VERB MOVEMENT AND INTERROGATIVES

O. lntroduction Verb movement is a phenomenon that has been studied extensively within the framework of Chomskyan generative grammar. The pioneering work by Pollock (1989) has been followed by a number of studies involving various languages, which has provided an important insight both into the language-specific and language-universal properties of verb movement. In most general terms, verb movement can be defined as movement of the verb from its base position in the (V)erb (P)hrase to some position higher in the clausal structure. In Government & Binding theory verb movement was motivated by the need of the bare Iexical verb to associate with the inflectional affixes hosted by the functional heads (Pollock 1989, Belletti 1990). By contrast, the Minimalist Program (Chomsky 1995) claims that ali types of movement are triggered by feature-checking requirements. In this system, items from Iexical categories are fully inflected in the lexicon. Thus the verb is inserted into its base position with ali its inflectional affixes and associated inflectional features. Functional heads do not contain any inflectional material; they carry only abstract features, which are checked against the corresponding features on the Iexical items. In order for feature-checking to take place the Iexical item ( e.g. the verb) must raise to the relevant functional head(s). This paper is an attempt at a syntactic account of the type of verb movement displayed in interrogative clauses containing a wh-element. In the generative literature, this type of movement, standardly known as 'I-to-C movement', has been related to the Wh-Criterion (Rizzi 1991 ). It has been claimed that the inflected verb must raise from I(nflection) to C(omplementizer) so as to be in a Spec(ifier)-head configuration with the wh-element in the specifier of the C(omplementizer) P(hrase). In this paper, verb movement in wh-interrogatives 1 in English, French and Slovenian will be examined from a comparative perspective, with special attention being paid to the following issues: (i) what are the general properties of verb movement in wh-interrogatives; (ii) how can this type of movement be analysed by adopting the basic concepts and tenets of the Minimalist Program (Chomsky 1995); and (iii) to what extent do the empirical observations follow from the general principles of the Minimalist Program. The paper is organised as follows. Section 1 provides a brief outline of the Minimalist theoretical framework adopted in this examination. Section 2 deals with verb

movement in English and French interrogatives with special emphasis on the familiar of root/embedded asymmetry of 1-to-C movement.In section 3 we focus on interrogative 1-to-C in Slovenian and propose an analysis of 1-to-C movement in terms of a modified Minimalist checking theory of movement which provides a uniform account of different properties of interrogative 1-to-C in languages of the English/French, Spanish and Slovenian type.In Section 4 we offer our conclusions on the research.l.Theoretical background

The Minimalist Program: basic concepts and assumptions
Minimalism is an attempt to reduce the theory of grammar to the essential three components any theory of grammar 2 must have: a lexicon, an interface with the articulatory-perceptual system and an interface with the semantic-conceptual system.The Minimalist Program thus postulates only two levels of structural representation, the interface levels of Phonetic Form (PF) and Logical Form (LF)3, a lexicon and a computational system which builds structures that constitute PF and LF representations.The Minimalist model is shown in (1 ). (1) Lexicon The structure-building process starts with Numeration -a set of lexical items from which a structure is to be formed.The computational system builds up structures in a step-by-step fashion, by selecting elements from the Numeration (the operation Select), combining the selected elements and partially formed structures (the operation Merge), and by moving elements that are already part of the structure (the operation Move).The computational system is constrained by economy conditions requiring that derivations be as economical as possible. 4At some point in the computation (Spell-Out), the dedvation splits and heads toward the two interface levels.At Spell-Out phoneti-2 A 'theory of grammar' in the Chomskyan sense, i.e. a theory of Iinguistic competence.
3 "This 'double interface' property is one way to express the traditional description oflanguage as sound with a meaning [ ... ]" (Chomsky 1995: 2). 4 "[ ... ] derivations and representations conform to an "economy" criterion demanding that they be minimal: no extra steps in derivation, no extra symbols in representation, etc" (Lasnik 1999: 26).cally relevant information is separated from semantically relevant information, so that ultimately two independent representations are formed: PF representation, containing phonetic information and an LF representation, containing semantic information.
A lexical item is defined as a set of phonological, semantic, and forma!features.Forma! features (e.g.tense and agreement features) occur both in lexical and functional categories.Since forma!features are relevant only to the computational system and play no role at the PF and LF interfaces, they must be eliminated in the course of the derivation.This is achieved by feature-checking -a matching ofthe features, which in effect cancels them out.This matching is brought about by the operation Move.For instance, in order for tense and agreement features to be checked, the verb has to raise and adjoin to the functional categories T(ense) and Agr(eement).
Features of the functional categories can be either strong or weak.Strong features are visible at PF and therefore have to be checked off before Spell-Out.If strong features are spelt out, the derivation crashes.Weak features, on the other hand, are invisible at PF and may therefore be checked after Spell-Out.Accordingly, there are two types of movement: overt (pre-Spell-Out) and covert (LF) movement.Movement is subject to the economy principle Procrastinate, which requires that it be delayed until after Spell-Out as long as this does not cause the derivation to crash.Overt movement thus occurs only when it has to (the principle of Last Resort), that is, in the presence of a strong feature.

