A remnant-correlate identity condition on TP

A remnant-correlate identity condition on TP deletion
Matthew Barros
Rutgers University
[email protected]
Luis Vicente
Universität Potsdam
[email protected]
&
April 16, 2015
1
State of the art and proposal
(2) Voice mismatches: active-passive
Someone assassinated JFK. . .
Identity conditions on ellipsis are usually formulated in semantic terms, i.e., LF equivalence (Sag 1976, Chung et al. 1995), mutual truth-conditional entailment modulo
F-closure (Merchant 2001), inquisitive entailment (AnderBois 2011, 2014), or QuD
equivalence (Ginzburg and Sag 2000, Krifka 2006, Reich 2007, Barros 2014, Weir
2014). The question is whether a purely semantic identity condition is enough. Chung
(2006, 2013) and Merchant (2013) have notably answered this question in the negative,
proposing to supplement semantic identity with a morphosyntactic condition requiring
identity of functional structure between the antecedent and the E-site.
a. * . . . but it is still a mystery who by [
b.
(3)
].
. . . but it is still a mystery who he was assassinated by.
Voice mismatches: passive-active
JFK was assassinated. . .
a. * . . . but it is still a mystery who [
b.
(1) Argument structure condition: if the remnant of sluicing is the argument of a
predicate in the ellipsis site, that predicate must have an argument structure
identical to that of the corresponding predicate in the antecedent clause.
[Chung 2013:30]
(4)
].
. . . but it is still a mystery who assassinated him.
Argument structure mismatches: psych verbs
Something worries Jack. . .
a. * . . . but he won’t say what about [
b.
This condition is meant to account for the ungrammaticality of voice and argument
structure alternations like the following, which demonstrably do not follow from semantic identity alone. We provide unsluiced grammatical examples as controls to show
that we are dealing with an ellipsis effect here.
(5)
].
. . . but he won’t say what he is worried about.
Argument structure mismatches: spray-load alternations
Jack embroidered something on the flag. . .
a. * . . . but I can’t remember what with [
b.
1
].
. . . but I can’t remember what he embroidered the flag with.
Matthew Barros & Luis Vicente
2
(6) Argument structure mismatches: spray-load alternations (II)
Limitations of the Argument Structure Condition
Jack embroidered something with peace signs. . .
2.1
a. * . . . but I can’t remember what on [
Pseudosluicing (sluices that stem from a cleft even though their antecedent is a noncleft) is a pervasive phenomenon both across and within languages (cf. Merchant 1998,
2001, van Craenenbroeck 2004, Rodrigues et al. 2009, Barros 2014). Within English
alone, pseudosluicing can be used to circumvent selectional restrictions. . .
b.
].
. . . but I can’t remember what he embroidered peace signs on.
(7) Argument structure mismatches: dative alternation
(10)
Jack sent someone a letter. . .
a. * . . . but he won’t say who to [
b.
].
. . . but he won’t say who he sent a letter to.
(11)
Either the kitchen is on fire or Jack is baking again.
It is hard to tell which [it is].
which it is.
‡ which the kitchen is on fire or Jack is baking again.
The window { broke / is still broken }. . .
b.
Sally has a new boyfriend, but she won’t tell us who [it is].
who it is.
# she has.
. . . to resolve ellipsis with disjoined clausal antecedents (AnderBois 2011, 2014). . .
(8) Argument structure mismatches: unaccusative/transitive alternations
a. * . . . but we haven’t found out who [
It is too strong
].
. . . but we haven’t found out who broke it.
. . . and to circumvent what would otherwise be movement violations (Barros et al. 2014
and references).
Here, we propose to replace (1) with a condition that regulates the relation between the
remnant of ellipsis and its correlate in the antecedent clause. The reason why we want
to do this is because, as we will see, a condition along the lines of (1) is both too strong
and too weak: it incorrectly predicts certain grammatical ellipses to be ungrammatical,
and vice versa. We will show that our condition circumvents these difficulties.
