<<

[Art i c l e ]

EXTRAPOSITION AND CYCLIC

Ma n a b u Mi z u g u c h i Nagano National College of Technology

This paper explores extraposition from NP and claims that it provides evi- dence for the cyclic model of derivations proposed in the recent Minimalist Program. We propose that an extraposed phrase, which is base-adjoined, is interpretively associated with its host in the semantic component via cyclic Transfer. We show that the proposed analysis can give a natural account to attested properties of extraposition. Through our investigation, we dem- onstrate that our proposal not only endorses a Minimalist view of syntactic derivations but also has theoretical consequences for the way CHL constructs derivations, proposing a more derivationally oriented phase theory. We also argue that the proposed analysis straightforwardly explains well-formed sub- extraction from subjects.*

Keywords: extraposition, base-adjunction, cyclic Transfer, phase syntax, sub- extraction from subjects

1. Introduction “Extraposition” is a name for the dislocation that moves a phrase to the right periphery of the sentence. Some illustrative examples of extraposition are the following:

(1) a. A review [PP of this article] came out yesterday. b. A review came out yesterday [PP of this article]. (2) a. John read a book [PP on German linguistics] yesterday. b. John read a book yesterday [PP on German linguistics]. In (1b) and (2b), the bracketed PPs are dislocated from the subject and the object, respectively, to the end of the sentence. This paper investigates PP extraposition from NP (subjects and objects) in English as exemplified

* For helpful comments, invaluable suggestions and constructive criticisms on earlier versions of this paper, I would like to thank Prof. Kinsuke Hasegawa, Prof. Hidekazu Suzuki, and especially, three anonymous EL reviewers. Needless to say, all remaining er- rors and inadequacies are my own.

English Linguistics 26: 2 (2009) 293–328 -293- © 2009 by the English Linguistic Society of Japan 294 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009) in (1) and (2), and considers its properties and implications in an explana- tory framework of the Minimalist Program (MP) (Chomsky (1995), in par- ticular, Chomsky (2000) and his subsequent writings).1 We will claim that extraposition is evidence for the cyclic syntax envisioned in the MP. We will defend a non-movement analysis of extraposition advocated in Culi- cover and Rochemont (1990) and advancing their insight in a derivational framework, will propose that an extraposed phrase and its host DP are in- terpretively associated with each other via cyclic Transfer to the external components. We will demonstrate that the syntax, with its interaction with these interpretive components, explains known properties of extraposition in a principled way. In the course of our discussion, we will see that the pro- posed analysis not only endorses a Minimalist model of syntactic derivations but also has theoretical consequences for the way CHL (the computational system of human language) produces derivations. This paper is organized as follows. In section 2, we will provide theo- retical and empirical arguments against a movement analysis of extraposi- tion. In section 3, we will argue that extraposition is base-adjunction and claim that extraposition interpretation is warranted by an interpretive prin- ciple in the semantic component. In section 4, we will consider two re- strictions on extraposition and demonstrate that they straightforwardly follow from the proposed analysis. This section claims that extraposition interpre- tation applies cyclically by way of cyclic Transfer, which we will argue pro- vides evidence for the phase-based cyclic syntax proposed in the MP. In this discussion, we will propose a more derivationally oriented phase theory, as a consequence of which phases as a whole are Transferred. In section 5, we will show that well-formed sub-extraction from subjects is given a nat­ ural account under the proposed analysis. In section 6, we will present the conclusion of this paper.

2. Extraposition from NP: Movement or Base-generation In this section, we will argue that movement is not involved in extraposi- tion. Since Ross’s (1967) formulation of extraposition as a transformational

1 Extraposition from NP includes PP extraposition as in (1) and (2), and CP extraposi- tion as in (i). In this paper, we restrict our discussion to PP extraposition (cf. fn. 27). (i) a. A man came into the room that no one knew. b. A report was made in public today that the ambassador was still in hiding. (Culicover and Rochemont (1990: 23)) EXTRAPOSITION AND CYCLIC SYNTAX 295 rule, a dominant view in the literature has been that an extraposed phrase is moved rightward to the end of the sentence, leaving its trace in the host DP, which allows the proper interpretation of the extraposed phrase. There are, however, theoretical (especially, in terms of the MP) and empirical argu- ments against this view. Below we will discuss them in turn.

2.1. Theoretical Arguments In the MP, where features are ineliminable and play a central role in determining the form and function of the basic operations of CHL, opera- tions are not free but motivated at each step and derivations proceed step by step through the local manipulation of uninterpretable features selected in functional heads; all syntactic operations follow the condition of last resort, which is an element of efficient computation. This shift of perspec- tive in the MP entails that there is no Move α (move anything anywhere anytime) but that every movement must have a trigger. Under this view of language design, rightward movement like extraposition goes against the last resort principle. Unlike other familiar instances of movement (e.g. subject raising, wh-movement), which are driven by morphological features such as φ, Case and wh, it is unclear what feature checking is involved in extraposition. Some may argue that there is an “extraposition feature.” It is unclear, however, to what extent this feature is independently motivated beyond extraposition, which casts doubt on the postulation of such a feature. To overcome this theoretical problem, one might resort to Fukui (1993), who claims that a movement operation which produces a structure consist­ ent with the value of the head parameter is costless. Given that English is a head-initial language, rightward movement will be costless under Fukui’s proposal. This proposal, however, does not provide a fundamental solu- tion. This is because Merge, a structure building operation in the MP, simply pairs two syntactic objects together and does not specify their linear order in the newly formed object (Merge(α, β) → {γ, {α, β}}). Put dif- ferently, the head parameter is no longer a property of the syntax (but see Saito and Fukui (1998)). In addition to this problem, as we have already noted, Move α is disallowed in the MP and every movement is thus costly, whether a phrase moves leftward or rightward.

2.2. Empirical Arguments We now turn to some empirical arguments against a movement analysis of extraposition. The first empirical argument comes from the fact that ex- traposition is possible from subjects (= (1)). As illustrated in (3), movement 296 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

(sub-extraction) from subjects is disallowed:

(3) a. * Whoi did [DP a picture of ti] please Bill? b. * Whoi did [DP stories about ti] terrify John? It has been well argued that the ungrammaticality of (3) is due to the Sub- ject Condition, which is descriptively stated as follows (Chomsky (1973), Huang (1982) among others): (4) Subject Condition Sub-extraction from subjects is prohibited. On the other hand, as we can see from (1b), extraposition is not subject to the Subject Condition. Since this condition seems to be an empirically robust generalization, the grammaticality of (1b), in contrast with the un- grammaticality of (3), suggests that movement is irrelevant to extraposition. The second argument is that extraposition, unlike subject raising and wh- movement, does not allow long distance movement and cannot be operated across clausal boundaries (Akmajian (1975), Ross (1967)). Consider the following examples: (5) a. * [Everyone believes [that [that a review will come out] is cer- tain] of this article]. b. * [Everyone believes [that [that a review will come out] is cer- tain of this article]]. c. [Everyone believes [that [that a review will come out of this article] is certain]].

(6) a. [Whati did you hear [that Sam thought [that Michael would sing ti]]]? b. [Garyi seemed [to appear [to be certain [to ti know the an- swer]]]]. Since extraposition cannot be considered as Case/φ-feature checking A- movement, it will be an instance of A′-movement under the familiar classification. Then if extraposition involves movement, long distance extraposition would be predicted to be possible just like wh-movement, thanks to successive-cyclic movement via [Spec, CP] (a so-called “escape hatch”). The ungrammaticality of (5a, b), however, shows that this is not borne out and that successive-cyclic movement is not available for extrapo- sition. A movement analysis would have to have some ad hoc stipulation to block long distance extraposition. On the other hand, if movement is extraneous to extraposition, the absence of long distance movement will be straightforward. Given that a long distance property is a hallmark of move- ment, the local nature of extraposition suggests that movement is irrelevant to extraposition. EXTRAPOSITION AND CYCLIC SYNTAX 297

The final argument comes from the directionality of movement. Extrapo- sition has been considered to be an instance of “rightward” movement. On the other hand, in typical instances of movement, XP moves leftward. On the assumption that linear orders are determined in the phonological compo- nent, this leftward-rightward distinction naturally follows if the movement of XP to the Spec of a head H in the syntax due to feature checking is an instruction to the phonological component that the XP must be linearized before H and its complement. Given this empirical association of move- ment and leftward pronunciation as exemplified by subject raising and wh- movement, the rightward nature of extraposition implies that movement is foreign to extraposition. In this section, we have argued that the Minimalist view of language de- sign favors a non-movement analysis of extraposition, which is endorsed by three empirical facts.

3. Extraposition as Base-adjunction 3.1. Base-adjunction to VP The theoretical and empirical discussions in the last section suffice to conclude that movement is irrelevant to extraposition and motivate an alter- native, non-movement analysis. Given our discussions, we suggest that an extraposed phrase (PP) is base-adjoined (Merged) to a maximal projection as an adjunct. In fact, this proposal has already been made independently in Culicover and Rochemont (1990) (also Rochemont and Culicover (1990)) in the pre-Minimalist framework. In this paper, following their original proposal, we assume as our first approximation the following structure for extraposition, where an extraposed phrase is indicated as “EXP”:2

(7) [TP Subj T [vP v [VP [V′ V Obj] EXP]]] The base-adjunction analysis of extraposition is not unreasonable given that PPs can freely behave as adjuncts to other XPs (typically, VP), as shown in (8):

(8) a. John fixed the car PP[ with that spanner]. b. They played tennis [PP in the park].

