Existential constructions: a syntactic predication approach

PDF

Download

Open document

Transcript (if available)

Content EXISTENTIAL CONSTRUCTIONS: A SYNTACTIC PREDICATION APPROACH by Aaron Hyde Walker University of Southern California ii Dedication I dedicate this dissertation to my Grandparents iii Acknowledgements First and foremost, I would like to thank my dissertation advisor, friend, and mentor, Hagit Borer. Without her help, guidance, and encouragement I would not be finishing this project. The long hours, the comments and discussions on drafts, re-‐writes, and re-‐re-‐writes, pushed me to produce beyond what I believed I was capable. No matter how much she had to do, or where she was in the world, she always made time to help and mentor me to become the best academic, and the best person I could be. It is truly a privilege to have been your student. I would also to like to thank my dissertation committee: Audrey Li and Scott Soames for taking the interest in my work, and for their input and help in the development of this project. Audrey Li deserves an extra thank you for suffering through my Chinese questions. On a separate but related note, I would like to thank Isabelle Roy for taking the time to meet with me and discuss the ideas in this dissertation. My experience in the Linguistics Department at the University of Southern California has been shaped by so many wonderful professors. I would like to specifically acknowledge Jean Roger Vergnaud whose kindness and wisdom will always be standards that I will aspire to. Also, thank you Maria Luisa Zubizarretta, Elsi Kaiser, Roumyana Pancheva, Elena Guerzoni, Barry Schein, Andrew Simpson, Jim Higginbotham and Hajime Hoji. I must also acknowledge that if it were not for Joyce Perez, I would never have completed the program (or gotten married). I would iv also like to thank Elly Van Gelderen at Arizona State University for taking me under her wing when I was an undergraduate and inspiring me to pursue my doctorate. My time at USC would not have been the same were it not for the support, camaraderie, and intellectual reinforcement from my fellow students. Without these people I do not believe that I could have thrived as I did, Thank you to: Phil Potamites, Ed Holsinger, Xiao He, Iris Ouyang, Heeju Hwang, Stephanie Huang, Canan Ipek, Mythili Menon, Erin Tavano, Janet Anderson, Wendy Cheng, Laura Tejada, and Sam Tilsen. Also, there are few people without whom my time in LA would have been nowhere near as amazing as it was. To my great friends from near and far, I could not have made it without you—you always helped me put it all in perspective: Nathan Saper, Julie Saper, Oliver White, John Kephart, Maggie Buckles, Matt Akana, Steven Rafferty, Corey Turoff, and Bill Prather. I would also like to thank my family for their years of emotional and financial support through everything it took to get here; what a long, strange trip it’s been. Thank you to my parents and grandparents for always believing in me. I’d like to thank my brother and best-‐man Grant Walker for always being ready to discuss, challenge, and question anything and everything. Thank you Socrates for dutifully sitting with me for years. To list all of the reasons for which I owe my wife, best friend, and fellow linguist, Yi-‐Hsien Liu Walker, thanks would fill volumes… and the list gets longer every day. v Sometimes there’s a man...and well…he’s the man for his time and place. -The Big Lebowski vi Abstract This dissertation is concerned with the specific relationship between copular sentences and existential sentences, and what about that relationship is responsible for the well-‐known restrictions on definiteness and predicates that are observed in the existential cases. Instead of starting with the well-‐known paradigms which illustrate restrictions in the existential constructions that result in ungrammaticality; this dissertation starts by trying to account for the more apparent flexibility the existential construction has when it comes to agreement. While previous accounts of these phenomena have sought characterize such agreement phenomena as marked or some sort of default agreement or just a PF/phonological phenomena with no semantic/pragmatic implications, I will argue the opposite. I will suggest that the singular/plural agreement is a result of the mass/count properties of the Predicational Structure from which the sentence is derived. Furthermore, I will suggest that it is precisely these mass/count properties that account for the observed restrictions existentials. Additionally, since the theory being advanced here is one that is dependent on mass/count properties, which are universally available in natural languages (albeit encoded and manifested in different ways), it allows us to understand related phenomena in typologically distinct languages, such as Mandarin. That is, we can explain why it is that in a “Classifier Language” which encodes mass/count differently than languages like English and French with plural morphology and articles, exhibits the definiteness vii effect but not the predicate restriction. I will propose that if we understand the existential constructions to be derivationally related to their copular counterparts, and we have a sufficiently fine-‐grained understanding of predication in copular constructions, then we can not only account for the behavior and restrictions of existentials vis-‐à-‐vis copulars within a language, but we can understand the difference between restrictions on existentials across typologically distinct languages in a unified and principled way. viii 9 Table of Contents DEDICATION II ACKNOWLEDGEMENTS III ABSTRACT VI CHAPTER 1: EXISTENTIALS, COPULARS, AND PREDICATION 11 SECTION 1. INTRODUCTION 11 SECTION 2: THE AGREEMENT PROBLEM IN THE ENGLISH EXISTENTIAL CONSTRUCTION 13 SECTION 3: THE EXISTENTIAL CONSTRUCTION, COPULARS, AND A SMALL CLAUSE APPROACH TO PREDICATION 18 SECTION 3.1: NON-‐VERBAL PREDICATION IN COPULAR CONSTRUCTIONS 19 SECTION 3.2: PREDP AND LEXICAL CATEGORIES 28 SECTION 4: DISSERTATION PREVIEW 33 CHAPTER 2: AGREEMENT VARIATION IN EXISTENTIAL CONSTRUCTIONS AS EVIDENCE OF STRUCTURAL VARIATION 36 SECTION 1: THE ‘OCCASIONAL CONSTRUCTION’ AND THE EXISTENTIAL CONSTRUCTION 37 SECTION 1.1 THE OCCASIONAL CONSTRUCTION 38 SECTION 1.2: THE OCCASIONAL CONSTRUCTION AND AGREEMENT VARIATIONS IN THE EXISTENTIAL CONSTRUCTION 42 SECTION 2: THE SYNTAX OF THE EXISTENTIAL CONSTRUCTION 51 SECTION 2.1: A STRUCTURAL ANALYSIS OF THE OCCASIONAL CONSTRUCTION DATA—AN INITIAL SKETCH OF THE TOPOGRAPHY OF EXISTENTIAL CONSTRUCTIONS 51 SECTION 2.2: ACCOUNTING FOR EC AGREEMENT VARIATIONS GENERALLY 60 SECTION 2.3: MASS/COUNT DISTINCTION: INTERPRETATIONS AND STRUCTURAL RAMIFICATIONS 62 SECTION 2.4: ANALYZING THE DERIVATIONS 64 SECTION 2.5: FLOATING QUANTIFIERS, AGREEMENT AND THE EC 66 SECTION 3: THE EXISTENTIAL CONSTRUCTION, LOCATIVES, AND PREDICATION 70 SECTION 3.1: RELATOR PHRASE (RP) VS. PREDP 70 SECTION 3.2: EXTENDING THE SITUATION-‐DESCRIPTIVE, CHARACTERIZING, AND DEFINING PREDICATIONAL SYSTEM: DPS AS PREDICATES 72 SECTION 3.3: PREDICATE INVERSION, LOCATIVE INVERSIONS, AND EXISTENTIALS 77 SECTION 4: THERE’S VS. THERE IS 90 SECTION 5: CONCLUSION 94 CHAPTER 3: PREDICATIONAL RESTRICTIONS IN THE ENGLISH EXISTENTIAL CONSTRUCTION (EEC) AND THE DEFINITENESS EFFECT 96 10 SECTION 1: RECONCEPTUALIZING THE PREDICATION RESTRICTIONS IN THE EEC 96 SECTION 1.1: THE IL/SL DISTINCTION AND THE EEC 98 SECTION 1.2: THETIC/CATEGORICAL PREDICATION, THE IL/SL DISTINCTION, AND THE EEC 100 SECTION 2: A THETIC PREDICATIONAL APPROACH TO THE IL/SL RESTRICTION IN EECS—BASILICO (1997) 105 SECTION 2.1: THERE AS A TOPIC MARKER (TM) AND THE IL/SL PREDICATE RESTRICTION 106 SECTION 2.2: BASILICO (1997)—THEORETICAL AND EMPIRICAL PROBLEMS 111 SECTION 3: THE DEVIL IS IN THE DETAILS—ABANDONING IL/SL AND SALVAGING THETICITY 113 SECTION 4: THE PREDICATE RESTRICTION EXPLAINED: THE EEC IS (A VERSION OF) SITUATION- DESCRIPTIVE 121 SECTION 5: THE DEFINITENESS EFFECT AND THE EEC—A NON-SMALL CLAUSE SEMANTIC APPROACH THE DE 125 SECTION 6: A SMALL-CLAUSE APPROACH TO THE DE—RETHINKING PRESUPPOSITION AND THETICITY 126 SECTION 6.1: DERIVING THE DE FROM SITUATION-‐DESCRIPTIVE PREDICATION 135 SECTION 7: CONCLUSION 141 CHAPTER 4: BEYOND THE ENGLISH EXISTENTIAL CONSTRUCTION: THE MANDARIN EXISTENTIAL CONSTRUCTION AND PREDICATION 143 SECTION 1: INTRODUCTION 143 SECTION 2: PREDICATE NOMINALS IN MANDARIN 150 SECTION 2.1: MANDARIN NOMINALS AND BARE NOUNS 154 SECTION 3: MANDARIN NOMINALS IN THE ROY SYSTEM 158 SECTION 4: THE MANDARIN EXISTENTIAL CONSTRUCTION 165 SECTION 4.1: NP AND CODA ARE A CONSTITUENT 169 SECTION 4.2 THE CODA IS A PREDICATE OF THE POST-‐VERBAL NP 171 5. THE SYNTAX OF THE MANDARIN EXISTENTIAL CONSTRUCTION 177 CONCLUSION 180 REFERENCES 182 11 Chapter 1: Existentials, Copulars, and Predication Section 1. Introduction This dissertation is concerned with the specific relationship between copular sentences and existential sentences, and what about that relationship is responsible for the well-‐known restrictions on definiteness and predicates that are observed in the existential cases (1-‐2): (1) a. A man/Some men is/are in the garden b. The man/Every man is in the garden c. Firemen are available d. Firemen are medics (2) a. There is/are a man/some men in the garden b.*There is the/every man in the garden c. There are Firemen available d. *There are firemen medics To be sure, these types of sentences and the restrictions they exhibit have been the subject of thousands of articles in the literature through the decades. This dissertation is different than these articles in one key sense. Instead of starting with the well-‐known paradigms in (1-‐2) which illustrate restrictions in the existential constructions that result in ungrammaticality; this dissertation starts by trying to account for the more apparent flexibility the existential construction has when it comes to agreement: 12 (3) a. Five cops/seem to be are at the door. b.*Five cops is/seems to be at the door c. There’s/are five cops at the door d. There seem/seems to be five cops at the door While previous accounts of the phenomena in (3) have sought characterize it as marked or some sort of default agreement or just a PF/phonological phenomena with no semantic/pragmatic implications, I will argue the opposite. I will suggest that the singular/plural agreement is a result of the mass/count properties of the Predicational Structure from which the sentence is derived. Furthermore, I will suggest that it is precisely these mass/count properties that account for the observed restrictions in (1-‐2). Additionally, since the theory being advanced here is one that is dependent on mass/count properties, which are universally available in natural languages (albeit encoded and manifested in different ways), it allows us to understand related phenomena in typologically distinct languages, such as Mandarin. That is, we can explain why it is that in a “Classifier Language” which encodes mass/count differently than languages like English and French with plural morphology and articles, exhibits the definiteness effect as in (2b) but not the predicate restriction illustrated in (2d): 13 (4). a. There is [SC *the/*every/a manp-‐v subject *(in the garden)] b. you [*zhe-‐ge/*mei-‐ge/ yi-‐ge ren *(zai huayuan-‐li)] have this-‐Cl every-‐Cl one-‐Cl person at garden-‐in (5). a. There is a student in the yard b. *There is a student (is a) doctor c. you yi-‐ge xuesheng zai huayuan-‐li have one-‐Cl student at garden-‐in d. you yi-‐ge xuesheng shi yisheng have one-‐Cl student be doctor I will propose that if we understand the existential constructions to be derivationally related to their copular counterparts, and we have a sufficiently fine-‐ grained understanding of predication in copular constructions, then we can not only account for the behavior and restrictions of existentials vis-‐à-‐vis copulars within a language, but we can understand the difference between restrictions on existentials across typologically distinct languages in a unified and principled way. Section 2: The Agreement Problem in the English Existential Construction The English Existential Construction (EC), or there-‐insertion construction, is a subgroup of a larger family of Expletive (EXPL) constructions that includes presentational/list constructions and deictics. The EC has been a topic of much discussion in linguistics covering an array of research interests ranging from its nuanced semantic restrictions such as: sensitivities to Definiteness and weak/strong quantification (Milsark 1974, 1977 a.o.; Safir 1985) and Stage level vs. Individual level predicates (Carlson 1977); to a test ground for early minimalist feature checking and movement schemas (Chomsky 1995, Groat 1999). Ultimately, the 14 interest in this construction is due to its semantic similarity with its non-‐expletive variants, its stable and restricted syntactic behavior, and the environments in which it appears. That is, there has long been the sense that there is a derivational relationship between the (a) and (b) variants in the following pairs 1 (e.g. Chomsky 1965): (6) a. Five cops are at the door b. There are five cops at the door (7) a. A cop is at the door b. There is a cop at the door From a syntactic standpoint, the question that immediately arises with the EC variants in relation to their non-‐EC counterparts is: what is the subject of the sentence? This question arises because there seems to be a contradiction if we apply the normal diagnostics of subjecthood to the EC sentences (Basilico 1997). Subjects of sentences are normally defined as in (8) (Chomsky 1995 a.o.): (8) DP is the subject of a sentence if DP satisfies both of the following conditions a. DP checks the EPP feature of finite T 0 (e.g. DP is in spec TP) b. DP agrees with finite T 0 for phi-‐features 1 I leave open at the moment which variant was derived from the other, but proposals exist both ways with varying flavors 15 However, when the two conditions are applied to the EC sentences, the element that meets condition (8a) seems to be the EXPL there, and the element that appears to meet condition (8b) (as manifested by the agreement morphology on the copula) is the post-‐copular DP five cops at the door or a cop at the door respectively. Many proposals exist in the literature that seek to link the two elements in some way so as to resolve this seeming contradiction while still having the system explain/predict the nuanced semantic/syntactic distinctions briefly mentioned above. Stowell (1978) argues for a particular conception of the copula be as being transitive (taking an event/situation complement) and thus linking the EXPL with the post-‐copular NP by the subcategorization frame of the copula itself. Chomsky (1986b) argues that the post-‐copular NP replaces the EXPL at LF; later proposals (Chomsky 1991, 1995) argue that the LF movement is either affixation of the NP to the EXPL, or simply covert feature movement of the relevant features at LF to the EXPL. Safir (1985) takes a different approach, instead arguing that the EXPL and the post-‐copular NP are linked by way of a binding chain at S-‐Structure. For Safir (1985) the D-‐structure of an EC is as in (9) with an S-‐Structure in (10): (9) e are [sc five cops [PP at the door]] (10) therei are [SC five copsi at the door] (modified from Safir (1985): 96-‐7) The EXPL there is inserted in the empty position in (9) and a binding chain is formed between the EXPL and the subject of the SC in (10). The binding chain serves to 16 syntactically link the θ-‐marking A-‐position and the case marking A’-‐position such that EC sentences will not be ruled by the Case Filter. However, these proposals and, to my knowledge, those subsequent to them, have largely left unmentioned the curious patterns of verbal agreement in ECs with raising predicates and their consequences for interpretation. That is, comparing the pairs of sentences in (11-‐12), one finds an interesting contrast: (11) a. There are five cops at the door b. *There is five cops at the door 2 (12) a. There seem to be five cops at the door b. There seems to be five cops at the door The contrast is that while singular agreement is ruled out in (11b) (although see footnote 2) the singular agreement morphology is grammatical in the raising context in (12). Moreover, data from native English speakers reveals that they observe a meaning distinction between (12a) and (12b) briefly (and non-‐ technically) summarized in (13): (13) a. Determiner Reading (12a)—There exist five cops, and they are at the door; the plural agreement gives the reading that what is being focused on is the cops specifically or the number of cops involved in the event of being at the door b. Prepositional Reading (12b)—There exists an event/situation of ‘being at the door’ and the participants in that event/situation happen to be five cops; the 2 The sentence in (7a) is rectified, however, if a contraction used as in (i): (i) There’s five cops at the door 17 singular agreement in (12b) yields the reading that what is at issue in the sentence is the state of affairs such that there are some cops at the door and the fact that there are five of them is just extra information Given this data, this chapter will seek to develop a syntactic account of ECs that not only predicts the already well-‐documented syntactic/semantic restrictions but also correctly predicts the availability of the readings in (13) in appropriately restricted contexts. Furthermore, the chapter will seek to clarify and more precisely define the readings in (13) as the interpretational consequences of principled structural variations in the predicates of ECs. There seem to be conditions, however, under which the availability of the non-‐agreeing pattern (12b) is not licensed, and only the option in (12a) is available. The conditions that limit the agreement options involve embedding ‘occasional DPs’ in the EC as in (14): (14) a. There seem to be the occasional girl scouts knocking on my door b. *There seems to be the occasional girl scouts knocking on my door The agreement licensing phenomena in these examples will offer a set of diagnostic tests to determine the structures of ECs in such a way that accounts not only for the well-‐known restrictions, but also the variable agreement patterns that are both grammatical albeit semantically distinct. It will be the contention of this chapter that the structure and quantificational properties of Predication Phrase (PredP) in ECs both determine the verbal agreement of the raising verb/copula and correctly account for the observed restrictions cited above as well as predict the availability of the readings in (13). 18 One significant difference between this proposal and the ones mentioned above is that it does not posit a link between the EXPL and the DP (or NP in the terms of the earlier proposals), rather it asserts that the link should be between the EXPL and the PredP head. Section 3: The Existential Construction, Copulars, and a Small Clause Approach to Predication There is by no means an agreement on the nature of copular structures. Furthermore, copular sentences are by no means a monolithic category. There are different interpretations of copular sentences, most notably those defined by Higgins (1977) 3 ; however, while the interpretational differences are fairly well agreed upon, the syntactic structures and/or lexical variations that are responsible for them are the subject of much debate. One line of reasoning has discussed the possibility of the copula itself being responsible for the interpretational differences. Two variations of this line of reasoning involve either: (a) postulating the existence of multiple copulas each with a particular semantic function and (perhaps) lexical subcategorization (Roy 1995: 15-‐7) or (b) assuming that only one form of the copula exists and locating the onus of interpretive difference into the semantic 3 (i) Identificational That animal is a dog (ii) Identity (Equative) The Morning Star is the Evening Star (iii) Predicational Bill is a scuba diver (iv) Specificational The person I like most is my father (Hegge 1988:5) 19 machinery of type shifting (see discussion of Chierchia (1985) above; Williams (1983)). The other line of thought locates the differences in the interpretation in structural differences in larger functional structures involving at minimum the SC (Stowell 1978; Moro 1996) and sometimes including PredP 4 (Bowers 1993; Roy 2005; Hegge 1988). The spirit of these proposals is that variations in the SC/PredP structures and various A’ movements can more economically account for the interpretation variations observed in copular sentences. The proposal of this dissertation falls within this second line of thought and seeks to explain the existential interpretation as a derivationally related structure to the copular structure and one whose semantics varies predictably with the syntax. Section 3.1: Non-‐verbal Predication in Copular Constructions Within the literature about copular sentences, there has been a tradition of attempting to account for the interpretational differences (and categorical restrictions) as a result of the type of predicate in the sentence. The distinction, in some way or another, usually reduces to whether or not the predicate denotes a transient or stable property (e.g. S-‐level/I-‐level predicate distinction of Carlson 1977). However, Roy (2005) argues that the traditionally assumed distinction 4 The formal/terminological realtionship between PredP and SC is somewhat idiosyncratic in the sense that various proposals, particularly Roy (2005), equate the two with the formal consequence of creating a properly headed SC. Other propoosals, Baker (2003, in press), allow PredP and SCs to exist simultaneously within the same derivation. His claim is that Pred is present even at the sub-lexical level, but it can also function in virtually the same way as in Roy (2005); See section 3.1.2 for discussion of Baker, 20 between I-‐level and S-‐level (permanent/transient) properties is insufficient in predicting and describing the interpretation of copular complements. Instead, languages make a ternary distinction between copular sentences: defining, characterizing, and situation-descriptive. Roy begins her analysis with a discussion of French copular constructions, first observing the status of copular predicates that vary with respect to the presence/absence of the indefinite determiner: (15) a. Paul est un acteur (Stable property/defining) P is an actor ‘P is an actor’ (Cf. English—Q: What does P do? A: P is an actor) b. Paul est acteur (eventive property) P is actor ‘P is an actor’ (Cf. English —Q: What does P do? A: P acts) There are supposedly two restrictions on the structures in (15): a) noun type and b) [+/-‐] human 5 . However, Roy illustrates that these distinctions are not absolute and can be accounted for by independent phenomena. Roy identifies a class of French Ns that have the property of an eventive reading when they appear as bare Ns in predicates. Ns that allow bare form structure include professions (medicin doctor, 5 The discussion of the noun-type condition will be recapitulated here given its event structure relevance to the current paper. Discussion of the +/- anim. distinction will be omitted 21 vigneron wine grower), hobbies (collectionneur collector, chasseur hunter), occupation (estudiante student, braconnier pilgrim), and “more generally with any predicate that entails and even in the pre-‐theoretical sense (cavalier horse rider, fumeur smoker...)” (Roy 2005: 32-‐33). To determine the status of a predicate as either eventive or stable, Roy suggests a “contradiction test” wherein a bare N predicate is compared to its article counterpart when they are embedded in conjoined sentences that result in a contradiction if the predicate in question is interpreted as eventive: Contradiction Test to determine eventive/stable 6 : (16) a. # Paul est medecin, mail ne pratique plus # Paul practices medicine, but he does not practice anymore b. Paul est un medecin, mail ne pratique plus Paul is a doctor, but he does not practice anymore (17) a. # Paul est invrogne , mais il n’a pas touche un verre depuis 15 ans. # Paul drinks, but he hasn’t touched a single glass in 15 years b. Paul est un invrogne , mais il n’a pas touche un verre depuis 15 ans. Paul is a drunkard, but he hasn’t touched a single glass in 15 years While these predicates seem well behaved with respect the I-‐level/-‐S-‐level distinction, when she analyzes the class of French Ns that is not supposed to allow for bare Ns 7 , there are cases in which bare N variants of these Ns are licensed. The 6 This test works well for professions, but the genrealization holds true for statives—see Roy 2005 pp. 34-5 7 Some Ns that seem to require an indef. D: (i) a. le lion est *(un) felin the lion is a feline 22 property reading of this class of Ns is strongest but the eventive reading turns out to be possible, in which case the indef. D can be omitted. The conditions under which the bare N structure is licensed with these verbs seem to be limited to intentional/conditional/possible-‐worlds contexts (Roy 2005 pp. 36-‐8). Ultimately, Roy illustrates that the ‘indefinite article + N’ construction is unique in the sense that it can only appear in postcopular position and no other predicative position (SC complement, antecedent to a non-‐restrictive relative clause, etc...). Furthermore, she notes, the ‘indefinite article + N’ structure is only licensed in the context of unbound lexical verbs and aspectually imperfect copulars. Roy proposes that these nominals constitute the defining predicates. The semantic test proposed is the Who is X? test since the answer to that question is a statement about the identity of X and therefore about a (fairly) permanent property (Roy, 2005: 51): b. Camille est *(un) homme, et non pas *(une) femme C is a man and not a woman c. Croc-‐Blanc est *(un) lupe White-‐Fang is a wolf (ii) a. Mon grand frere est *(un) heros My big brother is a hero b. Picasso est *(un) genie P is a genius c. Gabin est *(un) enfant G is a child 23 (18)Who is X? Test a. Qui est Paul? -‐Paul est un invrogne/* invrogne /*ivre Defining Who is Paul Paul is a drunkard/* drunkard/*drunk (adj) b. Qu’est Paul? -‐Paul est *un invrogne/invrogne /ivre Characterizing What is Paul? Paul is *a drunkard/ drunkard /drunk This test, however, only coarsely divides the types of copular predicates in the sense that it simply draws a distinction between bare Ns/As/PPs/other predicative categories besides N on one hand (Characterizing copular predicates) and ‘indefinite article + N’ constructions on the other 8 . That is, it appears from the simple test above that As and bare Ns have the same predicational distributions. However, while the bare N invrogne and the A ivre both constitute felicitous answers to the question ‘What is Paul?,’ the bare N cannot function as an answer to a question about a situation hence the delineation of the third type of copular predicate—the situation-‐descriptive predicate (Roy 2005: 67): (19) Qu’est ce qui s’est passe dehors, c’est quoi tout bruit? What’s going on outside, what is all this noise about? a. Paul est ivre, il a renverse la poubelle comme toujours Paul is drunk, he knocked the trash can over as always b. # Paul est invrogne, il a renverse la poubelle comme toujours Paul is a drunkard, he knocked the trash can over as always 8 This simple two way distinction is problematic because it also homogenizes three distinct subtypes of bare Ns (e.g. Transient candidat ‘candidate’ class, Permanent/stative fils de diplomat ‘son of a diplomat’ class, and Mixed medecin ‘doctor’ class) with different distributions and aspectual properties. 24 Essentially, Roy 2005 argues for a three way distinction between copular predicates that is really the result of two overlapping two way distinctions: (19) Predicational copular sentences ‘property-‐assigning’ sentences situation-descriptive sentences defining sentences characterizing sentences (Roy 2005: 90) Ultimately, in the Roy 2005 system, the interpretational differences, the aspectual/event-‐structure differences, and the categorical selectional distinctions between the three types of copular predicates are derived from the ways in which the predicates encode their semantic values of intensity and divisibility in the functional syntactic projections that give rise to the mass/count distinction. Roy argues that the semantic difference between situation-‐descriptive sentences and characterizing sentences is that: Situation-‐descriptive predicates denote necessarily properties that are intense, where intense means that they are true of an event only if the evidence that makes the predicate true holds for the duration of the event without any interruptions...In other words, all parts of the event are identical...Characterizing predicates on the other hand are not intense...while they hold of events in a certain interval, they typically allow for gaps, i.e. episodes where the evidence that makes the predicate true does not hold. (Roy 2005: 166) 25 She defines (non)-‐intensity formally as: (20) a. If a predicate P is intense, then ∃(e)vent ∧ P(e) = 1 in an (I)nterval iff for any I’, and subinterval of I, ∃(e’) s.t. P(e’) = 1 b. If a predicate P is non-‐intense, then P(e) = 1 in I iff P’(e) = 1 in a qualifying amount of non-‐overlapping subintervals of I ∧ ∀(P), P’ is a corresponding intense predicate Thus, a situation-‐descriptive sentence like John was at the door is intense because it is only true so long as John is actually at the door. If he was at the door for five minutes, then left to go get a sandwich for twenty minutes and then returned to the door and waited five more minutes, the sentence John was at the door would be infelicitous with respect to the half hour interval. However, a characterizing sentence such as John is an actor does not imply that John acts all the time. It implies that there are some times (subintervals) when the predicate actor could apply intensely—that is there are times where he is acting without interruption— and that the sum of these times is sufficiently large (a concept that is pragmatically determined) to warrant the label actor. While it is clear that the definitions of (non-‐)intensity make reference to the notion of sub-‐parts of events and subintervals, what is in unclear is how to translate this distinction of intense/non-‐intense into the syntactic mass/count architecture 26 without formal notions of divisibility and cumulativity to define sub-‐parts with respect to an event and an interval: (21) a. P is divisive iff (∀x) (∀y) y < x ∧ P(x) ∧ y ≠ 0 → P(y) and (∀x ≠ 0) [P(x) → ∃(y ≠ 0) y < x] b. P is cumulative iff ∀x[P(x) ∧ P(y) → P(x∪y)] This definition amounts to saying that divisive predicates are non-‐atomic since any division of them will result in an equivalent truth value when the predicate is applied to the result of the division. This means that intense predicates are divisive since they must be true of all sub parts of an event. Non-‐intense predicates are non-‐ divisive since they allow for an atomic event structure such that the event/property they describe may be discontinuous. With these notions established, a clear parallel emerges between the notion of mass/count and intense/non-‐intense, such that Roy concludes: “Intense predicates range over mass events...