Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Phrase structure and verb movement in Hebrew and English imperatives Strauss, Uri Neill 2000

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_2000-0152.pdf [ 3.62MB ]
JSON: 831-1.0089419.json
JSON-LD: 831-1.0089419-ld.json
RDF/XML (Pretty): 831-1.0089419-rdf.xml
RDF/JSON: 831-1.0089419-rdf.json
Turtle: 831-1.0089419-turtle.txt
N-Triples: 831-1.0089419-rdf-ntriples.txt
Original Record: 831-1.0089419-source.json
Full Text

Full Text

PHRASE STRUCTURE AND VERB MOVEMENT IN HEBREW AND ENGLISH IMPERATIVES by URI NEIL STRAUSS B . A , York University, 1997 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS in THE FACULTY OF GRADUATE STUDIES (Department of Linguistics) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA March 2000 © Uri Neil Strauss, 2000 UBC Special Collections - Thesis Authorisation Form In presenting t h i s thesis i n p a r t i a l f u l f i l m e n t of the requirements for an advanced degree at the University of B r i t i s h Columbia, I agree that the Library s h a l l make i t f r e e l y available f or reference and study. I further agree that permission for extensive copying of thi s thesis f or sc h o l a r l y purposes may be granted by the head of my department or by his or her representatives. It i s understood that copying or p u b l i c a t i o n of t h i s thesis f or f i n a n c i a l gain s h a l l not be allowed without my written permission. The U n i v e r s i t y of B r i t i s h Columbia Vancouver, Canada Date 1 of 1 3/15/00 7:06 PM Abstract This thesis investigates two questions about imperatives. The first is whether their phrase structure is similar to or different from the phrase structure of declaratives. The second is what the movement properties of imperative verbs are, and whether they are the same as the movement properties of other verb types. Both of these questions are investigated in English and Hebrew. It is shown that Hebrew has three distinct types of syntactic imperatives, each corresponding to a different morphological paradigm - one uniquely imperative, one identical to the future tense, and one identical to the infinitive. A variety of tests in the two languages, including licensing of negative polarity items, occurrence of VP ellipsis, and the presence of a subject/object asymmetry in embedded negatives, provide evidence that imperative phrase structure is essentially the same as the basic phrase structure of declarative clauses familiar from the syntactic literature. To test the extent of verb movement, a three-way distinction in adverb types is proposed, and the position of the verb is determined by its placement relative to each type of adverb. This test shows that in English, imperative verbs move to the same extent that other nonfinite verbs do, while in Hebrew, imperatives have unique movement properties, and are forced to raise higher than finite or infinitive verbs are. ii Abstract ii List of Tables v Acknowledgements vi CHAPTER 1 Introduction 1 CHAPTER 2 Indirect, True and Suppletive Imperatives 4 2.1 Overview 4 2.2 Defining 'Imperative' 4 2.3 Imperatives as Covert Performatives . . 6 2.3.1 Sadock (1974) 6 2.3.2 J.D. Fodor (1977) 8 2.4 True and Suppletive Imperatives 10 2.4.1 Differences in Hebrew 10 2.4.2 True/Suppletive vs. Direct/Indirect 12 2.5 Summary 14 CHAPTER 3 Basics of Verb Movement 16 3.1 Introduction 16 3.2 Verb Movement as Feature Checking 16 3.3 The Head Movement Constraint 16 3.4 Verb Movement Across Neg° 18 3.4.1 Hebrew and English 19 3.5 Conclusion 20 CHAPTER 4 The Phrase Structure of Imperatives 21 4.1 Introduction 21 4.2 Subjects and IPs 21 4.2.1 Beukema and Coopmans (1989) 21 4.2.2 Potsdam (1996) 23 Negative Polarity Licensing 23 Subject-deriving Operations 24 VP Ellipsis 25 4.3 NegP and IP in Hebrew 25 4.3.1 Agreement Blocking 25 4.3.2 NegP and IP in Hebrew Imperatives 29 4.3.3 Subject/Object Asymmetry 29 4.4 Conclusion 31 Appendix - Don 7 as a Particle 32 iii CHAPTER 5 Verb Movement in English Imperatives 36 5.1 Introduction 36 5.2 Potsdam (1996) 36 5.2.1 Negation Placement 37 5.2.2 Adverb Placement 39 5.3 Problems with Potsdam's Adverb Account 40 5.3.1 S-Adverbs and Negation 40 5.3.2 S-Adverbs and Movement to Pre-subject Position . . . 40 5.4 An Alternative Account of Adverb Positioning 41 5.5 Adverbs and Verb Movement 44 5.5.1 Finite Auxiliaries 45 5.5.2 Infinitive Auxiliaries 45 5.5.3 Subjunctive Auxiliaries 46 5.5.4 Main Verbs 47 Finite 47 Infinitive 47 Subjunctive 47 5.5.5 Imperatives 47 5.6 Conclusion 48 CHAPTER 6 Verb Movement in Hebrew 49 6.1 Introduction 49 6.2 Manner, Temporal and Speaker Adverbs 49 6.3 Finite Verb Movement 50 6.4 Infinitive Verbs 54 6.5 Imperative Verbs 56 6.5.1 True Imperatives 56 6.5.2 Future Suppletive Imperatives 57 6.5.3 Infinitival Suppletive Imperatives 59 6.6 Summary 62 CHAPTER 7 Summary 63 REFERENCES 65 iv List of Tables Page 48 - Verb Movement in English Page 62 - Verb Movement in Hebrew Page 64 - Verb Movement: A Comparison of English and Hebrew Acknowledgements I thank first and foremost my committee advisor, Henry Davis, for all of the help he provided. I also thank the other members of my committee, Rose-Marie Dechaine and Strang Burton, for valuable input. Many others at UBC contributed to this thesis as well - more than I can list or even remember. Cody Shepherd, Eun-Sook Kim, Ikuyo Kaneko, Marion Caldecott, Matt Ritchie. Sunyoung Oh, Suzanne Gessner, Tanya Bob, Tomio Hirose and Yumiko Nakamura contributed intellectually and also made the atmosphere at UBC a pleasant one in which to work. I am grateful to those who helped the process along when the going got tough. My parents - Ted and Rivka, were supportive throughout, as were my brother Ofer and sister Dafha. I also owe gratitude to Eric Sommer, Eun-Sook Kim, Henry Davis, Lisa Matthewson, Nora Danielson, Sabina Iseli-Otto, and Suzanne Gessner. I relied on a number of consultants for judgements on the Hebrew data, primarily Rivka Strauss and Avikam Hameiri. Hagit Borer and Edit Doron helped with judgements and also analysis. vi Chapter 1 - Introduction The purpose of this thesis is to investigate the phrase structure and verb movement properties of imperatives in English and Hebrew, focusing on two issues. The first is whether the phrase structure of imperatives is essentially different from the phrase structure of declaratives. The second is whether verb movement in imperatives can be subsumed under other types of verb movement, or whether it has unique movement properties. A discussion of imperatives will require some clarification of what exactly constitutes an imperative. Conceivably, imperatives might be limited to canonical forms like shut the door!, or at the other extreme they might include statements that only hint at a command or request, such as it's cold in here. Chapter 2 discusses this issue. What emerges is a three-way distinction in imperatives. The first division is between direct imperatives, which are syntactically distinct, and indirect imperatives, which syntactically resemble declaratives or interrogatives, but are interpreted as commands or requests. Since the discussion in this thesis centres around the syntax of imperatives, the rest of the thesis will focus on direct imperatives and exclude indirect imperatives. Within the class of direct imperatives, there is a two-way split between true imperatives and suppletive imperatives, which is familiar from the literature on a number of European languages (e.g. Han 1998, Rivero 1994, Zanuttini 1994, 1997). I introduce the distinction using Hebrew data. The question arises whether Hebrew suppletive imperatives are not in fact indirect imperatives, and I employ syntactic and pragmatic arguments to show that they are not. Chapter 3 introduces the basics of verb movement. The motivation for verb movement is accounted for using the feature-checking approach. The Head Movement Constraint, which is explained as a consequence of the Empty Category Principle following Baker (1988), is introduced to explain the fact that verbs only raise to the immediately c-commanding head. Chapter 3 closes with a discussion of an analysis that involves a case of verb raising in apparent violation of the Empty Category Principle. Chapter 4 deals with the phrase structure of imperatives, bringing evidence from English and Hebrew to bear against the hypothesis that imperatives are reduced clauses. Evidence includes the fact that subjects of imperatives license negative polarity items outside of the verb phrase, that they can be derived by raising to [SpecJP], that they are overt in clauses in which the verb phrase is ellipsed, that Hebrew imperative verbs have forms that must be licensed by raising to IP, and that complement clauses in Hebrew imperatives show an asymmetry between negative subjects and objects that is best accounted for by assuming a NegP in the higher clause. Chapter 5 investigates verb movement in English by considering the position of the verb with respect to not and adverbs. Adopting the assumption that adverbs are generally stationary specifiers of phrases (following Pollock 1989), it is shown that different classes of adverbs occur in different positions in the phrase structure. The position of verbs relative to the different adverb types can be used to determine which position the verbs occupy in the overt syntax. These tests are used on a range of finite and nonfinite verbs to determine the movement properties of each verb class. The conclusion, similar to the conclusion of Pollock (1989), is that imperative auxiliaries undergo a short movement, placing them in a class with auxiliary infinitives and subjunctives and in opposition to finite auxiliaries. 1 Chapter 6 uses the adverb positioning to investigate movement in Hebrew verbs. It is found that the relative positioning of adverb classes in Hebrew is the same as in English. The verb movement properties, however, are different, as shown by the different placement of verbs relative to adverbs. The different types of Hebrew imperatives form a class distinct from infinitives (there are no subjunctives in Hebrew) and possibly distinct from finite verbs as well. Since Hebrew imperatives can have morphology identical to infinitives or finite verbs, it follows that the movement properties of a verb are not determined solely by its overt morphology. 2 3 <9 Chapter 2 - Indirect, True and Suppletive Imperatives 2.1 Overview This chapter sets the domain of inquiry for the rest of the thesis by attempting to define what constitutes an imperative. We begin with a definition that takes account of the intuitive notion that imperatives sentences perform acts such as requesting or commanding and that they lack truth values. This implies that the answer to a sentence uttered in a discourse can be used as a diagnostic to determine whether or not the sentence is an imperative. Applying this test, we find that there is a class of sentences that cannot be neatly classified as imperative or declarative, because they are on the surface declaratives, while they are used to convey requests, commands, and other discourse notions associated with imperatives. This leads us to a review of the covert performative hypothesis, as presented in Sadock (1974) and Fodor (1977), which holds that such indirect imperatives - those that are not imperatives on the surface - form a single class with direct imperatives. We reject the covert performative hypothesis, arguing that the syntax distinguishes (direct) imperatives from other speech acts, and that the imperative-like interpretation of indirect imperatives is determined pragmatically. We then introduce a morphosyntactic distinction that exists in many languages between true and suppletive imperatives, and show that the distinction is not the same as the distinction between direct and indirect imperatives. Thus there are two divisions, resulting in three types of imperatives: one division separates direct from indirect imperatives, and a division in the class of direct imperatives separates true from suppletive imperatives. 2.2 Defining 'imperative' Informally, an imperative is a request, hope, wish, warning, command, instruction or some related speech act. As an initial characterization of imperative I use the following: an imperative sentence is one which has (i) a directive interpretation and (ii) no truth value. Right now I take no position on whether (ii) follows from (i) or is an independently stipulated fact. Directive, following Han (1998), means an instruction to the hearer of the sentence to change his or her plan set, a set of propositions which together form the state of affairs that he or she intends to bring about. This makes directive a discourse notion. The truth value of a sentence is usually understood to be determined from its semantic representation. Imperative is therefore a mixed discourse/semantic notion. As a test for imperatives, I suggest that the felicity of a range of responses to an utterance of a sentence can determine whether it is an imperative, declarative or interrogative. Appropriate responses to imperatives are "okay" and "I will", but not "yes", "I agree", "correct" or "true". "Okay" and "I wil l" indicate intention to comply with the direction, so they are a response to a statement with a directive interpretation. "I agree", "correct" and "true" are judgements about truth value, so they are appropriate responses to a statement that makes a claim about truth, i.e. a declarative. "Yes" is a response to a yes/no question. The judgements show (more or less) that the sentences in (1) are imperatives by these criteria. (1) A : (You/someone) go skiing! (a) Let's go skiing! (b) Go skiing or risk death! (c) Go skiing, will you! (d) B: Okay/I will/*yes/*I agree/*correct/*true. 4 A number of sentence types are compatible with imperative responses like "okay" as well as declarative responses like "correct". The following are some examples. (2) A: I order you to go skiing. (a) I want you to go skiing. (b) I request that you go skiing. (c) B: okay/I will/I agree/correct/true. The responses show that these sentences have a directive interpretation and a truth value. These do not correspond to the same reading of the sentence, though. Consider (2a). The "okay" response requires the reading in which the sentence is an order to go skiing, while "correct" is a response to the reading in which the sentence is a descriptive statement which claims that an order is being issued. The sentence is not simultaneously directive and truth-value bearing; rather it is ambiguous between imperative and declarative. (2b-c) are similar in this respect. Each of these sentences can be understood in one of two ways - on one hand either as a description of the speaker's performative use of the utterance (2a,c) or a description of the speaker's attitude (2b), or on the other hand as a request or command. In the rest of this thesis, I refer to such sentences as indirect imperatives, while the sentences in (1), which have only the imperative interpretations, I term direct imperatives. The sentences in (3) and (4) show a related ambiguity. (3) You will go skiing tomorrow, young man! (4) Hey, you there! Standing by the thermostat! It's pretty cold in here! Sentences (3) and (4) are never interpreted as assertions about the speaker's psychological state or discourse utterances. Instead, they can be implied commands or descriptive statements about the world. (3) can be a prediction that the addressee will go skiing the following day, or, if for instance uttered by a parent, it can be interpreted as a command. (4) can be interpreted as a claim about the temperature in the room, or as a request to change the temperature. Unlike the sentences in (2) and (3), however, an imperative response like "I will" seems inappropriate in (4). It seems that the request interpretation in (4) is the result of a Gricean implication, specifically from the maxim of relevance. Unless the room's temperature was the topic of conversation, the statement that it is cold is not relevant to the discourse. The addressee is meant to infer that the speaker would like him or her to adjust the temperature. Given a clear case where what is syntactically and semantically a descriptive statement can be interpreted as an imperative, the hypothesis that this is the case in all of the examples in (2-4) offers an attractive way to explain the fact that sentences can be used as commands or as statements. A number of other alternatives have been proposed, however, and are discussed in the rest of this chapter. One possibility is to reverse the analysis in the examples in (2) - that is, to assume that they are imperatives, and that their use as declaratives is derived. One way to do this is to assume that imperatives are typically embedded under a higher performative verb like / order, I want and I request like in (2), and that these performatives are either syntactically covert, as Sadock (1974) proposes, or present in the semantics but not in the syntax, as J.D. Fodor (1977) proposes. 5 A related alternative is that there is no higher clause in an imperative, but rather there is a marker in the phrase structure that appears both in imperatives and in the complements of performatives, and accounts for their semantic similarities. Varieties of this analysis include Katz and Postal (1964) and Han (1998). 2.3 Imperatives as covert performatives: the single imperative hypothesis There is a hypothesis about imperatives which claims that direct imperatives are really indirect • imperatives - imperatives that have a performative verb, which happens to be nonovert. I refer to this approach as the covert performative hypothesis. In this section, I investigate two analyses that adopt the covert performative hypothesis - those of Sadock (1974) and of J.D. Fodor (1977), both of which grew out of the analysis of Katz and Postal (1964). Katz and Postal criticized Klima's (1964) analysis of imperatives, in which imperatives like (5b) are derived from sentences containing will like (5a). (5) a. You [tense] will not be sitting there then b. Do not be sitting there then They pointed out that the sentences in (5) do not have the same range of interpretations. (5a), like the sentence in (3), can be interpreted either as a prediction or as a command, while (5b) is limited to a command interpretation. Katz and Postal were assuming a model of grammar in which meaning-changing rules were illegitimate. As a remedy, they proposed to add a marker / to the structure in (5a) with the meaning "the speaker requests (asks, demands, insists, etc.) that". Later researchers, such as Ross (1970) and Sadock (1974), proposed that instead of an invisible speech act marker, the illocutionary force of sentences can be characterized by an invisible higher clause. Thus imperatives are embedded underneath verbs like command and request, interrogatives under verbs like ask and declaratives under verbs like assert. I consider Sadock's approach in 2.3.1. In 2.3.2,1 consider a variant of the covert performative hypothesis, proposed and criticized by J.D. Fodor, in which the performative is present in the semantics but not the syntax. We will see that both of these approaches are flawed. 