Talk:Sluicing

Non-constituent?

This sentence puzzles me: "Sluicing is problematic in the field of syntax, as the elided content seems to form a non-constituent." As far I can tell this is simply not true: sluicing involves deletion of TP or IP (which are of course constituents). Is the author thinking of 'swiping'?:

Phoebe is eating rice, but I don't know what with.

This certainly seems to involve deletion of a non-constituent. Matve 18:40, 11 April 2006 (UTC)

Sluicing is also a type of mining - Hydralic mining used to mine Tin - at leaste in the 1930's. I'm not sure how to create a disambiguation but someone might want to do this?

This page needs some major work. It doesn't make any sense at all in its current form. 71.30.48.94 17:00, 17 February 2007 (UTC)

I think it's fine though I agree it should be fleshed out a lot more. As far as non-constituency is concerned, it might have been better if they had provided an example. I think the author was thinking of swiping (Ross discusses the non-constituent ellipsis possibility w.r.t. this). Also, some clausal ellipses in some languages (Turkish, see Ince 2007) leave more than one remnant which could plausibly be analyzed as non-constituent ellipsis Mattbarros (talk) 07:12, 1 March 2009 (UTC)

Some clarification for non-linguists?

I'm just wondering whether someone who is better-versed in linguistics than I might be able to provide links on the page for two details: firstly, what does that little italicised 'e' represent? And secondly, what does the word 'matrix' mean in this context? (I think maybe just some strategic links on the page might do.) Thanks for any help you can give. Edisk 04:40, 10 August 2007 (UTC)

neither of the words you mentioned have pages :( putting in some clarifying text, hopefully it'll be clear enough for non-linguists. Nyoro n (talk) 20:44, 14 June 2008 (UTC)

The 'e' is superfluous here, though it is used by some researchers to indicate an ellipsis site ('e' for ellipsis). I've removed it, since it's not necessary to understand the point. Also added some traditional terminology for 'matrix' questions. Will try to clean up and expand the rest of the entry when I have more time.Mundart (talk) 17:01, 30 April 2009 (UTC)

VP Ellipsis isn't just in English

I am going to remove the following quote from the introduction of this article: "unlike for instance VP-ellipsis, which occurs primarily in just English".

According to Goldberg (2005), VP ellipsis occurs in plenty of languages other than English. If this is a controversial point, and someone has a peer-reviewed source disputing Goldberg's claim, that would be interesting to discuss in the article on VP ellipsis. But for now, since I know of no such claim, I'm removing the aforementioned quote.

Goldberg, L. 2005. Verb-stranding VP ellipsis: A cross-linguistic study. Doctoral Dissertation, McGill University, Montreal. (available here)

--N-k (talk) 19:35, 10 December 2012 (UTC)

Modifications proposed

Two things occur to me in looking at this article. First, it is not possible for someone who doesn't know the phenomenon already or the data to understand the presentation of the elided material. We need to indicate that the subscripted material is intended to be not pronounced. I actually think the examples should just use English orthography, and then we can give hypothetical unpronounced material in the analysis section. This will help clarify what the data are, vs. what the claims about the data are: there are many analyses of sluicing (and of ellipsis more generally) that do not posit any unpronounced material, and so would object to defining the phenomenon with reference to such (Polly Jacobson 2014, Dalrymple et al 1995, Ginzburg and Sag 2000 are prominent exemplars). Second, the Osborne et al 2012 Syntax paper does not discuss multiple sluicing, so the extension of their analysis (of multiple remnants in fragment answers and gapping) to sluicing appears to constitute original research. Could someone find a citation to the lit where the analysis that is presented in this article is proposed? Then we can add that citation as an appropriate source for the analysis of multiple sluices as involving catenae. (It's still an open, and debated, question whether the remnants in gapping, fragment answers, and multiple sluices should all be analyzed the same way, of course: that's a separate question.) If anyone can make these changes in the next week or so, please do so! Otherwise, I may be able to get back to this and make the changes myself sometime after next week. Mundart (talk) 17:17, 3 January 2015 (UTC)

