Daria Protopopescu A SYNTACTIC APPROACH TO ADVERBS IN ENGLISH AND ROMANIAN TEMPORAL AND ASPECTUAL ADVERBS

Daria Protopopescu A SYNTACTIC APPROACH TO ADVERBS IN ENGLISH AND ROMANIAN TEMPORAL AND ASPECTUAL ADVERBS

Reproducerea integrală sau parţială, multiplicarea prin orice mijloace şi sub orice formă, cum ar fi xeroxarea, scanarea, transpunerea în format electronic sau audio, punerea la dispoziţia publică, inclusiv prin internet sau prin reţele de calculatoare, stocarea permanentă sau temporară pe dispozitive sau sisteme cu posibilitatea recuperării informaţiilor, cu scop comercial sau gratuit, precum şi alte fapte similare săvârşite fără permisiunea scrisă a deţinătorului copyrightului reprezintă o încălcare a legislaţiei cu privire la protecţia proprietăţii intelectuale şi se pedepsesc penal şi/sau civil în conformitate cu legile în vigoare.

Daria Protopopescu A SYNTACTIC APPROACH TO ADVERBS IN ENGLISH AND ROMANIAN TEMPORAL AND ASPECTUAL ADVERBS 2014

Referenţi ştiinţifici: Prof. univ. dr. Mariana Neţ (Academia Română, Institutul de Lingvistică Iorgu Iordan ) Conf. dr. Nadina Vişan (Universitatea din Bucureşti) Lector dr. Aura Gherguţ (Universitatea Tehnică de Construcţii Bucureşti) Şos. Panduri, 90-92, Bucureşti 050663, România Telefon/Fax: (0040) 021.410.23.84, E-mail: editura.unibuc@gmail.com, Librărie online: http://librarie-unibuc.ro Centru de vânzare: Bd. Regina Elisabeta, nr. 4-12, Bucureşti, tel. (0040) 021.305.37.03 Web: www.editura.unibuc.ro DTP & coperta: Florina FLORIŢĂ Descrierea CIP a Bibliotecii Naţionale a României PROTOPOPESCU, DARIA / A Syntactic Approach to Adverbs in English and Romanian Temporal and Aspectual Adverbs / Daria Protopopescu. Bucureşti: Editura Universităţii din Bucureşti, 2013 Bibliogr. ISBN 978-606-16-0399-2 811.111'367.624 811.135.1'367.624

TABLE OF CONTENTS Foreword... 5 1. Introduction... 9 1.1. Introduction and organization of the book... 9 1.2. Main issues; co-occurrence restrictions... 11 1.3. Linear order of VP-internal adverbs... 13 1.4. Word order in Romanian... 14 1.5. Summing up... 17 2. Aspects of Syntactic Theory. The Minimalist Framework... 19 2.1. Introduction..... 19 2.1.1. Merge, agree, move... 20 2.1.2. Derivation by phase... 24 2.2. Functional projections... 26 2.3. The functional structure of Romanian sentences... 28 2.3.1. The basic structure.... 29 2.3.2. The Romanian CP domain.... 31 2.4. Summing up... 32 3. Temporal and Aspectual adverbs across frameworks... 33 3.1. The cartographic approach... 33 3.1.1. The feature theory. Cinque (1999)... 33 3.1.2. The double specifier model. Laenzlinger (1998)... 37 3.1.3. Alexiadou (1997)... 40 3.2. In favour of left adjunction... 44 3.2.1. Haider (2000, 2004)... 46 3.2.2. The scope theory. Ernst (2002)... 49 3.3. Summing up... 54 4. Temporal and aspectual adverbs classifications, contexts of occurrence and licensing... 55 4.1. Classifications: English vs. Romanian... 56 4.1.1. Time orientation... 61 4.1.2. Frame adverbs or locating temporal adverbs... 62 4.1.3. Anchored vs. unanchored adverbs... 63 4.1.4. Aspectual adverbs... 66

6 DARIA PROTOPOPESCU 4.2. Sentence structure revisited... 68 4.2.1. English... 69 4.2.2. Romanian... 75 4.3. Positions and analysis... 82 4.3.1. The initial position... 86 4.3.2. The preverbal position... 88 4.3.3. The postverbal position... 90 4.3.4. The final position... 92 4.3.5. The parenthetical position... 95 4.4. Left/right asymmetries of temporal adverbs... 95 4.5. Temporal and aspectual adverbs in nominalizations... 98 4.6. Summing up... 109 Conclusions... 111 Appendix 1... 115 Appendix 2... 116 Appendix 3... 118 References... 214

FOREWORD The current book is the outcome of comparative research in the field of the syntax of temporal and aspectual adverbs. It is a natural follow-up to my doctoral dissertation that investigated the domain of manner adverbs in English and Romanian. The interest in the classes of temporal and aspectual adverbs is not new. It came from the large amount of literature dedicated to the study of adverbs in various languages, research that tried to provide either a syntactic or a semantic account of adverbs across languages. This book is an attempt of providing a unifying account of these two classes of adverbs from a comparative perspective. Since there are not that many studies on adverbs in Romanian, more particularly on these two classes of adverbs, the book also provides an extensive corpus of contexts containing temporal and aspectual adverbs, drawn from Romanian literary texts. This work was supported by a grant of the Romanian National Authority for Scientific Research, CNDI UEFISCDI, project number PN-II-IN-CI-2012-1-0366.

