<<

The Interpretation of PRO in the

2001년

서강대학교 대학원

영어영문학과

박 두 홍

The Interpretation of PRO in the Minimalist Program

지도교수 이 홍 배

이 논문을 문학석사 학위논문으로 제출함

2002년 6월 12일

서강대학교 대학원

영어영문학과

박 두 홍

논 문 인 준 서

박두홍의 문학석사 학위논문을 인준함

2002년 6월 12일

주심 김 영 석 (인)

부심 조 숙 환 (인) 부심 이 홍 배 (인)

The Interpretation of PRO in the Minimalist Program

by

Doo Hong Park

A Thesis

Presented to the Faculty of the Graduate School of

Sogang University in Partial Fulfillment of

the Requirement for the Degree of

Master of Arts

July 2002

ACKNOWLEDGEMENTS

Fear holds you in prisoner, but hope can set you free!

When you do something in your life, what is most dreadful to you? In my case, it is fear about something that does not happen yet. During my M.A course at Sogang University, I was in fear and trembling several times. Every time I felt fearful, it was hopes that set me free. The hopes were faith and numerous people around me. Were it not for them, I could not end up my MA course.

First of all, I would like to show my heartfelt gratitude to Prof. Hong-Bae Lee, my thesis adviser. Throughout my MA course, he has provided me insightful lectures on and persevering guidance. Without his considerate heart, I could not finish this thesis. I also express my sincere thankfulness to my committee members, Prof. Young- Seok Kim and Prof. Sook-Hwan Cho, who were willing to take pains in reading my thesis. Completion of my thesis is also indebted to their kind reading of the awkward draft and critical comments on it.

Beyond my thesis committee members, I would like to show my gratitude to Prof. Seo-Young Chae, who gave me warm and considerate lectures on . I also express my thankfulness to Prof. Sung-Bum Lee who showed me what a scholar’s life should be like. I am also grateful to Sung-Eun Cho, who read my thesis throughout and gave me critical comments on it.

In addition, I am very thankful to seniors, Jong-Un Park, Jin-Se Lee, Hong-Oak Yun, Eun-Hae Park, Seung-Koo Park, who were very friendly and helpful. Most of all, my special thanks go to Won-bin Lee, who always encouraged me and helped me to get a clue to my thesis. I am also indebted to other seniors and fellow students at Sogang: Jung- Ah Lee, Jay-Doe Ryu, Sung-Hee Park, Dong-Yon Lee, Yun-Ju Suh, Jun-Won Yoon, Hyun-Jeong Park, Nan-Ie Joo. Thanks also go to Jeong-In Lee, who gave me many helps and Seung-Won Paek, Mi-Young Jee, Eun-Jeong Yang, Young-Jin Jeon, Tae-Joo Roh,

-v- Sung-Eun Lee, Kyung-Hee Koo, Hyun Jin Choi, and Hyun Soon Kim. Special thanks go to my friends, Chan Su Park, Young Min Kim, schoolfellows at Gethsemane at SKKU as well as ones at Eden’s Tree at Sogang University.

Finally, I want to express my heartfelt thankfulness to my family. I cannot find appropriate words to express my gratitude to them. If it were not for their consistent support and affectionate concern for me, I could not continue my study till now. Again I would like to show my special thanks to my father, mother and elder brother Dea-Hong Park for their sacrifice and affection. Most of all, I would like to show my heartfelt gratitude to the one, who was, who is, and who is to come.

Doo Hong Park July, 2002

-vi- Table of Contents

Acknowledgements...... v Table of Contents ...... vii 국문초록...... ix Abstract...... x

ChapterⅠ Introduction ...... 1 1.1 What is ? ...... 1 1.2 Distribution of PRO ...... 2 1.2.1 Classical Approaches...... 2 1.2.1.1 Caseless PRO ...... 2 1.2.1.2 Ungoverned PRO ...... 4 1.2.2 Case-theoretic Approaches...... 6 1.2.2.1 PRO and Null case ...... 7 1.2.2.2 PRO and Enlightened Self Interest ...... 9 1.3 Two Types of Control ...... 10 1.4 A Brief Overview of This Thesis ...... 12

Chapter Ⅱ Framework of Minimalist Inquiries and Derivation by Phase: Chomsky (2000, 2001)...... 14 2.1 Introduction ...... 14 2.2 Core Functional Categories and Clausal Structure ...... 14

2.3 Operations of the Computational System (CHL) ...... 17 2.4 The Notion of Phase, PIC and Spell-Out ...... 19

Chapter Ⅲ Previous Analyses of OC PRO...... 23 3.1 Introduction ...... 23 3.2 Standard Approaches ...... 23 3.2.1 Rosenbaum’s (1967) Minimal Distance Principle ...... 24 3.2.2 Chomsky’s (1981) Intrinsic Properties of Predicates ...... 25 3.3 Minimalist Approaches ...... 27

-vii- 3.3.1 Manzini & Roussou’s (1999) Analysis ...... 27 3.3.2 Hornstein’s (1999) Analysis ...... 29 3.3.3 Problems ...... 35 3.3.3.1 Wanna-Contraction ...... 35 3.3.3.2 PRO ≠ NP-trace...... 37

Chapter Ⅳ A Minimalist Approach to the Interpretation of PRO ...... 39 4.1 Introduction ...... 39 4.2 A Minimalist Analysis based on Agree: Chomsky (2000, 2001)...... 39 4.3 Preliminary Assumptions for a New Analysis...... 41 4.3.1 Separate Operations (i.e., Agree and Move): Lee (1999) ...... 42 4.3.2 On Control ...... 43 4.3.3 Two Types of Non-finite T ...... 44 4.4 A New Analysis of the Interpretation of PRO...... 48

Chapter Ⅴ Conclusion ...... 58

References ...... 60

-viii- 국문초록

본 논문은 공범주인 PRO의 의무적 통제를 Chomsky(2000, 2001)가 제안한 통사적 작용인 일치(Agree)의 결과로 분석하는 데 그 목적이 있다. 통제란 비 시제절의 주어자리에 위치한 공범주 PRO(비통제자)의 지시적 특성이 통제자 의 지시적 특성에 의해 결정되는 것을 의미한다. 본 논문에서는 먼저, 최소주 의 이론에서 PRO의 분포는 지배()란 용어를 쓰지 않고도 설명될 수 있음을 보이고, 나아가, Chomsky(2000, 2001)가 제안한 현 최소주의 가정하에서 의무적 통제의 경우 PRO의 선행사 선택 (즉, 주어에 의한 의무적 통제와 목적 어에 의한 의무적 통제의 구별)이 어떻게 이루어지는 지를 알아볼 것이다. 이러한 목적을 위해, 현 최소주의 이론체계를 개관하고, 원리-매개변항 이론에 서부터 최소주의 이론에 이르기까지 있었던 의무적 통제현상에 대한 이전의 통사적 분석들을 다룰 것이다. 즉, Rosenbaum(1967)이 제시한 “최소 거리 원리 (Minimal Distance Principle)”와 Chomsky(1981)이 제시한 “술어의 내재적 특성 (Intrinsic Properties of Predicates)”에서부터, Manzini and Roussou(1999)가 제시한 “유인작용(ATTRACT)”과 Hornstein(1999)이 제시한 “의무적 통제는 명사구의 흔적과 같다(OC PRO = NP-trace)”란 분석들을 살펴 봄으로써 의무적 통제현상 의 선행사 결정이 어떻게 이루어지는 지를 살펴볼 것이다. 뿐만 아니라, 이러 한 통사적 분석들의 문제점들도 밝히고자 한다. 끝으로, 앞서 언급한 이전 분석들과는 다르게, PRO의 의무적 통제는 통사 적 작용인 일치에 의해 결정됨을 제시하겠다. PRO가 주어와 일치를 이룬 기능 핵 T와 일치를 이룰 경우, 주어에 의한 의무적 통제가 이루어 지고, PRO가 목 적어와 일치를 이룬 기능핵 v와 일치를 이룰 경우, 목적어에 의한 의무적 통 제가 이루어 지는 것으로 보고자 한다. 새로운 분석을 위해, 통제구문은 TP라 고 하는 Bošković(1997)의 주장과 일치(Agree)와 이동(Move)은 별개의 작용이 라는 Lee(1999)의 주장을 가정하겠다.

-ix- A Minimalist Approach to PRO and Control

Abstract

The aim of this thesis is to analyze obligatory control of PRO as a consequence of the syntactic operation Agree proposed by Chomsky (2000, 2001). What control means is that the referential properties of the PRO (controllee) occupied the position of the infinitival is determined by those of the antecedent (controller). In this thesis, I will first show how the concept of “government” can be eliminated in accounting for the distribution of PRO within the Minimalist Program. Secondly, I will show how the particular choice of the antecedent of OC PRO (i.e., the distinction between obligatory subject control and obligatory object control) is determined within the current minimalist assumptions proposed by Chomsky (2000, 2001). For this purpose, I will review the framework of the recent minimalist theory, and consider the previous analyses of obligatory control constructions from the principles and parameters theory to the recent one called minimalist theory: that is, Minimal Distance Principle in Rosenbaum (1967), Intrinsic Properties of Predicates based on Control Theory in Chomsky (1981), ATTRACT in Manzini and Roussou (1999) and OC PRO = NP-trace in Hornstein (1999). In addition, problems of these analyses will be reconsidered in this thesis. Finally, unlike these aforementioned analyses, I will propose that the particular choice of the antecedent of PRO should depend on the syntactic operation Agree. That is, when PRO agrees with the functional head T, which has already agreed with the subject DP (controller), we get obligatory subject control. When PRO agrees with the functional head v, which has already agreed with the object DP (controller), we get obligatory object control. For an alternative analysis, I will assume Bošković’s (1997) that control infinitives are TPs and Lee’s (1999) argument that Agree and Move are separate operations.

-x- ChapterⅠ

Introduction

1.1 What is Control?

As children grow up, they occasionally hear sentences like the following from their parents.

(1) a. No Billy, Kittie doesn’t like __ to be smacked. (Landau, 1999: 10) b. Tom! It’s too late. You know, time __ to go to bed. c. Jane! Please, try __ to pay attention to me whenever I tell you something.

Through the examples in (1), children are usually controlled by their parents in their behaviors. However, apart from the lesson in parental control, to be able to understand such sentences and produce similar ones by themselves, children should also learn a pattern of the innate grammar of their mind/brain from the examples in (1). For instance, in each of the sentences above, an element is missing from the underlined position. Nonetheless, the underlined position of the is considered as if some unexpressed element (i.e., PRO) occupies it. Moreover, the unexpressed element should be interpreted as referentially dependent on another NP in the same sentence for the right communication. This kind of referential inference is by and large dealt with by a pattern of the innate grammar of our mind/brain. Linguists call it a grammar of control, which deals with the process of referential inference. Bresnan (1982) illustrates the definition of a grammar of control as follows:

The term control is used to refer to a relation of referential dependency between an unexpressed subject (the controlled element) and an expressed constituent (the controller). The referential properties of the controlled element . . . are determined by those of the controller. (Bresnan, 1982: 372)

-1- In other words, control is a relation between some antecedent and the understood subject (i.e., PRO) of the infinitive. Therefore, mastering the properties of this relation is the task of acquiring a grammar of control. In the next section, I will examine where PRO appears.

1.2 Distribution of PRO

In this section, to find out where PRO appears, I will look through several approaches of the distribution of PRO. In Principle & Parameter theory, PRO, which is [+pronominal, +anaphoric], is considered to be placed in the ungoverned position since theoretic approach requires that the pronominal anaphor PRO should satisfy both Binding Principle A and Binding Principle B simultaneously.1 However, in the minimalist theory, PRO is argued to be in the (null)-case marked position since not only the concept of government is dispensed with the basic (local) relations, but also the structural case assignment by the head (e.g., INLF, V) government is replaced with a checking relation (i.e., Spec-Head relation). I will provide the detailed accounts below.

1.2.1 Classical Approaches 1.2.1.1 Caseless PRO

To find out the distribution of PRO, we may consider the several questions as follows:

(2) a. Can PRO be found in a direct object position? b. Is it possible to find PRO as the subject of a finite ? c. Does every infinitive have a PRO subject?

1 Chomsky (1981) proposes that PRO is [+pronominal, +anaphoric], subject to both principles B and A of the bindin g theory. BPA requires an anapho r to be bound in its domain (i.e.,governing category), BPB that a be free. Gi ven that a single expression cannot be both bound and free in its domain (i.e., go verning category), this entails that PRO never has a doma in. Hence, PRO can only appear in the ungoverned position.

-2- The following examples in (3)-(5) suggest that the answer to all three questions should be negative.

(3) a.*Billi wondered [CP whether [IP PROi to invite PRO]].

b.*Bill wondered [CP whether [IP he should invite PRO]].

(4) a.*[IP PRO should invite the sergeant].

b.*Billi wondered [CP whether [IP PROi to destroy someone]].