2.3.1 Sadock (1974) Sadock brings a range of syntactic evidence to support the covert performative hypothesis. For instance, the reflexives that Sadock (1974) calls 'spurious', such as myselfin (6a) and yourself in (6b), can be made to fall syntactically in line with other reflexives on a certain version of the covert performative hypothesis. (6) a. Linguists like myself speak an even number of languages b. Distinguished persons like yourself shouldn't carry their own luggage A reflexive generally requires an antecedent in the same clause, but reflexives in certain types of NP, such as himself'in pictures of himself, can have their antecedent in a higher clause, (as 7a shows). That an antecedent is obligatory is shown in (7b). If covert performatives include NPs referring to the speaker and hearer, 'spurious' reflexives like myself and yourself fail under this generalization. 6 (7) a. Malcohrij alleged that pictures of himself! were used in the blackmail b. *The Lichtensteinsi allege that pictures of himself, were used in the blackmail c. [I; assert to you that] linguists like myself speak an even number of languages d. [I; assert to you that] distinguished persons like yourself shouldn't carry their own luggage Sadock cites another argument for the covert performative hypothesis, originally presented by Robin Lakoff (1968), who notes that imperatives in Latin must occur in the subjunctive mood, which is otherwise prohibited in matrix clauses, being restricted to complements of certain types of verb, including verbs of ordering. This can be explained in the performative analysis by having Latin imperatives embedded under verbs of ordering. One of the weaknesses of this approach is that it has no explanation for the absence of overt subjects in imperatives in many languages. In English, subjects are unusual in imperative sentences. In other languages subjects are prohibited altogether (Zhang 1990). These facts must be stipulated under the covert performative hypothesis. Another problem with the performative hypothesis comes from logical facts about imperative sentences. If imperatives must be embedded under declaratives of command or direction, they should be able to participate in logical entailments the same way declaratives can. But as the arguments in (8) and (9) show, they do not. (8) I command you to eat more fish. Someone commands you to eat more fish. (9) Eat more fish! #1 command you to eat more fish. #Someone commands you to eat more fish. #You ought to eat more fish. #Eat more fish! (10) I command you to eat more fish/ someone commands you to eat more fish/ you ought to eat more fish. #Eat more fish! The performative declarative / command you to eat more fish in (8) entails someone commands you to eat more fish. (9) shows that the imperative eat more fish! does not entail any of the sentences that it might be expected to entail if it was linguistically identical to its performative counterpart, an assumption required by the performative hypothesis. It is impossible to argue eat more fish\ Therefore, 1 command you to eat more fish, or therefore, someone commands you to eat more fish, or therefore, you ought to eat more fish. Nor can any of these declaratives entail the imperative, as shown in (10). An imperative cannot even entail itself. It cannot be argued eat more fish! Therefore, eat more fish!. Given that direct imperatives never participate in entailment, while overt performatives always do (they always entail themselves), the claim that imperatives are always embedded under performatives, as the covert performative hypothesis claims, is problematic. 7 2.3.2 J.D. Fodor (1977) An approach similar to Sadock's is that of J.D. Fodor (1977), in which the performative is present not in the syntax, but in the logical form of the sentence. Fodor classifies as imperatives all sentences that are normally used to give orders, give instructions, etc. Specifically, she treats performatives, like the sentences in (2), as imperatives when it comes to semantic interpretation, and assumes that the ability to use such sentences as declaratives is derived. Assuming that the semantic component of the grammar interprets the syntactic component, and assuming that truth values are computed in the semantic component to yield the meaning of declarative sentences, Fodor (1977) comes up with three hypotheses about how the meaning of imperative sentences could be computed. i. Imperative sentences have truth values ii. Imperative sentences get their meaning from a function to a related declarative sentence iii. Instead of truth values, imperatives have some other semantic property P, for which the analogy truth : declarative as P : imperative holds. Let us consider the possibilities one by one. Imperative sentences have truth values If imperative sentences have truth values just like declaratives do, computing the meaning of an imperative sentence reduces to the task of computing the meaning of a declarative. This approach has the fatal problem that imperatives usually cannot be interpreted as statements of which truth holds, which can be seen from the fact that 'true' is not an appropriate response to an imperative. ii. Imperative sentences get their meaning from a function to a related declarative sentence The second approach holds that although imperatives have no truth value of their own, every imperative is related to a declarative that does, and gets its meaning through the truth conditions of the declarative1. Fodor notes that this approach assumes that the syntax feeds two separate pieces of information to the semantic component: the illocutionary force of the sentence, and the propositional content. Presumably, the semantic component first decides, based on the illocutionary force, which declarative sentence to compute truth conditions for, and then uses the propositional content of the sentence as the basis for the computation. Under this assumption of how the semantic component works, illocutionary force and propositional content are separate, and their interaction is limited. The propositional meaning is determined solely by the propositional content. We expect a propositional property like anomaly to be consistent across different illocutionary forces. It turns out that this is not the case. As Fodor points out, the anomalous sentence / command that the green vase broke yesterday is (in Fodor's covert performative analysis) interpreted as a simple pairing of the propositional content of the sentence the green vase broke yesterday and the specification that the sentence is a command. The anomaly cannot be based in the propositional content, which can be seen from the fact that the green vase broke yesterday is grammatical. We therefore cannot simply treat anomaly as a 1 A recent version of this idea is found in Han (1998), who connects the meaning of an imperative sentence to a set of propositions in the addressee's plan set. The meaning of the propositions in the plan set can be obtained by determining their truth conditions, so the meaning of the imperative would be obtainable in this system by a function from the meanings of the related propositions in the addressee's plan set. 8 characteristic of the proposition, but it must be a property that emerges when we compose the proposition and the illocutionary content. This undermines the model of the semantic component that we have just discussed, which leads Fodor to reject the assumption that the input to the semantic module consists of the proposition and its illocutionary force in separate packets. iii. Instead of truth values, imperatives have some other semantic property P.for which the analogy truth -.declarative as P : imperative holds Fodor settles on this method for determining the meaning of imperatives. Her particular implementation of it involves covert performatives. Both imperatives and declaratives are preceded by a covert performative. Finish your hamburger has the semantic form [I order you to] finish your hamburger, while the declarative / order you to finish your hamburger has the semantic form [1 assert that] I order you to finish your hamburger. The result is that a similar compositional device is used for all illocutionary forces. Declaratives have the special property of truth values which yield truth conditions, while imperatives instead have something like obedience conditions, which form the basis of the computation of imperative meanings. This solution, however, leads to the same problem that plagues Sadock's (1974) theory of covert performatives - it makes the wrong predictions for entailment, as Fodor recognized. Consider the following syllogism. (11) All men are mortal Socrates is a man Socrates is mortal This is a valid entailment. But according to Fodor's semantic hypothesis, the sentences above, because they are assertions, are actually preceded by / assert that in their semantic representations, which leads to the following false entailment: (12) I assert that all men are mortal I assert that Socrates is a man #1 assert that Socrates is mortal Another problem arises with sentences like the following (from Sadock 1974), which are ambiguous between imperatives and interrogatives by our criterion. (13) A: Can you bury the turtle? (a) Would you be willing to bury the turtle? (b) Will you bury the turtle? (c) B: Okay/yes/*I agree/*correct/*true If Fodor is to consistently treat indirect imperatives as imperatives and have their other uses derived pragmatically, she would need to include the sentences in (13) in her list of imperatives and derive their interrogative interpretation. It seems unlikely that it is some pragmatic effect and not the syntax that determines that the sentences in (2) can be used as statements and those in (13) as questions. 9 To conclude, it is Fodor's assumption that overt performatives enter the semantic module marked as imperatives that causes her hypothesis (ii), that imperative sentences get their meaning from a function to a related declarative sentence, to go awry. I claim that hypothesis (ii) can be made to work, i f we make the following assumption: imperatives like the ones in (1), which cannot be interpreted as anything other than imperatives, have their force marked in the syntax. Imperatives that are ambiguous between declaratives and imperatives, or between interrogatives and imperatives, are not marked as imperatives in the syntax and must be converted into imperatives by some pragmatic process. Considerations of relevance in the discourse context can be used to distinguish the use of the sentences in (2) as assertions from their use as commands, and the use of the sentences in (13) as inquisitive questions from their use as indirect requests. For instance, in the context of a job interview, it is assumed that information about the applicant's ability and willingness to perform certain tasks is needed by the interviewer, so the questions in (13) are interpreted as requests for information. If a zoo worker is asked one of the questions in (13) by her supervisor who has just informed her that the turtle has died, and if burying dead animals is part of the worker's job, it can be assumed that the supervisor knows that the worker is capable and willing to bury the turtle, and the questions in (13) are properly interpreted as requests. A further problem that none of Fodor's three approaches can rectify is the same problem that Sadock's theory could not handle - the fact that in English, the presence of subjects in direct imperatives is optional, while their presence in declaratives and interrogatives, including indirect imperatives, is obligatory. To summarize this section, we have seen the pitfalls of two hypotheses that attempted to explain the distinction between direct and indirect imperatives as a difference in the overtness of a higher performative clause. These approaches cannot account for the differences in logical properties of the two types of imperatives or for the fact that subjects are usually absent or optional in direct imperatives, but in many languages obligatory in declaratives, including indirect imperatives. We conclude what we proposed in 2.2 - that the best approach to the direct/indirect distinction is in the syntax, where direct imperatives are of an imperative form that is distinct from interrogatives and declaratives, and that the instruction or request element of the meaning of indirect imperatives comes from contextual pragmatic implication. 2.4 True and suppletive imperatives: a further distinction within the class of direct imperatives In sections 2.2 and 2.3 we have discussed the distinction between direct and indirect imperatives and we have seen the perils of trying to lump the two together in one syntactic class, as the covert performative hypothesis attempts to do. Unable to come up with a system that accounts for the semantic and logical properties of this syntactic class, we rejected the single imperative hypothesis. In this section, we will show that there is another division within the class of direct imperatives - the distinction between true and suppletive imperatives. This distinction is not found in English, but is present in many other languages, including Italian, Spanish, Modern Greek, Romanian and Hebrew. I will introduce the distinction using Hebrew data, after which I will consider the question of whether the true/suppletive distinction is the same as the distinction between direct and indirect imperatives. I will conclude on the basis of morphology and pragmatics that they are not the same, and that true and suppletive are two types of direct imperative. 2.4.1 Differences between true and suppletive imperatives in Hebrew Hebrew imperatives fall into three classes according to the morphological form of the highest verb. True imperative verbs have a unique form, which is typically uninflected or inflected only 10 with a class-marking affix or agreement. Future imperatives have the same form as the appropriate second person future tense verb. Infinitive imperatives have the form of infinitive verbs. The last two classes are collectively referred to as the class of suppletive imperatives, following the literature on imperatives in certain Romance languages, which exhibit a distinction between imperative verbs with unique morphology and imperative verbs that use the morphology normally associated with another sort of function (e.g. Han 1998, Rivero 1994, Zanuttini 1994, 1997). True imperatives (14) xake! "Wait!" Future suppletive imperatives (15) a. texake! "Wait!" b. Ata texake yamim, aval hu lo yavo you will-wait days but he neg come "You will wait for days, but he won't come" Infinitive suppletive imperatives (16) a. na lexakot please to-wait "Please wait" b. ani muxan lexakot I prepared to-wait "I am prepared to wait" The literature on true and suppletive imperatives (Rivero 1994, Zanuttini 1994, 1997, Han 1998) notes three primary syntactic differences between the two types. True imperatives typically cannot be negated, cannot be embedded, and precede pronominal clitics. In contrast, suppletive imperatives can be negated and embedded and occur following pronominal clitics, like finite verbs do. Hebrew pronominal clitics do not occur between the subject and verb as they do in the Romance languages, so the last property is not relevant for Hebrew. Like Romance imperatives, Hebrew true imperatives are not negatable or embeddable, while suppletive imperatives are. (17) a. *lo/*al/*eyn xake! neg (3 forms) wait b. al texake! neg wait "Don't wait" c. lo lexakot! neg wait "Don't wait" 11 (18) a. *civiti (se-)xake! I-commanded that-wait b. civiti se-texake! I-cornmanded that-wait "I commanded you to wait" c. civiti otxa lexakot! I-commanded acc-you to-wait "I commanded you to wait" 2.4.2 True/suppletive vs. direct/indirect This correlation of syntactic and morphological properties across languages and language families (Hebrew is a Semitic, not a Romance language) suggests that there is a grammatical distinction between true and suppletive imperatives. In the context of our discussion above, it makes sense to ask whether this distinction corresponds to the direct /indirect distinction. At first sight, it might seem that Sadock's (1974) conclusion identifying suppletive imperatives with indirect imperatives is justified, because of the contrast with true imperatives, which exhibit a set of unique morphological and syntactic properties. A closer investigation, however, reveals that Hebrew suppletive imperatives are actually direct rather than indirect. First, Hebrew has indirect imperatives like the English verbs in (2) or (3). (19) a. ani mecave otxa likbor et cav hayam I command you to-bury acc turtle the-sea "I command you to bury the turtle" b. ani roce se-tikbor et cav hayam I want that-you-will-bury acc turtle the-sea "I want you to bury the turtle" c. ani mevakes se-tikbor et cav hayam I request that-you-will-bury acc turtle the-sea "I request that you bury the turtle" d. ata muxan likbor et cav hayam? you prepared to-bury acc turtle the-sea "Are you prepared to bury the turtle?" e. ata yaxol likbor et cav hayam? you able to-bury acc turtle the-sea "Are you able to bury the turtle?" To any of the sentences in (19) beseder "okay" is an acceptable response, along with naxon "correct" for (19a-c), and ken "yes" for (19d-e). These sentences are exactly like the ambiguous English ones. They can be interpreted as imperatives, or as statements about the speaker's performance (19a,c), the speaker's attitude (19b), or the addressee's willingness (19d) or ability (19d,e) to perform the action of burying the turtle. 