The eclectic approach

The common ground of the literature on RNR advocates a single theory (i.e. exclusivist hypothesis), be it the ATB-movement, PF-deletion, MD, or external remerge. Nonetheless, each of the theories is not adequate to account for all the RNR properties. Barros and Vicente (2011) acknowledge the need for multiple approaches since asymmetric vehicle change and summative agreement effects/internal readings are in complementary distribution. Therefore, an eclectic theory of RNR is proposed: both the backward ellipsis and the MD are possible sources for RNR.

The examples below illustrate the interaction of summative agreement effects and vehicle change.

(89) No summative agreement, Vehicle Change OK

She_i fears [____], but Bob is not worried [____], that Alice_i might lose the election.

(90) No Vehicle Change, summative agreement OK

Alice_i fears that Beatrix [____], but Claire worried Diane [____], {have/*has} decided

48

to nominate Esther.

(91) Summative agreement, Vehicle Change blocked

She_*i/k fears that Alex [____], and I worry that Bob [____], {have/*has} decided to

nominate Clairei.

Instance (89) can be well-accounted for under PF-deletion, but not MD; oppositely, (90) is predicted under MD rather than PF-deletion. As for the conflict in (91), when there is summative agreement, vehicle change is blocked, indicating that summative agreement and vehicle change are in complementary distribution. However, Larson (2012a) notes an

important difference between the ellipsis example (89) and the test example (91): in (91), the proper name Claire is distanced from its antecedent by another proper name Alex in the first conjunt. Larson gives a similar test sentence in (92) prohibits asymmetric vehicle change, indicating that summative agreement effects may not be blamed for the lack of vehicle change.¹⁷

(92) *Shei fears that Alex thinks, but Bob is not worried, that Alicei might lose the election.

(Larson 2012a)

Barros and Vicente observe another set of conflict between asymmetric vehicle change and the availability of internal reading. The PF-deletion approach, but not the MD approach, explains (93); conversely, the MD approach, but not the PF-deletion approach, accounts for (94). Particularly, vehicle change and internal reading are not compatible in the same sentence as in (95). That is, when there is vehicle change, the internal reading is blocked.

Again, according to Barros and Vicente, the conflict of vehicle change and internal reading

17 Larson (2012a) brings out that the problem of (88) and (89) may be due to more R-expressions in the illicit sentences. An extra name seems to prevent the vehicle change reading in (i)

(i) *She_i fears, but Alex said Bob is not worried, that Alice_i might lose the election.

49

suggests an eclectic approach, rather than the exclusivist approach.

(93) No internal reading, Vehicle Change OK

Shei thinks that he must [____], but Bob fears that he won‘t [____], come up with a topic that satisfies Alice_i.

(94) No Vehicle Change, internal reading OK

Alice absolutely must [____], and Beatrix is obliged to [____], come up with different topics.

(95) Either Vehicle Change or internal reading blocked

Shei absolutely must [____], and Bob is obliged to [____], present different topics to Alice‘s_i supervisor.

However, contrary to the eclectic hypothesis, Larson (2012a) points out a well-formed example as given in (96), where both the asymmetric vehicle change and conjunct-internal reading are available.

(96) She_i thinks that she absolutely must, and Bill fears that he won‘t, present different topics

to Alice‘si supervisor. (Larson 2012a)

Besides, Larson (2012a) offers more pairs of examples which display the combination of morphological mismatch (assumed under PF-deletion) and summative agreement

effects/internal reading/scope ambiguities (assumed under the MD), shown in (97-99).

(97) With/without summative agreement, morphological mismatch is not OK

a. *Alice is happy that Iris [negotiated her salary with the manager], and Claire is proud that Daniel, have negotiated his salary with the manager.

50

b. *Alice is happy that Iris [can spell her name], and Claire is proud that Daniel, can spell his name.

(98) With/without morphological mismatch, internal reading is OK a. Alice must, and Iris should, work on different topics.

b. Alice must [work on different topics], and Iris ought to be, working on different topics.

(99) With/without morphological mismatch and scope ambiguities

a. Some woman must, and some man ought to be, working with every student.

b. Some woman hates, and some man loves, every dog in the pound.

Whether the summative agreement is present ((97a)) or not ((97b)), morphological mismatch is unacceptable, showing that summative agreement does not ban it. Whether morphological mismatch is not seen ((98a)) or seen ((98b)), the internal reading is always available. Whether morphological mismatch is shown ((99a)) or not ((99b)), scope

ambiguities (some> every; every>some) are observed.