CHAPTER 1 INTRODUCTION 1.1. Introduction and organization of the book Analyzing adverbs from a minimalist point of view is no easy task for any linguist and this is even more difficult when one attempts a comparative approach between two languages that behave so differently such as English and Romanian. The main endeavour of the current book is to provide a new classification for temporal, aspectual and sentence adverbs in English and Romanian while correlating it with the distribution of these adverbs in the sentence function of the position where they are properly interepreted. Since Jackendoff (1972), it has been argued in the literature that the different positions in which adverbs occur in a sentence trigger different interpretations. This book is going to explore this hypothesis about the class of temporal and aspectual adverbs. Adverbs, in general, are invariable and exhibit poor morphology. In earlier research, the only adverbs that were considered to enter relations of compatibility with the verb were temporal adverbs. The problem generally arises with sentences where more than one adverb occurs. The analysis of sentence structure has shown the existence of several functional projections: TP, AspP, VoiceP. Alexiadou (1994) proposes that temporal, aspectual and manner adverbs are licensed by means of Spec-Head agreement in the Specifier position of the respective TP, AspP and VoiceP. We are going a step further by following the adverb constraints discussed by other linguists, constraints imposed by the tense and aspect with which some of them have to combine. To this extent we will make use of functional projections that occur in the clause whenever a certain tense/aspect is activated. In the minimalist framework, the movement of a constituent occurs only if the constituent has to check a morphological or an operator feature. For instance, in some languages, the verb with a rich inflection raises from its base position to check its tense feature. Since the adverb has a very poor morphology, if at all in some cases, it is assumed not to move from its base generated position, serving as a fixed point, which indicates verb movement. As will be shown throughout the book, the structure of the sentence includes a series of functional projections representing the verbal categories of mood, tense, aspect and voice. Thus, certain adverbs, such as temporal, aspectual and manner adverbs could possibly enter matching relations with these verbal categories, either from a specifier position for the respective functional projection (the cartographic approach, defended by Cinque 1999, 2004), or as an adjunct (the scope-based approach, defended by Ernst 2002, 2004, 2006). It is expected that adverbs should have fixed positions and a relatively rigid order of occurrence as they can enter agreement relations with the relevant heads/projections. This is due to the fact that heads and maximal projections of functional categories are

10 DARIA PROTOPOPESCU themselves generated in a rigid order, if we are to follow the cartographic approach to the study of adverbs (cf. Alexiadou 1994, Cinque 1997, etc.) Whether this is so or not will be discussed throughout the book, since a second approach to the study of adverbs claims that it is possible to reconcile their syntactic positions with their semantic interpretations by means of a FEO Calculus (FEO, cf. Ernst 2002). According to this hypothesis, adverbs are generated as adjuncts rather than specifiers of the various functional projections in the clause. The data analyzed in this book are drawn from both English and Romanian. However, since there is a large amount of literature on English, there are two appendices at the end of the book providing a large corpus of examples from literary texts in Romanian for most of the adverbs under discussion. Chapter one introduces the data to be analyzed as well as the main issues to be dealt with in the book. Chapter two deals with the analysis of sentence structure in English and in Romanian and shows that the order of functional projections is the same in the two languages. It is concerned with movement and it gives an account of the minimalist framework.introducing the classification of adverbs from a traditional point of view, according to their function and syntactic behaviour but it proposes different classifications of aspectual and temporal adverbs as well. Chapter three offers an extensive analysis of the two theories that provide the framework for the study of temporal and aspectual adverbs. Since one approach, the cartographic approach is mainly syntactic and the other is mainly semantic, we will try to discuss both and come up with a solution that could accommodate the data in such a way that would reconcile the two views. Thus, under the specifier or cartographic approach, adverbs are treated as specifiers of functional heads (cf. Alexiadou 1994, Cinque 1997, Laenzlinger 1998, etc.). According to the cartographic approach, temporal and aspectual adverbs are licensed as specifiers of the functional projections of tense, and aspect (TP and AspP). Alexiadou (1994) proposes the following features in T 0 and Asp 0 features which agree or match adverb features and which we summarize below: Tense Aspect T 0 Temporal adverbs Asp 0 Aspectual +PERF -PERF adverbs +punctual +habitual +/-point +/-present +/-present +definite -definite +/-definite frequency +/-past (anteriority) +/-past (anteriority) +durative +/-durative Under the adjunction approach (cf. Ernst 2002, 2004, Haider, 2000, 2004 Costa 1998, 2004) adverbs are treated as left-adjoined as a result of a FEO Calculus, which offers a rather strict hierarchy for adjunction. Chapter four is concerned with the analysis of temporal and aspectual adverbs proper in the two languages. To this extent, we start by offering the different classifications available in the literature and try to come up with one that reconciles our analysis which is an attempt of correlating the interpretations of these adverbs to their syntactic placement. For English we start from the distributional classes of adverbs as discussed by Jackendoff (1972), whereas for Romanian, we start from the traditional classifications in the Grammar of the