I have made some minor changes to attempt to accommodate a couple of your comments. Concerning multiple sluicing and catenae, you are correct that Osborne et al. (2012) do not address multiple sluicing explicitly. Osborne et al. (2012) do, however, produce examples of many similar cases of ellipsis (from gapping, stripping, pseudogapping, answer fragments, comparative deletion) in which the elided material is a catena but not a constituent. Further, they produce an example of sluice stranding, which is similar to cases of multiple sluicing insofar as there is more then one remnant in the sluiced clause. Their example: He wanted to solve the problem, and she asked what _{he wanted to solve the problem} for. I therefore do not think that information about catenae and sluicing in the article "constitutes original research". The validity of the catena concept has been extensively established in numerous publications. See the list of catena publications here.--Tjo3ya (talk) 05:47, 4 January 2015 (UTC)

Since there isn't any published research using catenae to analyze multiple sluicing (and swiping is not the same as multiple sluicing, of course; I don't know a single published analysis that argues they are), any suggestion that the analysis using catenae for gapping be extended to multiple sluicing is by definition WP:OR. Even if it were a plausible or true analysis (not for us to decide or debate here, of course: that's for the journals), we can't have it here unless it's in a reliable source and, ideally, represents a fairly consensus view in the relevant literature. A separate, but equally important, issue is the heavy use of "catena" in this and related pages. Given the marginality of its use in the published literature, featuring it so prominently here in Wikipedia (in fact, in anything but a footnote) is clearly WP:UNDUE. I'd love to have a balanced article featuring phrase-structure approaches and non-structural approaches, etc. I doubt I'll have time to write such an article, but in the meantime, the article is better served by a silence on these matters. Mundart (talk) 03:46, 6 June 2015 (UTC)

Mundart, I have overlooked something that is directly relevant to our exchange here. My contribution to the Oxford Handbook on Ellipsis includes examples of multiple sluicing. Thus based upon your statement above, i.e. "

Mundart, your choice of terminology, i.e. "non-structural approaches", is curious. I think it illustrates too little exposure to other approaches to syntax outside of mainstream GB/MP. Dependency-based approaches can be just as structural as phrase structure approaches. Concerning journal articles and the catena unit, there are more manuscripts under review, so it is going to be increasingly difficult to ignore the potential of the catena unit going forward. The reason the catena unit is well represented in Wikipedia is of course because I put it there. I invested huge amonts of time expanding Wikipedia's coverage of syntax, and readers world wide are now benefitting from my efforts. If I had not been able to cite my own published works (in good peer-reviewed journals) when relevant, I would not have put forth the tremendous time and effort. These points should be considered when evaluating the appearance of the catena unit in some Wikipedia articles on syntax.

The key words in your comment are "I doubt I'll have time". I had the time to do the work, and I did do the work. Wikipedia is fantastic in this regard. In any case, I welcome your desire to contribute, and when you do decide to contribute, I will read your additions closely, attempting to ensure that you do not ignore the syntax world outside of the narrow GB/MP paradigm that I suspect consitutes the main chunk of your background in syntax. --Tjo3ya (talk) 11:20, 6 June 2015 (UTC)

The terminology I use here is from Merchant's article in the Oxford Handbook of Ellipsis (version [[1]]). It refers to a debate that has been at the center of theorizing about ellipsis since the 1970s (see the references in that handbook article). I do think dependency approaches would fall into the structural camp (since they seem to posit syntactic structure inside the ellipsis site); whether the identity relation is sensitive to that structure or not is a separate question (again, see the Merchant overview article for references and discussion). The Osborne et al paper posits that the identity relation is stated using the catena: a catena E can be elided only if there is a discourse-accessible identical catena A (A =/=E). Like all structural identity theories (including eg Merchant's 2013 LI paper version) it therefore seems to have trouble with exophoric ellipses (Miller and Pullum 2013). Far from ignoring non-GB work, I was the one who introduced Ginzburg and Sag to this article; and I promise I will take the time to ensure that the material on this page reflects the opinio communis in the field of ellipsis. I look forward to seeing your published work on multiple sluicing, at which point I'll be the first to refer to it and include its approach to multiple sluicing. Until that point, however, such conjectures (about how a catena-identity-based analysis of ellipsis would work for multiple sluicing) are WP:OR. Mundart (talk) 13:31, 6 June 2015 (UTC)