[and] non-‐intense predicates range over count nouns” (Roy 2005: 187). Having established the connection between intensity/non-‐intensity of predicates and the divisibility/cumulativity of their event interpretations with respect to the mass/count distinction, Roy’s three-‐way distinction is set to explain the categorical restrictions on what can constitute an intense or non-‐intense predicate: (22) a. Only Ns denote atomic eventualities b. All other non-‐verbal predicates denote mass eventualities 27 Roy argues that lexical categories must be embedded in some functional structure, and that the properties associated with that functional structure determine that item’s status as mass or count. Since only Ns can be embedded under the requisite functional structure (namely, Cl(assifier)P and Num(ber)P) to be interpreted as count, it follows that in the predicative context, Ns are the only category that could be non-‐intense. All other predicative categories are under functional structure (e.g. Deg(ree)P for A) but since all lexical categories enter the derivation with a mass interpretation and the functional structure in which they are embedded cannot yield a count interpretation these categories must be mass and therefore intense. The proposal for the structure of copular sentences with non-‐verbal predication in Roy (2005) is that the subject and post-‐copular predicate begin in a predication relation as a SC headed by Pred which licenses the external argument. Since the intensity and divisibility are governed by the available functional structure under which a given category can be embedded, it follows that only one SC structure is necessary and the variation in predicate types is purely based on the properties of the complement of Pred (Roy 2005: 214): (23) PredP = SC DP Pred’ Pred YP {event e} red in the garden doctor 28 The rest of the details of the derivation are not important here, but the rest of the story follows as one might imagine: DP moves up the tree for case and EPP, and BE is semantically vacuous and just a phonologization of tense. What is crucial about this analysis is the notion that there is one structure for copular sentences wherein all the interpretational and structural differences can be accounted simply by the structure of the complement of Pred. If, as I mentioned in section 1, there is a derivational relationship between copular and existential constructions and the account of copulars offered in Roy (2005) is on the right track, then it stands to reason that we should find a similar single existential structure from which we can derive the interpretational and categorical restrictions from a single point of structural variability. Moreover, if Roy’s notion of mass/count structure being responsible for both the notion of event structure and nominal plurality is correct, we will have exactly the type of tools to we need to articulate precise definitions of the heretofore non-‐technical descriptions of the Determiner and Prepositional readings that are obtained from ECs with different copular agreement morphology. Section 3.2: PredP and Lexical Categories Baker (2003) offers a systemic means by which to define lexical categories in terms of syntactic universals. Previous attempts to define lexical categories have generally failed to create definitions that make use of general syntactic features. One example of such a system is the classical Chomskian characterization of lexical categories as a set of +/-‐ valued features (e.g. Noun = +N, -‐V; Verb = -‐N, +V; etc...). 29 This system makes reference to N and V features that are unused and unhelpful anywhere else in syntax; in fact, this system seems to be an almost tautological definition of lexical categories. Baker (2003), in contrast, uses prevalent and simple syntactic/structural distinctions to define the categories in such a way that the definitions themselves make predictions about the cross-‐linguistic behavior of elements of a given lexical category. Baker (2003) defines the lexical categories 9 as in (25): (25) a. Verb has a specifier b. Noun has a referential index c. Adjective has neither specifier nor a referential index The definition of lexical categories in (25) has far-‐reaching syntactic consequences for a number of phenomena and a discussion of all of them is well beyond the scope of this paper. However, the definitions in (25) also raise some immediate questions. Indeed, if as (25) claims, Ns and As do not have specifiers—that is, if Ns and As are inherently saturated—then it is unclear how they could function as predicates. To address this Baker (2003) cites the following example: (26) a. Chris hungers b. Chris (*is) hungry c. Chris (*is) a skier (Baker 2003: 31) 9 In Baker (2003) only N, V, and A are assumed to be lexical categories contra the Chomskian system which says that P is also a lexical category (-N, -V). For Baker (2003) P is a functional category 30 While (26a) does not require copular support (e.g. Tns morphology appears directly on the V with [Chris] ending up in Spec, TP by way of Spec VP) both (26b-‐c) which have adjectival and nominal predicates respectively do require a copula 10 . The argument relies on an intuition from Chierchia (1985) which argues that lexical categories really refer to some property of individuals, but since those individuals and properties are in specific relationships to one another, semantic type-‐lifting functions can be applied to create one-‐place predicates of type <e,t>. In order to achieve this type-‐lifting function, however, there must be more functional structure; Baker (2003), in the spirit of Bowers (1993), argues that the functional projection PredP (with head Pred 0 ) is the locus of this type lifting. Baker thus offers the following (simplified) structures for (26a-‐b): (27) a. TP Chrisi T’ T VP ti hungers 10 This is certainly not the case in all languages but it is a prevalent pattern. Languages which don’t require copular support are argued to have other language specific structural restrictions that facilitate other patterns; these structural restriction, however, are argued to be consistent with the basic definitions given (31) 31 b. TP Chrisi T’ T PredP ti Pred’ Pred AP hungry The crucial difference between the structures in (27) is that in (27a) [Chris] originates in the specifier position (subject position) of the V (although the V has no complement since it is a stative V). In (27b), however, [Chris] is not an argument of a lexical category at all, rather it is introduced in the specifier of the functional projection PredP which licenses it as a subject of the type-‐lifted predicate [hungry]. 11 This allows the system to capture the fact that, “...‘objects’ can be followed by floating quantifiers if and only if they are followed by a predicative category. This predicative category can be an N, A or V” (Baker 2003: 37). These facts are illustrated in (28): (28) a. We consider the men all fools/crazy b. We made the children all cry c. * We saw the men all (Bowers 1993: 618) 11 Baker (2003) uses these facts to motivate his claim that Vs are really derivative categories resulting from the conflation of Pred 0 and an A such that [Pred + A] = V and is lexical (Baker 2003: 36) 32 Bowers (1993) argues that in (28a-‐b) the NP subject starts downstairs in Spec, PredP with the Q adjoined and the Q remains in-‐situ when the NP moves for Acc. case. In (28c), there is no PredP and thus no position for the Q, the sentence is ungrammatical therefore as a result of vacuous quantification. Interestingly, this data seems to be related to agreement patterns in ECs with raising verbs in the sense must be followed by a predicative category in order to license both agreement patterns/readings in (13): (29) a. There seem to be five cops at the door b. There seems to be five cops at the door c. There seem to be five cops d. *There seems to be five cops I will leave this as simply an observation here, but I shall return to it at the end of Chapter 2. Comparing the Roy and Baker systems described in these two subsections, it is unclear whether Pred has the type-‐lifting function in the Roy (2005) system. Also, the two models differ on whether Pred is always the head of an SC with Roy arguing that it must be and Baker arguing the Pred is crucial in lexical formation in the syntax 12 . Ultimately, I will conclude with Baker that Pred and SC can co-‐occur in the same derivation. It is crucial to note, however, that despite the (potentially theory internal) differences between the Roy (2005) and Baker (2003) proposals, they both advocate virtually exactly the same copular structure, and they both conclude 12 See Baker’s (2003) discussion of lexical Vs being formed by conflation of Pred 0 with an A such that the resulting complex is a V (with a specifier) 33 that there is only one and the interpretational differences and categorical restrictions are derived from the syntax of the predicates. Section 4: Dissertation Preview Overall, the dissertation is structured as two parts. In the first part (chapters 2-‐3) I will seek to establish a derivational relationship between copulars and their existential counterparts and I will propose a structure for the English existential construction that seeks to both describe and explain the observed restrictions and behaviors of the existential construction. Since the Roy (2005) system of predication that I am adopting to explain the behavior of English existentials is based on the mass/count properties of the predicate, the second section of this dissertation (Chapter 4) will explore the interactions of mass/count and predication in Mandarin which is typologically distinct and varies with Germanic and Romance languages specifically in how it encodes notions of mass/count. Section 2 will argue that the observed differences in restrictions on English Existential Constructions and Mandarin Existential Constructions can be understood as the manifestation of different mass/count encoding strategies operating in the same predicational framework. Chapter 2 of the dissertation will offer an in depth analysis of the existential construction and the agreement properties it exhibits. Specifically it will develop a set of diagnostic tests using agreement phenomena to probe the structural and semantic properties of the English Existential Construction. The second half of the chapter will closely analyze the special role of locatives in the existential 34 construction and examine this property in the context of proposals for the English copular construction. This section will both offer new and interesting data in support of the syntactic model of existentials advocated here, and it will allow us to extend the Roy (2005) system a bit further in the context of English. Finally, this chapter will conclude with a discussion on the There is vs. There’s distinction that has received fairly little coverage in the literature on the English existential construction. Chapter 3 will address what is commonly known in the literature as the IL- Predicate Restriction and the Definiteness Effect. This chapter will begin by examining the IL/SL distinction and suggesting, following Roy (2005), that it is insufficient to describe or explain the observed data. Next, I will discuss a related distinction, the thetic/categorical distinction as a source of the predicate restriction, but this distinction also fails to account for the data. Ultimately, I will contend that the English existential construction is derived from the situation-descriptive predicational structure in the Roy (2005) taxonomy, and as such is mass. I will then illustrate that both the IL-predicate restriction and the definiteness effect are direct consequences of existential constructions being derived from situation-descriptive predicational structures, and that this system also accounts for and predicts the availability of agreement variations that correspond to semantically distinct readings. Chapter 4 will extend the discussion of a mass/count based predicational system to Mandarin Chinese where I will show that the system can be applied in a language with different mass/count encoding strategies and that the system 35 accounts for the observed differences between copulars and existentials in that language while also explaining the differences in restrictions on existential constructions between the languages. This chapter will first discuss extending the system into Mandarin by examining predicate nominals in Mandarin in copular and existential contexts, as well as examining the syntactic and semantic properties of the Mandarin Classifier system. I will then propose a related distinction to the mass/count distinction that holds in Mandarin and accounts for the observed properties; that distinction is between counted/uncounted predicates. This distinction essentially creates a hierarchical binary distinction between mass/count where ‘mass’ is divided into separate sub-‐types with both distinct and overlapping properties. Then, I will propose a structure and analysis for the Mandarin Existential Construction within the predicational framework previously established. 36 Chapter 2: Agreement variation in Existential Constructions as Evidence of Structural Variation This chapter will be structured into four sections. Section 1 will seek to present some new data pertaining to the contrast in (13) in Chapter 1 by illustrating through quantifier scope and the definiteness effect that verbal agreement patterns are indicative of different structures. It will serve as an overview of the relevant test construction—the Occasional Construction (OC)—following Zimmerman (2003) and how this behaves when it is found in Existential contexts. Section 2 will introduce the syntactic proposal for the English Existential Construction in detail. Drawing on the discussion of predication in copulars from Chapter 1, the first section will account for the data about OCs in EC environments. From there, the section will discuss the availability, the structures, and the semantic differences in the agreement variations that are allowed in English existential constructions (cf. (12) in Chapter 1). This section will conclude with an analysis of the quantifier floating phenomena observed in existential constructions and set the stage for a discussion of secondary predication and locatives generally in copulars and existentials. Section 3 will further develop the proposal by more closely examining the role of locatives in the derivational syntax. Specifically, this section will analyze the Existential Construction in the context of Predicate-‐Inversion proposals (i.e. Den Dikken 2006) and examine syntactic agreement and interpretational consequences 37 of Locative Inversions in Existentials. Lastly, Section 5 will address the observed difference between *There is five cops at the door and There’s five cops at the door. Section 1: The ‘Occasional Construction’ and the Existential Construction Assuming that both sentences in (1) are grammatical, but structurally distinct, we predict the emergence of some contexts in which one structure, but not the other, is licensed 13 : (1) a. There seem to be five cops at the door b. There seems to be five cops at the door The task of this section is to devise a test, such that when it is applied to the two sentences in (1) one sentence is observed to be grammatical and the other sentence ungrammatical. It would then be the task of the structures proposed here for (1a) and (1b) respectively to account for that contrast. 13 However, for instance, the DE by itself cannot distinguish between the two structures in of (i) in (7) nor can the Individual vs. Stage Level predicate distinction (cf. Carlson 1977) of (ii), as both structures are sensitive to both effects in an identical way: (i) a. *There seem to be the/every cop(s) at the door b. *There seems to be the/every cop(s) at the door (ii) a. *There seem to be comets heavy b. *There seems to be comets heavy 38 Section 1.1 The Occasional Construction The construction that will distinguish (1a) and (1b) has been called the Occasional Construction (OC) (Zimmerman 2003; Larson 1999). The construction, illustrated in (2) with readings in (3), has the puzzling property of having a DP-‐ internal Adj that seems to take scope over the DP in which it is embedded at LF such that it binds an event variable: (2) The occasional sailor walked by (3) a. Occasionally, a sailor walked by b. The sailor who sails infrequently walked by c. *Occasionally, the sailor walked by (2) with the interpretation of (3a) means that some events of [a sailor walking by] occurred infrequently during the time span in question. The DP-‐internal Adj. is thus interpreted as binding an event variable outside the DP (see unselective binding— Davidson 1967). In (3b) the expected attributive reading obtains where the Adj’s scope is exactly as one would expect and the resulting interpretation is one where the Adj modifies the noun straightforwardly. That there are multiple readings of the same PF based on covert scopal ambiguities is of course nothing new. What is remarkable about the OC and its readings is that depending on the scope of the Adj, the definiteness of the DP is affected—hence, (3c) is ruled out as a felicitous paraphrase of (2). That is, if the PF DP-‐internal Adj is interpreted as taking scope outside of the DP the DP is crucially interpreted as indefinite. Conversely, on the attributive reading, the DP remains interpreted as a definite DP. 39 To explain the peculiar properties of the OC construction, Zimmerman (2003) following Larson (1999) argues in that in the case of (3a) the DP-‐internal Adj incorporates into D creating a complex quantifier INFREQ 14 (Zimmerman 2003: 256). Zimmerman (2003) concludes on the incorporation analysis over two competing analyses—all three analyses are represented structurally in (4): (4) a. Adji...[DP D [ti NP]] – Adj-‐only QR at LF b. [QP[Q D + Adj] [ti NP]] – Adj incorporation c. [DP Adji [D [ti NP]]] – Adj adjoins to DP (Zimmerman 2003: 250) The analysis in (4b) is empirically motivated given the four conditions that Zimmerman (2003) observes on the formation and interpretation of the OC—the conditions are succinctly represented in (5) – (8) 15 : (5) Adjacency constraint a. The occasional well-‐dressed sailor walked by [Occasionally, a well-‐dressed sailor walked by] b. The well-‐dressed occasional sailor walked by [NOT: Occasionally, a well-‐dressed sailor walked by] (6) No Coordination Constraint a. The occasional and well-‐dressed sailor walked by [NOT: Occasionally, a well-‐dressed sailor walked by] 14 Larson (1999) offers the semantics INFREQ IN (i): INFREQ<e,x> [part-of(e,e*) & sailor’(x)] (strolled by’ (x,e)) ‘For few pairs <e,x> such that e is part of some larger context e*, and x is a sailor, e consists of a strolling by of x’ 15 Constraint tag lines are mine, the examples are from (Zimmerman 2003:252-4) 40 (7) Head Restriction (*Cardinal Q, *Strong Q, *Demonstrative) a. The/An occasional customer entered the shop 16 [Occasionally, a customer entered the shop] b. Two occasional customers entered the shop [NOT: Occasionally, two customers entered the shop] c. Every occasional customer entered the shop [NOT: Occasionally, every customer entered the shop] d. This occasional customer entered the shop [NOT: Occasionally, this customer entered the shop] (8) *Frequency Restriction a. The frequent sailor walker by [NOT: Frequently, the sailor walked by] (Zimmerman 2003: 251-‐3) The structures (4a) and (4c) are ruled out because neither one of them can account in any principled way for the change in definiteness of the DP in the OC context. In the case of LF extraction of the Adj (4a), this structure should be ruled out independently by restrictions on extraction from definite DPs generally (Fiengo and 16 Zimmerman (2003) also notes that, “...the (semantically bleached) 2 nd singular possessive pronoun...” (252) functions as an acceptable D as in: (i) Well, your occasional sailor would also show up [Occasionally, a sailor would show up] 41 Higginbotham 1981) 17 . While it is the case that this particular concern is not an issue for the adjunction structure in (4c), (4c) still offers no principled reason why the reading in (3a) is obtained and not the reading in (3b). Furthermore, neither of the structures in (4a) or (4c) offer any kind of account for the semantic restriction in (8)—*Frequency (see Section 2.3 for discussion). The incorporation structure in (4b), however, accounts for the facts in (5-‐8) as well as offers a principled story for the definiteness discrepancy. As far as constraints (5-‐8), these are accounted for under the Head Movement Constraint in the case of (4b) since the Adj head cannot cross the intervening head in (5) and (6) is accounted for independently by the Coordinate Structure Constraint. In terms of the third restriction (7), Zimmerman (2003) argues: ...determiners with inherent semantic content (be it deictic or quantificational) cannot take part in CQF [complex quantifier formation] because their inherent semantics must not be overwritten. the underlying idea here is that complex syntactic items with particular semantic properties can only be formed if the properties of the complex item do not clash with the intrinsic semantic properties of its parts. According to this line of thinking, CQF is restricted to definite and indefinite articles because these determiners have no inherent semantic content, but only a discursive function (cf. Heim 1982). (258) Adopting the incorporation analysis also makes an interesting prediction about the availability of OC readings in ECs. Since the incorporation analysis assumes that (in the case of the OC reading) the Q that is formed is weak, one should be able to embed the seemingly definite DP—[DPthe occasional NP]—in an EC and 17 That is, if we take [the occasional sailors] to be a definite DP to begin with, then extracting out of it at LF should be illicit for the same reason (i) is bad (Zimmerman 2003:255): (i) *Who did you see [the picture of t]? 42 not have the EC be ruled ungrammatical under the DE. On the flip side, embedding the same DP in an EC should render the attributive reading unavailable/ungrammatical since, in the case of the attributive reading, the DP is assumed to be headed by +definite [the], which as discussed in section 2.1, should induce a DE violation. Indeed, this prediction is empirically validated: (9) There was the occasional sailor walking by (10) a. Occasionally, there was a sailor walking by b. *There was the sailor who sails infrequently walking by Section 1.2: The Occasional Construction and Agreement Variations in the Existential Construction Given the independent properties of the EC and OC constructions, it is now possible to return to the task at hand which is to derive a test that will be sensitive to the agreement patterns in (1). The point of such a test is to determine the underlying structural differences in the derivations that yield (1a) and (1b) respectively with the readings in (11) repeated here from Chapter 1: (11) a. Determiner Reading (1a)—There exist five cops, and they are at the door; the plural agreement gives the reading that what is being focused on is the cops specifically or the number of cops involved in the event of being at the door b. Prepositional Reading (1b)—There exists an event/situation of ‘being at the door’ and the participants in that event/situation happen to be five cops; the singular agreement in (12b) yields the reading that what is at issue in the sentence is the state of affairs such that there are some cops at the door and the fact that there are five of them is just extra information 43 Once a test exists that delineates and illustrates the structural differences of the variants by way of rendering one grammatical and one ungrammatical, one can begin to test hypotheses about the structural differences in the case of (1) where no ungrammaticality obtains in either case. The test therefore will embed identical OC type DPs (e.g. [the occasional NP]) in ECs with raising verbs in both agreement contexts. If one variant is grammatical and the other one is not then that will constitute evidence of structural differences between the two such that in the ungrammatical case, the ability of the complex Q to be formed and/or scope over the DP is blocked and the sentence is rendered ungrammatical as a result of the DE since the OC type DP will be interpreted as an attributive definite DP. Lastly, in order for the test to function properly, it is crucial that the OC type DP is plural. Unfortunately, one aspect of the OC that receives little attention in Zimmermann (2003) is the impact of plurality on the availability of the OC interpretation. It is here that a discussion of the *Frequency constraint (8) from the last section is in order. Zimmerman (2003) concludes that the *Frequency restriction is explained by the fact that the complex-‐quantifier functions as a pluractionality marker and that the semantic features of pluractionality markers preclude the possibility of frequency adjectives in these constructions. Lasersohn (1995) argues that the pluractionality marking on elements that denote events (e.g. VP) is parallel to its nominal counterpart insofar as they both change individual denoting elements in to pluralities: nominal plurals denote pluralities of individuals, pluarctionals denote pluralities of events. Furthermore, events can be pluralized in 44 three ways crucially involving elements that do not overlap in time or space as in (12) formalized in (13): (12) Type of Plural Semantic Function temporal iteration distribution of events over points in time spatially scattered distribution of event⏐s over points in space individually performed actions distribution of events over individuals (13) VPpluractional(X) ⇔ ∀ e,e’ ∈ X [VP(e) & ¬(τ(e) o τ(e’))] & ⏐X⏐> 2 τ = time of occurrence and o = overlap relation ( Lasersohn 1995: 251-253) As Zimmerman points out correctly, the OC reading with a singular NP as in (10) cannot mean that there were three sailors that walked by in a group one time such that there were multiple sailors and few events. Lasersohn’s pluarctional modification offers precisely this restriction on overlapping entities because of the nature of plural events generally. That is, Zimmerman offers a semantics of INFREQ that quantifies over pairs of individuals and events as in the formulation of Larson (1999), but also encodes the non-overlap condition of a pluractional marker. Zimmerman (2003) offers the following lexical entry for INFREQ: (14) a. [an/the occasional] → INFREQ b. INFREQ<e,x> = λQλS. ∃<e,x> [part-of(e, e*) ∧ Q(x)]: (S(e,x) ∧ ∀<e’,x’>, <e”, x”> [S(e”, x”) ∧ Q(x’) ∧ Q(x”)]: ((e’ = e”) ∨ (e’ ≠ e” & ¬(τ(e’) o τ(e”)))) c. “There are some pairs <e,x>, with e part of a contextually given event e* and x having property Q, such that e is an event S involving x, and all pairs <e’, x’>, <e”, x”> of events S involving an x having Q are such that the events do not overlap in time.” (Zimmerman 2003: 272) 45 Indeed, the formulation in (14) accounts for the *Frequency Constraint in (8) because since the semantics of INFREQ do not allow overlapping events as a result of restrictions of pluractionals generally, frequency adjectives should be ruled out for incorporating into D and CQF because they allow for overlapping events and CQF quantifiers must not overwrite the semantics of their constituent parts. Given the strict non-‐overlap conditions inherent in the pluarctional nature of the quantifier INFREQ, a question is raised about the availability of OC readings when the NP is plural. That is, while there is a robust finding that (10) will yield the readings (11a-‐b), the results for the sentence (15) with readings (16) are less stable: (15) The occasional sailors walked by (16) a. Occasionally, some sailors walked by b. The sailors who sail infrequently walked by For some speakers 18 , both (16a-‐b) are available readings; for others 19 , however, (16a) is strongly ruled out and only (16b) is available. However, for those speakers who dislike reading (16a) for sentence (15), when presented with (17) or (18), the OC-‐reading becomes available again: (17) The occasional girl scouts came through the neighborhood (18) The occasional Mormons knocked on my door 20 The explanation for this difference seems to be a pragmatico-‐semantic constraint that essentially allows plural NPs to get OC-‐readings if the activity denoted by the V 18 Barry Schein p.c. 19 James Higginbotham p.c. 20 The idea here being that. in general, both girl scouts and Mormons tend to knock on doors in groups, whereas (for the speakers that label (4a) *) sailors are not assumed to walk around in groups 46 is something that is normally done in groups (Higginbotham p.c.); thus Walker (2007) offers the following constraint: (19) Plural OC Constraint: An OC reading for plural NP is felicitous iff, in the structure [[D [adj [NP]] [VP]], the denotation of V is independently known by the hearer to be done by x participants and x > 1. The constraint in (19) is obviously more of a descriptive generalization than rule that one should expect to find in the syntax. In order to account for the data therefore, we must alter the lexical entry found in (14). As the pluractional lexical entry of the quantifier INFREQ is defined in (14), it specifically quantifies over pairs of individuals and events such that that the events cannot overlap in time and thus not more than event can occur at any given time. It seems, therefore, that (14) predicts that only the attributive reading is available in (15) because the plural NP means that the quantifier would have to bind pluralities of x’s to single e’s which would not satisfy the non-overlap condition. Another way to test this prediction is with embedding OC-type DPs into existential constructions. As noted in the previous section, adopting the incorporation analysis makes an interesting prediction about the availability of OC readings in existential constructions. On one hand, the incorporation analysis assumes a weak Q INFREQ, thus one should be able to embed the seemingly definite DP in an existential construction. On the other hand, embedding the same DP in an existential construction should render the attributive reading infelicitous since the DP is definite which induces the DE: 47 (16) There was the occasional sailor walking by (17) a. Occasionally, there was a sailor walking by b. *There was the sailor who sails infrequently walking by However, (14) incorrectly predicts (18) to be bad for the same reasons as (17b): (18) There were the occasional fishermen floating by. (19) a. Occasionally, there were/was [some fishermen]/[a group of fishermen] floating by b.*There were the fishermen who fish infrequently floating by These facts indicate that the semantics of INFREQ must be altered with a mechanism to transform pluralities of individuals into singular groups of individuals such that INFREQ can quantify not only over <x,e,> pairs but <<x 1 , x 2 , ... x n >,e> pairs as well. Recalling the definition in (14) given again below as (20), what I propose to modify is the nature of the pair of elements that that the quantifier quantifies over. (20) a. [an/the occasional] → INFREQ b. INFREQ<e,x> = λQλS. ∃<e,x> [part-of(e, e*) ∧ Q(x)]: (S(e,x) ∧ ∀<e’,x’>, <e”, x”> [S(e”, x”) ∧ Q(x’) ∧ Q(x”)]: ((e’ = e”) ∨ (e’ ≠ e” & ¬(τ(e’) o τ(e”)))) c. “There are some pairs <e,x>, with e part of a contextually given event e* and x having property Q, such that e is an event S involving x, and all pairs <e’, x’>, <e”, x”> of events S involving an x having Q are such that the events do not overlap in time.” (Zimmerman 2003: 272) 48 That is, instead of quantifying over pairs of individuals and events (<x,e> pairs), it should quantify over pairs of sets and events (<<x 1 , x 2 , ... x n >,e> pairs). I propose that instead of taking [[NP]] as its argument, INFREQ takes at minimum [[Cl(assifier)P]] as its argument and that the denotation of ClP is a plurality of n individuals, where n ≥ 1. The lexical entry for INFREQ includes a function that takes [[ClP]] and creates sets [Σ 1 , Σ 2 ... Σ k ] and that Σ plural = <x 1 , x 2 , ... x n >, Σ singular = <x 1 >. For , the Σ plural cases with <x 1 , x 2 , ... x n >, n = $ where $ is a contextually defined value in e*. In order to unify the representation of singular and plural sets, for all sets Σ, I will adopt the notation ! 