Clause structure
According to the X-bar theory of phrase structure5 the basic clause structure is as shown in (2a).A clause is a maximal projection IP headed by the functional category l.The specifier of IP is the subject of IP and the VP is the complement of l.The type of clause (i.e.declarative, interrogative, imperative) is determined by the functional category C(omplementizer), which takes IP as its complement, so that a full clause has the structure (2b ). (

Wh-movement and Verb movement -English and French
Wh-movement is involved in the formation of interrogative sentences where a whphrase raises from its base position to the specifier of CP [Spec, CP].Wh-phrases have operator-like properties and wh-structures are subject to the Wh-Criterion (3a), requiring configurations as in (3b).
(3) a) The Wh-Criterion A. A Wh-Operator must be in a Spec-head configuration with an x 0 1 +WHJ• B. An x 0 1 +WHJ must be in a Spec-head configuration with a Wh-operator.As a general well formedness condition on the scope of wh-operators, the Wh-Criterion applies universally at LF, but some languages may require it to be satisfied earlier.Under Minimalist assumptions, in languages with a strong [+wh] feature (English (4a), French (4b,b')), the movement of the wh-phrase will occur already in the overt syntax (i.e.before Spell-Out), whereas in languages with a weak [+wh] feature, the condition is metat LF, by covert wh-movement (Japanese -(4c)).In sentences with overt wh-movement (4a-b'), the wh-phrase is separated from the subject by a phonologically overt element, which can be a finite verbal form (4a,b), or a question particle ( 4b').
In the sections that follow, we will try to establish what licenses subject-finite verb inversion (Subj-Vfin) in (4a,b)as well as the presence ofthe question particle in (4b').

ENGLISH
In English, subject/finite verb inversion occurs in root interrogatives with the raised wh-phrase in [Spec, CP], unless the wh-phrase functions as subject (cf. (5a,b)).The inverted order is standardly assumed to be derived by the finite verb raising from its position in I to the position C.This movement is restricted to auxiliaries; main verbs do not move (5c).In embedded interrogatives, verb movement, resulting in the inverted order is not allowed (5d).

FRENCH
There are three different ways of forming interrogatives in French: (i) wh-phrase in situ (wh-movement applying covertly) (6a); (ii) wh-phrase overtly moved to [Spec, CP], C filled with the question particle estce que (6b); (iii) wh-phrase overtly moved to [Spec, CP], C filled with the finite verbal form (6c).As in the case ofEnglish, there is an asymmetry between root and embedded clauses.In the latter, wh-element in situ, verb movement to C, and est-ce que in C are not possible (cf.6d-f).