12 The question is whether suppletive imperatives fall into the same category as these indirect imperatives. For example, compare the English example in (20a) with the Hebrew example in (20b). (20) a. You will leave town tomorrow b. (ata) ta?azov et ha?ir maxar (you) leave.fut acc the-city tomorrow Both the English and the Hebrew sentences are ambiguous between a command and a prediction. I have claimed that the English sentence is syntactically marked as a declarative and that an imperative interpretation can be derived pragmatically. As for Hebrew sentences like (20b), there are two possible hypotheses. One is that taPazov in (20b) is a declarative verb, and that the sentence can derive its imperative interpretation pragmatically. The other hypothesis is that the two interpretations in (20b) are the result of generating separate syntactic outputs marked with different illocutionary force. To put the question in other terms, the issue is whether it is appropriate to gloss (20b) you will leave town tomorrow, and let the pragmatics derive the imperative interpretation, or to give it two separate glosses, you will leave town tomorrow and you leave town tomorrow] There are two reasons why the Hebrew sentence is different from the English one, and should receive two glosses, indicating different syntactic outputs. The first has to do with the pragmatic differences between the imperative interpretations of the sentences in (20). The second has to do with syntactic differences between the two interpretations of (20b). The imperative interpretation of the English sentence in (20a) has a pragmatic effect. If A tells B you will leave town tomorrow and intends it as an instruction, there is an implied hostility or threat. In addition to an instruction to B to leave town, the future tense in (20a) suggests that A has control over the situation and can make sure that B actually does leave town, whether he wants to or not - an interpretation that would be regarded by B as hostile or threatening. For instance, (20a) could be uttered by a powerful mafia boss to his less powerful rival, but would sound strange if it were uttered by the less powerful rival to the more powerful boss. In a syntactically marked imperative, on the other hand, there is no implication of threat or hostility. Leave town tomorrow could be uttered by the less powerful mafia boss to the more powerful one, if it is offered as friendly advice. To draw a clear scenario where a direct imperative is appropriate and an indirect imperative is inappropriate, consider a situation where the less powerful mafia boss learns that the police are planning a mass arrest of suspected gang members the following day, and wishes to pass the information on to the more powerful mafia boss. In this situation, it would be appropriate for him to say (21a) but not (21b). (21) a. Leave town tomorrow morning. The cops are carrying out mass arrests at 2 pm. b. ?You will leave town tomorrow morning. The cops are carrying out mass arrests at 2 pm. The Hebrew sentence in (20b) is appropriate in the scenario just described, if the subject is omitted, but not appropriate if the subject is present. I interpret this fact to mean that the future imperative with an overt subject is a declarative sentence that receives an indirect imperative interpretation, which comes together with the threatening/hostile reading, which makes it inappropriate in the scenario just described. When the subject is absent, the future imperative is direct, and does not have to be interpreted as threatening/hostile, and is therefore acceptable in our scenario. 13 There is also morphological evidence that the two interpretations for (20b) are associated with separate syntactic structures. The Hebrew sentence in (20) can be negated in two ways. (22) a. al ta?azov et ha-?ir maxar neg leave acc the-city tomorrow "(You) don't leave the city tomorrow!" b. Ata lo ta?azov et ha-?ir maxar You neg leave acc the-city tomorrow "You will not leave the city tomorrow" The sentence in (22a) can only be interpreted as an imperative. The sentence in (22b) has the two interpretations associated with the English sentence in (20a). It can be a declarative, or an imperative with the threatening/hostile interpretation. In our scenario, if the less powerful mafia boss found out that the police were not going to conduct the sweep after all, he could utter (22a), but not (22b). This is consistent with our earlier conclusion that future imperatives could be direct imperatives, in which case they would be subjectless, and as (22a) shows, will be negated with the marked negative al. (22b) is a future declarative which can be used as an indirect imperative, and has the threatening/hostile interpretation. These two pieces of evidence show that Hebrew sentences with future tense are ambiguous in the syntax between a declarative and an imperative. In other words, direct imperative sentences, in addition to true imperative verbs, can contain verbs that look like future tense or infinitive verbs. True and suppletive imperatives are both in the class of direct imperatives. 2.5 Summary In this chapter we have discussed two distinctions that are found in imperatives: the distinction between direct and indirect imperatives, and the distinction between true and suppletive imperatives. We have seen that direct imperatives are marked as such in the syntax, while indirect imperatives are not, and we have speculated that indirect imperatives obtain their command interpretation via pragmatic inference. True imperatives and suppletive imperatives are two different types of direct imperatives, which differ in their morphology and syntax: only true imperatives are associated with purely imperative morphology and syntax, while suppletive imperatives "borrow" their morphology from other verbal paradigms. 14 I. 15 Chapter 3 - Basics of Verb Movement 3.1 Introduction Before we move on to an investigation of the phrase structure of imperatives, we need to take a look at the nature of verb movement, which will be relevant in the course of our investigation of phrase structure, and the topic of chapters 5 and 6. In this chapter, we will look at two topics related to verb movement. In section 3.2 we will look at the motivation for movement. In 3.3 we will examine the principal constraint on head movement, the Head Movement Constraint (HMC) (Travis, 1984), and Baker's (1988) reduction of the H M C to the Empty Category Principle (ECP). In 3.4 we will look at an exception to the H M C involving what appears to be verb movement past a negative head, and provide an argument, originally from Belletti (1990), that such a movement is consistent with the ECP. In the remaining chapters we will rely on the framework set out in this chapter to interpret verb movement in imperatives, and on Belletti's analysis in particular to explain the difference between two negative heads in Hebrew - one of which, I will argue, raises, and the other one of which does not. 3.2 Verb movement as feature checking The Minimalist Program (Chomsky 1995) takes movement to be driven by features. Specifically, movement is motivated by the need for a substantive category (i.e. N, V, A or P) to have one or more of its features checked by the head of a functional category (e.g. I, D, C). For instance, a verb must raise to 1° to have its categorial V-feature checked. An unchecked feature causes the derivation to crash. Features may be strong, in which case they need to move overtly in order to be checked, or weak, in which case they move non-overtly. I assume that movement leaves behind a trace which is connected to the moved element via a chain, which must meet various well-formedness conditions, such as the ECP, discussed in 3.3. 3.3 The Head Movement Constraint Travis (1984) notes that both verbs and nouns appear to move only into the category that governs them. She postulates the Head Movement Constraint (HMC), to account for this fact: (1) Head Movement Constraint (Travis 1984): An X° may only move into the Y° which properly governs it. Baker (1988) argues that the Head Movement Constraint is not an independent grammatical principle, but that is a generalization that follows from the Empty Category Principle (ECP), which states that traces must be properly governed. In order to do so, he first defines a number of terms, which he mostly models on the definitions of Chomsky (1981, 1986). Here is his definition of proper government. (2) Proper government (Baker 1988): A properly governs B iff A governs B, and A and B are coindexed. Baker assumes that a trace can be coindexed with either its antecedent or with a head that assigns it a theta-role. He gives the following definition of government. 16 (3) Government (Baker, 1988): A governs B iff A c-commands B and there is no category C such that C is a barrier between A and B. This definition makes use of the notion barrier, which Baker defines as follows. (4) Barrier (Baker 1988) Let D be the smallest maximal projection containing A. Then C is a barrier between A and B iff C is a maximal projection that contains B and excludes A, and either: (i) C is not selected, or (ii) the head of C is distinct from the head of D and selects some WP equal to or containing B. Baker's basic idea is that government for the trace of a moved X° cannot be head government, since no head can assign a theta-role to another head (this follows from the fact that theta-roles are assigned under a sisterhood relationship, and that two heads are prevented by the X-bar schema from being sisters). Thus if a head moves, the ECP cannot be satisfied by the trace of the movement being properly governed by a different head - it must be satisfied by the trace and its antecedent being coindexed. This means that the ECP for traces of X° can be simplified to the following: (5) ECP applied to trace of X° (Baker 1988): An X° must govern its trace. It remains to be shown that an X° governs its trace if and only if it is united with the Y° which governs the XP that X° originally headed. Government, as shown in (3), has two parts: A must c-command B, and there must be no intervening barrier. The structure created by head movement is the following: (6) YP Xi Y t| ZP Baker uses the following definition of c-command, from Aoun and Sportiche (1983), to show that X c-commands the trace in this structure. (7) C-command (Aoun and Sportiche 1983): A c-commands B iff A does not dominate B and for every maximal projection C, if C dominates A then C dominates B. The second half of government, the non-intervention of a barrier between the governor and the element governed, can be rephrased as a minimality condition, as was proposed by Belletti (1990). We can assume that Y is the head that governs XP in (6), and that any head that is higher up will be blocked by Y from being a governor. Another way of phrasing this condition on government is to say that Y must be the nearest potential governor for the trace. This phrasing 17 will be used in the next section, in which I discuss an instance where the generalization expressed in the HMC appears not to hold. 3.4 Verb movement across Neg° The following sentences from Italian and French show the verb occurring between two negative elements. The negative elements non and ne are considered clitic-like in that they always occur adjacent to a verb, and have been analyzed as Neg°s, while the words that are glossed as "anymore" are taken to be negative phrases in [Spec,NegP]. (8) Gianni non parla piu Gianni neg speak anymore "Gianni doesn't speak anymore" (9) Jean ne parle plus Jean neg speak anymore "Jean doesn't speak anymore" Several studies (Moritz 1989, Belletti 1990) have suggested that this word order is derived when Neg0 raises to the highest inflectional head (1° in our structure), after which the verb raises to the same head, skipping over the trace of Neg0. Using our assumptions, this movement is motivated by the need for Neg0 to check one of its features in 1°. The movement yields the following structure. (10) IP NP Gianni Jean NegP Neg0 1° piu Neg' non ^/ \^plus ne V° parla parle I tNeg VP V tv This analysis does not conform to the generalization behind the HMC, because the verb raises to a position which does not properly govern its trace - Neg0 In order to explain this exception, I will use Belletti's (1990) account of the HMC. Like Baker, Belletti assumes that the HMC reduces to the ^ECP. She follows Moritz (1989) in arguing that the ECP is a representational condition applying not to traces, but to chains formed by movement. Formulated as such, it states that a trace of a movement must be antecedent governed by a member of its chain. Neg0 counts as a member of the same chain as the trace of the moved V° because Neg0 and V° end up in the same head, as shown in (10), which causes their chains to merge (indicated by co-indexation) 18 (11) IP NP Gianni Jean 1° NegP Negi0 1° piu Neg' non /^^xplus ne V,° parla parle I tiNeg VP V t;v The trace of negation fulfills all of the requirements that make it an antecedent governor for the trace of V°, under the revised ECP. It is the nearest potential governor of the trace, it c-commands it, and it is a member of the same chain. Thus it can be claimed that the verb bypasses Neg0 and moves directly to 1 ° , while still respecting the ECP. 3.4.1 Hebrew and English The derivation described above, in which Neg0 and V° independently raise to 1 ° , was originally motivated in order to account for languages like Italian and French, in which negation acts like a verbal clitic by always appearing directly in front of the verb (Moritz 1989, Belletti 1990). In these languages, the negation that acts as a clitic has been analyzed as the head of NegP, while negative adverbs such as plus in French or piu in Italian have been treated as specifiers of NegP, or else adjuncts of NegP. Hebrew, one of the languages discussed in depth in this thesis, is similar to Italian and French in this regard, as demonstrated by the fact that sentences can have a negative preceding the verb and another negative adverb following it. (12) Yoxanan lo axal af-pa?am pica Yoxanan neg ate never pizza "Yoxanan never ate pizza" Shlonsky (1997) notes that the negative lo is clitic-like in the exact same sense that ne and non are - they always appear adjacent to the verb. This means that the analysis that verbs raise directly to 1° applies to Hebrew. The situation in English is a bit more complicated. There is no consensus on whether not is a Neg0 or a specifier of NegP. If it is a Neg0 (Potsdam 1996), then its non-clitic-like behaviour would mean that Belletti's analysis fails, and V° -> 1° movement cannot be assumed. That not is not a clitic that attaches to a verb can be seen from the fact that it is possible for a parenthetical to intervene between not and the main verb, and for an adverb to intervene between not and a preceding auxiliary verb. (13) a. John did not, as far as I know, propose to Judy b. John is certainly not married to Judy 19 On the other hand, if not is a specifier of NegP, the analysis that applies to Italian, French and Hebrew can apply to English as well. I will assume that this is the case, which means that verb movement is movement from V° to 1°. 3.5 Conclusion We have explained movement as required by the need for constituents to have their features checked in a higher category, and examined the specific case of head movement, noting that heads typically raise only to the head that governs them. Baker and Belletti both derive this fact from the ECP, and Belletti is able to use the ECP as an explanation for the case of Neg-raising to 1° followed by verb raising to the same head, which explains the clitic-like nature of negatives in Italian, French and Hebrew. 20 Chapter 4 - The Phrase Structure of Imperatives 4.1 Introduction So far, we have drawn a three-way distinction in imperatives. There are direct imperatives and indirect imperatives, of which only the former are marked in the syntax as being imperatives. Within the class of direct imperatives there is a division between true and suppletive imperatives. In this chapter I will explore the phrase structure of direct imperatives, reviewing two hypotheses. The first hypothesis is that an imperative is nothing more than a clause with a missing or nonovert subject. The other hypothesis, advanced by Platzack and Rosengren (1994) and by Pirvulescu and Roberge (1999), is that imperative clauses are not full clauses, but rather reduced clauses, consisting of a VP only (Pirvulescu and Roberge) or a VP and one more projection (Platzack and Rosengren). In this chapter, I argue against such an approach by showing that higher projections are required. I will provide evidence from English and Hebrew that imperatives in general require the following three phrases: (1) i. NegP ii. A functional projection, with properties that have been attributed to IP, that dominates NegP m CP Other positions will be motivated in the next chapter, which discusses verb movement, but in this chapter I would like to show that the projections listed in (I) are present in most kinds of imperatives. 4.2 Subjects and IPs In this section I consider the hypothesis that imperatives are clauses without subjects, rejecting it and showing that imperative subjects in English are in [Spec,IP], just like declarative subjects. The hypothesis that imperatives are declaratives minus subjects could mean two things. It could mean that there is no subject in the imperative structure at all, or it could mean that an imperative sentence has a non-overt subject which does not appear in declaratives. In 4.2.1,1 present an argument from Beukema and Coopmans (1989) to show that imperatives without overt subjects have covert subjects. In 4.2.2,1 present arguments from Potsdam (1996), who shows that the overt subjects of imperatives are the same syntactically as the subjects of declaratives, and specifically that they are located in [Spec,IP]2. These two analyses show that imperative subjects are pro when they are covert and regular subjects when they are overt. 4.2.1 Beukema and Coopmans (1989) Beukema and Coopmans (1989) argue that imperatives that do not have overt subjects are not subjectless, but rather have a covert subject, which is the antecedent for reflexive anaphors and the controller for certain types of dependent clause. That the covert subject is the antecedent is evident from the fact that the controlled or bound element is understood to be the same second person referent that the covert subject of the imperative is understood to be. 2 This is not to deny that the form of the overt NP subject of imperatives is restricted. English imperatives, for instance, cannot be indefinites or first person pronouns. I have no account of this fact, but I will assume that it is not relevant to the position of imperative subjects. 