A SYNTACTIC APPROACH TO ADVERBS IN ENGLISH AND ROMANIAN - 11 TEMPORAL AND ASPECTUAL ADVERBS Academy. After establishing the distributional classes and the contexts of occurrence, we proceed to the analysis of their licensing mechanisms function of the framework introduced in the previous two chapters. We also discuss the left/right contrasts that occur in the case of temporal adverbs (cf. Shaer 2004). Additionally we discuss the occurrence of temporal and aspectual adverbs in nominalizations, again from a comparative perspective. Finally the conclusions are followed by three appendices which provide the classification of temporal and aspectual adverbs in Romanian as well as a small corpus of examples for temporal and aspectual adverbs in Romanian, taken from literary works. 1.2. Main issues; co-occurrence restrictions One of the most intriguing features of adverbs in general is their ability to co-occur in a sentence while at the same time being limited from co-occuring under certain conditions. (1) a. She frequently has wisely gone there on Sundays. b. She wisely has frequently gone there on Sundays. (2) a. She frequently was suddenly (being) rejected by publishers. b. She suddenly was (being) frequently rejected by publishers. (examples 3.85-3.86 in Ernst 2002: 119-120) (3) a. Habituellement ils regardent fréquemment la télé. usually they watch frequently the television. b. *Fréquemment ils regardent habituellement la télé. but c. Fréquemment ils ont regardé habituellement la télé. (examples 3.108-3.109 in Ernst 2002: 126) In examples (1)-(2), it appears that aspectual and temporal adverbs can freely co-occur. However, examples (3a-b) would indicate that Cinque s (1997) claim, that there is a rigid order of adjuncts, is true. Ernst (2002: 126) argues that although the rigid order claimed by Cinque is valid in the case of predicationals, what he calls participant adjuncts and most of the functional adjuncts, temporal and aspectual adverbs included here, appear not to observe the rigid order. Indeed, if we are to analyse the data, (3c) is identified by native speakers of French as possible if the time intervals referred to by the two adverbs are sufficiently different. Moreover, Travis (1988) suggests that manner adverbs and other categories that function as manner adverbials, such as, say, manner PPs are subject to different licensing conditions and surface in different structural positions. If this were correct, we would expect a manner adverb and a PP manner adverbial to be able to grammatically co-occur in one sentence. However, the examples below suggest that this is not possible. Co-occurrence yields ungrammaticality and the sentences are ill-formed. Therefore, it is safe to conclude that manner adverbs and other categories expressing manner adverbials compete for the same structural position. (4) a. *Maria vorbeşte [bine] Manner Adv [cu eleganţă] Manner PP. optional adverbials

12 DARIA PROTOPOPESCU Maria speaks-3sg. well with elegance Maria speaks well with elegance. b. *Maria vorbeşte [cu eleganţă] Manner PP [bine] Manner Adv. Maria speaks-3sg. with elegance well (5) a. *Ion se poartă [urât] Manner Adv [cu toane] Manner PP. obligatory adverbials Ion reflexive-3sg. behaves-3sg. badly with whims Ion behaves badly whimsically. b. *Ion se poartă [cu toane] Manner PP [urât] Manner Adv. Ion reflexive-3sg. behaves-3sg. with whims badly The sentences are well-formed if we apply coordination of the two manner adverbials, thus showing that in fact they are part of a larger constituent that occupies one structural position. (6) a. Maria vorbeşte [[bine] Manner Adv şi [cu eleganţă] Manner PP ]. Maria speaks-3sg. well with elegance Maria speaks well with elegance. b. Ion se poartă [[urât] Manner Adv şi [cu toane] Manner PP ]. Ion reflexive-3sg. behaves-3sg. badly and with whims Ion behaves badly and whimsically. A related question is that which refers to the impossibility of obligatory and optional manner adverbs to co-occur. Therefore, if they had different base positions as predicted by Travis (1988) then the data in (7) and (8) below should display grammaticality which it clearly does not. (7) a. *Maria vorbeşte bine fluent. optional adverbs Maria speaks-3sg. well fluently Maria speaks well fluently. b. *Ion citeşte repede corect Ion reads-3sg. quickly correctly Ion reads quickly correctly (8) a. *Ion se poartă urât capricios. obligatory adverbs Ion reflexive-3sg. behaves-3sg. badly whimsically Ion behaves badly whimsically. b. *Maria trăieşte bine confortabil. Maria lives-3sg. well comfortably Maria lives well comfortably. Again, as was the case above in (6a-b), if we apply coordination the sentences become well-formed. (9) a. Maria vorbeşte bine şi fluent. optional adverbs Maria speaks-3sg. well and fluently Maria speaks well and fluently.

A SYNTACTIC APPROACH TO ADVERBS IN ENGLISH AND ROMANIAN - 13 TEMPORAL AND ASPECTUAL ADVERBS b. Ion citeşte repede şi corect. Ion reads-3sg. quickly and correctly Ion reads quickly and correctly. (10) a. Ion se poartă urât şi capricios. obligatory adverbs Ion reflexive-3sg. behaves-3sg. badly and whimsically Ion behaves badly and whimsically b. Maria trăieşte bine şi confortabil. Maria lives-3sg. well and comfortably Maria lives well and comfortably. 1.3. Linear order of VP-internal adverbs Many languages exhibit a fixed linear order of direct objects with respect to VP internal manner, place and time adverbials. Thus, (11a-b) have neutral, syntactically unmarked word orders, while (12a-b) have syntactically marked word orders, in the sense that one of the constituents is interpreted as receiving stress or contrastive focus of some sort (see Zubizarreta 1998) for Romance: (11) a. Ion a deschis uşa repede. Ion has-3sg. opened door-the fast Ion opened the door fast b. Maria a citit poezia frumos. Maria has-3sg. read poem-the beautifully Maria read the poem beautifully (12) a. Ion a deschis repede uşa [nu geamul] Ion has-3sg. opened fast door-the [not window-the] It was the door [not the window] that Ion opened fast b. Maria a citit frumos poezia [dar nu şi piesa de teatru] Maria has-3sg. beautifully read poem-the [but not also play-the] It was the poem [not the play] that Maria read beautifully What is interesting, though, is that (14a-f) there appears to be no constraint on the ordering of sentence final manner, place and time adverbials. All these examples appear to have neutral, syntactically unmarked word orders in the sense that none of the constituents can be interpreted as receiving extra stress or focus. Cinque makes similar observations (1999: 15-16) for Italian, where in (13) below all mutual orderings seem possible, except for differences in scope: (13) a. Seguiro le lezioni tutti i giorni all universita diligentemente. I will attend classes every day at the university with great zeal. b. Seguiro le lezioni all universita tutti i giorni diligentemente. c. Seguiro le lezioni all universita diligentemente tutti i giorni.