(9) The Remnant Condition
For any variable assignment g, the (contextually restricted) alternative set of
the remnant must match the (contextually restricted) alternative set of the
correlate.
(12)
She met a rich man, but I don’t know how rich [that man is].
how rich that man is.
‡ how rich she met a man.
Moreover, in some languages (especially Austronesian), pseudosluicing is the only
possible sluicing strategy.
(13) Malagasy (Potsdam 2007)
Importantly, the Remnant Condition is meant to be a supplement to (rather than a
replacement of ) semantic conditions regulating identity between the E-site and the
antecedent. A successful ellipsis requires satisfying both the Remnant Condition and
E-site identity.
Nandoko zavatra
i Bao, fa hadinoko hoe inona [no nolokoin’ i Bao].
paint.at something Bao but forget
C what prt paint.tt Bao
“Bao painted something, but I forgot what it is that was painted by Bao”
And yet, the Argument Structure Condition predicts it should be ungrammatical. Whatever argument/functional structure we want to assign to a cleft, it is clearly different
from that of a non-cleft.
2 of 5
A remnant-correlate identity condition on ellipsis
2.2
It is too weak
2.2.2
There are some classes of TP deletion that are ungrammatical despite satisfying both
E-site semantic identity and the Argument Structure Condition.
2.2.1
Reverse pseudosluicing
Even though regular pseudosluicing is grammatical and pervasive, reverse pseudosluicing (an elliptical non-cleft with a cleft antecedent) is surprising ungrammatical (Vicente
2008).
Symmetric predicates
(17)
These are predicates whose two arguments are equal participants in the event in question
(Dimitriadis 2008). We can identify them through the following entailment test.
The person that Sally was talking to is someone from Accounting, but I don’t
know. . .
a.
. . . who (exactly) [it was].
b. * . . . to whom (exactly) [she was talking].
(14)
Jack is making out with Sally
Jack is making out.
Sally is making out
[other such predicates: resemble, have a conversation with, be related to. . . ]
c.
Our interest here lies on symmetric predicates whose second argument is a PP. Barros
(2014) notes that sluicing of such predicates is licit so long as the sluice maintains the
argument order of the antecedent (15a), but not if the argument order is reversed (15b).
Example (15c) is a control to show that unsluiced questions with different argument
orders are acceptable, so (15b) must be taken as an ellipsis effect.
(15)
Someone was making out with Jack. . .
a.
. . . but I don’t know who [was making out with Jack].
b. * . . . but I don’t know with who [Jack was making out].
c.
A:
Someone was making out with Jack.
B:
Yeah, Sally [was making out with Jack]!
(18)
B’: * Yeah, with Sally [Jack was making out]!
B”:
Details of the Remnant Condition
First off, we know that we can’t define the Remnant Condition in morphosyntactic
terms. The reason is that remnants and correlates support the kind of mismatches that
contraindicate a formulation along these lines.
The paradigm in (16) shows that the same effect holds for fragment answers.
(16)
Reverse pseudosluices do satisfy E-site semantic identity (short argument: regular pseudosluices do; given that E-site semantic conditions are symmetric, it doesn’t matter
whether the cleft is the sluice or the antecedent). The Argument Structure Condition
does predict their ungrammaticality, because clefts and non-clefts have different functional structures. However, it cannot distinguish regular from reverse pseudosluices:
either both classes of ellipses should be ungrammatical (what existing versions of the
Argument Structure Condition predict), or both should be grammatical (under any
hypothetical modification of the Argument Structure Condition that rules regular
pseudosluices in).
3
. . . but I don’t know with who Jack was making out.
. . . to whom exactly she was talking.
A: Did Jack eat natto [ ADV reluctantly]?
B: No, [PP with relish].
Yeah, Jack was making out with Sally!