21 (2) a. Yoiij'll kick yourself! when you hear who was at the party b. Trevor kicked yourself c. *He kicked yourself! d. subji kick yourself! In sentences with without-clauses, the understood subject of the without-dausQ must be given identity by a controlling antecedent, as in (4a). In imperative sentences, the 2nd person covert subject is also understood as the controller of the subject of the without-clause, as in (4b). This implies the syntactic presence of a covert subject in imperatives. (3) a. Trevor, kicked Luther [without PRO, really hurting him] b. subji kick Luther [without PROj really hurting him] These arguments show that the claim that imperatives do not have any syntactic subjects is false. Let us consider the hypothesis that imperatives have a covert subject that is not found in declaratives. The agreement facts in (2) show that the empty subject has pronominal content. This means that it must be the covert pronoun pro, rather than any other empty category. But pw can appear in declaratives in many languages. The following examples are from Hebrew. (4) a. pro ba?atti be-Yoram kicked, lsg in-Yoram "I kicked Yoram" b. (ata/proj) tiv?at be-?acmexaj ksetisma mi haya bamesiba you kick.fut in-yourself when-you-hear.fut who was in-the-party i. "You'll kick yourself when you hear who was at the party" ii. "Kick yourself when you hear who was at the party!" c. b?at be-?acmexaj ksetisma mi haya bamesiba kick.fut in-yourself when-you-hear.fut who was in-the-party "Kick yourself when you hear who was at the party!" d. pro; ba?atti be-Zalman [bli PROj mamas lehax?iv lo] kicked, l sg in-Zalman without really to-hurt to-him "I kicked Zalman without really hurting him" e. proj tiv?at be-Zalman [bli PRO: mamas lehax?iv lo]! kick in-Zalman without really to-hurt to-him "Kick Zalman without really hurting him!" f. pro; b?at be-Zalman [bli PROj mamas lehax?iv lo]! kick in-Zalman without really to-hurt to-him "Kick Zalman without really hurting him!" (4a) shows that pro exists in declaratives in Hebrew. (4b) shows that a.covert or overt subject pronoun in a suppletive imperative clause can license an anaphor, and (4c) shows the same for true imperatives. (4d) shows that pro can control the subject of a without-clause in declaratives, (4e) shows that the same is true in suppletive imperatives and (4f) shows that the same holds of true imperatives. There is no reason to assume that there is more than one kind of pro in (4). We 22 therefore conclude that there is nothing special about the null subject of imperatives - imperatives are just another possible context for pro. This discussion raises a question for English: why can pro be the subject of an imperative but not of a declarative? Following Potsdam (1996), I assume that it follows from Farkas' (1987) Recovery Condition on pro. (5) Recovery Condition on pro (Farkas 1987) The feature [person] must be recoverable, i.e. there must exist some algorithm which yields its value As claimed by Potsdam, it is an inherent property of imperatives that they propose that the addressee bring about an event. The feature [2nd person] is recoverable in imperatives because the subject must be the addressee. Declaratives, by contrast, do not have an inherent value for [person], which is why in English pro cannot be the subject of a declarative. We see that imperative clauses without overt subjects actually have pro as a covert subject. In the next section, I consider overt subjects of imperatives, repeating a number of arguments from Potsdam (1996) that show they are syntactically similar to the overt subjects of declaratives. 4.2.2 Potsdam (1996) If we assume the VP-internal subject hypothesis (e.g. Sportiche 1988), there are two potential positions for overt subjects of imperatives that are consistent with the subject properties ascribed to them by Beukema and Coopmans (1989) - [Spec,IP] and [Spec,VP]. Proponents of hypotheses that imperatives are reduced clauses claim that subjects are in [Spec,VP]. However, there is evidence that overt imperative subjects are actually in [Spec,IP] in English and Hebrew. If this is correct, an IP is implied in imperatives. Potsdam (1996) presents a number of arguments in favour of his hypothesis that the overt subjects of imperatives are located in [Spec,IP], which makes them no different from the subjects of declaratives in this respect. His arguments are from the syntactic regularity of the subject of imperatives and from VP ellipsis. To demonstrate the syntactic regularity of overt imperative subjects, Potsdam lists a number of subject properties originally compiled by McCloskey (1997), and shows that they hold of imperative subjects almost without exception. Many of these properties are consistent with the subject's being in [Spec,VP], since a subject in that position would still be the highest argument in the clause and c-command the other arguments. I will focus on two properties that appear to be related specifically to the position [Spec,IP]: negative polarity licensing and subject-deriving grammatical operations. Negative polarity licensing Among the properties that Potsdam lists as showing that subjects are more prominent than other arguments is the property of licensing negative polarity items in other argument positions, while not allowing a negative polarity item licensed by other arguments. Negative polarity licensing has been argued to be a c-command relationship (e.g. Progovac 1994, Uribe-Echevarria 1994), and since both [Spec,IP] and [Spec,VP] c-command all non-subject argument positions, this test does not distinguish between the two potential subject positions. However, besides licensing negative polarity items in argument positions, negative arguments like nobody can license negative polarity adverbs like ever. Since ever is a temporal adverb, it is outside the VP and falls outside 23 the scope of the element in [Spec, VP]. If the subject of imperatives is in [SpecJP], it should c-command and license ever. If the subject is in [Spec,VP], it should not c-command ever and therefore not license it. The direct object, which does not c-command outside of the VP, should also fail to license ever. (6) a. Nobody ever kick Brutus! b. *You ever kick nobody! (7) a. Nobody ever kicked Brums b. *You ever kicked nobody (6a) shows that the subject cannot be in [Spec,VP]. The contrast between (6a) and (6b) is precisely what we expect if imperative subjects are in [SpecJP], like they are in declaratives. The parallel between (6) and (7) confirms the similarity of declaratives to imperatives in this respect. Grammatical operations that derive subjects The following is a list of different movements in which subjects move to sentence-initial positions from lower positions. Most or all of these movements have been argued to be necessary for case assignment to the subject, which is assumed to take place in [SpecJP]. (8) i. Passive ii. Subject to subject raising iii. Unaccusative advancement iv. 7owg/?-movement v. Raising from predicate-internal position (evidence from floating quantifiers) As the following data show, imperatives with overt subjects are compatible with these movement rules3'4. (9) Passive Do/don't everyonei be examined t; by the doctor! (10) Subject to subject raising Do/don't everyonei stop ti writing when the bell rings! (11) Unaccusative advancement Do/don't everyonei come ti quickly! 3 Where possible, I have inserted don't in front of the subject everyone in order to ensure that the subject is not dislocated. Without don't, it is possible to argue that there is an intonational pause between the subject and the rest of the sentence, and that we might be dealing with an atypical subject in a non-[Spec,IP] position. The presence of don't makes it very difficult or impossible to separate the subject from the rest of the sentence by an intonational pause. It does leave open the possibility of focus intonation on the subject, however. 4 Zhang (1990) argues that don't is not a head that moves to precede the subject, but rather a construction-specific negative that is adjoined to TP, which he regards as the highest functional projection (equivalent to our LP). I argue against his claim in the appendix to this chapter. 24 (12) Tough-movement Do/don't everybody! be easy for us to spot ti! (13) Raising from predicate-internal position Rhett;, Scarletj and Lassiek all tyk get ready for the next scene! The simplest description of these movements is to assume that imperative subjects are subjects like any other, and raise to [Spec,IP] from whatever lower position they are generated in. It is commonly assumed that the motivation for many of these movements is the need to be assigned case, which is unavailable in the lower position but available in [Spec,IP], If the lower positions cannot be assigned case in declaratives, it is unlikely that they can in imperatives. Therefore if [Spec,IP] is unavailable then case is unavailable, and we expect sentences like (9)-( 13) to be unacceptable for case reasons. Since they are acceptable, there must be a case-assigning 1°. VP Ellipsis Potsdam brings further evidence from VP ellipsis that imperative subjects are in [Spec,IP]. VP ellipsis allows for the omission of an entire VP (Lobeck 1995). Therefore if an imperative subject can appear in the same clause as an ellipsed VP, it must be outside the VP. This is possible, as the following examples show. (14) a. Rick walked out of the lecture, but don't everyone else [e], please! b. Billy didn't tell mom what I did, so don't YOU [e], either! We see that imperative subjects are located outside of VP, which entails a functional projection outside VP. To summarize this section, our set of diagnostics show that the position of overt imperative subjects in English is i) a position that c-commands an adverbial outside of VP, and ii) the position that transformations place derived subjects in. The natural conclusion is that subjects are in [SpecJP], which supports the claim made at the beginning of this chapter that imperatives contain IPs. 4.3 NegP and IP in Hebrew This section aims to show that NegP and IP exist in Hebrew imperatives as well, which supports the notion that the phrase structure of imperatives is universally similar between languages, and similar to declarative phrase structure within languages. The arguments will be from agreement blocking and from a subject/object asymmetry in embedded negative quantifiers. 4.3.1 Agreement blocking Agreement blocking in declarative sentences shows that Hebrew has a NegP sandwiched between an IP and a VP. The argument will be structured as follows: it will be shown that present tense verbs and other verbs with identical morphology, known as benoni (following, e.g., Shlonsky 1997), are marked for gender and number agreement, but not person agreement, while past and future tense verbs are marked for person as well as gender and number. We will then see that in finite contexts where the main verb cannot raise, they can only be in the benoni form. We will assume, along lines similar to Shlonsky (1997), that finite verbs must raise to 1° to license person 25 agreement and past and future morphology, while other tense and agreement morphology can be licensed in situ. We will note that the negative word eyn occurs only with benoni main verbs, which will lead us to the conclusion that it blocks verb movement, preventing the verb from reaching 1°. This leads us to adopt the order IP > NegP > VP for Hebrew declaratives. Given the fact that non-benoni verbs with morphology identical to future tense finite verbs occur in future imperatives, and the fact that finite tense verbs must move to 1° to have their morphology licensed, it follows that future imperatives must move to 1° for the same reason. Thus we have evidence for 1° in imperatives. Given the fact that imperatives are negated, we have evidence for NegP. The sentences in (15) show simple past and future tense on the verb. The relevant morphology is highlighted. (15) a. ani axalti ugiot I cookies "I ate cookies" b. ata axalta ugiot cookies "You ate cookies" c. hu axal-0 ugiot he cookies "He ate cookies" d. ani ?oxal ugiot I eat. 1 cookies "I will eat cookies" e. ata toxal ugiot cookies "You will eat cookies" f. hu yoxal ugiot he cookies "He will eat cookies" Compare this with the present tense, in which person is not distinguished. (16) a. ani oxel ugiot I cookies "I eat/am eating cookies" b. ata oxel ugiot you.m.s cookies "You eat/are eating cookies" c. hu oxel ugiot he cookies "He eats/is eating cookies" 26 The sentences in (15) and (16) can be negated by placing lo in front of the verb (shown in 17), but sentences with the negative eyn can occur only in the present tense (shown in 18). (17) a. ani lo axalti ugiot I neg eat. 1 cookies "I didn't eat cookies" b. ata lo axalta ugiot neg cookies "You didn't eat cookies" c. hu lo axal ugiot he neg cookies "He didn't eat cookies" d. ani lo oxal ugiot I neg cookies "I won't eat cookies" e. ata lo toxal ugiot neg cookies "You won't eat cookies" f. hu lo yoxal ugiot he neg cookies "He won't eat cookies" g. ani lo oxel ugiot I neg cookies "I don't eat/am not eating cookies" h. ata lo oxel ugiot you.m.s neg cookies "You don't eat/aren't eating cookies" i. hu lo oxel ugiot he neg cookies "He doesn't eat/isn't eating cookies" (18) a. *Ofereyn-o axal ugiot Ofer cookies b. *Ofer eyn-o yoxal ugiot Ofer cookies c. Ofer eyn-o oxel ugiot Ofer cookies "Ofer doesn't eat/isn't eating cookies" 27 Shlonsky points out several contexts besides eyn in which only verbs or participles in benoni form occur, to the exclusion of past or future verbs. The contexts are: following an auxiliary (19a,b); in a small clause complement of a verb of perception (19c); in an adjunct gerund (19d); and in a semirelative (19e). The examples in (19) are all from Shlonsky (1997), chapter 2. (19) a. hayeladim hayu tofrim smalot the-children were sewing dresses "The children were sewing dresses" b. hayeladim yihiyu tofrim smalot the-children will-be sewing dresses "The children will be sewing dresses" c. ra?iti et hayladim tofrim smalot I-saw acc the-children sewing dresses "I saw the children sewing dresses" d. hayladim yasvu baxeder lo?asim mastik the-children sat in-the-room chewing gum "The children sat in the room chewing gum" e. hine is haxosev al kesef here man the-think about money "Here is a man who thinks about money" Shlonsky notes that the contexts in which only benoni verbs can occur either have a head above the VP, as in the case of eyn and auxiliary constructions, or occur in minimal types of clauses in which there is no independent tense. He then connects the agreement and tense morphology on past and future tense verbs with a functional projection that dominates both NegP and VPA ux, the head where he places auxiliaries. He identifies the head with AgrS 0, but in the present analysis it is 1° The facts in (17-19) now follow from a movement analysis in which verbs must raise to 1° in order for person morphology and past and future tense morphology to appear. Neg 0 is intermediate between V° and the target of the verb's movement. When the sentential negative is lo, the verb can reach its target. When the negative is eyn, verb movement is blocked. The facts about morphology and tense interpretation follow when the verb cannot reach 1°. I account for the difference between lo and eyn by assuming that Belletti's (1990) analysis of negative clitics (discussed in chapter 3) applies to lo, but not eyn. In Belletti's analysis, lo raises to the highest functional head, and the verb raises there independently. As a result of lo and the verb being in the same head, their chains are merged, which allows them to respect the Empty Category Principle. On the other hand, eyn does not raise, which means that if the verb raises past it, no chain merger takes place, and the ECP is violated. As a result, eyn is only compatible with verbs that do not need to raise past Neg 0 to check their morphology, i.e. benoni verbs. This analysis, which explains why eyn is possible only in present tense sentences, depends on the assumption that aside from the VP, there is a NegP and a higher phrase to which the verb moves to license agreement and tense morphology. At least in declaratives, then, Hebrew has a NegP intervening between IP and VP. 28 4.3.2 NegP and IP in Hebrew imperatives This section presents an argument that IP and NegP are present in imperatives. The argument for IP is that morphology for person agreement and future tense must be checked in IP, and such morphology is present on future imperatives (shown in (20)), which shows that they must raise to 1° to license the morphology. As in the last section, true imperatives cannot be used as diagnostics because main clause negation is required, but we can use future and infinitive suppletive imperatives. (20) a. toxal et hapica! acc the-pizza "Eat the pizza!" b. toxli et hapica! acc the-pizza "Eat the pizza!" c. toxlu et hapica! acc the-pizza "Eat the pizza!" d. toxalna et hapica! acc the-pizza "Eat the pizza!" The evidence for NegP derives from the fact that infinitive and future imperatives can be negated: infinitives are negated by lo, which we have seen is in Neg0. (21) lo lehikanes! neg to-enter "Do not enter!" Future imperatives are negated by al: (22) al tikanes! neg enter "Do not enter!" Like lo, al must be adjacent to the verb, which suggests that the raising that Belletti (1990) suggests for Neg0 applies to al as well. It follows that it, too, is in Neg0. We thus have evidence for NegP in imperatives. 4.3.3 Subject/object asymmetry Another piece of evidence for NegP in Hebrew imperatives comes from subject/object asymmetries like in (23), in which a negative quantificational subject of a clause embedded under a negative clause is unacceptable, while a negative quantificational object is acceptable. 2 9 (23) a.*lo lehagid se-af exad hirbic le-Yoram neg to-say that-nobody hit to-Yoram b. ?lo lehagid se-Yoram hirbic le-af exad neg to-say that-Yoram hit to-nobody "Don't say that Yoram hit anybody" c. *al tagid se-af exad hirbic le-Yoram neg say that-nobody hit to-Yoram d. ?al tagid se-Yoram hirbic le-af exad neg say that-Yoram hit to-nobody "Don't say that Yoram hit anybody " The standard account for negative asymmetries like those in (23) (e.g. Kayne 1981) involves assimilating subject/object asymmetries in negatives to subject/object asymmetries in wh-words, like the one in (24). (24) a. *Who do you deny that kicked Jake? b. Who do you deny that Jake kicked? The asymmetry in (24) is usually explained using the Empty Category Principle, the condition on traces of movement introduced in chapter 3, which states that traces must be properly governed. The ECP has had many different formulations over the years, but a consistent idea has been that it predicts an asymmetry between the subject position, which in many cases is ungoverned, and the object position, which is always governed. As a result, the ECP often introduces a subject/object asymmetry because the trace of movement is left ungoverned when the subject raises, but is governed when the object raises. Kayne's (1981) idea was that the negative subject/object asymmetry in (23) is a consequence of the ECP. Unlike questions, however, negative quantifiers do not raise visibly. If Kayne's analysis is correct, negatives must raise covertly at LF. In recent years, a number of researchers (e.g. Haegeman 1995, Acquaviva 1997) have argued that what motivates the movement of the negative is the need for negative quantifiers to be associated with the NegP of a higher clause in a negative concord language. In such languages, sentential negation cannot exist independently of NegP. As a consequence, every sentential negative, including negative quantifiers, must be associated with a NegP at LF. Whether this association requires the negative to be in a specifier-head relation with Neg0, as Haegeman (1995) argues, or merely requires it to be governed by an element in NegP, as Acquaviva (1997) argues, the subject or object quantifier must raise from its surface position, giving rise to the contrast in (23). This analysis, which assumes a NegP, predicts that such asymmetries arise in negative concord languages, in which multiple morphological negatives do not have the force of multiple logical negatives, but not in double-negation languages, in which multiple morphological negatives do have the force of multiple logical negations. Under this prediction, it is expected that Hebrew would be a negative concord language, contrasting with English, which should be a double negation language. Examining the meanings of sentences with multiple negations, we find that these predictions are correct. Thus, (23b,d) are instances of a single negation, and are glossed "don't say Yoram hit anybody". If this was a double-negation construction, we would expect the gloss "don't say Yoram hit nobody", in which the two negations cancel each other out. This can be seen more easily with the verb think in a sentence like / don't think you met nobody is equivalent to / think you met somebody. 