Spurious E:nsubj:xsubj

[nschneid]

• [x] EWT: http://universal.grew.fr/?custom=63a51efae73e0
• [x] GUM: http://universal.grew.fr/?custom=63a3c1ca1860c (first 3 may be OK but 4th one should be ccomp instead of xcomp)

[amir-zeldes]

OK, I changed the 4th GUM one to ccomp, but you agree with the xsubj, right? I mean, in "we owe it to future generations to make safety concerns come first", we are the ones who "owe", and we ware the ones who will "make", so the external subject of that second infinitive is indeed we, right? I think it's xcomp for that reason, but I recognize it's unusual because it's extraposed and replaced by the expletive object 'it' (we owe it = we owe to make..)

[nschneid]

No I think the relationship between "we" and "make" is a nonsyntactic inference. Paraphrase: "We owe prioritization of safety concerns to future generations". Unless you want to say "owe it to VP" is a syntactically special idiom with obligatory control.

Compare: "I leave it to you to decide". Syntactically similar, semantically different based on the meaning of "owe" vs. "leave".

[nschneid]

(At some point we may want to expand the types of things covered by enhanced edges to include adjunct control for instance—like purpose clauses. And maybe this, if we're willing to have a greater level of ambiguity in the enhancement procedure.)

[amir-zeldes]

Mm, actually that example convinces me more that it's control, since we see we have subject vs. object control based on the verb, just like in regular infinitive control. Isn't that 'it' an expletive object to facilitate postponing a heavy clausal object? If so, wouldn't that argue for structure sharing? Or otherwise, what's the deprel of 'it'?

[nschneid]

Mm, actually that example convinces me more that it's control, since we see we have subject vs. object control based on the verb, just like in regular infinitive control.

Well isn't it that subject control verbs never have a direct object? So it's not based just on the semantics of the verb, it's based on transitivity of the verb token. (I want to leave - subject control, I want him to leave - object control)

Isn't that 'it' an expletive object to facilitate postponing a heavy clausal object? If so, wouldn't that argue for structure sharing? Or otherwise, what's the deprel of 'it'?

My impression (which could be mistaken) was that xcomp was meant to represent matrix-verb-licensed controlled complement clauses, and one of the ways xcomps differ from ccomps is that they can't be targets of it-extraposition.

But "owe it to (somebody)" and "leave it to (somebody)" could be idioms with odd behavior. Not sure if this helps (CGEL p. 963 regarding verb + "it" + content clauses):

[nschneid]

^ may be relevant to some of the miscellaneous cases in #176

[amir-zeldes]

Well, they're not right about the 'it' being obligatory in ii. For "like" there are plenty of examples which sound fine to me:

• he didn’t like that I was an outcast dating someone from his world link
• he didn’t like that I called a timeout link

And if they agree it's redundant in iii., then it speaks strongly for this just being an expletive it in object position (otherwise we have proper obj+ccomp, which I doubt is the case here). For things like "taken (it) for granted that", weight makes a lot of sense for motivating such an expletive.

If we accept this is an expletive, then I think in "we owe it to X to Y" we could be looking at xcomp. I mean, that presupposes that without extraposition this would habe been:

• we owe to future generations to make safety come first

If so, then structurally it's a subject control equivalent to object control:

• we must help future generations to make safety come first

And owe is like want in your example.

[nschneid]

Agree about [ii] being grammatical—I think there's a register difference though: "didn't like that I was..." sounds casual to me, whereas "don't like it that I was..." sounds appropriate for both formal and informal registers.

By @jnivre's favored definition of expletive as the function of a nonreferential pronoun, it certainly fits. Whether it's extraposition is a different question. The usual test for extraposition is substitution of the extraposed element (the clause) for the pronoun. So, "We owe to VP to someone".

"We owe to VP to someone" sounds potentially unidiomatic but could be the historical origin, at least, of the idiom. Can you find any attestations of this?

Note that you can "owe someone money" (double object) but not "We owe someone to VP". Unlike "leave"—"We leave someone to VP" is object control.

[amir-zeldes]

Note that you can "owe someone money" (double object) but not "We owe someone to VP".

That does not appear to be true, I see "We owe them to be remembered", "We owe them to continue the education campaign" (WaPo), and others. I think if you don't have the 'it' then you can get the preposition-less recipient (like "we owe them a sandwich") but if you have the "it", it has a very strong pressure to be realized first (cf. the uncomfortable "?give John it"). So basically yeah, I think it is both non-referential (or coreferential with the VP really), and also substitutable. In fact I can even find "And we owe to them to make it better" and similar, but those are rarer.

[nschneid]

"We owe them to continue the education campaign"

Interesting! That was from a Harvey Milk speech in 1978.

"owe PRON to VERB" gives 14 hits in COCA. So it happens, but is marginal.

But that's with the recipient first. I don't see any matches of "owe to VERB to". If "owe it to someone to VERB" were it-extraposition we would expect that as the non-extraposed variant.

[nschneid]

Also, "to VERB" doesn't have to be explicit for the idiom to work: "I did that because I owed it to her." [I think I can get either a referential or nonreferential reading here.]

[amir-zeldes]

I don't see any matches of "owe to VERB to"

It's rare, but as I wrote above I found "And we owe to them to make it better".

I thought it might be a typo, but it recurs here and there, for example the Pennsylvania Republican Caucus:

• We also owe to them to do everything in our power to grow their ranks to ensure responders will always be there to answer the call link

In any case, I think expl is the most elegant thing to do with the 'it'.

[nschneid]

I don't see any matches of "owe to VERB to"

It's rare, but as I wrote above I found "And we owe to them to make it better".

I thought it might be a typo, but it recurs here and there, for example the Pennsylvania Republican Caucus:

• We also owe to them to do everything in our power to grow their ranks to ensure responders will always be there to answer the call link

That's not what I was asking: "owe to them to..." has the pronoun first. I am asking about the clause first.

[amir-zeldes]

Oh, OK. But that's really easy to explain away with weight, I mean, expecting a clause to beat a pronoun for serialization is a pretty tough ask. I don't think the fact that the realized order is the way it is really speaks against this being expl.

[nschneid]

I'm fine with "it" as expl because it's (arguably) nonreferential. But extraposition would require that the "it" can in principle be substituted with the clause.

[nschneid]

Here's why I think it's not extraposition: even if the NP is heavier than the clause, "it" is required:

• We owe it to everyone who worked so hard putting on this event to have a good time.
• *We owe to have a good time to everyone who worked so hard putting on this event. [clause + prepositional dative, no expl]
• *We owe everyone who worked so hard putting on this event to have a good time. [object + clause, no expl]

[amir-zeldes]

So if you accept expl for the pronoun, what label would you use for the clause? I don't think UD really distinguishes extraposition as a category. If you label the clause as a complement, it's not really possible to know if it's extraposed or not - you're just saying that it syntactically saturates the object slot in the verb's valency (or semantically, the thing owed). Do you see it differently?