14 DARIA PROTOPOPESCU d. Seguiro le lezioni diligentemente all universita tutti i giorni (Cinque 1999: 15-16 example (68)) Romanian (14) a. Ion a văzut-o pe Maria rapid acolo ieri. Ion has-3sg. seen-clitic-3sg.fem.acc. pe Maria quickly there yesterday Ion saw Maria quickly there yesterday b. Ion a văzut-o pe Maria rapid ieri acolo. Ion has-3sg. seen-clitic-3sg.fem.acc. pe Maria quickly yesterday there c. Ion a văzut-o pe Maria acolo rapid ieri. Ion has-3sg. seen-clitic-3sg.fem.acc. pe Maria there quickly yesterday d. Ion a văzut-o pe Maria acolo ieri rapid. Ion has-3sg. seen-clitic-3sg.fem.acc. pe Maria there yesterday quickly e. Ion a văzut-o pe Maria ieri rapid acolo. Ion has-3sg. seen-clitic-3sg.fem.acc. pe Maria yesterday quickly there f. Ion a văzut-o pe Maria ieri acolo rapid. Ion has-3sg. seen-clitic-3sg.fem.acc. pe Maria yesterday there quickly The question raised by these data is why the linear ordering of direct objects and VP internal manner, place and time adverbs is fixed, while the mutual order of these adverbs seems to be very free. Following this discussion of the main question, we move on to an analysis of the morphology of Romanian adverbs in order to gain a better insight into the kind of language Romanian is from the point of view of its adverbs. 1.4. Word order in Romanian In order to better capture the phenomena displayed by the placement of temporal and aspectual adverbs we should take a brief look at the word order of Romanian declarative sentences. It is argued that the neutral, syntactically unmarked word order for Romanian declarative sentences containing a subject, a verb and an object is VSO (cf. Dobrovie-Sorin 1994, Cornilescu 2000, Alboiu 2002 a.o.). However, under appropriate discourse conditions, all permutations of the SVO order are possible, making Romanian declarative sentences exhibit a relatively free word order. In (15) we illustrate all the different word orders in Romanian. The examples in (15c-f) are heavily marked and emphasize the assertion made in the sentence. (15) a. Ion a deschis o uşă. SVO Ion has-3sg. opened a door Ion opened a door b. A deschis Ion o uşă. VSO has-3sg. opened Ion a door Ion opened a door c. Ion o uşă a deschis (nu un geam). SOV

A SYNTACTIC APPROACH TO ADVERBS IN ENGLISH AND ROMANIAN - 15 TEMPORAL AND ASPECTUAL ADVERBS Ion a door has-3sg. opened (not a window) It was a DOOR that Ion opened (not a window) d. A deschis o uşă Ion (nu George). VOS has-3sg. opened a door Ion (not George) It was ION who opened a door (not George) e. O uşă, Ion a deschis-o (nu George). OSV A door Ion has-3sg. opened (not George) It was ION who opened a door (not George) f. O uşă a deschis Ion (nu un geam). OVS A door has-3sg. opened Ion (not a window) It was a DOOR that Ion opened (not a window) Neutral, or unmarked word order can be determined by means of a question/answer test. Thus, Zubizarreta (1998) a.o. argued that only sentences neutral, unmarked word order can usually serve as answers to the question What happened? (16) Ce s-a întâmplat? What happened? a. Ion a deschis o uşă. SVO Ion has-3sg. opened a door Ion opened a door b. A deschis Ion o uşă. VSO has-3sg. opened Ion a door Ion opened a door c. *Ion o uşă a deschis. SOV Ion a door has-3sg. opened d. *A deschis o uşă Ion. VOS has-3sg. opened a door Ion e. *O uşă, Ion a deschis-o. OSV A door Ion has-3sg. opened f. *O uşă a deschis Ion. OVS A door has-3sg. opened Ion The fact that only (16a-b) have a neutral syntactically unmarked word order is in concord with Cornilescu (2000), Alboiu (2002) a.o. who conclude that Romanian lacks a unique subject position, but not necessarily that both SVO and VSO are unmarked. This is also in agreement with Chomsky (1995) who draws a typology of languages function of the EPP feature. Thus, if a language has no EPP it is a VSO language; if it has EPP but Merge into [Spec,HP], where H is the probe, causes EPP to delete, the language is SVO without Multiple Subject Constructions; if it has EPP but allows one element to Merge without deleting EPP then the language is SVO with multiple subject constructions. In the case of SVO, that is with pre-verbal subjects, Cornilescu (2000) shows that the theta-role is assigned in SpecVP and case and theta-features are checked in AgrS or lower in TP, concluding that there is no pre-verbal argumental subject position in Romanian. Instead, the pre-verbal subject occupies one of the positions at the left periphery of the sentence, namely it is either Topic or Focus.