30 This analysis assumes that the negation in the higher clause is associated with a NegP, which provides a mechanism for the movement and association of other negative elements in the sentence to the matrix negation. A negative projection allows a feature-checking relationship between the negative head and the negative quantifiers, which drives the raising. Thus the fact that the negative imperatives in (23) show the subject/object asymmetry associated with raising to NegP is further evidence for the projection of NegP in suppletive imperatives. 4.4 Conclusion We have explored two hypotheses that claim that the phrase structure of direct imperatives is a substructure of declarative phrase structure. The first, which claims that imperatives are clauses lacking overt subjects, was shown to be incorrect using evidence from both English and Hebrew, because apparently subjectless imperatives have the same binding and control properties as declaratives with overt subjects, because constructions which require raising of the subject to [Spec,IP] in declaratives are grammatical in imperatives with overt subjects, and because overt subjects of imperatives can co-occur with VP ellipsis, hence are outside the VP. We then addressed the argument that imperatives are reduced clauses, specifically addressing Platzack and Rosengren's (1994) claim that they are VPs plus one functional head. In English, we have seen that these structures are inadequate for explaining the fact that subjects are quite regular syntactically, and appear to be in [Spec,IP], not [Spec,VP] even when they are preceded by do or don't. We have seen that Hebrew imperatives contain IPs, since the imperative verb forms have agreement morphology that depends on the verb's movement to IP. We have seen that NegP also provides for the best account for the subject/object asymmetry in negative quantifiers of complement clauses in imperatives - one which allows the asymmetry to be explained by an already existing principle, the ECP, and which predicts that the asymmetry is limited to negative concord languages. Our overall conclusion about the phrase structure of imperatives is the same as Potsdam's: they contain VP, NegP (if the sentence is negative), IP and CP, just like declaratives. This conclusion, however, applies more strongly to English imperatives and Hebrew suppletive imperatives than to Hebrew true imperatives, because the evidence for the latter is weak. Although we have seen that true imperatives have pro subjects and show overt person, gender and number morphology, both of which are evidence for an IP projection, the fact that true imperatives are incompatible with negation has made it impossible for us to apply the subject/object asymmetry test to them. If it could apply, we would expect to find the same asymmetry that we found in Hebrew suppletive imperatives. 31 Appendix to chapter 4 - Don't as a Particle Zhang (1990) argues, contrary to the conclusion in chapter 4, that don't is a negative imperative particle. We will examine his hypothesis here and argue that it is incorrect. Zhang begins by pointing out that other languages, such as Indonesian, can be analyzed as having a negative imperative element distinct from standard negation. In Indonesian, this can be seen from the form and position of the imperative negative, jangan, in (la,b), compared to the standard negative, tidak, shown in the declarative in (lc,d). (1) a. jangan (kamu) buka pintu itu! neg you open window the "Don't (you) open the window!" b. *kamu jangan buka pintu itu! (datum implied by Zhang 1990, p. 84) c. Bill tidak makan nasi itu Bill neg eat rice that "Bill didn't eat the rice" d. *tidak Bill makan nasi itu neg Bill eat rice that "Bill didn't eat the rice" The two negatives differ in form and in location relative to the subject: jangan must precede subjects while tidak must follow them. In addition, the two can co-occur in imperatives, yielding a double negation interpretation . (2) jangan (kamu) tidak makan nasi itu neg you neg eat rice the "Don't (you) not eat that rice!" Zhang assumes that in other languages, standard negation alone can be used to negate imperatives. He raises the question of which category English falls in, and concludes that English can negate imperatives either with the standard negation do not, or with the construction-specific don't. Rather than the sequence don't you being a result of an auxiliary don't moving past the subject, as is commonly assumed to be the case in interrogatives and negatives like in (3), Zhang proposes that don't is base-generated adjoined to TP. (3) a. Don't you care about me? b. Never had he felt so alone. This analysis is a priori less desirable than the alternative because it involves a construction specific rule, which adds to the complexity of the theory of grammar. Therefore it must be supported by strong empirical evidence. To support his hypothesis, Zhang brings three 32 arguments: the argument from topicalization, the argument from phonology, and the argument from conjoined subjects. I present and respond to each of his arguments in turn. The argument from topicalization Zhang, following Pollock (1989), splits IP into TP and AgrP, with TP being the highest inflectional head. Zhang cites Lasnik and Saito (1992) on location of topics. In order to account for the data in (4), they claim that topics are adjoined to TP. (4) a. I think that this book, you don't want to read b. *I think this book, that you don't want to read A topic can follow but not precede that. Since that is taken to be in C°, which takes TP as its complement, and since topics precede the subject, understood to be in [Spec, TP], there is only one position where the topic may be found, and that is in a position adjoined to TP. Zhang notes that this makes the wrong predictions for a subject-auxiliary-inversion analysis of don't: (5) a. *Don't that present, open until next week! b. That present, don't open until next week! If don't is in C°, as the auxiliary analysis predicts, the sentences in (5) should be parallel to those in (4), in which that is in C°. Instead, we find the opposite pattern of grammaticality. Zhang concludes that don't must be in the only position available between an adjunct of TP and [Spec, TP], where the subject is located; namely, in another adjunct of TP. There is a serious challenge to this argument from interrogatives. In interrogatives, it is uncontroversial that don't is an auxiliary that raises to C°, so if Zhang's approach is correct, we would expect interrogatives to pattern differently from imperatives. If we make the same assumption that Zhang made for imperatives - that topics are adjoined to TP - we predict incorrectly that don't should precede topics and not vice versa. (6) a. *Can't that classic novel, you read by next week? b. That classic novel, can't you read by next week? One of our assumptions - that don't is in C°, or that topics must be lower than C°, must be incorrect. The second assumption seems to be more susceptible to a challenge, and in fact it has been challenged by a couple of different approaches. The first is an analysis in which topics adjoin in some cases to IP (Zhang's TP) and in other situations to CP (e.g. Rochemont 1989). The second involves denying the claim that C° is the only head higher than the inflectional projections in a sentence. This is the approach of Rizzi (1997), who argues that the CP area has both a Force head, in which that is located, and a Finite head, which is the closest head to the inflectional layer of the sentence, and to which auxiliaries raise as a consequence. If such an analysis is adopted, topics do not need to be adjoined to TP, but could be in the specifier of FiniteP, or adjoined to FiniteP, or, as Rizzi suggests, in a separate projection headed by Topic0. This would account for the fact that topics must follow that in (4) but must follow don't in (5) and can't in (6). 33 Zhang does not ignore interrogatives completely. Although he does not present a judgement for (6a), which is a crucial counterexample to his analysis if the judgement presented here is correct, he does assert that (6b) is ungrammatical, which he takes as support for his claim that topics must occur in TP, and not in front of C°, which is the position he assumes for can't. He contrasts (6b) with (7), which contains a resumptive pronoun, and finds the contrast "quite clear", in his words. (7) That classic novel, can't you read it by next week? Whether topics are evidence for or counterevidence to his claim depends on which set of judgements is accepted. My informal survey of English speakers confirmed the judgements shown above, and not Zhang's. Accordingly, I take the argument from topicalization to be an argument against his claim that don't is an imperative particle adjoined to TP. The argument from phonology It has been argued that the phonological rule of contraction takes place between adjacent elements and cannot take place if a null syntactic element intervenes. (8) a. Do you want to step outside? b. Do you wanna step outside? c. Who; do you want e; to shoot you? d. *Whoj do you wan ti na shoot you? When want and to are adjacent as in (8a), they can merge into wanna. But if a trace intervenes between them, as in (8c), they cannot merge. Zhang cites Akmajian (1984), who reports the following contrast: (9) a. Doncha wanna go now? b. Cancha help me? c. Woncha come to the party? d. */?? Doncha hit me! The non-parallel between interrogatives and imperatives could be due to a difference syntactic configuration. Zhang suggests that since interrogative don't is in C°, this argues that imperative don't must not be in C°. Although Akmajian reports a sharp contrast in acceptability between the interrogatives and imperative in (8), the consultants polled for this thesis report a milder contrast. (10) ?Donchahitme! I suggest that the contrast between (9a-c) and (10) can be explained as a stylistic, rather than syntactic, contrast. Contractions like cancha and doncha are less formal than their non-contracted counterparts. Questions are much easier to interpret as informal than commands, which predicts that contractions are more harmonious in questions than in commands. If an imperative is uttered not as a formal command but in an informal context, doncha becomes perfectly acceptable, as shown in (11). (11) Doncha go breaking my heart, now (said playfully) 34 Another problem with this argument is that Zhang fails to account for the difference in terms of his hypothesis. He does not propose an empty category intervening between don't and you in imperatives. Both interrogative and imperative don't appear to be adjacent to the subject. It seems suspicious that his argument, modeled on a syntactic argument for a null category, fails to propose any syntactic difference to account for the difference in data. The argument from case Zhang reports the following contrast: (12) Zhang's judgement a. Don't you and him fight again! b. *Don't you and he fight again! Once again, Zhang's data are disputed by the speakers I consulted, who report that there is no such stark contrast between the two sentences. (13) Other speakers 'judgements a. Don't you and him fight again! b. Don't you and he fight again! There was some disagreement among the English speakers about which pronoun was preferred, but all agreed that either is acceptable. Zhang, like the speakers consulted for this thesis, finds both he and him acceptable in interrogative conjoined subjects. (14) a. Won't you and him come to my party? b. Won't you and he come to my party? Based on the contrast between imperatives and interrogatives that he reports, Zhang concludes that there is a structural difference between the sentences in (12) and those in (14). He offers no mechanism, however, for explaining the contrast in terms of the location of the auxiliary. There could easily be some alternative explanation for the contrast, if it exists. Considering the cost to the theory if don't is an imperative particle, it is better to assume that some alternative explanation exists for the contrast in (12) and (14), and that don't, like do, is in C°. 35 Chapter 5 - Verb Movement in English Imperatives 5.1 Introduction In chapter 4 I showed that the phrase structure of imperatives is the same as the phrase structure of declaratives. In this chapter I examine the movement properties of imperative verbs. I begin with a review of Potsdam's (1996) treatment of this issue. Potsdam argues on the basis of adverb placement that imperative verbs, as well as most verbs in English, do not undergo verb raising, the only exception being finite auxiliary verbs. I argue against Potsdam's analysis, showing that the facts of verb positioning with respect to adverbs can be better explained if we assume, with Cinque (1999), that adverbs fall into different classes, each of which is the specifier of a different projection. Based on evidence from English, I propose a phrase structure that is more modest than Cinque's but more elaborate than Potsdam's. I then use verb positioning with respect to different adverb types as a diagnostic for verb raising. I discover that the split in English is not simply between finite auxiliaries on the one hand and all other verbs on the other, as Potsdam argues, but rather a three way split between main verbs, for which there is no evidence of movement, nonfinite auxiliaries, which generally undergo a short movement, and finite auxiliaries which undergo a longer movement. This conclusion that there are three types of movement is identical to that reached by Pollock (1989), but the specific targets of verb movement differ. 5.2 Potsdam (1996) Potsdam (1996) investigates verb movement in imperatives and other sentence types, concluding that imperative verbs, like all verbs except finite auxiliaries, are stationary. He uses the phrase structure in (1), and applies a number of tests to determine whether the verb is located in 1° or in V°. I review two of the tests here: the negation placement test and the adverb placement test. (1) Pollock's phrase structure: CP C IP I NegP Neg VP V 36 5.2.1 Negation placement Contrary to our assumption (made in chapter 3) that not is a specifier of NegP, Potsdam assumes that it is a head in Neg0. The two assumptions make the same prediction for this diagnostic: an element in 1° will precede not, while an element in V° will follow it. It is generally assumed that it is sentential negation, and not constituent negation, that occurs on NegP, and this assumption is supported by approaches that build on it (e.g. Acquaviva 1997), so the diagnostic only works when not is used as a sentential negation. Sentential negation involves the logical connective that negates the entire proposition, while constituent negation negates just a part of the proposition. (2a) is an example of sentential negation, while (2b) is an example of constituent negation. John is not very tall John is very not tall John is not interested John is not very tall by any means *John is very not tall by any means (2a) is a negation of the proposition John is very tall, while in (2b) the negation applies only to tall. Sentences like (2c) are ambiguous - the negation may be sentential, in which case it is in NegP, or it could be constituent negation adjoined to AdjP. It is acceptable, although odd, to include sentential and constituent negation in a sentence, like in "John is not not interested". In many cases it is possible to distinguish sentential versus constituent uses of not by inserting a negative polarity item like any or even in an adjunct following the main part of the predicate, as is done in (2d,e). Negative polarity items need to be c-commanded by not (Uribe-Echevarria 1994), which is necessary if not is in NegP. When it occurs within the predicate phrase it is apparently not capable of c-command, because adding an adjunct with any takes away the constituent negation interpretation in sentences like (2c) or makes the sentence ungrammatical, as in (2e). Using our assumption that sentential negation is in NegP and the fact that not in (2a) is sentential negation, we can use the position of not as a diagnostic for positioning with respect to NegP. Thus very tall follows NegP, while is precedes it. More specifically, the position of an auxiliary verb between the subject and a sentential negative is a reliable indicator that it is in 1°, given the phrase structure that we have argued for in chapter 4. Using this diagnostic, Potsdam tests auxiliary and main verbs in finite, subjunctive and imperative sentences to determine whether they are in 1° or V 0 5 . 5 Throughout this chapter, I ignore support-fifo, which behaves differently from auxiliaries be and have in a way which makes it problematic for my analysis, if it is treated as an auxiliary. For instance, do differs from be and have by preceding negation in subjunctive clauses: I insist that you not be present We insist that you not have been convicted of a violent crime I insist that you do not get involved *I insist that you not do get involved The same is true in imperatives. I assume that do, when it is present, is generated in 1°, not moved from V°, which explains its pre-negation position in subjunctives and imperatives, as well as finite sentences, and makes it irrelevant to the discussion of verb movement in this chapter. 