(5) a.*Billi preferred very much [CP for [IP PROi to destroy something]].

b.*Billi believed [IP PROi to be the best detective].

The sentences in (3) show that the non-overt element PRO cannot appear in the object position. The ungrammaticality shown in (3) is due to the presence of PRO in the object position. If we replace PRO with an overt NP, the sentences become grammatical as follows:

(6) a. Billi wondered [CP whether [IP PROi to invite her]].

b. Bill wondered [CP whether [IP he should invite her]].

The sentences in (4) suggest that PRO cannot appear as the subject of the finite clause, whether they (sentences in (4)) are main clauses like (4a) or subordinate ones like (4b). If we replace PRO with an overt NP, the sentences in (4) become grammatical as follows:

(7) a. [IP You should invite the sergeant].

b. Billi wondered [CP whether [IP hei should invite someone].

The sentences in (5) finally provide evidence that although PRO may be the subject of some infinitival clauses, not all infinitival constructions allow PRO as its subject. Based on these facts, Bouchard (1984) generalizes the above three descriptive stipulations as one principle in (8).

-3- (8) PRO must not have case.

According to Bouchard’s principle (8), PROs crucially differ from lexical NPs, which must have case to avoid case filter violation.2 Hence, the ungrammatical examples in (3)- (5) are uniformly accounted for under the principle (8). As pointed out by Lasnik (1992b), however, the account of the distribution of PRO based on (8) has some empirical difficulties as the following examples show:

(9) a.*Jane believed sincerely [herself to be wise]. b.*My belief [Harry to be intelligent]

(10) a.*Jane believed sincerely [PRO to be wise]. b.*My belief [PRO to be wise]

In (9a), due to the lack of adjacency requirement on case assignment suggested by Chomsky (1981), the ECM verb believe cannot assign its accusative case to the subject of the infinitival clause herself, thus the sentence is ungrammatical. Similarly, the subject of the infinitival clause in (9b) is not case-marked due to the inability of nouns to assign structural case across a clausal boundary. Under the principle (8) proposed by Bouchard (1984), if PRO needs to satisfy only (8), the parallel sentences in (10) should be grammatical, contrary to fact. Thus, it appears that there is no obvious way to account for the distribution of PRO solely based on the principle in (8).

1.2.1.2 Ungoverned PRO

As seen in the previous section, the principle in (8) cannot explain the distribution of PRO. Instead, Chomsky (1981) argues that the crucial factor determining the distribution of PRO is government. In particular, the following is offered as a descriptive generalization.

2 Case filter: *[NP], where NP has a phonetic matrix but no Case. (Chomsky 1981: 49)

-4- (11) PRO theorem PRO must be ungoverned.

The requirement in (11) is deduced from the examples below.

(12) a.*Peter preferred very much [CP [IP the police to invite the sergeant]].

b.*[CP[IP Anyone to invite the police man]] would be regrettable.

(13) a. Peter preferred very much [CP for [IP the police to invite the sergeant]].

b. [CP For [IP anyone to invite the pliceman]] would be regrettable.

In (12a), the verb prefer cannot assign its accusative case to the subject of the infinitival clause the police due to the lack of adjacency. Likewise, in (12b), there is no case assignor to case-mark the NP anyone in the subject position of the infinitival clause. Hence, the non-case marked NPs the police and anyone violate case-filter.3 At this juncture, if the preposition for is inserted in C, the ungrammaticality of the sentences in (12) disappears as seen in (13) because the preposition for governs the relevant NPs and assigns accusative case to them. This is subject to the standard assumption that case- marking requires government. On the basis of these facts, the difference in between the sentences in (14) and the ones in (15)-(16) opens a way to derive the PRO theorem in (11).

(14) a. Peteri preferred very much [PROi to invite the sergeant]. b. [PRO to invite the policeman] would be regrettable.

(15) *Jane’s belief [PRO to be intelligent]

(16) *John doesn’t know [CP if [IP PRO to leave]]

The sentences in (14) illustrate that PRO is not only non-case-marked but also un-

3 See note1

-5- governed by any governor (e.g., N, V, P, A and tensed I4). Thus they are deemed to be grammatical. By contrast, the sentences in (15) and (16) indicate that PRO is only governed by the governor belief (i.e., N) or if (i.e., ), respectively. Thus the sentences are ungrammatical under the PRO theorem in (11). In the end, PRO must appear in the ungoverned position.

1.2.2 Case-theoretic Approaches

In the 1990s, the advent of a minimalist program imposes some problem on postulating the PRO theorem based on the notion of government. In the framework of the minimalist program, the basic relations are typically ‘local’. The local relations are presented as follows:

(17) XP ZP X’ X YP Y

In this tree diagram, we can derive three kinds of local relations: head-complement (X- YP), spec-head (ZP-X), and head-head relation (X-Y)5. To keep to relations of these kinds, the minimalist program requires that such notion as government by a head (i.e., head- government) should be dispensed with, and that the structural case licensing should be accomplished not under the government by V or INFL but under the spec-head relation (i.e., checking relation). Hence, the PRO theorem based on the notion of government has

4 According to Chomsky (1981), among INFLs, only finite I(NFL) ca n govern an NP (=subject) in [Spec, IP] position and thus assign nominative Cas e to the NP because of the reason why it is specified for both [+Tense] and [+AGR] features . On the contrary, the non-finite I(NFL) cannot be a governor for the NP in [Spec, IP] since it is negatively specified for both [Tense] and [AGR] features. Thereby, PRO appe ars in the subject position of the infinitival clause under the PRO theorem. 5 Head-head relation in (17) mean s the relation of a verb to (the head of) its complement.

-6- vanished as well. Therefore, in order to account for the distribution of PRO under the minimalism, Chomsky and Lasnik (1993) argue that PRO should be case-marked. In what follows, I will illustrate the problem of postulating the PRO theorem and a new analysis in Chomsky and Lasnik (1993).

1.2.2.1 PRO and Null Case

In order to maintain the PRO theorem in (11), it must be assumed that, in the examples like those in (18), PRO moves from its theta-position to an ungoverned position within the infinitival clause.

(18) a. John wants [TP to be rescued PRO].

b. John wants [TP PRO to be rescued t].

Under the PRO theorem in (11), the PRO in (18) moves from its theta-position to [Spec, TP] position in the infinitival clause to avoid being governed by the verb rescued. However, such an analysis is not compatible with the minimalist framework since the notion of government on which the PRO theorem is based does not exist anymore in the sense that the function of government (e.g., structural case-assignment) is replaced by the checking-relation (i.e., spec-head relation). Hence, PRO in (18) does not have to move elsewhere to avoid being governed. Then why does PRO in (18) do so? In general, it is assumed that the displacement phenomenon is morphologically driven under the minimalist program. Hence, the movement of a takes place to check the formal features of it. Chomsky (1993) calls it Last Resort (Greed).

(19) Last Resort (Greed) Movement of α targeting K is permitted only if the operation is morphologically driven, by the need to check some feature. (Chomsky 1995: 156)

According to the definition in (19), the motivation of the movement of PRO in (18)

-7- targeting [Spec, TP] in the infinitival clause is case licensing of the moved element PRO, which is to be deleted prior to LF. Based on this fact, Chomsky and Lasnik (1993) argue that PRO has null-case. Furthermore, on the basis of the contrast between (20) and (21), they propose that non-finite T can check null-case, distinct from other types of structural case, such as nominative, accusative, and so on.

(20) John tried [TP PRO to kiss Mary].

(21) *John tried [TP Billy to kiss Mary].

As seen in (21), the lexical NP Bill cannot be the subject of the infinitival clause since null-case of non-finite T cannot be checked against the NP Bill. Thus, Chomsky and Lasnik’s (1993) assumptions are listed in (22).

(22) a. Only PRO has null case b. Non-finite T checks null-case

In the end, the assumptions in (22) naturally restrict the possible occurrences of PRO as follows:

(23) a.*Pam believes [TP PRO solve the problem]. b.*Sarah saw PRO. c.*Sarah saw [pictures of PRO].

In (23a), PRO occupies the subject position to be nominative case-marked by the finite T in the embedded clause. In (23b) and (23c), it is in the position to be accusative case- marked by the lexical verb saw and the preposition of, respectively. Hence, the case- mismatch between PRO (null-case) and the functional categories (T, and v, P: nominative, accusatives, respectively) induces the ungrammaticality of the sentences in (23). By contrast, the following sentences are grammatical.

-8- (24) a. Kerry attempted [TP PRO to study linguistics].

b. Kerry persuaded Sarah [TP PRO to study linguistics].

Therefore, the contrast between sentences in (23) and the ones in (24) shows that the occurrence of PRO is limited to the subject position of the non-finite TP.

1.2.2.2 PRO and Enlightened Self Interest

In the previous section, we have looked into the distribution of PRO under the last resort and null case proposed by Chomsky (1993), and Chomsky and Lasnik (1993), respectively. However, the notion of Last Resort is replaced with Enlightened Self Interest in Lasnik (1995) and Chomsky (1995) since there are some cases as in (25) below, in which movement also takes place in order to check some features of the target.

(25) a. Jane wants to seem discreet.

b. Jane wants [TP to seem [SC PRO discreet]]

c. Jane wants [TP PRO to seem [SC t discreet]]

For a coherent argument, let us consider the following sentence first.

(26) a.*Jane wants to seem that he is discreet.

b.*Jane wants [TP PRO to seem [CP that [TP he is discreet]].

In (26), the verb seem is usually considered one-place predicate only with an internal argument. According to Burzio’s generalization (i.e., if a verb has an external argument, it also checks accusative case), the verb seem has neither accusative case nor an external argument. It means that the verb seem cannot have an external argument (subject) in the sentence at all. Otherwise, the sentence including the verb seem would be ungrammatical. Therefore, in order to account for the ungrammaticality of the sentence in (26a), we need to assume that PRO is required in [Spec, TP] like (26b). If so, PRO in (26b) fails to receive an external θ-role from the verb seem; consequently, the sentence in (26) induces

-9- the violation of θ-criterion6 thus becomes ungrammatical. Then, what is the principle that demands the presence of PRO in [Spec, TP]? An obvious candidate is the Extended Projection Principle (hereafter, EPP) in Chomsky (1981): every clause must have a subject. With this discussion in mind, let us go back to (25). In contrast to (26), (25) is perfectly grammatical. According to the EPP, the structure of (25a) is not equivalent to (25b) but (25c). Therefore, the movement of PRO in (25c) to [Spec, TP] is not only for a certain feature checking (i.e., null-case) of PRO itself as reviewed in section 1.2.2.1 but also for the EPP feature checking of the functional head T(to). Based on these facts, Lasnik (1995) and Chomsky (1995) argue that Last Resort in (19) is empirically inadequate, and that the movement of a category α targeting K is legitimate as long as it results in the feature checking of either the moved element α or the target K. Their arguments are defined as follows:

(27) Enlightened Self Interest An item can move to satisfy either (i) its own requirements or (ii) those of the target it moves to.

According to (27), it is said that the movement of PRO in (25) takes place to check the EPP feature of the non-finite T as well as null-case of PRO itself.

1.3 Two Types of Control

The type of control has been widely discussed in the literature: Bresnan (1982), Chomsky (1981, 1986a), Hornstein (1999, 2000), Larson (1991), Lasnik (1992b), Manzini (1983), Martin (1996), Rosenbaum (1967), Williams (1980), and so on. Although each linguist makes a slightly different analysis of control, the type of control is usually divided into two different parts. Williams (1980) initially argues that there are two types of control: obligatory control (hereafter, OC) and non-obligatory control (hereafter,

6 Theta-criterion: Each argument bears one and only one θ-role, and each θ-role is assigned to one and only one argument. (Chomsky 1981: 36)

-10- NOC). According to him, the distinction between OC and NOC is based on (i.e., closest c-commanding) between the antecedent and PRO. Consider the following OC examples.

(28) a. Johni tried [PROi to leave].

b. *Johni’s sisterj promised [PROi to behave himselfi]

c.*Johni thinks that it was expected [PROi to shave himself].

d.*Johni said that Janej tried [PROi+j to score a goal]. e.*It was expected [PRO to shave himself].

In the examples illustrated above, the contrast between (28a) and (28b) indicates that OC PRO must be c-commanded by the antecedent. The sentence (28c) shows that the antecedent (i.e., controller) must be local, and (28d) indicates that OC PRO cannot be controlled by both lexical NPs. Finally, (28e) demonstrates that OC PRO must have only an argument as an antecedent. Since an expletive, by definition, cannot bear a theta-role at all, it cannot satisfy a particular thematic requirement imposed on PRO. Hence, the antecedent of PRO must be an argument. Contrary to OC PRO, however, those properties of OC PRO are not shared by PRO in NOC contexts. The following examples illustrate the distinctive phenomena of NOC PRO.