2 A light verb v is added to the VP originally assumed in Culicover and Rochemeont (1990), in order to update verbal structure along recent Minimalist studies. For the mo- ment, we simply assume that an extraposed phrase is adjoined to VP. The light verb used in (7) represents both transitive v* (a strong phase) and intransitive v (a weak phase). For the strong/weak distinction of v, see our discussion in 4.1.1. 298 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

It should be noted that the proposed analysis overcomes the theoretical prob- lems pointed out in section 2 because adjunction, considered as an instance of Merge, is independently required in the syntax, and accounts for the three empirical cases we have considered. In addition, this analysis captures the traditional view that extraposition is an instance of adjunction. The structure (7) says that an extraposed phrase is adjoined to VP. This is a commonly held assumption for the PP extraposed from the ob- ject in such examples as (2b) (Baltin (1978), Guéron (1980), Nakajima (1984)). Culicover and Rochemont (1990) argue that the PP extraposed from the subject can also be adjoined to VP.3 We refer the reader to Cu- licover and Rochemont (1990) for detailed empirical arguments for this claim. Below, we provide additional supporting evidence for their original claim. The evidence comes from the following data. Consider (9): (9) a. A review didn’t appear in the New York Times [of any paper on international relations]. b. ? A picture wasn’t given to him [of any member of our party]. (Hirata (1996: 96)) In extraposition, a negative polarity item in an extraposed phrase can fall under the scope of sentential negation. It has been standardly assumed that sentential negation intervenes between TP and vP/VP as one ingredient of clausal architecture in the form of NegP (Pollock (1989)), the head of which is sentential negation (Neg) (thus, the clausal structure with NegP is [TP T [NegP Neg [vP v [VP V …]]]]). Given that scope relations are based on c- command relations (May (1985)), the fact that the negative polarity item any falls under the scope of sentential negation in (9) demonstrates that the ex- traposed phrase is attached lower than NegP; otherwise, any would be out- side of the scope of the negation. The licensing of a negative polarity item in an extraposed phrase thus provides another piece of evidence for (7). From this and other empirical arguments in Culicover and Rochemont (1990), we can conclude that an extraposed PP, whether it is from subjects

3 The assumption that the PP extraposed from subjects is adjoined to VP is not com- monly accepted. However, given the generalization (19), which we will discuss in 4.1, this assumption may not be so unnatural. In addition, there is no evidence that extrapo- sition from subjects may not be adjoined to VP, which suggests that adjunction to TP is not of necessity. As we will see immediately below, there is evidence that the PP extra- posed from subjects is in VP. EXTRAPOSITION AND CYCLIC SYNTAX 299 or objects, can be base-adjoined to VP.4

3.2. Interpretive Principle Given that extraposition is base-adjunction as represented in (7), there is no trace in the subject and the object which allows the proper interpretation of an extraposed phrase. Then some interpretive principle must be in place in order to assure extraposition interpretation. In this subsection, we will propose to reformulate the “Complement Principle” in a derivational model and claim that the proposed principle interpretively associates an extraposed phrase with its host phrase in the semantic component, an interface to the Conceptual-Intentional system, where a system of semantic interpretation comes into play and the semantic aspects of language are determined by the syntax, after the derivation is Transferred to this component. Considering a non-movement approach to extraposition in the pre- Minimalist framework, Culicover and Rochemont (1990) propose, adapting earlier proposals by Guéron (1980) and Guéron and May (1984), that the Complement Principle associates an extraposed phrase adjoined to XP with its host phrase. This principle is stated as (10), where β is an extraposed phrase and α is its host, and government is understood configurationally (Chomsky (1981)):5 (10) β is a potential complement of α (α, β = Xmax), only if α and β are in a government relation. (Culicover and Rochemont (1990: 41)) However, the Complement Principle is not a principled explanation in the MP since it crucially resorts to government, which is conceptually question- able for its arbitrariness, hence theoretically non-existent in the explanatory framework we are assuming. Instead, we claim that the original Comple- ment Principle can be reformulated as an interpretive principle of the se- mantic component, which applies to the output of syntactic derivations opti- mally executed via Merge. We propose the following Minimalist version of

4 Nakajima (1995) argues that CP extraposition targets TP. Then the fact that an ex- traposed PP comes before an extraposed CP in (i) suggests that the former is structurally lower than the latter, providing us with one more piece of evidence that an extraposed PP is adjoined to VP: (i) And then a man suddenly appeared at the door [PP from the CIA] [CP who I had seen the previous day]. (Nakajima (1995: 28)) 5 “Complement” here is intended to include not only an argument of N but also an ad- junct modifying NP. 300 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009) the Complement Principle:6 (11) The Minimalist Complement Principle (MCP) A phrase β is a potential complement of its host α iff α and β mutually c-command each other in the semantic component. The MCP associates an extraposed phrase with its host DP in the seman- tic component and its violation induces the violation of Full Interpretation at the semantic interface.7 But a fundamental question is why mutual c-com- mand is necessary, instead of (asymmetrical) c-command as found in scope and Agree relations.8 We answer this question by saying that a sisterhood relation, which underlies head-complement and modifier-modifiee relations, is a mutual c-command relation. To see this, consider (1), which is repeat- ed as (12) below:

(12) a. A review [PP of this article] came out yesterday. b. A review came out yesterday [PP of this article]. In (12a), [(a) review] and [of this article] are in a sisterhood relation, and this relation follows directly from the Merge of [(a) review] and [of this article]. On the other hand, there is no such relation in (12b), where the PP is extraposed and is base-adjoined to VP. However, if [of this article] c-commands [(a) review] and [(a) review] in turn c-commands [of this ar- ticle], the sisterhood relation can indirectly follow and the same relation can be established in both (12a) and (12b). Thus, provided that a sisterhood relation is a mutual c-command relation and is required for the proper in- terpretation of [(a) review] and [of this article] in (12a), it naturally follows

6 We assume the following definition of c-command, which is deducible from the two relations directly provided by Merge (viz. sisterhood, which holds of (α, β) in the object K(α, β), and immediately contain, which holds of (K, α), (K, β) and (K, K)): (i) C-command (= sister(contain)) α c-commands β if α is the sister of K that contains β. (ii) K contains β if K immediately contains β or immediately contains L that con- tains β. (Chomsky (2000: 116)) Thus, c-command is coded through Merge or is emergent from the history of Merge. C-command employed in the semantic (interpretive) component is representation- al. See Epstein (1999) and Epstein et al. (1998) for a derivational approach to c-com- mand, which we assume for the narrow syntactic computation. 7 More precisely, it is N (NP) that an extraposed phrase is a complement of. Given that DP is an extended projection of NP (Grimshaw (1991)), it can be safely said that an extraposed phrase and N (NP) are interpretively associated with each other through mutual c-command between the extraposed phrase and the extended projection of NP (i.e. DP). 8 I thank an anonymous reviewer for raising this question. EXTRAPOSITION AND CYCLIC SYNTAX 301 that mutual c-command, instead of asymmetrical c-command or Agree, is required for extraposition interpretation to be obtained.9 With the MCP in place, we now show that the proposed interpretive prin- ciple enables an extraposed phrase to be interpreted as a potential comple- ment of its host after the derivation is Transferred to the relevant compo- nent. Consider how extraposition in (1b) and (2b) is interpreted under the base-adjunction analysis in (7). The examples are repeated below as (13):

(13) a. A review came out yesterday [PP of this article]. b. John read a book yesterday [PP on German linguistics]. Let us start with (13a), a case of extraposition from subjects, and con- sider its derivation in detail, which is summarized in (14) below. As we have argued, in the course of the derivation, the extraposed phrase [of this article] is base-adjoined to VP (which then turns into, or is interpreted as, V′ under Bare Phrase Structure) (= (14a)). In the next stage, T is Merged with v/VP and Agrees with [a review], which has been Merged with V, in Case/φ-features when uninterpretable φ-features are inherited onto it when C is Merged (Chomsky (2007, 2008)). As a result of this Agreement, [a review] moves to [Spec, TP] from its base position for a reason that is per- tinent to feature checking (see below) (= (14b)). This derivation as a whole is handed over to outside components through the operation Transfer (TF) at CP phase as in (14c) (verb movement to v is omitted in (14) and elsewhere for ease of exposition).10

(14) a. [VP [V′ came out a review yesterday] of this article] b. [CP C [TP [a review]i T [vP v [VP [V′ came out ti yesterday] of this article]]]]

c. [CP C [TP [a review]i T [vP v [VP [V′ came out ti yesterday] of this article]]]] → TF (a review >< of this article) It should be noted that within this Transferred CP phase, [of this article] c-commands (the copy of) [a review] and that [a review], in turn, c-com- mands [of this article], which has been adjoined to VP, from its derived position ([Spec, TP]). The DP and the extraposed PP mutually c-command each other in the semantic component thanks to the movement of the

9 The MCP is thus a natural consequence of the local nature of semantic (i.e. argument structure, modifier-modifiee) relations. This implies that mutual c-command, which is a natural product of Merge, is a means to restore local relations from non-local ones. 10 “>” in (14) and elsewhere represents a c-command relation. Thus if “α > β,” then α c-commands β. 302 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

DP. Thus, it follows from the MCP that the extraposed PP is licitly inter- preted as a potential complement of the DP.11 We now turn to extraposition from objects in (13b) and consider its derivation step by step, which is summarized in (15). Suppose that the extraposed PP [on German linguistics] is adjoined to VP just like [of this article] in (13a) (= (15a)). In the next stage, transitive v* is Merged to this VP. Given that derivations are phase-based and uninterpretable features are inherited from phase heads to non-phase heads in the course of the deriva- tion, the uninterpretable φ-features in v* are inherited onto V when v* is Merged to VP (= (15b)), just like those in C are inherited onto T. The inherited uninterpretable φ-probe in V starts searching and Agrees with the object [a book] in φ-features. Suppose, along the lines of Epstein and Seely (2006), that φ-feature Agreement does not value in a free-ride way an uninterpretable Case feature in the object and that v* (and C) must also have a feature for Case valuation. Suppose, further, that uninterpretable features must function as a probe and probe matching features under lo- cal c-command in order to be valued (Bošković (2007), Epstein and Seely (2006); see also Epstein et al. (1998: Ch. 3)). These assumptions motivate the movement (Internal Merge) of the object to VP or [Spec, VP], a position that corresponds to [Spec, TP] in the verbal domain, in the same way that a subject is driven to move to [Spec, TP] for its uninterpretable Case.12 It should be noted that this movement is not counter-cyclic given that cyclic- ity is computed on uninterpretable features: the movement to [Spec, VP] (or Case feature checking) can be implemented only after v* is Merged and