1"$ # and in singular cases n = 1, and plural cases n = $ = e* value. Ultimately, under the new semantics INFREQ takes < ! 1"$ # , e> pairs. (21) a. [an/the occasional] → INFREQ b. INFREQ< ! 1"$ # , e> = λ#λQλS. ∃<x 1 , x 2 , ... x n > [#(<x 1 , x 2 , ... x n >) → ! 1"$ # ] ∃<e, ! 1"$ # > [part-of(e, e*) ∧ Q( ! 1"$ # )]: (S(e, ! 1"$ # ) ∧ ∀<e’, ! 1"$ # ’>, <e”, ! 1"$ # ”> [S(e”, ! 1"$ # ”) ∧ Q( ! 1"$ # ’) ∧ Q( ! 1"$ # ”)]: ((e’ = e”) ∨ (e’ ≠ e” & ¬(τ(e’) o τ(e”)))) c. The quantifier takes an argument that denotes n individuals and creates sets of size $ , ! 1"$ # where $ is contextually defined by e*. There are some pairs <e, ! 1"$ # >, with e part of a contextually given event e* and ! 1"$ # having property Q, such that e is an event S involving ! 1"$ # , and all pairs <e’, ! 1"$ # ’>, <e”, ! 1"$ # ”> of events S involving an ! 1"$ # having Q are such that the events do not overlap in time.” 49 Understanding the semantics of INFREQ as in (21) not only is theoretically superior since it accounts for a more nuanced functional structure within the DP, but it is empirically superior because it makes the correct predictions about both the availability of OC readings with plural NPs and about the available readings in existential contexts. Keeping this NumP based version of the semantics of INFREQ in mind, the resulting test structures and the reported grammaticality judgments are as in (22): (22) a. There seem to be the occasional girl scouts knocking on my door b. *There seems to be the occasional girl scouts knocking on my door The data in (22) thus indicates that (1a) and (1b), with readings as in (11) are structurally distinct. That is, no informant interpreted (22b) to mean (23) which is what might be expected given the patterns in (1) and (11): (23) Occasionally, there seems to be a state of affairs such that there are some girl scouts knocking on my door Ultimately, what this data indicates is that in the case of (22a) the formation of a weak complex Q occurs and the verbal agreement must be plural. If, as Zimmerman (2003) argues, the QINFREQ quantifies over some event variable, then it seems to be the case that in the ECs the QINFREQ quantifies over the predicate as a whole. Indeed, in the OC readings there are assumed to be multiple events of [some sailors walking 50 by] 21 and that these events are few and temporally distant from one another within a given time period yielding plural agreement. Furthermore, the flipside of this intuition also seems to be true—in the attributive reading, there is only one event of [the sailor who sails infrequently walking by]. The upshot of all this is that the matrix verb agreement seems to be governed not by the phi features on the DP alone but by the quantificational structure of the predicate as a whole. 21 For ease of exposition I opt here to use the examples about sailors (setting aside the plural OC constraint for the moment) since the attributive reading of the girl scouts is semantically awkward (e.g. [the girls scouts who are girl scouts infrequently]). But even if this semantic oddity exists, it would still occur in the context of definite attributive DP and therefore be ruled out on syntactic grounds and is therefore not a problem for the test as a whole. 51 Section 2: The Syntax of the Existential Construction Section 2.1: A Structural Analysis of the Occasional Construction Data—An Initial Sketch of the Topography of Existential Constructions Recalling the classic intuition in (6) and (7) from Chapter 1 that ECs and the copular counterparts are in some kind of close derivational/transformational/isomorphic relationship, it seems that a reasonable place to start a discussion of EC structure is with the copular structure of Roy (2005) and Baker (2003) as in (33b) from Chapter 1, repeated here for convenience as (24): (24) a. TP Chrisi T’ T VP ti hunger b. TP Chrisi T’ T PredP ti Pred’ Pred AP hungry 52 In (24b), the predicate is distributed over a functional projection so that the element merged into [Spec, PredP] is the subject of the type lifted lexical category. From a structural standpoint, if V is the only lexical category with a specifier, for the other categories (A and N) to become predicates they require the help of functional structure to facilitate the necessary subject position. In terms of the EC, it has been posited by many (Williams 1984, Hazout 2004, Stowell 1979 a.o.) that the EXPL and the postcopular (D/)NP are in a Subject-‐Predicate relationship. At first approximation this seems to suggest, following Roy (2005)’s copular structures, that the copular and EC variants of a given sentence might look like: (25) a. Copular TP DPi T’ T PredP ti Pred’ Pred XP b. EC option 1: EXPL as Subject of Predication TP EXPLi T’ T PredP ti Pred’ Pred XP 53 c. EC option 2: Internally Derived Predication TP therePROi T’ Tns VP ti V’ to be PredP [XP, or YP from [Spec XP] Pr’ Pr XP Taking (25a) to be the copular construction, it seems plausible that there are (at least) two option for deriving the EC. In (25b), the EXPL is simply inserted in [Spec, PredP] and the derivation proceeds as in the copular construction illustrated above. This is perhaps the simplest derivation and as such should serve as the null hypothesis. Entertaining the idea that (25b) is the appropriate structure, one is immediately confronted with a seeming semantic oddity. In the case of the copular construction in (25a), one XP is predicated of another YP (where it is possible for X = Y in category) such that an equivalence or categorical/taxonomical statement is created. In the structure in (25b), however, since EXPL is semantically vacuous, it is unclear what it means to predicate some XP of an EXPL. Indeed, this is part of the rationale behind LF movement schemas (cf. Chomsky 1993; 1995) that seek to unite the post-‐copular NP and the EXPL (either by way of affixation or replacement). In terms of the OC data reported in Section 1.3, adopting the structure in (25b) forces an analysis that is as in (26): 54 (26) a. There seem/*s to be the occasional sailors walking by TP EXPLi T’ seem/*s VP ti V’ to be PredP ti Pred’ Pred DP [D+Ak] NP tk N’ NP sailors VP walking by b. * There seems to be the occasional sailors walking by TP EXPLi T’ *seems VP ti V’ to be PredP ti Pred’ Pred DP the NP occasional N’ N VP sailors walking... 55 Adopting the structure in (25b) where There is the subject of PredP fails to block the adjective incorporating into D which makes the false prediction that the attributive reading is universally unavailable. Certainly in (26b) it is predicted that there will be a DE as opposed to (26a), in the latter a complex weak Q has been formed and therefore licensed in the scope of the existential quantification of the EXPL. In (26b), however, the A has not incorporated into D and therefore the attributive definite reading obtains and renders the sentence ungrammatical. What is not clear under this analysis is what the agreement patterns have to do with anything. That is, this structure gives us no reason to believe that the incorporation could not happen with [seems] as the verb and thus rectify the DE. Thus, these reasons taken together militate against the null hypothesis and should compel us to examine structure (25c) as a better option which offers a more principled and systematic explanation of the data. Assessing the viability of (25c), the first aspect to notice is the architecture of the PredP. In the copular construction in (25a) some XP is predicated of some other YP (where it is possible that X and Y are of the same category). In (25c), however, an XP is predicated of itself by way of movement. I will contend that this is a superior interpretation for three reasons. First, it immediately resolves the problem of EXPL being a semantically vacuous subject such that predicating something of it is meaningless until LF when it is either replaced or something is adjoined to it. Secondly, it gives a simple distinction between copular and existential sentences: copular sentences are ones that predicate something of something else, and 56 existential sentences are sentences that predicate something of itself. The third reason is that the difference that this distinction between copulars and existentials affords us offers us a pre-‐theoretical explanation of the DE. Simply put, if existential sentences seek to reaffirm or state the existence of something, that ‘something’ should be indefinite since if the ‘something’ were definite its existence would not be in question. This insight actually foreshadows our final analysis insofar is it prefaces a notion of theticity and the notion that existentials describe situations. From a syntactic standpoint, this system makes the prediction that agreement holds between the T 0 and the Pred 0 with Pred 0 inheriting φ-‐features and (in)definiteness from its specifier in a traditional Spec-‐Head agreement. It is also assumed that PredP must have a Spec that is filled—this assumption simply reduces to the idea that predicates must have subjects. Turning now to how the structure in (25c) would capture the facts from Section 1.3, consider the following structures: 57 (27) a. There seem to be the occasional girl scouts knocking on my door TP EXPLi T’ seem VP ti V’ to be PredP DPj Pred’ [D+Ak] NP Pred 0 tj tk N’ girl scouts VP... b. *There seems to be the occasional girl scouts knocking on my door TP EXPLi T’ *seems VP ti V’ to be PredP Pred’ VPj Pred 0 tj DP V’ the NP knocking PP occasional girl scouts on my door 58 With these structures the agreement patterns from the data in Section 1.3 can be explained in a principled way. In the grammatical case (27a), the XP that moves to [spec, PredP] is a DP that contains a SC. Stowell (1978) argues that SCs denote events or situations 22 (I take N’ to represent a SC here), which in the case of the OC reading is exactly what QINFRQ must quantify over. In this case, plural agreement on Tns is expected because the element in [Spec PredP] which is valuing the Features on Pred 0 is a plural DP. The creation of the weak QINFREQ ensure that the sentence will not be ruled under the DE. Again, recall that SCs denote events, when the EXPL raises to [Spec TP] out of the larger SC (the one whose daughter are EXPL and PredP) it is in precisely the right position to exert existential quantification over the SC and thus create the pre-‐conditions for deriving a DE in the first place. Turning our attention now to the ungrammatical structure in (27b), The difference between the two structures is the XP that moves to [Spec PredP]—in (27a) it is a DP and in (27b) it is a VP. Just looking at the architecture of the VP in (27b) it is clear that the subject DP does not contain an event in the same way that the DP in (27a) contains a SC. That is, there is no event variable that the complex Q 22 Stowell (1978) argues this based on the contrast between the following two paradigms where in (i) the sentences are ungrammatical because the sentences attempt to equate individuals to event or situations. The sentences in (ii) are grammatical because the subject is an event as well as the SC predicate resulting in a felicitous reading: (i) a. *Mabel was a pig roasted b. * John is an old man sleeping c. * Bill is a friend of mine sick (ii) a. What is saw was a pig roasted b. The classical situation is an old man sleeping c. The episode which I remember was John Wayne sick in bed (Stowell 1978: ex 40-1) 59 could quantify over since even if it were formed through incorporation and we assume that an event variable is introduced by the V, there is no way that QINFREQ could ever c-‐command the event variable. That is, while the structure in (25b) gives us no reason why the derivation could not be salvaged from the DE by incorporation as in (26), if we adopt the same strategy to rectify the DE violation in (27b) we are still left with a Q that has no variable to bind as in (28): (28) *There seems to be the occasional girl scouts knocking on my door (with CQF) TP EXPLi T’ *seems VP ti V’ to be PredP Pred’ VPj Pred 0 tj DP V’ [D+Ak] NP knocking PP tk girl scouts on my door Moreover, since the Pred 0 is valued in terms of both φ-‐features and Definiteness, in the case where VP is [spec PredP] the DE is predicted. It is the claim of this paper that if the subject of a PredP is +Def then the Pred is +Def. In the case of having a VP as [spec PredP] the Definiteness of Pred is derivative in the sense that if the [spec VP] is definite, then the VP itself is Definite and denotes a definite event. 60 Section 2.2: Accounting for EC Agreement Variations Generally With a better understanding of the structural basis of the variable agreement patterns and their syntactic implications, one can now turn to explaining the fact that both the agreeing and seemingly non-‐agreeing sentences in (29) are grammatical and that they receive the different readings in (30): (29) a. There seem to be five cops at the door b. There seems to be five cops at the door (30) a. Determiner Reading (29a)—the plural agreement gives the reading that what is being focused on is the cops specifically or the number of cops involved in the event of being at the door b. Prepositional Reading (29b)—the singular agreement in (29b) yields the reading that what is at issue in the sentence is the state of affairs such that there are some cops at the door and the fact that there are five of them is just extra information Taking the structure in (25c) to be the abstract structure of ECs with the properties and restrictions discussed in Section 2.2, the structures in (29) should be as in (31): 61 (31) a. There seem to be five cops at the door TP EXPLi T’ seem VP ti V’ to be PredP NumPj Pred’ Five ClP Pred 0 tj -‐s NP cop PP at the door b. There seems to be five cops at the door TP EXPLi T’ *seems VP ti V’ to be PredP Pred’ PPj Pred 0 tj NumP P’ Five ClP at DP -‐s cop the door 62 The analysis for these two structures is essentially the same as in the OC data, however, in this case the nature of the elements that are the moved XPs (a NumP in (31a) and a PP in (31b)) both result in grammatical readings. Section 2.3: Mass/Count Distinction: Interpretations and Structural Ramifications Before specifically analyzing each derivation, it is necessary to discuss the predicational and mass/count properties of the complements of Pred, namely NumP and PP, in (31a-‐b). Such a discussion will not only guide the technical explication of the derivations, but it will also facilitate translating the non-‐technical Prepositional and Determiner readings in (30) into technical definitions which not only more aptly explain the interpretational difference in question, but suggests a principled reason for the agreement variation as well. Ultimately, if the interpretational differences and categorical restrictions in copular sentences are the result of the mass/count characteristics of the predicates, then I predict that those same mass/count properties can account for the interpretational and agreement differences in the case of existentials. Taking first the complement of Pred in (31a), one should recall the discussion of mass/count predicates in Chapter 1. What Roy called situation- descriptive sentences were the result of mass predicates, in the sense that they were intense and divisive, and that they were mass by default because the complement of Pred (e.g. A, PP) did not license the functional structure responsible for count. Roy (2005) argues that (Cl)assifierP is the locus of the properties responsible individuation of a mass, and that only category N licenses ClP which means only Ns 63 can be interpreted as count. It is for this reason that only Ns are capable of being characterizing predicates. Recall that characterizing predicates are count in the sense that they are non-‐intense and non-‐divisive. That is, characterizing predicates can hold of discontinuous events. This is illustrated in (32) repeated here: (32) Qu’est ce qui s’est passe dehors, c’est quoi tout bruit? What’s going on outside, what is all this noise about? a. Paul est ivre, il a renverse la poubelle comme toujours Adj mass: sit.-desc. Paul is drunk, he knocked the trash can over as always b. # Paul est invrogne, il a renverse la poubelle comme toujours N count: charct. Paul is a drunkard, he knocked the trash can over as always In (b) above, the complement of Pred is argued to be a ClP with a null classifier heading it. In fact, Roy concludes that both NumPs and ClPs can be copular predicates and it is precisely that variation that is responsible for deriving the difference between characterizing sentences and defining sentences. While the details of the analysis are not relevant to the discussion here, it is crucial to notice that DP is not a predicative category in this system since it is the presence/absence of D that determines argumenthood (i.e. must satisfy Case Filter/theta-‐marking etc...) 23 . 23 Roy argues that English a starts as a classifier and either stays there (characterizing sentences) or moves to NumP (defining sentences) 64 In the case of the PP complement of Pred in (31b), Roy notes that the N+PP structure is unique in the sense that it has eventive properties, “the meaning of the post-‐copular expression N+PP differs slightly from the meaning of N without the PP: it is more dynamic and gives rise to the impression that the subject is taking active in [an event]” (Roy 2005: 38). Moreover, in terms of the mass/count properties of the PP it is mass as described above. In terms of the readings in (30) then, we can more precisely define these readings as: (33) a. Determiner Reading → Count reading b. Prepositional Reading → Mass reading Section 2.4: Analyzing the Derivations Starting with the case of (31a), a NumP merges in [Spec PredP] valuing the head Pred 0 with plural phi-‐features and a numeral. As the derivation progresses, the EXPL raises to check EPP on Tns but cannot satisfy the unvalued phi-‐features of Tns. The phi-‐features of Tns are valued in a subsequent AGREE relation that holds between Pred and Tns resulting in plural agreement. The reason that this structure obtains the Mass Reading is that the Pred-‐Tns agreement relation is informed/valued by the specifier of PredP such that the EC is really a NumP moved from the specifier of the lower PP, thus yielding the interpretation that what a speaker is referring to is the actual nominal elements primarily and not their spatio-‐ temporal orientation. 65 Turning now to the case of singular agreement in (31b), the XP constituent that is moved to [spec PredP] is a PP. Crucially, as discussed earlier 24 , while N, V, A are lexical categories in the system being adopted here, P is not. In terms of Baker’s system, then, P is a functional category and can have a specifier. Moreover in terms of prepositional phrases introducing predicative contexts and event variables Higginbotham (2000) offers this example: (34) John was in the kitchen, but it didn’t last long (Higginbotham (2000): (28) p. 56) In (34), it is understood that the event/state of being in the kitchen is what is binding the anaphor. In fact, Higginbotham concludes that despite the fact that, “the English PP resist[s] nominalization (*the being in London of John)...arguments from anaphora show that there must be states derivable from ordinary PP predicates” (2000: 55). The LF is given in (35): (35) (the e) in the kitchen(John, e) (Higginbotham 2000: 56) Ultimately, (34-‐5) constitute just one illustration of the notion that all PRED heads have an event-‐position (E-‐position). It will be the claim of this paper that PPs, by way of their head, denote a mass event. Thus when a PP occupies a [Spec PredP] the ensuing agreement between Tns and Pred will be singular with the resulting interpretation that the speaker is talking about a state of affairs such that the are some people at the door and they happen to be five cops. 24 Chaper 1, Section 3.2 66 The system proposed here thus captures not only the interpretational differences encoded in the different mass/count structures in (31), it also predicts the agreement patterns that should obtain for these different interpretations based purely on the known structural differences between mass/count structures generally. That is, the system predicts mass predicates (PPs) will trigger singular agreement with the copula by way of valuing the Pred as mass/singular, and the count predicates will trigger plural agreement—clearly, a desirable outcome. Also, the system proposed here maintains and exploits the derivational relationship between copular and existential sentences while not only preserving the interpretational differences and categorical restrictions in copular sentences but also accounting for the well known restrictions on existential sentences (which are ultimately derived from similar constraints on copular sentences). Section 2.5: Floating Quantifiers, Agreement and the EC It is at this point we can return to a fact that was left as simply an observation Chapter 1; namely, that the non-‐agreeing pattern in ECs is only licensed in cases that involve secondary predication as in (36): (36) a. There seem to be five cops at the door b. There seems to be five cops at the door c. There seem to be five cops d. *There seems to be five cops These facts are straightforwardly explained in the system here because in the infelicitous case (36d) what happens is that a plural DP moves to [Spec PredP] as a maximal projection which values PredP plural and creates a phi-‐feature mismatch. 67 In the felicitous case (36b), however, the plural DP moves as a specifier of the larger maximal projection PP of which values the head Pred 0 as singular. While the structure of the PP in (31b) may seem unorthodox there is good empirical evidence to justify it from the phenomenon of Quantifier Floating. Taking the old observation of Mailing (1976) about post nominal Quantifiers only being licensed in cases where there is secondary predication, one can derive a straightforward structural test for the PP being proposed here. Mailing observed the following: (37) a. we considered the men all fools b. *we considered the men all The structure posited was something along the lines of: 68 (38) TP wei T’ T vP ti v’ v VP DPj V’ the NP consider PredP men DP Pred’ All tj Pred AP fools The argument was that in these cases the A could not function as a predicate on its own and therefore needed some functional structure (here I call it PredP for consistency) and that the DP [the men] was in the specifier of this functional projection with the Q [all] adjoined to it. When the DP moved to get case in the ECM context it stranded the Q in the specifier of the functional projection yielding the surface order in (37a) and ruling out (37b). Applying this logic to the structure in (31b), one should make the prediction that a floating Q is grammatical. The paradigm in (39) is robust and makes exactly the predictions that one would expect: (39) a. There seem to be five cops all at the door b. *There seems to be five cops all at the door 69 In (39a), the exact same movement occurs as in (38) albeit from [spec PP] to [spec PredP] instead of [spec PredP] to [Spec VP]. Since the NumP that moves is plural and thus values Pred 0 as such, agreement on the Tns is expected to be plural. In the case of (39b), however, the agreement is expected to be ungrammatical since it is the same NumP that is valuing Pred 0 as in (39a). Thus, in the case of a PP, the option seems to be either move the NumP to [Spec PredP] (see (40) for structures of (39a-‐ b)) or move the whole PP to [Spec PredP] (see (31b)): (40) a. There seem/*s to be five cops all at the door TP EXPLi T’ seem/*s VP ti …. PredP NumPj Pred’ Five NP Pred 0 PP cops QP P’ All tj @ DP the door Since NumP is a predicative category and does not bear D it does not need case as in the Mailing’s ECM example so one might wonder what motivates the movement of NumP to Spec, PredP. Even though the NumP is not moving case in (40), something 70 must fill the spec, PredP as noted earlier—recall that this is simply the notion that predicates must have subjects. Section 3: The Existential Construction, Locatives, and Predication This section will seek to summarize the relevant parts of the proposal in Den Dikken (2006) (hereafter DD06) concerning the formal structure of predicates and compare it the structures proposed in earlier in this chapter. Specifically, it will address the DD06 treatment of copulars and expand a bit on the earlier discussion of Roy (2005). It is in the course of this discussion that I will address the predicative role and function of DP—essentially, DD06 will expand the notion of predicate such that DPs are predicative categories and I will abandon the constraint cited from Roy (2005) that DPs are restricted to argument positions. Finally, I will summarize DD06’s discussion of locative inversion constructions and I will offer some evidence that these structures make some false predictions vis-‐à-‐vis locative inversion in existential contexts. This section will also introduce some new evidence about existentials and agreement in the context of locative inversion and attempt to extend the DD06 proposal to account for the EC data in a way consistent with the structures advocated in the current work. Section 3.1: RELATOR Phrase (RP) vs. PredP DD06 begins the discussion of the syntax of predication with two claims: (a) the relationship of predication is facilitated by functional head in such a way that it creates a standard binary branching phrase marker, and (b) that the order of the constituents is variable. These claims are structurally represented in (41) and (42): 71 (41) Structure of Predication RP XP R’ R YP (DD06: 11) (42) Order of Predication a. [RP [XP subject] [R’ RELATOR [YP predicate]]] b. [RP [XP predicate] [R’ RELATOR [YP subject]]] c. predication is nondirectional (emphasis in original) (DD06:12-‐14) The structure in (41) is essentially the same structure as the PredP proposed in Roy (2005), Baker (2003, 2006) and Bowers (1993), as well as the one advocated in the present work. DD06 is different from the Bowers (1993) and Roy (2005) proposal in terms of the lexical status of the head Pred 0 vs. RELATOR. Bowers and Roy both assume Pred 0 to be a functional head, and it is not be identified as an other already existing category (e.g. v ≠ Pred 0 ). The DD06 proposal is much more akin to the Baker (2003, 2006) proposals which argue, in some form or another, that Pred 0 (for Baker) and RELATOR (for DD06) can exist both as an abstract functional head or be instantiated by various closed class morphemes or lexical items (e.g. copulas, adpositions, particles LINKERS, etc...). Indeed, just as Baker (2003) argues that lexical categories do not have specifiers (except verbs, but those are really covert PredPs and Adjs), DD06 says: ...a head H is a θ-‐assigner or a mediator of predication relationships but never both at the same time—an approach, in other words, that strictly divides the labor of θ-‐assignment and mediation of predication relationships. From the perspective of this work 72 [DD06], no lexical projection ever has a specifier, while every RELATOR must have a specifier. 25 (DD06: 22) In terms of the proposal of the current work, I will remain agnostic about the lexical status of Pred/ RELATOR in general. However, in the context of the English there-‐existential construction I contend that Pred 0 is not phonologized. I will also say that given the empirical evidence offered in Baker (2003, 2006) about Kinande morphology and in DD06 as well, it seems plausible that Pred/ RELATOR can, under the appropriate circumstances be phonologized by an existing closed class morpheme/lexical item. Section 3.2: Extending the Situation-‐Descriptive, Characterizing, and Defining Predicational system: DPs as predicates This section will now turn comparing the Small Clause (SC) structures proposed in my work (specifically, the proposals made in answer to Question 1) to the theory of Small Clauses adopted in DD06. First, I will summarize the proposal in DD06. I will then compare the proposal in DD06 regarding the structure and typology of copular sentences to that of Roy (2005). In this discussion, I will illustrate that the insights in DD06 can be used to extend the scope of the Roy (2005) proposal to include Equative copular sentences (which she excludes as non-‐predicational). Lastly, in the course of the Roy (2005) discussion I will specifically address the NumP vs. DP discrepancy in the structures in Chapter 1 (e.g. I will answer Question 4). DD06 defines a small clause as an RP lacking tense (DD06: 60). This has a number of implications. First, it means that small clauses are elements involved in a subject-‐predicate 25 It is important to note, however, that the DD06 system allows head movement, which is ultimately the way that RELATORs are phonologized. Thus while a phonologized RELATOR may look as though it is doing double-duty (e.g. assigning a theta-role and mediating a predication relation), in reality those functions are being jointly carried out by a head adjunction structure. 