I have been assuming all this time that there are languages (like Russian, but also just found a paper on another language which has this kind of data) which allow in situ questions generally but require fronting in embedded clauses.
(1) Ты где работаешь?
2SG where work.2SG
Where do you work? [rus] -- this is a very high frequency sentence, though there is an emphasis on *where*, the stress certainly goes there.
(2) *Я знаю, ты где работаешь
1SG know.1SG 2SG where work.2SG
Intended: I know where you work [rus]
I am not sure however how to model this.
I can say that the insitu-phrase is [ MC + ] which will disallow it from embedding but that won’t prevent it from taking the entire complex clause as its daughter!
I cannot say that the embedding verb requires a complement that’s QUE-empty (because we need question phrases to cross the clause boundary).
I was hoping to reuse the L-QUE feature, perhaps turning it into a luk and doing something clever, but I don’t see how to make it work so far (though I may simply be missing the right logic).
Now, the theoretical literature does not like the idea that fronting can be optional, so much so they will insist the in Russian sentences like Ivan where goes? both Ivan and where have moved :).
I wonder how deep of a rabbit hole I am going into here. Does it sound like modeling different possibilities for embedded and main clauses was a bad idea? Generally, thoughts on this?
…it seems like I want to say that the embedding verb takes something that either contains a gap or does not contain a question (is QUE-empty). Would it be crazy to create such a type (is it even possible)?
Oh wait, maybe I can in fact say that the embedding verb wants QUE-empty complements because that does not in fact contradict extraction from said complements? An empty SLASH would contradict it but not an empty QUE? (as discussed here)
The above seems to work, anyway…
Is this something actually attested in the typological literature, or just a logical possibility given the parameters of your typology? If the latter, I’d say not to spend time trying to implement it…
It is attested (see first post); it is the syntactic literature that objects to it. I can kind of see why though, given that there is special intonation involved in such cases. But I have kind of been ignoring that across the board and still modeling in situ in e.g. russian.
Ah sorry, I was led astray by your statement that you’ve been assuming some language type to exist. Attested data are attested
Anyway: my concerns with insisting on QUE empty complements would be:
I think this would rule out cases of multiple wh questions where only one moves:
Who did Kim think saw what?
Are those allowed in Russian?
It also predicts that you don’t get echo questions in embedded contexts (maybe you want this, though?) unless echo questions don’t involve non-empty QUE:
A: Kim said you saw a beautiful lampyrid beetle
B: Kim said I saw a beautiful WHAT?
Yeah, you are right…
Disallowing echo is what I wanted in this case (I just keep calling them in situ but they probably are echo), but I certainly wouldn’t want to rule out Who did Kim think saw what…
So, perhaps it sounds like modeling something like this would be too tricky and maybe that points in the direction where, if you have fronting, you have fronting, and stuff that appears in situ is actually not really a question in the same sense (echo etc.) and can be considered out of scope?
…and a semantic variable indicating that the question is echo would be the “slippery slope of messages”? Because a variable like that could solve this problem easily, I guess (it would be introduced by the in situ phrase).
I hope you don’t end up leaving questions like Who did you say Kim saw out of scope … that’s just a case of partial fronting from an embedded clause.
Not sure what you mean by a semantic feature indicating echo, unless it’s a different SF value, which isn’t on the messages slippery slope…
I meant leaving disallowing in-situ questions in embedded clauses out of scope, which would not include partial fronting. Rather, it would mean removing the choices like “There is optional fronting in main clauses but there is obligatory fronting in embedded clauses”, from the questionnaire. Which Russian (and this other language I found) might look like it has it but the “optional fronting” clearly has something special going on, probably echo.
Yes, I meant an additional SF value. Does that sound like a good idea?
Leaving out those options would be fine.
For an additional SF value, it might work, but that’s not obvious to me. Would it be a subtype of ques or a sister type to it? What would you want to see on echo questions whose ‘base’ is a regular question?
A: Who saw the lampyrid beetle?
B: Who saw the WHAT?
I was thinking that ques could have two children types, something like canonical-ques and echo.
Good question about example B. I think I would like to see just echo on the SF there?
…It’s interesting in example B that you need a determiner. I wonder if it means there is something different about this question? Or is a determiner just optional in such situations?
As for the SF feature: creating two children for ques won’t work of course because I need the clause-embedding verb to take not only canonical questions but also propositions :). But maybe a type prop-or-canonques, and then canonques is its child, not ques’s (or not only ques’s)…
OK well, I think that disallowing in situ questions in embedded clauses doesn’t make sense after all. I can sort of see why people insist that they sound bad, but I also do think that it all comes down to context and that echo questions should be possible everywhere, you can always construct a sentence with appropriate context and intonation and then an echo question can be embedded and what not. Sort of by definition of echo. So I think I will abandon this.
In particular, my SF solution will still not prevent the full sentence like I know Ivan saw WHAT? from going through an insitu rule at the end, and I think it’s just fine, you can say this with a certain intonation.
Interloper here: I think “Who saw WHAT?” is grammatical, no determiner required.