37 Finite clauses (3) Auxiliaries a. John was not afraid of vampires by any means b. *John not was afraid of vampires by any means c. John has not been wearing a cross by any means d. *John not was wearing a cross by any means (3) shows that the finite auxiliaries was and has precede, but do not follow, sentential not. By the reasoning above, finite auxiliaries are in 1°. In contrast, finite main verbs, as in (4), follow but do not precede sentential not. (4) Main verbs a. John did not see the truck coming by any means b. *John not saw the truck coming by any means Subjunctives Potsdam shows that in subjunctives, both auxiliary and main verbs follow, and do not precede, sentential not. This is shown in (5-6). (5) Auxiliaries a. My dentist requested that I not be late for my appointment by any means b. *My dentist requested that I be not late for my appointment by any means c. The workers requested that I not have ruined their new sidewalk before it even dries (Potsdam 1996, p. 162) d. *The workers requested that I have not ruined their sidewalk before it even dries (6) Main verbs a. I insist that you not involve yourself in human sacrifice by any means b. *I insist that you involve not yourself in human sacrifice by any means Since preceding sentential not is evidence for verb movement in Potsdam's framework, (5-6) is evidence that subjunctive verbs, including auxiliaries, do not move. Imperatives Imperatives, whether auxiliary or main verb, also follow sentential not, as shown in (7-8)6. 6 It is not clear to me why do needs to appear in these sentences, when it is not necessary in infinitival or subjunctive contexts. It seems unlikely, given the lack of inflection on imperative do, that the reason it is necessary is the same as in finite clauses, where do is inserted to support inflectional features in place of the main verb. 38 (7) Auxiliaries a. Do not be late for your appointment by any means! b. *Be not late for your appointment by any means! c. Do not have ruined our sidewalk before it even dries! d. *Have not ruined our sidewalk before it even dries! (8) Main verbs a. Do not involve yourself in human sacrifice by any means! b. *Involve not yourself in human sacrifice by any means! Potsdam's conclusion about imperatives is the same as his conclusion about subjunctives: they do not raise. 5.2.2 Adverb placement Following Jackendoff (1972), Potsdam divides adverbs into VP-adverbs and S-adverbs. He characterizes their distribution descriptively as in (9) and in terms of a phrase structure theory as in (10). Some of the data he bases this distribution on are given in (11) (Potsdam 1996, p. 23-25). (9) a. Positional distribution of VP-adverbs i. to the immediate left of the main verb ii. clause-final b. Positional distribution of S-adverbs i. clause-initial ii. immediately following the subject iii. to the immediate right of a modal or finite auxiliary (10) a. Syntactic distribution of VP-adverbs i. left or right adjunction to main verb V b. Syntactic distribution of S-adverbs i. left adjunction to IP ii. left adjunction to I' iii. left adjunction to topmost VP 39 (11) a. VP adverbs i. Stanley completely/easily/handily ate his Wheaties ii. Stanley ate his Wheaties completely/easily iii. George was completely/rapidly ruined by the tornado iv. George might have completely/entirely lost his mind v. Peter could have been safely/effortlessly rescued vi. *Completely/*easily Stanley ate his Wheaties vii. George *completely/*rapidly was ruined by the tornado viii. George might *completely/*rapidly have lost his mind ix. Peter could *safely/*effortlessly have been rescued b. S-adverbs i. Horatio evidently/probably/ has lost his mind ii. Evidently/probably Horatio has lost his mind iii. Horatio has evidently/probably lost his mind iv. George has been *probably/*certainly ruined by the tornado v. George is being *certainly/*apparently tailed by the FBI Because S-adverbs can occur as low in the clause as an adjoined-to-VP position, Potsdam finds it useful to use them as a diagnostic for verb positioning. A verb that must follow an S-adverb must be in the VP, presumably remaining in V°. A verb that can precede an S-adverb can be in 1°. There are some facts about verb positioning, however, that are problematic for Potsdam's model of adverb adjunction. I discuss them in the next section. 5.3 Problems with Potsdam's adverb account 5.3.1 S-adverbs and negation Potsdam's model of adverbial syntax allows S-adverbs to adjoin to VP, IP, or F. This predicts that S-adverbs should be able to precede sentential negation when they are adjoined to IP or F, or to follow sentential negation when they are adjoined to VP. In fact, S-adverbs cannot follow sentential negation under regular intonation, as seen in (12c). (12) a. Mickey certainly was prepared for his championship match b. Mickey was certainly not prepared for his championship match c. *Mickey was not certainly prepared for his championship match 5.3.2 S-adverbs and movement to pre-subject position An asymmetry between S-adverbs that precede the subject (adjoined to IP, in Potsdam's account) and S-adverbs that follow the subject (adjoined to I' or VP, in Potsdam's account) is evident in the following contrasts: (13) a. Who did John probably kiss? b. *Who did probably John kiss c. Bill, John certainly kicked d. *Bill, certainly John kicked 40 This asymrnetry is unexpected in Potsdam's account of adverb placement. It might be possible to claim that wh-movement and topicalization are movements to [Spec,CP], and that such a movement may not take place across an element adjoined to IP, but may take place across an element adjoined lower in the clause, but there is no obvious explanation for such a claim, and it would just be stipulation. 5.4 An alternative account of adverb positioning A different approach to adverbs assumes that each class of adverbs are associated with a separate position, and that different word orders are normally derived by different movements of items around the adverbs. This approach is advocated, for instance, by Cinque (1999), who assumes a rigid universal hierarchy of many functional positions which take adverbs as specifiers. I adopt the basic assumptions of this approach - that an adverb is in the specifier of a specific projection, and that adverbs are generally stationary - although I will argue that they may move to [Spec,CP]. I propose three classes of adverbs, each in the specifier of a different syntactic projection. The first class consists of what Potsdam (1996) calls S-adverbs and I call speaker adverbs, shorthand for Jackendoffs (1972) label speaker-oriented adverbs. These adverbs appear either before the subject or between the subject and the verb, and are characterized semantically by the fact that they involve implicit reference to the speaker of the sentence. Jackendoff points out that speaker adverbs can often be paraphrased in terms of a related adjective as it is Adj (to me) that S or 1 consider it Adj that S. For example, obviously Frank has graduated has as part of its meaning it is obvious to me that Frank has graduated, and Frank has unfortunately graduated has as part of its meaning I consider it unfortunate that Frank has graduated. These paraphrases show that the reference to the speaker involves the speaker's attitude towards the proposition expressed in the rest of the sentence. The second class of adverbs consists of what Potsdam calls VP-adverbs, and have traditionally been called manner adverbs. Jackendoff (1972) notes that these adverbs can typically occur between the subject and the verb or following the VP, although some of them can only occur in one position or the other. The third class of adverbs contains the temporal adverbials. This class is not discussed by Jackendoff or Potsdam, but it will be useful to separate them into a distinct class in order to investigate the positioning of verbs relative to this class. I propose the following phrase structure: There is a projection between CP and NegP, which I assume is AgrP, following Pollock (1989). The subject appears in [Spec, AgrP]. Speaker adverbs occur in the specifier of a phrase intermediate between AgrP and NegP, which I will call speaker phrase and symbolize S P ; temporal adverbs are specifiers of TP, which is between NegP and VP; and manner adverbs are specifiers of VP. The tree in (14) summarizes the phrase structure I propose. 4 1 (14) CP Evidence for the structure in (14) comes from the relative word orders of different elements. SP, as we have noted in (12), must precede sentential negation, but it can follow the subject, suggesting that a projection exists between AgrP and NegP. More supporting evidence comes from the fact that finite auxiliaries which follow the subject can either precede or follow speaker adverbs, but must precede negation. (15) a. Max has certainly not been to India b. Max certainly has not been to India c. *Max certainly not has been to India The fact that the auxiliary can precede or follow the adverb in [Spec, BP] can be accommodated in the phrase structure in (14), if we assume that it can be either in Agr° or in S ° . Aside from appearing in the specifier of S P , speaker adverbs are capable of moving to [Spec, CP]. Movement of adverbs to [Spec,CP] has been suggested by Cinque (1999) as an explanation for first-position adverbs in verb-second languages, and is supported in this instance by a couple of arguments. First, consider the following data. (16) a. Probably Bill is dead b. Bill probably is dead c. Bill is probably dead 42 The variation in order between Bill and probably can either be accounted for by allowing the subject to appear in two possible positions or else allowing speaker adverbs to occur in two possible positions. If we assume that it is the subject that can occur in two places, we are forced to complicate the phrase structure, while two positions for speaker adverbs can be accommodated in the current structure. To see this, suppose the subject has two potential positions, sandwiching a single position for the speaker adverb. The higher subject position cannot be [Spec,CP], because the subject can occur between a wh-word in [Spec,CP] and a speaker adverb, and between a topic and a speaker adverb. (17) a. Who did John probably kick? b. What did John certainly say? c. Bill, John certainly kicked d. The Star-Spangled Banner, John evidently sang Therefore [Spec,AgrP] is the higher position for the subject. The lower position for the subject cannot be [Spec,NegP], since the subject can occur between a speaker adverb and not: (18) Evidently John is not missing So assuming that the subject moves entails that another specifier position, hence another phrase, is necessary. If, on the other hand, we assume that speaker adverbs can occur either in [Spec, SP] or [Spec, CP], we do not run into this problem. When the adverb precedes the subject, it is compatible neither with a wh-question nor with topicalization, as we have noted in (13). The simplest explanation for this fact is that the adverb occupies [Spec,CP] and does not allow any other element to move there. In keeping with our claim that adverbs are generated in a unique position, we will assume that adverbs move to [Spec,CP], rather than being generated there. This is consistent with the analyses of adverb movement in Cinque (1999) and Gundel (1988). Potsdam characterizes the positions of VP adverbs as immediately to the left of the main verb, and clause-final. Cinque (1999) devotes some discussion to the question of whether these adverbs have two positions where they can be base-generated, or whether some movement takes place. His conclusions are that different explanations are required in different languages. In Italian, he argues that all VP adverbs are generated above the VP, and the VP can optionally raise to precede them. In English, however, there are some syntactic facts that raise problems for the movement analysis. Consider for instance the following contrasts (judgements are from Jackendoff 1972)7. ' It has been pointed out to me by Henry Davis (p.c.) that (19b) and (20b) sound much better if material is added following the direct object, like in the following: i. Joe carefully worded the letter so as to not implicate himself ii. ?The bread easily cuts with a moderately sharp knife I believe that the reason these sentences are more acceptable than (19b) and (20b) is that in i and ii, the adjunct following the object fulfills the selectional restriction that is not met in (19b) and (20b). In fact, the adverbs are optional when another adjunct follows the object: iii. Joe worded the letter so as to not implicate himself ii. ?The bread cuts with a moderately sharp knife 43 (19) a. Joe worded the letter carefully b. *Joe carefully worded the letter (20) a. The bread cuts easily b. *The bread easily cuts (19) shows that a clause-final VP-adverb can satisfy the selection requirement of a verb like worded, but not a VP adverb immediately to the left of the main verb. (20) shows a similar contrast in licensing middle constructions. If one of the two positions was derived by movement, we would expect verbs in both positions to behave the same way with respect to these properties. Cinque concludes that the two positions described by Potsdam must be independent positions where VP adverbs are base-generated - one preceding the VP and one following the VP. I will only talk about the adverbs preceding the VP. I assume that they are located in [Spec,VP] . This means that any verb raising at all will force the verb to precede manner adverbs. We will see in the next section that this demonstrates that manner adverbs are lower in the clause than temporal adverbs. Temporal adverbs like always, often, and today follow sentential negation. Consider the pair of sentences in (21). (21) a. Cynthia is not always at the bar at this time of day, by any means b. *Cynthia is always not at the bar at this time of day, by any means When not precedes the temporal adverb, as in (21a), it can license a negative polarity item in an adjunct, indicating that it is sentential negation. When not follows the temporal adverb, it does not license negative polarity items in the adjunct, which indicates that the scope of the negation is not the sentence. With our assumption that not is sentential only when it is in NegP, these facts show that temporal adverbs follow NegP. 5.5 Adverbs and verb movement The elaborated phrase structure in (14) forces us to propose more specific targets of verb movement. Potsdam (1996) claims that finite auxiliaries raise to an inflectional position preceding not, and that otherwise English verbs do not raise. In our phrase structure, the 1° to which finite auxiliaries move could be S ° or Agr°. In 5.5.1 - 5.5.3 I argue that there is a distinction between finite and nonfinite auxiliaries. Finite auxiliaries must raise at least as far as 9° overtly, and possibly to Agr°. Infinitive and subjunctive auxiliaries, which I class together as nonfinite auxiliaries, raise to T°. The essence of this claim is that nonfinite auxiliaries undergo a shorter movement than finite auxiliaries, as in Pollock (1989). In 5.5.4 I investigate main verbs, concluding with Pollock and Potsdam that English main verbs do not raise at all in standard constructions. In 5.5.5 I investigate imperatives, pointing out that imperative auxiliaries have the exact same movement properties as nonfinite auxiliaries, and concluding that they are nonfinite verbs. Cinque's point maintains its force: the adverbs preceding the verb must not have been moved from a postverbal position, or the trace of the movement will fulfill the verb's selectional requirements. 44 5.5.1 Finite auxiliaries In Potsdam's analysis, finite auxiliary verbs move to 1 ° , which entails that they follow S-adverbs that are adjoined to IP or I', and precede S-adverbs that are adjoined to the topmost VP, all of which are permissible positions for S-adverbs in his analysis. Thus finite auxiliaries can precede or follow S-adverbs. In our account, this fact is derived by allowing the auxiliary verb to move either to S ° or Agr°. If it moves to S ° , it follows the speaker adverb. If it moves to Agr° and the speaker adverb does not raise to [Spec,CP], the auxiliary precedes the adverb. Finite auxiliaries always precede manner adverbs (Potsdam's VP-adverbs), derived in Potsdam's phrase structure by having manner adverbs adjoin to VP, and in ours by placing them in [Spec,VP]. Let us turn to temporal adverbs, which are not discussed by Potsdam. Our account, in which temporal adverbs are in [Spec,TP] and finite auxiliaries raise past them, to S ° or Agr°, predicts that finite auxiliaries (and modals) should appear to the left of temporal adverbs. This appears to be the pattern, at least under neutral intonation. This can be seen in (22-23). (22) a. I am always happy to see you b. Max might sometimes misbehave, but he's usually a good boy c. Mindy has often remarked that she would like to dump Marcus d. Fido is frequently missing e. Jane had usually gone to linguistics conferences (23) a. I always am happy to see you (acceptable only with emphasis on always or am) b. Fido sometimes might misbehave, but he usually is a good boy (acceptable only with emphasis on (sometimes or might) and usually) c. Mindy often has remarked that she would like to dump Marcus (acceptable only with emphasis on often or has) d. Morris frequently is missing (acceptable only with emphasis on frequently or is) e. Jane usually had gone to linguistics conferences (acceptable only with emphasis on usually or had) Under neutral intonation, the sentences in (22) are grammatical, but those in (23) are ungrammatical. This is explained if English finite auxiliaries must raise at least to fi° under normal intonation. If either the auxiliary or the temporal adverb is emphasized, the auxiliary may stay in T°. 5.5.2 Infinitive auxiliaries Pollock (1989) analyzes infinitive auxiliaries and concludes that they involve short movement -movement to T°. This is based on sentences like the following (p. 382). (24) a. I believe John to be often sarcastic b. The English were then said to have never had it so good c. (?)John is said to be seldom on time at his appointments In the structure introduced earlier in this chapter, these movements would count as movements at least to 9°, since the auxiliaries precede the temporal adverb in [Spec,TP]. In my judgement, 45 however, these sentences are less than fully acceptable, especially when contrasted with similar sentences in which the auxiliaries follow the temporal adverb. (25) a. I believe John to often be sarcastic b. The English were then said to never have had it so good c. John is said to seldom be on time at his appointments Consulting a number of English speakers, I found that judgements were variable. Most consultants strongly preferred the sentences in (25), but some agreed with Pollock and found only a slight contrast. It is tempting to assume that for the former, movement of infinitival auxiliaries is to T°, while for the latter, there is the possibility of moving them to 9°. There is a problem, however, for this approach. Movement to 8 ° entails that the auxiliaries will linearly precede not. Pollock (1989, p. 376) reports that such sentences are unacceptable, and my consultants agreed for the most part8. (26) a. ?To be not happy is a prerequisite for writing novels b. (?)To have not had a happy childhood is a prerequisite for writing novels c. ?To be not arrested under such circumstances is a miracle Given these judgements and the structure in (13), it follows that movement of the infinitive auxiliary cannot be past T°. The auxiliaries also precede manner adverbs as shown in (27), indicating that they are not in the VP. (27) a. *Simon expects to devastatingly be poor b. Simon expects to be devastatingly poor c. ?I expect Simon to forcefully be kicking Bob d. I expect Simon to be forcefully kicking Bob The best position for movement of the infinitive auxiliary is thus T°, although the judgements are murky with respect to placement relative to temporal adverbs. The same facts that are found with infinitive auxiliaries can be found in subjunctive auxiliaries. 