16 DARIA PROTOPOPESCU Preverbal NPs, i.e. NPs appearing to the left of the verb are constrained by interpretation to the extent that (16f) would be ungrammatical with the direct object in preverbal position, unless contrastively focussed: (17) a. O UŞĂ a deschis Ion (nu un geam). OVS A door has-3sg. opened Ion (not a window) It was a DOOR that Ion opened (not a window) b. *O uşă a deschis Ion (nu un geam). OVS A door has-3sg. opened Ion (not a window) It was a DOOR that Ion opened (not a window) All direct objects above are indefinite NPs, which according to Zubizaretta (1998), and Alboiu (2002: 31) affect the grammaticality of the examples in (16, 17). Alboiu concludes that in Romanian, preverbal NPs are constrained by a specificity requirement which states that unless they are contrastively focused preverbal NPs should be specific or strong, i.e. they should be definite NPs (e.g. prietena mea), indefinite NPs with either a referential (e.g. o prietenă de-a mea), a partitive (e.g. Doi peşti sunt negri, al treilea e roşu) or a generic collective reading (e.g. Trei peşti sunt mai scumpi decât doi.) (examples from Alboiu 2002:32). Her conclusion is that these semantic restrictions on NPs in the preverbal domain are indicative of a topical domain in Romanian. Furthermore, Alboiu claims that given empirical data, Romanian is a V-type EPP language, where Spec, IP is not the canonical subject position and the language actually lacks a unique subject position. As shown by previous studies (Cornilescu 1997, Dobrovie-Sorin 1994), the nature of this EPP feature in Romanian is that of a strong [+V] feature, which leads to the lexical verb always raising to I 0. Alternatively, the marked orders of (15c-f) suggest that (15c-f) cannot be base-generated. In this case the question that arises is whether these sentences can be generated by movement. If they can be generated by movement, the question is what kind of movement operations are we dealing with: scrambling or movement to a focus-related projection, i.e. a position associated with stress or contrastive focus. In the case of the VOS order, the traditional analysis for other Romance languages favours the right-adjunction of the subject to the VP or base-generation in this position. Cornilescu argues that in Romanian it is impossible to adjoin the post-verbal subject to the VP, since such an analysis would predict that sentences with the subject appearing after an object clause would have to be grammatical which is contradicted by the data below from Cornilescu (2000): (18) a. *I-a spus lui Ion că vremea va fi frumoasă Petre. him has said to Ion that weather-the will be nice Petre b. Petru i-a spus lui Ion că vremea va fi frumoasă. Petru him has said to Ion that weather-the will be nice Petru told Ion that the weather will be nice Thus, the VOS order is argued to involve object movement (scrambling) instead of subject movement. More specifically, the object moves past the subject in SpecVP, to a SpecAgrO. According to Alboiu (2002:28), Romanian exhibits two types of scrambling, both semantically restricted and both of which represent non-feature driven movement: vp

A SYNTACTIC APPROACH TO ADVERBS IN ENGLISH AND ROMANIAN - 17 TEMPORAL AND ASPECTUAL ADVERBS scrambling, i.e. de-focusing, which has A-movement properties, and IP-scrambling i.e. topicalization, which has A-bar movement properties. Alboiu (2002) follows Cornilescu (2000) in arguing against right-adjunction of the subject in the VOS order and for an object raising analysis. In conclusion, the VS(O) word order is the unmarked one in Romanian, so any derivation departing from the basic VS(O) has to be accounted for, therefore examples (15c-f) are created by means of movement to some focus-related position. In the following chapters dealing with the analysis of sentence and manner adverbs, I will be dealing mostly with the VSO and SVO word orders. 1.5. Summing up At this point we have set out a few directions for study and throughout this book we are going to address these issues and try to offer a pertinent analysis and hopefully good solutions to the puzzles raised by the class of temporal and aspectual adverbs in English but mostly in Romanian.

CHAPTER 2 ASPECTS OF SYNTACTIC THEORY. THE MINIMALIST FRAMEWORK The analysis developed in this book is based on the Minimalist Inquiries framework of Chomsky (2000, 2001). The purpose of this chapter is to introduce those aspects of the theory that exhibit relevance for the study of temporal and aspectual adverbs and the VP structure of Romanian. The discussion is relatively detailed; thus, in section 2.1 we discuss the main ideas of the Minimalist Inquiries framework, and in section 2.2 we take a closer look at functional projections since there are many studies on the syntax of adverbs that are centered on this issue. Section 2.3 provides an overview of the structure and properties of Romanian sentences which is based on several existing analyses of the language. 2.1. Introduction Within Chomsky s (20001, 2001) Minimalist framework, a human language L represents a cognitive system storing information about sound, meaning and structural organization. This information is accessed by the performance systems by the sensorimotor systems and the systems of thought which are external to L. The cognitive system provides information to the performance systems in the form of interface levels: the sensorimotor systems access the phonological form interface level PF, while the systems of thought access the logical form interface level LF. Thus, a language L is a device generating sets of expressions Exp = <PF, LF>, where PF provides instructions to the sensorimotor systems and LF to the systems of thought. The theory of syntax proposed by Chomsky (2000: 100-101) consists of a lexicon and a computational system. Lexical items stored in the lexicon are specified for phonological, semantic and formal features: phonological features are interpreted at PF by the sensorimotor systems, while semantic and formal features, which are further divided into categorical, φ- and case features (Chomsky 1995: 277) are interpreted at LF by the systems of thought. In their simplest form, features are instructions to the performance systems about interpretation. The computational system accesses items of the lexicon and maps them into expressions Exp = <PF, LF> in a manner specified by the computational and economy principles of UG. The former constrain the structure building operations merge and move, while the latter ensure that the derivations formed by the structure building operations are not only convergent but also optimal. At some point, an operation called spell-out splits the

20 DARIA PROTOPOPESCU derivation into two parts: one of them contains elements which are relevant only to interpretation at LF (the path from the lexicon to LF is known as narrow syntax), the other elements which are relevant only to interpretation at PF. (1) PF LF Spell-out Lexicon Spell-out and narrow syntax According to Chomsky (2000: 95) the computation of an expression converges at an interface level if it consists solely of objects which are interpretable to the performance systems accessing that level. To ensure convergence, all uninterpretable features (i.e. features that cannot be interpreted by the performance systems) must be eliminated in the course of the computation, before the derivation reaches the interface levels. Economy conditions require derivations to be optimal. This means, first, that instead of accessing the lexicon continuously at every step of the derivation, speakers reduce operative complexity by making a one-time selection of a lexical array LA from the lexicon which is then mapped to expressions Exp = <PF, LF> (cf. Chomsky 2000: 101). Thus, a lexical array is a collection of lexical items (and their features) from the lexicon, roughly equivalent to the numeration in Chomsky 1995. Second, the economy conditions state that simple operations must be favoured over more complex and hence also more costly operations and, third, that all unnecessary operations must be eliminated. Failure to meet the economy conditions results in wrong interpretations. Operations are allowed if and only if they have effects at the interface levels. 2.1.1. Merge, agree and move The computational system takes derivations to the PF and LF interface levels, by selecting lexical items and generating derivations, in a manner specified by the computational and economy principles of UG. The syntactic component of the computational system is a series of structure building operations. The most elementary of these operations is merge: it takes two syntactic items α and β, and combines them into a new syntactic item χ. If α projects, then the merger of α and β produces a new syntactic item χ with a label K such that K = {α{α, β}} (cf. Chomsky 1995: 243-245): (2) α ru α Set-Merge β