Mundart, who are you? You know who I am, but I do not know who you are. Why not reveal your identity? I might be more willing to take your points seriously if you were to do so. Remaining anonymous allows one to be bolder than one would otherwise be. I was not yet aware that the Oxford Handbook is yet out. In fact I think it is not yet out. I think the editors are having difficulties gathering all the contributions, so it may be quite awhile before Merchant's article is published. Your point about multiple sluicing and the catena unit is not fair. Extending the catena analysis to mulitiple sluicing is obvious, and an article that did just that would likely have difficulty getting published due to the redundancy of the message. I'm going to wait to allow you to respond before I revert your most recent edits. In the scheme of things, it looks like we may need third party arbitration here. --Tjo3ya (talk) 14:28, 6 June 2015 (UTC)

Mundart, I've just taken a look at Miller and Pullum (2012). Exophoric ellipsis is not a problem for the catena-based analysis, since nowhere does the catena-based account claim that there must be an overt antecedent that is itself a catena. The catena analysis claims merely that the elided material should correspond to a catena; it says nothing about a "discourse-accessible identical catena". --Tjo3ya (talk) 14:50, 6 June 2015 (UTC)

Mundart, I have overlooked something that is directly important to our exchange here. My contribution to the Oxford Handbook on Ellipsis includes examples of multiple sluicing. Hence based upon your statement, i.e. "I look forward to seeing your published work on multiple sluicing, at which point I'll be the first to refer to it and include its approach to multiple sluicing" and your trust in the validity of the Oxford Handbook, you should now be good with including statements about the catena-based account of sluicing and multiple sluicing. The trees I originally included can be reinstated. Actually, though, I will wait to do this, since the trees would be more effective if they are compared with phrase structure trees. In other words, I intend to include phrase structure trees as well. I am now going to revert the article back to how it was before you undid my contributions. Please respond here before making additional changes. We want to avoid an editing war. --Tjo3ya (talk) 00:42, 7 June 2015 (UTC)

That's great. Definitely put them in, then. I don't see why that would mean you should revert my edits, though. Also, you've written "movement does not occur, but rather the elided material is a catena". Setting aside terminological differences, surely the analysis you have does use "movement", of the first wh-phrase to its position in specCP (whether by transformation or slash-feature or multi-dominance, etc., is irrelevant). So the puzzle is precisely the same as for Ross and Merchant: how are the island effect voided? (Note that the upside is that the P-stranding effects would have the same explanation.) The problem is with the non-initial wh-phrase in multiple sluicing: as Rodrigues, Nevins, and Vicente have shown, the noninitial wh-phrase strongly respects Merchant's generalization about the P-stranding: even in languages like Spanish and Brazilian Portuguese, the second wh-phrase, if its correlate is in a PP, must be a PP as well. The catena-based analysis has no obvious way to make this cut, especially since you claim the second wh-phrase hasn't moved at all, making the appearance of a movement-sensitive constraint like P-stranding a mystery.

Mundart (talk) 02:02, 7 June 2015 (UTC)

Movement does not occur in the sense that the remnant of multiple sluicing (or gapping or pseudogapping or etc.) need not move out of an encompassing constituent, TP for instance, so that that constituent can then be deleted. Yes, the first wh-phrase does move in the sense you suggest on our account. We prefer to conceive of movement along the lines of feature passing (slash feature). Island effects are voided in terms of an abbreviated cleft-type analysis along the lines suggested by van Craenenbrok (2010). P-stranding is no problem on the catena-based account since the elided material in such cases is a catena, as mentioned and illustrated above with the example: He wanted to solve the problem, and she asked what _{he wanted to solve the problem} for. Merchant's account of P-stranding in terms of remnant movement is plausible only if one overlooks the numerous other cases where movement of the remnant does not work, such as in cases of answer fragments with negative polarity items, e.g. What didn't you try? _{I didn't try} Any of the oysters vs. *Any of the oysters _{I didn't try}. We (my coauthor and I) currently have a major manuscript under review that is examining the nature of answer fragments closely. Finally, the last sentence in the article is not accurate. It's not difficult to create examples of sluicing in which the remnants appear in separate clauses, e.g. A: Some boy said he has a crush on some girl. B: Which boy on which girl? --Tjo3ya (talk) 12:09, 7 June 2015 (UTC)