5.5.3 Subjunctive auxiliaries (28) a. I insist that John be forcefully kicking Bob b. ?I insist that John forcefully be kicking Bob c. I insist that John be completely honest d. *I insist that John completely be honest (29) a. I insist that the men always be kicking Bob b. I insist that the men be always kicking Bob (judgements variable) (30) a. *I insist that John be not forcefully kicking Bob (on sentence negation reading) b. *I insist that John be not completely honest (on sentence negation reading) (28) shows that subjunctive auxiliaries must precede manner adverbs and may not follow them. This indicates obligatory movement at least to T°. (29) shows that when subjunctive auxiliaries precede temporal adverbs, judgements are variable, which leads us to the same conclusion as we were led to in the case of infinitive auxiliaries, which is that there is a preference for movement to be limited to T°, but for some people movement can be to 9°. (30) suggests that movement is Some consultants found (26b) acceptable, but all agreed on the ungrammaticality of (26a,c). 46 limited to T° by showing that the auxiliary may not precede sentential negation. Subjunctive auxiliaries pattern like infinitive auxiliaries, not like finite auxiliaries. We can conclude that while finite auxiliaries raise to 1° or nonfinite auxiliaries generally only raise to T°. 5.5.4 Main verbs As Potsdam (1996) showed, main verbs appear not to raise when their position relative to speaker adverbs, floating quantifiers and not is considered. The conclusion does not change when their position with respect to temporal and manner adverbs is considered. Finite (31) a. Francis always ate his dessert before dinner b. *Francis ate always his dessert before dinner c. Francis didn't so much eat his dessert as voraciously devour it d. *Francis didn't so much eat his dessert as devour voraciously it Infinitive (32) a. I expect Francis to always eat his dinner first b. *I expect Francis to eat always his dinner first c. I expect Francis to quietly eat his dessert d. *I expect Francis to eat quietly his dessert Subjunctive (33) a. I insist that Francis always eat his dinner first! b. *I insist that Francis eat always his dinner first! c. I insist that Francis quietly eat his dessert! d. *I insist that Francis eat quietly his dessert! The data in (31-33) show that a main verb, regardless of whether it is finite or nonfinite, will fail to raise past the temporal adverb (the a,b examples) or the manner adverb (the c,d) examples. The evidence that motivates movement in auxiliaries - the fact that they may precede one or more classes of adverb - is missing in this case, suggesting that main verbs do not move at all. The division between finite and nonfinite clauses, therefore, is evident in the degree of movement of auxiliary verbs. We finally turn to imperatives, finding that they pattern like nonfinite verbs - the auxiliaries precede manner adverbs (34) but follow temporal adverbs (35) and sentential negation (36), indicating movement to T°, while the main verbs follow temporal (37a,b) and manner (37c,d) adverbs, indicating a lack of movement. 5.5.5 Imperatives (34) a. Be forcefully kicking Bob when I look your way! b. *Forcefully be kicking Bob when I look your way! c. Be completely honest! d. *Completely be honest! (35) a. Always be kicking Bob! b. *Be always kicking Bob! 47 (36) a. *Be not kicking Bob! b. *Be not completely honest by any means! (37) a. Always eat your dinner before dessert! b. *Eat always your dinner before dessert! c. ?Quietly eat your dinner! d. *Eat quietly your dinner! The data in (34-37) show that imperatives pattern more or less like other nonfinite verbs in English. Like subjunctive and infinitive auxiliaries, and in contrast to finite auxiliaries, the imperative auxiliaries in (34-36) precede manner adverbs but follow temporal adverbs and sentential negation. Interestingly, there is agreement that imperative auxiliaries cannot precede temporal adverbs, in contrast to the variability in judgements on infinitive and subjunctive auxiliaries. Main imperative verbs pattern with finite, infinitive and subjunctive main verbs, being unableto precede manner adverbs, indicating the absence of movement. 5.6 Conclusion We have used sentential negation and the classes of manner, temporal and speaker adverbs as position diagnostics to investigate the extent to which different classes of verbs move in English. We have discovered that English verbs fall into three classes. The first class consists of all main verbs, none of which raise. A second class consists of finite auxiliaries, which move to a position preceding NegP but following the subject - either Agr°, or the head of another category that we have introduced, S P . The third class is made up of nonfmite auxiliaries, including imperative auxiliaries, which raise to T°, the head of a position between NegP and VP. Within the third class, it was clear for imperative auxiliaries that the limit of the movement was to T°, but for infinitive and subjunctive auxiliaries, there is some variation, which apparently allows movement past T° for some speakers. The table in (38) summarizes the possible verb positions in English, indicating with question marks where there is some variation. (38) V° -pO ®° A R T 0 Main verbs / X X X Finite Auxiliaries X X / / Infinitive Auxiliaries X / X ? X ? Subjunctive Auxiliaries X / X ? X ? Imperative Auxiliaries X / X X 48 Chapter 6 - Verb movement in Hebrew 6.1 Introduction We have investigated the interaction between various verb movements and the different types of adverbs in English. We now turn to Hebrew, a language that has different verb movement properties, to see which of the results of the previous chapter can be replicated. Unlike English not, the Hebrew negatives al, lo and eyn are heads, as argued in chapter 4, and none of them can be used as reliable diagnostics for a verb's position, eyn prohibits verbs from raising past it, but because of this property, it is restricted to benoni environments, and does not allow a comparison across clause types. The other two negatives, lo and al, raise overtly to precede the verb, as argued in chapter 3, so they cannot be used to determine verb positions. Since negation is unusable, this chapter focuses on adverbs as positional diagnostics. In 6.2 I provide examples of each adverbial category. In 6.3 I investigate the position of finite verbs relative to each of the adverbial classes, which leads to two conclusions: first, that the position of adverbs in Hebrew is the same as in English, and by extension the phrase structure of Hebrew is the same as in English; and second, that finite verbs may raise to either 9° or T°. Sections 6.4 and 6.5 investigate the movement properties of infinitive and imperative verbs. We will see that imperative verbs in Hebrew are not allowed the same range of positions that infinitive verbs are, which means that the classification proposed in the last chapter, in which imperatives are a kind of nonfinite verb, is not a universal fact. More specifically, I find that verbs fall into two classes in Hebrew, with finite and infinitive verbs on one side and true and suppletive imperative verbs on the other. 6.2 Manner, temporal and speaker adverbs Hebrew has the three classes of adverbs that we have investigated in English. The following are some of the adverbs that are found in each class. (1) Manner adverbs besimxa "happily" bekoax "forcefully" be?itiut "slowly" bimhirut "quickly, speedily" (2) Temporal adverbs hayom "today" tamid "always" af pa?am "never" kol yom "every day" bederex klal "usually" kvar "already" 49 (3) Speaker adverbs lemazal-U-xaJ-ex/-o/-a/ -enu/-xem/-xen/-am/-an "luckily" (literally, 'to my/your(^ fortune') be?emet "truly, really" kanir?e "apparently, probably" ulay "maybe" batuax "certainly" This classification will be justified in the following sections, by the fact that verbs occur in different positions relative to the different adverbial classes. At the same time, we will also see that the positions of the adverb classes in the phrase structure fits the model that has been presented for English in chapter 59. 6.3 Finite verb movement Manner adverbs (4) a. David xazar besimxa habayta David returned happily home "David returned home happily" b. ?David besimxa xazar habayta (4) shows us that unlike English, where verbs precede or follow manner adverbs with equal felicity, Hebrew verbs prefer to precede rather than follow manner adverbs. The pattern in (4) holds for the other examples as well. (5) a. David hif?il bekoax et hamexona David activated forcefully acc the-machine "David forcefully activated the machine" b. ?David bekoax hif?il et hamexona 9 The data used in the following sections were judged by two native speakers of Hebrew. The first, Rivka Strauss, is a woman in her early 50's who grew up in Netanya, Israel, and currently resides in Toronto. The second speaker, Avikam Hameiri, is a man in his early thirties who also grew up in Netanya. Throughout this chapter, definite judgements of ungrammaticality are preceded by an asterisk, while a question mark indicates a weaker judgement of deviance. A question mark in parentheses indicates speaker divergence on the grammaticality of the sentence, with one consultant judging the sentence grammatical and the other judging it at least somewhat deviant. The sentences involving finite verbs (4-17) were tested by both consultants, and the data involving infinitive and imperative verbs (19-31) were given by one (Rivka Strauss). 50 (6) a. David zaxal be?itiut cafona David crawled slowly northward "David slowly crawled northward" b. ?David be?itiut zaxal cafona (7) a. hasus dahar birhhirut lekivun hame?ara the-horse galloped speedily towards the-cave "The horse speedily galloped towards the cave" b. ?hasus bimhirut dahar lekivun hame?ara Temporal adverbs (8) a. David axal hayom tiras David ate today corn "David ate corn today" b. David hayom axal tiras (8) demonstrates that Hebrew finite verbs can precede or follow temporal adverbs with apparent preference. The examples in (9-13) confirm the pattern. (9) a. David oxel tamid tiras David eats always corn "David always eats com" b. David tamid oxel tiras (10) a. David lo oxel af pa?am tiras David neg eats never com "David never eats com" b. David af pa?am lo oxel tiras (11) a. David oxel kol yom tiras David eats every day corn "David eats com every day" b. David kol yom oxel tiras (12) a. David oxel bederex klal tiras David eats usually com "David usually eats com" b. David bederex klal oxel tiras 51 (13) a. David axal kvar et hatiras David ate already acc the-corn "David already ate the corn" b. David kvar axal et hatiras Speaker adverbs (14) a. David lemazalo axal tiras David luckily ate corn "David luckily ate corn" b. *David axal lemazalo tiras (14) shows that finite verbs follow but do not precede speaker adverbs. The pattern is confirmed in (15-17). (15) a. (?)David be?emet axal tiras David really ate com "David really ate com" b. *David axal be?emet tiras (16) a. (?)David kanir?e axal tiras David apparently ate com "David apparently ate some com" b. *David axal kanir?e tiras (17) a. (?)David ulay axal tiras David maybe ate com "David might have eaten com" b. *David axal ulay tiras The verb placement data in (4-17) have confirmed that the three classes of adverbs that we have distinguished in English are distinguished in Hebrew as well. Additional evidence from co-occurrence of two types of adverbs shows that the order of adverbs in Hebrew is the same as in English. (18) a. David lemazalo tamid oxel tiras David luckily always eats com "Luckily, David always eats com" b ??David tamid lemazalo oxel tiras David always luckily eats com 52 (19) a. ?David oxel tamid besimxa tiras David eats always happily corn "David always eats corn happily" b. *David oxel besimxa tamid tiras David eats happily always corn These facts support the following linear order for adverbs: (20) Speaker adverbs > Temporal adverbs > Manner adverbs Given this evidence, it makes sense to propose that the same phrase structure that we have proposed for English in Chapter 5 applies in Hebrew as well. The structure is repeated in (21) (with English not removed). (21) CP It is clear from (21) where finite verbs move to. They must always precede manner adverbs, so they must raise overtly at least as high as T°. They may follow temporal adverbs, so it is possible for them to remain in T°. They may also precede temporal adverbs, so they may raise as high as 9°. They may not precede speaker adverbs, so the highest they can raise is 9°. So verbs raise overtly to either T° or 9°. As we have seen in chapter 4, Hebrew finite verbs generally must raise to Agr° at LF for their tense and person morphology to be licensed. 53 We now turn to nonfinite verbs, of which I will investigate two types - the infinitive and imperative. Unlike English, Hebrew has no subjunctive verb form. 6.4 Infinitive Verbs Manner adverbs (22) a. bikasti miDavid laxzor besimxa habayta I-requested from-David to-return happily home "I asked David to return home happily" b. *bikasti miDavid besimxa laxzor habayta c. bikasti miDavid lehaf?il bekoax et hamexona I-requested from-David to-activate forcefully acc the-machine "I asked David to forcefully activate the machine" d. *bikasti miDavid bekoax lehaffil et hamexona e. bikasti miDavid lizxol be?itiut cafona I-requested from-David to-crawl slowly northward "I asked David to slowly crawl northward" f. *bikasti miDavid be?itiut lizxol cafona g. bikasti mehasus lidhor bimhirut lekivun hame?ara I-requested from-the-horse to-gallop speedily toward the-cave "I asked the horse to gallop speedily towards the cave" h. *bikasti mehasus bimhirut lidhor lekivun hame?ara Temporal adverbs (23) a. bikasti miDavid le?exol hayom tiras I-requested from-David to-eat today corn "I asked David to eat corn today" b. bikasti miDavid hayom le?exol tiras c. bikasti miDavid le?exol tamid tiras I-requested from-David to-eat always corn "I asked David to always eat corn" d. bikasti miDavid tamid le?exol tiras e. bikasti miDavid lo le?exol af pa?am tiras I-requested from-David neg to-eat never corn "I asked David never to eat corn" f. bikasti miDavid af pa?am lo le?exol tiras g. bikasti miDavid le?exol kol yom tiras I-requested from-David to-eat every day corn "I asked David to eat corn every day" h. bikasti miDavid kol yom le?exol tiras 54 i. bikasti miDavid leexol bederex klal tiras I-requested from-David to-eat usually corn "I asked David to usually eat corn" j . bikasti miDavid bederex klal leexol tiras Speaker adverbs (24) a. *bikasti miDavid le?exol lemazalo tiras I-requested from-David to-eat luckily corn b. *bikasti miDavid lemazalo le?exol tiras c. *bikasti miDavid le?exol be?emet tiras I-requested from-David to-eat really corn d. *bikasti miDavid be?ement le?exol tiras e. *bikasti miDavid le?exol kanir?e tiras I-requested from-David to-eat apparently corn f. *bikasti miDavid kanir?e le?exol tiras g. *bikasti miDavid le?exol ulay tiras I-requested from-David to-eat maybe corn h. *bikasti miDavid ulay le?exol tiras From the data in (22-24) we have no evidence that infinitive verbs have raising properties that distinguish them from finite verbs. The pairs in (22) illustrate that infinitive verbs precede manner adverbs. The pairs in (23) show that they can precede or follow temporal adverbs. The ungrammaticality of all of the infinitives with speaker adverbs shows that speaker adverbs are apparently not felicitous in infinitive contexts. So we have evidence that infinitive verbs may be located in T° in the overt syntax, and that they may also be located in a head higher than T°, plausibly in 9°. We cannot use speaker adverbs to test the verb's location because they are not compatible with infinitive verbs10. Hebrew does not have any element that is obviously in [Spec,NegP] overtly, so no test using such an element is possible. Subjects cannot be used as a test unless we know with certainty the position of infinitive subjects, which we do not. In the absence of a reliable marker high in the phrase, we are left to speculate about how high infinitive verbs can raise. I conjecture that they do not differ from finite verbs and raise to 9° just like finite verbs do, since their behaviour with respect to the two lower classes of adverbs is identical. 1 0 This is true in English as well, as the following examples show: i. *Ned wants to evidently cause some trouble ii. *Dorothy wished Joan to probably love Fred We cannot attribute the ungrammaticality of these sentences to the unavailability of syntactic phrases where speaker adverbs are, because we have argued that infinitive verbs can be in S ° (from ex. 23). There is presumably a discourse-based explanation that would connect this fact about speaker adverbs with the more general fact that the speaker's point of view is usually connected to the main clause, but I cannot provide a full explanation here. 55 6.5 Imperative verbs As I have expressed in previous chapters, Hebrew imperatives come in different forms - true imperatives, future imperatives and infinitive imperatives. I examine the movement properties of each type of verb, finding that they pattern identically with respect to the adverb diagnostics. 6.5.1 True imperatives Manner adverbs (25) a. xazor besimxa habayta! return happily home "Return home happily!" b. *besimxa xazor habayta! c. hafPel bekoax et hamexona! activate forcefully acc the-machine d. *bekoax hafPel et hamexona! e. zxal beitiut cafona! crawl slowly northward "Slowly crawl northward!" f *beitiut zxal cafona! g. dhar bimhirut lekivun hame?ara! gallop speedily towards the-cave "Gallop speedily towards the cave!" h. *birnhirut dhar lekivun hame?ara! Temporal adverbs *• (26) a. exol hayom tiras! eat today com "Eat com today!" b. * hayom exol tiras! c. exol tamid tiras! eat always com "Always eat com!" d. *tamid exol tiras! e. exol kol yom tiras! eat every day com "Eat com every day!" f. *kol yom exol tiras! 