Let us look at Sabbagh‘s (2012) argument for the eclectic approach, which claims that the ATB-movement may also be a possible way to derive English RNR. Following Barros and Vicente‘s (2011) logic, Sabbagh proposes that scope ambiguity is in complementary distribution with some other properties that can only be accounted for under the PF-deletion or the MD approach.

(100) a. The lieutenant will either arrest or shoot every suspected arsonist with his rifle.

(Sabbagh 2012) b. The lieutenant will either arrest or shoot with his rifle, every suspected arsonist.

(Sabbagh 2012)

51

Example (100a) has only a surface scope: every suspected arsonist will be arrested or will be shot; besides the surface scope reading, example (100b) has an inverse scope reading (i.e. every scopes over or): some suspected arsonist might be arrested while others might be shot. The scope ambiguity of (100b) indicates that the RNR target is external, just as the assumption of the ATB-movement approach. Therefore, Sabbagh claims that this approach might also be a possible way to derive RNR. Nevertheless, (100b) involves the interaction of surface and inverse scopes, and Sabbagh does not specify what properties are in

complementary distribution with scope ambiguity and can only be derived under the PF-deletion or the MD approach.

In conclusion, Barros and Vicente‘s eclectic theory provides a possible multiple analysis of RNR, including the PF-deletion and the MD approaches. This approach, however, is weakened with the counterexamples found in Larson (2012a). Moreover, Sabbagh claims that in addition to the PF-deletion and the MD approaches, the ATB-movement approach is at least among the possible ways to derive RNR.

Table 6 Properties explained by the eclectic approach

1. Right Edge Restriction (RER) 

2. Nonconstituency ?

3. Coordinate Structure Constraint (CSC) ? 4. Right Roof Constraint (RRC) violation ?

5. Island insensitivity ?

6. P-stranding ?

7. Strict/sloppy readings ?

8. Asymmetric vehicle change 

9. Asymmetric licensing of NPI ? 10. Licensing of relational modifiers 

11. Scope ambiguity 

12. Summative agreement 

52

2.7 The sparse approach

Due to the inadequacy of the previous RNR theories, Larson (2011a) and his subsequent works argue for a sparse approach with respect to the inherent syntactic incompleteness of RNR. In this analysis, a c-command relation only exists between the second conjunct and the RNR target. The gap in the first conjunct¹⁸ becomes fully interpretable by extra-grammatical inferencing. The RNR target is presupposed as the background material (Hartmann 2000) and thus restricts the relevant events in each conjunct. Following Herburger (2000), the

interpretation of each event is based on the event structure of the verb and what it is lacking.¹⁹ For instance, the example (101) is represented in (102).

(101) Ivan bought, and Ivy read, Pale Fire. (Larson 2015) (102) [Ivan bought] [and [Ivy read Pale Fire]]. (Larson 2015)

In (101), the second conjunct is a complete clause with the obligatory argument (i.e. the RNR target) contained in it whereas the first conjunct is an incomplete clause due to the lack of the obligatory argument. However, the first conjunct is interpretable via inferencing, which is restricted by the shared target Pale Fire: There are event-A and event-B; both events

involve Pale Fire such that Ivan is the agent of event-A which is a past-buying and Ivy is the agent of event-B which is a past reading.

Based on the extra-grammatical inferencing, Larson argues that the sparse approach captures a much freer interpretation of the RNR target. Concretely, this approach accounts for

18 Larson assumes that the first conjunct with a missing part is acceptable under Predicate Function (Pietroski 2005), allowing a nonlocal semantic relation—a lexico-pragmatic concept which composes the semantic meaning through the conceptual well-formed of the event (with appropriate thematic roles).

19 For instance, the restrictor, the unfocused element wrote poetry in (i) is obligatorily interpreted as: Some event of past poetry writing was such that its agent was Rosalia. She is the agent of events limited to those of past poetry writing. The restrictor is interpreted irrespective of not being explicitly represented as complements to their respective verbs.

(i) ROSALIA wrote poetry (Larson 2015)

53

(i) no island effects since the RNR target does not move, (ii) the asymmetric NPI licensing since only the second conjunct and the RNR target bears a c-command relation, (iii) the scope ambiguity since the only one instance of the RNR target is not coordinated and thus is

allowed to take covert movement and scope high, and (iv) asymmetric vehicle change since nothing in the first conjunct c-commands the target, the coindexation of the pronoun (in the first conjunct) and the target (in the second conjunct) is possible. Note that even though Larson does not discuss about P-stranding, we think the sparse approach may also account for it. Without a c-commanding relation between the RNR target and the first conjunct, the preposition can be stranded in the first conjunct and be fully interpretable via inferencing with its complement contained in the RNR target.