P-stranding and catena/nonstructural claims

I added the following sentence in the section on Theoretical Approaches back in June: "Neither the nonstructural nor catena-based analysis has been shown to be able to handle the preposition-stranding facts.". Tjo3ya deleted it in November, and I noticed and reverted this deletion this week (21 Dec 2015), based on my reading of the scholarly literature on the question: as justification for my restoring this sentence, I wrote: "The sentence is an accurate statement of the state of the field. The P-stranding facts (how German differs from English, eg) have not yet been analyzed in print. Whether you or I or anyone thinks such an analysis might work is not the issue."

The very next day, Tjo3ya reverts my edit, with both a declaration of war and an admonition as justification (instead of actually engaging with either the sentence or my justification of it): "We are now going to enter an editing war! -- unless you identify yourself, in which case we can negotiate."

Rather than respond to such tactics, I issue a simple call: show us where in the published literature anyone has analyzed the difference between e.g. German and English with respect to P-stranding under sluicing in a nonstructural approach. Cite a paper that derives the fact that German doesn't allow it, but English does. This is core observation or discovery (part of a larger generalization), and the one that persuaded many authors to adopt a deletion approach to sluicing over the alternatives (including the LF-copy version of Chung, Ladusaw, and McCloskey 1995, which also couldn't account for these facts). Unless and until such a reference can be cited, I don't see how the sentence is in any way objectionable: it is a simple statement of fact. It is not prejudicial against nonstructural accounts, nor does it presume to state that such accounts would never be capable of handling the facts (if one thought so, the place to make such a claim is in a journal article, not on Wikipedia). Instead, it is a simple statement, a universal negative, that would be easy to disprove: name one article that "handle[s] the preposition-stranding facts". The facts are those in the section on Preposition-stranding, and what some have called "Merchant's Generalization", and are much discussed in the literature on sluicing: not the fact that English allows P-less remnants under sluicing (every approach can handle this simple fact), but that German (etc.) doesn't. (Again, it's a separate question whether we can imagine or even come up with a plausible such analysis: the sentence at issue is about the published literature, not about our own analytical abilities.)

I'll leave this as is for a day, but unless someone can cite such a paper and we can agree that it "handles the facts", I'm going to put the sentence back in tomorrow. Mundart (talk) 19:21, 23 December 2015 (UTC)

Mundart and I have been wrangling about the content of this article for about two years now. This is in general a good thing for the content of the article, since it is likely to make it better. The situation is, though, lopsided. My identity, motivations, and agenda are entirely clear -- see my user page. Mundart, in contrast, is secretive. He or she is choosing to remain anonymous throughout. This in itself is not a problem -- until there is significant disagreement. We now have such a disagreement, and my willingness to negotiate and reach a compromise is reduced, precisely due to the fact that Mundart is choosing to remain anonymous. Anonymity allows one to be bolder than one would otherwise be. My suspicion is that Mundart is a run-of-the-mill Chomskyan syntactician, interested in promoting the Chomskyan line. If he or she were to make his/her identity known, we would likely get a sense about his or her motivations and agenda. The lack of transparency reduces my willingness to seek compromise.

Now on to the issue. The disputed sentence is "Neither the nonstructural nor catena-based analysis has been shown to be able to handle the preposition-stranding facts". The statement is very broad, but within the context of the article on sluicing, it is wrong; it misrepresents the catena-based account. An example:

(1) Larry is working with someone, but we don't know who <he is working> with.

