Sentential Objects and Ergativity

[dhok]

I know that sentential objects- in other words, independent clauses used as arguments for a verb- are treated as accusatives in Nom-Acc languages: I think+ He's crazy I think [he's crazy], with "he's crazy" being interpreted as the object of "think". How do ergative languages- or languages with other alignments in general- deal with these?

[Astraios]

It's not that complement clauses are interpreted as being the objects of verbs, that's what they are.

"I think X" = S + V + DirObj, whatever the DirObj is. Obviously if a language has a different way of constructing that meaning, then it'd be formed different, but as it is, complement clauses are direct objects.

[Whimemsz]

In Algonquian languages, which mostly have a direct-inverse alignment system, complement clauses are basically treated as "objects" (though the category "object" doesn't really play an important role in Algonquian grammar generally). Afaict, the matrix verb is normally a VTI verb (transitive verb with inanimate object -- where in this case the inanimate (singular) object is the complement clause itself), e.g. Ojibwe ingikendaan (i'iw) wii-kimiwang = "I know it's going to rain" (ingikendaan = 1-know.TI-INAN:SG = "I know it/that"). Except when the subject of the complement clause is animate, the general pattern seems to be that the matrix verb is inflected as a VTA (transitive verb with animate object), with the subject of the complement verb serving as the object of the matrix verb, e.g. Ojibwe gigikenimin wii-tagoshinan = "I know you'll arrive" (gigikenimin = 2-know.TA-LOCAL:INV = lit. "I know you", with the whole sentence literally meaning something like "I know [with respect to] you that you will arrive"). I'm not sure exactly how prevalent this pattern is across all Algonquian languages, however; it's sometimes referred to as the "copying-to-object" construction.

[zompist]

It's not that complement clauses are interpreted as being the objects of verbs, that's what they are.

You can have sentential subjects too.

[ná'oolkiłí]

Georgian, despite having a pretty exotic MSA, doesn't really do anything exotic in regards to complement clauses; they're just treated like direct objects.

ივანემ იცის, რომ მოხვალ, და ქეთევანმაც ის იცის.
ivane-m icis, rom moxval, da ketevan-ma=c is icis
Ivane-ERG knows, that you.will.come, and Ketevan-ERG=also 3S.NOM knows
"Ivane knows you'll come, and Ketevan knows it/this too"

ივანეს ჰგონია, რომ მოხვალ, და ქეთევანსაც ის ჰგონია.
ivane-s hgonia, rom moxval, da ketevan-sa=c is hgonia
Ivane-DAT thinks, that will.come, and Ketevan-DAT=also 3S.NOM thinks
"Ivane thinks you'll come, and Ketevan thinks so/this too"
(this verb takes an experiencer (ie, dative-marked) subject)

Alternatively, for certain verbs, the complement clause can be expressed as a nominalization.

შენი მოსვლა მინდა / მინდა მოხვიდე
šeni mosvla minda / minda moxvide
your come.NOMLZ.NOM I.want / I.want you.come.OPT
"I want you to come"
(this verb too has a dative subject, so the object is nominative)

[vec]

It's not that complement clauses are interpreted as being the objects of verbs, that's what they are.

You can have sentential subjects too.

That he is right comes as no surprise.

Correct me if I'm wrong but I would assume an ergative language works the same, except with different underlying "cases":

I.erg think [he's crazy].abs
It is thought [he's crazy].abs
I.abs think.antipassive
[That he is right].erg comes as no surprise.

[Salmoneus]

It's not that complement clauses are interpreted as being the objects of verbs, that's what they are.

You can have sentential subjects too.

Can you?
You can have that-clauses as either subjects or objects, but is that the same as having an independent clause as subject or object? That-clauses are not independent. You can have independent clauses as objects - can you have them as subjects? I assumed so, but I can't think how, so maybe you can't?

[zompist]

I missed "independent" in the original question. You may be right, but it's very alien to how clauses are handled in generative grammar. There, it's all just underlying S's (sentences).

S = subject:

That John came surprised me. [simplest case— add a subordinator]
It's surprising that John came. [supply dummy subject]
For John to come took a lot of nerve.
John's coming surprised me. [gerund form]

S = object:

I know that John came.
I know he came.
I'm surprised by John's coming.

Historically, so far as I know, omitting "that" is recent— you can't omit it in most of the European language I know. Since the OP was talking about different alignments, I don't think this rather English-specific distinction is that relevant to the question.

[ná'oolkiłí]

It's not that complement clauses are interpreted as being the objects of verbs, that's what they are.

You can have sentential subjects too.

Can you?
You can have that-clauses as either subjects or objects, but is that the same as having an independent clause as subject or object? That-clauses are not independent. You can have independent clauses as objects - can you have them as subjects? I assumed so, but I can't think how, so maybe you can't?

I'm not sure I understand. By independent you mean unsubordinated, right? What's an example of an independent clause as an object? Even when you drop that in something like John said [ Ø you'd come ] the object clause is still subordinated. If you're getting at the fact that predicates don't seem to ever be subcategorized to take a clausal subject, I think that is a correct observation — and an interesting one.

I missed "independent" in the original question. You may be right, but it's very alien to how clauses are handled in generative grammar. There, it's all just underlying S's (sentences).

What do you mean by alien? Depending on what theory you subscribe to, all clauses are CPs (complementizer phrases). Some verbs, like say or want, can take CPs as their complement, parallel to how others take DPs (kill), PPs (come), or (apparently) nothing (decay). In some cases, a CP (perhaps dominated by a silent DP) can fulfill the syntactic role of subject.

[Astraios]

Correct me if I'm wrong but I would assume an ergative language works the same, except with different underlying "cases":

I.erg think [he's crazy].abs
It is thought [he's crazy].abs
I.abs think.antipassive
[That he is right].erg comes as no surprise.

Basque at least does (though ergative + 'come' = no):

Zoro dela uste dut.
[zoro da-el]-a uste dut
[crazy AUX-CMPL]-ABS think AUX
I think he's crazy.

And Lakota being active-stative just nominalizes it with the article čha and then the nominalized bit is stative:

Witkó čha iblúkčaŋ.
witkó=čha i<wa>yúkčaŋ
crazy=CMPL think<1S.ACT(;3S.STA)>
I think he's crazy.