Thursday, September 29, 2005

3 sluicing puzzles: Number one

I don't actually know anything much about sluicing or ellipsis, but this past year I've learned a lot more than I had known, notably from a talk I heard at MIT in March by Sandy Chung, reporting one of the neatest observations I've heard in a long time, and now from a supercool seminar that Andy Barss is giving here at Arizona on the topic. Also, last year, a student in my fieldwork class at Harvard was interested in sluicing, and now this year I'm on the prelim committee of Dave Medeiros, one of our graduate students, and he's been thinking about it. Plus recently Rob Stainton sent me a paper, and then kindly invited me to come and kibitz at an ellipsis Erniefest he's organizing next year. So I'm gradually getting up to speed.
In the course of learning about it, I've run across a few related phenomena that (I think) are surprising or interesting. They may be discussed in the literature -- as I say, I'm no expert -- but so far I haven't seen 'em talked about. I don't have any immediate plans to investigate, but I'd like to tell about them, in case some one else might find them interesting.

The first one is a discovery from our fieldwork class on Finnish at Harvard last year, where undergrad Jeremy Hartman was interested in sluicing, and elicited some surprising things. The main surprise was that Finnish seems to violate Merchant's generalization concerning sluicing and preposition-stranding. This may have been noticed before, but this was the first I'd learned of it.

Merchant's generalization is this: A language L will allow preposition-stranding under sluicing iff L allows preposition stranding under regular wh-movement.

That is, in English, we can say

1. John talked to someone, but I don't remember who.

because we can say

2. Who did John talk to ?

... whereas in French, you can't say

3. *Marie a parlé à quelqu'un, mais je ne sais pas qui.
Marie has spoken to someone, but I neg know not who.

because you can't say

4. *Qui a Marie parlé à ?
Who has Marie spoken to?

In French, you can't strand a preposition when you wh-move the NP associated with it; rather, you have to pied-pipe the preposition along with the question word. That is, you have to say

5. a. À qui a Marie parlé?
To who has Marie spoken?

b. Marie a parlé à quelqu'un, mais je ne sais pas à qui.
Marie has spoken to someone, but I neg know not to who.

(Excuse any mistakes in my French, 'tis been a while. But the point about preposition stranding is true).

Anyway, the point is that the possibility of a sentence like 2 or 4 in a language is supposed to predict the possibility of a sentence like 1 or 3 (and vice versa). It's a very robust generalization, it seems. But Finnish, according to our consultant Santeri Palviainen and another Finnish friend of his, does not allow sentences like 2 or 4 but does allow sentences like 1 and 3.

Here's the illustrative examples, taken from Jeremy's paper for the course. First, examples which show you can't strand an adposition in a wh-question in Finnish:

6. a. Kene-n kaa sä leiki-t
who-gen with you.nom play-2sg
With who(m) are you playing?

b. *Kene-n sä leiki-t kaa
who-gen you.nom play-2sg with
Who are you playing with?

Because 6b is ill-formed, while 6a is ok, it's clear that the adposition cannot remain in its base position, but has to move to the front of the sentence with the wh-word.

And now, here is the sluice:

(7) Se leikki-i jonku-n kaa, mutt-en tiiä kenen (kaa)
He.nom play-3sg sb-gen with, but-neg.1sg know who-gen (with)
He’s playing with someone, but I don’t know (with) who.

The adposition kaa is perfectly omissible, indeed, Santeri suggested that the sluice is somewhat more natural without it.

Finnish is of course probably somewhat special in terms of its adposition inventory, since it's got a zillion nominal cases to play with, but it definitely has lexical items that qualify as adpositions (both pre- and postpositions, and some that are both), and they definitely can't strand in wh-questions.

Some, though, seem to do something funny when they pied-pipe along with their wh-NP. Compare (8) and (9):

(8) T tule-e ennen uu-ta
T come-3rdsg.pres before U-part
‘The T comes before the U.’

(9) Mi-tä ennen T tule-e
What-part before T come-3rdsg.pres
‘Before what does T come?’

That is, a preposition like ennen appears after the wh-phrase, rather than before it, when it pied-pipes in a wh-question. This might be the place to look in trying to reconcile Finnish with the generalization, I suspect -- it's maybe like Finnish does mandatory swiping in wh-questions, without the deletion part. (It's worth noting that in other respects, such as in exhibiting case-matching, Finnish sluicing is quite well behaved).