(29) a. Bill’si mother believes that PROi encouraging himself is necessary for the social life.

b. Johni thinks that it is believed that PROi shaving himself is important.

c. Johni told Maryj [that [PROi+j washing each other] would be fun]]. d. It was believed that PRO shaving was important.

What is indicated in the example (29a) is that the antecedent need not c-command the NOC PRO. In addition, (29b) shows that the antecedent need not be local under the NOC environment, and (29c) indicates that the split antecedents are available in the NOC environments. Finally, (29d) shows that NOC PRO does not require an antecedent. Thus, by comparing the sentences in (28) with the ones in (29), we can generalize the salient properties of OC PRO as follows:

-11- (30) Obligatory Control Convention a. The controller must c-command PRO. b. The controller must be local. c. PRO cannot be controlled by both lexical NPs. d. The controller must be a syntactically realized argument.

Therefore, it is concluded that obligatory control is subject to the convention in (30).

1.4 A Brief Overview of This Thesis

The main aim of this thesis is to eliminate (i) the need for government in accounting for PRO (already explained in section 1.2.2), as well as to find out (ii) how the particular choice of the antecedent of OC PROs (i.e., obligatory subject control and obligatory object control) is determined within the minimalist framework of Chomsky (2000, 2001). As many linguists argue, it is hard to account for control constructions in syntactic analysis. Hence, to find out a new analysis of PRO and control within the current minimalist framework, I will examine the previous syntactic analyses and the problems, as well as the minimalist assumptions in Chomsky (2000, 2001). And then, I will propose a new analysis of PRO and control This thesis is organized as follows: Chapter 1 demonstrates the classification and the distribution of PRO as already reviewed above. In chapter 2, I will introduce the framework of Minimalist Inquiries and Derivation by Phase i.e., Chomsky (2000, 2001), respectively. In chapter 3, I will consider the previous syntactic analyses of OC PROs: Minimal Distance Principle in Rosenbaum (1967), Intrinsic Properties of Predicates based on Control Theory in Chomsky (1981), ATTRACT in Manzini & Roussou (1999) and OC PRO = NP-trace in Hornstein (1999). And I will also point out their problems. In chapter 4, I will propose that the particular choice of the antecedent of OC PROs depends on the syntactic operation Agree between the functional head (T, v) and PRO, which is subject to the Optional EPP Requirement on Control Verbs. That is, when PRO agrees with the head T, which has already agreed with the subject DP, we get obligatory subject control. When PRO agrees with the head v, which has already agreed with the object DP,

-12- we get obligatory object control. Finally, in chapter 5, I will make a brief conclusion as to a new analysis of the interpretation of PRO.

-13- ChapterⅡ

Framework of Minimalist Inquiries and Derivation by

Phase: Chomsky (2000, 2001)

2.1 Introduction

In chapter 2, I will look into the framework of Minimalist Inquiries and Derivation by Phase as follows: In section 2.2, I will summarize the Core Functional Categories (CFCs). And then, based on CFCs, I will derive the Clausal Structure in Chomsky (2000, 2001).

In section 2.3, I will summarize the basic operations of Computational System (CHL), i.e., Merge, Agree, and Move, and introduce the probe-goal system inducing Agree. After that, I will apply these new notions to real linguistic data. Finally, in section 2.4, I will summarize the notion of Phase, Phase Impenetrability Condition and Spell-Out.

2.2 Core Functional Categories and Clausal Structure

Lexical Items (LIs) are divided into two main categories: substantive (e.g., N, V, P, A) and functional (e.g., C, T, v, D). Among them, Chomsky (2000, 2001) is concerned with C, T, and v, which are called ‘Core Functional Categories’ (CFCs), in the sense that the (uninterpretable) φ-features these CFCs may have constitute the core of the systems of (structural) case- and dislocation (i.e., Move).7 According to Chomsky (2000), among CFCs, C expresses ‘force/mood’, T does ‘tense/event structure’, and v(*) does the “light verb” head of transitive constructions8. In addition, due to the semantic (s-) selection of CFCs, they (C, T, v) can select each other.

7 φ-features are obligatory for T, v. 8 Here v means the light verb of unaccusative or passive whereas v* means that of the with full argument (i.e., active)

-14- For example, C is selected by the substantive categories,9 v only by a functional category (i.e., T), and T is selected by C or V. If T is selected by C, it has a full complement of φ- features. However, if T is selected by V, it is defective (Tdef). Usually, C selects T, while T and v select verbal elements. The light verb v may also select a nominal phrase NP/DP as its external argument EA = SPEC-v. Each CFC also allows an extra SPEC beyond its s- selection: for C, a raised wh-phrase; for T, the surface subject; for v, the phrase raised by Object Shift (hereafter, OS).10 Especially, for T, the property having a SPEC is related to the Extended Projection Principle (EPP). The corresponding properties of C and v are EPP-features, which are uninterpretable. Based on the account above, the basic structural properties of CFCs can be illustrated in the configuration as follows:

(1) α = [XP [(EA) H YP]]

In the configuration in (1), H is the CFC, XP is the extra SPEC selected by its EPP feature, and EA is the external argument selected by H = v. Now the configuration in (1) is represented as P(hrase)-markers as follows:

(2) HP XP H’(= HP) EA H’ H YP

9 In addition, C can be unselected (root), while v and T cannot. 10 Usually, English is considered as [-OS] language. However, at l east, in the case of wh- question, English seems to show the property of [+OS] language as follows:

(i) a. Who did you invite?

b. [CP who C-did [TP you[v*P twho [v*P tyou [v*-invite twho]]]]].

As seen in (ib), who undergoes movement twice. The fi rst movement to the outer spec of v*P is caused by the EPP feature of v*. The second one to the spec of CP is induced by the [Q]-feature in C. The former is considered as the property of OS language. Like this, in wh-question, English seems to show the property of [+OS] language.

-15- On the basis of the P-markers in (2), we can now derive a clausal structure. At first, if H is v, then the phrasal structure is as follows:

(3) (vP) (OB) vP(= v’) Sub v’ v VP

V tOB (OS)

Next, if H is C, then the phrasal structure is represented as follows:

(4) CP wh-element C’ C TP Sub T’ T . . .

Finally, bringing (3) and (4) together, we can derive a clausal structure as follows:

(5) CP C’ C TP Sub T’ T vP

tSub v’ v VP V OB/complement

In what follows, I will look into the basic operations of the computational system (CHL) i.e., Merge, Agree, and Move.

-16- 2.3 Operations of the Computational System (CHL)

In Minimalist Inquiries (2000) and Derivation by Phase (2001), Chomsky introduces the alternative operation for feature checking in Chomsky (1993, 1995)11: Agree, one of the basic operations of the computational system deriving a sentence. Before proceeding further, let us consider what the operations of the computational system (CHL) are.

The first operation of the CHL is Merge, which takes two syntactic objects (α, β) and forms K (α, β) from them. The second one is Agree that establishes a relation (agreement, case-checking) between an LI α and a feature F in some restricted search space (i.e., its domain). Unlike Merge, this operation causes the uninterpretable features of LI α to be deleted in the course of derivation since the derivation otherwise would be canceled. Finally, the third operation is Move, combining Merge and Agree. Chomsky (2001) assumes that Move establishes ‘agreement’ between α and F and merges P(F) to αP, where P(F) is a phrase determined by F and αP is a projection headed by α. Thus, Chomsky (2001: 10) claims the definition of Move as follows:

(6) The combination of Agree/pied-piping/Merge is the composite operation Move, preempted where possible by the simpler operations Merge and Agree.

As seen in (6), Merge is over Move (MoM); the former is one-operation, but the latter is the composite operation in the sense that it involves an extra step determining P(F) (i.e., “pied-piping”). Furthermore, to account for the syntactic operation Agree, Chomsky (2000) proposes the notion of Probe and Goal. Matching of probe with goal induces Agree, eliminating uninterpretable features that activate them.12 Chomsky (2001: 6) further proposes the condition on Agree as follows:

11 Basically, Chomsky (1993, 1995) assumes that feature-checking t o eliminate the un- interpretable features is accompli shed in the structural configu rations: spec-head relation and head-head relation, respectively. See section 1.2.2. 12 Chomsky (2000, 2001) assumes the following three uninterpretable features. (i) φ- set of T (ii) EPP-feature of T (iii) structural case of nominal

-17- (7) a. Goal as well as probe must both be active for Agree to apply. b. α must have a complete set of φ-features (it must be φ-complete) to delete uninterpretable features of the paired matching element β.

In (7b), matching means a relation that holds of a probe P and a goal G. However, not every matching pair induces Agree. Hence, to sort out the possibility of matching of probe P and goal G, Chomsky (2000: 122) argues that G be in the domain D(P) of P and satisfy locality conditions. The simplest assumptions for the probe-goal system Chomsky proposes are (8):

(8) a. Matching is feature identity b. D(P) is the sister of P c. locality reduces to “closest c-command”

In (8), D(P) is the c-command domain of P, and a matching feature G is closest to P only if there is no G’ in D(P) matching P.

On the basis of the discussion so far, now I will illustrate the process of Move as follows:

(9) αP NP α’ ①Agree α XP ③Move P(F) X’ [φ-F] X YP ②pied-piping

In (9), at first, the φ-feature set of α, acting as a probe, seeks the closest matching goal, φ- feature set of F in its domain (i.e., X and spec of XP) and establishes agreement between LI α and a feature F, deleting uninterpretable features of both probe P and goal G. This is Agree. Next, for “pied-piping”, an uninterpretable feature F determines P(F) as its candidate for the raising to the spec of αP. Finally, P(F) is merged to the spec of αP to

-18- satisfy the EPP feature of the head α. Therefore, the composite operation Move is Agree, pied-piping and Merge.13 Now we apply these new notions to real linguistic data below.

(10) a. John seems to be honest. b. LA = { be, honest, to, John, seems, T, v} c. . . . [ T(to)- be honest]

d. . . . [TP John T(to)-be honest].

e. John T(probe) [vP seems [TP tJohn(goal) to be honest]]. ①Agree +②pied-piping ③Merge

The sentence in (10a) has the Lexical Array in (10b) at the beginning of the derivation. At some stage of the derivation, the sentence (10a) has the structure in (10c). At this juncture, the DP John is purely merged to [Spec,TP] to satisfy the EPP-feature of T as (10d) shows. As seen in (10e), the already formed single syntactic object TP merges with seems and v, forming vP. After that, Tmatrix is merged to vP. At this point, by the definition of probe-goal system, the probe, φ-feature set of T, locates the goal, φ-feature set of the DP John, thus agreement between them takes place. By the operation (Agree), the uninterpetable feature (case) of the DP John determines the DP John as its candidate for the raising to [Spec,

TP]. And then, the DP John is merged to the spec of TPmatrix to satisfy the EPP feature of

Tmatrix (see (10e)). As a result, the derivation of the sentence in (10) is grammatical since there is no uninterpretable feature at LF level.

13 Lee (1999) objects to such assumption and claims that Agree and Move should be regarded as separate operations; the task of Agree is to erase u ninterpretable features of both probe and goal, and that of Move is to satisfy the EPP-feat ures. I will adopt his claim that Move and Agree are separate operations. See section 4.3.1.

-19- 2.4 The Notion of Phase, PIC and Spell-Out

Chomsky (2000, 2001) basically assumes that L(anguage) is the recursive definition of a set of expressions EXP = . Hence, Chomsky construes L as a step-by-step procedure for constructing expressions EXPs. This kind of assumption leads to adopt a derivational approach to L. So, Chomsky (2000: 101) proposes that a language L should follow the procedure (i, ii) of (11) to specify the language, then apply (iii, iv) to derive a particular EXP.

(11) (i) Select [F] from the universal features set F (ii) Select Lexicon (LEX), assembling features from [F] (iii) Select Lexical Array (LA) from LEX (iv) Map LA to EXP, with no recourse to [F] for narrow syntax

Universal Grammar usually makes available a set F of features and operations (CHL) that access F to generate expressions. And the language L maps F to a particular set of expressions EXP. If L makes a one-time selection of a subset [F] of F, dispensing with further access to F, then operative complexity is reduced. It is further reduced if L includes one-time operation that assembles elements of [F] into a lexicon LEX, with no new assembly as computation proceeds. However, unlike this, Chomsky (1993, 1995) argues that the derivations access the lexicon to generate a particular EXP, at every point. This proposal means that the derivations must carry along the huge operative complexity. By that, in order to simplify the operative complexity (i.e., computational system), Chomsky (2000, 2001) proposes that derivations make one time selection of a lexical array LA from LEX, then map LA to expressions, dispensing with further access to operative complexity. This means that derivations should proceed phase by phase. To find out how derivations proceed phase by phase, we need pay attention to what Chomsky says as follows: Chomsky (2000: 106) assumes, “[W]e select LA as before, under (iii) of (40): the computation need no longer access the lexicon. Suppose further that at each stage of the derivation, a subset LAi is extracted from Lexicon, and submitted to the procedure L.