73 relationship that is tenseless. Second, it commits us to the idea that all SCs are properly headed phrase markers that obey the projection principle. Finally, it opens the possibility for Small Clauses to be recursive (e.g. there is no a priori reason RPs cannot embedded within other RPs). To illustrate these points, DD06 (pp. 62-‐3) starts with the Moro (2000) idea that be takes a bare SC as its complement such that the following sentences are ungrammatical because both as/for are lexicalizations of the functional heads of RPs (=SCs): (43) a. Brian is (*as) a singer b. Jane is (*for) a woman While the observation that these are ungrammatical is certainly correct, the proposal in Moro (2000) is seemingly too strong since, given the appropriate amount of predicational relationships, we can say sentences such as: (44) a. This butterfly is big for a butterfly b. Brian is popular as a singer Indeed, the sentences in (44) illustrate all the properties of SCs: subject-‐predicate relationships in both directions, proper headedness, and recursion of RPs: (45) a. [RP2 [DP this butterfly] [R’ RELATOR2=is [RP1 [AP big] [R’ RELATOR1=for [DP a butterfly]]]]] b. MATRIX: [RP2 [subject] [R’ RELATOR2=is [RP1 [predicate]] ⇒ {Subject, Predicate} c. EMBEDDED: [RP1 [predicate] [R’ RELATOR1=for [DP subject]] ⇒ {Predicate, Subject} The very nature of SCs, then, in DD06 ensures that there is only 1 RELATOR per predicative relationship (hence explaining the ungrammaticality of (43)) as well as offer a compositional 74 semantics to compute the denotation of the sentences in (44). One should also note that RP2 does include a tensed element is, this is not problematic however since the definition of SC is an RP without Tense. The definition has nothing to say about Tensed RPs; in fact, DD06 argues that copular sentences are tensed RPs, and by extension is committed to a system where existential constructions are also RPs involving Tense although the nature and structure of those will be discussed later. So far, in terms of the theory of small clauses and the nature/structure of copular sentences not much different has been said than the discussion in Roy (2005). The three-‐way distinction between Defining, Characterizing, and Situation-Descriptive sentences of Roy (2005) that was designed to account for the interpretational differences of copular sentences adopts a single PredP structure that is virtually identical to the one adopted in DD06. The differences in interpretation of these sentences is not a difference in structure, but rather a difference in the content of the terminal nodes of one basic predicational structure however they choose to label it. However, Roy (2005) argues that DPs are not predicational categories, thus, sentences such as (46) are not predicational sentences and do not figure into the copular typology advocated by Roy (2005). Instead, she (as well as others e.g. Bowers (1991), Stowell (1978)) argue that the sentences in (46) are the result of BE2 that is transitive-‐esque in the sense that it assigns two θ-‐roles: (46) a. Cicero is Tully b. The Morning Star is the Evening Star c. Your attitude toward Jones is my attitude toward Davies Indeed, there are a number of facts that indicate the sentences in (46) are neither predicational nor are they transitive or θ-‐governed by the same matrix verb, as the sentences in (47) and (48): 75 (47) Your attitude toward Jones, which you got from your parents, is my attitude toward Davies, which I got from mine (48) a. *Whose attitude toward Davies would you say your attitude toward Jones is t? b. *Whose attitude toward Jones would you say t is my attitude toward Davies? (DD06: 71) The sentence in (47) illustrates that both DPs can take nonrestrictive relative clauses which means they are both referential and hence not predicates. The sentences in (48) illustrate that be is not transitive, since if it were, one of the two DPs should be able to be wh-‐extracted. However, while these facts seem convincing that these sentences are derived from different underlying structures than the copulars discussed in Roy (2005), DD06 points out that they share the same restrictions as inverse copular sentences (particularly inverse specificational sentences, characterizing sentences to Roy (2005)): (49) a. Brian’s arrest was the biggest upset b. The biggest upset was Brian’s arrest (50) a. Whose arrest do you think t was the biggest upset? b. How big an upset do you think t was Brian’s arrest? (51) a. *Whose arrest do you think the biggest upset was t? b. *How big an upset do you think t was Brian’s arrest? The sentences in (50) represent the relevant wh-‐extractions from the sentence (49a) which is in canonical Subject-‐Predicate order (e.g. it is specificational) while the sentences in (51) represent the same extractions from the sentence in (49b) 76 which is in the state of predicate inversion. Indeed, the failure of the sentences in (51) is the same as the failure in the equative sentences. DD06, therefore, offers an analysis of these sentences that capitalizes on these extraction phenomena and also unifies the structure of these equative sentences with the single structure proposed in Roy (2005) and elsewhere for predicative copular sentences based on a free-‐ relative analysis bound by a PRO-‐predicate: (52) Cicero is Tully TP Predj T’ PRO-‐PREDICATE CP T+ RELATORK RP Opi C’ be DP R’ C∅ RP Tully tk tj DP R’ Cicero R∅ ti (DD06: 73) The idea here is that the Predj constituent is, in fact, a reduced free-‐relative— something like what Cicero is. In the derivation, the beginning RP starts with Tully as the subject, and the Reduced Free Relative as the predicate. The first RELATOR moves to T to be phonologized as be. The next step is the process of PREDICATE-‐ INVERSION, which can be summed up as: 77 (53) Predicate Inversion involves A-‐movement to a subject position triggered by the need to license an empty predicate head (DD06: 98) In this case, the empty predicate head is found in the reduced free relative. I will defer discussing the technical and interpretational details of predicate inversion for the next section. For the present discussion, the crucial aspect of this structure is that it unifies equative copular sentences as predicational constructions with all the other types of copular sentences. Moreover, it makes compelling claims about the possibility of DPs functioning in predicational structures. Also, given the general recursion possibilities afforded by the structure in DD06 we can see that RPs may well occur within DPs in such a way as to make them predicative (e.g. Bingo is the cutest dog of all). Thus, if we abandon the notion that DPs cannot appear in predicational structures, then the motivation to allow only nominal projections extended as far as NumP is no longer valid. In this new light, the NumPs in Chapter 1 and (31) from Section 2 above can be understood as DPs. However, it is important to note that the requirement of Cl(assifier)P and/or NumP be present to facilitate plural agreement and the corresponding count reading for existentials is not lost. Section 3.3: Predicate Inversion, Locative Inversions, and Existentials In the last section, I briefly introduced (52), a fairly technical structure for the equative copular Cicero is Tully where Cicero is in fact a reduced free relative that begins in the complement position of the lowest RP and then by way of predicate inversion somehow crosses the subject of the RP Tully to land in Spec TP. The motivation for this movement (which presumably should be ruled a priori on 78 relativized minimality grounds (Rizzi 1991)) was the stipulation in (53). Clearly, there are a number of issues yet to be discussed with respect to this model. This section will therefore begin with a discussion of what DD06 terms the ‘how and why’ of predicate inversion. First I will finish reviewing the equative copular sentence analysis presented in the section above. I will then discuss this system with respect to Locative Inversions. This is where the review of DD06 will end, and I will begin to analyze the DD06 system in terms of there-‐existentials. In this last section, I will propose a potential interpretation of there in these type of ECs, as well as offer some interesting new evidence from agreement contrasts in existentials with locative inversions. The first question we might ask of the structure in (52) is not ‘how did this exotic movement happen?’ but ‘why would we want to posit this move in the first place?’ Indeed, DD06 links the operation of predicate inversion to the scope of focus in the sentence: What all predicate inversion constructions have in common is the fact that the postcopular/postverbal constituent is a FOCUS—a presentational focus in locative inversion constructions...and in copular inversion constructions in which the preposed predicate nominal is indefinite, and an identificational/exhaustive focus in copular inversion sentences with definite predicate nominals...(DD06: 82) Essentially, Predicate inversion affords the grammar a syntactic means of assigning narrow focus to post-‐verbal subjects. A similar observation is made by Kuroda (1972) about the postcopular subjects in ECs in the sense that they are interpreted as detopicalized subjects. But even if predicate inversion is a syntactic mechanism 79 for mediating sentential focus, it still does not really provide a justification for the structure in (52). Given the rule that predication is non-‐directional (42c), it seems much more reasonable to assume a base RP of the form in (54b), rather than the one proposed which in (52) which is (54a): (54) a. [RP [XP subject] [R’ RELATOR [YP predicate]]] b. [RP [XP predicate] [R’ RELATOR [YP subject]]] If we hypothesize the base RP as (54b) there is no movement violation because the complement of RP does not cross the subject of RP to land in the subject position of TP. However, despite the attractiveness of this alternative, Predicate Inversion structures and the sentences with a base RP of the type (54b) such as (45) repeated here as (55) have different properties: (55) a. [RP2 [DP this butterfly] [R’ RELATOR2=is [RP1 [AP big] [R’ RELATOR1=for [DP a butterfly]]]]] b. MATRIX: [RP2 [subject] [R’ RELATOR2=is [RP1 [predicate]] ⇒ {Subject, Predicate} c. EMBEDDED: [RP1 [predicate] [R’ RELATOR1=for [DP subject]] ⇒ {Predicate, Subject} (56) A’-‐Extraction from (55) ok (sort of...) a. ?I cant figure out what Im looking at—a butterfly, this creature is too small for t; but a dragonfly, I think it would be too big for t (57) Predicate Inversions/Locative Inversion a. I think the best candidate is this man b. I said that on this wall hung a picture of John 80 (58) A’-‐Extractions from (57) bad a. *Which man do you think the best candidate is t? b. *Whose picture did you say on this wall hung t? (59) Consider Obligatory Copula Distinction a. I consider Imogen’s article about Brian (to be) my key example of good journalism b. I consider my key example of good journalism *(to be) Imogen’s article about Brian (DD06: 84-‐93) That is, sentences of the type in (55) seem to allow for A’-‐extraction of the postcopular subject with a violation on par with something like mild subjacency effects. In (58), the A’-‐extractions of postcopular/postverbal subjects from the Predicate/Locative Inversion constructions in (57) are totally ruled out, on par with Subject-‐Island violations of the familiar type. Also, in (59), the predicate inversion order of the embedded clause (e.g. my key example of good journalism *(to be) Imogen’s article about Brian) can only be embedded under consider with an overt copula, whereas in the subject-‐predicate version of the embedded clause, the copula is optional. Given these extraction facts, it seems reasonable to assume that the sentences in (57) involve a unique type of syntactic operation. The task now is to determine the formal nature of the procedure of predicate inversion. The answer to the puzzle, is essentially that all inverse copular sentences have the same structure as the equative copular sentences in the sense that the engine for predicate inversion is the need for the empty head of the complement of the base RP (a 81 predicate in this case) to be licensed by T. DD06 succinctly summarizes that, “the need for inversions indicate[s] that a reduced free relative is not licensed to stay in the predicate position of a small clause; it must raise to spec TP in order to be properly licensed” (DD06:93). Thus, the same structure is employed for a regular predicate inversion copular as for the equative copulars in (52): (60) The best candidate is Brian TP Predj T’ PRO-‐PREDICATE CP T+ RELATORK RP Opi C’ be DP R’ C∅ RP Brian tk tj ti R’ R∅ DP the best candidate (DD06: 93) It is from the empirical facts concerning these sentences and the logical arguments in support of these structures that DD06 argues for the overarching mechanism of predicate inversion summed up in (61): (61) Predicate Inversion involves A-‐movement to a subject position triggered by the need to license an empty predicate head (DD06: 98) However, if (61) is true, it seems strange to apply to the notion of predicate inversion to locative inversions, because it is not prima facie clear that there is any empty head to be licensed. In (57b) the apparent predicate that is fronted is a full 82 PP. However, upon closer inspection, DD06 argues that the fronted PP predicate is in fact not just a full XP moved from the complement position of RP to spec TP as in (60). Rather, the overt PP is in fact base-‐generated in an A’ Topic position (Spec TopP) and co-‐indexed with a pro-‐predicate in Spec TP (which was moved there from Spec RP by way of predicate inversion). Evidence that locative inversions are topics comes in three forms: ECM differences, Aux inversion possibilities, and extraction patterns: (62) ECM a. Imogen expects the best candidate to be Brian b. The best candidate Imogen expects to be Brian c. *Imogen expects on this wall to be hung a portrait of Brian d. On this wall Imogen expects to be hung a portrait of Brian (63) Aux Inversion a. The best candidate seems to be Brian b. Does the best candidate seem to be Brian c. On this wall seems to have hung a picture of Brian d. *Does on this wall seem to have hung a picture of Brian? (64) Extraction a. The best candidate was claimed by Imogen at the party to Brian b. This is the girl by whom the best candidate was claimed at the party to be Brian c. On this wall was claimed by Imogen at the party to have hung a portrait of Brian d. *This is the girl by whom on this wall was claimed at the party to have hung a portrait of Brian 83 In (62), the inverse copular version (62a) has no problem being embedded in the ECM domain in (62c), the locative inversion of (62c) cannot be embedded directly under the ECM verb without raising the inverted to locative to sentential Topic position. The data in (63) show clearly that in the inverse copular construction, aux inversion is acceptable, while in the locative inversion case it is not. The extraction evidence in (64) once again reaffirms the dichotomy between locative inversions and copular predicate inversion sentences. The upshot of all this is that, while the operation ‘predicate inversion’ is assumed to be at work in both of these constructions, the formal instantiations of that operation are quite different and unique to the other formal properties of these constructions. DD06 succinctly summarizes the process and restrictions of locative inversion constructions: The locative PP in locative inversion constructions is base-‐ generated in a topic position strictly local to the pro-predicate [emphasis in the original] in Spec TP. That is, the base position of the topic-‐PP must be the SpecTopP immediately outside the TP whose specifier serves as the landing site of the A-‐movement operation affecting the locative pro-‐predicate. It is impossible for the topic-‐PP to be base generated in the topic position in a higher clause or in Spec CP; nor can the topic-‐PP raise from its base position local to the pro-‐predicate to a higher position...movement from SpecTopP is impossible (DD06: 102) 26 I offer the following tree to model the DD06 locative inversion structure: 26 This does leave an open question about the status of (62d): (i) On this wall Imogen expects to be hung a portrait of Brian Where the solution to the problem is to move the locative PP to a matrix higher specifier TopP position. Presumably this is allowed because the Spec TP pro-predicate also raises with it in the ECM case. 84 (65) In the room hung a picture of Gandhi TopP PPi TP in the room pro-predicatei T’ pst VP hung RP DP R’ a pic of... RELATOR ti This system, and particularly the strict locality condition between the base generated locative PP and the pro-‐predicate make some apparently false predications about locative inversions in there-‐existential constructions. They wrongly predict that (66) is ungrammatical: (66) On the wall, there hung a picture of Gandhi On any account of existentials, there is in Spec TP which is where the pro-‐predicate would need to be to be licensed. What is even more interesting is that native speakers express a preference for non-‐agreement with the post copular DP in cases movement by predicate inversion coreference 85 of locative inversions in existentials as in (67) (with the topicalization reading in (66)): (67) a. * In the room, there seem to be hanging pictures of Gandhi (*room-‐focus reading) b. In the room, there seems to be hanging pictures of Gandhi For those that don’t find (67a) ungrammatical, they take hanging to be a modifier of pictures of Gandhi (i.e. not free-‐standing pictures of Gandhi). In the case of (67b), however, people report that the sentence is about a room in which there are pictures of Gandhi hanging—i.e. the same room-‐focus reading as in (66). If we incorporate the insights of the DD06 proposal into the proposal for the structure of existentials in this work, I believe we can account for the observed differences as well reconcile the fact that as it stands now, the DD06 cant account for locative inversions in existentials. I will first illustrate the proposed structures for the sentences in (67), then I will discuss how to address the locality issue of the topic-‐PP and the pro-‐predicate. Lastly, I will illustrate how the structures can account for the different available readings. The structures are as in (68) and (69)—I use PredP instead of RP: 86 (68) In the room, there seems to be hanging pictures of Gandhi TopP PP TP In the room Therei T’ seems VP ti V’ to be VP ti V’ hanging PredP ti PredP PPk Pred’ DP pro-‐pred Pred 0 tk 87 (69) In the room, there seem to be hanging pictures of Gandhi TopP PP TP In the room Therei T’ seems VP ti V’ to be VP ti V’ hanging PredP ti PredP DPk Pred’ pics of... Pred 0 PP tk pro-‐ pred The first issue to be addressed in the structures in (68-‐9) is that the topic-‐PP and pro-‐predicate are not in a local relationship at all. Moreover, no predicate inversion has occurred as demanded by the constraint that all empty headed predicates A-‐move to Spec TP for licensing purposes a la (61). However, if we examine the reason that movement of the empty predict head must occur in the first place combined with the particular semantics of the existential construction, it may be the case that establishing a coreference relationship between the locative pro-‐ 88 predicate and the expletive there could serve to fulfill both licensing functions of the pro-‐predicate, check the EPP of Tense, and let the pro-‐predicate remain in Spec PredP. Recall the reason for predicate inversion was two-‐fold: (a) to mediate sentential focus—e.g. too ensure narrow focus on the post-‐copular/postverbal subject and (b) to license empty predicate heads that were, as such, unfit to be in complement position of an RP. In the existential constructions, as noted by Kuroda (1972) the postcopular subject acts like a detopicalized subject and receives narrow focus. In the preferred non-‐agreeing case, this is exactly the intonation that is produced: (70) In the room, there seems to be hanging PICTURES of Gandhi 27 The existential structure itself has already fulfilled, if only be default, the focus justification for predicate inversion. The question, therefore, is how is the locative inversion in existentials licensed as a base-‐generated topic-‐PP that must be bound be a local pro-‐predicate, when Spec TP is filled by there? Reed (1996) makes just this proposal, arguing that there is in fact an abstract locative therePRO which: ...must be linked to a locative in the discourse...In this analysis the anaphoric relationship between the XP constituent and there parallels that defined for definites: an accessible discourse entity must be available from the context or explicitly stated in the sentence. the explicit discourse locative occurs in those there-‐ sentences which contain and adjunct PP or VP which defines in the discourse context relevant to the assertion (Reed 1996: 157) 27 Note that in the agreeing case, the intonation changes as well: (i) In the room, there seem to be HANGING PICTURES of Gandhi 89 In keeping with the general intuition that existential sentences are really statements about predicates 28 , the Reed (1996) proposal imposes its form of the coda condition familiar from Keenan (2003) 29 . He determined that the thetic predication relationship could be encoded in a restriction on determiners such that the determiner had to be “conservative on its second argument” insofar is it used the coda constituent as the realm in which to evaluate truth. Note also that in the work proposed here (e.g. the structure in (68)) the movement of the PP from the complement of Pred to Spec PredP, also ensures that the existential construction is ultimately a statement about a predicate. The option of what to move from the complement to specifier of PredP is what determines not only the agreement on Tense, but the object-‐central or event-‐central readings (Sasse 1987) that thetic predication structures are predicted to have. From a syntactic standpoint, this solution also solves the need for the empty head predicate to move, therePROi is in an anaphoric relationship with pro-predicatei which is supplied by the locative PP that is a constituent of PredP 30 . The subsequent movement of the co-‐indexed therePROi to Spec TP means that the D feature of there can check the EPP tense and the coreferential anaphor PROi component can serve 28 e.g. for Basilico (1997) the subject of the existential SC is just a predicate modifier that is incorporated; or for Kuroda (1972) and Ladusaw (2000) the structure of existential sentences as thetic predication which are not property denoting, but statements about predicates as a whole; or Kennan (2003) who encodes the coda condition which denotes the primacy of the Coda XP in determining the universe of the sentence; or McNally (1992) that existentials are statements about nominalized functions (predicates) and not property denoting etc... 29 See Chapter 3 for a detailed freview of the Keenan (2003) semantics 30 Whether the PP moves entirely as in (28) to first Spec of PredP or whether only the DP specifier of PP moves to first Spec of PredP as in (29) is irrelevant since there PRO can be bound by the locative in either case, since it would c-command it in either position from where it is base generated in the second specifier of PredP 90 not only to license the empty headed predicate in its lower PredP position, but it can also serve as to license the topic-‐PP. Interestingly, it seems that the locality requirement need not be followed, however, in the locative inversions in existentials. DD06 says that the topic-‐pp may not move out of a strictly local relationship with whatever licenses it in Spec TP. However, in footnote 2 about example (62) I said that perhaps the Spec TP element moved with the topic-‐PP because of the ECM nature of the example. Here is (62) again: (62’) ECM a. Imogen expects the best candidate to be Brian b. The best candidate Imogen expects to be Brian c. *Imogen expects on this wall to be hung a portrait of Brian d. On this wall Imogen expects to be hung a portrait of Brian However, even if we give the benefit of the doubt on that one, the existential cases leave no doubt that the requirement is too strict: (71) ECM – Existential a. Imogen expects on this wall there to be hung a portrait of Brian b. On this wall Imogen expects there to be hung a portrait of Brian In the case of (71b) therePROi is clearly still in Spec TP and the locative has moved to a higher TopicP. Thus, in the case of existentials, where there is an overt anaphor therePROi in Spec TP forming a proper chain, with pro-predicate in a lower XP, licensing between the topic-‐PP and the Spec TP anaphor obeys the principle of minimal compliance (Boskovic 1999). Section 4: There’s vs. There is While there is considerable literature on the nature and significance of agreement vs. non-‐agreement in the English Existential Construction, there is 91 relatively little attention given to the contrast between *There is five cops at the door and There’s five cops at the door. Common solutions to this problem involve some idea that the agreement relation is between tense and there, but since there doesn’t have its own features, the ‘s is some kind of default agreement (Schutze 1999). Others, such as Sobin (1997), have argued for the agreement phenomenon under a ‘virus’ model where it is understood purely as a surface PF effect. I will say that whatever solution I offer here is speculative at best. Given the prior work on this project, I do take it to be the case that singular vs. plural agreement in ECs is meaningful and indicative of structural differences, contra Sobin (1997) and the virus model. However, I will contend that the difference between there is and there’s is in some sense a featural difference, such that the lack of actual [+3sg] on seeming non-‐agreeing DPs (which in reality are raised XPs of which the plural DPs are just a specifier) results in a default weak agreement ‘s. Indeed, the evidence would suggest that the phenomenon is not just limited to there in ECs, as the data in (72) suggest, there are a number of other Deictic pronoun and Wh-‐ cases where non-‐agreeing ‘s is licensed. The data, however, indicate that the full is is ungrammatical: (72) a. Where’s your books (DeWolfe 1992) b. Here’s the books you ordered c. How’s the horses? d. When’s the races? (b-‐d Sparks 1984) (73) a. *Where is your books b. *Here is the books you ordered 92 c. *How is the horses? d. *When is the races? In the cases in (1), I argue that there is an implicit adverbial or PP which functions as the phrase which triggers weak agreement. That is, I offer the following structure for (72a) as well as the structure for the agreeing version for reference: (74) Where’s your books? CP Where’si C’ ∅ TP ti T’ T+Vk VP ti V’ tk PredP ti Pr’ Pr 0 PP DP P’ P∅ ti weak agreement between there and T+Vk creates the where’s that ultimately surfaces in SpecCP. Weak agreement bc where is not [+3sg], thus default weak agreement. 93 (75) Where are your books? CP Wherej C’ arek TP DPi T’ your books tk VP ti V’ tk PredP ti Pr’ Pr 0 PP P∅ tj In both derivations, the matrix copular begins as a PredP consistent with the current work as well Bowers (1993), Den Dikken (2006), Roy (2003), among others. The difference is that in (74) the complement to Pred 0 is a PP with a null P and DP specifier your books. In this structure, where is the complement of the null P and is raised to the spec of PredP. The wh-‐ where however, is not marked specifically for 94 [+3sg] so when it raises to SpecTP to check the EPP, a default weak agreement occurs. The resulting output of that weak agreement is where’s. In the structure in (74) C is labeled as ∅, but I am agnostic as to whether this could just be T raising to C with no phonological reflex since weak agreement is already overt on the where’s. I should also note that the base PredP in (75) slightly different in the sense that the DP your books is base generated in Spec PredP which gives it the structure of a copular absent the complement to spec movement found in existentials and non-‐agreeing cases of existentials as well as those of the type in (74). In terms of how this explanation carries over to the proposal for existentials here, I am NOT proposing that there agrees with T in the same way that where is agreeing with T (i.e. I am not saying there is weak phi-‐featured there, I am still committed to the notion that there is [+D] and PRO bound by the locative). Rather, the weak agreement is a manifestation of the T probe finding the Pred goal which is valued by the XP of which the plural DP is a specifier. Section 5: Conclusion This paper has argued that the EC and the copular variants of sentences are related but structurally distinct. Furthermore, while both variations are licensed by an initial PredP a la Baker (2003), the nature of these PredPs is what distinguishes both ECs from copulars and different ECs from one another. In order to establish the nature of the PredP in ECs, this paper reviewed new empirical data from the English EC which focused on the structural ramifications of different agreement paradigms in ECs in raising contexts. Specifically, three phenomena were analyzed: (1) the scopal properties and their implications for grammaticality in ECs with 95 embedded OC constructions; (2) the availability of both singular and plural agreement on Tns for what appear to be the same surface structures and the different meanings they obtain; and (3) Quantifier float facts from post-‐nominal positions in ECs. Ultimately, this paper concludes, that ECs are the result of complement-‐to-‐Spec movement in the initial PredP and that what is assumed to be canonical S-‐V agreement really obtains between Tns-‐Pred. Thus, the agreement patterns that are licensed in ECs and the different readings they engender are a function of the type of complement that moves from complement-‐to-‐Spec where it values the phi-‐features and definiteness features of Pred. 96 Chapter 3: Predicational Restrictions in the English Existential Construction (EEC) and the Definiteness Effect Section 1: Reconceptualizing the Predication Restrictions in the EEC The English EC has long been noted to have sensitivities to both the type of quantification of the postcopular DP (Milsark 1974, 1977) and the property-‐ denoting qualities of the predicate (or ‘coda’ in existential literature). Definite/Strong-‐quantified postcopular DPs are (generally) ungrammatical, whereas weak/indefinite postcopular DPs are licit. In terms of the codas that are licit in existentials, predicates denoting transient properties (Stage Level) are acceptable, while predicates denoting permanent properties (Individual Level) are ungrammatical. The full paradigm is given below: (1) Def/Indef & Strong/Weak OK as normal subjects a. John is in the garage b. The man is in the garage c. Every man is in the garage d. A man is in the garage e. Some man is in the garage (2) *Def/Strong EC subjects a. *There is John in the garage b. *There is the man in the garage c. *There is every man in the garage (3) OK Indef/Weak EC Subj a. There is a man in the garage b. There is(/are) some man(/men) in the garage 97 (4) Predicate Restrictions a. There are comets visible (SL) b. *There are comets heavy (IL) Since Milsark (1974) and Carlson (1977) the characterization of the problem has been in terms of the now familiar distinction between “stage-‐level” (SL) and “individual-‐level” (IL) predicates. Indeed, these concepts and their entailments for their respective subjects have generated numerous proposals and extensions to account for the distributions of predicates found in EECs and the quantificational/definiteness restrictions observed cross-‐linguistically with respect to post-‐verbal determiners and nominals. Most notably, the distinction illustrated in Kuroda (1972) between thetic and categorical predication/judgment-‐forms has been argued to be an underlying epiphenomenon regarding information-‐structure that both naturally characterizes ECs as a specific type of predication (thetic) and therefore accounts for the predicate distribution and DE facts (e.g. Ladusaw 2004). While these lines of research have yielded interesting results and insights in a number of domains of linguistics, they have failed to accurately account for the distribution of predicates, determiners, and nominals found in ECs cross-‐ linguistically. This section, therefore, will extend the proposal in Roy (2005) to abandon the IL/SL distinction in favor of a three-‐way distinction between characterizing, situation-descriptive, and defining predicates that are derived from a single syntactic configuration and are distinguished by the structural properties of the XPs that inhabit the terminal nodes of that Small Clause/PredP structure. To that end, the first part of this chapter will introduce the IL/SL distinction and its 98 relevance to the EEC. The next part of the section will introduce the distinction of thetic and categorical predication and examine how it has been argued to interact with IL/SL predication and the EEC. Finally, the section will end with a discussion of the short-‐comings of the IL/SL distinction and the proposal of a more syntactically rigorous alternative approach that both accounts for the distribution of predicates in EC environments and encodes the information-‐structure properties of thetic predication. Section 1.1: The IL/SL Distinction and the EEC While the terminology IL and SL has its roots in Carlson (1977), the distinction is based on work in Milsark (1974). Milsark (1974) identified two crucial distinctions: one between types of DPs and one between types of predicates. DPs can be either cardinal, quantificational, or ambiguous between the two. It should be noted that Milsark’s distinction between types of DPs was later reformulated as distinctions between types of determiners (or generalized quantifiers) in Barwise and Cooper (1980). This development reconceptualized the distinction as one between “weak” (cardinal) and “strong” (quantificational) quantified DPs (with some determiners remaining ambiguous) as in (5): (5) a. Cardinal/Weak: a cat, some people b. Ambiguous: several women, many cars, bare plurals c. Quantificational/Strong: the man, this cat, each, all… In Milsark (1974), predicates also were divided into two classes, one class described (semi-‐)permanent properties and was called “Properties.” The other class of 99 predicates described more transient properties