Asymmetric 1-to-C movement
As evidenced by the data in the previous section, both English and French root interrogatives with overt wh-movement require the presence of a phonologically overt element in C.This element can be either a finite verbal form (English, French) or a special interrogative particle (French).In the case ofthe former, the requirement is met by the verb raising to C. This movement involves only those verbal forms that are within the IP-domain; it does not affect VP-intemal verbal elements.Following standard terminology, we will refer to this type of verb movement as I-to-C movement.
Both languages exhibit a root/embedded asymmetry ofl-to-C movement: it occurs in root interrogatives but is absent from embedded interrogatives.Given the Wh-Criterion (3), the root/embedded asymmetry is unexpected.lf 1-to-C movement must apply to establish a Spec-head configuration involving the wh-element and the inflected verb (Rizzi 1991: 1 ), why does it apply only in root and not also in embedded questions?Rizzi (1991: 3-4) argues that in the case of embedded questions, the matrix verb (e.g.wonder in (5d)) selects an indirect question, hence a CP whose head C is marked by the feature [+wH].Since the specification [+wH] fills the embedded C, the latter is not available asa landing site for I-to-C movement.Therefore the verb stays in I and there is no subject-finite verb inversion.The Wh-Criterion is satisfied by wh-movement, which creates the required Spec-head configuration (cf.(3b)).
Rizzi's account raises two questions.
(i) A filled C is claimed to prevent the verb from moving into it.However, the verb could raise and adjoin to C. There is no apparent reason why adjunction to C should be excluded, hence a filled C as a motivation for the absence of verb movement is problematic.(ii) Root questions involve two different types of movement: movement of an XP (wh-phrase) and head movement (I-to-C).Assuming the Minimalist checking theory of movement, the [ +wH] feature of C induces wh-movement and is checked in the Spec-head configuration.The feature triggering I-to-C must be checked in a head-head configuration, but what is this feature?In the next section we will address the question of trigger for I-to-C movement and the related issue of root/embedded asymmetry from the Minimalist perspective and propose an account ofthe asymmetry phenomenon based solely on the feature-checking requirements of the relevant functional head.

[+QUESTION] feature as a marker of interrogative force
As is well-known, differences in clause types are often mirrored in specific syntactic properties of individual clause types.For example, root declaratives in the majority of Germanic languages display the Verb-Second (V2) phenomenon (8a); root questions often require subject-verb inversion in SVO languages (8b ); in directives the subject usually stays unexpressed even in non-pro-drop languages such as English (8c)8: Clause type is determined by the formal features hosted by the functional head C (Chomsky 1995), or, under the Split-CP Hypothesis, (Rizzi 1997), by the functional head Force ofthe highest projection in the CP-domain, the Force P(hrase)9.
Adopting the assumption that C (Force) is the locus of clause-type features, we suggest that in the case ofroot interrogatives C (Force) contains the feature [+QUESTION].This feature is a marker of the interrogative illocutionary force of the clause, and, as will be argued below, is not identical with the [+wH] feature.
There are severa!pieces of empirical evidence that support the postulation of the [+QUESTION] feature: (i) in the Chamorro language there is a special question verb paradigm (Chung 1982); lO (ii) the presence of question particles (est-ce que in French); (iii) special question affixes (the suffix -ne in Latinll).
Based on the empirical evidence and the assumptions presented above, we propose that root interrogatives are clauses with the clausal head C (Force) containing the feature specification as in ( 9): are two distinct features, thus lending additional support to the proposed analysis.

2.3.l VERBALQUESTIONS: NO [+WH] FEATURE
Verbal questions (10) have no wh-element, hence no [+wH] feature.They exhibit 1to-C movement, which is licensed by the presence of the [ +QUESTION] feature.Si111ce this feature is strong in French (IOa) and in English (IOb), 1-to-C movement must apply overtly.

EXCLAMATIVES: NO [+QUESTION] FEATURE
English exclamatives are restricted to the type of exclamatory utterance introduced by what or how (Quirk et al., 1999: 833).These clauses are not interpreted as questions although they involve wh-movement to [Spec, CP].In languages with a strong [+wH] feature, such as English, wh-movement is overt (11 a).The presence of the [ +wH] feature in exclamatives, however, does not license 1-to-C movement (cf. the ungrammaticality of (11 b) ).Exclamatives, being devoid of interrogative force, Jack the [ +QUES-TION] feature, hence 1-to-C is ruled out by the principle of Economy.