Although this sparse approach justifies the above RNR properties, Husband (2015) points out a problem as in (103).

(103) a. [John bought] and [Mary read the book yesterday]. (Husband 2015) b.*[John bought] and [Mary slept yesterday]. (Husband 2015)

Without a grammatical relation between the first conjunct and the RNR target, the

grammatical contrast between (103a) and (103b) cannot be accounted for. It is unclear how an argument in the second conjunct is able to make the first conjunct fully interpretable without any c-command relation.

Furthermore, we find problems in terms of the available sloppy reading, the

interpretation of asymmetric NPI licensing, licensing of relational modifiers and summative agreement effects. Recall that a RNRed pronoun can be a bound variable simultaneously referring to the subject in the first and second conjuncts.

(104) John_i likes but Bill_j hates his_(i/j)

father.

54

As in (104), the RNRed his can refer to both John and Bill. It is not clear how this property can be explained under Larson‘s account. Specifically, how to derive the coindexation of the RNR target and John as Larson assumes that the first conjunct can be complete and

interpretable via inferencing with the target in the second conjunct.

Also, if the RNR target is shared by inferencing between the two conjuncts, then this approach has to specify how to interpret the sequence of the first conjunct (which lacks a licensor) plus the RNRed NPI, as in (105): John bought + any books about insects.

(105) John bought, but Bill didn‘t read, any books about insects.

With respect to the licensing of relational modifiers, RNR requires a plural NP or plural predicate as in (106), meaning that the car John sold and the one Bill bought are identical.

(106) John sold and Bill bought the same car.

Nevertheless, via inferencing, we can only obtain the external reading of the first conjunct.

That is, an additional mechanism might be needed to derive the internal reading of (106) under the sparse approach. Just as relational modifiers are licensed in plural contexts, summative agreement needs a plural antecedent to be licensed. If there is no c-commanding relation between the first conjunct and the RNR target, we would expect the singular verb, not the plural verb, to be licensed as in (107). Obviously, the prediction is contrary to fact.

(107) [Sue is proud that Bill], and [Mary is happy that John],*has/have traveled to England.

As summarized in Table 7, the sparse approach directly accounts for certain properties of RNR while having trouble explaining other properties.

55

Table 7 Properties explained by the sparse approach

2.8 Chapter summary

In this chapter, we have shown the general properties of English RNR constructions and have reviewed previous significant literature on the RNR. As can be seen from Table 8, Sabbagh‘s (2007) version of ATB-movement approach, unlike the traditional versions, explains the properties of constraints on typical syntactic movements. The PF-deletion approach is argued to solve the difficulties which would arise under the traditional

ATB-movement approaches. Assuming two copies of the RNR target and no overt movement involved, the PF-deletion approach easily avoids violations induced by the movement

constraints, sloppy reading, and vehicle change. Nevertheless, neither the ATB-movement approach nor the PF-deletion approach can analyze the summative agreement in RNR. This property of RNR is specifically explained under the MD approach. Yet, it fails to explain sloppy reading and asymmetric properties. The external remerge approach captures

asymmetric licensing of NPI and verbal inflection mismatch; nevertheless, the movement of the RNR target to a non c-commanding position is a theoretic problem. The eclectic approach is proposed since certain properties of RNR are in complementary distribution. However, 1. Right Edge Restriction (RER) 

2. Non-constituency 

3. Coordinate Structure Constraint (CSC)  4. Right Roof Constraint (RRC) violation 

5. Island insensitivity 

6. P-stranding 

7. Strict/sloppy readings ?

8. Asymmetric vehicle change 

9. Asymmetric licensing of NPI 

10. Licensing of relational modifiers ?

11. Scope ambiguity 

12. Summative agreement ?

56

such claims seem to be refuted by empirical evidence—the occurrence of one property cannot be blamed for the absence of another property. Last but not the least, the sparse approach claims an asymmetric c-command relation and the RNR is explained by extra-grammatical inferencing. Still, it needs to specify how the process of the inferencing can account for RNR properties.

In the next chapter, we will first present the general properties of Chinese RNR and make comparison between Chinese RNR and English RNR. Then we will discuss two previous approaches proposed to the Chinese RNR in the literature.

