Ha, Seungwan. 2009. "Multiple dominance CAN'T, but ellipsis CAN account for Right Node Raising." Proceedings from the 42nd Annual Meeting of the Chicago Linguistic Society - Main Session. Chicago: Chicago Linguistic Society, pp. 17-31.

Multiple Dominance CAN’T, but Ellipsis CAN account for Right Node Raising∗ SEUNGWAN HA Boston University

0. Introduction Right Node Raising (RNR) refers to a type of ellipsis where the gap occurs not in the second conjunct but in the first, as we see in the title of this paper. The gap of the first conjunct corresponds to account for Right Node Raising, so the fully reconstructed sentence should be Multiple Dominance CAN’T account for Right Node Raising, but Ellipsis CAN account for Right Node Raising. This paper concerns two issues connected to RNR. First, we discuss current analyses of RNR and argue that RNR is an ellipsis phenomenon. After providing evidence that RNR is a type of ellipsis, we will discuss licensing conditions for RNR. 1. Representation in syntax 1.1. Rightward movement analysis There have been largely two types of analyses of RNR, differing with respect to the role played by syntactic movement. As for movement analyses, Ross (1967), Bresnan (1974), Postal (1974, 1998), Sabbagh (2003) claim that the RNR target constituent in each conjunct undergoes rightward ATB movement, shown in (1). The DP every student moves and adjoins out of the conjunct.

∗

I would like to thank my advisor, Paul Hagstrom, for his helpful discussion, judgments, and comments. Also, many thanks to Michela Ippolito, Kyle Johnson, Jason Merchant, Shigeru Miyagawa, and Mark de Vries for comments. All errors are, of course, mine.

Seungwan Ha (1) Kim bought an apple for ____, and Jane gave it to every student. &P qp &P DP qp 4 CP1 & CP2 every student 6 6 Kim bought Jane gave it to [t] an apple for [t]

However, a number of problems for the movement analysis have been pointed out in the literature. Consider (2). The supposed landing site for RNR movement is dominated by the temporal adjunct in the second conjunct, contrary to what ATB movement would have predicted. Rightward ATB movement should yield (2b), where the DP moves out of the whole conjunct. Additionally, if there is a constraint against Vacuous Movement, an option like (2a) should also be banned. (2) Mary ignored t (last year), and John talked about t [the man you met in Paris] yesterday. a. *[… t (last year) TP], and [… t TP]... [the man you met in Paris] yesterday TP]. b. [… t (last year) TP], and [… t

yesterday TP] [the man you met in Paris]&P].

A movement analysis of RNR also yields preposition stranding even in languages that otherwise disallow it. According to McCloskey (1986), Irish is not a preposition stranding language, so extraposition cannot strand a preposition, as in (3a). However, RNR allows preposition stranding, as shown in (3b), which is puzzling for the movement analysis. (3) a. *Bhí mé a éisteacht le ti inné Was I listen(prog) with yesterday [DP clár mór fada ar an ráidió faoin toghachán]i program great long on the radio about-the election ‘I was listening yesterday to a great long program on the radio about the election.’ b. Nil sé in aghaidh an dlí a thuilleadh a bheith ag éisteacht le Is-not it against the law anymore be(-fin) listen(prog) with nó ag breathnu ar [DP ráidió agus teilifís an Iarthair] or look(prog) on radio and television the West(gen) ‘It is no longer against the law to listen, or to watch, Western radio and television.’ (Irish, McCloskey 1986: 184-185)

18

MD CAN’T but Ellipsis CAN account for RNR Examples (4-6) show RNR is insensitive to various island constraints. (4) is a whisland, (5) is a Complex NP island, and (6) is an adjunct island. If RNR arises from a movement operation, it is not ordinary movement. (4) Susan wonders [when John ordered t], and Bill wants to know [when he returned t] the tickets for the opera. (Wh-island) (5) John likes [a professor who lectured on t], and Mary likes [a graduate student who debunked t] a recent theory of Right Node Raising. (Complex NP island) (6) Josh was happy [after he heard of t], but Willy got angry [after finding out about t] the news that the food chain will no longer carry live lobsters. (Adjunct island) 1.2. Non-movement analysis 1.2.1. Multiple Dominance In response to the problems with movement analyses of RNR, non-movement analyses have been proposed, such as Multiple Dominance and phonological deletion. Let us discuss Multiple Dominance hypothesis, as proposed by McCawley (1982), and Wilder (1999), among others. Under the MD account, the target constituent is simultaneously dominated by each conjunct. Consider (7), for example, where the node inside the object DP an article are asymmetrically c-commanded both by the verb wrote in the first and reviewed in the second conjunct. (7) Chris wrote, and Kate reviewed an article.

