Island Evasion Explains Variability in Island Repair Matthew Barros, Patrick D. Elliott and Gary Thoms Manchester Forum in Linguistics 1.

16th November 2013

Introduction • The issue: apparent island violations under ellipsis. Sluicing, relative clause island (Ross 1969): (1)

Abby speaks the same Balkan language that one of the new hires speaks, but I don’t know which new hire. a. ... which new hire she speaks the same Balkan language that t speaks b. *Which new hire does she speak the same Balkan language that t speaks?

• So island violations are repaired by ellipsis, somehow... right? But wait! In other situations they’re not repaired, e.g. contrastive fragments (Merchant 2004). (2)

A: Does Abby speak the same Balkan language that B EN speaks? she speaks the same Balkan language that t speaks. (cf. (1)) B: *No, C HARLIE x

• What explains the variation in island repair effects? Some approaches: Repair by ellipsis: Ross (1969), Chomsky (1972), Lasnik (2001), Merchant (2008) – Repair: Island-escaping movement leaves “starred traces” which lead to ungrammaticality unless deleted at PF. (3)

CP ...

Sluicing C’

remnanti C[E]

<...> ... *ti ... [island ... ti ... ] ...

– Crucial to this approach: variation in the syntax of the different kinds of ellipsis environments, i.e. starred traces being inside or outside ellipsis site, i.e. fragments move to higher projection. – Merchant’s (2004) analysis of difference in (1)-(2): starred trace is outside E-site with fragments, crucial difference is in cartography of CP. Barros (Rutgers): [email protected]; Elliott (University College London): [email protected]; Thoms (University of Edinburgh): [email protected]. We thank Klaus Abels, Jeroen van Craenenbroeck, Zoltan Galsi, Anik´o Lipt´ak, Jason Merchant, Yasutada Sudo, the audience at CLS 49, the audiences at the Edinburgh Ellipsis Workshop, and the Edinburgh Meaning and Grammar Reading Group for comments and discussion.

2

Island Evasion Explains Variability in Island Repair (4)

FocP ... remnanti

Fragments FocP’

Foc

CP C’

*ti

<...>

C[E]

... *ti ... [island ... ti ... ] ... • Island evasion: Fukaya (2007), Abels (2011), Barros (to appear) – Evasion: no island in the ellipsis site. One possibility for (1): a “short source,” i.e. nonisomorphism. Mechanisms for this worked out in Merchant (2001) for some islands. (5)

They want to hire someonei who speaks a Balkan language, but I don’t know which one. ... which one hei speaks t / hei should speak t Merchant (2001)

– Crucial to this approach: variation in the availability of evasion strategies. It’s assumed there is little or no repair, and so when evasion is not available island effects resurface. • Also: Mixed approach (some islands repaired, others evaded - see Merchant (2001)), and nonmovement approaches (whP base-generated in surface position, no movement out of island - see, e.g. Chung et al. (1995), Barker (in press), Jacobson (under review)). • In what follows we show that repair and non-repair are attested with both slucing and fragments, not expected under “repair by ellipsis” approach. • We’ll explain this in terms of the use of evasion strategies, concentrating on ways in which this can be achieved by having a smaller non-isomorphic structure in the ellipsis site. When evasion isn’t possible, island effects resurface even with ellipsis, thus indicating that • We thus argue that evasion is pervasive and that island repair is not real. 2. Non-repair in “repair” contexts 2.1 “Left branch extraction” in sluicing • In English, the DP is an island for left-branch modifiers, such as possessors (6a) and attributive APs (6b) (see Corver (1990) for discussion). (6)

a.

*Whose did Jane kiss [t husband]? x

b.

*[Howxtall] did Mary marry a t man?

Island Evasion Explains Variability in Island Repair

3

• Focus on adjectival LBE; Possessors and demonstratives independently licensed nominal ellipsis (7a) - introducing a confounding factor - whereas attributive APs don’t (7b). (7)

a. b.