11 Copies can be employed for the purposes of semantic interpretations, which is evi- denced by (i) (italics are intended to show binding relations): (i) [TP1 Johni seems to Mary [TP2 ti to appear to himself [TP3 ti to be ti happy]]]. An anaphor must have a clause-mate c-commanding phrase as its antecedent for its in- terpretation (Binding Condition A — see Lasnik and Hendrick (2003)). This is shown in (iia), where John in the matrix clause cannot be an antecedent for the reflexive. As is well known, the DP embedded in an experiencer PP can c-command out of this PP (= (iib)): (ii) a. *[TP1 Mary seems to John [TP2 to appear to himself [TP3 to be happy]]]. b. *[TP1 Susan seems to him [TP2 to appear [TP3 to like Charlie]]]. Given this, the legitimate antecedent for the reflexive in (i) is not John in [Spec, TP1] but its copy (ti) in [Spec, TP2]. Hence we can see that copies can be used for semantic interpretations, even though they are inaccessible to the narrow syntax. 12 Given that X′ is invisible (Chomsky (1995)), Merge, including Move (= Internal Merge) and adjunction, targets XP (Xmax): a phrase can only be Merged with XP. Thus, [Spec, XP] created by movement is the highest position within the XP. EXTRAPOSITION AND CYCLIC SYNTAX 303 uninterpretable features including Case are inherited onto V. From [Spec, VP], the uninterpretable Case feature in the object probes its matching feature in V. This derivation yields (15c). Note that the uninterpretable Case feature cannot probe from its in-situ position because it no longer c-commands V after the stage in which V and the object are Merged has passed in the derivation (Epstein (1999)). The rest of the derivation goes as shown in (14b) and the finally derived structure is (15d).

(15) a. [VP [V′ read-V [a book{uCase}] yesterday] on German linguis- tics]

b. [v*P John v* [VP [V′ read-V{uφ} [a book{uCase}] yesterday] on German linguistics]]

c. [v*P John v* [VP [a book{uCase}]i [V′ [V′ read-V{uφ} ti yesterday] on German linguistics]]]

d. [CP C [TP Johnj T [v*P tj v* [VP [a book{uCase}]i [V′ [V′ read- V{uφ} ti yesterday] on German linguistics]]]]] → TF (a book >< on German linguistics) When the derived structure (the VP that contains [a book] and [on Ger- man linguistics]) is Transferred to the semantic component at CP phase, the extraposed phrase [on German linguistics] c-commands (the copy of) the object [a book], which also c-commands VP-adjoined [on German linguis- tics] from [Spec, VP] since it became higher than the PP through its move- ment (fn. 12). As in (13a), there is a mutual c-command relation between the extraposed phrase and the object in the semantic component. Thus the MCP, coupled with independently motivated object movement in the syntax, explains the licit extraposition interpretation in (13b). As we have said, objects move to [Spec, VP] for Case feature checking, just as subjects move to [Spec, TP]. It is argued that the assumptions we have introduced are well motivated. Theoretically, they dispense with a mysterious and stipulative EPP (or more generally, edge) feature, which is considered to be a feature specifically designed for movement in the MP, and allow us to deduce movement (the EPP effects) from independently motivated mechanisms of ineliminable morphological features. Empirically, object movement to [Spec, VP] for Case feature checking is suggested by the fact that objects can c-command VP adjuncts. Consider (16) ((16a–c) are cited from Lasnik (1995: 11, 12)): (16) a. ? The DA accused the defendants during each other’s trials. b. The DA accused none of the defendants during any of the trials. 304 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

c.?* Joan believes him even more fervently than Bob’s mother does. d. I saw every boy in his house. These examples show that objects are higher than VP adjuncts. Given that objects are base-Merged below such adjuncts, (16) endorses the derivation in which they move to [Spec, VP]. To summarize the discussion in this section, we have argued that an ex- traposed phrase is base-adjoined to VP and proposed that it is interpretively associated with its host phrase in the semantic component through the inter- pretive principle (the MCP) we have advocated.13 Before we leave this section, we should emphasize that the MPC can explain extraposition without any restrictions on adjunction. Thus far, we have followed Culicover and Rochemont (1990) and assumed as our first hypothesis that an extraposed phrase in extraposition from subjects is also base-adjoined to VP (= (7)). The MPC, however, can correctly rule in ex- traposition from subjects even if an extraposed phrase is base-adjoined to TP, a possibility also mentioned in Culicover and Rochemont (1990). In the derivation in which an extraposed phrase is base-adjoined to TP (which then turns into T′), it c-commands a subject copy within VP and the subject, in turn, c-commands the extraposed phrase from its derived [Spec, TP] posi- tion.14 Consider this with (13a):

(17) [CP C [TP [a review]i [T′ [T′ T [vP v [VP came out ti yesterday]]] of this article]]] → TF (a review >< of this article) Likewise, problems do not arise even if an extraposed phrase is adjoined to vP, a possibility available in the current MP. On the other hand, TP-adjunc- tion is ruled out for extraposition from objects. This is also straightforward from the MCP because a raised object cannot c-command a TP-adjoined phrase from [Spec, VP]:

13 An anonymous reviewer has pointed out that the proposed analysis of extraposition predicts the following example will be ambiguous in scope. Consider (i): (i) Some review came out yesterday of every article. This prediction is correct. As expected, some review and every article can take scope over each other (some review >< every article). 14 A tacit assumption is that the adjunction of an extraposed phrase precedes the move- ment of the subject. Given that Move is more complex than Merge (Chomsky (2000) et seq.), this assumption follows from the requirement reducible to the efficiency principle that simpler operations take precedence over more complex ones in application. EXTRAPOSITION AND CYCLIC SYNTAX 305

(18) [CP C [TP Johnj [T′ [T′ T [v*P tj v* [VP [a book{uCase}]i [V′ read-V{uφ} ti yesterday]]]] on German linguistics]]] → TF (on German linguistics > a book; *a book > on German linguistics) Thus, adjunction, one instance of Merge, is unconstrained, and the refer- ence to VP (or any other XPs) is unnecessary. Where an extraposed phrase can be adjoined (VP, TP or elsewhere) for its proper interpretation follows from the MCP, not from adjunction restrictions.15

4. Restrictions on Extraposition In this section, we will claim that the proposed analysis of extraposition, with its interaction with the semantic component via cyclic Transfer, can straightforwardly explain attested restrictions on extraposition without any unmotivated assumptions, demonstrating that extraposition is strong evidence for the cyclic syntax envisioned in the MP. Along the way, we will con- sider theoretical consequences of our proposal for the way CHL constructs derivations, refining the mechanisms of syntactic derivations.

4.1. Extraposition from D-structure X0-sister Positions 4.1.1. Surface Subjects It has long been noted in the literature that PP extraposition is allowed only from D-structure object positions (the so-called “Unaccusative Hy- pothesis”) (Coopmans and Roovers (1986), Johnson (1985) and Nakajima (1995), among others). Hirata (1996) generalizes this observation and puts forward the following generalization: (19) Extraposition is possible only from D-structure X0-sister posi- tions. (Hirata (1996: 97)) Recall (1b) and (2b). Extraposition is possible in these examples because both [a review] and [a book] are in V-sister positions in D-structure. Since (19) is straightforward for extraposition from objects such as (2b) (but see 4.1.2), we will focus on extraposition from subjects in this subsection. The generalization says that only unaccusative verbs (including passives) and er- gative verbs allow extraposition. This is shown by the following examples:

15 In the rest of this paper, unless otherwise mentioned, we assume that an extraposed phrase is adjoined to VP for the purposes of our illustration. 306 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

(20) a. * [A man ] hit Mary [PP with hostility toward her]. b. * [A new book ] has attracted many people [PP about the origin of human language]. (Nakajima (1995: 21))

(21) a. * [A man ] whispered [PP from Nuie]. b. * [A man ] ran [PP from the EPA]. (Johnson (1985: 109)) (22) a. [A man ] was called [PP with blond hair]. b. [A book ] was published [PP by Updike]. (Hirata (1996: 94))

(23) a. [Men ] appeared [PP from Tanzania]. b. [Books ] arrived at the store [PP about Hammett’s life]. (Johnson (1985: 111))

(24) a. [A storm ] followed [PP from the north]. b. [A picture ] stands in the hallway [PP by Picasso]. (Johnson (1985: 111)) The subjects of transitive verbs and unergative (“true” intransitive) verbs in (20) and (21) are D-structure subjects and are not sisters to V or any other head in D-structure. On the other hand, the subjects of passives, unac- cusative and ergative verbs in (1b), (22), (23) and (24) are all D-structure objects and are sisters to V. How is this restriction on extraposition explained? Nakajima (1995) claims that an extraposed PP moves out of the internal arguments of unac- cusative and ergative verbs before the arguments raise to [Spec, TP] as illus- trated in (25). Extraposition takes place from the object position, not from the subject position, and hence does not violate the Subject Condition (4).