&P qp TP1 &’ 3 3 DP VP & TP2 Chris t 3 V DP VP wrote Kate ry V DP reviewed 5 an article

One of the main concerns of the MD hypothesis is how to linearize a RNR structure. Consider the contrast between (8a) and (8b). (8) a. Bill recorded [t], and Jason watched in December [every Red Sox game]. b. *Bill sent [t] a present, and Mary congratulated [all the winners].

19

Seungwan Ha In both (8a) and (8b), the ellipsis site has its antecedent in the second conjunct, but only (8a) is possible. Ross explains this contrast by proposing the Right Roof Constraint in (9). (9) Right Roof Constraint (McCloskey’s version 1999) Rightward movement may move an element X to the right edge of the cyclic-node that most immediately contains X, but no further. The Right Roof Constraint is a constraint on movement, so the MD hypothesis needs an alternative explanation for the RRC effect that does not depend on movement. Wilder (1999) proposes deriving the word-order using Kayne’s Linear Correspondence Axiom (LCA). Kayne’s LCA states that A precedes B when A asymmetrically c-commands B. Let us consider how the word order would be derived. In the first conjunct in (7), Chris asymmetrically c-commands wrote, therefore Chris precedes wrote in linear order. And wrote asymmetrically c-command the object DP, so wrote precedes an article. The second conjunct is linearized in the same way. The first conjunct also asymmetrically ccommands the second conjunct, so every element in the first conjunct precedes every element in the second. The result of linearization for (7) is (10). (10) Asymmetric c-command & linearization: Reflexivity violation (*x >> x) Chris >> wrote >> an >> article >> & >> Kate >> reviewed >> an >> article However, this results in an article preceding itself, thus unlinearizable due to reflexivity violation (Kayne 1994). Wilder proposes to interpret the LCA as applying only to the terminals fully dominated. For example, a shared constituent z in A cannot be linearized in (11). (11) Modified LCA (Wilder 1999:594) &P A

&

B

x, y, z α, β, z

The target object DP in the first conjunct is not in the image of linearization because it is not fully dominated, so only Chris and wrote can be linearized in TP1. An article is fully dominated in the second conjunct, so all the elements of the second conjunct are linearized as Kate, reviewed, an, article. TP1 asymmetrically c-commands TP2, so every element inside TP1 precedes every element in TP2. The overall linearization is (12).

20

MD CAN’T but Ellipsis CAN account for RNR (12) Chris >> wrote >> & >> Kate >> reviewed >> an >> article. Wilder claims that his modified LCA can explain why SVO languages do not allow backwards gapping, but SOV languages do. In (13a), a red wine precedes drank because TP1 asymmetrically c-commands TP2, but drank also asymmetrically c-commands a red wine. Then, we have a conflict in linear order. On the other hand, (13b), a Korean backwards gapping example, presents no problems for linearization. Basically the linearization is the same with SV & SVO order in English. (13) a. *Jane [e] a red wine, and Mary drank an orange juice. (*a red wine >> drank & drank >> a red wine)

&P qp TP1 &’ 3 3 DP VP & TP2 Jane t 3 DP DP VP 5 Mary ry a red wine V DP drank 6 an orange juice

b. Jane-un wine-ul [e], kuliko Mary-nun orange juice-lul masi-ess-ta. (Korean) J.-Top wine-Acc [e], Conj M.-Top orange juice-Acc drink-Past-Dec ‘Jane drank wine, and Mary drank an orange juice. Under the MD hypothesis, insensitivity to island violations is expected because movement is not involved, shown in (14-15). (14) wh-islands [[CP1Susan wonders when John ordered [CP2Bill wants to know when he returned

], and [DP the tickets for the opera]]].

(15) Complex NP constraint [[CP1John likes a professor who lectured on [CP2Mary likes a graduate student who debunked

], and [DP a recent theory of RNR]]].