Jane kissed Mary’s husband, and Susan kissed Helen’s husband *Mary married a tall man, and Jane married a short man

• Merchant (2001) argues based on data like (8) that LBEs are repaired under ellipsis. (8)

Mary married a tall man, but i’m not sure [howxtall] she married a t man

• But data from case concord in German/Dutch suggests that the remnant AP is not merged as an NPmodifier, and hence that the e-site is not, in fact, isomorphic1 . The adjectival wh-remnant must show up in its bare form, patterning with an AP predicate.2 (9)

Der Mann ist groß The man is tall

(10)

Lena hat einen groß*(en) Mann geheiratet tall(. ACC) man married. L. has a

(11)

Lena hat einen großen Mann geheiratet, aber ich weiß nicht wie groß(*.en) tall. ACC man married, but I know not how tall(. ACC). L. has a “Lena married a tall man, but I don’t know how tall”

• Correlate/remnant mismatch: Correlate = großen,

Remnant = groß(*en)

• Baker (2008) claims German/Dutch are exceptions to wider generalisation that predicates show richer agreement than modifiers. He argues case concord in German/Dutch isn’t true agreement. Correlate/remnant mismatches artifact of property unique to German/Dutch? • Hungarian conforms to Baker’s generalisation. Predicates agree in number, modifiers do not. Crucially, the remnant must display number agreement, even when the correlate is a modifier (Elliott (2013)), suggesting that mismatches in the adjectival domain are not an isolated phenomenon. (12)

1 Merchant

a.

John ismer n´eh´any magas(*ak) l´anyt girls J. knows some tall(.PL) “John knows some tall girls.”

- No number agr. on modifier

acknowledges these facts, but resists the conclusion that this is evidence for a predicational analysis, speculating that “...this lack [of adjectival agreement] may open an interesting window into the nature of the inflection itself” p233). Merchant provided two reasons for not adopting a predicative analysis: (i) it would apparently require loosening of the identity condition on ellipsis, and (later on) (ii) we see repair by deletion with LBE in the domain of attributive comparative subdeletion as well (see e.g. Kennedy and Merchant 2000). Here we’ll argue that the loosening of the identity condition is inevitable. For alternative analyses of the comparative subdeletion facts, see Izvorski (1995) and Kennedy (2002). 2 Many languages which seem to show matching effects on APs, such as Greek, independently allow post-adjectival nominal ellipsis, making it difficult to draw any conclusion regarding island evasion vs. repair.

4

Island Evasion Explains Variability in Island Repair b.

(13)

A l´anyok magasak The girls tall.PL “The girls are tall.”

- Number agr. on predicate

John ismer n´eh´any magas l´anyt, de nem tudom milyen *magas/magasak. J. knows some tall girls, but not know.I how tall/tall.PL “John knows some tall girls, but I don’t know how tall.”

• Correlate/remnant mismatch: Correlate = magas,

Remnant = magas*(ak)

• Not just ‘starred trace’ approaches to island repair that encounter difficulties here: Also a major obstacle for semantic accounts of clausal ellipsis, such as Culicover and Jackendoff (2005), Barker (in press) and Jacobson (under review). Both implement ‘brute force’ mechanism for ensuring featural matching between remnant and correlate - Not clear why this should apply to nominal but not adjectival remnants. • Observation here: difference between predicative and non-predicative adjectives in “left-branch sluices” provides further evidence against isomorphic e-site and for predicative source. • Predicative adjectives (e.g. diligent) and non-predicative adjectives (e.g. hard as in hard-working) appear as modifiers (14)a, but only predicative adjectives appear as predicates (14)b. Both kinds can be wh-fronted, so long as they’re gradable (15). (14)

a. b.

Billy hired a diligent/hard worker. This worker is diligent/*hard.