Anyway, that's the first thing. I'll post the other two (much less involved) observations sometime in the next week, but I'm not sure exactly when.


Update: Francophones -- check out this survey on sluicing posted by Chris over at Serendipity. How bad are those prepositionless wh-words?


Blogger Chris W said...

I'm not entirely sure about your statement about French. Not the part about the impossibility of preposition-stranding in relative clauses -- that's impossible, outside some Canadian dialects, the syntax of which is clearly influenced by English. But about the ungrammaticality of your sentence 3.: sure, the rule books would frown upon it, but I believe I'm hearing that sort of thing quite a bit regardless.

I've posted an informal survey on my own blog -- maybe a few francophone readers will help clear this up.

As for German, I have to mull that one over, even though it's my native language.

4:53 AM  
Blogger Bridget said...

Yay for Jeremy! I'll have to tell him he's famous when I see him in bioling on Monday. :-)

7:08 AM  
Blogger Lance said...

Oy, sluicing. As you likely know (weren't you following my vaguely-linguistic, vaguely-dissertationy blog?), sluicing's one of those things I wish I had the time right now to be interested in. I swear, I don't even understand the generalizations at this point, never mind the theories that explain them. (Pseudosluicing? Swiping? Sheesh.)

Karlos Arregui, the accuracy of the spelling of whose name I do not guarantee, recently gave an interesting Ling-Lunch about sluicing. Naturally, I've forgotten everything he said, but I've got the handout. Somewhere.

4:08 PM  
Anonymous Anonymous said...

I think the name is "Santeri Palviainen" instead of "Santieri".

I wonder if "kaa" is cognate with Estonian case ending -ga. It's comitative, I believe.

3:56 AM  
Blogger Luis said...

For what is worth, I can optionally have P-stranding in Spanish sluicing, even though P-stranding is not possible in Spanish otherwise.

Maria ha salido con muchos chicos, pero no se (con) cuantos
"Mary has gone out with several boys, but I don't know (with) how many)"

The version with the preposition sounds better than the one without. One more thing is that P-stranding is impossible for dative "to"

Juan les ha dado un libro a muchos estudiantes, pero no se *(a) cuantos
"John has given a book to many students, but I don't know (to) how many"

Incidentally, I think that the P-stranding version of this sentence sounds kind of weird in English too, but I'm not a native speaker, so...

7:42 AM  
Blogger hh said...

Eek! Of course it's Santeri, not Santieri. I just get so into throwing extra 'i's around in Finnish. It's fixed now. Sorry about that, Santeri, if you saw it...

W/r to dative 'a' being non-strandable in Spanish -- the English equivalent ('...I don't know how many') sounds perfect to my ear. But dative 'a' is morphologically special at least in French; Miller 1992 uses various similar tests to argue that 'a' is not a real preposition but rather a case marker.

10:16 AM  
Blogger Luis said...

True, my mistake, though the generalisation for Spanish, if I remember from Cuervo's thesis right, is that "a"-datives are DPs when clitic-doubled and PPs when not doubled. So yesterday's example doesn't count because it had a clitic ("les"). Here goes a cliticless version, so it should be a PP, yet P-stranding is still out for me, in contrast to the "dating" example.

Juan ha mandado su ultimo articulo a muchas revistas, pero no se *(a) cuantas
"John has sent his latest paper to several journals, but I don't know how many"

I will try this out with other types of datives...

1:36 AM  
Anonymous Anonymous said...

Hi Heidi! Thanks, again, for forwarding me the link to these data.

BTW: It may be worth bearing in mind that Merchant's own data in his 2001 book are kind of "noisy" as regards his P-stranding generalization. He even has a few pages devoted to discussing what could account for this "noisy-ness", under the assumption that the P-stranding generalization is true.

Also, there are now a variety of other persons now claiming for a variety of other languages (Brazilian Portuguese, Chinese, Malagassy, Serbo-Croatian) that Sluicing *can* generally save P-stranding.

I'm interested in cases like this, as I have an account that (perhaps wrongly, perhaps correctly) predicts that sluicing *should* save P-stranding. Of course, under this account, the analytic challenge is to understand the cases where slucing seems not to save P-stranding. From what I've seen, though, there's no prima facie reason for one or the other being the "exceptional" case.


10:38 AM  

Post a Comment

<< Home