Because of the entailment conditions of symmetric predicates, the ungrammaticality of
(15b) and (16B’) doesn’t follow from existing semantic identity conditions. Additionally,
because both arguments of a symmetric predicate bear the same participant role wrt
the event, these alternations also satisfy the Argument Structure Condition.
[Weir 2014]
(19) Either [T P the kitchen is on fire or Jack is baking again]. It’s hard to tell [ DP
which].
Similarly, we can have the Remnant Condition compare ordinary semantic values
across the board. The reason is that wh- phrases, which are both licit remnants and licit
correlates, lack a defined ordinary semantic value (Beck 2006, Cable 2010).
3 of 5
Matthew Barros & Luis Vicente
(20) I don’t want to know [which students] failed the exam, I just want to know [how
many].
Focus semantic values won’t do either, because it non-F-marked indefinites and disjunctions (which have a undefined focus semantic value) are licit correlates.
(21)
A: Someone here has [F the key to the liquor cabinet].
4
4.1
Applications of the Remnant Condition
Argument Structure Condition effects
Voice and argument structure mismatches can still be ruled out through the Remnant Condition, given that the remnant and the correlate give rise to non-equivalent
alternative sets.
B: Who?
[Weir 2014]
B’: Yeah, Mary.
(22)
* Someone assassinated JFK, but I don’t know who by [JFK was assassinated].
[alt-set(someone) x alt-set(by who)]
(26)
* Something worries Jack, but he won’t tell us what about [he’s worried].
[alt-set(something) x alt-set(about what)]
A: Either Ryan or Dexter will [F play center field].
B: I wonder which one.
B’: Probably Ryan.
[adapted from AnderBois 2014]
We can get around these limitations with a disjunctive definition of the RC, but this
misses the generalization that we only care about the ordinary semantic value of the
correlate if it is an inquisitive expression. We propose that we can capture this much by
appealing to alternative sets rather than ordinary/focus semantic values.
(23)
(25)
Either J corr K f
J remn K f , or J corr Ko
Compare to the grammatical, non-mismatched cases.
J remn K f
We propose to calculate alternative sets with the following algorithm. Admittedly, it is a
bit of a hack. It works for our current purposes, but clearly we need a better formulation.
(24) if the remnant/correlate has a defined focus semantic value
then its alternative set is equal to its focus semantic value (a set of contextually restricted, type-theoretically identical expressions, see Rooth 1992)
(27)
Someone assassinated JFK, but I don’t know who [assassinated him].
[alt-set(someone) alt-set(who)]
(28)
Something worries Jack, but he won’t tell us what [worries him].
[alt-set(something) alt-set(what)]
We need to note two things here. First, our account predicts that the following parse
(with a voice mismatch) should be grammatical. We haven’t figured out how to test this
prediction in English, but Potsdam’s (2007) analysis of Malagasy sluicing suggests that
such mismatches are actually allowed.
else, if the remnant/correlate is not F-marked but it is an inquisitive expression,
(29)
then pretend it is F-marked and has a defined focus semantic value.
JFK was assassinated by someone, but I don’t know who [assassinated him].
[alt-set(someone) alt-set(who)]
else, if the remnant/correlate is not F-marked and it is not inquisitive,
then it is not a licit remnant/correlate and ellipsis will fail.
Second, we need to assume that prepositions are invariably semantically contentful.
This is uncontroversial for directional, locative, and instrumental ones, but not so for
purely functional ones, especially of and by. We will assume that, even if they lack an
obvious lexical meaning, they still have a non-trivial semantic contribution (cf. Adger
2013), so that J DP K x J P DP K.
4 of 5
A remnant-correlate identity condition on ellipsis
4.2
Effects beyond the Argument Structure Condition
The Remnant Condition correctly captures the contrast between regular and reverse
pseudosluicing. As above, it is the presence of an unmatched preposition in the remnant
that makes reverse pseudosluicing fail.