(15)

[How diligent/hard] a worker did Billy hire t? x

• A non-predicative AP remnant is markedly degraded compared to a predicative counterpart (with a modifier correlate)3 . Suggests the AP is merged as a predicate, and that an isomorphic e-site not possible, as it would give rise to unrepaired LBE. (16)

Billy hired a diligent/hard worker, but I don’t know how diligent/*?hard.

• Same can be shown for e.g. heavy (as in drinks heavily): (17)

a. b.

(18)

[How ugly/heavy] a drinker did Mary ignore t? x

(19)

Mary ignored a heavy/ugly drinker, but i’m not sure how ugly/*?heavy.

3 An

Mary ignored a ugly/heavy drinker. The drinker was ugly/*heavy

online questionnaire study of 24 English speakers corroborated our intuitions regarding the status of (16) (Elliott (2012))

Island Evasion Explains Variability in Island Repair

5

• Summary: – Left-branch sluicing remnants pattern morphosyntactically with predicates, not modifiers. – Non-predicative adjectives can’t be remnants at all. • Simplest explanation: sluicing never repair LBE; ‘repaired’ LB sluices have a predicational structure. • LB sluices are a case of island evasion; when the evasion strategy isn’t available, extraction from an island is necessary, and this can’t be repaired, even in sluicing. Importantly, this fits with the picture from previously: non-clausal islands aren’t amenable to repair. 2.2 Predicational sources • Recap: Correlate/remnant mismatches in (10) and (13) and the degraded status of non-predicative AP remnants in (16) and (19) suggest a non-isomorphic source. • Predicational source4 with e-type pronoun subject (Merchant 2001) captures the data we have seen. (20)

Mary married a tall man, but I don’t know [howxtall] hee−type was t

• Independent evidence for predicational source elsewhere: Sluicing in headed unconditionals (Rawlins (2008)) (Discussed briefly by Merchant (2001, p236) in a footnote as ‘concessive’ sluices). (21)

Mary would marry ANYONE, no matter WHO.

• No repair here, so we can see quite clearly than an isomorphic source is ruled out independently: (22)

#Mary would marry ANYONE, no matter WHO Mary would marry.

• (22) not an accurate paraphrase of (21). An explanation for this is given by ?. • A predicational source gets the right interpretation, and also allows a felicitous unelided continuation. (23)

Mary would marry anyone, no matter who x they were t. = If they were a shoe salesmen, then Mary would marry them & If they were a fan of Slade, then Mary would marry them

• The argument extends to unconditional sluices with a left branch remnant5 , but this is harder to see, since an isomorphic continuation is independently out due to illicit LBE. 4 It’s

more common in the lit. to invoke a left-source to account for otherwise mysterious phenomena (see e.g. Vicente (2008) on apparent P-stranding violations in Spanish.) but they won’t work for left branch sluicing since predicates generally are degraded as pivots, e.g. ?/??It is TALL that John is., and they don’t work in unelided wh-questions, i.e. *but I don’t know how tall it was (as noted by Merchant). 5 Note also that unconditional sluices violate the general ban on sprouting AP modifiers, i.e. Mary married a man, but I don’t know how tall, suggesting that this is a semantic, rather than a syntactic restriction.

6

Island Evasion Explains Variability in Island Repair (24)

# Mary would marry any man, no matter how tall Mary would marry the t man. = #If Mary would marry the 6ft tall man, then Mary would marry that man & If Mary would marry the 5ft tall man, then Mary would marry that man, etc.

(25)

Mary would marry any man, no matter how tall he was. = If the man was 6ft tall, then Mary would marry him & If the man was 5ft tall, then Mary would marry him

3. “Repair” in Non-Repair Contexts: Subject Islands 3.1

Subject island repair with fragments? • Recall: Merchant (2004) claims sluicing repairs islands, whereas Contrastive Fragments (CFrs) do not, and gives a cartographic explantion within starred-trace framework (intermediate traces of islandescaping movement marked with PF-uninterpretable *-feature). • Under starred-trace account, CFrs should never give rise to repair illusions, at least if we assume strict syntactic isomorphism. • We refute this putative generalisation on basis of repair illusions with CFrs involving subject islands. The following seem to indicate that subject islands are repaired with fragment answers: (26)

a. Has a biography of H ARPO just been published? b. ?No, G ROUCHO

CFr

(27)

A biography of a Marx brother has just been published, but I don’t know which Marx brother Sluicing

(28)

*Who has a biography of t just been published?