(25) [VP [V′ V [DP … ti …]] EXPi]

This analysis, however, also predicts that extraposition is possible with non-unaccusative verbs. Given the predicate or VP-internal subject hypoth- esis, the external arguments (subjects) of non-unaccusative verbs are Merged to [Spec, v*P] and then raise to [Spec, TP]. If extraposition is possible when internal arguments are in the D-structure object position, then it would also be possible when the external arguments of non-unaccusative verbs are in [Spec, v*P] and the examples in (20) and (21) would be predicted to be grammatical. But this is empirically incorrect. Some may argue that extraposition from [Spec, v*P] violates the Subject Condition because [Spec, v*P] is the “D-structure subject” position. This argument, however, does not go through. There is much evidence from a number of languages showing that sub-extraction from the subject of a transitive sentence is grammatical if it stays in [Spec, v*P]. For instance, consider German ex- EXTRAPOSITION AND CYCLIC SYNTAX 307 amples in (26):

(26) a. [CP Wasj haben [TP denn [vP [tj für Ameisen] what have indeed for ants einen Postbeamten gebissen]]]? a postman bitten ‘What kind of ants have bitten a postman?’

b. * [CP Wasj haben [TP [tj für Ameisen]i denn [vP ti einen Post- beamten gebissen]]]? (Diesing (1992: 32)) Subject raising is optional in German (Grewendorf (1989) and Wurmbrand

(2006), among others). In (26a), the transitive subject [tj für Ameisen] re- mains in [Spec, v*P], which is indicated by the fact that it is placed to the right of the sentential adverb denn ‘indeed.’ In this case, the sub-extraction of was ‘what’ is grammatical, unlike in (26b), where the transitive subject raises to [Spec, TP]. The German data in (26a) show that Nakajima’s anal- ysis cannot distinguish between grammatical and ungrammatical examples of PP extraposition and spoil the generalization (19). Furthermore, as we have already argued in section 2, it is unlikely, especially in the perspective of the MP, that movement is relevant to extraposition. The central claim in this section is that our analysis in the last section, with its interaction with derivational syntax, explains the generalization (19). We will take (20a) and (21a) as our examples, which are repeated below as (27a) and (27b), respectively:

(27) a. * [A man ] hit Mary [PP with hostility toward her]. b. * [A man ] whispered [PP from Nuie]. Let us begin with (27a). Given that extraposition is base-adjunction, suppose that the PP [with hostility toward her] is extraposed to VP in the derivation of (27a). However, this will not satisfy the mutual c-command requirement in the MCP and extraposition interpretation will not be estab- lished in the semantic component: the PP cannot c-command [a man] from its VP-adjoined position since the subject is Merged in [Spec, v*P] and is higher than the PP:

(28) [v*P a man [v*′ v* [VP [V′ hit Mary] with hostility toward her]]] (*with hostility toward her > a man) But consider another possibility available in a theory with a layered VP structure: that is, the PP is base-adjoined to v*P, not to VP. In this alterna- tive analysis, the PP can c-command [a man] in v*P from its adjoined posi- tion. In the course of the derivation, the subject raises to [Spec, TP] for Case/φ-feature checking and will in turn c-command the PP from its derived position. Then it would be expected that the mutual c-command required 308 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009) for extraposition interpretation is established between [a man] and [with hostility toward her] in the semantic component, and (27a) would be well- formed under the intended interpretation. It is argued that the ill-formedness of (27) follows from the dynamics of derivations. We claim that the mutual c-command relation must be estab- lished locally within the same single phase. In other words, the MCP must apply cyclically at phases via cyclic Transfer in the course of the deriva- tion. Our claim is summarized as follows: (29) The Cyclic Application of the MCP Extraposition interpretation is obtained iff an extraposed phrase and its host are interpreted in the same phasal domain through mutual c-command. Recall that the mutual c-command required for the proper interpretation of an extraposed phrase is a condition in the semantic component. Given that semantic and phonological interpretations are cyclically executed via cyclic Transfer as derivations proceed without the mediation of distinct levels like LF and PF (Chomsky (2000, 2001, 2004), Uriagereka (1999)), it follows from this that extraposition interpretation through the MCP must come cy- clically phase by phase and that the mutual c-command must be established within the same phase or TF (Transferred) domain, in which semantic and phonological interpretations are cyclically determined and assigned to Trans- ferred representations. Thus, (29) falls out as one natural consequence of the Minimalist model of derivations. In well-formed examples such as (13a, b), the extraposed PPs and their hosts mutually c-command each other in the same TF domains (i.e. CP and VP, respectively, under conventional cyclic Transfer), and extraposition interpretation is successfully obtained in the semantic component, as we have discussed. On the other hand, this is not the case in (27a), for the subject and the extraposed PP are both Transferred at CP phase (Chomsky (2004, 2007, 2008)) and the mutual c-command is established in this domain when the PP is adjoined to v*P, which would make a wrong prediction that (27a) is well-formed. In order to explain the ill-formedness in question, we will first discuss phases and the mode of cyclic Transfer as the preliminaries to our explanation.

4.1.1.1. Phases and Cyclic Transfer In this subsection, we will propose a more derivationally oriented phase theory, which has a consequence that a phase as a whole, including its edge and its head, is Transferred in the next higher phase; the phase edge and the EXTRAPOSITION AND CYCLIC SYNTAX 309 head are not Transferred as part of the next higher phase. It has been as- sumed that C and v* are (strong) phases. In the traditional cyclic Transfer (Chomsky (2000, 2001)), a phase as a whole is assumed to be Transferred upon the Merge of the next higher phase head, at which point its comple- ment domain becomes invisible for the remainder of the derivation due to the PIC (Phase Impenetrability Condition), with only the edge and head of a phase accessible to computation in the immediately higher phasal domain for head raising and successive-cyclic movement.16 In Chomsky (2004, 2007, 2008), on the other hand, only the complement of a (strong) phase head is Transferred, with the effects of the PIC being derived from Transfer itself. Thus, the non-complement positions are Transferred as part of the next phase. We will argue that Chomsky’s original view of phasal Transfer is correct but that the PIC follows derivationally, as proposed in his later work. Adapting to the present theoretical contexts the line of research pursued in Atkinson (2000), Chomsky (2000), Felser (2004), Svenonius (2004), among others, which maintains that uninterpretable feature checking/valua- tion plays a role in the construction of phases, we propose that a phase is derivationally enhanced as such through feature checking by a phase head; otherwise, it is repressed. Under this proposal, enhanced phases are con- sidered as “strong” phases in that C and v* engage in feature checking with uninterpretable features. On the other hand, repressed phases are regarded as “weak” phases in that a non-transitive light verb v, even though it is a phase head, is free from feature checking. This proposal derives from the availability of feature checking the strong/weak distinction of phases in the course of the derivation, and there is no strong/weak specification of phase heads inherent in the lexicon: phase heads are phase heads and there is no inherent v*/v distinction.17

16 The PIC is stated as follows: (i) In phase α with head H, the domain of H is not accessible to operations out- side α, only H and its edge are accessible to such operations. (Chomsky (2000: 108)) 17 Under this proposal, the strong/weak distinction of v is not determined by a full argument structure with a specific theta role (agent or experiencer). This consequence is a welcome advantage in that the association of a full argument structure with a strong phase has been called into question (Epstein and Seely (2002, 2006)). This proposal opens up a possibility that the strong/weak distinction extends to other phases such as C. In fact, it has been suggested in the literature that C can be either strong or weak, just like v (Kanno (2008), Ndayiragije (2005), and Ura (2007)). We just 310 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

Given this derivational enhancement/repression theory of phase, it follows that unlike the conventional approaches to cyclic Transfer, the edge and head of the lower phase need not be exempted from being inaccessible to the computation outside this phase. Consider this with the following sche- matic structure, where αP and βP are phases:

(30) [αP α … [βP XP [β′ β YP]]] In (30), αP-phase is neither strong nor weak before its feature checking: αP is a strong phase only after α is derivationally enhanced via its feature checking; otherwise it is a weak phase. This has a consequence that not only β and its edge XP but also its complement YP is visible, hence acces- sible to α, for the Transfer of the lower phase does not come immediately upon the Merge of α, and there is no need to distinguish a phase edge and a phase head from a phase complement as an interface to the next higher phase. These domains can also be made invisible to further computation together with the complement of β after αP is derivationally enhanced as a strong phase. Thus, the proposed theory of phase, which derivationally follows from the mechanism of feature checking, enables phase categories as a whole to be Transferred to the interfaces even when they are non-roots, and receive se- mantic and phonological interpretations. This endorses Chomsky’s original proposal on cyclic Transfer but improves it in that the PIC is eliminated; just as in Chomsky (2004, 2007, 2008), TF domains are literally sent to the interfaces, with the PIC effects derived from Transfer itself. In a transitive construction with v*P and CP phases, Transfer takes place as follows, where boxed areas represent TF domains: v*P, a strong phase, is Transferred at CP, the next higher strong phase (cf. fn. 17), and CP, a root strong phase, is Transferred as a default:

(31) [CP Spec [C′ C [TP Spec [T′ T [v*P Spec [v*′ v* [VP V …]]]]]]]

Transfer to SEM and PHON Transfer to SEM and PHON There are empirical and theoretical arguments for this mode of Trans- fer under the proposed phase theory. Below, we will provide such argu- ments to show that (31) is not contrived only for the purpose of extra-

note this and leave for the future the detailed investigation of the strong/weak distinction of C and other phases under the phase theory proposed here. EXTRAPOSITION AND CYCLIC SYNTAX 311 position. Empirical arguments come from the accessibility of non-edge positions. The proposal allows, for instance, a C-probe to access in the VP domain, and a phrase in this domain can be a goal for this probe without its intermediate movement to the v*P edge. This is evidenced by French examples in (32): (32) a. Qui as-tu vu? whom have-you seen ‘Who did you see?’ b. Tu as vu qui? (non-echo reading) (Bošković (2000: 54)) In French simple clauses, wh-movement is optional. The absence of this movement in (32b) is shown by (i) the lack of Subject-Aux (= (33a)) and (ii) the availability of a single-pair answer (= (34a)), which is unavailable in wh-movement (= (34b)): (33) * As-tu vu qui? have-you seen whom (34) a. Il a donné quoi à qui? (single-pair answer available) he has given what to whom ‘What did he give to whom?’ b. Qu’a-t-il donné à qui? (pair-list answer only) (Bošković (1998: 18)) Thus, we can conclude that qui in (32b) remains in its in-situ position throughout the derivation. On the reasonable assumption that an inter- rogative C Agrees with qui in Q-feature in both (32a) and (32b), the French example (32b) suggests that the VP domain is visible to this probe and that the C-probe directly Agrees with the in-situ wh-object as shown in (35). Under our proposal, this is straightforward because v*P is Transferred only after C achieves its feature convergence through its feature checking and is enhanced as a (strong) phase:

(35) [CP C [TP tu as-T [v*P v* [VP vu qui]]]] Agree (C, qui) Another piece of evidence for (31) is provided by Passamaquoddy. Con- sider (36), in which the embedded clause is an indirect question:18

18 The abbreviations used in the glosses are the following: 1, 3 = first, proximate third person, App = Applicative, Conj = Conjunct, Dir = Direct voice, Emph = Emphatic par- ticle, Inv = Inverse voice, Neg = Negative, Obv = Obviative, Perf = Perfect, Ta = Transi- tive verb with animate object, Pl = Plural. 312 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

(36) a. Tihtiyas ma=te wewitaham-a-wiy-il wen-il Tihtiyas Neg=Emph remember-Dir-Neg-Obv who-Obv amsqahs kis-aqosom-uw-iht kiwhosu. first Perf-cook-App-3.Conj.Inv muskrat.Obv.Pl ‘Tihtiyas doesn’t remember who first cooked muskrat for her.’ b. N-kosiciy-a-k keq nuhuw-ok muqinuw-ok 1-know.Ta-Dir-3.Pl what three-3.Pl bear-3.Pl kis-temu-htit. Perf-eat-3.Pl.Conj ‘I know what the three bears ate.’ (Bruening (2001: 177)) In this language, a certain class of verbs shows morphological object agree- ment with the embedded subject. This suggests that the matrix v* Agrees with the embedded subject in φ-features in the syntax. In (36a), this Agree relation is straightforward: since the embedded wh-subject moves to the edge of the embedded CP, it is accessible to this probe under conventional as well as our proposals on Transfer. Now consider (36b). This example demonstrates that the matrix v* Agrees with the embedded non-wh-subject across the CP boundary. It is reasonably considered that the subject remains in [Spec, TP] (non-edge) in the embedded clause, which is signaled by keq in [Spec, CP] (edge). This Agree relation is ruled in by (31): the embedded CP phase, which includes TP, is still accessible to the matrix v* because the phase head is not en- hanced as a strong phase before its feature checking and the CP phase has not been Transferred. Thus the matrix v* can see the TP domain and can Agree with the embedded subject.19 The non-edge accessibility shown by the above examples is correctly ruled in by (31) but is ruled out by the traditional proposals on cyclic Transfer because VP and TP domains are rendered invisible to higher probes. Thus, empirical support is given to (31). The mode of Transfer entailed by the proposed phase theory is also theoretically advantageous in the following ways. First, as we have already mentioned, it can derive the effects of the PIC and can eliminate it from the grammar. Second, it solves the problem of Transferring to the interfaces

19 For one possible solution to how this Agree relation evades a potential intervener keq, which shares φ-features with the embedded subject, see among others, Obata and Epstein (2008). EXTRAPOSITION AND CYCLIC SYNTAX 313 non-phase categories such as TP and VP, which are not interface-motivated (Abels (2003), Epstein and Seely (2006)). Thirdly, it removes an asymme- try between root and non-root phases in that phases are always Transferred in full. Finally, it strengthens phase-based cyclic derivations by removing a loophole in edge and head positions within phases and warranting the integ- rity of phases. From these arguments, we can conclude that (31) is empirically and theo- retically favorable over the previous approaches to cyclic Transfer.20, 21

4.1.1.2. The MCP and Cyclic Transfer With the proposed theory of phase and cyclic Transfer in mind, we now go back to (27a), in which an extraposed PP is adjoined to v*P.22 Its step- by-step derivations are summarized in (37) below. In the course of the der- ivation, transitive v*P as a whole is Transferred at CP phase, and CP phase, since it is a root, is Transferred as a default (= (37c, d)):

(37) a. [v*P [v*′ [a man] v* [VP hit Mary]] with hostility toward her] b. [CP C [TP [a man]i T [v*P [v*′ ti v* [VP hit Mary]] with hostility toward her]]]

20 See also Fox and Pesetsky (2005), who independently argue for this mode of Trans- fer based on cyclic linearization. If their discussion is on the right track, it provides evidence for (31) from the phonological interface. 21 An anonymous reviewer has pointed out that if a strong phase as a whole is Trans- ferred as we have proposed, a moved XP in the next higher phase and its copy in the edge of the immediately lower phase will not be Transferred together and that the move- ment relation (dependency) will not be interpreted within a single phase (TF domain) in the semantic component. This, he/she argues, contrasts with extraposition, whose relation, as we have claimed, must be established in the same phase for its interpretation (see (29)), saying that this requires an unwanted stipulation. It is argued that this is not a problem and that the difference naturally follows from the nature of movement. Pro- vided that a copy of L is identical to L except that it lacks the phonological features of L (Chomsky (2004: 111)), a moved XP and its copy can be interpreted as identical and the movement relation will thus be established in the semantic component even if the two are Transferred in different phases. On the other hand, in the case of extraposition, an extra- posed phrase is not identical to its host and consequently, for its proper interpretation, the two must be Transferred and interpreted in the same phase. 22 Recall that v* is simply v in the lexicon and is derivationally enhanced as a strong phase via feature checking. For simplicity and illustration, we use v* for transitive sen- tences from the beginning. 314 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

c. [CP C [TP [a man]i T [v*P [v*′ ti v* [VP hit Mary]] with hostility toward her]]] → TF of v*P (only with hostility toward her > a man)

d. [CP C [TP [a man]i T [v*P [v*′ ti v* [VP hit Mary]] with hostility toward her]]] → TF of CP (*a man > with hostility toward her) This derivation, however, does not result in the intended extraposition inter- pretation: mutual c-command will not be established between [a man] and [with hostility toward her] in the semantic component, which is due to the fact that the two phrases are cyclically Transferred in different TF domains (v*P and CP, respectively) and thus interpreted separately. True, [a man] may c-command the adjoined PP in CP phase thanks to its movement to [Spec, TP] and the two phrases c-command each other in the whole repre- sentation. However, they fail to do so in the same single TF domain, in which semantic and phonological interpretations are cyclically determined and extraposition interpretation is computed by the MCP. By the time CP is Transferred, v*P, which contains the extraposed PP, has already been se- mantically and phonologically interpreted for its earlier Transfer. Thus, the MCP is not cyclically satisfied. The same argument applies to unergative (true intransitive) sentences like (27b), on the assumption that unergatives are hidden transitives and are derived through the incorporation of a noun in the object position into V (Hale and Keyser (2002)). In the present contexts, the unergative structure is translated into the one with transitive v*, which is to be enhanced as a strong phase for feature checking and is hence a point of cyclic deriva- tions. Thus, whether [from Nuie] is adjoined to VP or to v*P, extraposition interpretation will not be obtained for the absence of mutual c-command between [a man] and the PP in the same TF domain.23 On the other hand, in unaccusative/non-transitive examples such as (23) and (24), an extraposed phrase and its host (the subject) can mutually c- command each other in the same TF domain. vP is not independently

23 We have standardly assumed that arguments are Externally Merged before adjuncts are Merged (adjoined). However, even if the PP in (27) is adjoined to v*P before the Merge of the subject as in (i), this does not raise a problem for our analysis: (i) [v*P Subj [v*′ [v*′ v* [VP V …]] PP]] Since the extraposed PP is Merged lower than the subject and hence cannot c-command it in v*P, mutual c-command cannot be established between the two phrases in this TF domain, and (29) is not satisfied. EXTRAPOSITION AND CYCLIC SYNTAX 315

Transferred at CP phase because v is not a strong phase for the lack of feature checking and fails to be a point of cyclic Transfer; there is only one phase or TF domain: CP. Thus, the MCP is cyclically satisfied. This is also confirmed by long distance raising examples like (38), where extraposi- tion is possible across raising infinitives:24 (38) a. A review seems to have come out yesterday of this article. b. A review seemed to appear to be certain to come out yester- day of this article. Since vPs are weak phases and the matrix CP is the only TF domain in (38), the raised DP [a review] and the extraposed PP [of this article] can mutually c-command each other in this TF domain. The well-formedness of (38) is correctly explained by our analysis. In this subsection, we have discussed the generalization (19) with extra- position from subjects and have claimed that the MCP, with its interaction with the cyclic syntax, gives a principled explanation to (19).