In addition, non-constituent RNR is no problem for the MD account, either. In (16), we can have a “multiple” MD structure. DP and PP each can be multiply dominated in an independent location of the tree. (16) John borrowed, and Bill [stole [a large amount of money] [from the bank]]

21

Seungwan Ha &P qp TP &’ 3 3 DP VP & TP John r 3 VP DP VP r B  ill 3 V VP PP borrowed 3 6 V DP from the bank stole 6 a large amount of money

However, the MD hypothesis faces some insurmountable problems. First, MD can account for strict and third-party readings, but not a sloppy reading. The pronoun his can be bound by the DP Bill for strict identity in (17a). Similarly, the pronoun can be bound by a third-person in (17c). However, to get sloppy reading in (17b), the pronoun must be bound by each subject DP of both conjuncts simultaneously. (17) Chrisi likes, and Billj loves hisi/j friend.

&P qp TP1 &’ 3 3 DP VP & TP2 Chrisi t 3 V DP VP likes Billj ry V DP loves 5 hisi/j father

a. Chris likes Bill’s friend, and Bill loves Bill’s friend. b. Chris likes Chris’ friend, and Bill loves Bill’s friend. c. Chris likes Jason’s friend, and Bill loves Jason’s friend.

(Strict identity) (Sloppy identity) (Third-party reading)

Secondly, under the MD account, Vehicle Change (VC) effects are not expected, either. Yet VC allows a proper name to be shifted into a pronoun, thereby avoiding a Principle C violation (Fiengo & May 1994). For example, the ellipsis site in (18a) should violate Principle C, but the sentence is good. Fiengo & May show that the same holds for pronoun and reflexive, as shown in the example (18b). (18) a. Mary loves Johni, and hei thinks Sally does , too b. Mary did , but Joshi didn’t vote for himselfi. (Fiengo & May 1994:220)

22

MD CAN’T but Ellipsis CAN account for RNR Consider (19). (19a-b) show that the VC effect is observed in RNR as well. The proper name must have shifted into a reflexive in (19a), since no Principle C violation occurs. And the pronoun can be shifted into the reflexive in (19b), as no Principle B violation occurs. (19) a. Mary heard that Johni submitted , but Sue said that Bill actually wrote . b. Johni couldn’t , so I nominated himi A similar VC effect can be found in the alternation between some and any. Klima (1964) claims that some and any are different morphological realizations of the same underlying element. In (20), the Negative Polarity Item (NPI) in the ellipsis clause can be reconstructed with a Positive Polarity Item (PPI). Therefore, it is not (20a) but (20b) that is reconstructed for the sentence. (20) Mary didn’t read any article, but John did. a. *Mary didn’t read any article, but John did read any article. b. Mary didn’t read any article, but John did read some article. MD cannot account for this alternation in RNR. Consider (21), where the first conjunct is not an environment that can license the NPI. Under the MD account, we predict (21) to be ruled out because the NPI cannot be licensed in the first conjunct, as shown in (21a). However, the fact that the reading in (21b) is possible shows the VP in the first conjunct should nevertheless be able to host the NPI. (21) John read, but he hasn’t understood any of my books. a. *John read any of my books, but he hasn’t understood any of my books. b. John read some of my books, but he hasn’t understood any of my books. One might think that it all depends on the view of NPI licensing. Let us suppose that the NPI licensing under coordination can be satisfied if it is licensed by at least one conjunct. However, this assumption cannot be right, because if so, (22) should also be possible. In the case of (22), the NPI is overtly pronounced in the non-NPI-licensing environment. The first conjunct has negation that can license the NPI, but the sentence is ruled out, which indicates both conjuncts must be able to license the NPI. (22) * John hasn’t understood, but he has read any of my books. a. *John hasn’t understood any of my books, but he has read any of my books. b. *John hasn’t understood some of my books, but he has read any of my books. A similar problem arises with number agreement in (23). The numeral two in the first conjunct requires a plural object to check its [+plural] feature, and the singular NP essay

23

Seungwan Ha in the antecedent clause cannot check the [+plural] feature. Under the MD account, we predict the numeral in the first conjunct cannot merge the NP in the antecedent clause because there is a number mismatch between probe and goal. The prediction is not borne out since (23) is perfectly grammatical. (23) Bill has to read two, and Mary must write one essay by tomorrow. Wilder’s (1999) analysis does not rely on a parallelism requirement, so even when the two conjuncts are not structurally parallel, a sentence should be acceptable so long as it is linearizable. In support of this, Wilder gives (24), which he judges to be acceptable. (24) John should fetch and give the book to Mary.