(29)

a. b.

Did a picture of F RANCO arrive last week? No, S ADDAM

(30)

A picture of a dictator arrived last week, but I don’t know which dictator

(31)

*Who did a picture of t arrive last week?

Baseline extraction

CFr Sluicing Baseline extraction

• (26) and (29) are a lot better than starred-trace account predicts. According to Merchant’s analysis, the island-escaping remnant Groucho in (26) leaves a PF-uninterpretable *-marked trace in specCP crucially, external to the e-site - as in (32): (32)

[ FocP G ROUCHO [ Foc’ Foc [ .... [ CP t* [ C’ C[E] < [ IP [ island a biography of t ] has just been published] > ]]]]]

• What’s going on here? Our answer: Island evasion.

Island Evasion Explains Variability in Island Repair

7

• Relevant observation: Subject-island sensitivity re-emerges under contrastive sluicing: (33)

?*I know a biography of H ARPO has just been published, but I don’t know which Marx brother

(34)

?*I know a picture of F RANCO has just arrived, but I don’t know which OTHER dictator6

OTHER

• CFrs involve corrective focus, whereas contrastive sluicing involves additive focus. Can we blame this for the difference in acceptability between (26)-(29) and (33)-(34)? • We can test to see whether the additive/corrective distinction is responsible by considering what we dub additive fragments (AFrs). It turns out that the acceptability judgements are not so clear-cut. Subject island repair is attested with AFrs, but only relative to a certain reading, which does not seem to correspond to a syntactically isomorphic ellipsis site: (35)

a. b.

A biography of Harpo has just been published. Yeah, G ROUCHO too. # a (possibly distinct) biography of Groucho has just been published too Xthe biography in question was of Groucho too

(36)

a. b.

A picture of Franco has just arrived. Yeah, S ADDAM too. # a possibly distinct picture of Saddam has just arrived too Xthe picture in question was of Saddam too

• Readings that emerge as acceptable correspond to an e-site containing an e-type pronoun7 , as in (37)b and (38)b. The readings corresponding to a fully isomorphic source, as in (37)c and (38)c, are unavailable. (37)

a. b. c.

A biography of Harpo has just been published. Yeah, [the biography of Harpo that has just been published]/it was of G ROUCHO too. Yeah, a biography of G ROUCHO has just been published too.

(38)

a. b. c.

A picture of Saddam has just arrived. Yeah, [the picture of Saddam that has just arrived]/it is of F RANCO too. Yeah, a picture of F RANCO has just arrived too.

• Readings corresponding to fully isomorphic sources in (37)-(38)c unavailable because they involve (unrepaired!) extraction from a subject island. 6 This

is marginally possible on a “which other dictator arrived” interpretation, with no island violation to speak of. assume a d-type analysis of e-type pronouns in which an e-type pronoun is underlyingly a definite description (see, e.g. Elbourne (2005), a.o., but c.f. Moltmann (2006) a.o.). We shall occasionally spell-out the content of the definite description in question. Although this occasionally sounds cumbersome, it makes the relationship between the antecedent and the ellipsis site wrt. the identity condition more salient. 7 We