and were dubbed “State Descriptions.” Carlson (1977) coined the now-‐used terms ‘Individual-‐Level (IL)” for ‘Properties’ and “Stage-‐Level (SL)” for “State Descriptions”: (6) a. Properties/IL: tall, intelligent, a lawyer b. State Descriptions/SL: sick, drunk, open (Ladusaw 2004: 2) Given these distinctions, there are four possible combinations: (7) Possible Combinations: Weak/Strong & IL/SL Strong-‐IL The man is tall Weak-‐IL *Some man is a lawyer Strong-‐SL The man is sick Weak-‐SL Some man is sick Given the data in (7), we can formulate a version of the famous Milsrak’s Generalization: (8) IL predicates require strong subjects (i.e. ‘properties’ can only be predicated of quantificational subjects (or ambiguous subjects interpreted quanitificationally)). To account for the distribution of predicates and the DE in the EEC, Milsark (1974) proposed that existential sentences be interpreted by taking “There BE” to be an existential quantifier that takes an non-‐referential predicative NP (or property-‐ denoting NP in terms of McNally 1992, 1998) as its restrictor. The reason that the NP must be non-‐referential is that it must be able to be existentially quantified as the restrictor of the existential quantifier. If the NP were Strong/Quantificational then the NP would be double quantified and the sentence would be ungrammatical. A corollary to this analysis is that since the NP can never be quantificational, i.e. it 100 can not be a strong NP, then we should not expect IL predicates in the coda of existentials since IL codas can only be predicated of strong subjects as per (8). Section 1.2: Thetic/Categorical Predication, the IL/SL Distinction, and the EEC Building on the IL/SL analysis, some researchers (Ladusaw 2004, Leonetti 2002) have attempted to articulate the IL/SL distinction as a natural consequence of another more fundamental distinction between the information structures of two types predication: categorical predicational structures and thetic predicational structures. To keep the issues clear, let me define two operational concepts: (9) a. Predicate-‐ an unsaturated expression that denotes some property b. Predicational Structure-‐ A fully saturated structure that expresses some relationship between a subject and predicate The distinction in (9) will allow us to conceptualize the IL/SL distinction as a distinction among predicates, and the categorical/thetic distinction as a distinction among predicational structures. To be concrete, if permanence and transience are the only characteristics we use to describe the difference between SL/IL predicates, the description risks being purely dependant on the lexical semantics of predicates. Encoding the IL/SL distinction structurally is a superior option. Indeed, if we leave the distinction as one of permanence/transience then in cases like (4a), we are left to conclude that there must be 2 visibles in the lexicon, on visibleSL as in (4a) and another visibleIL which would lead to an ungrammatical sentence in the case of (4a) (Cf. the reading that there exist comets that are visible as opposed to comets that are 101 invisible) 31 . However, regardless of how one tries to structurally represent IL/SL predicates, the representation will only be one of predicates. That structural representation might entail that it can only be predicated of certain types of subjects, however, the structural description of IL and SL predicates is constructed without reference to the role of the subject. Subject restrictions therefore fall out as consequences of the structures of predicates (9a). Thus, while the differences between SL and IL predicates are to be found within the structures of the predicates’ themselves, there is an overlapping bifurcation of predicational structures (9b) which divide these predicational structures into either: (a) those that assert the existence of a subject and assign a property to it—categorical predicational structures or (b) those recognize or reject a ‘material judgment’—thetic predicational structures (Kuroda 1972, Ladusaw 2000, Brentano 1879). To illustrate this take Kuroda’s (1972) examples: (10) a. neko ga asoko de nemutte iru (Thetic) a/the cat SM there sleeping is b. neko wa asoko de nemutte iru (Categorical) the cat TM there sleeping is In (10a), the Subject Marker ga is used to convey the perception of a situation where a cat is sleeping in a particular place. The (10b) variant, with the Topic Marker wa, 31 For instance, Kratzer (1989) offers a structural account of IL and SL predication illustrates that SL predicates contain and event(uality) variable and IL predicates do not. 102 has a categorical interpretation in the sense that it draws attention to the cat and then says of that cat that it is sleeping. While these two predicational structures could be used to describe the same situation, there is a difference in their referentiality: In [(10a)], only the existence of the eventuality is affirmed, but commitment to that description will indirectly...commit the judger to the existence of the cat. But the cat has a much fuzzier existence that in the thetic judgment that it does in the categorical version [(10b)], where it is presented and affirmed directly, first of all, before moving onto the predication...the thetic mode of judgment nicely represents the effects of unselective existential closure over the description, giving priority to one object, while obliquely entailing the existence of subparts of the complex object (Ladusaw 2000: 237) In (10a) the fact that the definiteness of the cat is ambiguous (i.e. a/the cat) is what is crucial. The subject status of the NP (ga subject marker) means that the NP is lower within the structure. Ladusaw (2000) suggests that means the entire description of the event is existentially bound 32 , and to the extent one is committed to the truth that the event of [a/the cat sleeping] at some place [there] is happening, one is committed to the entailment that a/the cat exists. Note that this licensing of either strong or weak construals of the NP is similar to that found in the types of subjects that are interpretable with SL predicates. The distinction in the type of referentiality that thetic/categorical predicational structures encode results in complex interactions with SL and IL predicates. In essence, when it comes to predication, at the very least, there is a macro-‐distinction between thetic and 32 The story is that the NP-ga originates in the VP and can reconstruct. VP is the domain of existential closure. In the case of the NP-wa the element is base-generated in a topic position in the left periphery higher than VP and cannot be existenitally closed. Also, we know independently that topics are generally definite. 103 categorical predicational structures and a micro-‐distinction between SL/IL predicates such that a decision at the macro-‐level will determine the possibility of a choice at the micro-‐level. The Categorical/Thetic distinction is taken as the macro-‐distinction precisely because it is a distinction that is determined by the relationship between the subject and the predicate; it is therefore a distinction manifested at the matrix level. The IL/SL distinction, however, is manifested in the predicates themselves. Following Ladusaw (2000), if we assume that semantic components available in making statements are: individuals, event(ualitie)s, descriptions of individuals, descriptions of event(ualitie)s, and properties, then we can create a clearer understanding of the differences between thetic and categorical predicational structures. At the macro-‐level, we can say that thetic judgments are about accepting/denying descriptions and categorical predications are about individuals and properties where the existence of the individual is asserted directly and the property is predicated of it. At the micro-‐level, IL predicates must be interpreted relative to a strong quantified subject; otherwise, the result is a generic where the indefinite or bare plural subject is interpreted as a kind. For SL predicates, the same strong quantification/presuppositionality requirement does not hold. In these cases, the existence of the subject is not as relevant since the predications can be understood as event(ualitie)s. It is important to note that there is nothing ruling out the possibility of strong subjects in SL contexts. In the following examples, Somewhere in the world can be placed in front of SL predicates with strong and weak 104 subjects, while it is impossible in front of IL predicates, also weak subjects are out generally in IL sentences: (11) SL c. A man is thirsty d. John is thirsty c. Somewhere in the world, a man is thirsty d. Somewhere in the world, John is thirsty (12) IL a. *a man is a lawyer b. John is a lawyer c. #Somewhere in the world, a man is a lawyer d. *Somewhere in the world, John is a lawyer Here we have illustrated Milsark’s (1974) generalization that IL predicates must have strong subjects. But rather than leaving this as a generalization, it is correlated with the referentiality differences inherent in the macro-‐level of the predicate distinctions. That is, IL predicates are a type of categorical predication which must have strong subjects, where strong can be understood as presupposed (e.g. definite, strong-‐quantificational). Predicates cannot receive an IL interpretation in thetic predication constructions because strong subjects are object-‐denoting and thetic predications are based on descriptions of objects not objects qua objects. Moreover, existential constructions have a special discourse function. According to McNally (1992) they introduce new discourse referents into the discourse. Thus, put simply, the logic is that indefinites by their nature introduce new referents but definites/strong-‐quantified DPs generally do not. 105 Section 2: A Thetic Predicational Approach to the IL/SL Restriction in EECs—Basilico (1997) Basilico (1997) offers a preliminary structure for the EEC very similar in spirit to the one I adopt in Chapter 1: (13) a. There are comets visible/*heavy b. [TP Therei [T’ arej [VP tj [DP [D’ ti [SC comets visible]]]]] (Basilico 1997: 281) Basilico says of this structure, “’there’ functions as a topic quantifier, which binds an event variable supplied the predicate head of the SC complement. The SC functions to provide the range for the topic quantifier, just as any predicative constituent provides the range for the quantificational determiner associated with DP” [e.g. every planet] (Basilico 1997: 279). Following Stowell (1983, 1991 a.o.), he argues that the SC constituent is composed of elements in a subject-‐predicate relationship such that the predicate is the head and the subject occupies a specifier a position. As far as the licensing of the head of the SC is concerned, Basilico (1997) argues that: In copular sentences without ‘there’, the head of the SC would be licensed by T 0 . However,...when the expletive is present, ‘there’ is a D 0 whose trace is in a proper locality configuration with the SC predicate head, functioning to license the SC head. In effect, the entire is SC is functioning as an argument expression. (Basilico 1997: 280) We can assume from his description that a copular structure would be: (14) [TP cometsi [T’ arej [VP tj [SC ti visible]]]] Both (13b) and (14) seem to suggest that there is some V-‐T movement of the copula. These are the basics of the Basilico proposal which he argues accounts for both the 106 strange subject behaviors of the postcopular XP in ECs and the semantic restrictions on predicates in those XPs. Section 2.1: There as a Topic Marker (TM) and the IL/SL Predicate Restriction Crucial to the Basilico (1997) analysis, following Sasse (1987) and McNally (1992), is the notion that there are different types of predication and that ECs are thetic predicational structures as opposed to categorical predicational structures: (15) a. thetic predication-‐ “single presentation is made about the existence of some event of entity”(Basilico 1997: 281) b. categorical predication-‐ an object is presented and a property is ascribed to it The idea is that if we treat ‘there’ as a quantifier in D 0 with the SC as its restrictor, then there is a DP that is the complement of the matrix ‘be’ but the quantifier and the restrictor act as separate parts of the same argument. This has the intuitive benefits of explaining the distributed subject effects (e.g. ‘There’ in [spec TP] but postcopular NP dictating agreement on T) as well as making the predicate part of the only argument in the sentence and thus being a single presentation about the existence of some event or entity—thetic predication. It is this notion of thetic vs. categorical predication that presumably allows Basilico (1997) to account for the IL/SL distinction. That is there is a relationship between IL/SL predicates and categorical/thetic predication such that sentences with SL predicates allow for thetic judgments and categorical judgments, and sentences with IL predicates only 107 allow categorical judgments (Kuroda 1995). Furthermore, Raposo and Uriagareka (1995) argue that: categorical SCs [SCs which contain IL predicates] are not separated from the verb that selects them. The verb carries the Tense that enters into a governing relation with the SC. If the SC is displaced out of the governing domain of this verb, the [categorical judgment] will not be sanctioned (205). Crucially, the difference between the structure in (13a) and the copular structure in (14) is that in (13a) the SC is licensed by D 0 and thus cannot have an IL predicate in the SC and in (14) the SC is licensed by T 0 which does allow IL predicates. The IL/SL distinction falls out as a result of the fact that in ECs T cannot govern the head of SC. To account for the peculiar subject-‐like effects of the post-‐copular NP (i.e. that it generally determines agreement with the copula), Basilico invokes this same notion of the [DP there [SC ... ]] as a single DP unit to account for the seeming agreement with the lower NP. He begins by noting that in other languages an agreement relation is morphologically realized between the D and the NP (i.e. Spanish las (fem. pl.) sillas (fem. pl.) ‘the chairs’). This observation prompts the following generalization: (16) A D 0 agrees with the features of its predicative complement (pp 292) The ultimate claim is that the D 0 there agrees with its predicative element SC and then moves to [Spec TP] to check the agreement features on T 0 . However, in order for this explanation to work, Basilico must say something about the structure of the SC complement. Recall that the first assumptions we made about the SC were that the predicate was the head and the subject was in a specifier position. However, if 108 we leave the structure at that, then we make the prediction that the post-‐copular NP is in a subject position (albeit the subject position of the SC) and therefore should behave as such. If we do a simple extraction test, we find that we can extract elements from the post copular NP which we should not be allowed to do if it were the subject of a SC as in (17): (17) a. which planet is there a picture of on the screen b. *which book did you find the author of very eloquent To get around this problem, Basilico again turns to the notion of thetic predication. The idea is that since the sentence is a single statement about the existence of a entity or event, and that entity or event is structurally represented by the predicate as a whole, then the NP is in fact a predicate modifier. He says, “thus, the post-‐ copular NP becomes part of the description of the event expressed by the SC predicate head” (Basilico 1997: 293). He then asserts (without much technical explanation) that the NP is incorporated into the SC head and as a result of this incorporation the NP’s number features are available to the head and percolate up to the SC maximal projection. Despite this somewhat technical solution to the agreement problem, there are still at least two major problems that remain with the Basilico (1997) analysis as it stands now: (1) If There is a D head but then moves to check T head phi-‐features, then we seem to have a chain-‐uniformity violation, because [spec TP] is reserved for maximal projections and (2) there would be the only D head that had the ability to do this. Take the passive for instance: 109 (18) a. *The was destroyed asteroid b. The asteroid was destroyed c. *[TP thei [T’ was [VP destroyed [DP ti [asteroid]]]]] (285) The answer to both of these problems lies in the incorporation of Tense into the topic marker by way of adopting a more articulated structure of the TP. To motivate the first part of the solution, Basilico notes that the sentence There are comets visible is not just a comment on an event of comets being visible, it is crucially a non-‐past event. The second part of the solution appeals to the idea that specific subjects overtly move to a position above TP, in this case [Spec RefP] (similar to some versions of AgrSP). The specifics of the solution come from Kiss (1996) who observes that there behaves like a specific subject and must therefore be [+specific]. Evidence of the specificity of there comes from the fact that it patterns with specific subjects in tag questions (Gueron 1980) and Only-‐modification: (19) a. *A man arrived, didnt he? b. *A storm occurred, didn’t it? c. There shouldn’t be any quarrels, should there? (20) a. *[Only[JOHN READ A NOVEL BY KARL MAY]]; nothing else happened b. [Only[A BABY WAS BORN]]; nothing else happened c. *[Only[THERE WAS AN ACCIDENT]]; nothing else happened Based on this evidence, Kiss (1996) argues that [+specific] subjects are raised to a higher specifier position in RefP so the head of RefP can check that feature. If a subject is non-‐ specific it remains in [Spec TP]. Thus, since there is [+specific] it will raise to [spec RefP] to get checked. The resulting structure is: 110 (21) [RefP Therei [Ref’ [TP ti [T’ arej [VP tj [DP [D’ ti [SC comets visible]]]]]]] To answer the problem mentioned above of there being the only D capable of this exotic movement, Basilico invokes the special lexical semantics of there and the chain uniformity condition. Basilico argues that other Ds cannot take TPs as restrictors so the solution presented here could never work for them and they would be ruled out a chain uniformity violations. Basilico (1997) also uses this structure to argue for his chain uniformity solution. The argument is that from the structure in (21), the entire TP raises to adjoin to there in [Spec RefP] (presumably at LF) thus creating a proper DP to govern the trace in the DP below (i.e. it started out as a [DP D [restrictor]] and ends up as [DP D [restrictor]]) so proper government is maintained. There are however, two more structural problems that arise from this move: (1) even though the head and foot of the chain are both [DP D [restrictor]], the trace in [spec TP] is still just a trace of D there, and (2) if the unit that is formed in [spec RefP] is [DP there + TP] then presumably TP which is complement to the head D will no longer c-‐command its trace. The answer to this problem, according to Basilico (1997) is to destroy the evidence of the crime; that is, the answer is to delete the offending traces. As far as the trace in [spec TP], he argues that since it only moved to check features it can be eliminated, and for the trace of the TP it can be deleted because it does not license anything and all its relevant information has been incorporated into the restrictor of the DP which is now in [Spec RefP]. The final structural proposal then is: (22) [RefP Therei [TP are [DP [D’ ti [SC comets visible]]] Ref] (Basilico 1997: 297) 111 Section 2.2: Basilico (1997)—Theoretical and Empirical Problems The Basilico (1997) proposal, while at times technically awkward or exotic, encodes important insights about the existential construction. Namely, that ECs are thetic as opposed to categorical predicational structures, the licensing properties of SCs and the relationship between Tense and the predicate. Unfortunately, the Basilico proposal runs into serious problems with chain uniformity and thus relies on a complex system involving rule violations rectified by convenient trace deletion. There are at least two significant problems with the Basilico (1997) proposal as it stands now; one is theoretical, and the other is empirical. In terms of theory, the crucial movement that facilitates the derivation-‐saving trace deletion is unmotivated, and in terms of empirical inadequacy, it makes questionable judgments about subject extractions and cannot account for the agreement variations discussed in Chapter 1. To first address the theoretical problem, while not explicitly saying it, the Basilico proposal relies on the Chomsky (1991) proposal(s) that LF or covert movement is necessary to combine the EXPL there and its associate. For Chomsky, this is a way of getting the appropriate phi-‐features in [spec TP] to satisfy the agreement requirements of the copula, since the featural composition of EXPL there is only [+D] which only suffices to check the strong EPP feature on finite Tense (Chomsky 1991). Without mentioning it however, Basilico (1997) uses the same mechanism of LF movement, but for an entirely different reason. In Basilico’s system, both the phi-‐feature agreement and the EPP on Tense are entirely checked by EXPL there. Recall that in Basilico (1997), the postcopular NP is simply a predicate modifier that incorporates into the SC head (the SC 112 predicate) such that all its phi-features are absorbed and percolated up to the SC maximal projection. The D head there then enters into an agreement relation with the SC and is thus imbued with the necessary phi-‐features. Thus, in Basilico (1997), the LF/covert movement of TP cannot be for the same reasons. In fact, it is unclear what reason there is for the crucial movement of the TP to adjoin to D there in [Spec, RefP] as in (22). Without this move, the violating traces will still be visible and computed at the interfaces and the derivation will crash. The empirical problems for the proposal are two-‐fold. First, the evidence that Basilico (1997) cites for the ability to extract out of the postcopular NP is questionable. While in (17) he argues that the (17a) variant is good, many native speakers disagree with this judgment or find it to be significantly degraded. The other empirical problem with the Basilico (1997) proposal is central to the discussion at hand of clarifying and further developing the notion of SC in Chapter 1. That is, there is no way that the Basilico (1997) proposal can account for the different agreement patterns and their respective readings in: (23) a. There seem to be the occasional sailors walking by b. *There seems to be the occasional sailors walking by c. There seem to be five cops at the door (count reading) d. There seems to be five cops at the door (mass reading) In fact, the Basilico (1997) proposal wrongly predicts (12d) to be wrong. According to the Basilico (1997) model, the NP five cops is only a predicate modifier and will incorporate into the PP at the door and D there will agree with the plural SC five cops at the door and then move to check/value T 0 : 113 (24) [TP Therei [T’ seem [VP to be [DP [D’ ti [SC [NP5 cops] [PPat the door]]]]]] Section 3: The Devil is in the Details—Abandoning IL/SL and Salvaging Theticity It would appear, given the distinction between IL/SL and thetic/categorical, that we should have a fairly complete and compelling story for both the DE and the predicate restrictions we observe in EC environments. However, when one attempts to systematically implement these distinctions in such a way as to make predictions and create natural classes of predicates and determiners/quantifiers, the system breaks down. Immediately, one is confronted with the fact that while permanent/transient may seem like an intuitive way to distinguish predicates, both the definitions of these concepts and the syntactic tests designed to illustrate them fail to make stable predictions across contexts. The case of the thetic/categorical distinction suffers a similar fate upon implementation. While the insight that there are two types of predicational structures seems to be warranted, the actual detail of these structures is still a mystery. The first difficulty one encounters when trying to implement the restrictions described in terms of IL/SL or Thetic/Categorical is that it becomes unclear which predicates are which. Milsark (1974) grapples with this problem and in the end concludes: 1. incorporate 2. Agree 3. move 114 The best I can do is to suggest some tendencies and rules of thumb, plus an imprecise definition or two. Properties are those facts about entities which are assumed to be, even if they are not in fact, permanent, unalterable, and is some sense possessed by the entity, while states are conditions which are, at least in principle, transitory, not possessed by the entity of which they are being predicated, and the removal of which causes no change to the essential qualities of the entity (212) Obviously, this is not a technical definition, but if one were to try to implement this definition, it seems that it would start with the assumption that NPs are the best candidates for prototypical properties/IL predicates, PPs are the best state/SL predicates and adjectives are somewhere in the middle depending on their lexical semantics. While this seems like a reasonable grouping of predicates by syntactic category, it immediately falls apart when applied to natural language. To exemplify this, one need only examine the first assumption (NPs make IL predicates, NPs are ‘permanent’ properties) in the context of predicative NPs in French. Recalling the discussion of Roy (2005) from Chapter 1, we are immediately confronted with a problem with the first assumption. In (25), we see that even within predicative NPs there are eventive/transitory possibilities and those possibilities are mediated by the syntactic structure of the NP: 115 (25) a. Paul est un acteur (Stable property/defining) P is an actor ‘P is an actor’ (Cf. English—Q: What does P do? A: P is an actor) b. Paul est acteur (eventive property) P is actor ‘P is an actor’ (Cf. English —Q: What does P do? A: P acts) Already the assumptions that the IL/SL system led us to are in trouble. However, even if we adjust our assumptions and suggest that (for whatever reason) NPs do not behave as a homogenous group and that, in French, Bare NPs are transient/SL predicates and [det + NP]s are property/IL predicates, the system still fails. While the adjusted assumption that “[det + NP]s are property/IL predicates” turns out to be correct, our bare NP predictions are false. Roy (2005) identifies four canonical tests that are used in the literature to determine permanence/transience of a predicate: (26) Permanence/Transience Tests a. Temporal Modification-‐ If a predicate is transient then it can be temporally modified (Kratzer 1995) b. Spatial Modification-‐ If a predicate is transient then it can be spatially modified (Kratzer 1995) c. Lifetime Effects-‐ If a predicate is in the past tense, and that entails the subject is dead, then the predicate is permanent (Musan 1995) d. When adjuncts-‐ Transient predicates are licensed in when-‐adjuncts with referential subjects, permanent predicates are not (Kratzer 1995) 116 However, when we apply these tests to bare NPs in French we are left with a very unsatisfying picture. BNPs in French do not pattern together with respect to these tests; in fact, three overlapping classes of BNPs emerge ranging in the extent to which they are interpreted as transient or permanent leaving us with no useful classification of predicates. The first type of French BNP that emerges is the “medecin” class a hybrid class that sometimes patterns as a transient property and permanent property. The second class that emerges is the “fils de diplomate” class which patterns exclusively as permanent properties. Finally a third class—the “candidat” class—patterns as transient properties. The data is summarized in the table below: (27) French BNP Classes The above facts motivate Roy (2005) to propose that the relevant distinction should not be one based on the transience/permanence of the property denoted by the predicate, rather the distinction should be a syntactic distinction between the BNPs which are characterizing predicates, and the [det+NP] predicates which are defining predicates. At this point, there are two ways one can conceptualize Roy’s 117 distinction: (a) defining predicates are necessarily permanent, and characterizing predicates may or may not be or (b) defining predicates have count structure and characterizing predicates do not. If we conceptualize the distinction as it stands now as in (a) then we are left to wonder why the “fils de diplomate” class is not a defining predicate. Also, (a) is simply a French specific restatement (or notational derivative) of the IL/SL distinction. Conceptualizing the distinction as in (b), however, begins to give us some useful structural insight into the properties and structures of the various types of predicates. However, if we simply leave the distinction as it is now (between defining and characterizing predicates) and we understand that difference to be the result of (the presence or absence of) count structure, then we are led to a peculiar conclusion. Once we take into account other predicative syntactic categories (most notably AP and PP), we are lead to the conclusion that they should behave in the same way as BNPs because they cannot have count structure (see Chapter 1 for details; also Borer 2005). While it is true that both a BNP and an AP can be predicated as properties of an entity as in (28), the BNP and AP are in complementary distribution in eventive contexts as in (29): (28) Qu’est Paul What is Paul? a. Paul est invrogne Paul is a drunkard b. Paul est ivre Paul is drunk 118 (29) Qu’est ce qui s’est passé dehors, c’est quoi tout de bruit? What is going on outside, what is all this noise about? a. Paul est ivre, il a renverse la poubelle comme toujours Paul is drunk, he knocked the trash can over as always b. #Paul est ivrogne, il a renverse la poubelle comme toujours Paul is a drunkard, he knocked the trash can over as always (Roy 2005: 66) The contrast between (28) and (29) illustrates that there must be one more distinction between predicates. Roy (2005) proposes that we must split the characterizing predicates and the situation-descriptive predicates. Roy’s taxonomy of copular sentences is as in (30): (30) Predicational copular sentences ‘property-‐assigning’ sentences situation-descriptive sentences defining sentences characterizing sentences (2005: 90) The taxonomy could, however, be represented as a syntactic taxonomy as well: (31) Non-‐Verbal Predicates Nominals AP PP BNP Count ClP NumP 119 When we compare the taxonomies in (30) and (31), interesting correlations come to light. First, ‘property-‐assigning’ sentences (i.e. both characterizing and defining) from (30) are just various flavors of nominal predication, while situation-descriptive sentences are those with any non-‐verbal predicate that is not nominal. Informally, Roy (2005) suggests that the interpretational differences are that, “the characterizing reading [is] a predication of a characteristic property of an individual, while the situation-‐descriptive reading describes a situation that the individual is involved in” (2005: 171). Ultimately, recapping Chapter 1, the proposal for the structure of copular sentences with non-‐verbal predication in Roy (2005) is that the subject and postcopular predicate begin in a predication relation as a SC headed by Pred which licenses the external argument. Since the intensity and divisibility are governed by the available functional structure under which a given category can be embedded, it follows that only one SC structure is necessary and the variation in predicate types is purely based on the properties of the complement of Pred (Roy 2005: 214): (32) PredP = SC DP Pred’ Pred XP [=DegP, PP, ClP, NumP, BNP] all have and e variable red in the garden doctor 120 In the case of a copular construction that is ambiguous between a characterizing reading and a situation-‐descriptive reading, the structural point of ambiguity is reduced to the content of the XP, and not a fundamentally different SC structure: (33) PredP sick(e) & Subj (e, DP) DP Pred’ λx Subj (e,x) Pred ClP/DegP sick(e) sick Given the semantics in (36), the reading that is obtained is purely a function of whether the event introduced by the complement of Pred is count as in the case of the ClP resulting in a characterizing reading, or whether it is mass as in DegP which gets a situation-‐descriptive reading. While the Roy (2005) system obtains the correct results, there is one puzzling aspect to it that is never mentioned. The Roy (2005) system is totally committed to the idea that lexical categories must be introduced into the structure by functional heads. This is certainly well accepted within the literature, but Roy (2005) never discusses what might occupy the specifiers of these functional projections. Indeed, following Baker (2003), the key difference between lexical categories and functional categories is that functional categories introduce specifiers and lexical categories do not. 33 To address this issue, I will propose that the specifiers of PP and DegP are occupied by a pro that is bound by the DP base 33 Baker (2003) does say Vs have specifiers but only because they are the result of incorporating a functional Pred head into the A such that a [A+Pred]=V. V thereforeis a derived category in this system and not a counter example. 