4.1.2. Surface Objects As we have mentioned, the generalization (19) is straightforward for objects. It has been noted, however, that extraposition from indirect ob- jects, unlike extraposition from direct objects, is disallowed (Takami (1990), which is further developed in Hirata (1996)). Consider extraposition in the Double Object Construction (DOC) in (39) and (40): (39) a. I sent Mary [a book ] yesterday [by an American novel- ist]. b. I gave Mary [a puppy ] yesterday [with big cute ears]. (Takami (1990: 198)) (40) a. * I sent [a man ] an interesting book yesterday [with blond hair]. b. * I gave [a man ] an interesting book yesterday [with green eyes]. (Takami (1990: 198)) We demonstrate that this contrast naturally follows from our proposal. Be- fore we do this, we will spell out the structure of the DOC. Richards (2007) argues that given the current model of syntactic deriva- tions incorporating phases (Chomsky (2007, 2008)), the basic unit of phrase structure should be the one in which one phase head, maximally and mini- mally, pairs with one non-phase head, as in C-T and v*-V, with no succes-

24 I thank an anonymous reviewer for directing me to these data. 316 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009) sive co-occurrence of either phase heads or non-phase heads. Given this Minimalist view of phrase structure, the structure of the DOC can be updat- ed as follows (cf. Larson (1988), Marantz (1993) and McGinnis (2004)):25, 26 (41) v*P

Subj v*′

v* VP

VApplP

Indirect ObjAppl′ (goal/source) Appl VP

VDirect Obj (theme)

In (41), Applicative (Appl) is a functional, hence phase head selecting lower VP, forming a phase + non-phase pair with this VP (Appl-V), and has un- interpretable φ-features and Case feature for feature checking with a direct object. This head selects an indirect object and assigns a goal/source theta role, just like v* selects an external argument and assigns an agent theta role. On the other hand, v* forms a phase structure with upper VP (v*-V) and has uninterpretable φ-features and Case feature, as commonly assumed in transitive structures. Given this background, let us take (39a) and consider extraposition from direct objects. Suppose that the extraposed PP [by an American novelist] is base-adjoined to lower VP, which yields the VP structure in (42a). When Appl is Merged with this VP, its uninterpretable features are inherited onto V and start searching, as a result of which they Agree with the direct object [a book] (= (42b)). In the next step, this object is Internally Merged with the VP and moves to [Spec, lower VP] for its uninterpretable Case in order

25 For arguments that goal/source is Merged higher than theme as assumed in (41), see Hirata (1996) and references cited therein. 26 We follow McGinnis (2004) in assuming that the relevant functional head in the DOC is Appl, which is morphologically realized in Austronesian and Bantu languages. EXTRAPOSITION AND CYCLIC SYNTAX 317 to probe its counterpart in V (= (42c)). The first phase that is cyclically Transferred is ApplP, which is Transferred when v*P is enhanced as a strong phase. It should be noted that the extraposed phrase and the direct object mutually c-command each other in this TF domain thanks to the movement of the object, just as in (13b) (= (42d)). This explains why extraposition (interpretation) is possible in (39a): the MCP is satisfied in the same single phase as required in (29).

(42) a. [VP [V′ sent-V [a book{uCase}] yesterday] by an American nov- elist]

b. [ApplP Mary Appl [VP [V′ sent-V{uφ} [a book{uCase}] yesterday] by an American novelist]]

c. [ApplP Mary Appl [VP [a book{uCase}]i [V′ [V′ sent-V{uφ} ti yester- day] by an American novelist]]]

d. [v*P v* … [ApplP Mary Appl [VP [a book{uCase}]i [V′ [V′ sent- V{uφ} ti yesterday] by an American novelist]]]] → TF of ApplP (a book >< by American novelist) (39b) shares the derivation in (42) and exactly the same explanation applies to this example. Let us now turn to the ill-formedness of (40) and consider why extra- position from indirect objects is impossible. We take (40a) as our exam- ple. There are a number of possibilities for where the PP [with blond hair] is adjoined. For illustration, suppose that it is adjoined to ApplP, which generates the partially constructed structure in (43a) (irrelevant details are omitted for simplicity). In this structure, the extraposed PP can c-command [a man] from its adjoined position. Then upper V is Merged to ApplP and then v* is Merged to this upper VP. At this point, the uninterpretable φ-features and Case feature in v* are inherited onto upper V and the V Agrees with [a man]. The DP moves to [Spec, upper VP] for Case feature checking. This yields the structure (43b). This movement, however, has the effect that the extraposed PP and its host are Transferred in different TF domains and hence interpreted separately: the former is Transferred in Ap- plP (= (43b)) and the latter in v*P (= (43c)). As a result, [a man] and [with blond hair] do not mutually c-command each other in the same single TF domain, and the mutual c-command will not be established cyclically in the semantic component. Thus (29) is not satisfied and the intended extraposi- tion interpretation cannot be obtained. 318 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

(43) a. [ApplP [Appl′ [a man] Appl [VP sent an interesting book yester- day]] with blond hair]

b. [v*P I v* [VP [a man]i V [ApplP [Appl′ ti Appl [VP sent an inter- esting book yesterday]] with blond hair]]] → TF of ApplP (only [with blond hair] > [a man])

c. [CP C [TP Ij [v*P tj v* [VP [a man]i V [ApplP [Appl′ ti Appl [VP sent an interesting book yesterday]] with blond hair]]]]] → TF of v*P (*[a man] > [with blond hair]) The proposed analysis of extraposition, with its interaction with the cyclic syntax, explains the ill-formedness of extraposition in (40a). (40b) is ac- counted for in the same way. In the derivation summarized in (43), the extraposed PP is base-adjoined to ApplP. It should be noted that other possible adjunction sites (say, up- per VP and v*P) also lead to the failure of mutual c-command between [a man] and [with blond hair] in the same TF domain. If the PP is adjoined to upper VP, it is c-commanded by [a man] moved to [Spec, upper VP] in the v*P domain. However, it cannot c-command the DP from its adjoined position in this TF domain. On the other hand, if it is adjoined to v*P, it can c-command the moved DP in the relevant Spec but it cannot be c- commanded by this DP within v*P. Thus, even though [a man] and [with blond hair] are Transferred in the same TF domain, extraposition interpre- tation will not be computed in the semantic component for the absence of mutual c-command. Thus, given the structure of the DOC in (41), which follows from inde- pendent Minimalist assumptions on phrase structure, the cyclic applications of the MCP can explain not only why extraposition from indirect objects is disallowed but also the parallelism between indirect objects and transitive subjects in the impossibility of extraposition from these positions. To summarize our discussion in 4.1, we have claimed that the MCP, with its interaction with cyclic Transfer in the phase-based derivational syntax (that is, (29)), can give a principled answer to the generalization (19). We have argued that impossible extraposition examples follow from the failure of mutual c-command in the same TF domain. On the other hand, in pos- sible cases of extraposition, an extraposed PP and its host DP mutually c- command each other in the same TF domain. In our discussion, we have also proposed a more derivationally oriented phase theory, as a consequence of which phases as a whole are Transferred (as originally proposed in Chomsky (2000, 2001)). EXTRAPOSITION AND CYCLIC SYNTAX 319

4.2. Extraposition Is Upward Bounded In this subsection, we will discuss another curious property of extraposi- tion, which we have already noted in 2.2: extraposition is upward bound- ed. Unlike wh-movement, extraposition cannot be unbounded and long distance extraposition is disallowed. Recall examples in (5), which are repeated as (44): (44) a. * [Everyone believes [that [that a review will come out] is cer- tain] of this article]. b. * [Everyone believes [that [that a review will come out] is cer- tain of this article]]. c. [Everyone believes [that [that a review will come out of this article] is certain]]. To explain this restriction, Ross, adopting a movement analysis, proposes that S is a boundary (or cyclic node) that restricts extraposition. Akmajian (1975) argues that not only S but also NP is such a cyclic node and pro- poses (45) as a condition on extraposition: (45) No element may be extraposed more than one cycle up from the cycle containing it. (Akmajian (1975: 119)) This condition accounts for (46) and (47) below as well as (44). In (46a), the PP [about French cooking] cannot be extraposed from the DP [a new book about French cooking] because this DP is further embedded in another DP and two cyclic nodes must be crossed for extraposition. On the other hand, in (46b), the PP crosses only a single cyclic node:

(46) a. * [DP A review of [DP a new book ]] came out yesterday [about French cooking].

b. [DP A review ] came out yesterday [of a new book about French cooking]. (Akmajian (1975: 118)) In (47a), in which [by three authors] is extraposed, the sentence is unam- biguous, allowing only the interpretation in which three authors reviewed a book ([by three authors] cannot modify the embedded DP [a book] but the whole DP [a review of a book]). The other reading (three authors wrote a book and the review is of that book) is impossible, which is because the PP would have to cross two DP boundaries for this interpretation to be ob- tained. On the other hand, (47b) is ambiguous, having the two readings (the PP [by three authors] can ambiguously modify the DP [a book] or the whole DP [a review of a book]): (47) a. A review of a book appeared last year [by three authors]. b. A review of a book [by three authors] appeared last year. (Akmajian (1975: 122)) 320 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

We claim that the condition (45) straightforwardly follows from the pro- posed analysis of extraposition. The examples above show that an extra- posed PP cannot be interpretively associated with the DP that is embedded in CP or another DP. Consider (44a, b) first. Recall that in order for extraposition interpretation to be obtained, an extraposed phrase and its host DP must mutually c-command each other in the same single TF domain. In these examples, the extraposed PP [of this article] is base-adjoined within the matrix CP (= (44a)) or within the intermediate CP (= (44b)). Given that as we have proposed in 4.1, extraposition interpretation is cyclically executed through cyclic Transfer in the course of the derivation (= (29)), the intended interpretation will be impossible because the PP and its host [a review] are Transferred in different TF domains and hence, interpreted separately. In addition, since [a review] is in the most embedded clause, it will not be able to c-command [of this article] within the higher CPs. The ill-formedness of (44a, b) is explained under our analysis. Let us now consider (46a) and (47a). We take (46a) for illustration. Its derivation is summarized in (48):

(48) a. [VP [V′ came out [a review of a new book] yesterday] about French cooking]

b. [CP C [TP [a review of a new book]i T [vP v [VP [V′ came out ti yesterday] about French cooking]]]] c. [CP C [TP [a review of a new book]i T [vP v [VP [V′ came out ti yesterday] about French cooking]]]] → TF of CP (about French cooking > a new book, *a new book > about French cooking) In the derivation of VP, the extraposed phrase [about French cooking] is base-adjoined to VP (= (48a)). From this adjoined position, it can c- command the DP [a new book] embedded in the object position. When C/T is Merged, the DP [a review of a new book] moves to [Spec, TP] for Case/φ-feature checking and becomes higher than the extraposed phrase (= (48b)). Even after this raising, however, the embedded DP [a new book], which is supposed to be the host of the extraposed PP [about French cooking], is too deeply embedded to c-command the PP. Then the mutual c-command will not be established and the extraposed PP cannot be inter- pretively associated with the embedded DP [a new book] when the whole derivation undergoes Transfer at CP phase, even though the two phrases in question are Transferred in the same phase (= (48c)). This explains the ill- formed interpretation in (46a). Exactly the same explanation applies to (47a) under the reading in which EXTRAPOSITION AND CYCLIC SYNTAX 321