(30)
Sally has a new boyfriend, but she won’t tell us who [it is].
[alt-set(a new boyfriend) alt-set(who)]
(31)
* The guy that Sally was talking to was someone from Accounting, but I don’t
know who to (exactly) [she was talking].
[alt-set(someone from Accounting) x alt-set(to who (exactly))]
References
Adger, David. 2013. A syntax of substance. Cambridge: MIT Press.
AnderBois, Scott. 2011. Issues and alternatives. Doctoral Dissertation, University of California,
Santa Cruz.
AnderBois, Scott. 2014. The semantics of sluicing: beyond truth conditions. Language 90:887–926.
Barros, Matthew. 2014. Pseudosluicing and identity in ellipsis. Doctoral Dissertation, Rutgers
University.
Barros, Matthew, Patrick Elliott, and Gary Thoms. 2014. There is no island repair. Ms., Rutgers
University, University College London, and University of Edinburgh.
Beck, Sigrid. 2006. Intervention effects follow from focus interpretation. Natural Language
Semantics 14:1–56.
Cable, Seth. 2010. The grammar of Q: Q-particles, wh- movement, and pied-piping. Oxford: Oxford
University Press.
Chung, Sandra. 2006. Sluicing and the lexicon: the point of no return. In Proceedings of BLS 31,
ed. Cover and Kim, 73–91.
Chung, Sandra. 2013. Syntactic identity in sluicing: how much and why? Linguistic Inquiry
44:1–44.
Chung, Sandra, William Ladusaw, and James McCloskey. 1995. Sluicing and Logical Form. Natural
Language Semantics 3:239–282.
van Craenenbroeck, Jeroen. 2004. Ellipsis in Dutch dialects. Doctoral Dissertation, Leiden
University.
Dimitriadis, Alexis. 2008. Irreducible symmetry in reciprocal constructions. In Reciprocals and
reflexives: theoretical and typological explorations, ed. Ekkehard König and Volker Gast. Berlin:
Mouton de Gruyter.
Ginzburg, Jonathan, and Ivan Sag. 2000. Interrogative Investigations. Stanford, CA: CSLI.
Krifka, Manfred. 2006. Association with focus phrases. In The architecture of focus, ed. Valeria
Molnar and Susanne Winkler, 105–136. Berlin: Mouton de Gruyter.
Merchant, Jason. 1998. Pseudosluicing: elliptical clefts in Japanese and English. In ZAS Working
Papers in Linguistics 10, ed. Artemis Alexiadou. Berlin: Zentrum für Allgemeine Sprachwissenschaft.
Merchant, Jason. 2001. The syntax of silence: sluicing, islands, and the theory of ellipsis. Oxford:
Oxford University Press.
Merchant, Jason. 2013. Voice and ellipsis. Linguistic Inquiry 44:77–108.
Potsdam, Eric. 2007. Malagasy sluicing and its consequences for the identity requirement on
ellipsis. Natural Language and Linguistic Theory 25:577–613.
Reich, Ingo. 2007. Toward a uniform analysis of short answers and gapping. In On information
structure, meaning, and form, ed. Kerstin Schwabe and Susanne Winkler, 467–484. Amsterdam:
John Benjamins.
Rodrigues, Cilene, Andrew Nevins, and Luis Vicente. 2009. Cleaving the interactions between
sluicing and P-stranding. In Romance languages and linguistic theory 2006, ed. Daniéle Torck
and Leo Wetzels, 245–270. Amsterdam: John Benjamins.
Sag, Ivan. 1976. Deletion and Logical Form. Doctoral Dissertation, MIT.
Vicente, Luis. 2008. Syntactic isomorphism and non-isomorphism under ellipsis. Ms., University
of California, Santa Cruz.
Weir, Andrew. 2014. Fragments and clausal ellipsis. Doctoral Dissertation, University of Massachusetts, Amherst.
5 of 5