When LAi is exhausted, the computation may proceed if possible. Or it may return to LA

-20- and extract LAj, proceeding as before. The process continues until it (LA) terminates. If so, operative complexity in some natural sense is reduced, with each stage of the derivation accessing only part of LA.” What Chomsky illustrates above can be represented as follows:

(12) Lexicon L(anguage) LA

LAj(subset) LAi(subset)

② ① CHL

Phase2 Phase1 Spell-Out Derivations

As seen in (12), the subset (or subarray) LAi or LAj is called a phase (PH), which refers to a verbal phrase with full argument structure (i.e., v*P) and a full clause including tense and force (i.e., CP). Furthermore, Chomsky (2000: 107) argues that phases should satisfy the cyclicity condition to reduce operative complexity as follows:

(13) The head of a phase is “inert” after the phase is complete, triggering no further operations.14

According to the cyclicity condition in (13), the head of a phase cannot trigger Merge or Attract in a later phase. Chomsky calls it “Phase Impenetrability Condition” (hereafter, PIC) as follows:

(14) In phase α with head H, the domain of H is not accessible to operations α, but

only H and its edge. [ZP Z . . .[HP α [H YP]]] (Chomsky 2000:108)

14 See Chomsky (2000:107)

-21- The PIC in (14) can be represented as the following P-markers.

(15) ZP (= Phase2) Z’ Z . . .

HP (= Phase1) Agree(O) α H’ Spell-Out H YP Agree(X)

According to the PIC in (14), in (15), except for YP (= complement of H), only head H and its edge α are accessible to Z since Spell-Out applies at the phase level.15 That is, in (15), when a later phase (e.g., ZP) is completed, the complement of the lower phase (e.g., HP), YP is by definition sent to each of the PF and LF components for interpretation. Therefore, YP is spelled out at the lower phase HP. It means that agreement between the probe (i.e., Z) and the goal (i.e., YP) cannot take place over a phase.

15 Spell-Out is an operation that applies to the structure ∑ (i.e. , a set of phrase markers) already formed, reducing ∑ to ∑ <-Pf> (phonetic features stripped away from ∑) and sending ∑ to PHON. See Chomsky (1995: 229).

-22- Chapter Ⅲ

Previous Analyses of OC PRO

3.1 Introduction

The main goal of this chapter is to introduce the previous analyses of the obligatory control constructions in English. In this respect, I will consider the previous analyses associated with PRO. Those are as follows: (i) Minimal Distance Principle (hereafter, MDP) in Rosenbaum (1967); (ii) Intrinsic Properties of Predicates based on Control Theory in Chomsky (1981); (iii) ATTRACT in Manizi & Roussou (1999); (iv) OC PRO = NP-trace in Hornstein (1999). In reviewing these analyses, I will focus on how the obligatory (subject/object) control constructions are derived in the syntactic analysis, and discuss the problems of those analyses. This chapter is organized as follows. Section 3.2 is largely concerned with the standard approaches such as MDP and Intrinsic Properties of Predicates based on Control Theory within and GB theory, respectively. Section 3.3 is pertained to the minimalist approaches such as ATTRACT analysis and OC PRO = NP- trace analysis. According to those minimalist approaches, PRO doesn’t exist in the . Finally, section 3.4 demonstrates the problems of Hornstein’s (1999) analysis. Although all of these syntactic analyses are attractive, they have some difficulties in explaining how the particular choice of the antecedent of the OC PRO is determined and why the passive tense of the subject control verb promise contains ungrammaticality.

3.2 Standard Approaches

This section is subdivided into two parts as follows. In subsection 3.2.1, I will focus on Minimal Distance Principle in Rosenbaum (1967). According to the MDP, the closest c-commanding referential NP can be the controller of the OC PRO. Next, in subsection

-23- 3.2.2, I will show another approach, Intrinsic Properties of Predicates based on Control Theory in Chomsky (1981). Based on Chomsky (1980), Chomsky (1981) proposes that the crucial factor drawing a distinction between obligatory subject control and obligatory object control be the intrinsic properties of the main predicates.

3.2.1 Rosenbaum’s (1967) Minimal Distance Principle

Rosenbaum (1967) observes that typically, if a verb with an infinitival complement alternates between intransitive and transitive forms, obligatory control shifts from subject to object. Let us consider the examples below.

(1) a. Johni got to shave himselfi/*oneself.

b. Johni got [PROi to shave himselfi/*oneself].

(2) a. Johni got Fredj to shave himselfj/*i/*oneself.

b. Johni got Fredj [PROj to shave himselfi/*i/oneself].

The sentences in (1) and (2) illustrates that the covert syntactic elements PROs are obligatorily controlled by the antecedents. For instance, PRO in (1b) is controlled by the subject DP John, whereas PRO in (2b) is controlled by the object DP Fred. Like this, the contrast between (1b) and (2b) implies that the motivation of control shift from subject to object is alternation between intransitive and transitive verbs. However, it is not always true since there is a counterexample below.

(3) a. Johni persuaded Fredj to shave himselfj/*i/*oneself.

b. Johni persuaded Fredj [PROj to shave himselfj/*i/*oneself].

In (3), the matrix verb persuade doesn’t show alternation between intransitive and transitive verbs like in (1) and (2). Nonetheless, as shown in (3b), the verbs like persuade that occurs with a direct object and a control infinitival require the direct object (Fred) to be the controller of the OC PRO. Thus, the covert syntactic element PRO in (3) is obligatorily controlled by the object Fred. Based on these facts, Rosenbaum proposes the

-24- Minimal Distance Principle below.

(4) Minimal Distance Principle PRO is controlled by the closest c-commanding referential DP/NP.

The MDP in (4) means that if there is no direct object, PRO in the subject position of the infinitival clause is obligatorily controlled by the closest DP, the subject, and that if there is a direct object, PRO in the same position is obligatorily controlled by the closer DP, the object rather than the subject. On the basis of the MDP in (4), Rosenbaum (1967) accounts for control shift between (1) and (2). However, there is also a counterexample to Rosenbaum’s (1967) analysis based on the MDP in (4). It is the verb promise that exhibits the following pattern.

(5) a. Johni promised [PROi to shave himselfi/*oneself].

b. Johni promised Fredj [PROi to shave himselfi/*j].

In (5a), there is no direct object. Thus the closest c-commanding referential NP, the subject John obligatorily controls PRO in the infinitival clause under the MDP. Unlike (5a), a direct object Fred is added in the sentence (5b). Therefore, on a par with (5a) observing the MDP in (4), PRO in (5b) should be controlled by the closer c-commanding NP, the object Fred rather than the subject John. However, PRO in (5b) is obligatorily controlled by the subject John, contrary to what we expect. This indicates that Rosenbaum’s (1967) analysis based on the MDP in (4) cannot explain how the particular choice of the antecedent of the OC PRO is determined in the syntactic analysis.

3.2.2 Chomsky’s (1981) Intrinsic Properties of Predicates

Chomsky (1981) proposes the several modules of grammar to account for the distribution and the interpretation of the NP constituents in the sentence: X’-theory, θ- theory, case theory, movement theory, binding and control theory. Among them, Chomsky (1981) argues that control theory is mainly concerned with the interpretation of PRO. To make a distinction between obligatory subject control and obligatory object control, he

-25- proposes that the distinction between obligatory subject control and obligatory object control is dependent upon the intrinsic properties of the matrix verbs. Consider the examples below.

(6) a. Peteri decided finally [[PROi to go on hisi/*one’s own]].

b. Peteri tried [[PROi to love himselfi/*herself]].

c. Peteri promised Maryj [[PROi to buy a new watch]].

(7) a. Peteri ordered Miss Kimj [[PROj to go to market to buy some milk]].

b. Peteri allowed Miss Kimj [[PROj to go on her/*his/*one’s own]].

c. Peteri persuaded Miss Kimj [[PROj to go on her/*his/*one’s own]].

The sentences in (6) illustrate the obligatory subject control constructions, whereas the ones in (7) demonstrate the obligatory object control constructions. The sentences in (7) are all subject to the MDP in (4) since the closest c-commanding referential NP (i.e., Miss Kim) obligatorily controls PRO. The sentences in (6a) and (6b) are also subject to the MDP in (4). In contrast to them, the sentence (6c) is not subject to the MDP because PRO in (6c) is obligatorily controlled by the further distant NP, the subject DP Peter rather than the object DP Mary. To account for this, Chomsky (1981) argues that the intrinsic properties of the matrix verbs should be the essential elements in making a distinction between obligatory subject control and obligatory object control. For instance, the subject control verbs like try, decide, prefer and promise are assumed to impose obligatory subject control on PRO. By contrast, the object control verbs like tell, force, allow, order and persuade are assumed to impose obligatory object control on PRO. These assumptions follow from Chomsky (1980). Chomsky (1980) argues that the subject control verbs like promise, try, etc. be marked in the lexicon with the feature [+SC], which indicates “assigns subject control”, and that the object control verbs like persuade, allow, etc. be marked in the lexicon with the feature [-SC/+CC],16 which indicates “assigns complement (object) control”. Hence, the distinction between obligatory subject control and obligatory object control is deemed to be accounted for under Chomsky’s

16 [+CC] means an abbreviation of complement (object) control.

-26- (1980, 1981) analysis. However, Chomsky’s (1981) analysis is largely based on the semantic analysis. Therefore, to be compatible with the current minimalist theory of Chomsky (2000, 2001), the obligatory control constructions should be analyzed in the syntactic analysis. In section 3.3, other syntactic analyses of the OC PRO will be illustrated.

3.3 Minimalist Approaches

This section is composed of two subsections as follows. In subsection 3.3.1, I will illustrate ATTRACT in Manzini & Roussou (1999). In subsection 3.3.2, I will show that OC PRO is the residue of NP-movement as proposed by Hornstein (1999).

3.3.1 Manzini & Roussou’s (1999) Analysis

As one of the minimalist accounts of the OC PRO, Manzini & Roussou (1999) regard control as ATTRACT. ATTRACT means an abstract relation between a single DP and two predicates (normally, the matrix one and the embedded one). The relation is labeled ATTRACT. However, ATTRACT is not considered as movement since Manzini & Roussou (1999) basically regard only copy-theory of movement in Chomsky (1993) as movement. That is, the only candidate for the movement operation is not a feature but a category. To improve their theoretical adequacy, Manzini & Roussou (1999) propose the several assumptions below.

(8) a. PRO does not exist due to the lack of D-structure.17 b. θ-roles are features on verb to be checked. c. A surface subject is directly merged to the surface position: [Spec, IP]. d. A DP attracts more than two θ-roles of the predicates. e. ATTRACT observes Socopal Minimal Link Condition.

17 D-structure disappears within the minimalist program due to the lack of the independent conceptual support. For a detailed account, see Chomsky (1995: 188).

-27- (9) Scopal Minimal Link Condition (Scopal MLC)

Feature F attracts feature FA only down to the next F’ that also attract FA.

=> F F’ FA

The Socopal MLC in (9) is based on the Minimal Link Condition, requiring the controller to be the closest available DP. It means that the closest DP attracts more than two θ-roles of the predicates. Let us examine the following examples.

(10) a. John persuaded Mary to drink.

b. [+D] θ2 [+D] θ1

(10a) illustrates obligatory object control in the active tense. (10b) demonstrates the representation of ATTRACT operation of (10a). There are two arguments (John, Mary) and two predicates (persuaded, drink) in (10a). Therefore, by assumption (8d), the subject

DP Mary in the embedded clause attracts the θ1-role of the verb drink, and the matrix subject DP John attracts the θ2-role of the verb persuade, respectively. However, by the

Socpal MLC in (9), the matrix subject DP John attracts the θ1-role of the verb drink only down to the next DP Mary that also attracts it. Hence, the object DP Mary becomes a controller, thus we get obligatory object control from (10). In the same way, the passive tense of the obligatory object control construction is represented as follows:

(11) a. John was persuaded to eat.

b. [+Df] θ2 θ1

In (11), the only candidate for an attractor is the matrix subject DP John. By assumption

(8d), the subject DP John attracts θ2 and θ1 at the same time (see (11b)), thus we get obligatory object control from (11). However, Manzini & Roussou’s (1999) analysis cannot account for the following examples.

-28- (12) a. *Johni was promised tJohn [PROi to leave]. b. *John was promised to leave

[+Df] θ2 θ1

(12) demonstrates an obligatory subject control construction in the passive tense. In (12b), by assumption (8d), the only DP John attracts all theta-roles (θ2, θ1) of the predicates (i.e., promised, leave), and the derivation (12b) observes the Scopal MLC in (9). Hence, Manzini & Roussou’s (1999) analysis incorrectly predicts (12) to be grammatical. Next, consider another counterexample below.

(13) a. Johni promised Mary [PROi to return home early]. b. John promised Mary to return home early.

