Split-Ergativity in Proto-Yonut: How realistic is this alignment?

[dɮ the phoneme]

So, my newest project is the Yonut language family, which must of course begin with Proto-Yonut. Proto-Yonut features a split-ergative alignment, whereby clauses containing 1st and 2nd person pronouns are marked accusatively, à la many Australian languages. However, there are also some complications involving a constraint against marking inanimate nouns in the ergative. Honestly, I'm not really sure how realistic this system is, so I'd appreciate any feedback anyone has on it.

----------------------------------------------------------------------

The ergative case is used principally for animate, 3rd person agents in transitive clauses, while the absolutive case is used for associated patients:


biɟipaheq senampuso sosqilles

b-iɟi-pahe-q s-e-nampu-s-o s-os-qille-s

N-PL-person-ERG 3pl.N.A-3sg.INAN.P-till-ACT-PST INAN-SG-soil-ABS

“(Some) people were tilling the soil.”


However, when the agent is a 1st or 2nd person pronoun, the agent takes its least marked form, the absolutive. In such clauses, the patient takes the accusative:


tus hiɟinnampuso sosqillehi

tu-s hiɟin-nampu-s-o s-os-qille-hi

1pl.N-ABS 1pl.P-till-ACT-PST INAN-SG-soil-ACC

“We were tilling the soil.”


The only agreement marking on the verb is the “patient” agreement (with the pronoun, which is semantically an agent), while there is no agreement with the semantic patient, which is marked in the accusative. This is because these constructions originate historically from intransitive constructions, where the pronominal subject was marked like a patient (since Proto-Yonut is ergative), and the semantic patient was marked in a locative case of some sort, which later specialized into this accusative. IIRC, a similar path is attested for the development of split-ergativity in Basque.

As one would expect for an ergative language, intransitive subjects take the absolutive case, whether pronominal or otherwise:


tus hiɟinpiɻoso

tu-s hiɟin-piɻo-s-o

1pl.N-ABS 1pl.P-go-ACT-PST

“We went.”


Proto-Yonut has a moderately strong constraint against marking inanimate nouns as ergative. In order to uphold this constraint, the antipassive voice is used. The antipassive demotes an ergative agent to an absolutive subject, and is used to place an inanimate noun in the semantic role of agent:


sosbodos heχuɻnuko hijastaʈ

s-os-bodo-s he-χuɻnu-k-o hija-staʈ

INAN-SG-rock-ABS 3sg.INAN.P-hit-ANTIP-PST 1sg.F-LAT

“A rock hit me/I was hit by a rock.”


Here, the antipassive affix -k has turned the verb -χuɻnu-, which is typically transitive, into an intransitive, with a semantic agent (the rock) marked in the absolutive, and a semantic patient marked in a non-core case, the lative. Since this is an antipassive construction, the inclusion of the patient here is optional.

[Bessunire]

It looks plausible to me. It makes sense if Proto-Yonut's speakers see the ergative as marking action with conscious thought behind it, and the active voice as describing such actions. Inanimate objects are incapable of conscious thought, therefore it wouldn't make sense to use the ergative or the active voice with them. If the ergative and active voice are linked to conscious thought, however, it may also lead to the antipassive being used for when animate agents perform actions involuntarily, such as tripping or falling asleep.

You might also want to look at the grammar of Australian Aboriginal languages, since I believe some of them may make choices of whether to use the active voice or the antipassive based on animacy or similar criteria.

[dɮ the phoneme]

One of the worries I've been having about this system, though, is that if ergative marking is so closely associated with animacy (or conscious thought), is it plausible that discourse participants (1st and 2nd person pronoun), typically the most animate entities, are always absolutive?

[bbbosborne]

i'm a bit confused so this might be completely useless, but you could make the split a bit more precise, e.g. 1st person with 2nd person animates.

[dɮ the phoneme]

i'm a bit confused

What I mean is that many languages seem to consider 1st and 2nd person pronouns to be "more animate" than even animate 3rd person referents.

1st person with 2nd person animates.