This example involves the stranded preposition with. The issue at hand concerns the string he is working. This string is a catena in a straightforward way, but it certainly is not a constituent. Thus the p-stranding facts of English are easily accommodated in terms of the catena unit. Sluicing elides a catena regardless of whether p-stranding does or does not occur, just like most other ellipsis mechanisms elide catenae.

If Mundart reinserts the sentence, I am going to remove it again. My hope is that third-party participants will come in and help us decide the issue. --Tjo3ya (talk) 14:23, 24 December 2015 (UTC)

The English example in (1) above is not the problem. The problem is the P-stranding generalization. To take a concrete example (but see the lit on these things for a huge amount of discussion, e.g. Merchant 2001 et seq.): German does not allow the equivalent of (2a), as illustrated in (2b):

(2)a. Larry is working with someone, but I don't know who.

b. *Larry arbeitet mit jemandem, aber ich weiss nicht, wem.

Instead, German and other non-P-stranding languages require the preposition in such cases:

(3) Larry arbeitet mit jemandem, aber ich weiss nicht, mit wem.

The deletion analysis of sluicing reduces the ill-formedness of (2b) to that of (4) (i.e., they have the same source; they violate the same ban on P-stranding in German, however formulated; see Abels 2002 for some suggestions):

(4) *Larry arbeitet mit jemandem, aber ich weiss nicht, wem Larry mit arbeitet.

A nonstructural analysis can make no connection between 2b and 4, since by hypothesis 2b lacks any structure internal to the ellipsis site.

Until the published literature gives us a nonstructural or catena-based analysis of the ill-formedness of 2b, my sentence is accurate. To avoid potential confusions, though, I'll alter it to specify which P-stranding facts are at issue here, and hope that resolves it. Mundart (talk) 18:08, 27 December 2015 (UTC)

I think there is a devotion here to Merchant's movement-then-deletion approach that is not supported by the state of the science. The movement-plus-deletion approach is assailed in the literature from various angles. See for instance here:

https://depts.washington.edu/hpsg2011/pp/Nykiel-Sag.pdf

But one can overlook the strengths or weaknesses of Merchant's approach in terms of movement-plus-deletion. There is no relevant sense in which Merchant's movement-plus-deletion approach has an advantage over the catena-based approach in this area. The catena-based approach produces the same sort of explanation for the badness of (2b). Both approaches are assuming movement of the wh-element before ellipsis. (2b) is bad on both approaches because German does not allow wem to be fronted without the preposition mit before ellipsis occurs.

But more importantly in my view, the catena-based approach has the clear advantage over Merchant's movement-plus-deletion approach because it can deal with sentence (1) above in a straightforward manner (by eliding a catena), whereas Merchant's movement approach must somehow vacate with from the TP, so that the TP can then be deleted.

I am removing the sentence again.

As far as the "state of the science" goes, I guess we'll just have to disagree: see Luis Vicente's excellent overview article on sluicing in the van Craenenbroeck and Temmerman Handbook on Ellipsis for a recent summary of the state of the science. The link you provided is to an abstract of a 40-minute talk at a conference from 2011: this barely qualifies as "assail[ing] in the literature": we should always assign greater weight to published, peer-reviewed work. As for my original sentence, its truth stands, and is undisputed: nowhere *in the published literature* has the catena-based analysis been shown to handle these facts. This is simply because the published lit on sluicing from a catena standpoint hasn't addressed the relevant facts, period. You and I may agree (or not) that "the catena-based approach produces the same sort of explanation", but that's original research, not in the lit. Once this claim has been vetted by peer-review, it is welcome in a generalist encyclopedia article, I think: we're not here to make arguments and claims, but rather to try to give a representative view of the common opinion in the published scholarship--we're here to report on that scholarship, not add to it. Mundart (talk) 17:20, 5 January 2016 (UTC)

I disagree! A third party needs to decide this. For that third party, please note that Mundart continues to hide by anonymity.--Tjo3ya (talk) 01:03, 6 January 2016 (UTC)

Upon a re-reading of the relevant paragraph, I am comfortable with its current formulation. Please forgive my last reversion. --Tjo3ya (talk) 03:27, 7 January 2016 (UTC)