[+Df] θ2 [+Df] θ1

(13a) demonstrates an obligatory subject control construction in the active tense. (13b) illustrates the representation of the ATTRACT operation of (13a). As shown in (13b), the matrix subject DP John attracts the θ2-role of the verb promise, and the object DP Mary attracts the θ1-role of the verb return, respectively. This derivation (13b) also observes the Scopal MLC in (9). If so, the object DP Mary should become the controller of PRO, thus (13a) should be obligatory object control, contrary to fact. Like this, Manzini & Roussou (1999) have some difficulties in accounting for those cases such as (12) and (13). In what follows, I will consider another minimalist approach to OC PROs.

3.3.2 Hornstein’s (1999) Analysis

At first, let us consider the sentence below.

(14) John tried to leave early.

Since the traditional approach in GB theory and Minimalist Program assumes that the

-29- infinitival has a covert syntactic subject PRO that receives its θ-role by ordinary principles, the structure of the sentence in (14) looks something like (15).

(15) [TP Johni T-[ vP v-tried [TP PROi [T’ to [vP v-leave early]]]]].

In (15), the external θ-role of the verb leave is assigned to the covert syntactic subject PRO. However, Hornstein (1999), departing from this standard approach, offers a minimalist theory of syntactic control by reducing certain cases of control to movement. That is, Horstein (1999) proposes that OC PRO should be the residue of NP movement. His proposal is based on Wanna Contraction below.

(16) a. I’m going to stay.

b. I’m going [TP tI to stay]. c. I’m gonna stay.

(17) a. Who do you want to stay?

b. Who do you want [TP twho to stay]. c.*Who do you wanna stay?

In (16), contraction between the matrix verb (going) and the functional head T (to) is possible as shown in (16c), whereas it is not in (17). The explanation for the contrast between (16) and (17) is originally based on Jaeggli’s (1980) argument: only case-marked trace blocks contraction between the matrix verb and the functional head T. For instance, in (16), the pronoun I cannot check its case in the embedded TP since the head T is 18 defective. Instead, the pronoun I raises to the spec of TPmatrix to check its case.

According to Jaeggli (1980), the NP trace (tI) in (16b) is not case-marked, thus it cannot block contraction between going and to as shown in (16c). By contrast, in (17), the wh- element who in [Spec, TP] is (accusative) case-marked by the ECM verb want, and it raises to the spec of CPmatrix to check [Q]-feature in C. Since the wh-trace (twho) is (accusative) case-marked in [Spec, TP] as ECM construction does, it blocks contraction

18 See note 5

-30- between want and to as shown in (17c). On the basis of Jaeggli (1980), let us consider the wanna-contraction of the obligatory subject control example below.

(18) a. I want to buy a car.

b. Ii want [TP PROi [T’ [vP tPRO v-buy a car]]]. c. I wanna buy a car.

According to the case-theoretic approach to the distribution of PRO proposed by Chomsky and Lasnik (1993) in section 1.2.2.1, PRO in (18b) is (null) case-marked by the head T of control infinitival. If so, in (18b), the (null) case-marked PRO in [Spec, TP] must block contraction between want and to under Jaeggli’s (1980) analysis. However, wanna-contraction takes place as shown in (18c). This means that Chomsky and Lasnik’s (1993) analysis cannot explain why wanna-contraction in (18c) is allowed under Jaeggli (1980). To resolve this problem, Hornstein (1999) proposes to reduce obligatory control to the movement phenomenon based on the several assumptions below.

(19) a. OC PRO is the residue of movement: (OC PRO = NP-trace). b. θ-roles are features on verbs. c. A NP/DP “receives” a θ-role by checking a θ -feature of a verbal/predicative phrase that it merges with. d. There is no upper bound on the number of θ-role a chain can have. (Hornstein 1999:78)

Based on (19), the operations Merge and Move build up the derivation of obligatory subject control in (18) as follows:

(20) a. Ii want [TP PROi [T’ to [vP tPRO [v-buy a car]]]]. b. Array = {I, want, to, buy, a, car}

c...... [vP I [v-buy a car]]

+θexternal of buy

d...... [TP I [T [vP tI [v-buy a car]]]]

+EPP

-31- e...... want [TP I [T [vP tI [v-buy a car]]]]

f...... [vP I [v-want [TP tI to [vP tI [v-buy a car]]]]]

+θexternal of want

g. [TP I [T’[vP tI [v-want [TP tI to [vP tI [v-buy a car]]]]]]].

+EPP /+Case h. I wanna buy a car.

(20a) is the representation of the obligatory subject control within the minimalist framework. (20b) is the relevant array including the lexical items plus assorted functional elements (not listed). The representations in (20c-g) illustrate the deriving process of obligatory subject control (20a) within Hornstein’s (1999) analysis. The ‘embedded clause’ is constructed first. As seen in (20c), the verb buy and the NP I from the array (21b) merge together to form vP, thus by assumption (19c), the external θ-role (i.e., [+θ-f]) of the predicate (buy) is transferred to the pronoun I. After that, the functional head T merges with vP to form TP (i.e., infinitival clause), and the NP I raises to the spec of TP to satisfy the EPP of T (see (20d)). At this point, since the NP I in [Spec, TP] of the infinitival clause is not case-marked, wanna-contraction can take place (see (20h)). Then, as shown in (20e), the matrix verb want merges with the infinitival clause TP, and the NP I raises to the spec of vP to check the external θ-role of want (see (20f)).

After that, Tmatrix is merged to vP, and the NP I raises to the spec of TPmatrix to check the

EPP feature and case of Tmatrix (see (20g)). As shown in the whole derivation (20g), the initial position (i.e., [Spec,vP]) of the OC PRO in (20a) is the same as the initial NP-trace of the moved NP, the subject I in (20g). Thus, (20) exhibits obligatory subject control. Like this, Hornstein proposes that when PRO is an NP trace, wanna-contraction of (18c) can be explained under Jaeggli’s (1980) analysis. Then how can the obligatory object control construction be analyzed in the syntactic analysis under Hornstein’s (1999) analysis? Let us consider the example below.

(21) a. Johni persuaded Maryj [TP PROj [T’ to [vP tPRO [v-leave early]]]]. b. Array = {John, persuaded, Mary, to, leave, early}

c...... [vP Mary [v-leave early]]

+θexternal of leave

-32- d...... [TP Mary [T [vP tMary [v-leave early]]]]

+EPP

e...... persuaded , [TP Mary [T [vP tMary [v-leave early]]]]

f...... [VP persuaded+Mary [TP tMary [T [vP tMary [v-leave early]]]]]

+ θinternal of persuaded

g...... [vP John [v-persuaded Mary [TP tMary [T [vP tMary [v-leave early]]]]]]

+θexternal of persuaded

h. [TP John [T’[vP tJohn [v-persuaded Mary [TP tMary [T [vP tMary [v-leave early]]]]]]]].

+EPP /+Case of T

In (21), the ‘infinitival clause’ is constructed first. The verb leave and the light verb v from the array (21b) merge together to form vP. To check the external θ-role of leave, an NP Mary from the array (21b) is purely merged to the spec of vP (see (21c)). Next, T merges with vP, and the NP Mary raises to [Spec, TP] to check the EPP-feature of T (see (21d)). After that, as show in (21e), the verb persuaded is selected from the array (21b). Next, by assumption (19c), the NP Mary merges with the verb persuaded to check the internal θ-role of the verb persuaded, forming the VP (see (21f)). And then, the light verb v is merged to the VP, and the other NP John from the array (21b) is purely merged to

[Spec, vP] matrix to check the external θ-role of the verb persuaded (see (21g)). Tmatrix merges with vP, and the subject NP John raises to [Spec, TP] matrix to check the EPP feature and (nominative) case of Tmatrix (see (21h)). As seen in the whole derivation (21h), the initial position (i.e., [Spec, vP]) of the OC PRO in (21a) is the same as the initial NP- trace of the moved NP, the object Mary in (21h). Hence, (21) exhibits obligatory object control.

Finally, let us consider an obligatory control construction in the passive tense below.

(22) a. Johni was persuaded [PROi to leave here].

-33- b. John was persuaded tJohn [tJohn to leave here]. c. TP John T’

Tfinite vP

+Case/+EPP v VP

was tJohn V’

+θinternal V TP

persuaded tJohn T’

T(to) vP

selected leave here Array = { . . . persuaded...... }

(22a) is the representation of obligatory object control in the passive tense within the current minimalist theory, and (22b) is the corresponding one within Hornstein’s (1999) analysis. The derivation of the infinitival clause in (22b) is the same as that of (21) (thus, it is not illustrated here). After the infinitival TP is formed, as seen in (22c), the matrix verb persuaded is selected from the array, and merged to the infinitival TP. After that, the object NP John raises to [Spec, VP] to check the internal θ-role of the verb persuaded as shown in (22c). Then, the light verb v and the finite T are merged to the VP, sequentially. Next, the object NP John raises to [Spec, TP] to check the EPP feature and case of the finite T. Thus, (22) exhibits obligatory object control in the passive tense. So far, I have considered Hornstein’s (1999) analysis of the OC PRO: PRO = NP trace. Like what he asserts, if PRO is replaced by a non-case marked NP-trace, then we will have explanatory adequacy as follows: (i) wanna-contraction in (18c) can be accounted for under Jaeggli’s (1980) analysis; (ii) obligatory control constructions like (20-22) are the results of A-movement (e.g., the DP I and John ). Especially, the latter means that control theory can be analyzed into movement theory, thus the independent control theory exists no more in the generative grammar. However, Hornsetin’s (1999) analysis has several problems. In what follows, I will consider them.

-34- 3.3.3 Problems 3.3.3.1 Wanna-Contraction

In the previous section, I have considered the reason why Hornstein (1999) regards OC PRO as NP-trace. His argument is based on the possibility of wanna-contraction between the matrix verb and the functional head T. However, there is a different account of wanna-contraction in Bošković (1995). In contrast to Jaeggli’s (1980) analysis, Bošković (1995) argues that the case-marked wh-trace should not be the obstacle for contraction between the matrix verb and the functional head T since wh-traces are invisible at PF. Let us consider (23).

(23) a. Whati do you think’s ti happening there tomorrow?

b. Whati do you think [TP ti is happening there tomorrow]?

c. Whati do you think’s ti been happening there tomorrow?

d. Whati do you think [TP ti has been happening there tomorrow]?

The examples in (23a) and (23c) are the representations of contraction between the matrix verb and the head T in the sentences (23b) and (23d), respectively. As shown in (23b) and (23d), the wh-phrase what is (nominative) case-marked by the head, finite T in the embedded clause, and it raises to the spec of CP to check [Q]-feature of C in the matrix clause. Hence, contraction in (23a) and (23c) should be impossible under Jaeggli’s (1980) analysis. However, the (nominative) case-marked wh-traces in (23b) and (23d) do not block contraction in (23a) and (23c), contrary to Jaeggli’s (1980) analysis.19 Then how can the impossibility of wanna-contraction in (17) be accounted for? Bošković (1995) argues that a null complementizer C serves as an obstacle for PF contraction. The sentence (17) is repeated here as (24).

19 The argument that the finite forms, is in (23a) and has in (23c), are contracted onto the matrix verb think is called encliticization. For a detailed account, see Bošković (1995).

-35- (24) a. Who do you want to stay? b. *Who do you wanna stay?

c. *Who do you want [CP [ C [TP twho to stay]]]?

blocking wanna-contraction at PF

According to Bošković’s (1995) IP analysis,20 the structure of (24a) is (24c). Hence, in (24b), the reason why contraction between the matrix verb (want) and the head (to) is impossible is that a null complementizer C serves as an obstacle for PF contraction (see(24c)). In the end, contrary to Jaeggli’s (1980) analysis, the case-marked trace is not an obstacle for PF contraction any more. If so, the wanna-contraction of obligatory subject control in (18) is not supporting evidence for Hornsetin’s (1999) analysis any longer. Now, let us consider how wanna-contraction of obligatory subject control in (18) can be explained. The sentence (18) is repeated here as (25).

(25) a. Ii want [PROi to buy a car].

b. Ii want [TP PROi [T-to [vP tPRO v-buy a car]]]. +null-case/+EPP c. I wanna buy a car.

Bošković (1995, 1997) assumes that PRO has null-case proposed by Chomsky and Lasnik (1993), and that control infinitivals are TPs.21 The first assumption implies that PRO moves to [Spec, TP] to check its null case as well as the EPP feature of T. The second assumption means that there is no null complemetizer C between the matrix verb (want) and the head (to) in (25b), thus wanna-contraction is allowed as shown in (25c). Therefore, without Hornstein’s assumption (PRO = NP-trace), the wanna-contraction of obligatory subject control (18) can be explained in the alternative analysis by Bošković.