ROOT/EMBEDDED ASYMMETRY REANALYSED
As noted in section 2.2, Rizzi's (1991) account ofthe root/embedded asymmetry of 1-to-C movement in interrogatives is problematic from the perspective ofthe Minimalist checking theory of movement.In particular, if the root C contains only the [ +WH] feature, which triggers wh-movement, there is no feature to license 1-to-C movement.
Our analysis of root interrogatives (9) provides a straightforward answer to the problem of trigger for 1-to-C.The root C contains two features, [+wH] and [+QUES-TION], the former licensing wh-movement and the latter 1-to-C.Now consider embedded interrogatives.The information about the illocutionary force of the clause is encoded in the highest functional projection in the CP-domain.In the case of embedding, this is the matrix CP; embedded clauses themselves have no illocutionary force, 12 hence their C contains no illocutionary-force features.The [+QUESTION] feature, being an illocutionary-force feature, is thus not present in the C of embedded interrogatives.
The root/embedded asymmetry can be fully accounted for in terms of presence/absence ofthe features [+wH] and [+QUESTION].In root interrogatives, [+wH] triggers wh-movement to [Spec, CP] and [+QUESTION] triggers I-to-C movement.In embedded interrogatives, [+wH] triggers wh-movement to [Spec, CP], while I-to-C does not take place due to the absence of the [ +QUESTION] feature.
To sum up, we have argued that the presence of the [ +wH] feature licenses only one type of movement, namely the raising of a wh-element to [Spec, CP], where the [ +wH] feature of C is checked against the corresponding feature of the raised wh-element.Ito-C movement is licensed by another feature on C, the [+question] feature determining the interrogative force of a clause, which is present in the C of root interrogatives only.On this analysis, the root/embedded asymmetry is the result of different featural content ofthe clausal head C, as shown in ( 12). ( 12)

Subject -Finite Verb+Non-finite Verb inversion in French
The analysis of interrogatives proposed in 2.3 above is based on the <lata involving only English interrogatives and those French interrogatives where the subject is weak.13 In this section we will take a look at French interrogatives containing a strong subject.
when goes phone Yves b) quand Yves va telephoner.when Yves goes phone 'when Yves will phone.'c) quand il va telephoner.d) *quand va telephoner il.e) *quand va-t-il telephoner.'when he will phone.'An adequate account ofthe data presented above would have to provide answers to the following questions: (i) Why is Subj -Vfin+Ynon-fin inversion possible in interrogatives with a strong subject but not in those with a weak subject (cf.(13a), (14a))?(ii) What motivates the movement of both finite and non-finite verbal forms (resulting in Subj -Vfin+Ynon-fin inversion) in root and embedded interrogatives with strong subjects?These questions will be addressed in the section below.

STRONG AND WEAK SUBJECTS
In his analysis of Italian interrogatives Rizzi (1991) suggests that pre-verbal subjects are assigned nominative Case under Spec-head agreement with the inflectional head of the highest functional projection in the IP-domain, whereas post-verbal sub-jects are assigned nominative Case by the inflectional head of a lower functional projection under government.15 Extending Rizzi's proposal to French and assuming that only strong subjects can occur in post-verbal position (Cardinaletti 1997), the contrast in grammaticality between (13a) and (14a) can be accounted for as follows.(14a) contains a weak subject, which should be assigned nominative Case under Spec-head agreement.With the moved participle intervening between the subject and the finite verb, the required configuration is destroyed.Since Case cannot be assigned, the Case Filter is violated, and ungrammaticality results.In (13a) the subject is strong and occurs in post-verbal position.As such, it is assigned nominative Case under government.Therefore the moved participle does not interfere with Case assignment, and the construction with Subj -V fin+ Vnon-fin inversion is well-formed.In this way, an answer to question (i) above is provided: the presence/absence of Subj -Vfin+Vnon-fin inversion in interrogatives with strong and in those with weak subject respectively is due to two different mechanisms of Case assignment associated with the two subject types.