(Wilder 1999: 595)

However, a very similar structure in (25) is much worse, although it should be linearizable as well. (25) *Mary congratulated , and Bill gave the winner a prize. (p.c. Norvin Richards) One possible explanation for why (24) sounds acceptable is that it involves a kind of “reanalysis” of fetch and give as a complex predicate. In support of this, if we pair two verbs that are not as easily understood as a single action, such as critique and give in (26), the sentence is dramatically degraded. (26) ?*John should critique and give the book to Mary. In an SOV language, like Korean, the object of the first conjunct can be gapped, instead of the verb, as in (27), even though this should face the same problem for linearization as backwards gapping does in English. (27) Bill-i sse-ess-ko, Mary-ka ku chayk-ul caymiisske ilk-ess-ta. B.-Top the book-Acc wrote-CONJ M.-Top the book-Acc joyfully read-Past-Dec. ‘Bill wrote the book, and Mary enjoyed reading the book.’ 1.2.2. Strict phonological deletion Another type of non-movement analyses is the strict phonological deletion account. Wexler & Culicover (1980), Hartmann (2000), Bartos (2001), and Abels (2004) have argued that RNR is possible by deletion under strict phonological identity (both above and under the word-level) between parallel structures under coordination. Examples in (28) each show that the identical DP a new car and part of the word rated can be deleted. (28) a. Jennifer bought , but Mary leased a new car.

24

MD CAN’T but Ellipsis CAN account for RNR b. The Red Sox players have been under-rated, and the Yankees players have been over-rated. However, similar types of problems occur with the MD account. Regarding VC, it is puzzling that a form different from its antecedent can be deleted. For example, in (19a) and (21b), repeated in (29a-b), the deletion site does not contain a corresponding antecedent in the second conjunct; that is, the reflexive in (29a) and the PPI in (29b) do not have their antecedent under the strict phonological deletion. (29) a. Mary heard that Johni submitted , but Sue said that Bill actually wrote . b. Mary didn’t read any article, but John did read some article. It has been observed in the literature that morphological mismatch is possible in VPellipsis (Warner 1986, Lasnik 1999, and Lightfoot 1999). For example, the tense can be mismatched in an ellipsis sentence, John slept and Mary will , too. The same phenomenon is observed in RNR examples (30) and (31), which is unexpected if RNR obeys a strict PF identity condition. (30) John couldn’t , so I helped Judyi work on heri homework. (31) John couldn’t , so I had his hair cut. Strict PF-deletion accounts also rely on structural parallelism. However, the following example (32) shows that RNR does not have to observe structural parallelism. (32) Bill might wish he had , but this isn’t a car he has driven ta car. (Johnson 1996: 6) To summarize this section, I have argued that the current analyses of RNR face several challenges. In the next section, I will propose an alternative account for RNR. 2. The ellipsis account I propose that RNR is a type of ellipsis. An ellipsis view does not face the same problems as Multiple Dominance and strict phonological deletion accounts, such as i) Sloppy Identity, ii) Vehicle Change, and iii) Parallelism. Sloppy Identity is easily explained under ellipsis, since there are two occurrences of pronominal variables in (17), repeated in (33). Each occurrence of the pronoun can be independently bound by the subject of each conjunct, resulting in sloppy identity. (33) Chrisi likes , and Billj loves hisj friend. = Chris likes Chris’ friend, and Bill loves Bill’s friend.

25

(Sloppy reading)