A SYNTACTIC APPROACH TO ADVERBS IN ENGLISH AND ROMANIAN - 21 TEMPORAL AND ASPECTUAL ADVERBS (2) above illustrates what Chomsky (2000: 133-143; 2001: 3-4) has called set-merge. In set-merge, one of the syntactic items, α or β, acts as a selector and is the projecting element. Set-merge is, then, an obligatory operation satisfying some property of the selector. In pair-merge, on the other hand, neither α nor β acts as a selector: pair-merge is an optional operation which adjoins α to β to form an ordered pair <α, β> and which always leaves the category type unchanged. Adjoining α to β thus forms a new syntactic item with a label K such that K = {α, <β, α>}: (3) α ru β Pair-Merge α Features can either be interpretable or uninterpretable. All uninterpretable features have to be eliminated from the derivation before it reaches interface levels. Uninterpretable features are eliminated or deleted by means of another operation, agree. Agree establishes an agreement or checking relation between two items α and β where α has uninterpretable features and β has interpretable features and the features of α are eliminated (cf. Chomsky 2001: 3-6). The structure in (4) below illustrates an instance of long distance agreement between α and β: (4) αp ru α YP [uf] ru Y β [F] agreement A third operation, move, combines agree with merge. Move establishes an agreement or checking relation between an uninterpretable feature of α and a corresponding interpretable feature of β, by merging β to a projection headed by α. The difference between agree and move is that while in agree, the features of α are eliminated when both α and β remain in their original positions, in move they are eliminated by raising β from inside of αp to αp. This is illustrated in (5) where β moves to αp: (5) αp ru β αp [F] ru agreement α [uf] YP ru Y t movement

22 DARIA PROTOPOPESCU In Chomsky (1995), movement of β to αp was driven by the uninterpretable features of β, which is the moving element, while within the Minimalist Inquiries framework, movement is crucially driven by the uninterpretable features of α, which is the target. Chomsky (2000: 102) assumes β to move to αp only when α has an uninterpretable EPP-feature, i.e. an extra feature which allows α to take an extra specifier beyond its s(emantic)-selection. Thus, the EPP determines the positions which are not forced by the projection principle, i.e. positions in which no theta roles are assigned, and which can serve as landing sites for movement, and as positions where expletive pronouns are merged, because expletives do not need a theta role. In (6a) below, we illustrate a situation where α has both a selectional and an extra EPP feature, while (6b-c) illustrate situations where α has one or the other: (6) a. [XP[YP[α ZP]]] α selects YP, ZP and has EPP for XP b. [YP[α ZP]] α selects YP, ZP and has no EPP c. [XP[α ZP]] α selects ZP and has EPP for XP The situation in (6a) allows for α to have both a selectional and an extra EPP feature, with multiple specifiers, whereas the structures in (6b-c) do not allow for that. Since movement involves pied-piping of phonetically overt material to αp, it is a more costly operation than agree or merge, so it should only take place as last resort. If the lexical array contains an expletive, merge of that expletive is preferred over move. If, on the other hand the lexical array does not contain an expletive, move must take place, to provide convergence at the interface. Both situations are illustrated in (7a-b): (7) a. There were three puppies in the basket. b. Three puppies were in the basket. Chomsky (2000: 103; 106; 127) argues that pure merge is only possible to theta positions and therefore, it is restricted to arguments. Non-theta positions can only be filled by movement or by merge of an expletive. Following this line of reasoning, movement or merge of an expletive can never take place to a theta position. Since adverbs are non-theta marked, it follows that they are licensed by movement not merge, and only Alexiadou s complement-like adverbs may be merged being argument-like in their nature. To see how the system outlined so far works, let us assume that the lexical array contains the items and their features: (8) The lexical array: T, D, John, v, boiled, D, eggs Merge applies in pairs, i.e. at every stage, one of the items acts as a selector, so that the operation satisfies some property of the selector, in this case its c- and/or s-selection: (9) a. Merge D, N: [ DP John] b. Merge D, N: [ DP eggs] c. Merge V, DP: [ VP boiled [ DP eggs]] d. Merge v, VP: [ vp v [ VP boiled [ DP eggs]]]