Table 8 Summary of different approaches to RNR

Approaches

57

Chapter 3

Properties and Previous Analyses of the Chinese RNR

In chapter 2, we have seen the general properties of English RNR and have reviewed three main approaches (i.e. TB-movement, PF-deletion, and MD) and two recently proposed approaches (i.e. the eclectic approach and the sparse approach.) In this chapter, we aim to show that (i) Chinese RNR shares many similarities with the English one except for the asymmetric island effects and asymmetric P-stranding, (ii) no asymmetric vehicle change, summative agreement, and scope ambiguity are observable in Chinese RNR, and (iii) two analyses of Chinese RNR, namely Cheng‘s (2012) multidominance approach and Wang‘s (2014a, b) approach with TEC plus the leftward focus movement. The chapter is organized as follows. Section 3.1 presents the general properties of Chinese RNR. Specifically, there are some restrictions on Chinese RNR, which diverges from the English counterpart. Section 3.2 reviews two previous works on Chinese RNR. Finally the summary is given in section 3.3.

3.1 General Properties of Chinese RNR

3.1.1 Similarities between English RNR and Chinese RNR Right Edge Restriction (RER)

As in English RNR, the RER is also observed in Chinese RNR (Wang 2014b). The RNR target shangban ‗work‘ in (1a) is at the rightmost position of each conjunct, following RER.

The illicit (1b) results from the RNR target mingtian ‗tomorrow‘ not being at the right peripheral of each conjunct.

(1) a. [Zhangsan mingtian hui qu ____], keishi [Lisi mingtian bu hui qu ____]

Zhangsan tomorrow will go but Lisi tomorrow not will go

58

shangban

work

‗Zhangsan will, but Lisi won‘t go to work tomorrow.‘

b. *[Zhangsan __ hui qu shangban], keishi [Lisi ___ bu hui qu shangban], Zhangsan will go work but Lisi not will go work

mingtian

tomorrow

‗Zhangsan will, but Lisi won‘t go to work tomorrow.‘

In addition, Wang (2014b) observes that RER is not enough to account for the cases of Chinese RNR. Seen in (2), the RNR target is at the rightmost in each conjunct, yet, the sentences are unacceptable.

(2) a. Descriptive Complement

*[Zhangsan kan shu kan-de ____], [Lisi ye kan shu kan-de ____], hen kuai Zhangsan read book read-DE Lisi also read book read-DE very fast

(Wang 2014b)

b. Resultative Complement

*[Zhangsan kan shu kan-de ____], [Lisi ye kan shu kan-de ____], hen lei Zhangsan read book read-DE Lisi also read book read-DE very tired

(Wang 2014b) c. Duration Phrase

*[Zhangsan kan shu kan-le __], [Lisi ye kan shu kan-le ___], san-ge xiaoshi Zhangsan read book read-Asp Lisi also read book read-Asp three-Cl hour

(Wang 2014b)

d. Frequency Phrase

59

*[Zhangsan kan shu kan-le ____], [Lisi ye kan shu kan-le ____], san-ci Zhangsan read book read-Asp Lisi also read book read-Asp three-time

(Wang 2014b)

However, we do not agree with Wang‘s judgment for (2). Instead, there are acceptable examples of the RNR containing the above four types of elements. Just like English RNR, Chinese RNR does obey RER, the precondition for deriving RNR as in (3-6). Particularly, if we follow Huang, Li and Li (2009), the phrase following descriptive and resultative de is also subcategorized.

(3) Descriptive Complement

a. [Zhangsan kan shu kan-de ____], [Lisi ye kan shu kan-de ____], jinjin Zhangsan read book read-DE Lisi also read book read-DE interest

you wei

have interest

b. [Zhangsan kan shu kan-de ____], [Lisi ye kan shu kan-de ____], quan Zhangsan read book read-DE Lisi also read book read-DE all

sheng guan zhu

spirit focus attention (4) Resultative Complement

a. [Zhangsan da-de Lisi ___], [Wangwu ye da-de Lisi______],tou po Zhangsan hit-DE Lisi Wangwu also hit-DE Lisi head break

xie liu

blood flow

b. [Zhangsan kan shu kan-de ____], [Lisi ye kan shu kan-de ____], tou Zhangsan read book read-DE Lisi also read book read-DE head

60

hun yan hua

dizzy eye woozy (5) Duration Phrase

[Zhangsan kan shu kan-le ____], [Lisi ye kan shu kan-le ____], zheng Zhangsan read book read-Asp Lisi also read book read-Asp completely

zheng ba-ge xiaoshi

completely eight-Cl hour (6) Frequency Phrase

[Zhangsan kan shu kan-le ____], [Lisi ye kan shu kan-le ____], chaoguo Zhangsan read book read-Asp Lisi also read book read-Asp over

san-ci

three-time

Non-constituency

The RNR target in Chinese may not be a traditional syntactic constituent which can undergo leftward movement or be clefted (cf. chapter 2). The RNR target can be a constituent as a DP in (7a), or a clause in (7b). Besides, non-constituent RNR targets are allowed as in (7c), containing two objects of the double complement construction.