Seungwan Ha As observed in the ellipsis literature, the VC effect is not a problem for RNR. Reconstruction is not sensitive to the change of value, so a different value of the antecedent can be realized in the ellipsis site. Let us reconsider (21-22), repeated in (3435). Under the ellipsis account, it is clear that there is a contrast in grammaticality between (34) and (35). In (34), what we observe is a VC effect, but in (35), the overtly pronounced NPI in the second conjunct cannot be under consideration of any feature value change, but must be licensed by its local licensor. (34) John read, but he hasn’t understood any of my books. (35) *John hasn’t understood, but he has read any of my books. Lack of Morphological Identity is also common in ellipsis. In a RNR example (36), the tense of the verb have does not have to be phonologically identical with that of its antecedent had. (36) John couldn’t , so I had his hair cut. Parallelism between the two conjuncts is crucial for the ellipsis to be licensed. We will discuss what kind of parallelism (i.e. syntactic or semantic) should be met in RNR in section 4. Let us assume for now that parallelism should be observed in RNR as well. Then, the reason that (25), repeated in (37), is ruled out is because the two conjuncts are not parallel. (37) *Mary congratulated , and Bill gave the winner a prize. 3. Potential problems & Solutions 3.1. Relational Modifiers In this section, we will discuss some problems potentially posed against the ellipsis account. Abbott (1976), Jackendoff (1977), and Gazdar (1981) have claimed Relational Modifiers, such as similar, together, different, respectively, are problematic for nonmovement analyses. Consider (38). If (38b) is the underlying structure for (38a), it is not clear why only the ellipsis example (38a) is possible. (38) a. Peter sings and Mary whistles a similar tune. b. *Peter sings a similar tune and Mary whistles a similar tune. Hartmann (2000) acknowledges but dismisses this problem, relying on crosslinguistic and individual variation. Hartmann observes that the equivalent German example is not grammatical, shown in (39), and that not every English informant accepted (38a).

26

MD CAN’T but Ellipsis CAN account for RNR (39) *Hans singt und Maria pfeift ein ahnliches lied. (= (38a))

(Hartmann 2000:79)

I argue instead that the first verb in (38a) is intransitive. (38a) is not derived from (38b), so it is not truly a RNR example. Rather, it means something closer to Peter sings, and then Mary starts whistling with a similar tune. This hypothesis predicts that a structure that contains a solid transitive verb in the first conjunct will be bad. The prediction appears to be borne out in (40). (40) Mary and I have a very different taste in fine art. We have seen a lot of paintings in a last couple of days. For example, *?I like, but Mary absolutely hates a similar painting. Abels (2004) also acknowledges (41) would be problematic for the non-movement analyses. The reading we are interested in is the distributive reading (41a), although the collective reading is also possible (41b). In addition, the intransitive use of sang for the first conjunct is available in (41c). (41) Carrie sang, and Mike recorded two very different songs. (Abels 2004) a. Carrie sang a song, and Mike recorded a song, and they are different from each other. b. Carrie sang two different songs, and Mike recorded two very different songs. c. Carrie sang, and Mike recorded two songs independently. That is, what Carrie sang and what Mike recorded are different. The distributive reading in (41a) is comparable to (42), and this reading cannot be explained by the ellipsis account (see de Vos & Vicente (2005) for an interesting proposal about (42)). (42) John likes, and Bill loves Susan and Mary. = John likes Susan, Bill loves Mary.

(de Vos & Vicente 2005)

Abels compares (41) with the VP ellipsis example (43), and claims that the distributive reading is not available in (43), but only the collective reading is available. I argue that this shows that it is the collective reading in (41b) that is derived by ellipsis, not the distributive reading in (41a). The intransitive use of the verb sing is not possible in (43), which indicates that that reading is not derived from ellipsis, either. The ellipsis account for RNR is compatible with the same environment with the ellipsis, and the readings of RNR where the forward ellipsis does not allow should be explained by something else, such as de Vos & Vicente (2005). (43) Carrie sang very different songs, and Mike did , too. ≠ Carrie sang a song, and Mike sang a song, and they are different from each other.

27

Seungwan Ha 3.2. Scope ambiguity Sabbagh (2003) poses challenges for non-movement views based on the scope differences between the RNRed version (44a) and the overtly pronounced version (44b). He claims that (44a) allows an inverse scope reading, but (44b) does not. Sabbagh’s claim is if the underlying structure of the deletion view is (44b), and the DP bearing the universal quantifier in the first conjunct is deleted under the identity with its antecedent, then the deleted version must take the same scope with the underlying structure, meaning only surface scope reading should be possible in (44a) too. (44) a. Some policeman arrested t, but ended up releasing t, every teenager who was near the crime scene. ( ∃ > ∀ or ∀ > ∃) b. Some policeman arrested every teenager, but ended up releasing every teenager. (Only ∃ > ∀) Bošković (2004) argues that Rightward ATB Heavy NP Shift needs to be distinguished from RNR. I claim that RNR is involved in (44b), but not in (44a). The NP in (44a) undergoes Rightward ATB movement because it is heavy. Since the nature of the movement is A-bar movement, inverse scope is available. On the other hand, in (44b), the target NP does not move since it is not heavy, so only surface scope reading is available. The NP in the first conjunct can be deleted in (44b) by RNR. This distinction is clearer if we insert a temporal adjunct in (45). The inverse scope reading is unavailable when no movement occurs. (45) a. Some policeman arrested t, but ended up releasing t last night, every teenager who was near the crime scene. (= HNPS, ∃ > ∀ or ∀ > ∃) b. Some policeman arrested , but ended up releasing every teenager last night. (= RNR, Only ∃ > ∀) 4. Licensing condition for RNR 4.1. Alternative semantics for Focus If the ellipsis account is correct, the next question to ask is if RNR has the same licensing conditions with the other types of ellipsis phenomena. Hartmann (2000) has discussed semantic licensing conditions for RNR, based on an alternative semantics for focus (Rooth 1992). (46) gives Hartmann’s licensing conditions for RNR. (46) Hartmann’s licensing conditions for RNR: a. Both conjuncts must be structurally identical, b. The pre-RNR elements in the first conjunct and the elements with which they contrast in the second must be focused, c. The focused elements create sets of alternatives, and the sets of alternatives for both conjuncts must be identical.