A SYNTACTIC APPROACH TO ADVERBS IN ENGLISH AND ROMANIAN - 23 TEMPORAL AND ASPECTUAL ADVERBS e. Merge vp, DP: [ vp [ DP John] [ vp v [ VP boiled [ DP eggs]]]] f. Merge T, vp: [ TP T [ vp [ DP John] [ vp v [ VP boiled [ DP eggs]]]]] The φ-features of functional T heads are uninterpretable since they have no semantic content, while the φ-features of DPs are interpretable, i.e. systems of thought can make use of the information that a particular DP, such as John, is [3 rd person, singular, +human], while another DP, such as eggs, is [3 rd person, plural, -human]; on a functional T, this information is irrelevant and the only reason for T to have a φ-feature is to show that it enters a relationship with a particular DP. According to Chomsky (2000: 122), uninterpretable φ-features of T act as a probe which seeks a matching goal, i.e. a matching set of features, within T s minimal domain/c-command domain. When the probe has located its goal and the uninterpretable features have been eliminated, the probe erases because of the matching. In (9f) the closest matching goal for T s probe is the interpretable set of φ-features of John, so the uninterpretable features of T can be eliminated in two ways: a) by long-distance agreement between John and T, or b) by raising John to Spec,TP. The choice between these two options depends on the presence of an EPP feature on T: in English, T has an EPP property which requires that something should be merged with the category that it heads. In languages like Romanian, T lacks EPP and a long distance relationship is established between T and the relevant DP. This is in agreement with Cornilescu (2000) and Alboiu (2002) who claim that Romanian is a V-type EPP rather than T-type EPP language, therefore, the VSO order of Romanian is a direct result of T heads lacking EPP and of the finite V raising to T: (10) a. [ TP [ DP John] i [ TP T] [ vp t i [ vp v [ VP boiled [ DP eggs]]]]] b. [ TP A deschis j [ vp [ DP Ion] i [ vp t j [ VP t j [ DP o uşă]]]]] has-3sg. opened Ion a door Ion opened a door According to Chomsky (2000: 123-124), the φ-features of v heads are also uninterpretable. In (9f), the closest matching goal for v s probe is the corresponding feature of the DP eggs : the features of v, just like the features of T can be eliminated either by means of long-distance agreement, or, in languages where v also has an EPP property, by means of raising eggs to Spec,vP. Besides their interpretable φ-features, the DPs John and eggs have an uninterpretable structural case feature. Following Chomsky (2000: 122-124), case features make DPs active so that they can be identified as the goals of some functional head s probe. Although T and v heads do not have case and there are no reasons to believe their probes would manifest such features either, they are in Chomsky s system nevertheless able to eliminate the structural case feature of their goal, i.e. assign nominative or accusative case to the DPs. According to Chomsky (2000:122), structural case is a reflex of an uninterpretable φ-set so that it, too, erases under matching with a probe. Chomsky (2001: 6) argues that structural case, although not a feature of the probes is assigned a value under agreement, then removed by spell-out from the narrow syntax. The question that arises here is what it means exactly for a feature to be eliminated. Following Chomsky (2000),

24 DARIA PROTOPOPESCU Pesetsky & Torrego (2001), uninterpretable features enter the derivation without a value and they receive their values under agreement from a set of corresponding interpretable features. After they are valued, these features must be eliminated from the narrow syntax, because otherwise they become indistinguishable from interpretable features at LF but they are left at the same time available for phonology. The value is then transformed into an actual overt case morpheme in the morphological component of the grammar. This means that the situation in (1) must be replaced by the one in (11) below: (11) PF Morphology LF Spell-out Lexicon Spell-out and narrow syntax Figure (11) above illustrates a situation where elements consist of various kinds of phonological, semantic and formal features up until morphology. The features play a crucial role in determining how the element is finally pronounced, i.e. which affixes it contains, which phonological processes it undergoes, etc. 2.1.2 Derivation by phase So far, we have assumed that the full sentence is constructed before spell-out applies and material is sent to the PF and LF interface levels. Chomsky (2000: 106; 2001: 11-12) assumes that derivations proceed in phases rather than constructing the full sentence at once. It means that after the lexical array LA is selected, a subarray LA i is selected and used to construct a syntactic object SO i. When this happens, another subarray LA j is selected to construct another syntactic object SO j. This proceeds until the LA is exhausted. Chomsky argues that each subarray must determine a syntactic object that is independent with respect to interface properties. Therefore, in this system, each subarray must contain an occurrence of C or little v, determining a clause or a verb phrase. Chomsky (2001:106) takes phases of a derivation to be to be syntactic objects derived by the choice of subarrays and concludes that a phase can be either CP or vp, but not TP or lexical VP headed by H lacking φ-features and therefore not entering into case-agreement checking. Phases are divided into weak and strong phases. Every strong phase must contain (cf. Chomsky 2001:12-13) a complete set of φ-features, which means that each strong phase must contain a φ-complete C or v head, so strong phases are finite CPs and transitive vps. Strong phases are also subject to the Phase Impenetrability Condition (cf. Chomsky 2000: 107-110, Chomsky 2001: 11-14).

A SYNTACTIC APPROACH TO ADVERBS IN ENGLISH AND ROMANIAN - 25 TEMPORAL AND ASPECTUAL ADVERBS (12) Phase Impenetrability Condition In phase α with head H, the domain H is not accessible to operations outside α; only H and its edge are accessible to such operations. In [ ZP Z [ HP α[h β]]], where HP is a strong phase and ZP is the next strong phase, only H and its edge are accessible to operations at ZP. In this case α is a hierarchy of specifiers or elements adjoined to HP constituting the edge of H while β constitutes its domain. The Phase Impenetrability Condition states that only H and α are accessible to operations outside of HP, and they are only accessible to operations within the next highest strong phase ZP. Operations at ZP cannot look into HP beyond its head. So if there is a feature F of Z which acts as a probe seeking a matching goal, then an agreement relation can only be established between the probe and a feature of H, or the probe and a feature of α but not between the probe and a feature of β. According to Gallego (2010), interpretation is ideally done piece by piece, that is, phase by phase, without any backtracking or recycling. For instance, theta-role assignment is taken to operate in the vp cycle (cf. Hale & Keyser 1993), whereas binding rules require a bigger domain, that affects at least the whole clause, and therefore operates in the CP cycle (cf. Chomsky 1995). For Chomsky, the system is designed to reduce complexity, which is done by means of phase-by-phase derivations. Chomsky (2004: 124) argues that [phases] should have a natural characterization in terms of [Interface Conditions]: they should be semantically and phonologically coherent and independent. Gallego (2010) notices that the problem with this assertion is that we do not know what a natural characterization is in advance and proposes the following possible characterizations: Morpho-Phonological Component - syllable structure - stress - intonation - phonological phrasing - categorization - agreement/case - ellipsis Semantic Component - theta roles - binding - control - predicate composition - discourse-oriented interpretations (focus, topic, etc.) Chomsky s intuition is that phases are small: phases should be as small as possible, to maximize the effects of strict cyclicity, hence computational efficiency. (Chomsky 2007: 17) Gallego (2010) questions how small a phase should be. He provides several answers to that question: either they are every lexical items (every feature), in which case the option is too strong since it entails no structure and no compositionality; or, every application of Merge, which triggers two possible solutions: a. Merge (α,β) = {α,β} Spell-Out {α,β} (Epstein & Seely 2002) b. Merge (α,β) = {α,β} Spell-Out {β} (Boeckx 2009)