20 Bo šković (1995) assumes that the structure of that-less non-finite declarative complement in (24) is CP but that that of that-less finite declarative one is TP as shown in (23b) and (23d). Hence, the former has a null complementizer C b ut the latter does not. This is called IP (TP) analysis of that-less finite declarative complement. For a similar analysis, see Radford (1997: 147). 21 This argument will be discussed in section 4.3.2.

-36- 3.3.3.2 PRO ≠ NP-trace

In this section, I will examine empirical evidence against Hornstein’s (1999) analysis. Let us consider the following example (e.g., obligatory control construction in the passive tense) below.

(26) a.*Johni was promised [PROi to leave]. b. Array= { John, was, promised, to, leave}

c...... [vP tJohn [v-leave]]

+θexternal of leave

d...... [TP John [T’ to [vP tJohn [v-leave]]]]

+EPP of T

e...... promised, [TP John [T’ to [vP tJohn [v-leave]]]]

f...... [VP promised John [TP tJohn [T’ to [vP tJohn [v-leave]]]]]

+θinternal of promised

g...... [vP [v-was [VP promised John [TP tJohn [T’ to [vP tJohn [v-leave]]]]]]

h. *[TP John [T’ [vP[v-was [VP promised tJohn [TP tJohn [T’ to [vP tJohn [v-leave]]]]]]]].

+Case/+EPP of T

(26a) is the passive tense of the obligatory control construction, and (26b) is an lexical array of (26a). The representations in (26c-h) demonstrate the deriving process of (26a) within Hornstein’s (1999) analysis. By assumption (19b,c,d), the NP John is purely merged to [Spec, vP] in the infinitival clause, and checks the external θ-role of the verb leave (see (26c)). After that, the NP John raises to the spec of TP in the infinitival clause to check the EPP feature of T (see (26d)). As shown in (26e), the matrix verb promised is selected from the array (26b), and the NP John merges with the verb promised to check the internal θ-role of the promised, forming VP (see (26f)). Next, the light verb v (be) is merged to the VP (see (26g)). Finally, the finite T is merged to the vP, and the NP John raises to [Spec, TP] matrix to check the EPP feature of the finite T (see (26h)). During the whole derivation (26h), the movement of the DP John does not violate any economy principle. Hence, (26) is incorrectly predicted to be grammatical. If the argument by Hornstein (1999) that OC PRO be an NP-trace is correct, then the sentence (26) should be

-37- grammatical contrary to fact. Therefore, the claim by Hornstein (1999) that OC PRO is the residue of the moved NP is not reasonable.

-38- Chapter Ⅳ

A Minimalist Approach to the Interpretation of PRO

4.1 Introduction

In this chapter, I will discuss how the obligatory control constructions are analyzed in the current minimalist theory of Chomsky (2000, 2001). Chomsky assumes that the operation Move is a composite operation of Agree (α, β), generalized pied-piping that determines the candidate of the phrase P(β) to be merged, and Merge (P(β), αP), where P(β) is merged as the of α. However, I will follow Lee’s (1999) argument that Move and Agree are separate operations, thus Move is allowed to check the EPP feature of a head (T, v). In addition, to avoid the PIC effect, mentioned in the current minimalist theory of Chomsky, I will also follow Bošković’s (1997) argument that control infinitives are TPs. In this framework, I will explain how the particular choice of the antecedent of the OC PRO is determined as well as why the passive tense of the subject control verb promise contains ungrammaticality within the minimalist assumptions of Chomsky (2000, 2001).

4.2 A Minimalist Analysis based on Agree: Chomsky (2000, 2001)

According to Chomsky (2000, 2001), a sentence is derived phase by phase. That is, spell-out is cyclic at the phase level. Hence, the obligatory control constructions have the following problem under the framework of Chomsky (2000) and Chomsky (2001). Consider the following sentence.

(1) a. Johni tried [CP [TP PROi to go home early]].

b. John [v*P/strongphase[tJohn v*-tried[CP/strongphase[TP PRO to [vP tPRO go home early]]]]].

The derivation of the sentence (1a) is as follows. In (1b), the non-overt NP PRO is base-

-39- generated in the argument structure (i.e.,[Spec, vP]) of the embedded clause. After agreeing with the probe, φ-set of control T,22 PRO raises to the spec of TP to satisfy the EPP feature of control T. And then C merges with TP, forming a phase CP. At this point, we have a structure like (2).

(2) CP (= PH1) C’ C TP PRO T’

T(control) vP

to tPRO go home early

Next, the matrix verb try is merged to CP and forms VP. And then, the light verb v* merges with VP and the subject DP John is merged to [Spec, v*P], forming another phase v*P. We have the intermediate structure (3).

(3) v*P (= PH2) John v*’ v* VP

V CP (= PH1) try - - -

Finally, finite T is merged to v*P, and the subject DP John raises to the spec of TP to satisfy the EPP feature of the finite T after agreeing with the probe, φ-set of the finite T. Thus, we have the whole derivation of (1a) as follows:

22 According to Chomsky (2000: 105), control infinitives (i.e., co ntrol T) fall together with finite clauses, headed by C selecting nondefective T.

-40- (4) TP John T’

T(probe) v*P (= PH2)

EPP of T tJohn v*’ Agree v* VP

V CP (= PH1) try C’ Spell-Out C TP Agree (X) PRO T’

T(control) vP

to tPRO go home early

According to CFCs mentioned in section 2.2, the strong phases in (4) are v*P and CP. Hence, when the higher phase v*P is formed, the complement TP of the head C of the lower phase CP is spelled-out (sent) to each of the PF and LF components for interpretation on the assumption that spell-out is cyclic at the phase level. Hence, the probe, φ-feature set of T, which agrees with the controller John is not accessible to the goal, PRO. Thus, how PRO in (1) is obligatorily controlled by the antecedent John cannot be explained within the current minimalist theory of Chomsky (2000, 2001). In what follows, to resolve this problem, I will illustrate the preliminary assumptions for a new analysis.

4.3 Preliminary Assumptions for a New Analysis

In this section, I will illustrate several assumptions to resolve the problem of obligatory control of PRO within the current minimalist theory. In section 4.3.1, I will suppose Lee’s (1999) argument. Lee (1999) proposes that Move and Agree are separate operations and thus movement operation should be allowed to check the EPP-feature of a functional head (e.g., T, v). In section, 4.3.2, I will assume Bošković’s (1997) argument to avoid the PIC effect. Contrary to what Chomsky argues, Bošković (1997) argues that

-41- control infinitivals are TPs. On the basis of his argument, we can avoid the PIC effect, which is problematic within the minimalist framework. In section 4.3.3, I will claim the need for dividing the non-finite T into two parts.

4.3.1 Separate Operations (i.e., Agree and Move): Lee (1999)

Basically, Chomsky (2000, 2001) supposes that the displacement of an element takes place due to the composite operation Move (i.e., Agree, pied-piping and Merge). But there is another argument about the motivation of displacement. Lee (1999) proposes that unlike Chomsky’s (2000, 2001) argument, Agree and Move should be regarded as separate operations.23 That is, Agree applies to delete “uninterpretable” formal feature, but Move applies to meet the selectional properties (i.e., the EPP-feature) of a head. Consider the following examples.

(5) a. There is a man in the room. b. A man is in the room.

At some stage of the derivation, the sentences in (5a) and (5b) have the same intermediate structure like (6) below.

(6) [TP T-be a man in the room]

There are two ways to derive both sentences (5a) and (5b). On the one hand, the expletive there may be purely merged to T directly. On the other hand, the associate NP a man may raise to the spec of T, yielding (5a) and (5b), respectively. Lee (1999) insists that as Chomsky (2000: 105) indicates, the choice should depend on whether or not the expletive there is included in the sub-lexical array for the phase under consideration. So if it is in the sub-lexical array, (Pure)-Merge applies, but if it is not, Move applies. Based on this fact, in (5a)24, the expletive there is directly merged to [Spec, TP] to satisfy the EPP

23 For a different analysis of Agree and Move, see Park (2001). 24 In this case, the phrase (i.e., there) selected to satisfy the EPP-feature has nothing to

-42- feature of T, whereas in (5b), the probe, φ-feature set of T, agrees with the goal, φ-feature set of the associate NP a man, and the uninterpretable features between them are valued and deleted. Then, the associate NP a man moves to the spec of T to satisfy the EPP feature of T. It means that the EPP feature of T is satisfied by (Pure) Merge (e.g., (5a)) or Move (e.g., (5b)). Like this, Lee (1999) claims that Move should be considered as an independent syntactic operation. This is consistent with the assumption by Lasnik (1992a) that the strong NP feature of Tense (i.e., EPP) renders movement necessary.

4.3.2 On Control Infinitives

Following Stowell (1982), Bošković (1997) argues that the distribution of the empty can be accounted for if they are subject to the Empty Category Principle (hereafter, ECP).25 The ECP satisfaction supports the possibility that control infinitives are TPs. Let us consider the examples below.

(7) a. It is believed [CP C [TP he is crazy]].

b.*[CP C [TP he would buy a car]] was believed at that time.

c.*It was believed at that time [CP C [TP you would fail her]].

In (7a), the empty complementizer C is properly governed by the verb believe. Hence, (7a) satisfies the ECP, thus it is grammatical. In contrast, the empty Cs in (7b) and (7c) are not properly governed by a lexical verb. Hence, (7b) and (7c) violates the ECP, thus they are ungrammatical. However, the infinitival clauses behave quite differently from the finite clauses as seen in (8).

(8) a. I tried at that time [CP C [TP PRO to fail her]].

b. [CP C [TP PRO to buy a car]] was desirable at that time.

According to the ECP, the grammaticality of the infinitival clauses in (8a) and (8b) is not do with Agree. 25 The Empty Category Principle (ECP)

: [α e] must be properly governed. (Chomsky 1981: 250)

-43- expected. Since the empty Cs in (8) are not properly governed by a lexical verb, they should be ruled out by the ECP violation on a par with (7b) and (7c). However, (8a) and (8b) are grammatical. Hence, it is concluded that there are no CP projections in the control sentences like (8). There is another example below, which shows that control infinitivals are TPs.

(9) a.*Mary believed Peter finished school and Bill [α Peter got a job].

b. Mary tried to finish school and Peter [α PRO to get a job].

(9a) is ungrammatical because α that is CP is not properly governed. Hence, the α induces the ECP violation. By contrast, (9b) does not induce the ECP violation, thus it is grammatical. As a result, the contrast between (9a) and (9b) demonstrates that α in (9b) is not CP but TP. Based on these facts, Bošković (1997) argues that unlike the standard analysis, the control infintives are TPs. In this thesis, I assume Bošković’s argument to avoid the PIC effect.

4.3.3 Two Types of Non-finite T

In section 1.2.2.1, I have illustrated the distribution of PRO proposed by Chomsky and Lasnik (1993). According to them, PRO can appear only in the subject position of the non-finite T as shown in (10).

(10) a. Billy attempted [TP PRO to steal the luxurious fountain pen].

b. Billy persuaded Mary [TP PRO to steal the luxurious fountain pen].

c. *John believes [TP PRO solve the problem]. d. *Mary saw PRO. e. *Mary saw [pictures of PRO].

Their arguments, based on the grammatical examples in (10), imply that PRO appears in the subject position of every non-finite T. However, Martin (2001) argues that there are some counterexamples as follows:

-44- (11) a. Naomi tried [PRO to solve the problem]. b.*Naomi believes [PRO to have solved the problem].

(12) a. It was difficult [for Naomi to solve the problem]. b.*It seems to Naomi [PRO to have solved the problem].

In the (b) examples in (11) and (12), the embedded clauses are all non-finite Ts like the ones in the (a) examples in (11) and (12). Hence, if Chomsky and Lasnik’s (1993) argument is correct, the occurrence of PRO in (11b) and (12b) should be possible contrary to fact. This indicates that unlike Chomsky and Lasnik’s (1993) argument, not all non-finite Ts permit the occurrence of PRO in its spec-position. We have another counterexamples below.

(13) a. I believe [TP him to [vP thim [v-have read the book]]].

+EPP

b. He is believed [the to [vP the [v-have read the book]]].

+Case/+EPP +EPP

(14) a. John seems [tJohn to [vP tJohn [v-be honest]]].

+Case/+EPP +EPP

b. She seems to Kim [TP tshe to [vP tshe [v-have solved the problem]]].