8

Island Evasion Explains Variability in Island Repair

3.2 Subject Island Evasion • As in LBE cases, we propose an evasion source involving an e-type pronoun + copular underlies repair illusions. AFrs were useful, since they rendered isomorphic vs. evasion sources truth-conditionally distinct. • Prediction: where e-type pronoun is de-licensed in e-site, subject island-sensitivity should re-emerge. • E-type pronouns introduce an existential presupposition, just like definite descriptions (Elbourne (2005)); if this presupposition isn’t satisfied, e-type pronoun is de-licensed. Consider the following: (39)

a. Has any book about B OWIE been published this week? b. #No, it was a book about S YD BARRETT c. No, (but) a book about S YD BARRETT has been published this week

• Since the correlate Bowie is embedded in a subject island, we predict an evasion source underlies fragment, BUT evasion source independently predicted to be infelicitious, due to the failure of the existential presupposition in (39)b. Prediction is borne out: (40)

a. Has any book about B OWIE been published this week? b. #No, S YD BARRETT it was about t

• Furthermore, note that a quantifier headed by no- fails to be resumed by an e-type pronoun (Evans (1980), Romero (1998)). (41)

a. #No friend of Mary gets along with John, since they drive him crazy. b. A few friends of Mary get along with John, since they don’t drive him crazy.

• We predict that where the correlate is embedded in a subject island, and an e-type pronoun is delicensed due to the subject being headed by a no-quantifier, the fragment should fail (island violation): (42)

a. No friend of M ARY gets along with John b. #and Sally, as well c. #...< they are a friend of t > as well (e-type is de-licensed) d. *...< no friend of t > as well (subject island violation)

• Note that the entire quantificational subject is an acceptable fragment, which is predicted under our analysis since an e-type evasion strategy is unnecessary, and there is no subject-island violation. (43)

a. No friend of M ARY gets along with John b. ?and no friend of S ALLY, as well

• We assume that an indefinite licenses an e-type pronoun in the ellipsis site (Merchant (2001) and Romero (1998) for mechanisms underlying this). We predict both e-type and indefinite readings should be available in absence of islands. This seems to be what we find.

Island Evasion Explains Variability in Island Repair

9

• Context: Two car salesmen discussing how many cars have been sold that day. Salesman 1 utters (44)a, and salesman 2 responds with (44)b (example due to Yasutada Sudo): (44)

a. b.

Somebody bought a blue car. Yeah, a GREEN car too.

• Salesman 2’s response can mean that someone (possibly a different person to the one who bought the blue car) bought a green car. This corresponds to an ellipsis site with an indefinite: (45)

a GREEN car < someone bought t > too

• Consider a scenario in which the two salesmen are discussing how many cars specific customers bought. Salesman 1 utters (46)a, and salesman 2 responds with (46)b (repeated from (44)): (46)

a. b.

Somebody bought a blue car. Yeah, a GREEN car too.

• Salesman 2’s response can mean that the particular customer who bought the blue car also bought a green car. This corresponds to an ellipsis site with an e-type pronoun: (47)

a GREEN car < hee−type bought t > too

• A final point to make: our subject island evasion sources have the satisfying property of being compatible with case-connectivity between the correlate and the remnant. 4.

More on short sources • Preceding discussion shows non-isomorphic sources for clausal ellipsis must be possible. This is not new: Merchant (2001, ch.5) argues this happens in other cases of sluicing. • For (1) he proposes that the correct underlying structure is a “short source” like (48), with the identity condition being satisfied by modal subordination (for subjunctive relatives) and an E-type pronoun subject. Another case of evasion. (48)

They want to hire someonei who speaks a Balkan language, but I don’t know [which x one] hei should speak t

• As argued by Merchant (2001) we need a short source here, rather than a predicational source/cleft, since examples like (48) show connectivity effects - see below. Different contexts call for different evasion strategies. (49)

Sie wollen jemanden anstellen, der [einen britischen Dialekt] spricht... They want someone hire, who a. ACC british. ACC dialect speaks...