26 DARIA PROTOPOPESCU It appears that only the b) option (cf. Gallego 2010) provides what Chomsky calls EDGE, i.e. a memory buffer and seems adequate in order to avoid having to restart the computation after every application of Merge. As far as what triggers Spell-Out is concerned, Chomsky (2000) argues that it involves the existence of uninterpretable morphology, and φ-features are in v and C. Moreover, what the Phase Impenetrability Condition captures is the fact that compositionality imposes no going back to previous stages, i.e. no changes to what has already been assembled. Therefore, derivations proceed in phases and phases are in turn subject to the Phase Impenetrability Condition acting as barriers that restrict movement to higher positions. Elements inside one and the same phase may sometimes be equidistant for movement to a higher structural position. Only phase-mates (v and V, or T and C) can interact for compositionality purposes: C, T and external arguments interact for form/ interpretation effects, while v, V and internal arguments interact for form/interpretation effects (cf. Gallego 2010). 2.2. Functional projections As known, linguistic items are divided into two main types: substantive (lexical) and functional. Substantive items have descriptive content, nouns, adjectives and lexical verbs. Functional items are made up of elements having purely grammatical functions, auxiliary verbs and grammatical morphemes, such as tense, mood, etc. The distinction between lexical and functional categories is important to the extent that, in line with Chomsky (2000: 127), theta role assignment is only possible inside of substantive/lexical categories, i.e. all the arguments of a noun, adjective or lexical V are merged inside a projection of that noun, adjective or lexical V. Agreement and movement are only possible inside functional categories, i.e. in configurations where one of the elements is functional and the other one lexical. The distinction between lexical and functional categories was first discussed in Stowell (1981) and adopted by Chomsky (1986) and related work. By extending X -theory, Chomsky (1986) assumed English sentences to have the following base generated structure: (13) [ CP Spec [ C C [ IP Subject [ I I [ VP Spec [ V V Object]]]]]] Fukui & Speas (1986), Sportiche (1988) and Koopman & Sportiche (1991) brought the first major revision to this structure, namely the VP-internal subject hypothesis, which stated that the subject of the sentence was base generated in Spec,VP, moving in those languages where no case was assigned to the Spec,VP position, to Spec,IP to receive nominative case from I under Spec-Head Agreement. The advantage of this proposal was that it allowed all arguments of V to be realized inside a projection of V: (14) [ CP Spec [ C C [ IP Subject i [ I I [ VP t i [ V V Object]]]]]]

A SYNTACTIC APPROACH TO ADVERBS IN ENGLISH AND ROMANIAN - 27 TEMPORAL AND ASPECTUAL ADVERBS Pollock (1989) proposed that the tense and agreement features located under I should be treated as independent functional categories heading T(ense)P and Agr(eement)P. According to Pollock, Tense hosts features for finiteness and tense/mood, while Agreement is the locus of subject-verb agreement. He further argued that when present, negation was another functional category heading its own maximal projection. In Pollock s system, therefore, English sentences had the following structure: (15) [ CP [ C C [ TP Subject i [ T T [ NegP [ Neg Neg [ AgrP [ Agr Agr [ VP t i [ V V Object]]]]]]]]]] Within these theories, functional categories were closely associated with inflectional morphology and word formation processes because functional categories were supposed to host inflectional morphemes which the lexical stems then picked up when moving to a higher position, or which were lowered onto lexical stems in their base positions. Within early Minimalist frameworks, including those proposed by Chomsky (1995), it was assumed that lexical items entered the derivation in their fully inflected forms, but were associated with features that needed checking in the syntax against the corresponding features of the appropriate functional head. Within these frameworks, the number of clausal Agreement projections was increased by adding AgsSP responsible for subject-verb agreement and for checking nominative case on the subject DP, and AgrOP responsible for checking accusative case on the direct object DP. The strength of the feature also determined whether it was checked overtly or covertly, thus, strong features driving overt movement, while weak features driving covert movement to the relevant Spec,AgrP position: (16) [ CP [ C C [ AgrSP Subject i [ AgrS AgrS [ TP [ NegP [ Neg Neg [ AgrOP Object j [ AgrO AgrO [ VP t i [ V V t j ]]]]]]]]]]] According to Chomsky (1995: 24), the postulation of functional categories must be justified either by output conditions (phonetic and semantic interpretation) or by theory internal arguments. Functional categories like T and D are justified by output conditions, to the extent that T has interpretable features such as [±Finite] and [±Tense], while D has features like [±Referential]. Functional categories such as Agreement, on the other hand, are only associated with uninterpretable features, so their presence in the structure must be justified by some theory-internal arguments. Chomsky (2000: 102) argues that all functional categories must be justified by output conditions. According to him, the core functional categories are C (expressing mood/force), T (expressing tense/event structure) and v (the little or light verbal head of transitive constructions). He also assumes each of these heads to contain an uninterpretable φ-feature set which must be checked and valued under agreement, against the corresponding features of the closest DP, and an EPP-feature, which allows heads to be targets of movement (Chomsky 2001:7) states that C, T and v are only cover terms for a richer array of functional categories. The presence of an EPP feature on a functional head is subject to parametric variation. C is the target of Wh movement, while T checks nominative case and hosts the derived grammatical subject. Little v, according to Chomsky (2000: 123-124), checks accusative case and hosts the derived