LICENSING SUBJECT-FINITE VERB + NON-FINITE VERB INVERSION
In section 2.3 we argued that the trigger for ( overt) I-to-C movement is a strong [+QUESTION] feature in C.This feature is checked by the finite verb raising to C. In the case of Subj -Vfin+Ynon-fin inversion both the finite and the non-finite verb forms move.Given that in the Minimalist system movement is driven by feature-checking requirements and that the feature-checking requirement imposed by the strong [ +QUES-TION] feature is satisfied by the finite verb raising to C, the non-finite verb should not move unless there is another strong feature to be checked.This raises a few questions.What is this feature?Where is it located?Does Subj -Vfin+Ynon-fin involve head movement or some other type of movement?
It seems impossible to answer such questions with any certainty without a detailed analysis of constructions with Subj -Vfin+Ynon-fin inversion.However, this is beyond the scope ofthis paper.Therefore let us mention only the most recent research concerning the left periphery of Romance interrogatives ( cf.Pollock 2001a, Pollock 2001b, Pollet-to&Pollock 2002, among others).This research suggests that various aspects of the syntax ofRomance questions derive from the highly split structure ofthe interrogative Romance CP field and the morphological properties of individual wh-elements which are responsible for the features they can check in the left periphery.Based on this, severa!types of interrogative constructions in Romance ( e.g.subject clitic inversion, Subj -Vfin+Ynon-fin inversion, wh-in situ constructions) have been explained in terms of generalized remnant movement to the CP area.As we will see in the next section, the Subj -Vfi 0 +V 00 n-fin inversion occurs also in Slovenian, which shows that the !5 Assuming the Split-IP Hypothesis of Pollock (1989) and the relative order ofprojections ofBelleti (1990), the two functional projections concemed are Agr( eement) P(hrase) and T( ense) respectively.(Rizzi 1991: 17) phenomenon is not language-group-specific (Romance).How Subj -V fin+ V non-fin in Slovenian is to be analysed and whether or not the remnant movement analysis could be extended to Slovenian are matters we leave for future research.

Verb movement in Slovenian interrogatives
Slovenian belongs to the class of languages with overt interrogative wh-movement (Golden 1996a(Golden , 1996b;;and Marvin 1997).Under Minimalist assumptions this means that the [ +wH] feature of C in Slovenian is strong (as in English and French -cf.section 2 above ).It must be checked overtly by the wh-phrase raising overtly to [Spec, CP]; if no movement occurs, ungrammaticality results (16a,b).Slovenian shares another characteristic with English and French: as seen in ( 16a), the verb precedes the subject (Subj-V fin inversion), which suggests overt 1-to-C movement.
Let us now examine the properties ofl-to-C movement in Slovenian interrogatives in more detail and compare them with those of 1-to-C in English and French.

Data III
In Slovenian, auxiliaries as well as lexical verbs may undergo 1-to-C raising (17a-b).In this respect Slovenian resembles French (cf.4b, 6c), but differs from English, where only auxiliaries are allowed to raise (cf.Sa, c)).

Asymmetric and symmetric 1-to-C
In English and French, I-to-C movement shows a clear asymmetry between root and embedded interrogatives.As argued in section 2.3, this asymmetry is due to the presence/absence of the [+QUESTION] feature of the root and embedded C respectively.In Slovenian, however, the root/embedded distinction is weakened; I-to-C can optionally apply in embedded interrogatives.Furthermore, in some Romance languages (Spanish, Catalan, Romanian 17 ), this distinction disappears altogether; I-to-C is obligatory in both (23) a) strong feature: a feature that is visible at PF b) weak feature: a feature that may be checked only by a pure feature (Collins 1997:117;(4a,b)) Since, as pointed out by Collins (op.cit.: 118), these two definitions are independent, they can be cross-classified, allowing for a third type of feature: a feature that is neither strong nor weak.Such a feature would not be visible at PF and would not have to be checked by a pure feature.Asa result it could be checked either overtly or covertly, l9 hence movement could be optional.
Applying the modified theory of features as proposed by Collins to verb movement in Slovenian interrogatives, the optionality of Slovenian 1-to-C in root and embedded interrogatives can be straightforwardly accounted for in the following way: in Slovenian the features [+QUESTION] and [+FIN] are neither strong nor weak and may therefore be checked either by overt or by covert 1-to-C.
To conclude, adopting the core idea of the Minimalist Program that cross-linguistic variation in the surface order of clausal constituents is the result of parametric variation in feature strength of functional heads, and a theory of features which postulates three types offeatures, allows for a uniform account of both obligatory (asymmetric and symmetric) as well as optional 1-to-C in interrogatives.Specifically, assuming the relevant features of the clausal head C to be (+QUESTION] and [+FIN], the typological differences in interrogative I-to-C in the languages under consideration can be ac-counted for in terms of different feature values (strong/weak/neither strong nor weak) -cf.(24).In this paper we have examined verb movement in wh-interrogatives in English, French and Slovenian against the background of recent approaches to this phenomenon within the framework of Chomskyan generative grammar.We have argued that 1to-C verb movement in interrogatives occurs independently of wh-movement and have identified the [ +QUESTION] feature of root C, denoting interrogative illocutionary force, as the licenser of 1-to-C raising in root interrogatives.Adopting the Minimalist checking theory of movement (Chomsky 1995), modified as in Collins (1997), we have proposed an account of the typological differences in 1-to-C (overt/covert, asymmetric/ symmetric, obligatory/optional) in terms of the featural content of the functional head C and the strength ofthe relevant features [+QUESTION] and [+FIN].