+Case/+EPP +EPP

The sentences in (13) are classical ECM constructions. The ones in (14) are raising constructions. In (13a), according to Lee (1999), the NP him purely merged in [Spec, vP] raises to [Spec, TP] in the same clause to satisfy the EPP feature of the non-finite T. At that position, the NP him is accusative-case marked by Agree with the probe, φ-feature set of the verb believe. In (13b), the subject NP he is first purely merged to [Spec, vP] in the embedded clause, and it raises to [Spec, TP] in the same clause to satisfy the EPP of the non-finite T. After the φ-feature-set of the subject NP he agrees with the probe, φ-feature set of Tmatrix, the subject NP he raises to the spec of TPmatrix to satisfy the EPP feature of

Tmatrix (see (13b)). Similarly, in (14a) and (14b), the subject NPs John and she first raise to

-45- [Spec, TP] in the infinitival clause to check the EPP feature of the non-finite T. After that, the φ-feature-set of John and She agrees with the probe, φ-feature of Tmatrix. Then, John and she raise to the spec of TP in the matrix clause to satisfy the EPP feature of Tmatrix, respectively. In (13) and (14), the movement of the NPs (John, she) to [Spec, TP] in the infinitival clause is allowed only to check the EPP feature of the non-finite T. This indicates that the non-finite T of ECM and Raising does not agree with the subject NP of the non-finite TP due to the lack of its case.26 This fact is supported by Martin’s (2001) argument. Following Saito and Murasugi (1990), Martin (2001) proposes that only complement of an agreeing head can undergo . Consider the examples below.

(15) a. Pam [VP likes soccer] and Rebecca [T does][VP e] too.

b. Bill believes Sarah is [AP honest], and he believes Kim is [AP e] as well.

In (15), VP-ellipsis is possible since the finite T including case agrees with the subject (e.g., Rebecca, Kim) of a finite clause. In contrast with (15), VP-ellipsis is not possible in the ECM/Raising infinitives like (16) and (17) below, since the head, non-finite T does not have case, thus cannot undergo agreement with the subject of the non-finite TP.

(16) a.*I consider Pam to [VP like soccer], and I believe [Rebecca [T to][VP e]]as well.

b.*Bill believes Sarah to be [AP honest], and he believes [Kim [Tto][VPe]] as well.

(17) a.*John does not like math but Mary seems to [VP e].

b.*Harry may not be as happy as he appears to [VP e].

Next, consider the examples below.

(18) a.Kim isn’t sure she can [VP solve the problem], but she will try [PRO [Tto][VPe]].

b.Rebecca wanted Jill to [VP join the team], so Pam persuaded her [PRO [T to]

[VP e]].

26 Chomsky (2000, 2001) assumes that uninterpretable features acti vate the matched goal. Hence, for agree to apply, uninterpretable features (e.g., case) are necessary to the head T. For uninterpretable features, see note 12.

-46- In contrast with ECM/Raising infinitives in (16) and (17), control infinitives allow VP- ellipsis (see (18)). Hence, it is concluded that the non-finite T of the control infinitives has case, thus can undergo agreement with the subject PRO in its spec-position. In other words, the asymmetry between ECM/Raising infinitives and control infinitives in terms of VP-ellipsis demonstrates that the non-finite T of ECM/Raising infinitives has no case, thus never allows Agree and VP-ellipsis as shown in (16) and (17), but that the non-finite T of control infinitives has case, thus allows Agree and VP-ellipsis as shown in (18). Based on the contrast between ECM/Raising infintives and Control infinitives reviewed so far, we can generalize the status of the non-finite as follows:

(19) The Revised Alternative [1] There are two types of non-finite T in infinitivals.

(i) Non-finite T is selected by V: Tdefective1 (no case, but EPP, person) e.g. ECM/Raising verbs (e.g., believe, expect/ seem. . .)

(ii) Non-finite T is selected by Vcontrol: Tdefective2 (null-case, EPP and person) e.g. Obligatory control verbs (e.g., try, promise, persuade. . .)

The generalization (19) derived from the Alternative [1] in Chomsky (2001) and Bošković’s (1997) argument (i.e., control infinitives are TPs). On the one hand, according to the Alternative [1] in Chomsky (2001), C selects T-complete, but V selects T-defective. This means that when T is selected by C, it is complete, thus has a full complement of φ- features, and that when T is selected by V, it is defective, thus has person and the EPP feature but no case. On the other hand, according to Bošković’s (1997) argument, control infintives are TPs. Hence, based on those two arguments, it is generalized that control T is also selected by V, thus is defective like ECM/Raising T. However, due to the asymmetry with respect to VP-ellipsis, which has already been shown in the contrast between (16-17) and (18), the non-finite T of control infinitives and that of ECM/Raising ones need to be distinguished from each other. Therefore, I propose that the non-finite T of ECM/Raising infinitivals should be defective 1 (see (19-i)), and that the non-finite T of control infinitives should be defective 2 (see (19-ii)). According to the Revised Alternative [1] in (19), the structures of the ECM, Raising, and control infinitives can be represented as follows:

-47- (20) a. I believe Bill to know the way to go there.

b. I believe [TPdef1 Bill [Tdef1’ to [vP tBill v-know the way to go there]]].

+EPP of Tdef1

(21) a. John seems to be honest.

b. John seems [TPdef1 tJohn [Tdef1’ to [vP tJohn v-be honest]]].

+Case/+EPP of T +EPP of Tdef1

(22) a. Hei promised me [PROi to return home early].

b.Hei promised me [TPdef2 PROi [Tdef2’ to [vP tPRO v-return home early]]].

+Case/+EPP of Tdef2

4.4 A New Analysis of PRO and Control: OC (=Agree)

In this section, I will propose a new analysis of PRO and control. As a solution to the problem of the obligatory control of PROs, I first propose the Optional EPP Requirement on Control Verbs in (23). Secondly, I propose that PRO should move and then it agree with a functional head (T, v). That is, when PRO agrees with T, we get obligatory subject control. When PRO agrees with v, we get obligatory object control. The Optional EPP Requirement on Control Verbs is as follows:

(23) Optional EPP Requirement on Control Verbs27 (i). Subject control type verbs: matrix v* is specified for EPP feature e.g. Promise, try, decide, . . . (ii). Object control type verbs: matrix v* is unspecified for EPP feature e.g. Persuade, force, order, allow, . . .

27 Here positing the EPP feature of the matrix v* is subject to Chomsky’s (2000) proposal : the Core Functional Categories (i.e., C, T, v) may have the EPP features.

-48- In (23), I assume that the light verb v* of the subject control type verbs like promise has the EPP feature to be checked by Merge, but that the light verb v* of the object control type verbs like persuade does not. The proposal is based on the difference of the lexical properties between promise and persuade proposed by Larson (1991). In the same way, I suppose that the optionality of the EPP feature of control verbs is due to the difference of their lexical properties. Furthermore, this supposition has empirical supporting evidence as follows:

(24) a. Whati did John promised Mary ti?

b. [CP What [C did[TP John [v*P t2 [v*P tJohn [v*-promise Mary t1]]]]]]

EPP-feature of C driven EPP-feature of v* driven

(24b) is the representation of the derivation of (24a). In (24b), the wh-element what undergoes movement twice. The first movement is driven by the EPP feature of the matrix v*, and the second one by the EPP feature of C. Since Chomsky (2000: 109) assumes that the EPP feature is optionally assigned to the matrix v* to avoid the PIC effect, the optional EPP assignment proposed by Chomsky supports the proposal in (23-i). Accordingly, the matrix v* of the subject control type verbs like promise has the EPP feature to avoid the PIC effect.

(25) a.*Whati did John persuaded Mary ti?

b.*[CP What [C did[TP John [v*P/strong phase tJohn [v*-persuade Mary twhat]]]]] PIC violation

(25b) is the representation of the derivation of (25a). In (25b), unlike (24b), the wh- element what cannot move to the outer spec of the matrix v* of the verb persuade since the matrix v* has no EPP feature. So, the one fell swoop of what to [Spec, CP] in the matrix clause induces the violation of the PIC (see (25b)). Thus, the contrast between (24) and (25) shows that unlike the matrix v* of the subject control type verbs like promise, that of the obligatory object control type verbs like persuade does not have the EPP feature.

-49-

From now on, I will talk about how the particular choice of the antecedent of OC PRO is determined within the current minimalist theory of Chomsky (2000, 2001). Firstly, let us consider the obligatory subject control construction in the active tense (e.g., try).

(26) a. John tried to go home.

b. John tried [TPdef2 [T’ to [vP PRO [v-go home]]]].

(26a) illustrates obligatory subject control, and (26b) is the structure of the initial stage of the derivation of (26a). As shown in (26b), PRO is base-generated in [Spec, vP] in the 28 non-finite clause. By the Revised Alternative [1] in (19-ii), Tdef2 is merged to vP. Upon agreeing with the probe, φ-set of Tdef2, PRO moves to [Spec, TPdef2] to satisfy the EPP of

Tdef2. At this point, case of PRO and the uninterpretable features (person, null-case) of

Tdef2 are valued and deleted. Thus, we have a structure like (27).

(27) TPdef2 PRO T’

Tdef2 vP

EPP of Tdef2 tPRO v’ v VP Agree go home

Next, the subject control type verb try and the light verb v* are merged to the TPdef2 sequentially. After that, the subject DP John is merged to [Spec, v*P]. Since by (23-i), the matrix v* has the EPP feature, PRO raises to the outer spec position of v*P to check the EPP feature of v*. Thus, we have the intermediate structure below.

28 The merge of PRO to [Spec, vP] in the embedded clause is due to the semantic- selection in Grimshaw (1979).

-50- (28) v*P PRO v*’(=v*P: Phase) John v*’ v* VP

V TPdef2

EPP of v* try tPRO T’ - - -

The raising of PRO in (28) makes it (PRO) locate outside of a phase (v*P). Then, the finite T is merged to v*P. At this point, the probe, φ-feature set of T, agrees with the goal, φ-feature set of the subject DP John, since the subject DP John is in the minimal domain of Tprobe. Next, the subject DP John raises to [Spec, TP] matrix to check the EPP feature of

Tprobe. Thus, we have the whole derivation as follows:

(29) TP John T’ Agree

T(probe) v*P PRO v*’(=v*P)

Merge tJohn(goal) v*’ Agee v* VP

V TPdef2 - - -

In (29), the probe T, which has already agreed with the controller John, agrees with PRO because Chomsky (2000: 131) assumes that the checked features of the probe T do not erase right away when they are checked, but are accessible until the end of the phase. Accordingly, we get obligatory subject control of PRO from (26). Next, I will consider another obligatory subject control construction with three place predicate (e.g., promise).

(30) a. Bill promised me to return home early.

b. Billi promised me [TPdef2 PROi [T’ to [vP tPRO [v-return home early]]]].

-51-

According to the Revised Alternative [1] in (19-ii), the subject control verb promise selects TPdef2 as its complement. Hence, the structure of the infinitival clause of (30a) is

(30b). The derivation of the embedded TPdef2 is the same as that of (26). Thus, we have a structure like (31).

(31) TPdef2 PRO T’

Tdef2 vP

EPP of Tdef2 tPRO v’ v VP Agree return home early

Next, the subject control verb promise, the object DP me, and the light verb v* are merged to the TPdef2 sequentially. After that, the subject DP Bill is merged to [Spec, v*P]. The intermediate structure is as follows:

(32) v*P Bill v*’

v*(EPP) VP v* V me V’

tV TPdef2 promise - - -

In (32), the matrix verb promise is adjoined to the light verb, v*. By (23-i), the matrix v* of the verb promise has the EPP feature, thus PRO moves to the outer spec (XP)-position of v*P to satisfy the EPP feature of the matrix v*. This is consistent with Lee’s (1999) analysis. After that, the φ-feature set of the verbal complex [v*-V v*] agrees with the object DP me, and the uninterpretable features between them are valued and deleted. Thus, we have a structure like (33).

-52- (33) v*P

PRO(=XP) v*’ (= v*P: Phase) Bill v*’

v*(EPP) VP v* V me V’

promise tV TPdef2

EPP of v* tPRO T’

Tdef2 vP - - -

In (33), PRO in [Spec, TPdef2] crosses over another NP me since PRO and the object DP me are in the same distance from the XP-position of v* according to Equidistance29 in Chomsky (1995). That is, PRO and the object me are in the same minimal domain of the verbal complex [v*-V v*]. As a result, by Equidistance, PRO can cross over another DP me. And then, the finite T is merged to (33). At this point, agreement between the probe, φ-feature set of T, and the goal, φ-feature set of Bill, takes place. After that, the subject

DP Bill raises to [Spec, TP] to check the EPP feature of Tprobe. The whole derivation of (30) is as follows:

(34) TP Bill T’ Agree

T(probe) v*P

PRO(=XP) v*’ (= v*P: Phase)

EPP of T tBill v*’

Agree v*(EPP) VP v* V me V’

promise tV TPdef2 - - -

29 If α, β are in the same minimal domain, they are equidistant from γ. See Chomsky (1995: 184).

-53- As shown in (34), the head Tprobe, which has already agreed with the subject DP Bill, agrees with PRO. Thus, we get obligatory subject control of PRO from (30). Now, I will examine the obligatory object control construction with three-place predicate (e.g., persuade).

(35) a. He persuaded me to leave here.

b. He persuaded mei [TPdef2 PROi to leave here].