28

MD CAN’T but Ellipsis CAN account for RNR Following Rooth, structural isomorphism is crucial in Hartmann (2000), so both conjuncts must be structurally identical, and the pre-RNR elements in the first conjunct, for example, peeled and the elements with which they contrast in the second, cooked, must be focused in (47). The focused elements create sets of alternatives, and the sets of alternatives for both conjuncts must be identical. Consider the formula in (47). (47) [TP1[JOHN]F [PEELED]F ], and [TP2[MARY]F [COOKED]F the squash]. a. [[ TP1]]f = λp [∃z∈ALT(John′) [∃Q∈ALT(peeled′) [p=Q(ιx.squash(x))(z) ]]] b. [[ TP2]]f = λp [ ∃z∈ALT(Mary′) [∃Q∈ALT(cooked′) [p=Q(ιx.squash(x))(z) ]]] In (47), TP1 and TP2 are structurally identical. The subject and the verb in each conjunct are focused and generate sets of alternatives, shown in (47a-b). Since (47a) and (47b) are the same, RNR is licensed, so PF-deletion follows the focused verb in the first conjunct. However, Hartmann’s licensing conditions have a couple of problems. (48) Bill might wish he HAD , but this ISN’T a car he has driven. Hartmann predicts (48) to be ungrammatical, contrary to fact, since the conjuncts are not syntactically identical, and consequently, the antecedent and the elided part are not identical. The focused pre-RNR element is an auxiliary had in the first conjunct, so the ellipsis part is vP driven a car. The focus in the second is assigned on a copular isn’t, so antecedent must be DP a car that he has driven. 4.2. Semantic licensing condition The main problem of Hartmann’s licensing condition is the insistence on structural parallelism. To address the problems above, I claim that semantic licensing condition, such as Merchant (2001)’s e-GIVENness, should be observed.1 With e-GIVENness, we can account for the contrast between deaccented RNR and ellipsis RNR. Merchant claims that mutual entailment relationship must be established between the two conjuncts for ellipsis to be licensed, but deaccenting can be licensed if A entails F-clo (E), regardless of whether E entails F-clo (A). The same contrast holds for RNR, as shown in (49). In (49a), ellipsis RNR can be licensed because A and E mutually entail each other. When E does not entail F-clo (A), ellipsis RNR cannot be licensed, as shown in (49b), but instead, deaccented RNR is licensed in (49c). 1

Merchant (2001)’s e-GIVENness An expression E counts as e-given iff E has a salient antecedent A and, modulo ∃-type shifting, (i) A entails F-clo (E), and (ii) E entails F-clo (A) (Merchant 2001:26) F-clo (A) is the proposition derived by replacing all focused constituents by variables and existentially binding them.