121 generated in the specifier of PredP. While this changes nothing about the semantics (33) (i.e. pro is type <e> and is bound by the type <e> DP base generated in [Spec,PredP]) it will be a crucial structural point as we extend the Roy (2005) proposal to existential constructions. Section 4: The Predicate Restriction Explained: The EEC is (a version of) Situation-‐Descriptive Returning now to the case of the existential construction, it is clear that the IL/SL distinction is insufficient to handle the data, and that even when it is enhanced with an epiphenomenal notion of thetic predicational structures, it cant capture all the right generalizations. However, if we follow the intuition in Chapter 1 that existential constructions are derivatives of their copular counterparts and if we adopt the Roy (2005) proposal, we immediately see striking similarities between existential constructions and situation-descriptive copular sentences. From an information-‐structure standpoint, thetic predicational structures exhibit the same oblique entailment of the subject as situation-descriptive sentences. Even more strikingly, Situation-‐Descriptive sentences have the exact same predicate restrictions as existential constructions. Lastly, if we take the EEC to be derived from a situation-‐descriptive predicate, we can derive the DE straightforwardly (see the next section). The basic structure that I am proposing for the English EC is the following: 122 (34) Structure of the English EC TP therePROi T’ Tns VP ti V’ to be PredP [XP or YP from [Spec XP] Pr’ Pr XP This structure starts out with some XP (e.g. X = Deg, P, V) 34 with a functional head Pr to form a properly headed small clause, PredP. The structure in (34) is virtually identical to the structure proposed for situation-‐descriptive copulars in Roy (2005) (cf. (32)). However, whereas in Roy (2005) the subject DP was base generated in PredP and in a coreference relationship with a pro in the specifier of the XP complement of Pred; in (34), the subject is base generated in the position where the pro would go and then moved to Spec PredP (or alternatively the entire XP is moved to [Spec PredP]). Indeed, the next step of the derivation involves movement of one of two things; either (a) the specifier of XP moves to the [Spec PredP] or (b) the XP itself moves to the [Spec PredP]. It is the option that is exercised here that will ultimately be responsible for the agreement flexibility and availability of multiple readings in ECs illustrated in Chapter 1. 34 X ≠ N, D because DP/NP codas would be IL predicates since they would represent permanent properties (although see Roy 2001 for detailed discussion). Hence *There is a man a hero 123 It is precisely in this flexibility that we are able to capture one of the most tantalizing facets of the thetic predicational structure vis-‐à-‐vis the agreement data in Chapter 1. Sasse (1987) notes that there are generally two readings available in thetic predication constructions—entity-‐central and event-‐central. It is the contention of this paper, that the thetic predicational structure established in existential sentences can, generally speaking, be expressed in two ways. Either the entire XP moves to [Spec PredP] resulting in the mass reading (Sasse’s event-‐central reading) or the [spec, XP] can move to [spec PredP] in which the count reading obtains (Sasse’s entity-‐central reading). Note that in cases where [spec XP] is not plural, these options are indistinguishable since both options would trigger singular agreement on the copula. Thus, in the sentence there is a cop at the door both the entity and event-‐central readings are available, they are just not morphologically distinguished. 35 Whichever movement one makes, however, crucially therePRO is not in the derivation at this point. That is to say, therePRO does not originate in the PredP and is therefore not in a subject/predicate relationship with the post-‐copular DP or the coda. The therePRO is merged at the edge of the VP in [spec VP]. TherePRO is subsequently moved to Spec TP, the therePRO when in [Spec TP] is a locative subject which asserts the existence of a location L 36 (at a time T), the complement of that TP is a fully saturated predicational structure P of a (post-‐verbal) subject and a coda 37 : 35 Categorical predication is established by base generating a subject in [spec PredP] such that a predication relation holds in the traditional sense. 36 The details of that location are provided by the discourse or the coda explicitly 37 The coda is, to some extent, optional when the location is generic or can be implied from the discourse 124 P(subject-‐coda). The role of the verb be is to express affirmation (or negation in the case of not be) of whether P(subject-‐coda) = 1 at T in L. Indeed, this thetic judgment of whether a predication relation is true/false at T in L is a natural consequence of the EEC being derived from a situation-‐descriptive predicate since situation-‐ descriptive predicates are intense and non-‐divisive. That is, one does need to determine some extent to which the statement is true, 38 it is either true of (an instantiation of) that subject at T in L or it is not. The reason for the element therePRO is that it must be bound by some locative that is either supplied by the coda (e.g. a VP or PP) or is supplied by the discourse as some kind of relevant default. The EXPL therePRO : ...must be linked to a locative in the discourse...In this analysis the anaphoric relationship between the XP constituent and there parallels that defined for definites: an accessible discourse entity must be available from the context or explicitly stated in the sentence. the explicit discourse locative occurs in those there-‐ sentences which contain and adjunct PP or VP which defines in the discourse context relevant to the assertion (Reed 1996: 157) As far as the derivational syntax of the EC is concerned, the story is that therePRO is [+D] but has no phi-‐features. That means that the EPP probe on Tns can be satisfied by AGREE plus movement of the EXPL goal following Chomsky (2001). However, this procedure still leaves Tns unvalued for phi-‐features. Instead of adopting the LF affixation or LF movement solutions of recent Minimalist literature wherein features and/or lexical items move to join/replace the EXPL at LF, I argue that the remaining unvalued phi features on Tns are an active probe that AGREEs with the head Pr which inherits phi-‐features from its specifier. 38 For instance, in the case of a characterizing predicate just because soemone sings does not make them a singer per se, we must determine whether they sing enough to be labeled a singer and have that label be true when they are not singing 125 Section 5: The Definiteness Effect and the EEC—A Non-‐Small Clause Semantic Approach the DE To explain the DE McNally (1998) adopts a neo-‐Milsarkian approach to the semantics of the EEC. Rather than try to locate the source of the DE in the IL/SL restriction, McNally (1998) suggests that the DE is in fact two different restrictions: one is a semantic restriction imposed by the existential predicate, and the other is a pragmatic restriction imposed by the discourse model. In this proposal, rather than taking the Barwise and Cooper (1980) semantics which treats EECs as existential predicates which take quantifiers over particulars as arguments (36), we should instead understand EECs to be equivalent to some predicate to be instantiated which takes a property (or non-particular denotation) as its argument (37): (35)There was snow (36)λP[∃x[snow(x) ∧ P(x)]( λy [exists(y)] = [∃x[snow(x) ∧ exists(x)] (37)a. is-instantiated(^λx[snow(x)]) b. λP[∃x[P(x)]( λy [snow(y)] = [∃x[snow(x)] The semantics in (36) and (37) result in the same interpretation but they do it in a slightly different way. (36) actually says that there is some snow that exists in the sense that it quantifies over particulars. In (37) however, the predicate just says some class of stuff snow is instantiated, such that if any example of this stuff exists then the sentence is true. (37) therefore quantifies over non-‐particulars in a place at a time and the truth of sentence is determined when one can verify some 126 instantiation of that class of stuff exists (in this case snow). The predicate is- instantiated has a sortal restriction that requires that its argument denote (or quantify over) non-‐particulars 39 . McNally suggests that the DE is really in fact two restrictions. The strong quantification DPs are ruled out by a sortal requirement imposed by the predicate. The sortal requirement is that any post-‐verbal NP must range over non-‐particulars which means that necessarily quantificational DPs are only licit in cases where they quantify over kinds (McNally 1998: 358-‐9). The second restriction is a pragmatic one that restricts the actual definites (Proper names and pronouns etc…). She argues that because Existential Construction introduce new elements into the discourse, that definites and proper names are ruled out a priori for pragmatic reasons following Zucchi (1995). Section 6: A Small-‐Clause Approach to the DE—Rethinking Presupposition and Theticity As briefly discussed earlier, there is not simply one form of predication, rather there are two types of predicational structures: (i) thetic predication is, “a single presentation is made about the existence of some event of entity” (Basilico 1997: 281) and (ii) categorical predication involves positing the existence of some entity and then assigning it a property. Kuroda (1972) succinctly describes the nature of these two-‐forms or predication: 39 This type of sortal restriction is well attested, McNally (1998; 356) suggests that “gather” exhibits a similar predicate sortal restriction by selecting plural predicates, or predicates that quantify over groups (i) {*A girl/the girls/a crowd} gathered outside 127 this theory assumes, unlike either traditional or modern logic, that there are two different fundamental types of judgments, the categorical and the thetic. Of these, only the former corresponds to traditional paradigm of subject-‐predicate, while the latter represents simply the recognition or rejection of a material judgment. Moreover, the categorical judgment is assumed to consist of two acts, one the act of recognition of that which is to be made the subject, and the other, the act of affirming or denying what is expressed by the predicate about the subject. (Kuroda 1972: 154) Crucially, the argument goes, categorical predication requires a subject that is “presupposed”. That is, in order to ascribe a property to something by way of a predicate you must first assert that it exists as a subject. In the section describing the Basilico (1997) proposal, I mentioned that there is a relationship between the observed restriction concerning IL/SL predicates in ECs and the nature of thetic vs. categorical predication. Specifically, it was argued that only IL predicates require categorical judgments. Interestingly, this is totally consistent with Milsark’s observation: (38) Properties may only be predicated of strong NPs (= IL predicates must have strong subjects) (Ladusaw 2000) Indeed, Ladusaw (2000) offers a derivation of Milsark’s generalization using the notion of thetic predication: (39) a. Properties cannot be the basis of a thetic judgment, only part of the basis of a categorical judgment b. The subject of the basis of a categorical judgment must be an object, not a description of an object c. Strong construals are either object-‐denoting or quantificational d. Therefore, the subject of an IL predicate must have a strong construal 128 I will now first briefly present a proposed explanation rooted in the notion of thetic vs. categorical predication and presupposition (Zucchi 1995), then I will present a modified semantics of the EC (a la (Keenan 2003)) that avoids some of the pitfalls of the purely thetic/categorical predication based analysis. Zucchi (1995) offers an explanation of the DE that is based on the independent status of the coda of the sentence (the predicate of the post-‐copular NP) and a semantico-‐pragmatic ‘Felicity Condition’ concerning the presuppositionality on the postcopular NP. Zucchi (1995) adopts an EC structure as in (40) and the associated Coda Condition: (40) a. There [VP be, Det+N, Coda] b. Coda Condition-‐ The coda provides the domain of evaluation of there-‐ sentences The DE is ostensibly explained by Zucchi’s three rules summarized briefly by Ladusaw (2000) in (41): (41) a. DPs unacceptable in there-‐sentences are those built from presuppositional Dets (det+N’ is felicitous only in contexts whose common ground contains the proposition that the denotation of the N’ is not empty) b. the interpretation of there-‐sentences incorporated the coda property into the scope of the Det in the postverbal DP, so There is every student in the garden presupposes that the denotation of every student in the garden is not empty c. the felicity conditions of there-‐sentences require that the common ground include neither the proposition that N’+coda is empty or that it is non-‐empty (195) 129 The DE is accounted for in (41) by resorting to some discourse model where information is mutually shared or new such as to impose information structure requirements on sentences. There are however, at least three problems with the Zucchi proposal and others like it. The first problem is learnability, the conditions in (41a-‐c) don’t tell us what is allowed in an EC (what property licit postcopular NPs share), rather it just tells us what is excluded. This means that children would have to learn EC structures by some kind of elsewhere-‐condition. The second problem with this approach is that the notion of presupposition wrongly predicts that certain quantifiers should be illicit in ECs when in fact they are fine. One test of presuppositionality proposed by Keenan (2003) is that presupposed material is implied by the original sentence as well as its negation. If that is true, then just n (and only n) are presuppositional since both (42a-‐b) imply (42c) and (43a-‐b) imply (43c): (42)a. He found just two mistakes in the proof b. He didn’t find just two mistakes in the proof c. There were at least two mistakes in the proof (43)a. Only John ran b. Not only John ran c. John ran (adapted from Keenan 2003: 196) According to Zucchi (1995) the presuppositional nature of the determiners just n/only n should mean that they are bad in ECs, however, (44) illustrates that this is not the case and these determiners are in fact fine: 130 (44) a. There were just two mistakes in your paper b. Aren’t there only two undergraduates in your logic class? Moreover, as Keenan (2003) points out, the presuppositionality requirement is too restrictive from a discourse standpoint as well. That is, existential sentences do not just question existence, they can question/comment on cardinality as in the following example: (45) A: Wow! There’s gonna be hundreds of people at my lecture on existential constructions. B: No, there’ll be 3 or 4, the others are just taking a shortcut to the spa. (Keenan 2003: 197) In this exchange, both speaker A and B agree on the existence of some people at the lecture, the question is just how many are going to be there. However, according to the felicity constraint proposed in (18c) which operates over discourses and shared information between speakers, this exchange should be impossible. Given these problems which stem from the formulation of rules based on presuppositionality, Keenan (2003) proposes a formal truth-‐conditional semantics that offers us a systematic way to determine appropriate EC determiners (both unary and k-‐placed) while still maintaining the insights of thetic-‐predication and the coda-‐constraint. Keenan begins with the observation from Milsark (1977) that cardinal Det(erminer)s are the most common in EC sentences, and thus makes the following generalization: 131 (46) Det Generalization Whether a DP built from a Det and an appropriate number of nominals is natural in there-‐sentences is decided by the choice of det 40 (Keenan 2003:189) The task then becomes to define a Det as a generalized quantifier (GQ) over a domain E of objects, such that we create functions from PE, properties of individuals, to {True, False}. To generalize, GQE, X is the set of functions that map PE into set X (if X = [true, false] then just GQE): (47) a. A Function D from PE into GQE, X is cardinal iff for all A, A’, B, B’ ⊆ E if ⏐A ∩ B⏐= ⏐A’ ∩ B’⏐then DAB = DA’B’ b. A det d is cardinal iff in all models with domain E, d denotes a cardinal function c. Let the set of all cardinal dets = DetTHERE 41 (190) However, given the appropriateness of sentences such as (48) we must also say something about Boolean combinations of Dets which is the theorem in (49): (48) There at least two dogs and more than five cats in the yard 40 This generalization does not hold true for partitives and kind-NPs such as: (i) There were both kinds of tomatoes on the counter (ii) *There were both tomatoes on the counter These Kind-NP, partitive, and psudeo-partitive cases will be addressed in another of the dissertation. 41 The definition of k-placed Dets (e.g. more...than...) is: (i) A map D from k-tuples of sets into GQ E, X is cardinal iff DAB = DA’B’ whenever ⏐A i ∩ B⏐= ⏐A i ’ ∩ B’⏐, all 1≤i≤k (ii) “a cardinal Det k decides truth by checking the cardinality of the intersection of the predicate property with each of the k noun properties” (Keenan 2003: 192) 132 (49) Boolean Theorem: Boolean functions of cardinal functions are themselves cardinal. That is, (D ∧ D’) (D ∨ D’) and ¬D are each cardinal whenever D and D’ are. With this groundwork laid out, we are ready to establish a definition of DetTHERE that will properly delineate those determiners that are acceptable in ECs based a definable semantic characteristic of the lexical determiner itself. The Coda Condition from Zucchi (1995) served to limit the evaluation of truth of the entire there sentence to the domain of the denotation of the coda. In the Keenan (2003) system, the Coda Condition is recast as a structural condition on the semantics of the Det vis-‐à-‐vis the coda, in essence creating thetic predication since the there sentences are limited to statement about the entirety of the predicate. That is, “[the] idea is that DetTHEREs are ones that decide by limiting their universe to the coda set” (Keenan 2003: 199). The semantic characteristic that allows us to define the set DetTHERE of cardinal functions is (distributed) conservativity. A D is conservative with respect to an argument if that argument serves to define the local universe of the D. In this sense, we can relativize conservativity to of a D to specific arguments. Thus, we can define a D as ‘conservative on the first argument’ or cons1 as: (50) A map D from PE into GQE, X is cons1 iff A ∩ B = A ∩ B’ ⇒ DAB = DAB’ for all A, B, B’ ⊆ E What (50) means is that the D will use the first argument A (a noun) as its local universe to determine whether D(A) maps a B to True, the only thing that must be 133 evaluated are Bs in A. Keenan offers the concrete example of Most poets daydream which is conservative on the first argument because to determine the truth of this sentence, we only need to know about poets and whether or not they daydream, we don’t need to know anything about daydreamers who are not poets. Hence, the first argument poets constitutes the local universe for evaluation of the truth of the sentence. Put simply, “to say that det is conservative on any given argument is to say that the value that it assigns is determined by that argument and its intersections with other arguments” (Keenan 2003: 200). We can define a D that is conservative on the second argument cons2 as follows: (51) A map D from PE into GQE, X is cons2 iff A ∩ B = A’ ∩ B ⇒ DAB = DA’B for all A, A’, B ⊆ E This definition in (51) simply says that the argument B is what is used to create the local universe for the evaluation of the truth conditions. In the case, of an existential sentence, the second argument is taken to be the predicate XP of the postcopular NP. Importantly, the definitions of cardinality and cons1 /cons2 allow us to derive the conclusion that: (52) For all maps D from PE into GQE, X D is cardinal ⇒ D is cons1 and cons2 Given (52) we can make the conclusion that DetTHERE is the set of dets that are lexically cons2. Take the sentence in (53): (53) a. *There are all students in the park b. *There are most students in the park 134 In (53a), if it were the case that D all ∈ DetTHERE then it must be cons2. That means we should only have to know about Ns (in this case students) in the park. However, in order to evaluate the truth of D all applied to the N, we must know about the people not in the park such that we can verify they are not students. In (53b), we need to now about the cardinality of students outside the park such that it is smaller than the cardinality of students in the park in order to judge the sentence true. Basically, these Dets are ruled out because they do not obey the coda condition. Another corollary to these definitions is that we have a test to determine which determiners are existential determiners. A Det is in the set of DetTHERE iff [Det NP VP] = [Det NP which VP exist(s)] = 1. To apply the test, take a bare numeral three which is in DetTHERE because: three cats are in the tree iff three cats which are in the tree exist. Now assume a world with 5 cats and three of them are in a tree, if we apply the test to every then: [Every cat is in the tree] = 0 and [Every cat which is in the tree exists] = 1 (Partee 2006). Lastly, it is important also to compare the role of there in both the Basilico (1997) and the Keenan (2003) proposals. It would be hard to find two more different proposals concerning the role of there in the syntax and the semantics. For Basilico (1997) there is a crucial driving force in the syntactic derivation of the sentence (there is [+specific], agrees with NP, then checks Tense, then moves to [Spec, RefP]...). There, for Basilico, also fulfills a semantic role in that it licenses the SC thus taking it out of the scope of Tense which, according to him is why IL predicates are disallowed (although, recall that Ladusaw (2000) derives the IL 135 predicate restriction by using thetic predication in the postcopular clause with no dependence on Tense). It also heads the DP that creates the framework for establishing the crucial relationship of thetic predication. Indeed, for Basilico (1997) the thetic predication relation is between there and [SC NP XP]. For Keenan (2003) however, there is truly a dummy expletive; it literally serves no semantic function. From a syntactic standpoint, it would appear that it just checks EPP on Tense (agreement phenomena notwithstanding). Perhaps the most striking difference between the two proposals is that, for Keenan (2003) the thetic predication relation is between the postcopular DP and the Coda XP—there is not involved in any predication relation at all. Section 6.1: Deriving the DE from Situation-‐Descriptive Predication Taking stock of where we are in the derivation of the DE, Keenan (2003) adopts an approach that suggests that semantic restrictions on the classes of Determiners account for the unavailability of strong quantified DPs in EECs. Specifically, that Dets which determine their domain evaluation with reference to the coda in the EC are licit. This characterization is argued to be superior to other approaches because: it structurally encodes the primacy of the coda in the interpretation of ECs; it offers a truth-‐conditional semantic calculation for ECs rather than vague pragmatic calculation; and it makes the right predictions with respect to Boolean compounds of cardinal determiners in ECs. McNally (1998), 136 however, suggests that an exclusively Determiner based account of the DE fails to account for the fact that strong-‐quantified kind DPs are licit in ECs: (54) a. *There was every candy in the bowl b. There was every kind of candy in the bowl Therefore, for McNally (1998) what is crucial is to adopt a semantics that ensures that the post-‐verbal DP denotes a “non-‐particular” (e.g. property). Whereas Keenan (2003) suggests the DE is a single effect (i.e. it is the result of violating a single restriction), McNally (1998) suggests that the DE is really two effects (i.e. it is the result of violating on of the two simultaneous restrictions on ECs). For McNally (1998), the restriction on strong quantified DPs (all, both, every, etc…, but not the) is a result of a sortal restriction imposed by the existential predicate There BE. The definites (the, pronouns, proper names etc…) are ruled out by a separate pragmatic restriction which precludes presuppositional DPs in ECs (presumably because ECs exist to introduce new referents into the discourse). While the McNally (1998) proposal accounts for the kind DP data, there are two major drawbacks to this proposal: (a) It commits us to a non-‐small clause based syntax which can not account for the data in Chapter 2, and (b) it invokes a vague and empirically false pragmatic restriction about presuppositionality and discourse-‐new/hearer-‐new information that seems somewhat ad hoc as an explanation for non-‐grammaticality. While both of these proposals have their pros and cons, it is clear that an effective solution to the DE in ECs must meet the all the following parameters: 137 (55) A Successful DE solution will… a. Avoid an exclusively Determiner central account because it cant explain Kind DP Data b. Maintain a small-‐clause syntax to account for the agreement/’occasional’ data c. Ensure ‘property-‐denoting’ post verbal EC subjects (most of the time) d. Capture the intuitions of thetic predicational structures e. Replace pragmatic presupposition/felicity conditions with a truth-‐conditional version of the coda condition Adopting the structure in (34) above offers a structural account of the DE that meets the five criteria articulated in (55 a-‐e). The structural solution to the problem is a result of the critical addition to the Roy (2005) proposal at the end of Section 2 in this chapter. Specifically, it was argued that, in the case of copulars, the specifier of the functional structure (FP) that is the complement of Pred hosts a pro that is bound by the base-‐generated DP in Spec PredP. In the case of existentials, the DP is generated in Spec FP and then either moves from Spec FP to Spec PredP, or alternatively, FP moves to Spec PredP 138 (56) (Modified) Roy (2005) Situation Descriptive Copular PredP 5 cops<e>i Pr’ Pr PP (= FP) proi P’ at the door (57) Existential Plural/Count Agreement (Entity Central Reading) PredP 5 cops<e,t>i Pr’ Pr PP (=FP) t<e>i P’ at the door (58) Existential Singular/Mass Agreement (Event Central Reading) PredP PPi (=FP) Pr’ Pr ti The key difference between the copular structure in (56) and the existential structures in (37-‐8) is that since the DP is base generated in the lower FP, it is subsequently moved to a higher position that results in type-‐lifting. Whereas McNally (1998) used a sortal restriction imposed by the Predicate “There BE” to ensure that the post-‐verbal NP was property-‐denoting, the structure in (57) ensures that the post-‐verbal DP will be property-‐denoting without reference to some higher existential predicate. Deriving the property-‐denoting post verbal subject in the EC 139 as in (57) accounts for burden (55c) since it ensures type lifting but it also solves burden (55b) because (56) is derived from a situation-‐descriptive copular which ensures that there is an underlying PredP structure. However, something must be said of (58). If type-‐lifting is the mechanism that ensures the property-‐denoting post-‐verbal DP, then it is unclear how such type-‐ lifting could occur in (58) since the subject doesn’t move in a way that would type-‐ lift it; rather, in (58) the entire PP is moved of which the post-‐verbal DP is the subject. Indeed, McNally (1998) imposes a strict sortal restriction on the PVDP by way of an existential predicate. It is the prediction of her system then, that any non-‐ property denoting DP should be illicit in ECs as a violation of the sortal restriction. Her arguments for the sortal restriction mechanism over a Determiner restriction mechanism stem from the putative inability of exclusively Determiner mechanisms to account for the Kind DP data in (54). Essentially, she argues that if we exclusively focus on determiner restrictions then we cant explain why “every N” is bad but “every kind of N” is good because they both “every”. While this argument is reasonable, the mechanism she uses to rectify it (i.e. the sortal restriction of the existential predicate) is too strong in the opposite direction. McNally (1998) takes an approach to solving the DE that focuses exclusively on the denotation of the DP as whole such that regardless of the determiner, as long as the DP is property-‐denoting it will not violate the sortal restriction of the predicate. While Keenan (2003) and McNally (1998) take mutually exclusive approaches to deriving the DE, the mutual exclusivity of the approaches is in fact an illusion perpetrated by the assumption that the DE must be 140 the result of a single point of variance: either the determiner or the denotation of the DP. If we adopt the proposal here, both of these proposals turn out to contribute to a comprehensive explanation. In (57) the type-‐lifted DP satisfies McNally’s claim that the DP should be property-‐denoting albeit in a completely different way than she envisioned. (58), however, violates McNally’s strong restriction on non-‐property denoting post-‐ verbal subjects. However, if we carefully analyze the data about Kind DPs, it becomes clear that the justification for the total prohibition on non-‐property denoting DPs is misguided. (59)Mini-‐Discourse I Speaker A: Wow, there seem to be a lot of people trying to get into that bar Speaker B: Yeah, there seem to be five more customers in line every 2 minutes Speaker B’: Yeah, there seems to be five more customers in line every 2 minutes (60) Mini-‐Discourse II Speaker A: At the world bank, there are all forms of currency available all the time Speaker B: Well at my bank, there seem to be all forms of currency available most of the time Speaker B’: *Well at my bank, there seems to be all forms of currency available most of the time It is the contention of this paper that the Speaker B’ variants in Mini-‐Discourses I and II (which are structures as in (58)) are not property denoting post-‐verbal subjects and more importantly, property-‐denoting DPs are actually bad in structure (58). In (59), in the sentence of Speaker B’, there is no sensible way to interpret ‘five 141 more customers’ as a property 42 . Here, the post-‐verbal subject is interpreted as an amount and the semantics of Keenan (2003) indicate that as a cardinal Det ‘five’ is lexically cons2 and therefore all we have to do is evaluate the line every 2 minutes, and if there are five more customers in it, then the truth conditions have been met. For McNally, it is unclear how one would compute the semantics for this sentence. What is even more interesting, however, is the data in (60). When ‘all forms of currency’ is in the non-‐agreeing event-‐central reading structure, the result is ungrammatical. The important thing to note for now, is that the sortal restriction imposed by McNally’s (1998) predicate on the post-‐verbal subject is too strong and a determiner based approach is necessary to explain the full variability of the agreement data with respect to kinds, partitives, and psuedopartitives in existentials. Maintaining the determiner restriction semantics also entails that we will have a syntactically and semantically coherent and precise instantiation of the Coda Condition instead of a pragmatic restriction—for those still keeping score, that satisfies burden (55e). Lastly, both the coda condition and the fact that the structures in the current proposal are derivatives of the situation-‐descriptive copular sentences of Roy (2005) ensure that the thetic nature of the predicational structure is preserved, hence satisfying the last burden (55d). Section 7: Conclusion Ultimately, the system proposed here suggests that existentials are one variation of the situation-descriptive predicational structure. In the copular case, the 42 Perhaps one could try some way to interpret it is a property of the line, but that would be inconsistent with the proposal in McNally (1998) since for her the predicate is irrelevant in determining the DE 142 subject is base generated in [Spec PredP] and binds a pro in [Spec XP]. In the existential case, there is no pro and the subject is either moved from [Spec XP] to [Spec PredP] or the entire XP is moved to [Spec PredP]—these two options are the reason we observe the entity-‐level and event central interpretations with different agreement manifestations. The DE we observe is a result of a coda condition that is the manifestation of restriction on the type of determiner that can appear as the subject of PredP—namely, the Determiner must be weak (cons2). The restriction on IL predicates in this system is in reality two restrictions: (1) nominal predicates (including NumP and ClP predicates) are ruled out a priori since they are never in situation-descriptive predicational structures, and (2) predicates traditionally referred to as IL predicates are known to require strong subjects independently (Carlson 1977 a.o.) which means that since strong subjects are ruled out under the Coda Condition (Keenan 2003) these predicates will not be licit either. Note also, the fact that the IL-‐predicate restriction is no longer a unified phenomenon is not a deficiency of the proposal, rather it is an explanatory advantage in that it militates the IL/SL distinction generally which has been shown to be empirically insufficient (Roy 2005). 