10 Island Evasion Explains Variability in Island Repair a.

...aber ich weiß nicht [welchen britischen dialect] er sprechen soll t. ...but i know not which. ACC British. ACC dialect he speak should t.

b. *...aber ich weiß nicht [welcher britische Dialekt] es ist t. ...but i know not which. NOM British. NOM dialect it is t. • So evasion may account for many cases of apparent island violation in sluicing. Much subsequent work dismisses this and maintains, presumably due to reactions in Lasnik (2001), but see Barros (to appear) for responses to some points made by Lasnik. • Variable availability of short sources account for variation in contrastive fragments with relative clause islands. Short source possible for (50). (50)

A: Did they hire someonei who works on F RENCH (last year)? hei works on t. B: No, G ERMAN x

• What about RC extractions that fail? Don’t seem to have plausible short source parses (Abels 2011). (51)

A: Does Abby speak the same Balkan language that B EN speaks? B: *No, C HARLIE ... = ??? cf. #No, C HARLIE speaks the same Balkan language that Abby speaks. #No, C HARLIE speaks it.

• Full answer here would not answer the Question Under Discussion (cf. Roberts 1996, ?) so it would not work as a source for the fragment either. • Similarly consider definite DP islands: (52)

A: Did you give Mary’s picture of P RINCE to her?

B: ?*No, E LVIS.

• No obvious short source. Indefinite variant I gave a picture of Prince to her would be incongruent. • Conclusion: RCs often don’t show repair effects, and this correlates with evasion via short sources. 5.

Discussion and conclusion • We’ve seen repair with fragments and non-repair with sluicing, a set of facts which is best explained by island evasion with non-repair. We’ve seen two evasions strategies in action: (i) predicational sources, and (ii) short sources. • We propose that there is at least one other evasion strategy that is responsible for evasion: pseudosluicing, or hidden clefts. (53)

Abby speaks the same Balkan language that one of the new hires speaks, but I don’t know which new hire it is t

Island Evasion Explains Variability in Island Repair

11

• Merchant 2001 shows that all sluicing does not reduce to pseudosluicing, but this doesn’t rule out pseudosluicing (van Craenenbroeck (2010)); moreover Postdam (2007) shows it’s necessary to allow it. Other evidence from tags, P-stranding indicate clefts are employed more often than Merchant envisaged. Note clefts seem to be incompatible with contrastive sluicing/fragments: (54)

a. *John spoke to JANE, but I don’t know who else it was. b. A: John spoke to Jane. B: *Yeah, it was A NNA, too.

• We think that these evasion strategies carve up the whole repair landscape pretty well. Some of our other findings (and some from Merchant (2001)): Environment English adjunct island sluice English RC island sluice English CSC sluice English adjunct island contrast sluice Case-marked clausal isl. sluice (eg German) English corrective fragment, adjunct island English additive fragment, adjunct island Dutch contrastive fragment, definite DPs Dutch fragments, APs Slavic multiple sluicing Turkish non-contrastive fragment, arguments Turkish non-contrastive fragment, APs Turkish contrastive fragment, APs Japanese non-contrastive fragment, arguments Japanese non-contrastive fragment, APs

Short source X X ?? * X * * * * * X * * X *

Predicational source * * * * * X * * * * * X * X *

Cleft X X X * * * * * * * X * * * *

Judgments X X X * X X * * * * X X * X *

A tick in any of the first three columns guarantees well-formedness, by evasion; when evasion is not available, judgments are *. Right result in all these cases! • Conclusion: What we have been referring to as ‘repair’ isn’t real - it’s an illusion based on a faulty premise (isomorphism). Adopting evasion simplifies our theory, and captures the pervasive variability of repair phenomena. References Abels, Klaus. 2011. Don’t fix that island! It ain’t broke, Paper presented at the Islands in Contemporary Linguistic Theory conference, Vitoria-Gasteiz. Baker, Mark C. 2008. The Syntax of Agreement and Concord. Cambridge University Press. Barker, Chris. in press. Scopability and sluicing. Barros, Matthew. to appear. Short sources and pseudosluicing: a non-repair approach to island sensitivity in contrastive TP ellipsis. In Proceedings of CLS 48. Chomsky, Noam. 1972. Some empirical issuees in the theory of transformational grammar. In Paul Stanley Peters, ed., Goals of linguistic theory, Englewood Cliffs: Prentice-Hall. Chung, Sandra, Ladusaw, William, and McCloskey, James. 1995. Sluicing and logical form. Natural Language Semantics 3:239– 282. Corver, Norbert. 1990. The syntax of left branch extractions. Ph.D. thesis, Katholieke Universiteit Brabant.