What do you mean exactly?

[bbbosborne]

1st person with 2nd person animates.

What do you mean exactly?

spreading out the agentivity hierarchy:
1st person animate
1st person inanimate
2nd person animate <--- the split is here
2nd person inanimate
etc...

so the entire 1st person and 2nd person animates get nom/acc alignment

[akam chinjir]

If I'm understanding right, pre-Proto-Yonut was (morphologically speaking) purely ergative, but then there were two changes. First, transitive clauses with 1p and 2p pronoun subjects were replaced with intransitive clauses, with the semantic object demoted to an oblique case. Second, transitive clauses with inanimate subjects were replaced with intransitive clauses (with an antipassive-marked verb), with the semantic object demoted to an oblique case (a different one). And the question is whether it makes sense to have more or less the same thing happen twice, but with subjects at the opposite ends of the language's animacy hierarchy.

I've got two reasons for thinking there's no problem here, but also some things I'd worry about if I wanted to worry. I hope none of it is too far off base.

One thing is that if the new pattern with 1p and 2p pronoun subjects is fully established by the time the second change gets started, there's no synchronic problem since speakers will no longer need to attach any particular significance to the distinction between those pronouns and other subjects. Put another way, a rule against ergative-marked inanimate subjects need not have anything to say about non-ergative-marked non-inanimate subjects.

Also, a common function of antipassives/detransitives is to raise the discourse salience of a subject (by demoting the object), and his doesn't have to intersect with animacy in any particular way. 1p and 2p pronouns could get this treatment because of the obvious salience of discourse participants; inanimate agents could get it (especially if acting on an animate patient) because that's not the stereotypical role for inanimate arguments.

That said, if I wanted to worry, I guess I could worry about the similarities between the two constructions. Both involve a subject in absolutive case, a detransitivised verb, and a semantic object marked in something other than one of the usual core cases, with the verb not agreeing with that object (though it would agree with an absolutive object).

One difference is that in the newer construction, the verb is marked as an antipassive, but in the older (1p/2p pronoun) construction there's no analogous marking. First question: was it marked somehow in earlier forms of the language? Or did pre-Proto-Yonut just allow zero derivation of unergatives? If it was marked, did the marking just get sound-changed away?

The other difference between accusative and lative objects, I take it, is that the accusative is supposed to be a core case, whereas the lative is not, though neither triggers agreement on the verb. Are there differences elsewhere in the syntax between the two? E.g., are there adverbs that can come between the verb and a lative object but must follow an accusative object? Is it possible to gap the object in relative clauses or question it in content questions with an accusative object but not a lative one? That sort of thing. (Does the accusative show up anywhere else?)

...reading over that it looks like I didn't end up with any real worries, just a couple things I enjoyed thinking about. Oh, well.

[Travis B.]

1st person with 2nd person animates.

What do you mean exactly?

spreading out the agentivity hierarchy:
1st person animate
1st person inanimate
2nd person animate <--- the split is here
2nd person inanimate
etc...

so the entire 1st person and 2nd person animates get nom/acc alignment

Umm typically 1st and 2nd persons are treated as inherently animate, or as more animate than other animate things. Speaking of "2nd person inanimate" is almost inherently a contradiction in terms.

[mèþru]

2nd person inanimates mightbe useful for when addressing inanimate objects as if they are animate. Then again, addressing them as animates should mean they are treated as animates for syntactic purposes.

[Frislander]

Also "animacy heirarchy" is a bit of a misleading term, because typically it's about who the speaker cares most about/who is most important to the people in the conversations, which is why 1st and 2nd persons are at the top (because they're of course the people those involved in a conversation most care about), then animate 3rd persons then inanimates. Also with regards to inanimate SAPs, 1st persons are undoubtedly going to be animate all of the time in natural human language, while with 2nd persons if someone for whatever reason finds themselves talking to an inanimate object, the more typical thing I would expect would be for them to treat the object like any other normal (animate) 2nd person. The idea of having a separate "inanimate 2nd person" seems rather unnatural to me.