(7) a. Zhangsan dong-le dan Lisi bu mingbai,

zhe ge daoli

Zhangsan understand-Asp but Lisi not understand this Cl reason

‗Zhangsan understands but Lisi doesn‘t understand this reason.‘

b. Zhangsan zuotian cai faxian, dan Lisi zhiqian jiu zhidao, Zhangsan yesterday just discover but Lisi before already know

Wangwu xihuan dubo

Wangwu like gambling

61

‗Zhangsan just discovered yesterday but Lisi already knew before that Wangwu likes gambling.‘

c. Zhangsan zengsong er Lisi ye zengsong -le, Wangwu yi-ben

shu

Zhangsan give and Lisi also give-Asp Wangwu one-Cl book

‗Zhangsan gave and Lisi also gave Wangwu a book.‘

In particular, the non-constituent involving in (7c) can be seen from the ungrammaticality of the leftward movement as in (8a) and of the pseudo-cleft construction as in (8b) (cf. Huang, Li, and Li 2009).

(8) a. *[Wangwu yi-ben shu] Zhangsan g zengsong-le Wangwu one-Cl book Zhangsan give-Asp

‗[Wangwu a book] Zhangsan gave.‘

b. *Zhangsang zengsong-le de shi Wangwu yi-ben shu Zhangsan give-Asp DE be Wangwu one-Cl book

‗What Zhangsan gave was Wangwu a book.‘

Coordinate Structure Constraint (CSC)

Both English and Chinese RNR respect Coordinate Structure Constraint. The

representation in which a movement chain starts from a single conjunct and ends out of the coordinate structure is banned. Extraction of the RNR target in two conjuncts (rather than all three conjuncts) as in (9) and (10) is prohibited. In contrast, the grammatical (11) indicates that the CSC is observed.

(9) *[Zhangsan jia fujin you zahuo dian], [Lisi gongsi pangbian you _____], Zhangsan home near have grocery store, Lisi office next to have

62

[Wangwu xuexiao pangbian ye you_____], bianli

shangdian

Wangwu school next to also have convenience store

‗There‘s a grocery store near Zhangsan‘s home, a convenience store next to Lisi‘s office and one next to Wangwu‘s school.

(10) *[Zhangsan jia fujin you _____], [Lisi gongsi panbian you zahuo dian], Zhangsan home near have Lisi office next to have grocery store [Wangwu xuexiao pangbian ye you_____], bianli

shangdian

Wangwu school next to also have convenience store

‗There is a convenience store near Zhangsan‘s home, a grocery store next to Lisi‘s office, and a convenience store next to Wangwu‘s school.‘

(11) [Zhangsan jia fujin you _____], [Lisi gongsi panbian you_____], Zhangsan home near have Lisi office next to have

[Wangwu xuexiao pangbian ye you_____], bianli

shangdian

Wangwu school next to also have convenience store

‗There‘s a convenience store near Zhangsan‘s home, one next to Lisi‘s office, and one next to Wangwu‘s school.

Strict/sloppy readings

Both English and Chinese RNR exhibit strict/sloppy readings in RNR.

(12) Zhangsan xihuan, dan Lisi bu xihuan, ta

de zhaopian

Zhangsan like but Lisi not like he DE picture

‗Zhangsan likes, but Lisi doesn‘t like, his picture.‘

a. Zhangsan likes Zhangsan‘s picture, but Lisi doesn‘t like Lisi‘s picture. (sloppy) b. Zhangsan likes Zhangsan‘s picture, but Lisi doesn‘t like Zhangsan‘s picture. (strict) c. Zhangsan likes Lisi‘s picture, but Lisi doesn‘t like Lisi‘s picture. (strict)

63

A RNRed pronoun can be a bound variable which simultaneously refers to the subject in the first and in the second conjunct as in (12a). Besides, it can refer to only the subject in the first or in the second conjunct, as in (12b) and (12c). Note that a strict reading of the RNR target can be seen clearly by forcing the two subjects to differ in gender, as in (13).

(13) Zhangsan xihuan, dan Mali bu xihuan, ta

de zhaopian

Zhangsan like but Mary not like he/she DE picture

‗Zhangsan likes, but Mary doesn‘t like, his/her picture.‘

a. Zhangsan likes Zhangsan‘s picture, but Mary doesn‘t like Zhangsan‘s picture. (strict)

a. Zhangsan likes Zhangsan‘s picture, but Mary doesn‘t like Zhangsan‘s picture. (strict)

在文檔中 漢語右部節點提升結構 (頁 54-0)