dichotomy is not perfect.]

Is there a good source outlining the two types in depth? I am not familiar with the Indonesian type system.

Kotanian, which has subject/object swapping in passive constructs. It was derided for being non-naturalistic

Yes. Over the last two years, my reading of the board's pulse is that conlangs are judged by naturalistic criteria unless the author asks otherwise.

[Salmoneus]

Hunh! Excellent food for thought. Thanks everyone.

[Western austronesian languages are traditionally split functionally between Phillipine and Indonesian types - P-type systems tend to have multiple voices, whereas I-type languages have only the two symmetrical voices, but also use applicatives. Apparently this dichotomy is not perfect.]

Is there a good source outlining the two types in depth? I am not familiar with the Indonesian type system.

Sorry, can't give you a single cite. However, the terms are very established, so it's worth searching for them. Or just get a grammar of Indonesian/Malay, Javanese, or Balinese, the three main examples of Indonesian-type languages (though in the case of Indonesian/Malay, this apparently only applies to the standard language and educated registers - colloquial versions apparently follow a different voice system. Also, I/M confuses the matter by having two symmetrical voices but then also having a normal demoting passive (don't know whether that's also true in Javanese and Balinese).


Wait: just found http://pubman.mpdl.mpg.de/pubman/item/e ... 2009_s.pdf - an analysis of voice in Minangkabau. Of interest, however, may be the 22-page section on austronesian voice systems in general, distinguishing acehnese and sundic systems in addition to the main two. I haven't read it yet, but it might be a good starting point for you?

[2+3 clusivity]

Ah! Yes. That is good. Thanks Sal.

Edit: the Sundic-type is pretty fascinating I must note.

[roninbodhisattva]

... but it would probably not be described as such:

I was gonna say the same thing. There's at least one language, Flathead Salish, in which the object in the anti-passive construction is marked with the same particle/preposition as the subject in the fully transitive construction. That preposition has generally been called an `oblique' marker. So you get a pattern of ergative-like marking where the O/P in anti-passives is marked the same was as the `ergative' marker. But I'd still hold off calling it them both being marked by ergative case. For reference, both transitive and anti-passives are marked by overt morphology on the verb. So the pattern looks like this:

Transitive V- [Ø Pat] [t Agt]
Anti-passive V- [Ø Agt] [t Pat]

The oblique marker is t. Also, in the transitive construction, the both the Agent and the Patient control agreement on the verb whereas in the anti-passive only the Agent does (referenced with intransitive marking.) I have a handout on this if you want more data.

[Richard W]

Regarding situation two and the basic assumption of A -> S, do all languages with an anti-passive demote O/P to an non-core oblique case, or can the O/P reemerge as an ergative? I.e. A-> O/P.absolutive and O/P -> A.ergative without a change in valence.

There's a third option - A and P can remain, but in the absolute case. This construction can be seen on p9 of a discussion of Basque resultatives at http://dare.uva.nl/document/339745 and in Wikipedia at http://en.wikipedia.org/wiki/Antipassiv ... rom_Basque .

And, I suppose paralleling that, do any languages have a similar "non-valence adjusting passive voice" which rearrange A and P?

Depending on how you want to interpret it, there's the Thai adversative passive (example and discussion with implicit past time):

Active: mǎː kàt dèk 'dog bite child' = 'The dog bit the child'
Passive: dèk tʰùːk mǎː kàt 'child hit dog bite' = 'The child was bitten by the dog'

There's an alternative interpretation that /tʰùːk/ introduces a subordinate clause explaining the misfortune. The phrase 'mǎː kàt' on its own should mean 'the dog was bitten'; one can view Thai transitive verbs as being neutral as to voice, with the the number of arguments present often resolving the ambiguity. I've a feeling the rearrranged agent isn't accessible in a relative clause.