29

Seungwan Ha (49) a. Mary COULDN’T , but John COULD eat the snail at the French restaurant. (√ Ellipsis RNR: A & E mutually entail each other) b. * Mary COULDN’T , but John COULD eat the snail at the French restaurant. (Ellipsis RNR: A
E, but *E
A) c. Mary COULDN’T eat a French dish, but John COULD eat the snail at the French restaurant. (√ Deaccented RNR: A
E) RNR does not have to observe structural (form) identity between the two conjuncts as shown in (48). The antecedent has several logically equivalent possibilities (50a-c). 2 Among those, we choose one of them that can serve as an antecedent for the ellipsis site. And (50c) is chosen and taken to be an antecedent for ellipsis in the first conjunct. (50) Bill might wish he HAD , but this ISN’T a car he has driven. Logically equivalent possibilities for TP(A): a. This isn’t a car he has driven ta car. b. It’s not the case that this is a car he has driven ta car. c. √ It’s not the case that he has driven this car. The Verum focus on had and isn’t generates a set of polarity alternatives in (50). The two alternatives for TP(A) are {It’s the case that Bill has driven this car, It isn’t the case that Bill has driven this car.}. TP(E) also generates a set of alternatives, such as {It’s the case that Bill had driven this car, It isn’t the case that Bill had driven this car.}. RNR can be licensed because TP(A) and TP(E) mutually entail each other. PF-deletion follows the focused pre-RNR element in the first conjunct. 5. Conclusion To conclude, I reviewed challenges for current analyses on Right Node Raising, in particular the commonly adopted Multiple Dominance hypothesis. As an alternative, I proposed that RNR is best explained under the ellipsis account. I argued that RNR is licensed by semantic constraints (i.e. mutual entailment relationship between the antecedent and the RNRed clause), rather than syntactic isomorphism. References Abbott, Barbara. 1976. Right node raising as a test for constituenthood. Linguistic Inquiry 7: 639-642. Abels, Klaus. 2004. Right Node Raising: Ellipsis or ATB movement? North East Linguistics Society 34: 44-59. Bartos, Huba. 2001. Sound-Form Non-Insertion and the Direction of Ellipsis. Acta Linguistica Hungarica 48: 3–24. 2

What triggers this accommodation is beyond the focus of this paper.

30

MD CAN’T but Ellipsis CAN account for RNR Bošković, Željko. 2004. Two notes on right node raising. University of Connecticut Working Papers in Linguistics 12: 13-24. Bresnan, Joan. 1974. The position of certain clause-particles in phrase structure. Linguistic Inquiry 5: 64-619 Fiengo, Robert & Robert May. 1994. Indices and identity. Cambridge, MA: MIT Press. Gazdar, Gerald. 1981. Unbounded dependencies and coordinate structure, Linguistic Inquiry 12: 155-184 Hartmann, Katharina. 2000. Right Node Raising and Gapping: interface conditions on prosodic deletion. Amsterdam: John Benjamins. Jackendoff, Ray. 1977. X' syntax: A study of phrase structure. Cambridge, MA: MIT Press. Johnson, Kyle. 1996. When VP go missing. Glot International 2: 2-9 Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Klima, Edward. 1964. Negation in English. In The Structure of Language, J. A. Fodor and J. J. Katz (eds.). Englewood Cliffs, NJ: Prentice Hall. Lasnik, Haward. 1999. Verbal morphology: Syntactic Structures meets the Minimalist Program. Lasnik, H., Minimalist Analysis, 97-119. Malden, MA: Blackwell. Lightfoot, David. 1999. The Development of Language: Acquisition, Change, and Evolution. Malden, MA: Blackwell. McCawley, James. 1982. Parentheticals and discontinuous constituent structure. Linguistic Inquiry 13: 91-106 McCloskey, James. 1986. Right node raising and preposition stranding. Linguistic Inquiry 17: 183-186 McCloskey, James. 1999. On the right edge in Irish. Syntax 2: 189-209 Merchant, Jason. 2001. The syntax of silence. Oxford: Oxford University Press. Postal, Paul. 1974. On raising. Cambridge, MA: MIT Press Postal, Paul. 1998. Three investigations of extraction. Cambridge, MA: MIT Press. Rooth, Matt. 1992. A theory of focus interpretation. Natural Language Semantics 1: 75116. Ross, John. 1967. Constraints on variables in syntax. MIT dissertation. Sabbagh, Joseph. 2003. Ordering and Linearizing Rightward movement. WCCFL 22: 436-449. Somerville, MA: Cascadilla Press. de Vos, Mark, & Luis Vicente. 2005. Coordination under Right Node Raising. WCCFL 24: 97-104. Warner, A. 1986. Ellipsis conditions and the status of the English copula. York Papers in Linguistics 12: 153-172. Wilder, Chris. 1999. Right Node Raising and the LCA. WCCFL 18: 586-598. Wexler, Ken & Peter Culicover. 1980. Formal principles of language acquisition. Cambridge, MA: MIT Press.

31