[by three authors] is intended to be extraposed from [a book]: the DP can- not c-command the adjoined PP even after the DP [a review of a book] moves to [Spec, TP]. Thus, it follows from the MCP that (47a) is unam- biguous. In this subsection, we have shown that the proposed interpretive principle (the MPC) can explain the upward bounded property of extraposition (hence, Akmajian’s condition (45)). To summarize our discussion in section 4, we have claimed that the at- tested restrictions on extraposition fall out from the MCP and its cyclic ap- plication without any unmotivated assumptions. A theoretical payoff of our proposal is that the restrictions we have considered, which used to be treat- ed as two different and separate restrictions, receive a unified derivational account: they are both deducible from the MCP executed in a derivational model of the syntax. In the course of our discussion, we have endorsed the derivational syntax incorporating phases and proposed a phase theory in which feature checking plays a crucial role. It should be noted that the discussion in this section not only provides strong evidence for the syntax interacting with interface components cycli- cally, phase by phase; but it also argues against the view that interpretive procedures cyclically come into play upon the application of Merge/Move as proposed in Epstein et al. (1998) and Epstein and Seely (2002, 2006). As we have discussed, extraposition interpretations must be calculated deriva- tionally at phases.27, 28

27 CP extraposition noted in fn. 1 shows different properties from PP extraposi- tion. For example, CP extraposition is possible with transitive verbs, which suggests that (29) is irrelevant and hence that a different analysis is required. Consider (i) (cf. (20)): (i) a. [A man ] hit Mary [CP who had hostility toward her]. b. [A new book ] has attracted many people [CP which is concerned with the origin of human language]. (Nakajima (1995: 21)) We follow Nakajima (1995) and assume that CP extraposition is mediated by a predi- cation relation between an extraposed CP, which is base-adjoined to TP, and its host DP. Given that predication is available for CP extraposition thanks to an operator in an extraposed CP, the difference in question follows, and CP extraposition is possible even without (29). 28 An anonymous reviewer has asked whether the MCP and its cyclic application can extend to other “rightward” phenomena such as heavy NP shift, right-dislocation and sec- ondary predicates, which seem to share with PP extraposition some of its properties. For example, a heavy-NP-shifted object must enter into a sisterhood relation with the verb selecting it for its interpretation, and right-dislocated phrases and secondary predicates are base-adjoined to XP, which suggests that the MPC and its cyclic application could ex- 322 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

5. Sub-extraction from Subjects Before we conclude this paper, we argue in this section that the base-ad- junction analysis of extraposition can give a natural account to well-formed sub-extraction from subjects. It has been observed that sub-extraction from subjects is possible with pied-piping (Chomsky (1986, 2008), Kuno and Takami (1993), Ross (1967)). Consider (49) and (50): (49) a. Of which article did a review come out yesterday? b. It was these cars of which the hoods were damaged by the explosion. (50) a. Of whom was the biography published yesterday? b. It was the boy (and not the girl) of whom a story appeared in the newspaper.

(51) a. * Which articlei did [DP a review of ti] come out yesterday? b. * Whomi was [DP the biography of ti] published yesterday? In contrast with (51), the Subject Condition is circumvented if a preposition is sub-extracted together with DP. Note that D-linking is irrelevant to the above contrast, which is suggested by (50) and (51a). Well-formed sub-ex- traction from subjects is straightforward under our analysis: a sub-extracted PP is extraposed, hence base-adjoined to VP and is moved to [Spec, CP] from its adjoined position, as illustrated in (52):

(52) [CP Of which articlej C [TP [DP a review]i T [vP v [VP [V′ come out

ti yesterday] tj]]]]

Since the movement in question does not take place from the subject in [Spec, TP], it does not violate the Subject Condition. In other words, the movement of [PP P Wh] in (49) and (50) is analyzed on a par with wh- movement in (53) and (54), where adjoined wh-adverbs and adjoined wh- PPs move without any problem:

tend to these cases. The details are still unclear, however, and other properties must be considered as well. We must leave for the future the investigation of these phenomena under the proposed framework. I thank the reviewer for directing me to the discussion of other related phenomena. EXTRAPOSITION AND CYCLIC SYNTAX 323

(53) a. Howi did you fix the car ti? b. Whyi did you go there ti? (54) a. [In which park]i did they play tennis ti? b. [With whom]i did you go to the museum ti? Thus, the base-adjunction analysis explains apparent violation of the Sub- ject Condition in examples like (49) and (50), and a long-standing observa- tion since Ross (1967) that sub-extraction and pied-piping are correlated.29 If grammatical sub-extraction from subjects is due to extraposition (base-adjunction), then it is predicted that the sub-extraction is illicit under those circumstances where extraposition is disallowed. This prediction is borne out. We have seen two restrictions on extraposition in the last sec- tion. Consider, first, extraposition from D-structure0 X -sister positions. As the following examples clearly illustrate, sub-extraction from subjects is disallowed when predicates do not select surface subjects as D-structure ob- jects: (55) a. * It was the CAR (not the TRUCK) of which the (driver, pic- ture) caused a scandal. b. * Of which car did the (driver, picture) cause a scandal? cf. c. * The (driver, picture) caused a scandal of the car. (Chomsky (2008: 153–154)) (56) a. * Of whom did a picture hit Mary? b. * Of whom did a friend introduce you to Bill? (Wexler and Culicover (1980: 325)) (57) a. * Of which car did the driver run in the park? b. * Of which actors would beautiful pictures cost too much? (Rochemont and Culicover (1990: 33)) The verbs employed in (55) and (56) are transitive verbs while those in (57) are unergative (“true” intransitive) verbs. In these examples, surface subjects are not D-structure objects and extraposition, as we have already argued, is ruled out by (29). Given that well-formed sub-extraction from subjects presupposes extraposition, the ill-formedness of (55)–(57) is given the same account as that of (27) and endorses an extraposition analysis of (49) and (50). Likewise, sub-extraction from indirect objects is ill-formed, which is ex-

29 For much related discussion on well-formed sub-extraction from subjects, see among others, Boeckx (2008), Chomsky (2008), and Gallego and Uriagereka (2006). 324 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009) pected given our discussion in section 4.1.2. Consider (58):30 (58) a. * Of which book did the foundation give the author a prize? b. * Of what did you give pictures a finishing touch? On the other hand, sub-extraction from direct objects is unproblematic, as expected. This is confirmed by the following examples: (59) a. Of which book did you give Mary a review? b. Of whom did you send George a picture? Next consider that extraposition is upward bounded. If the apparent vio- lation of the Subject Condition in (49) and (50) is due to extraposition, then the following examples will be interpreted in the same way that (46a) and (47a) are interpreted. Consider (60): (60) a. About which subject did a review of a new book come out yesterday? b. By how many authors did a review of a book appear last year? If our analysis is on the right track, it is predicted that the wh-counterpart of [about French cooking] in (46a) cannot be sub-extracted with the in- tended reading in which it modifies [a new book] and the sentence will be ill interpreted. Likewise, the sub-extraction of the wh-counterpart of [by three authors] in (47a) should allow only unambiguous reading, a reading in which it modifies the whole DP [a review of a book] and the other reading, a reading in which it modifies the embedded DP [a book], is ruled out.

30 Chomsky (2008: 147) points out that (i) is well-formed: (i) Of which car was the (driver, picture) awarded a prize? In (i), [of which car] is sub-extracted from the indirect object. It is argued that the well- formedness of (i) does not pose a problem to our discussion in 4.1.2; instead, it in fact endorses our proposals in 4.1.1. Suppose that [of which car] is base-adjoined to upper VP, a possibility available under free adjunction. In the derivation of (i), the indirect ob- ject, which is Merged to [Spec, ApplP], raises to [Spec, TP] for Case/φ-feature Agreement while [of which car] moves to [Spec, CP] from its adjoined position for its checking with C. Since v in (i) is a weak phase for the lack of Case/φ-feature Agreement and does not constitute a strong phase, hence is not a point of cyclic Transfer, the TF domains in (i) are CP and ApplP. We should note that in this CP phase, the indirect object c-commands (the copy of) [of which car] from [Spec, TP], which in turn c-commands [the (driver, picture)] from [Spec, CP]: (ii) [CP [of which car]i C [TP [the (driver, picture)]j T [vP v [VP [V′ V [ApplP tj Appl [VP awarded a prize]]] ti]]]] → TF of CP ([the (driver, picture)] >< [of which car]) As we can see, the MCP can be satisfied cyclically within this CP. Thus, the sub-extrac- tion in (i) is well-formed under the intended interpretation. EXTRAPOSITION AND CYCLIC SYNTAX 325

This prediction is, indeed, corroborated. The sentence in (60a) cannot ask which subject a new book that is reviewed is about, and the one in (60b) only asks the number of authors who reviewed a book; it cannot ask the number of authors who wrote a book. To summarize this section, we have argued that licit sub-extraction from subjects is naturally explained if extraposition (i.e. base-adjunction) is pre- supposed. We have validated this argument by demonstrating that sub- extraction is ungrammatical under those circumstances where extraposition is ruled out.

6. Conclusion In this paper, we have considered extraposition in the MP and have claimed that it is evidence for the cyclic syntax envisioned in this theoreti- cal framework. We have argued against a movement analysis of extraposi- tion through theoretical and empirical arguments. Instead, building on Cu- licover and Rochemont (1990), we have advocated a non-movement analysis in which an extraposed phrase is Merged as an adjunct and claimed that it is interpretively associated with its host via the MCP executed derivation- ally through cyclic Transfer at phases (= (29)). We have shown that our proposal can give a natural and unified account to the restrictions observed in extraposition. Through our discussion in this paper, we have not only demonstrated that extraposition endorses a Minimalist model of syntactic derivations, where the syntax interacts with interface components cyclically, phase by phase; we have also proposed a more derivationally oriented phase theory, in which phases are enhanced or repressed depending on the avail- ability of feature checking, as a consequence of which phases as a whole are cyclically Transferred. Finally, we have argued that the proposed analy- sis naturally accounts for well-formed sub-extraction from subjects, explain- ing why the sub-extraction is possible in some cases but not in others.