143 Chapter 4: Beyond the English Existential Construction: The Mandarin Existential Construction and Predication Section 1: Introduction The last chapter addressed the DE and *IL restriction in English ECs by abandoning the IL/SL distinction and replacing it with Roy’s (2005) cross-‐ linguistically motivated trifurcation of predicational structures: defining, situation- descriptive, and characterizing. These distinctions are the result of the presence or absence of count structure in the predicate XP. The presence of count structure means that the predicate is non-‐intense (i.e. discontinuous; not necessarily true of all sub-‐intervals of the event), and only nominal predicates can have count structure. We observed that the predicational restrictions and the DE in the EEC fall out naturally from this distinction if we understand that the EEC is derived only from Situation-‐Descriptive predicational structures (i.e. predicates that are intense/mass/-‐count). This finding makes sense since existential sentences affirm or disaffirm the existence of some entity in a particular location at a certain time— that is, they describe situations. Roy’s three-‐way distinction between defining, situation-‐descriptive, and characterizing sentences in English/French is: (1) Predicational copular sentences-‐ English/French ‘property-‐assigning’ sentences situation-descriptive sentences defining sentences characterizing sentences Roy 2005: 90 144 But we can restate (1) in two ways: a syntactic formulation as in (2) or a semantic formulation as in (3): (2) Predicational copular sentences-‐ English/French:Syntax Count/Nominal predicates Mass/Non-‐Nominal Predicates NumP ClP AP/PP (3) Predicational copular sentences-‐ English/French:Semantic Count/Non-‐Intense predicates Mass/Intense Predicates Defining Characterizing Sit-‐Descriptive Given that the Roy (2005) system is based on the mass/count distinction in the predicate, it is interesting to examine the proposal in the context of languages that vary typologically in precisely how they encode the mass/count distinction. Mandarin Chinese as a “classifier language” is problematic for the predictions made by the Roy system and the current extension of it to existential constructions. For instance, in the case of copulars, The Roy ontology of nominal predicates is: (4) Roy Ontology of Nominal Predicates -‐[Num-‐Cl-‐NP], [Num NP] [Cl NP] In Mandarin, there exists a very different ontology which includes bare NPs which are excluded in (4) and disallows [Num NP] prima facie. 145 (5) Chinese ontology of nominal predicates -‐-‐[(Num)-‐Cl-‐NP], [NP] (6) Chinese a. Zhansang shi yi ge yihseng/shagua John BE 1 CL doctor/fool “John is a doctor/fool” b. Zhansang shi ge yihseng/shagua John BE CL doctor/fool “John is a doctor/fool” c. Zhansang shi yihseng/shagua John BE doctor/fool “John is a doctor/fool” d. *Zhansang shi yi yihseng/shagua John BE 1 doctor/fool “John is a doctor/fool” Moreover, when comparing Existential Constructions (ECs) in English (There- constructions) and Mandarin (you existentials), it is interesting to note the striking similarities between the two: both have post-‐verbal subjects, both exhibit the Definiteness Effect and strong quantification restrictions, both seem to contain a small clause, both have interesting relationships to copular analogues, and both require a locative of some sort: (7) a. There is [SC *the/*every/a manp-‐v subject *(in the garden)] b. you [*zhe-‐ge/*mei-‐ge/ yi-‐ge ren *(zai huayuan-‐li)] have this-‐Cl every-‐Cl one-‐Cl person at garden-‐in 146 However, there is also a conspicuous difference between the two structures: no *Individual (IL)-‐predicate restriction in Mandarin: (8) *IL-‐predicate Restriction in English a. There is a student in the yard b. *There is a student (is a) doctor c. you yi-‐ge xuesheng zai huayuan-‐li have one-‐Cl student at garden-‐in d. you yi-‐ge xuesheng shi yisheng have one-‐Cl student be doctor The overwhelming similarity between the two constructions and the specific point of variation are of significant interest to the proposal at hand. The proposal advocated in the previous chapters relies on the insight that the mass/count properties of predicates are the key to understanding the observed facts in the EEC and copulars. That is, by extending the Roy system, we hypothesize that existential constructions are derived from mass/situation-descriptive predicates. Also, in this system, nominal predicates must be dominated by functional count structure to be interpretable, so there is no possibility of something like “situation-descriptive nominals.” The data in (8d) is therefore surprising since it shows that Mandarin allows nominal predicates in existentials. One more related challenge to the current extension of the Roy (2005) proposal is that Mandarin allows the possibility of some count predicates in existential codas (8b) but not others (8c), while the last chapter predicted only the presence of mass predicates: 147 (9) a. you xuesheng shi yisheng have student be doctor b. you xuesheng shi ge yisheng have student be CL doctor c. *you xuesheng shi yi ge yisheng have student be one CL doctor Specifically, this point of variation between the EEC and the MEC is an insight into a crucial typological distinction between Mandarin and English in the nominal domain. The different ontological nature of the nominals in Mandarin and English and the strategies and structures they adopt to encode plurality and mass/count properties will force us to operationalize the important insights in the Roy (2005) system in a very different way in Sinitic and other so-‐called Classifier languages. The representations of copular sentence types in (1-‐3), the syntactic instantiations of the various predicate types, and the semantic properties associated with those instantiations are the result of a few key structural properties of English, French, and the other languages for which this trifurcation makes true predictions (including Spanish and Russian to name a few). The previous chapter of this dissertation extended the Roy system to explain English existential constructions by proving: (a) that existential constructions and copular constructions are related and (b) that the restrictions observed in existential constructions are a direct result of existential constructions being derived from Situation-‐Descriptive predicational structures. Namely, this system assumes the following: 148 (10) Axioms underlying (1-‐3) a. bare lexical categories (e.g. NP) cannot be direct predicates b. Non-‐D headed nominals cannot be arguments c. Only Nominals can have count structure d. Mass/Count correlates to +/-‐Predicate Intensity Therefore, trying to use the Roy (2005) system to account for Mandarin, we would predict that the sentence-‐type, syntactic formulation, and the semantic formulation of the types of non-‐verbal predication in Mandarin might be as in (11): (11) Roy Model for-‐ Mandarin Count Mass NumP ClP AP/PP/Bare-NP There are three undesirable results of the model in (11). The first is that when compared to the representation in (2), it is clear and somewhat surprising that the representation in (11) breaks the categorical homogeneity of nominality with respect to predicate distribution. That is, nominals no longer constitute a natural group of count/non-‐intense predicates. The second problem is that if existentials are derived from mass predicates, then (11) makes the false prediction that Bare NPs (BNP) should be the only type of nominal coda that is licit in MECs (viz. (9b)). Another way of conceptualizing this problem is to say that the 149 Mandarin EC (and copular construction as we shall see) has the possibility of a bare mass noun that is property-‐denoting which violates axiom (10a). If we are to extend the Roy (2005) system to Mandarin, we will have to make some serious adjustments to the axioms in (10). Specifically, we will have to abandon axioms (10a) and (10b). Indeed, Mandarin bare NPs can in fact be predicates (Wei 2007), and Non-‐D headed NumPs can be arguments (Li 1999). It is important to note, however, that while the argument/predicate status of various levels of nominal-‐structural representation can vary by language, the key theoretical insight that (10c) and (10d) shed on the interrelationship between mass/count and Predicate Intensity is taken to be linguistically universal and is maintained in the description of Mandarin non-‐verbal predication. In order to understand predication and existential constructions from a more cross linguistic and typologically diverse perspective the rest of the chapter will first develop a formal review of the properties of Chinese predicate nominals and the implications of the classifier system. While the Roy (2005) system suggests that [Num-‐Cl-‐NP] is a Defining predicate, [CL-‐NP] is a characterizing predicate, and [NP] is not available as a predicate, Mandarin allows all three of these options as predicates however they do not always obtain the same interpretations as predicted by Roy (2005). The next section of the chapter will explore the consequences of the Mandarin nominal data for the Roy (2005) system and present a typologically appropriate ontology of predicates for Mandarin. Finally, the last section will 150 address the Mandarin EC in detail and discuss the similarities and differences between it and the English EC discussed in previous chapters. Section 2: Predicate Nominals in Mandarin It has long been observed that Mandarin and English differ typologically with respect to how their nominals are individuated and counted. Chinese is a classic example of a numeral-‐classifier language in which direct numeral modification/quantification of nouns is prohibited and such a relationship must be established by means of an intervening classifier. It is interesting to note that while a classifier must be present for a numeral to appear, it is not the case that a numeral must be present for a classifier to occur (12). English on the other hand is a language with plural morphology and (in)definite articles: (12) Chinese a. yi ge ren 1 CL people “one/a person” b. san ben shu 3 CL book “three books” c. ge ren CL person “a person” d. ben shu “a book” e. *yi ren 1 people “1 person” 151 f. *san shu three book “three books” That there are two types of classifiers in Mandarin is well-‐documented. The first type are the kind which subcategorize objects based on certain intrinsic properties of the noun referents such as shape or some other (potentially abstract) property (Allan 1977); in the literature, these have been called: the sortal classifiers (Senft 2000), count classifiers (Cheng & Sybesma 1998) or individual classifiers (Chao 1968). For instance, bi ‘pen’ in (13a) must occur with zhi, a classifier for elongated objects, not pian, a classifier for flat objects. In (13b) wei ‘tail’ is an intrinsic property of the head noun yu ‘fish’. Ke, a classifier for round small object, hence cannot be used with fish. Sortal classifiers combine with specific nominal based on the essential property of the head noun (Her & Hsieh 2010). (13) a. yi zhi/*pian bi one CL pen “a pen” b. san wei/*ke yu three CL fish “three fish” The other type of classifier is the mesural classifiers (Senft 2000), also called “measure words” or “massifiers” in the literature (Tai & Wang 1990; Cheng & Sybesma 1998). These elements define a unit of measure and individuate masses in terms of quantit(ies) (Liang 2006; C. Hsieh 2009; S. Yang 2005). In (16a), what is being counted are boxes, and in (14b) we are counting kilograms. This type of 152 classifier contributes important semantic information to the nominal phrase unlike the sortal classifiers. (14) a. yi xiang juzi one box orange “a box of oranges” b. liang jin mi two kilogram rice “two kilograms of rice” (Liu 2012: ms) While mesural classifiers exists prevalently in the world languages (e.g. a box of candy, two bottles of water, two pounds of meat), sortal classifiers are found only in the so-‐called classifier languages (Tai 1994). Chierchia (1998b) offers a systematic typology of nominals that at first glance seems to both accommodate the data and offer explanatory insight into the behavior of Mandarin nominals. Chierchia (1998b) suggests that languages have some sort of default setting for [+/-‐ arg] and [+/-‐ pred] parameters for nominals. While all languages have the ability to use NPs as both arguments and predicates by way of type shifting, it is not necessarily the case that all NPs must start as predicates in all languages. The conclusion is that if a language is [-‐arg, +pred] then the category NP will always project something that is a predicate of type <e,t> unless it is type-‐shifted to denoting the atomic individuals of which that predicate is true. Ultimately, this is the case for English and Romance languages although there are still variations across those typological groups that will not be expounded on here. Mandarin on the other hand is taken to have the opposite valence for the 153 nominal parameters with [+arg, -‐pred]. Thus, in the case of Mandarin the NP generally denotes kinds of type <e>. Furthermore, because of the nature of the type-‐ shifting operation, the inevitable outcome is that the output is mass. Ultimately, however, Borer (2005) established that contra Chierchia (1998b), the mass denotation of nominals is true cross linguistically. The differences in the language typology are a result of the way in which the nouns combine with the count structure. That is, both types of languages have mass denotations as default interpretations, but mass is simply the absence of count. Each language has count structure, specifically a Cl(assifier)P which individuates the mass denotation of bare nouns and makes available the atomic units (or the singleton subsets from the join semi-‐lattice) and a Num(ber)P which counts the output of the ClP (i.e. counts the units/atoms of stuff it individuates). Together ClP and NumP are the components of ‘count structure’ and they may occur independently or together in different ways in different languages depending on their morphological status (‘free’ as in Mandarin Classifier or ‘inflectional/bound’ as in the English plural Morpheme) 43 . Following Roy (2005) I will also adopt this view of nominal denotation and count structure. Roy (2005) however formulates her trifurcation of predicational structures based on evidence from languages that have both plural morphology and (in)definite articles as well morphological restrictions that require functional structure. This section will explore the properties of a numeral-‐classifier type language which will require a reformulation of the predictions in Roy (2005). 43 The English indefinite article is taken to be a morphologically ‘free’ element that can appear in ClP as well 154 Section 2.1: Mandarin Nominals and Bare Nouns While the Borer (2005) system makes the claim that all nouns have a default mass interpretation when they are bare, that does not mean that bare nouns are licit themselves as arguments or predicates in all languages. Indeed, this is a distinction clearly illustrated by predicate nominals in English and Mandarin: (15) English a. John is a/the doctor/fool b. *John is doctor/fool (16) Chinese a. Zhansang shi yi ge yihseng/shagua John BE 1 CL doctor/fool “John is a doctor/fool” b. Zhansang shi ge yihseng/shagua John BE CL doctor/fool “John is a doctor/fool” c. Zhansang shi yihseng/shagua John BE doctor/fool “John is a doctor/fool” Rullman and You (2003) adopt an analysis of Mandarin nominals wherein they are assumed to have General Number (Corbett 2000) which means that Mandarin bare nouns are simply unspecified for number and therefore they are neither plural nor singular: 155 (17)a. Zuotian wo mai le shu Yesterday I buy ASP book Yesterday I bought one or more books b. Zuotian wo mai le yi ben shu Yesterday I buy ASP one CL book Yesterday I bought one book c. Zuotian wo mai le yixie shu Yesterday I buy ASP some books Yesterday I bought some books In (17a) the noun shu is unspecified for number, while in (17b-‐c) the NP is interpreted as either singular (17b) 44 or plural unambiguously (17c). To account for this, Rullman and You (2003) assume that the domain of type <e> entities is structured as a free atomic join semi-‐lattice then both singular atoms and their sums constitute members of the set of pluralities with individual atoms being singleton subsets (Link 1983, Chierchia 1984, 1998a, 1998b, a.o.). They represent the denotation of a Mandarin NP as: (18) Denotation of Mandarin nominals Rullman and You (2003: 5) The inherent plurality encoded in the denotation of Mandarin Bare Nouns with GENERAL NUMBER is best exhibited by the fact that they behave in precisely the 44 It is important to note that in (1b) a stressed yi (one) results in the English counterpart of ‘I bought one book’ and an unstressed yi is akin to English ‘I bought a book’ 156 same way as English bare plurals. That is, in English we know that bare plurals always take narrow scope and in Mandarin there appear to be the same scopal limitations. Rullman and You (2003) following Carlson (1977) illustrate the parallelism in (19) and (20): (19)a. Minnie wishes to talk with a young psychiatrist (wide or narrow) b. Minnie wishes to talk with young psychiatrists (narrow only) (20) a. Mini xiang gen yi ge nianqing de xinlixuejia tantan (wide/narrow) Minnie wish with one CL young MOD psychiatrist talk “Minnie wishes to talk with a young psychiatrist” b. Mini xiang gen nianqing de xinlixuejia tantan (wide or narrow) Minnie wish with young MOD psychiatrist talk i. Minnie wishes to talk with young psychiatrists (narrow scope) ii. Minnie wishes to talk with the young psychiatrists (definite) Rullman and You (2003:13) To illustrate the simultaneously singular nature of the Mandarin bare nouns, we must turn to evidence from binding. While Mandarin generally lacks plural morphology, it does have a morphologically realized plural/singular distinction (-men) in the domain of pronouns that are referring to human antecedents: (21) a. Zhansang hen chao. Ta/*Tamen cong bu bizui John very loud. He/they never not open-‐mouth “John is loud. He/*they never shuts up” b. Zhansang he Lisi hen chao. *Ta/Tamen cong bu bizui John and Lisa very loud. *He/they never not open-‐mouth “John and Lisa are loud. *He/they never shut up” 157 When we use this diagnostic of plurality on Mandarin bare nouns an interesting pattern obtains: (22)a. Zuotian wo yudao le tongshi. Wo quing ta/tamen chifan le. 45 Yesterday I met ASP colleague. I invite him/her/them eat ASP “Yesterday I met one or more colleagues. I invited him/her/them to dinner” b. Zuotian wo yudao le yi ge tongshi. Wo quing ta/*tamen chifan le. Yesterday I met ASP 1 CL colleague. I invite him/her/*them eat ASP “Yesterday I met a colleague. I invited him/her/*them to dinner” c. Zuotian wo yudao le yixie tongshi. Wo quing *ta/tamen chifan le. Yesterday I met ASP some colleague. I invite *him/*her/them eat ASP “Yesterday I met some colleagues. I invited *him/*her/them to dinner” Rullman and You (2003:18-‐9) In (22a) the fact that both the plural and singular pronouns are licit is taken to be evidence that the denotation of the bare noun must be composed of a join semi-‐ lattice that includes both atoms and pluralities. In (22b) only the singular pronoun is licensed as a result of the quantized NP; interestingly, it seems to be the classifier here also implies singularity since the same pattern is found even when there is no numeral element present: (23) Zuotian wo yudao le ge tongshi. Wo quing ta/*tamen chifan le. Yesterday I met ASP CL colleague. I invite him/her/*them eat ASP “Yesterday I met a colleague. I invited him/her/*them to dinner” 45 It is worth noting that the data in (7a) seem somewhat problematic for Chierchia’s analysis of English Mass nouns. That is, if they have the same denotation as Mandarin bare nouns (i.e. they have their plurality “built in”) then the following data become difficult to explain: (i) Yesterday I bought (some) furniture and I had it/*them delivered. 158 Finally, the (22c) variant requires the plural pronoun since the quantifier ‘some’ necessarily denotes pluralities in Mandarin. 46 Section 3: Mandarin Nominals in the Roy System If we recall the Roy (2005) schema for understanding the behavior of non-‐ verbal predications in copulars repeated here as (23): (24) a. Predicational copular sentences-‐ English/French ‘property-‐assigning’ sentences situation-descriptive sentences defining sentences characterizing sentences Roy 2005: 90 b. Predicational copular sentences-‐ English/French:Syntax Count/Nominal predicates Mass/Non-‐Nominal Predicates NumP ClP AP/PP It becomes clear that the system makes some very specific predictions with respect to the availability of types of predicates and their interpretations. To briefly 46 This is different than English ‘some’ which can have an indefinite singular interpretation as in ‘Every girl kissed some boy’ where the indefinite takes scope over the universal quantified DP. These inverse scope relations do not happen in Mandarin (Huang 1984). 159 summarize, in the case of nominals, defining predicates were NumPs and they denote permanent stable properties while ClPs were Characterizing predicates which denoted “eventive” properties. As of now, the system is silent on bare nouns since the system axiomatically rules them out as discussed in Section 1. At the moment, if we were going to incorporate the Mandarin bare noun predicate into the Roy system it would somewhat awkwardly be predicted to be a situation-descriptive predicate since those are the all the Mass/non-‐count predicates. Alternatively one may suggest that what appear to be bare nouns are really nouns that have moved up to CL, but this seems unlikely. This strategy is similar to what Roy argues for what appear to be French bare nouns, except instead of movement to CL, Roy suggests that there is a null CL in French. This proposal seems incredibly unlikely in Mandarin given the dependency between CL and Num in this language. Setting aside the issue of the Mandarin bare nouns for the moment, the Roy system does offer some concrete diagnostics/predictions about Characterizing vs. Defining sentences which are based on ClP and NumP predicates respectively. Since we have already demonstrated that Mandarin has these predicates in its inventory we should be able to test the system on count structure side straightforwardly. Roy suggests four ways in which the Characterizing and Defining sentences are distinct: Eventive Readings (25) a.#Paul est medecin,mais il ne pratique plus b.Paul est un medecin, mais il ne pratique plus Paul is a doctor, but he does not practice anymore Roy 2005:31 160 Who/What (26) a. Qui est Paul? -‐Paul est {un chanteur/*chanteur} Who is P P is a singer b. Qu’est Paul? -‐Paul est {*un chanteur/chanteur} What is P P is a singer Roy 2005:39 Predicative Contexts (27) Je croyais Matisse {violoniste/*un violoniste} I believed Matisse a violinist Roy 2005:40 Pronominal Subjects (28) a. Personne n’est (*un) acteur Nobody is an actor b. Quelqu’un est (*un) acteur Somebody is an actor c. Qui est (*un) acteur Who is an actor? Roy 2005:45 For the eventive readings data in (25), the interpretation is that what appears to be a bare noun in (25a) is actually a ClP with a null CL and it is ruled ungrammatical. The intuition is that characterizing sentences involve events that are more dynamic, therefore in a case where the agent is clearly no longer participating in the event, the characterizing predicate results in a mismatch with the assertion that “he no longer practices.” In (25b) the defining NumP predicate is a timeless (non-‐intense) predicate that asserts that being a doctor is a property of Paul whether or not he is engaged in the practice of medicine or retired. Looking at the data in (26), following 161 Higgins (1979), Roy notes that answer to the who question is identificational and the answer to the what question is predicational. As the data in (26) suggest, the NumP predicate is a felicitous answer to the who question and thus represents an identificational structure as compared to the ClP predicate which appears in the canonical predicational position. Again, in (27) it is observed that NumPs do not make good predicates, this time in Small-‐Clause environments. Finally, indefinite pronominal subjects be restricted by nominals with an article (i.e. NumPs) in French. When we look at the same diagnostics for characterizing and defining sentences in Mandarin the results are: (28) Eventive Readings a. Zhangsan shi yi ge yisheng, danshi ta yijing bu xing-‐yi le Zhangan be one CL doctor but he already NEG practice-‐medicine PFV “Zhangsan is a doctor, but he doesn’t practice medicine anymore.” b. Zhangsan shi ge yisheng, danshi ta yijing bu xing-‐yi le Zhangan be CL doctor but he already NEG practice-‐medicine PFV “Zhangsan is a doctor, but he doesn’t practice medicine anymore.” c. *Zhangsan shi yisheng, danshi ta yijing bu xing-‐yi le Zhangan be doctor but he already NEG practice-‐medicine PFV “Zhangsan is doctor, but he doesn’t practice medicine anymore.” d. Zhangsan shi yi ge fuqin, danshi tade erzi yijing si le Zhangsan be one CL father but his son already die PFV “Zhangsan is a father, but his son is already dead.” e. Zhangsan shi ge fuqin, danshi tade erzi yijing si le Zhangsan be CL father but his son already die PFV “Zhangsan is a father, but his son is already dead.” f. *Zhangsan shi fuqin, danshi tade erzi yijing si le Zhangsan be father but his son already die PFV “Zhangsan is a father, but his son is already dead.” 162 (29) Who Questions a. Zhansang shi shei? Zhansang be who “Who is Zhansang?” b. Zhansang shi yi ge yisheng Zhansang be one CL doctor “Zhansang is a doctor” c. Zhansang shi ge yisheng Zhansang be CL doctor “Zhansang is a doctor” d. *Zhansang shi yisheng Zhansang be doctor “Zhansang is a doctor” e. Zhangsan zuotian shi yi ge shagua Zhangsan yesterday be one CL fool “Zhangsan was a fool yesterday.” (30) Predicative Contexts a. * Wo dang Zhansang yi ge shagua I consider Zhansang 1 CL fool “I consider Z a fool” b. *Wo dang Zhansang ge shagua I consider Zhansang CL fool “I consider Zhansang a fool” c. Wo dang Zhansang shagua I consider Zhansang fool “I consider Zhansang a fool” Wei 2007: 96 (b added here) 163 (31) Pronominal Subjects a. You yi ge ren shi yisheng Have one CL person be doctor “Someone is a doctor.” b. *You yi ge ren shi yi ge yisheng Have one CL person be one Cl doctor “Someone is a doctor.” c. *You yi ge ren shi ge yisheng Have one CL person be CL doctor “Someone is a doctor.” d. Mei you ren shi yisheng NEG have person be doctor “No one is a doctor.” e. *Mei you yi ge ren shi yi ge yisheng NEG have one CL person be one Cl doctor “No one is a doctor.” f. *Mei you yi ge ren shi ge yisheng NEG have one CL person be Cl doctor “No one is a doctor.” When we take a general impressionistic look at the comparisons of the data, it seems that the Roy system is on the right track. It appears that the mass/count distinction is a crucial factor in the contexts she predicted it would be; however, upon a finer grain comparison, it seems that the defining/characterizing split along the NumP/ClP line does not bear out as predicted in Mandarin. Instead, what the data shows with incredible consistency, is that where the Roy system predicts the presence of NumP to the exclusion of ClP, Mandarin accepts both options with 164 interpretations that are virtually impossible to differentiate (and if they are different that difference is not in how “eventive” the predication is understood to be). Also, wherever the Roy system predicts ClP to the exclusion of NumP, Mandarin licenses bare nouns, which should not exist altogether in the Roy system. Starting with the Eventive Readings (28), the data are in Agreement with the French data insofar as [Num-Cl-N] is grammatical, but contra the French data, [Cl-N] is also grammatical. What appears to be a bare noun in French patterns with what is demonstrably a bare noun in Mandarin. In the who question data, we again observe the same pattern. Unfortunately, what questions do not work as a diagnostic in Mandarin: (32) (a) Zhansang shi shen me Zhansang be what “What is Zhansang?” (b) Zhansang shi (yi ge) ren Zhansang be one CL people “Zhansang is a person” (c) #/??Zhansang shi (yi ge) yisheng Zhansang be one CL doctor Cf. “Zhansang is a doctor” The what questions are infelicitous when answered with a profession so we cant generate a similar minimal pair as in the French case. Also, even when we use person/human as a felicitous answer, the construction is not sensitive to mass/count distinctions at any level from bare noun to full NumP. (29e) is an attempt to create a context similar to the one created by the what-‐question. Namely, the goal was to create a situation where a nominal predication could be eventive in 165 the sense that it was bounded and participatory/volitional. The data in (29e) also pattern with what is predicted for what-‐questions in the Roy system in the sense that full NumPs are bad there, but we find no verification for the ClP being any better. The data in (30) and (31) show the same patterns as well. Section 4: The Mandarin Existential Construction Before introducing the Mandarin you-‐EC specifically, the initial portion of this section, largely following Mei (1999), will introduce the inventory and order of functional projections generally in Mandarin. To be specific, this chapter will assume that Mandarin functional structure exists in the following order: (33) TopP Top TP T AspP Asp PredP Pred VP The Topic Phrase TopicP is host to a syntactic topic about which the rest of the sentence (i.e. the totality of the functional projections below TopicP) denotes a comment which says something about the topic. It is crucial to note that a topic is different from a subject in the sense that it, “sets a spatial, temporal, or individual framework within which the main predication holds” (Mei 1999:21). A subject, on the other hand, is a semantic argument of the predicate in that it either experiences the state denoted by the predicate or participates in the action denoted by it. This 166 relationship can be observed in (34) where fuyuan is the subject of laiokuo, but zhongguo is the topic on which the denotation of [fuyuan laiokuo] is a comment: (34) zhongguo (a) fuyuan laiokuo (Mei 1999:21) China PRT territory vast “China, (its) territory is vast” It is also important to note that Topics can be raised elements or simply base generated in Topic position. This distinction which can be observed in (35) will remain an empirical observation here, but be discussed in great detail and theoretically motivated in the subsequent discussion of ECs in Mandarin with IL and/or perfective predicates: (35) a. nei jian shii Wangli zhidao ti (Mei 1999:22-‐3) that matter Wangli knows “That matter, Wangli knows.” b. [nei chang huo] xingkui xiaofangdui lai de zao that CL fire fortunately fire-‐brigade come DE soon “In terms of that fire, fortunately, the fire brigade came soon” While the Topic phrase is widely accepted, the presence of the Tense projection in Mandarin is more controversial. Mandarin generally does not have overt tense (although the future marker jiang is said to be a true tense marker and hence evidence for a TP). There is often indirect evidence cited for its existence in terms of blocking and licensing various movements or interpretations . Also, from the standpoint of the Minimalist approach to syntax, the presence of a Tense node is crucial in case checking and driving movement, as well as having a (seemingly) 167 language independent) selectional relationship with C (Chomsky, Li, Pesetesky and Torrego a.o.). However, Lin (2002, 2005, 2010) suggests an analysis of Mandarin with no Tense and hence no TP. His claim is that the Mandarin aspectual system combined with the mass/count system and aksionsart of verbal lexical semantics is all that is needed to compute temporal reference in Mandarin. Ultimately, while the semantic insight in Lin’s proposal will be featured later in the chapter, the syntax adopted here, following the Minimalist program will posit the existence of TP. In contrast to Tense, Aspect in Mandarin is morphologically encoded. Mandarin has four aspect morphemes: two perfective (-le and –guo) and two imperfective (-zhe and zai): (36) a. Lifan he le yi bei cha (Mei 1999:29-‐31) Lifan drink PF one cup tea “Lifan has had a cup of tea” a’.