As for the embedded clause in (35b), we have a structure like (36).

(36) TPdef2 PRO T’

Tdef2 vP - - -

Next, the object control verb persuade, the complement DP me, and the light verb v* are merged to TPdef2 in order. Consecutively, the subject DP he is merged to [Spec, v*P], and the matrix verb persuade is adjoined to the matrix v*, forming the verbal complex [v*-V v*]. After that, the probe, φ-feature set of [v*-V v*], agrees with the goal, φ-feature set of me, and the uninterpretable features between them are valued and deleted. At this point, we have the intermediate structure below.

(37) v*P he v*’

v*(No EPP) VP

v* V me V’

persuaded Agree tV TPdef2 PRO T’

Agree Tdef2 vP (Phase) - - -

In (37), PRO does not move to any higher position and remains in [Spec, TPdef2] instead

-54- since by (23-ii), the matrix v* of the object control type verbs like persuade does not have the EPP feature. Hence, at that position, PRO agrees with v*, which have already agreed with the NP me. Therefore, we get obligatory object control of PRO from (35).

So far, I have considered the obligatory control constructions in the active tense within the current minimalist theory. Now, I will examine the passive tense of the subject control verb promise, which is problematic within the previous syntactic analyses introduced in chapter Ⅲ.

(38) a. *Johni was promised ti [TPdef2 PROi to leave here]. b. TP John T’

Tprobe vP Agree

PRO(XP) v’ (= vP)

Ø(Sub) v’

MLC violation v(EPP) VP

be tJohngoal V’

V TPdef2

promised tPRO T’

EPP of v Tdef2 vP

(39) a. Johni was persuaded ti [TPdef2 PROi to leave here]. b. TP John T’

Tprobe vP Agree

Ø(Sub) v’

v(No EPP) VP

be tJohngoal V’ c-command

V TPdef2 persuaded PRO T’

Agree Tdef2 vP

-55- (38) and (39) are all passives. The former is ungrammatical but the latter is not. In (38), the DP John cannot move to the spec of Tprobe because the movement of the DP John induces the violation of the Minimal Link Condition (hereafter, MLC) in (40), which is proposed by Chomsky (1995: 296).

(40) Minimal Link Condition α can raise to target K only if there is no legitimate operation Move β targeting K, where β is closer to K.

=> K(target=probe) < β < α

( < means “c-command”)

(40) indicates that α cannot cross over β when α targets K, which also attracts β. Based on this condition, let us return to (38). In (38b), by (23-i), the matrix v of the verb promise has the EPP feature, thus attracts PRO in [Spec, TPdef2] to satisfy itself. Hence, PRO raises to the outer spec-position (XP) of vP. After that, the DP John agrees with the probe,

φ-feature set of Tprobe. At this level, the EPP feature of Tprobe attracts both PRO and the DP

John. So, the EPP licensing mechanism of Tprobe in (38b) is represented as follows:

(41) T(target=probe) < PRO < John

However, as shown in (41), the DP John cannot raise to the spec of T directly since the raising of John induces the MLC violation in (40). Accordingly, (38) cannot be grammatical. By contrast, in (39b), the DP John agrees with the probe, φ-feature set of T, thus raises to the spec of Tprobe to satisfy the EPP feature of Tprobe. However, PRO does not move any more out of the spec of TPdef2 since the matrix v of the verb persuade has no

EPP feature (see (23-ii)). At that position, PRO agrees with Tprobe, which has already agreed with the DP John, thus it is eventually controlled by the DP John (see (39b)). So, the EPP licensing mechanism of Tprobe in (39) is represented as follows:

-56- (42) T(target=probe) < John < PRO

(42) indicates that the NP John can raise to the spec of TP since the raising of John does not violate the MLC in (40). Therefore, the passivized sentence in (39) is grammatical in contrast with the one in (38). So far, I have explained how the particular choice of the antecedent of OC PROs is determined and why the passive tense of the subject control verb promise in (38) contains ungrammaticality within the current minimalist theory of Chomsky (2000, 2001). This particular example (38) has not been explained by the previous syntactic analyses introduced in chapter Ⅲ. In addition, I have also proved that my proposal is compatible with the recent minimalist framework. There have been many theories of PRO and control, but I conclude that the particular choice of the antecedent of OC PROs depends on the syntactic operation Agree, which is subject to Optional EPP Requirement on Control Verbs (23).

-57- ChapterⅤ

Conclusion

From the beginning, we started this thesis with the following two questions:

(1) a. Can the conception of government be eliminated in accounting for PRO? b. How is the particular choice of the antecedent of the OC PRO determined?

To answer the question (1a), I have first considered the reason why the notion of government should be dispensed with within the Minimalist Program (Chomsky 1993, 1995). Owing to the local relation of the Minimalist Program, the notions of government and PRO theorem based on it have vanished together. Hence, instead of the PRO theorem, I have illustrated Chomsky and Lasnik’s (1993) analysis: case theoretic account of the distribution of PRO. To answer the question (1b), I have examined (i) the minimalist assumptions proposed by Chomsky (2000, 2001) in ChapterⅡ; (ii) the previous analyses of the OC PRO (i.e., Minimal Distance Principle, Intrinsic Properties of Predicates based on Control theory, ATTRACT, and PRO = NP-trace analysis) in Chapter Ⅲ; and (iii) the preliminary assumptions for a new analysis and the Optional EPP Requirement on Control Verbs in Chapter Ⅳ. The problems of the several analyses of the OC PRO are also illustrated in Chapter Ⅲ. In Chapter Ⅳ, as a new analysis of the interpretation of PRO, I have proposed that the choice of the antecedent of the OC PRO depends on the syntactic operation Agree between a head (T, v) and PRO, which is subject to the Optional EPP Requirements on Control Verbs (23). That is, when PRO agrees with T, we get obligatory subject control. When PRO agrees with v, we get obligatory object control. For a new analysis, I have supposed Bošković’s (1997) argument (i.e., control infinitivals are TPs) and Lee’s (1999) argument (i.e., Agree and Move are separate operations). In this framework, contrary to the previous analyses, I have shown that my proposal can verify

-58- the difference between obligatory subject control and obligatory object control as well as the ungrammaticality of the particular example (i.e., the passive tense of the subject control verb promise) within the current minimalist theory of Chomsky (2000, 2001).

-59- References

Abraham, Werner, Samuel D. Epstein, Hoeskulder Thráinson, and C. Jan-wouter Zwart. 1996. Minimal Ideas: Syntactic Studies in the Minimalist Framework. Amsterdam: John Benjamins. Boecks, Cedric. 2000. A Note on Contraction. Linguistic Inquiry 31, 357-365. Bošković, Željko. 1995. Principles of Economy in Nonfinite Complementation. Doctoral dissertation, University of Connecticut. Bošković, Željko. 1997. The Syntax of Nonfinite Complementation. Cambridge: MIT Press. Bouchard, Dennis. 1984. On the content of empty categories. Foris, Dordrecht. Bouchard, Dennis. 1985. PRO, Pronominal or Anaphor. Linguistic Inquiry 16, 471-477. Bresnan, Joan. 1982. Control and Complementation. Linguistic Inquiry 13, 343-434. Chierchia, Gennaro and Pauline Jacobson. 1986. Local and Long Distance Control, NELS 16, 57-74. Chomsky, Noam. 1980. On Binding. Linguistic Inquiry 11, 1-46. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Kluwer. Chomsky, Noam. 1986a. Knowledge of Language: Its Nature, Origin and Use. Praeger, Westport, Connecticut. Chomsky, Noam. 1986b. Barriers. Cambridge: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist Inquiries: the Framework. In Roger Martin, David Michaels and Juan Uriagereka (eds.). Step by step: Essays on minimalist syntax in honor of Howard Lasnik. Cambridge, 89-155. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2001. Derivation by Phase. In Michael Kenstowicz (eds.). Ken Hale: A Life in Language, 1-52. Cambridge, Mass.: MIT Press. Chomsky, Noam And Lasnik, Howard 1993. The theory of principles and parameters. In J. Jacobs, A. Von Stechow, W. Sternefeld and T. Vennemann. eds. Syntax: An International Handbook of Contemporary Research. Berlin: Walter de Gruyter. Grimshaw, Jane. 1979. Complement selection and the lexicon. Linguistic Inquiry 10, 279- 326. Groat, Erich M. 1995. English Expletives: A minimalist approach. Linguistic Inquiry 26,

-60- 354-365. Haegeman, Liliane. 1994. Introduction to Government and Binding. Oxford: Blackwell. Higginbotham, James. 1992. Reference and Control, in R. Larson, S. Iatridou, U. Lahiri, and J. Higginbotham (eds.) Control and Grammar, Kluwer, Dordrecht, 79-108. Hornstein, Norbert. 1996. Control in GB and Minimalism. Glot International 2, 3-6. Hornstein, Norbert. 1999. Movement and control. Linguistic Inquiry 30, 69-96. Hornstein, Norbert. 2000. Move! A Minimalist Theory of Construal. Oxford: Blackwell. Jaeggli, Osvaldo. 1980. Remarks on to contraction. Linguistic Inquiry 11, 239-245. Larson, Richard. 1991. Promise and the Theory of Control. Linguistic Inquiry 22, 103- 139. Lasnik, Howard. 1988. Subjects and the θ-Criterion. NLLT 6, 1-17. Lasnik, Howard. 1992a. Case and Expletives: Notes toward a Parametric Account. Linguistic Inquiry 23, 381-405. Lasnik, Howard. 1992b. Two Notes on Control and Binding, in R. Larson, S. Iatridou, U. Lahiri, and J. Higginbotham (eds.) Control and Grammar, Kluwer, Dordrecht, 235-51. Lasnik, Howard. 1995. Last Resort and Attract F. In Papers from the Sixth Annual Meeting of the Formal Linguistics Society of Midamerica, eds. L. Gabriele, D. Hardison, and R. Westmoreland, 62-81. Indiana University Linguistics Club, Indiana University, Bloomington. Lee, Hong Bae. 1999. On defining move. Korean Journal of Linguistics 24, 619-637. Lee, Hong Bae. 2001. Lectures on Minimalist Syntax. Seoul: Hanshin Publishing Co. Manzini, Maria Rita. 1983. On Control and Control Theory. Linguistic Inquiry 14, 421- 446. Manzini, Maria-Rita, and Anna Roussou. 1999. A Minimalist Theory of A-movement and Control. Lingua 110, 409-447. Martin, Roger. 1996. A Minimalist Theory of PRO and Control. Doctoral dissertation, University of Connecticut. Martin, Roger. 1999. Case, the extended projection principle, and minimalism, In S. D. Epstein and N. Hornstein (eds.), Working minimalism, Cambridge, Mass.: MIT Press, 1-25. Marin, Roger. 2001. Null Case and the Distribution of PRO. Linguistic Inquiry 31, 141-

-61- 166. McCloskey, James. 1991. There, it, and agreement. Linguistic Inquiry 22, 563-567. Mohanan, K. P. 1985. Remarks on Control and Control Theory. Linguistic Inquiry 16, 637-648. O’Neil, John Herbert. 1997. Means of Control: deriving the properties of PRO in the minimalist program. Doctoral dissertation, Harvard University. Park, Du-Hong. 2001. A Minimalist Approach to Control Constructions. Language and Information Society 2, 31-43. Park, Sung-Hyuk. 2001. Agree and Move. Korean Journal of English Language and Linguistics 1, 561-585. Pesetsky, David. 1992. syntax, Vol 2. Ms., MIT. Peter W. Culicover and Ray Jackendoff. 2001. Control is not Movement. Linguistic Inquiry 32, 493-512. Radford, Andrew. 1997. Syntactic Theory and the Structure of English: A minimalist approach. Cambridge: University of Cambridge Press. Randau, Idan. 1999. Elements of Control. Doctoral dissertation, MIT. Roberts, Ian. 1997. Comparative Syntax. London: Edward Arnold. Rosenbaum, Peter. 1967. The Grammar of English Predicate Complement Constructions. Cambridge, Mass.: MIT Press. Rosenbaum, Peter. 1970. A Principle Governing Deletion in English Sentential Complementation. Readings in English Transformational grammar. R. Jacobs and P. Rosenbaum. Waltham, Ginn. Růžička, Rudolph. 1983. Remarks on Control. Linguistic Inquiry 14, 309-324. Saito, Mamoru, and Keiko Murasugi. 1990. N’-deletion in Japanese. In Uconn working papers in linguistics 3, 87-107. Department of Linguistics, University of Connecticut, Storrs. Stowell, Tim. 1982. The Tense of Infinitives. Linguistic Inquiry 13, 561-570. Ura, Hiroyuki. 2001. Case. In M. Baltin and C. Collins, eds., The Handbook of Contemporary Syntactic Theory. Oxford, UK: Blackwell Publishers. Williams, Edwin. 1980. Predication. Linguistic Inquiry 11, 203-308.

-62-