12 Island Evasion Explains Variability in Island Repair Culicover, Peter and Jackendoff, Ray. 2005. Simpler Syntax. Oxford: Oxford University Press. Elbourne, Paul D. 2005. Situations and individuals, volume 90. MIT Press Cambridge. Elliott, Patrick D. 2012. Illusory island repair in sluicing: Evidence from non-predicative adjectives, paper presented at the Syntax and Semantics Research Group, University of Edinburgh. Elliott, Patrick D. 2013. Towards a theory of radical island evasion under ellipsis, paper presented at LEL Postgraduate conference, University of Edinburgh. Evans, Gareth. 1980. Pronouns. Linguistic inquiry 11:337–362. Fukaya, Teruhiko. 2007. Sluicing and stripping in Japanese and some implications. Ph.D. thesis, University of Southern California. Izvorski, Roumyana. 1995. A solution to the subdeletion paradox. In Jos´e Camacho, Lina Choueiri, and Maki Watanabe, eds., West Coast Conference on Formal Linguistics (WCCFL) 14, 203–219, Stanford, CA: CSLI Publications. Jacobson, Pauline. under review. The short answer: Implications for direct compositionality (and vice-versa), ms. Kennedy, Chris. 2002. Comparative deletion and optimality in syntax. Natural Language and Linguistic Theory 20:553–621. Kennedy, Chris and Merchant, Jason. 2000. Attributive comparative deletion. Natural Language and Linguistic Theory 18:89– 146. Lasnik, Howard. 2001. When can you save a structure by destroying it? In M. Kim and U. Strauss, eds., Proceedings of the North Eastern Linguistic Society 31, volume 2, 301–320, GLSA. Merchant, Jason. 2001. The Syntax of Silence: Sluicing, Islands, and the Theory of Ellipsis. Oxford: Oxford University Press. Merchant, Jason. 2004. Fragments and ellipsis. Linguistics and Philosophy 27:661–738. Merchant, Jason. 2008. Variable island repair under ellipsis. In Kyle Johnson, ed., Topics in Ellipsis, Cambridge: Cambridge University Press. Moltmann, Friederike. 2006. Unbound anaphoric pronouns: E-type, dynamic, and structured-propositions approaches. Synthese 153:199–260. Rawlins, Kyle. 2008. (Un) conditionals: An Investigation in the Syntax and Semantics of Conditional Structures. ProQuest. Roberts, Craige. 1996. Information structure: Towards an integrated theory of formal pragmatics. In Jae-Hak Yoon and Andreas Kathol, eds., OSU Working Papers in Linguistics, volume 49, OSU: The Ohio State University Department of Linguistics. Romero, Maribel. 1998. Focus and reconstruction effects in wh-phrases. Ph.D. thesis, University of Massachusetts Amherst. Ross, John. 1969. Guess who? In Robert Binnick, Alice Davison, Georgia Green, and Jerry Morgan, eds., Chicago Linguistic Society (CLS) 5, 252–286, University of Chicago, Chicago, Illinois: Chicago Linguistic Society. van Craenenbroeck, Jeroen. 2010. Invisible last resort: A note on clefts as the underlying source for sluicing. Lingua 120:1714– 1726. Vicente, Luis. 2008. Syntactic isomorphism and non-isomorphism under ellipsis, under review for publication in Lingua.