Rizzi
(4) a) Where did she go? b) Que veux-tu?what want-you b') Qu' est-ce que tu veux?what Q 7 you want 'What do you want?' c) John-ga doko-ni ikimasa (ka) John where gone-has -k 'Where has John gone?' (5) a) What have you done?b) Who has done it?c) *What want you? d) I wonder what *{have} you {have} done.
ne sais pas qui elle a vu *{qui}.I ne know not who she has seen e) Je ne sais pas qui *{a} elle a vu.I ne know not who she has seen f) Je ne sais pas qui * { est-ce que} elle a vu.I ne know not who she has seen 'Ido not know who she has seen.' (8) a) Denne film har b0mene set.Danish (Rohrbacher 1999: 13, (2d)) this film have children seen 'The children have seen this film.'b) Is she happy?c) Go home.
If both features in (9) are strong, they have to be checked in the overt syntax.The[+WH]feature is checked in the Spec-head configuration, with the wh-element raising to[Spec, CP].The[+QUESTION]  feature can be checked either by a lexical element carrying the corresponding feature: a verbal element (1-to-C movement) or a special question particle -cf.section 2.1.9 Under the Split-CP Hypothesis ofRizzi (1997), the formerly uniform CP-Iayer ofprojection is analysed as consisting of several distinct projections, ForceP>Top(ic)P>Foc(us)P>Fin(ite )P. 10 The relevant examples: a) Hay f-11m-a'gasi i kareta?(Chung 1982: 49, (30a)) who ?-washed the car 'Who washed the car?' b) Ha-fa'gasi si Juani kareta.(Chung 1982: 49, (30b)) washed case Juan the car 'Juan washed the car.' l I The suffix -ne can attach to either verbal or non verbal elements.Consider: a) Domine, dominae domine sunt?b) Domine, dominae domi suntne?Lord-voc ladies home are 'O lord, are the Iadies at home?' Now let us consider two clause types which clearly show that [ +WH] and [ +QUESTION]

(
II) a) What a tirne we have had today!b) *What a tirne have we had today!
je Marko prinesel.what is-cl Marko brought b) kaj bere Marko.what reads Marko 'Tell me what has Marko brought/ is Marko reading.' I-to-C in English and French (in the case of interrogatives with a weak subject and Subj -V fin inversion) is obligatory.In contrast, Slovenian I-to-C is optional: cf.(l 7b) and (18b), repeated below as (19a,b), and their respective counterparts without Subj-Vfin inversion (20a,b), which are equally acceptable.16(19) a) Kaj bere Marko?what reads Marko 'What is Marko reading?' b) Povej mi, kaj bere Marko.teli me what reads Marko 'Tell me what Marko is reading.(20) a) Kaj Marko bere?what Marko reads 'What is Marko reading?' b) Povej mi, kaj Marko bere.teli me what Marko reads 'Tell me what Marko is reading.'As in French interrogatives with strong subjects, both finite and non-finite verbal forms may raise across the subject (Subj -Vfin+Vnon-fin inversion) -cf.(13a), (15a,b) and (21 a,b).Note that in Slovenian, unlike in French, Subj -Vfin+Vnon-fin inversion is optional both in root and in embedded clauses (cf. the well-formed (l 7a), (18a), where only the finite verb has raised.)(21) a) Kaj je prinesel Marko?what is-cl brought Marko 'What has Marko brought?' b) Povej mi, kaj je prinesel Marko.teli me what is-cl brought Marko 'Tell me what Marko has brought.'