* Lifan he yi bei cha le Lifan drink one cup tea PF “Lifan has had a cup of tea” b. Lifan kan guo zhe ben shu Lifan read GUO this CL book “Lifan has read this book” b’.* Lifan kan zhe ben shu guo Lifan read this CL book GUO “Lifan has read this book” c. Ta tou shang dai zhe yi ding hong maozi 3sg head top wear ZHE one CL red hat “He was wearing a red hat on his head” c’.*Ta tou shang dai yi ding hong maozi zhe 3sg head top wear one CL red hat ZHE 168 “He was wearing a red hat on his head” d. Lifan zai da dianhua Lifan PRO call phone “Lifan is making a telephone call” d’. Lifan zai gei ta mama da dianhua Lifan PRO to his mother call phone “Lifan is making a telephone call to his mother” I will discuss the detail of the semantics and the actual aspectual properties of the morphemes later, but for now what should be noted is that in the a/a’-‐c/c’ examples, the morpheme must be located directly after the verb whereas in the d/d’ example the morpheme appear pre-‐verbally and can be separated. Lastly, just as posited in English in the last chapter, Mei (1999) following Bowers (1993) suggests that there is a PredP projection in Mandarin based on adjunct ordering data as well hen support required in adjectival copulars. It is crucial to note, however, that in order to maintain consistency, since the PredP projection is between the VP and the AspP projection, in order for the covert-‐LF movement/AGREE relation Asp and the aspectual morpheme to occur, there must be V to Pred movement in cases where the relevant morphology appears. With the basic assumption of word order and the topography of the functional projection hierarchy in Mandarin established, we can now turn our attention to elucidating the crucial properties of the mandarin you-‐EC as in: (37) (Locative) you (num-‐CL-‐)NP XPCoda 169 Section 4.1: NP and Coda are a Constituent The first and most basic constituency test to determine whether two elements are constituents of the same type is whether or not they can be coordinated. In the case at hand, if the post-‐verbal NP and the coda form a constituent [NP Coda], they should be able to coordinate with other such [NP Coda] structures . (38) shows that coordination is possible with the [NP Coda] structures: (38) a. you yi-‐ge nusheng zai sao-di, yi-‐ge nansheng zai ca chuanghu have one-‐CL girl PROG sweep-‐floor one-‐CL boy PROG wipe window “There is a girl sweeping the floor, a boy wiping the window.” If the two elements being coordinated have internal predication relationships within them resulting in something like: [[Subj. Pred.] ∧[Subj. Pred.]], then a slightly more elaborate test along the same lines as the coordination test can be conducted. The ellipsis test is whether or not a constituent can be pronominalized or elided when it is repeated in the second conjunct. Moreover, this test offers more detailed structural insights into the nature of the coordinate conjuncts themselves because depending the type of ellipsis that is licensed across constituents the scope of the pronominalization will different. This difference in scope is a property that must be indicative of the conjuncts structural topography. In the English example in (39a), did replaces the whole VP expects to finish his homework quickly, while in (39b), did so replaces only the lowest VP finish his homework. Therefore, (39a) cannot have a different adjunct in the second conjunct while (39b) can. 170 (39) a. Bill expects to finish his homework very quickly, and John also does *(very slowly). b. Bill expects to finish his homework very quickly, and John expects to do so very slowly. In Chinese, the equivalent pronominal form of English do is shi ‘be’ (Li 1998), and it replaces the whole VP conjunct including the adjunct. (40) Zhansan yuqi neng hen-‐kuai-‐de wancheng tade gongke, Lisi *(manman-‐de) ye shi Zhangsan expect able very-‐fast-‐DE finish his homework Lisi slowly also be “Zhangsan expects to finish his homework very quickly, and Lisi also does *(very slowly). ” The Chinese null object constructions (NOCs) (Hoji 1998) are the Chinese counterpart to the English do so ellipsis. Xu (2003) illustrates that (41) does not exclude the possibility that Lisi brushed his teeth sloppily. Thus, (42) is grammatical with a different adverbial in the second conjunct just as in (39b). (41) Zhangsan zixide shua-‐le ya, Lisi ye shua-‐le. Zhangsan carefully brush-‐PFV teeth Lisi also brush-‐PFV “Zhangsan carefully brushed his teeth, and Lisi also brushed (it).” (42) Zhangsan zixide shua-‐le ya, Lisi ye suibiande shua-‐le Zhangsan carefully brush-‐PFV teeth Lisi also sloppily brush-‐PFV “Zhangsan carefully brushed his teeth, and Lisi also sloppily brushed (it).” 171 In the case of the Chinese EC, if the coda is an adjunct then a different coda should be allowed in the second conjunct. However, following Liu (2010), (43b) and (44b) prove that different codas are ungrammatical. She therefore concludes that the post-‐verbal NP and the coda must form a constituent that is a complement of the existential verb. (43) a. jiaoshi-‐li you yi-‐ge shizhong hen da, litang-‐li ye you classroom-‐in have one-‐CL clock very big auditorium-‐in also have “There is a clock which is big in the classroom, and also (one which is big) in the auditorium.” b. *jiaoshi-‐li you yi-‐ge shizhong hen da, litang-‐li ye you hen xiao classroom-‐in have one-‐CL clock very big auditorium-‐in also have very small “There is a clock which is big in the classroom, and also (one) which is small in the auditorium.” If the post-‐verbal NP and the coda form a constituent, they should be able to coordinate. Section 4.2 The Coda is a Predicate of the Post-‐Verbal NP While the evidence from the ellipsis test is some indirect evidence that the coda is a predicate, more direct evidence that it is the predicate of the post-‐verbal NP can be illustrated. In Chinese, the morpheme hen ‘very’ is a predicative marker when used post nominally indicating that the adjective is a predicate (44). Furthermore, adjectival codas in the Chinese EC must have hen 'very', while the prenominal modifier counterpart does not necessarily require it (45). 172 (44) a. zhe-‐ge nuhai *(hen) piaoliang this-‐Cl girl very beautiful ‘This girl is very beautiful.’ b. wo *(hen) e I very hungry “I’m very hungry.” (45) a. wode ban-‐shang you yi-‐ge nuhai *(hen) keai my class-‐in have one-‐CL girl very cute “There is a girl who is very cute in my class.” b. wode ban-‐shang you yi-‐ge (hen) keai de nuhai my class-‐in have one-‐CL girl very cute DE girl “There is a (very) cute girl in my class.” More evidence that the coda is a predicate comes from the lexical restriction on post-‐nominal codas. The coda in the Chinese EC is restricted to predicative adjectives, but prenominal modifiers have no such restriction. Liu (2010) offers the following example where yigu 'late' and gongtong 'common' are non-‐predicative adjectives and therefore cannot appear in the EC. (46) a. meng-‐li you yi-‐ge yigu de qingren dream-‐in have one-‐CL late DE relative “There was a late relative in the dream.” b. *meng-‐li you yi-‐ge qingren yigu (EC) dream-‐in have one-‐CL relative late 173 (47) a. zhe liang-‐ge buluo you yi-‐ge gongtong de yuyan this two-‐CL tribe have one-‐CL common DE language “There is a common language between the two tribes.” b. *zhe liang-‐ge buluo you yi-‐ge yuyan gongtong (EC) this two-‐CL tribe have one-‐CL language common Section 4.3: Movement and Binding Effects in the Chinese EC Lastly, Liu (2010) catalogues the results of the canonical movement and binding tests to determine where elements are base generated and where they end up. Initially, she notes the fact that binding patterns in the EC suggest that the post-‐ verbal NP is base-‐generated in the subject position of the coda. Second, she shows idioms can be separated which illustrates that the idiom (which has a structure of [subj pred]) is base generated as the complement of the existential verb. Finally, she notes that extractions out of codas are subject to island violations. In the case of binding, according to Principle A, taziji ‘himself/herself’ must be locally bound. In (48), taziji ‘himself’ can only be bound by Lisi, the local subject, not Zhangsan, the matrix subject. (48) Zhangsani shuo Lisij hen xihuan taziji*i/j. Zhangsan say Lisi very like himself “Zhangsani said that Lisij like himself*i/j very much.” Turning to the ECC examples in (48a), taziji 'himself' can be successfully bound by Zhangsan by reconstructing yi-ben taziji de shu ‘a book of himself’s’ back to the object position of the coda since kan ‘read’ is a two-place verb that takes an external and an internal argument. On the other hand, shui-zhao ‘fall-asleep’ in (49b) is a one-place verb which does not have an extra argument position for yi-ben taziji de shu ‘a book of 174 himself’s’ to reconstruct back to. Taziji ‘himself’ cannot be properly bound, and hence (49b) is ungrammatical. Also, crucially notice that (49c) with a bare noun is ungrammatical, but (72d) with the in-situ reflexive is fine as a bare noun. (49) a. you [yi-ben taziji i de shu] j Zhangsan i bu xiang kan t j have one-Cl himself DE book Zhangsan not want read Lit: “There is [a book of himself i ' s] j Zhangsan i does not want to read t j ” “There is a book Zhangsan i wrote himself i , which he i does not want to read.” b. *you [yi-ben taziji i de shu] Zhangsan i dou shui-zhao le. Have one-CL himself DE book Zhangsan even fall-asleep LE Lit: “There is a book of himself i ' s which even Zhangsan i fell asleep.” c. *you [taziji i de shu] j Zhangsan i bu xiang kan t j have himself DE book Zhangsan not want read d. Zhangsan i bu xiang kan taziji i de shu Zhangsan not want read himself-DE book The same point can be illustrated by the binding of pronouns. In (50a) the optional pronoun ta 'he' can be understood to be referential to yi-ge ren 'one-Cl person', while the pronoun in (50b) cannot. The ungrammaticality of (50b) can be easily accounted for by Principle B violation if we reconstruct yi-ge ren 'one-Cl person' back to the object position. Notice in (50), however, that a BNP is licit in (50b). 175 (50) a. you [yi-ge ren] i (ta i ) zou le have one-Cl person he leave LE “There is a person i who i left.” b. you [ren] i (ta i ) zou le have people they leave LE c. *you [yi-ge ren] i (ta i ) bu renshi t i have one-Cl person he not know “There is a person i who i does not know t i ” Liu (2010) also shows that reconstruction effects can be observed using idiom expressions. Since an idiom is one unit in the lexicon, if some part of the idiom is separated from the rest of it, movement must have taken place. She notes that in (51), the morpheme pairs chi ciu ‘eat vinegar’, which mean being jealous; zhan pianyi ‘occupy cheap’, taking advantage; and kai dao ‘open knife’, operating on (someone), are not adjacent. Their status as idioms, despite the inverted order is evidence that the topics in (51) were base generated in a lower position in the and subsequently moved. (51) a. tade ciui, wo bu gan chi ti his vinegar I not dare eat Lit: “His vinegar, I do not dare to eat.” “I do not dare to be jealous of him.” b. [[ta zhan ti de] pianyii] bi shei dou duo. he occupy DE cheap compare who all much Lit: “The advantage he occupies is more than anyone else’s.” “He took more advantages than anyone else’s.” 176 c. daoi, wo tiantian dou zai kai ti. Knife I everyday all ZAI open Lit: “Knife, I open (it) everyday.” “I have operations (on patients) everyday.” (52) a.you yi-‐zhong ciui ni bu neng chi ti have one-‐Cl vinegar you not can eat Lit: 'There is a kind of vinegar that you cannot eat.’ “There is one kind of jealousy you cannot have.” b. you yi-‐zhong pianyii wo bu gan zhan ti have one-‐Cl cheap I not dare occupy Lit: 'There is a kind of advantage that I do not dare to occupy. “I do not dare to take this type of advantage.” Finally, while the binding and idiom evidence shows the robust possibility of movement of constituents within the EC, Liu (2010) uses island violations on movements out of codas to illustrate the restrictions on movements in the EC and thus offer insight into the structural composition of the Chinese EC. While extraction out of an island is generally not possible (Ross 1967), (40) illustrates that Chinese has a strategy to rescue such extractions in the form of the resumptive pronoun. Topicalizations of indefinite NPs will result in you-‐existentials since Chinese does not allow indefinite subjects or topics (the fact that these type of predicate inversions result necessarily in existential constructions is an important to which we will return in the analysis section that follows). However, topicalizations from within a complex NP (53a) or adjunct (53b) are only acceptable when a resumptive pronoun is in the trace. 177 (53) a. you [yi-‐ge xuesheng]i [wo hen xihuan [[*(tai) changge de] shengyin]] (CNPC) have one-‐CL student I very like he sing DE voice “There is a student who I really like the voice with which *(he) sings.” b. you [yi-‐ge laoshi]i xiaozhang [yinwei *(tai) mei lai] hen shengqi (AC) have one-‐CL teacher principal because he not come very angry “There is a teacher who made the principal very angry because *(he) did not come.” 5. The Syntax of the Mandarin Existential Construction With the properties of the English EC (EEC) and Mandarin EC (MEC) established, we can immediately note some striking similarities. First, the constituency of the post-‐ you NP and the coda are reminiscent of the data for the EEC. Second, the fact that the coda is a predicate of the post-‐you element also shows similarity with the EEC. However, as noted in the introduction, there are in fact subtle distinctions in the types of predicates that are licensed in the EEC vs. the MEC: (54) a. there is a student tired SL ok b. *there is a student intelligent/ a doctor *IL adj/nominal c. you yi ge xuesheng hen lei SL ok have one CL student very tired d. you yi ge xuesheng hen conming IL-adj ok have one CL student very intelligent e. you yi ge xuesheng shi yisheng IL-‐nominal ok have one CL student be doctor As in (54), it is clear that non-‐verbal IL predicates are licit in MEC contexts but illicit in EEC contexts. Specifically, this section will argue that Mandarin does allow IL 178 predicates, because it essentially has a strong subject (pro) that is the subject of the coda predication; the strong subjects are licit in Chinese because the DE holds at the level of the matrix clause where the post-‐verbal NP is actually base generated and in a co-‐reference relationship with a lower pro. I will propose the structure for the MEC is: (55) you yi ge xuesheng shi yisheng have one CL student BE doctor TP LocSubj AsPP youk VP yi ge xueshengj V’ tk DP *NPi + D= proj PredP ti Pr’ … shi yisheng In this structure, a *NP is base generated in the Spec PredP, the *NP is similar in content as to what is in the object position of an English –wh question before the element is moved to a phase edge to adjoin to its –wh phonology and take an A-‐bar scope position (Vergnaud P.C.). This element then moves to D, the resulting complex of the *NP + D is pro. This pro is bound by a coreference relation with the element that is base-‐generated in Spec VP. The fact that it is a pro in D means that it 179 functions as a strong subject and can therefore support an IL predicate. Evidence that there is a pro there comes from the fact that an overt pronoun can be inserted: (56) you yi ge xuesheng ta shi yisheng have one CL student 3rdP BE doctor Note that the structure is virtually identical to the structure proposed in Aoun and Li (2003) for English –wh relative clauses. 180 Conclusion Through a crosslinguistic analysis of existential constructions and their copular counterparts, the following generalizations were obtained: (i) Existential Constructions in English are derived from the situation-‐descriptive predicational structure, and their theticity is a result of being derived from mass-‐ predicates (Roy 2005). The structure is as in (1): (1) TP therePROi T’ Tns VP ti V’ to be PredP [XP, or YP from [Spec XP] Pr’ Pr XP (ii) There are generally two interpretations of thetic structures—entity-‐central and event-‐central (Sasse 1987)—and in the existential construction both are available (unless ruled out for independent reasons a la the OC construction) and structurally distinct with different agreement patterns in the plural. (iii) The Definiteness Effect in English is a result of a Coda Condition (Keenan 2003) that semantically follows from the existential being derived from the situation- descriptive and precludes the possibility of definite/strong quantified subjects in PredP/XP 181 (iv) The Predicate Restriction in the English is really two overlapping effects. The first part is that nominal predicates are ruled out because they can never be in existentials since they would always be characterizing or defining predicational structures. The second part of the restriction against “IL adjectives” in the English EC follows from the definiteness effect since subjects of these predicates must be strong and strong subjects are ruled-‐out by the Coda Condition. (v) Extending the predicational system to Mandarin requires a reinterpretation of mass/count as counted/uncounted as in (2) (2) Predicational copular sentences-‐ Mandarin-Semantics Count Non-‐Counted counted countable uncountable (vi) The structure for the Mandarin EC is similar to the English –wh relative clause structure and as such has a strong pro subject in PredP. This difference has two consequences: (a) it means that the DE holds at the Matrix level as a result of a general ban on double quantification (YOU existentially quantifies the nominal and therefore if it were also strong-‐quantified it would be ungrammatical—viz. Barwise and Cooper (1981)) and (b) since there is a pro which is itself strong, there is no problem with IL-‐adj predicates in the coda. 182 References Abbott B. 1995. Some Remarks on Specificity. Linguistic Inquiry 26(2), 341-347 Abush, D. 1994. The Scope of Indefinites. Natural Language Semantics 2(2), 83-136 Aoun J. and Li A. (2003) Essays on the Representational and Derivational Nature of Grammar MIT Press Baker M.C. (2003) Lexical Categories: Verbs Nouns, and Adjectives Cambridge Universtiy Press Baker M.C. (2006) The Syntax of Agreement Cambridge Universtiy Press Barrie M.J.M. (2006) Dynamic Antisymmetry and the Syntax of Noun Incorporation. PhD. Diss. University of Toronto Barwise, Jon & Robin Cooper 1981. Generalized quantifiers and natural language. Linguistics & Philosophy 4: 159–219 Basilico D. (1997) “The Topic is ‘There’ “ Studia Linguistica 51:3 pp 278-‐316 Bianchi V. (2000) Some issues in the syntax of relative determiners Syntax of relative clauses eds. Alexiadou et al pp. 53-‐82 Bock K. and Miller C.A. (1991) “Broken Agreement” Cognitive Psychology 23 pp. 45-‐ 93 Borer H. (2005a) Structuring Sense Vol. I: In Name Only Oxford University Press Borer H. (2005b) Structuring Sense Vol. II: The Normal Course of Events Oxford University Press 183 Bowers (1993) “The Syntax of Predication” Linguistic Inquiry 24: pp 591-‐656 Brentano F. (1874) Psychology from an Empirical point of View transl. A.C. Rancurello (1973) London: Routeledge Carlson, Gregory 1977a. Reference to Kinds in English. Ph.D. Dissertation.University of Massachusetts, Amherst. Published 1980 by Garland Press, New York Carlson, Gregory 1977b. Amount relatives. Language 53: 520-‐542. Cheng, Lisa L. -‐S. 1991. On the Typology of Wh-questions. Cambridge: MIT dissertation. Cheng, Lisa & Sybesma, Rint 1999. Bare and Not So Bare Nouns and the Structure of NP. Linguistic Inquiry 30 (4): 509-‐542 Chierchia G. (1988) “Pluality of Mass Nouns and the Notion of “Semantic Parameter”” in S. Rothstien ed. Events in Grammar pp. 53-‐103 Kluwer Publications Chomsky (1981) “A note on non-‐control PRO” Journal of Linguistic Research Chomsky N. (1986) Barriers MIT Press Chomsky N. “On Phases” Unpublished Manuscript MIT Chomsky N. (2000) “Derivation By Phase” in M. Kenstowicz Ken Hale: A Life in Language MIT Press pp 1-‐52 Chomsky (1995) The Minimalist Program MIT Press Chomsky N. (1993) “The Minimalist Program for Linguistic Theory” in Hale K. and Keyser S.J. eds. The View From Building 20 MIT Press Chung, Sandra 1987. The syntax of Chamorro existential sentences. In: E. Reuland & A.G. B. ter Meulen (eds.). The Representation of (In)definiteness. Cambridge, MA: MIT Press, 191-‐225. 184 Davidson D. (1967) “The Logical Form of Action Sentences” in N. Rescher ed. The Logic of Decision and Action University of Pittsburgh Press Diesing M. (1992) Indefinites MIT Press Dechaine R-‐M. (1993) Predeicates Across Categories: Towards a Category Neutral Syntax. PhD. Diss. UMASS Amherst den Dikken M. (2006) Relators and Linkers: the syntax of predication, predicate inversion, and copulas MIT Press DeWolf G.D. (1992) Social and Regional factors in Cannadian English Tornto: Cannadian Scholar’s Press Enç M. 1991. The Semantics of Specificity. Linguistic Inquiry 22(1), 1-25. Fodor, J. and I. Sag 1982. Referential and Quantificational Indefinites. Linguistics and Philosophy 5, 335-398. Guerton J. (1980) On the syntax of PP extraposition Linguistic Inquiry 11 637-‐678 Groat E.M. (1999) “Raising the Case of Expletives” in Epstein S.D. and Hornstein N. eds. Working Minimalism MIT Press Grice, H.P. 1975. Logic and Conversation. In: D. Davidson & G. Harman (eds.). The Logic of Grammar. Encino, CA: Dickenson. Grosu A. and Landman F. (1998) Strange relative of the third kind. Natural Language Semantics 6 pp.125-‐170 185 Hazout I. (2004) “The Syntax of Existential Constructions” Linguistic Inquiry 34:3 pp 393-‐430 Hegge L.A. (1988) The Syntax of Copular Structures. PhD. Diss USC Heim I. (1982) The semantics of definite and indefeinite noun phrases PhD Diss UMASS Amherst Heim I. (1987) “Where does the Definiteness Restriction Apply? Evidence from the Definiteness of Variables” in Reuland E.J. and Ter Meulen A.G.B. eds. The Representaion of (In)Definiteness MIT Press Heim I. and Kratzer A. (1998) Semantics in Generative Grammar Blackewell Publishing Malden MA. Higginbotham J. (1987) “Indefinites and Predication” in Reuland E.J. and Ter Meulen A.G.B. eds. The Representaion of (In)Definiteness MIT Press Higginbotham, J. and G. Ramchand (1996): The Stage-Level/Individual-Level Distinction and the Mapping Hypothesis, ms, University of Oxford. Higginbotham J. (2000) “On Events in LInguistic Semantics” in HIgginbotham J. Pianesi F. and Varzi A.C. eds. Speaking of Events Oxford Universtiy Press Hu J. & Pan H. 2002. Focus and the Basic Function of Chinese Existential You-‐ Sentences Huang C.-‐T. 1987. Existential Sentences in Chinese and (In)definiteness. In: Eric J. Reuland and Alice GB ter Meulen (eds.) The Representation of (In) definiteness. Cambridge, MA.: MIT Press, 227-‐254 Huang, S.-‐Z. (2006). Property theory, adjectives, and modification in Chinese. Journal of East Asian Linguistics, 15, 343–369 186 Kayne R. (1994) The antisymmetry of syntax MIT Press Keenan E.L. (2003) The Definiteness Effect: Semantics or Pragmatics? Natural Language Semantics 11 pp. 187-‐216 Kiss K. (1996) Two subject positions in English. The Linguistic Review 13 119-‐142 Kratzer, Angelika 1988. Stage-‐level and individual-‐level predicates. Genericity in Natural Language, ed. By Manfred Krifka, 247-‐284. Tübingen: University of Tübingen Kratzer, Angelika 1996. Severing the external argument from its verb. Phrase Structure and the Lexicon, ed. By John Rooryck and Laurie Zaring, 109-‐137. Dordrecht: Kluwer Kuroda S. (1972) The Categorical Judgement and the Thetic Judgement Foundations of Language 9 pp. 153-‐185 Kuroda S. (1995) Sentences, Judgement and Propositions. unpublished ms. Ladusaw W.A. (2000) Thetic and Categorical, Stage and Individual, Weak and Strong Negation and Polarity eds. Horn L.R. and Yasuhiko K. London: Oxford Press pp. 232-‐242 Larson R.K. (1999) “Semantics of Adjectival Modification” LOT Winterschool Class Notes, Amsterdam Lin, J.-‐W 2005. Time in a language without tense: the case of Chinese. Journal of Semantics 23: 1-‐53 Lin, T.-‐H 2004. Aspect, Distributivity, and Wh/QP Interaction. Language and Linguistics 5.3:615-‐642 187 Lin, T.-‐H & Fang, H.-‐L 2008. The Mandarin You existential: a verbal Analysis. USTWPL 4: 43-‐56 Lewis D. (1975) “Adverbs of Quantification” in E Keenan ed. Formal Semantics of Natural Language pp. 3-‐15 Cambridge Universtiy Press McCloskey J. (1991) “There, It, and Agreement” LInguistic Inquiry 22:3 563-‐7 McNally L. (2008) DP-‐internal ONLY, Amount-‐realtives, and relatives out of existentials. Linguistic Inquiry MIT Press pp. 161-‐169 McNally L (1992) “An interpretation for the English Existential Construction” Diss. UCSD McNally, Louise 1998. Existential sentences without existential quantification. Linguistics & Philosophy 21: 353-‐392 McNally, Louise 2008a. DP-‐internal only, amount relatives, and relatives out of existential sentences. Linguistic Inquiry 39, 161-‐169 Milsark G. (1974) Existential Sentences in English. Doctoral Diss. MIT Milsark G. (1977) “Toward an Explanation of Certain Peculiarities in the Englsih Existential Construction” LInguistic Analysis vol. 3 pp. 1-‐29 Moro A. (2000) Dynamic Antisymmetry MIT Press Pasrons T. (1990) Events in the Semantics of English: A Study in Subatomic Semantics MIT Press Rando E. and Napoli D.J. (1978) “Definites in There-‐Sentences” Language 54:2 pp. 300-‐313 188 Rapposo E. and Uriagareka J. (1995) Two types of small clauses. Syntax and Semantics 28: Small Clauses eds. A. Cardinaletti and M. Guasti: NY, Academic Press pp. 153-‐177 Reed A. (1996) Pratitives, existentials, and partitive determiners. Parititves ed. J. Hoeksma Mouton: Hague pp. 143-‐179 Rizzi L. (1990) Realtivized Minimality MIT Press Roy. I.A. (2005) Non Verbal Predications: A Syntactic Analysis of Predicational Copular Sentences. PhD. Diss. USC Sasse H.J. (1987) “The Thetic/Categorical” Distinction Revisitied” Linguistics 25 Safir, K.J. (1985) Syntactic Chains Cambridge University Press Schutze C. (1999) “Englsih Expletive Construction are not Infected” LInguistic Inquiry 30 pp. 467-‐484 Sobin N. (1997) “Agreement Default Rules and Grammatical Viruses” Linguistic Inquiry 28 pp. 318-‐343 Sparks, R.B. (1984) Here’s a few more facts. Linguistic Inquiry 15 pp. 179-‐83 Stowell T. (1978) “What There was before There was There” in CLS XIV Williams E. (1984) There-‐Insertion Linguistic Inquiry vol. 15 pp 131-‐153 Williams E. (1994) Thematic Structure in Syntax MIT Press Williams E. (2006) “The Subject-‐Predicate Theory of There” Linguistic Inquiry 37:4 pp. 648-‐651 Zhang, Ning 2008. Existential coda construction as Internally Headed Relative Clause Construction. The Linguistics Journal 3 : 8-‐54 189 Zimmerman M. (2003) “Pluractionality and Complex Quantifier Formation” Natural Language Semantics vol. 11 pp. 249-‐287 Zucchi A. (1995) “The ingredients of Definiteness and the Definiteness Effect” Natural Language Semantics 3 pp. 33-‐78

Abstract (if available)

Abstract This dissertation is concerned with the specific relationship between copular sentences and existential sentences, and what about that relationship is responsible for the well-known restrictions on definiteness and predicates that are observed in the existential cases. Instead of starting with the well-known paradigms which illustrate restrictions in the existential constructions that result in ungrammaticality

Linked assets