56 g. *exol bederex klal tiras! eat usually corn h. *bederex klal exol tiras! Speaker adverbs (27) a. *exol lemazalxa tiras! eat luckily corn b. *lemazalxa exol tiras! c. *exol be?emet tiras! eat really corn d. *be?emet exol tiras! e. *exol kanir?e tiras! eat apparently corn f. *kanir?e exol tiras! g. *exol ulay tiras! eat maybe corn h. *ulay exol tiras! True imperative verbs behave differently from finite verbs and infinitive verbs, in contrast with English, in which imperatives pattern with nonfmite verbs. Like finite and nonfinite verbs, true imperatives must precede manner adverbs. Unlike finite and infinitive verbs, however, true imperatives cannot be in T°, but must move higher, as indicated by the fact that they must precede temporal adverbs, shown in (26). Regarding the ungrammaticality of (26g), it must be noted that my consultant expressed difficulty with assigning a meaning to the sentence, indicating that its unacceptability is perhaps due to semantic factors rather than syntactic ones. (29g) and (32g) below, which also try to effect this combination, are likewise ungrammatical. True imperatives, like othe nonfinite clauses, are incompatible with speaker adverbs, making it impossible to test whether they may or must be in 9°. 6.5.2 Suppletive future imperatives Manner adverbs (28) a. taxzor besimxa habayta! return happily home "Return home happily!" b. *besimxa taxzor habayta! c. taf?il bekoax et hamexona! activate forcefully acc the machine "Activate the machine forcefully!" d. *bekoax taffil et hamexona! e. tizxal be?itiut cafona! crawl slowly northward "Crawl northward slowly!" f. *be?itiut tizxal cafona! 57 g. tidhar bimhirut lekivun hame?ara! gallop quickly towards the-cave "Gallop quickly towards the cave!" h. *bimhirut tidhar lekivun hame?ara Temporal adverbs (29) a. toxal hayom tiras! eat today corn "Eat corn today!" b. *hayom toxal tiras! c. toxal tamid tiras! eat always corn "Always eat corn!" d. *tamid toxal tiras! e. toxal kol yom tiras! eat every day corn "Eat corn every day!" f. *kol yom toxal tiras! g. *toxal bederex klal tiras! eat usually corn h. *bederex klal toxal tiras! Speaker adverbs (30) a. *toxal lemazalxa tiras! eat luckily corn b. *lemazalxa toxal tiras! c. *toxal be?emet tiras! eat really com d. *be?emet toxal tiras! e. *toxal kanir?e tiras! eat apparently com f. *kanir?e toxal tiras! g. *toxal ulay tiras! eat maybe com h. *ulay toxal tiras! Future tense imperative verbs behave identically to true imperative verbs. They obligatorily precede manner and temporal adverbs, indicating that they raise at least to 9°. They are incompatible with speaker adverbs, so whether they raise further cannot be tested by this method. 58 6.5.3 Suppletive infinitival imperatives M a n n e r adverbs (31) a. ?bevakasa laxzor besimxa habayta please to-return happily home "Please return home happily" b. *bevakasa besimxa laxzor habayta c. bevakasa lehaf?il bekoax et hamexona please to-activate forcefully acc the-machine "Please activate the machine forcefully" d. bevakasa bekoax lehafPil et hamexona e. bevakasa lizxol be?itiut cafona please to-crawl slowly northward "please slowly crawl northward" f. bevakasa be?itiut lizxol cafona g. bevakasa lidhor bimhirut lekivun hame?ara please to-gallop speedily towards the-cave "Please speedily gallop towards the cave" h. ?bevakasa bimhirut lidhor lekivun hame?ara Temporal adverbs (32) a. bevakasa le?exol hayom tiras please to-eat today com "Please eat com today!" b. *bevakasa hayom le?exol tiras c. *bevakasa le?exol tamid tiras please to-eat always com d. *bevakasa tamid le?exol tiras e. bevakasa le?exol kol yom tiras please to-eat every day com "Please eat com every day" f. bevakasa kol yom le?exol tiras g. *bevakasa le?exol bederex klal tiras please to-eat usually com h. *bevakava bederex klal le?exol tiras 59 Speaker adverbs (33) a. *bevakasa le?exol lemazalxa tiras please to-eat luckily corn b. *bevakasa lemazalxa le?exol tiras c. *bevakasa le?exol be?emet tiras please to-eat really corn d. *bevakasa be?emet le?exol tiras e. *bevakasa le?exol kanir?e tiras please to-eat apparently com f. *bevakasa kanir?e le?exol tiras g. *bevakasa le?exol ulay tiras please to-eat maybe com h. *bevakasa ulay le?exol tiras The tokens using infinitival imperative verbs was elicited with the word bevakasa "please", which made it much easier for my consultant to interpret the sentence as a request. The data are much less regular than the corresponding examples with true imperative or future imperative verbs, so conclusions must be drawn hesitantly. (3 Id) and (3 If), in which the manner adverb precedes the imperative, are judged acceptable, while (3 lb) and (3 Ih), in which the same situation occurs but different lexical items are used, are judged as deviant. In (32c-h), there is no difference in judgement whether the verb precedes or follows temporal adverbs - either both pairs are grammatical or they are both ungrammatical. In (32a,b), however, there is a contrast between the sentence in which the verb precedes the adverb and the sentence in which the verb follows the adverb. Although the data are not as neat as the rest of the data in this chapter, a useful generalization can be drawn. Whenever different judgements were offered for two sentences in a pair, they always favoured the sentence in which the verb precedes the manner or temporal adverb, never the other way around. This provides some evidence that the same raising past T° that takes place with the other imperatives takes place with infinitival imperatives as well. In addition, infinitival imperatives are incompatible with speaker adverbs, just like true and future imperatives. We can conclude that despite their differing morphological forms, Hebrew imperatives have a uniform set of properties when it comes to verb raising and compatibility with speaker adverbs. This result reveals a fact about the role of morphology in raising: examining the data for suppletive imperatives and their morphological correspondents in the infinitive and future tense, we find that the raising properties of verbs do not depend on the verb's overt morphology, but rather on the syntactic type of the verb - that is, whether the verb is imperative or infinitive or finite. This can be seen by comparing the position relative to temporal adverbs of future tense verbs and suppletive imperatives with future form (29, repeated below). 60 (34) Future tense verbs a. ata hayom toxal tiras you (sg) today eat corn "you will eat corn today" b. ata toxal hayom tiras c. ata tamid toxal tiras you always eat com "You will always eat com" d. ata toxal tamid tiras e. ata kol yom toxal tiras you every day eat com "You will eat com every day" f. ata toxal kol yom tiras . g. ata bederex klal toxal tiras you usually eat com "You will usually eat com" h. ata toxal bederex klal tiras (29) Suppletive future verbs a. toxal hayom tiras! eat today com "Eat com today!" b. * hayom toxal tiras! c. toxal tamid tiras! eat always com "Always eat com!" d. *tamid toxal tiras! e. toxal kol yom tiras! eat every day com "Eat com every day!" f. *kol yom toxal tiras! g. *toxal bederex klal tiras! eat usually com h. *bederex klal toxal tiras! The difference between these two sets of data is the same difference noted earlier between finite verbs and imperatives: the future finite verbs may be either in T° or in 9°, while imperative verbs with future morphology may not be in T°, and taking into account the fact that they must precede manner adverbs (shown in (28)), must be at least as high as S ° . This means that suppletive future imperatives must undergo obligatory overt movement to a position where future tense verbs are not obligated to move. Since the verbs have identical overt morphology, it cannot be overt morphology alone that drives movement. 61 6.6 Summary We have seen that the structure that we have proposed to account for the various positions of English adverbs works nicely in Hebrew as well. Using the structure, we have investigated verb movement in finite, imperative and infinitive contexts, discovering that infinitive verbs appear to pattern with finite verbs, which can raise to T° or 9°. In contrast, the various forms of the imperative verb appear to raise farther, being unable to stop at T°. Whether Hebrew imperative verbs raise past 9° is not known because they are not compatible with speaker adverbs. The fact that imperative verbs pattern differently from other verb types in Hebrew indicates that what we discovered in the last chapter about English imperatives, that they pattern identically with other nonfinite verbs, is not universally true. The table in (35) summarizes our discoveries and the remaining uncertainties about the possible overt locations of different Hebrew verbs: (35) V° j O Agr u True imperatives X X ? Suppletive (infinitive) imperatives X? X? / ? ? Suppletive (future) imperatives X X / ? Finite verbs X / / X Infinitive verbs X / / ? Comparing suppletive imperative verbs with their counterparts in the infinitive and future tense, we were able to draw the conclusion that the movement properties of verbs do not correlate with the verb's overt morphology, but rather with the force of the sentence. To the extent that we can abstract from the somewhat fuzzy data in the case of suppletive infinitival imperatives, imperative verbs appear to all share the same movement properties, regardless of whether the verb's morphology is uniquely imperative, or whether it is shared with a finite or infinitive verb. 62 Chapter 7 - Summary We have distinguished direct imperatives, the class of imperatives that are syntactically recognizable as imperatives, from indirect imperatives, which have the form of declarative or interrogative sentences, but which are interpreted as commands, requests, or other speech acts canonically associated with imperatives. In languages like Hebrew, there is another main distinction that can be made - between true imperatives, which have a unique morphology, and suppletive imperatives, which use verbal morphology from a different paradigm - in the case of Hebrew, from the infinitive and future paradigms. We have compared the future suppletive imperative with an indirect imperative based on future tense declaratives, and have determined that they are not the same case, using as evidence the special negative al that is used in Hebrew future imperatives but not future indirect imperatives, and from the pragmatic implication of threat or hostility that is associated with the indirect imperative but not the direct one. Focusing on direct imperatives, we turned to their phrase structure and presented arguments that despite the fact that imperatives typically contain less material than declaratives, they have the full set of clausal projections that are found in declaratives. Evidence for this conclusion was drawn from the syntactic regularity of covert and overt subjects that suggests they are in [Spec,IP], from VP ellipsis, and in Hebrew from an agreement analysis and from a subject/object asymmetry in negative complement clauses. We then investigated verb movement in English and in Hebrew, using the overt position of verbs relative to different kinds of adverbs - manner, temporal and speaker - as a diagnostic for the position of verbs, and by assumption, a diagnostic for the extent of verb movement. We found that English imperatives pattern with nonfinite declaratives, in that main verbs do not move at all, while auxiliary verbs move only to T°, the head of the phrase immediately dominating VP. In Hebrew, we found that finite and infinitive verbs pattern together, being able to precede or follow temporal adverbs, indicating that they can be located in T° or in S ° - the head of a proposed phrase between AgrP and NegP. In contrast, imperatives of all types, true or suppletive, are not able to appear following a temporal adverb, indicating that they must be at least as high as S ° . The patterning of English imperatives with nonfinite verbs, and the distinctness of Hebrew imperatives, are shown in the following table. 63 V° •jf) g o Agr° , English All main verbs / X X X Imperative auxiliaries X / X X Nonfinite auxiliaries X / X X Finite Auxiliaries X X y / Hebrew Imperatives X X / ? Finite/Infinitives X / / X/? The table illustrates the fact that although imperatives may show the same movement properties as other nonfinite verbs in English, this is not a universal fact. A comparison of suppletive future imperatives with finite future tense verbs shows that the form of the verb alone does not determine its movement properties. From our conclusion in chapter 4 that the syntax of imperatives and finite declaratives is similar, a movement analysis that predicts the different behaviours of the two verb types cannot rely on syntactic triggers either. An analysis of movement that distinguishes interpretive possibilities, a proposal made by Fox (2000), offers a more promising approach for future research into this question. 64 References Acquaviva, Paolo (1997). The Logical Form of Negation : A Study of Operator-Variable Structures in Syntax. New York: Garland. Akmajian, Adrian (1984). Sentence Types and the Form-Function Fit Natural Language and Linguistic Theory 2: 1-23. Aoun, Joseph and Dominique Sportiche (1983). On the Formal Theory of Government. Linguistic Review 2: 211-236. Belletti, Adriana (1990). Generalized Verb Movement: Aspects of Verb Syntax. Torino: Rosenberg and Sellier. Beukema, Frits and Peter Coopmans (1989). A Government-Binding Perspective on the Imperative in English. Journal of Linguistics 25: 417-136. Borer, Hagit (1995). The Ups and Downs of Hebrew Verb Movement. Natural Language and Linguistic Theory 13: 527-606. Chomsky, Noam (1981). Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam (1986). Barriers. Cambridge, MA: MIT Press. Chomsky, Noam (1995). A Minimalist Program for Linguistic Theory, p. 167-217 in Chomsky, Noam, The Minimalist Program. Cambridge, MA: MIT Press. Cinque, Guglielmo (1999). Adverbs and Functional Heads: A Cross-Linguistic Perspective. New York: Oxford University Press. Emonds, Joseph (1978). The Verbal Complex V'-V in French. Linguistic Inquiry 9: 151-175. Fodor, Janet (1977). Semantics: Theories of Meaning in Generative Grammar. Hassocks: Harvester Press. Fox, Danny (2000). Economy and Semantic Interpretation. Cambridge, MA: MIT Press. Grice, H.P. (1968). Logic and Conversation. Manuscript. 65 Haegeman, Liliane (1995). The Syntax of Negation. New York: Cambridge University Press. Haegeman, Liliane and Raffaella Zanuttini (1996). Negative Concord in West Flemish, p. ???? in Belletti, Adriana and Luigi Rizzi (eds). Parameters and Functional Heads: Essays in Comparative Syntax. New York: Oxford University Press. Han, Chung-hye (1998). The Structure and Interpretation of Imperatives: Mood and Force in Universal Grammar. PhD dissertation, University of Pennsylvania. Iatridou, Sabine (1990). About AGR(P). Linguistic Inquiry 21: 551-577. Jackendoff, Ray (1972). Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press. Johnston, Michael (1994). Because-dauses and Negative Polarity Licensing. In Proceedings of the Eastern States Conference on Linguistics. Ithaca: DMLL. Katz, Jerrold and Paul Postal (1964). An Integrated Theory of Linguistic Descriptions. Cambridge, MA: MIT Press. Kayne, Richard (1975). French Syntax. Cambridge, MA: MIT Press. Kayne, Richard (1981). ECP Extensions. Linguistic Inquiry 12: 93-133 Kayne, Richard (1994). The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Klima, Edward (1964). Negation in English. In Fodor, Jerry and Jerrold Katz (eds.), The Structure of Language: Readings in the Philosophy of Language. New Jersey: Prentice Hall. Kratzer, Angelika (1989). An Investigation of the Lumps of Thought. Linguistics and Philosophy 12: 607-653. Lakoff, Robin (1968). Abstract Syntax and Latin Complementation. Cambridge, MA: MIT Press. Lasnik, Howard and Mamoru Saito (1992). Move-a: Conditions on its Application and Output. Cambridge, MA: MIT Press. Lobeck, Anne (1995). Ellipsis: Functional Heads. Licensing, and Identification. New York: Oxford University Press. McCloskey, James (1997). Subjecthood and Subject Positions, p. 197-235 in Haegeman, Liliane (ed.), Elements of Grammar: Handbook of Generative Syntax. Dordrecht: Kluwer. Pirvulescu, Mihaela and Yves Roberge (1999). Objects and the Structure of Imperatives. To appear in Authier, Marc, Bullock, Barbara and Lisa Reed (eds.), Formal Perspectives on Romance Linguistics. Amsterdam: John Benjamins. 66 Platzack, Christer and Inger Rosengren (1994). On the Subject of Imperatives: A Minimalist Account of the Imperative Pronoun and Negated Imperatives. Manuscript, Lund University. Pollock, Jean-Yves (1989). Verb Movement, Universal Grammar, and the Structure of IP. Linguistic Inquiry 20: 365-464. Portner, Paul (1997). The Semantics of Mood, Complementation, and Conversational Force. Natural Language Semantics 5: 167-212. Potsdam, Eric (1996). Syntactic Issues in the English Imperative. PhD dissertation, University of California, Santa Cruz. Progovac, Ljiljana (1994). Negative and Positive Polarity: A Binding Approach. New York: Oxford University Press. Roberts, Ian (1992). Verbs in Diachronic Syntax: A Comparative History of English and French. Dordrecht: Kluwer. Rohrbacher, Bernhard (1994). The Germanic VO Languages and the Full Paradigm: A Theory of V to I Raising. PhD dissertation, University of Massachusetts, Amherst. Rivero, Maria-Luisa (1994). Negation, Imperatives and Wackernagel Effects. Rivista di Linguistica 6: 39-66. Rizzi, Luigi (1990). Relativized Minimality. Cambridge, MA: MIT Press. Rizzi, Luigi (1997). The Fine Structure of the Left Periphery, p. 281-337 in Haegeman, Liliane (ed.), Elements of Grammar: Handbook of Generative Syntax. Dordrecht: Kluwer. Rochemont, Michael (1989). Topic Islands and the Subjacency Parameter. Canadian Journal of Linguistics 34: 145-170. Ross, John (1970). On declarative sentences. In Jacobs, Roderick and Peter Rosenbaum (eds.), Readings in English Transformational Grammar. Waltham: Ginn. Sadock, Jerrold (1974). Toward a Linguistic Theory of Speech Acts. New York: Academic Press. Sadock, Jerrold and Arnold Zwicky (1985). Speech Act Distinctions in Syntax, p. 155-196 in Shopen, Timothy (ed.) Language Typology and Syntactic Description, Vol. I.Cambridge, England: Cambridge University Press. Shlonsky, Ur (1997). Clause Structure and Word Order in Hebrew and Arabic : An Essay in Comparative Semitic Syntax. New York: Oxford University Press. Sportiche, Dominique (1988). A Theory of Floating Quantifiers and Its Corollaries for Constituent Structure. Linguistic Inquiry 19: 425-449. Travis, Lisa (1984). Parameters and Effects of Word Order Variation. PhD dissertation, MIT. 67 Uribe-Echevarria, Maria (1994). Interface Licensing Conditions on Negative Polarity Items: A Theory of Polarity and Tense Interactions. PhD dissertation, University of Connecticut. Vikner, Sten (1995). Verb Movement and Expletive Subjects in the Germanic Languages. New York: Oxford University Press. Zanuttini, Raffaella (1994). Speculations on Negative Imperatives. Rivista di Linguistica 6: 119-142. Zanuttini, Raffaella (1997). Negation and Clausal Structure: a comparative study of Romance languages. New York: Oxford University Press. Zhang, Shi (1990). The Status of Imperatives in Theories of Grammar. PhD dissertation. University of Arizona. 68 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items