SIG on Wh Extraction, pied-piping, superiority, Wh questions library

Present: Olga Zamaraeva (OZ), David Inman (DI), Emily Bender (EB), Antoine Venant (AV), Kristen Howell (KH), Guy Emerson (GE), Berthold Crysmann (BC)

Scribe: Emily Bender

OZ: Phenomena where other constituents can come along with the wh word in extraction (required or optional). Russian has optional pied piping of NOM with wh determiner.

DI: Can you also say: “Good Masha gave to Ivan book”?

OZ: Yes.

DI: So this might just be part of Russian allowing discontinuous noun phrases.

EB: But what’s interesting here is that something can come along with the wh word.

OZ: When looking at embedded questions, things get more complicated — ex (3) v (4). Could “you say” in (3) be something different?

EB: Could be a parenthetical, which would put “at which” and “hour” in the same clause in (3), and so the discontinuous NP thing is clause bounded, and pied piping of NOM with “which” becomes required.

OZ: So my library says…

EB: Discontinuous NPs are not your problem (for this library), so you under generate (disallowing (3) and (2b)).

GE: Especially if you can get that discontinuous NP with a different determiner, like “at that you say they arrived hour”.

OZ: Yes.

[Moving on to superiority effects]

OZ: Why aren’t we contrasting (1a) with (5)?

EB: Because non-subject wh questions require subj-aux inversion. You can’t say “What did Kim?” either.

DI: English specific?

OZ: Usual claim is that Bulgarian and Serbo-Croation have superiority effects, but it’s always the same few examples. Russian doesn’t. ‘What who did’ looks worse in Russian than ‘Who what did’, so maybe those judgments in Bulgarian and S-C are iffy.

DI: Ask a non-linguist to say it, and if they feel weird, then don’t put it in your paper.

GE: But it might depend on context.

BC: Set up context where the order of patient-agent is salient in the context.

BC: German happily fronts objects anyway, and don’t have superiority effects. English has a strong clustering of subject properties with topic…

EB: But wh words are always foci, so topic isn’t relevant here.

BC: Point taken. But you get strong subject first ordering preference out of that association…

OZ: Berthold, do you agree with these German examples? (slide 6)

BC: Both are okay, but “der Peter” isn’t my dialect.

OZ: So I shouldn’t do superiority in my library?

EB: Indeed — allow multiple wh questions with the right flexibility about how many can extract, but don’t worry about modeling superority. Include in your testsuites, but just say I don’t get that far.

OZ: Because the data isn’t clear anyway.

EB: Yes, and if someone were to try to model it on top of your library, it’s probably a question of adding constraints, so that’s fine.

BC: How about: “Helen tried to figure out what device which patient ordered.” Also similar cases in German. From paper Häussler et al 2015 “Superiority in English and German”.

GE: Multiple wh questions already require a lot of context to be sensible.

[Discussion scribe missed about people wanting distinction in meaning between “who what did” and “what who did” in Russian but the contrast seeming to have to do with information structure.]

BC: Those S-C examples also have suspiciously short wh phrases (just a single word) and longer ones might work better.

OZ: Adding the (general) focus particle makes some of these better. “Who arrived where” If you put “where” before “who”, maybe it’s not so good but adding the focus particle next to “where” it sounds a bit better, but even without it it’s okay.

OZ: Back to German example (from Yehuda Falk 2012 “Superiority Effects” )

BC: (b) is worse, as I said before, but wh words from the same clause can go in either order, but here the object wh word is from the embedded clause. Want to remove complementizer and proper noun….

EB: As a general rule, these kind of fine-grained distinctions aren’t compatible with a Matrix library.

GE/DI: What about resumptive pronouns? Fun but more work. Eg: “Who did Kim say Sandy saw them?” (EB: star for me!)

EB: I think cross-linguistically this is more common with relative clauses than wh questions, with some languages doing “The dog they saw it barked.” but you don’t like “Who did you see them?” right?

GE/DI: Yeah, that’s out.

OZ: Displaying current questionnaire … remove stuff on superiority.

EB: Yep! Also, you might want to clarify that the first two sections are about wh questions where the wh word stays in its own clause.

OZ/EB: Is the subj-aux inversion in matrix non-subj questions only pattern too English specific to include? Maybe. Are there other matrix v. embedded differences that motivate keeping those two separate in the questionnaire?

[Back to pied-piping]

OZ: Clausal pied-piping: (81a) in Szendroi on Basque: who write aux-comp book say aux Peio, meaning “Who did Peio say wrote the book?”

OZ: I’m suspicious about examples with ‘say’, even with that translation. “Who wrote the book, Peio said?”

EB: What’s interesting here is how to associate the wh-ness of ‘who’ with the matrix clause. Pied-piping of the clause would get this.

OZ: Should I allow pied-piping of clauses?

GE: 81a could just be in situ? Clause fronting + wh in situ would get you that.

BC: I would in general steer clear of whatever’s been said about focus on Hungarian. There’s something dogmatic about there being only one specific position…

[Back to questionnaire}

EB: Pied-piping question should be something like:

Do you get pied-piping of NOMs with which? [ ] optionally [ ] obligatorily [ ] not at all Do you get pied-piping of Ps with wh complement? [ ] optionally [ ] obligatorily [ ] not at all

OZ: Should that be once, or separate for matrix/embedded/extracted from embedded?

OZ: Back to (3) & (4), do we need do we need to say that Russian requires pied piping for out of embedded, but optionally in same clause?

GE: But treating discontinuous NPs as a separate phenomenon, which intersects.

BC: Serbo-Croation: Gerald Penn 1999 paper on domain structuring. Also clause-bounded

OZ: So it sounds like pied-piping choices should be across the board? Current version allows more flexibility, but but less theoretically satisfying.

GE: Also more user-friendly.

OZ: And developer-friendly, too! Need far fewer regression tests…

BC: If English prime was good enough for Montague, then Russian prime should be good enough for you :)

EB: The design principle of capturing generalizations serves us well in single grammar grammar engineering, helping us to build grammars that are maintainable over time. Good to apply to Matrix libraries as well.

CambridgeWhQLibrary (last edited 2019-07-18 09:56:31 by EmilyBender)

(The DELPH-IN infrastructure is hosted at the University of Oslo)