REFERENCES Abels, Klaus (2003) Successive Cyclicity, Anti-locality, and Adposition Stranding, Doctoral dissertation, University of Connecticut. Akmajian, Adrian (1975) “More Evidence for an NP Cycle,” Linguistic Inquiry 6, 115–129. Atkinson, Martin (2000) “Uninterpretable Feature Deletion and Phases,” Essex Re- 326 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

search Reports in Linguistics 34, 91–122. Baltin, Mark (1978) Toward a Theory of Movement Rules, Doctoral dissertation, MIT. Boeckx, Cedric (2008) Bare Syntax, Oxford University Press, Oxford. Bošković, Željko (1998) “Wh-Phrases and Wh-Movement in Slavic,” Position paper, Comparative Slavic Morphosyntax, Bloomington, Indiana. Bošković, Željko (2000) “Sometimes in [Spec, CP], Sometimes in Situ,” Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, ed. by Roger Martin, David Michaels and Juan Uriagereka, 53–87, MIT Press, Cambridge, MA. Bošković, Željko (2007) “On the Locality and Motivation of Move and Agree: An Even More Minimal Theory,” Linguistic Inquiry 38, 589–644. Bruening, Benjamin (2001) Syntax at the Edge: Cross-Clausal Phenomena and the Syntax of Passamaquoddy, Doctoral dissertation, MIT. Chomsky, Noam (1973) “Conditions on Transformations,” A Festschrift for Morris Halle, ed. by Stephen R. Anderson and Paul Kiparsky, 232–286, Holt, Rinehart & Winston, Inc., New York. Chomsky, Noam (1981) Lectures on Government and Binding: The Pisa Lectures, Foris, Dordrecht. Chomsky, Noam (1986) Barriers, MIT Press, Cambridge, MA. Chomsky, Noam (1995) The Minimalist Program, MIT Press, Cambridge, MA. Chomsky, Noam (2000) “Minimalist Inquiries: The Framework,” Step by Step: Es- says on Minimalist Syntax in Honor of Howard Lasnik, ed. by Roger Martin, David Michaels and Juan Uriagereka, 89–155, MIT Press, Cambridge, MA. Chomsky, Noam (2001) “Derivation by Phase,” Ken Hale: A Life in Language, ed. by Michael Kenstowicz, 1–52, MIT Press, Cambridge, MA. Chomsky, Noam (2004) “Beyond Explanatory Adequacy,” Structures and Beyond: The Cartography of Syntactic Structures, Volume 3, ed. by Adriana Belletti, 104–131, Oxford University Press, Oxford. Chomsky, Noam (2007) “Approaching UG from Below,” Interfaces + Recursion = Language? Chomsky’s Minimalism and the View from Syntax-Semantics, ed. by Uli Sauerland and Hans-Martin Gärtner, 1–29, Mouton de Gruyter, Berlin. Chomsky, Noam (2008) “On Phases,” Foundational Issues in Linguistic Theory: Es- says in Honor of Jean-Roger Vergnaud, ed. by Robert Freidin, Carlos P. Otero and Maria Luisa Zubizarreta, 133–166, MIT Press, Cambridge, MA. Coopmans, Peter and Irene Roovers (1986) “Reconsidering Some Syntactic Proper- ties of PP-Extraposition,” Utrecht Formal Parameters Yearbook II, 21–35. Culicover, Peter and Michael Rochemont (1990) “Extraposition and the Complement Principle,” Linguistic Inquiry 21, 23–47. Diesing, Molly (1992) Indefinites, MIT Press, Cambridge, MA. Epstein, Samuel D. (1999) “Un-Principled Syntax: The Derivation of Syntactic Rela- tions,” Working Minimalism, ed. by Samuel D. Epstein and Norbert Hornstein, 317–345, MIT Press, Cambridge, MA. Epstein, Samuel D., Erich M. Groat, Ruriko Kawashima and Hisatsugu Kitahara EXTRAPOSITION AND CYCLIC SYNTAX 327

(1998) A Derivational Approach to Syntactic Relations, Oxford University Press, Oxford. Epstein, Samuel D. and T. Daniel Seely (2002) “Rule Applications as Cycles in a Level-free Syntax,” Derivation and Explanation in the Minimalist Program, ed. by Samuel D. Epstein and T. Daniel Seely, 65–89, Blackwell, Oxford. Epstein, Samuel D. and T. Daniel Seely (2006) Derivations in Minimalism, Cam- bridge University Press, Cambridge. Felser, Claudia (2004) “Wh-Copying, Phases, and Successive Cyclicity,” Lingua 114, 543–574. Fox, Danny and David Pesetsky (2005) “Cyclic Linearization of Syntactic Structure,” Theoretical Linguistics 31, 1–46. Fukui, Naoki (1993) “Parameters and Optionality,” Linguistic Inquiry 24, 399–420. Gallego, Ángel and Juan Uriagereka (2006) “Conditions on Sub-extraction,” ms., Universitat Autònoma de Barcelona and University of Maryland. Grewendorf, Günther (1989) Ergativity in German, Foris, Dordrecht. Grimshaw, Jane (1991) “Extended Projections,” ms., Brandeis University. Guéron, Jacqueline (1980) “On the Syntax and Semantics of PP-Extraposition,” Lin- guistic Inquiry 11, 637–678. Guéron, Jacqueline and Robert May (1984) “Extraposition and Logical Form,” Lin- guistic Inquiry 15, 1–31. Hale, Ken and Samuel J. Keyser (2002) Prolegomenon to a Theory of Argument Structure, MIT Press, Cambridge, MA. Hirata, Ichiro (1996) “Extraposition and VP-Shell,” English Linguistics 13, 93–120. Huang, C.-T. James (1982) Logical Relations in Chinese and the Theory of Gram- mar, Doctoral dissertation, MIT. Johnson, Kyle (1985) A Case for Movement, Doctoral dissertation, MIT. Kanno, Satoru (2008) “On the Phasehood and Non-Phasehood of CP,” English Lin- guistics 25, 21–55. Kuno, Susumu and Ken-Ichi Takami (1993) Grammar and Discourse Principles: Functional Syntax and GB Theory, University of Chicago Press, Chicago. Larson, Richard (1988) “On the Double Object Construction,” Linguistic Inquiry 19, 335–391. Lasnik, Howard (1995) Minimalist Analysis, Blackwell, Oxford. Lasnik, Howard and Randall Hendrick (2003) “Steps toward a Minimalist Theory of Anaphora,” Minimalist Syntax, ed. by Randall Hendrick, 124–151, Blackwell, Oxford. Marantz, Alec (1993) “Implications of Asymmetries in Double Object Constructions,” Theoretical Aspects of Bantu Grammar, ed. by Sam Mchombo, 113–151, CSLI Publications, Stanford. May, Robert (1985) Logical Form: Its Structure and Derivation, MIT Press, Cam- bridge, MA. McGinnis, Martha (2004) “Lethal Ambiguity,” Linguistic Inquiry 35, 47–95. Nakajima, Heizo (1984) Eigo no Idoo Genshoo Kenkyu (A Study of Movement Phe- nomena in English), Kenkyusha, Tokyo. 328 ENGLISH LINGUISTICS, VOL. 26, NO. 2 (2009)

Nakajima, Heizo (1995) “Shugo kara no Gaichi: Toogoron to Goyooron no Sumi- wake (Extraposition from Subjects: Division of Labor between Syntax and Prag- matics),” Nichieigo no Uhoo Idoo Koobun: Sono Koozoo to Kinoo (Rightward Movement Constructions in English and Japanese: Their Structures and Func- tions), ed. by Ken-Ichi Takami, 17–35, Hituzi Syobo, Tokyo. Ndayiragije, Juvénal (2005) “On Phases and Cyclicity,” MIT Working Papers in Lin- guistics 49, 265–282. Obata, Miki and Samuel D. Epstein (2008) “Deducing Improper Movement from Phase-Based C-to-T Phi Transfer: Feature-Splitting Internal Merge,” WCCFL 27, 353–360. Pollock, Jean-Yves (1989) “Verb Movement, Universal Grammar and the Structure of IP,” Linguistic Inquiry 20, 365–425. Richards, Marc D. (2007) “On Feature Inheritance: An Argument from the Phase Im- penetrability Condition,” Linguistic Inquiry 38, 563–572. Rochemont, Michael and Peter Culicover (1990) English Focus Constructions and the Theory of Grammar, Cambridge University Press, Cambridge. Ross, John R. (1967) Constraints on Variables in Syntax, Doctoral dissertation, MIT. Saito, Mamoru and Naoki Fukui (1998) “Order in Phrase Structure and Movement,” Linguistic Inquiry 29, 439–474. Svenonius, Peter (2004) “On the Edge,” Peripheries: Syntactic Edges and Their Ef- fects, ed. by David Adger, Cécile de Cat and George Tsoulas, 259–287, Kluwer, Dordrecht. Takami, Ken-Ichi (1990) “Remarks on Extraposition from NP,” Linguistic Analysis 20, 192–219. Ura, Hiroyuki (2007) “Long-Distance Case-Assignment in Japanese and Its Dialectal Variation,” Gengo Kenkyu 131, 1–43. Uriagereka, Juan (1999) “Multiple Spell-Out,” Working Minimalism, ed. by Samuel D. Epstein and Norbert Hornstein, 251–282, MIT Press, Cambridge, MA. Wexler, Ken and Peter Culicover (1980) Formal Principles of Language Acquisition, MIT Press, Cambridge, MA. Wurmbrand, Susi (2006) “Licensing Case,” Journal of Germanic Linguistics 18, 175–236.

[received September 19 2008, accepted April 8 2009]

Department of General Education Nagano National College of Technology 716 Tokuma Nagano-shi Nagano 381–8550 e-mail: [email protected]