<<

Advanced Generative

Michael Barrie Sogang University

December 2, 2020 2 Contents

1 Foundations of 7 1.1 Historical Background ...... 7 1.2 Methodological Concerns ...... 11 1.3 Syntactic Theorizing ...... 15 Further Reading ...... 17

2 The , Features, and Phrase Structure 19 2.1 The Lexicon ...... 19 2.2 Features ...... 20 2.2.1 Types of Features ...... 21 2.2.2 Subcategorization ...... 27 2.2.3 Agree ...... 28 2.3 Bare Phrase Structure ...... 31 2.4 Clausal Structure ...... 32 2.5 Configurationality ...... 35 Further Reading ...... 38

3 Ditransitives and the vP Layer 39 3.1 VP-Internal Hypothesis ...... 39 3.2 Severing the External ...... 44 3.3 Ditransitives ...... 47 3.4 vP Refined ...... 52 Further Reading ...... 55

4 Nominal Syntax 57 4.1 DP Theory ...... 57 4.2 Nominalizers and Roots ...... 59 4.3 Number ...... 65

3 4 CONTENTS

4.4 Case ...... 68 Further Reading ...... 73

5 The Architecture of Syntax 75 5.1 Phrasal Movement and Head Movement ...... 75 5.2 Transfer and the Interfaces ...... 79 5.3 Phase Theory ...... 83 Further Reading ...... 92

6 Case and A-Movement 93 6.1 Case Assignment ...... 93 6.1.1 Nominative Case ...... 94 6.1.2 Accusative Case ...... 97 6.2 A-Movement ...... 103 6.3 Unaccusativity ...... 106 6.4 Ergativity and the Dependent Theory of Case ...... 110 6.5 Properties of A-Movement ...... 111 Further Reading ...... 113

7 A-Bar Movement 115 7.1 Introduction ...... 115 7.2 Constraints on A-Bar Movement ...... 115 7.3 The Mechanism of A-Bar Movement ...... 117 7.4 Wh-in-situ ...... 121 7.5 Multiple Wh Movement ...... 127 7.6 Partial Movement and Marking Constructions ...... 131 Further Reading ...... 134

8 and 135 8.1 Introduction ...... 135 8.2 Diagnostics for Raising and Control ...... 137 8.2.1 Idiom Tests ...... 138 8.2.2 Expletive Subjects ...... 140 8.2.3 Transparency ...... 140 8.2.4 Scope Reconstruction ...... 141 8.3 The Analysis of Raising and Control ...... 142 8.3.1 GB Analysis of Raising and Control ...... 142 8.3.2 Raising to ...... 144 CONTENTS 5

8.3.3 Control as Movement ...... 146 8.4 Partial Control ...... 148 8.5 Control into Subjunctive ...... 154 8.6 Non-obligatory Control ...... 154 Further Reading ...... 157

9 Linearization 159 9.1 Introduction ...... 159 9.2 The Headedness Parameter ...... 159 9.3 Antisymmetry ...... 161 9.4 Dynamic Antisymmetry ...... 168 Further Reading ...... 170 Index ...... 170 Bibliography ...... 173 6 CONTENTS Chapter 1

Foundations of Generative Grammar

1.1 Historical Background

This book is an advanced introduction to generative grammar. It assumes a prior familiarity with general linguistic theory and some familiarity with basic generative grammar. In this introduction we briefly cover some historical aspects of how generative linguistics came into being followed by a discussion of the core properties of generative linguistics. Until the mid nineteenth century linguistic scholarship was dominated by . Scholars such as Sir William Jones, Grimm, and others were concerned with the historical relation- ships among languages and reconstructing proto-languages. In the mid-nineteenth century Ferdinand de Saussure proposed that language is an abstract system consisting of units of sound and units of meaning. This view of language grew into Structuralist Linguistics (구조주의 언어학). Structural- ist linguistics attempts to understand languages and dialects in synchronic terms, divorced from their historical predecessors (although structuralists certainly believed that both synchronic and diachronic approaches to the study of language were necessary). To be concrete, when students take introduction to linguistics at most universities and are introduced to and , most of the procedures covered are the same procedures developed by structuralists. Structuralist linguistics is the study of how the units of sound and the units of meaning are put together to form language. This view continued to grow both in Europe and in the United States. The structuralist approach began to mesh with similar approaches in China, although similar approaches had already been established in China much earlier (Wang and 王士元, 1989). In North America Structuralist Linguistics blossomed under the work of scholars such as Bloom- field, Haas, Hockett, and Martinet. In Europe, the Prague School led by Jakobson and Trubetzkoy took hold. The two approaches were largely similar, although American Structuralism was influ- enced by advances in behavioural psychology, notably by B. F. Skinner. Meanwhile, in China, Chao Yuen Ren worked with Chinese scholars on Historical Chinese linguistics, integrating both West-

7 8 CHAPTER 1. FOUNDATIONS OF GENERATIVE GRAMMAR ern and Eastern approaches to the study of language. What all these approaches had in common was a methodical approach to understanding how the bits of sound and the bits of meaning were put together. Structuralist phonology and morphology came to full fruition under these approaches and was applied to hundreds of languages around the world. Here are two Korean examples of the Structuralist approach to language.

(1.1) Allophones in Korean

(a) /s/ has two allophones: [s] and [S] (b) /s/ → [S] / __ i

(1.2) Allomorphs in Korean

(a) nom has two allomorphs: /-i/ and /-ka/ (b) nom → [-i] / C __, elsewhere /-ka/

A major shift in the field of linguistics took place in the 1950s with Chomsky’s The Logical Struc- ture of Linguistic Theory (Chomsky, 1955).1 The much shorter Syntactic Structures is a published version of LSLT, which gives a concise discussion of Chomsky’s work. Chomsky harshly refuted the behaviourist approach to language (Chomsky, 1959) and proposed an innate (UG) as a model for . Concomitantly, he proposed that a model of grammar should be able to account for all and only the grammatical sentences of a language. We say that a model consists of a set of rules that generates sentence, hence this approach is called generative grammar (생성문법). The first instantiation of Generative Grammar is now referred to as Standard Theory (Chom- sky, 1957, 1965). It consists of a language specific Lexicon and a set of language-specific Phrase Structure Rules (PSR). Here is a basic example for English.

(1.3) (a)S → NP VP (b)NP → (Det) N (c)VP → V (NP) (d)N → {Susan, John, apple} (e)V → {sees, eats, cries} (f) Det → {the}

These rules are capable of generating the following sentences.

1LSLT was written in 1955, but was not published until 1975. 1.1. HISTORICAL BACKGROUND 9

(1.4) (a) S

NP VP

N V NP

Susan sees N

John (b) S

NP VP

N V

John cries (c) S

NP VP

N V NP

Susan eats Det N

the apple

The following data are problematic for the mini-grammar in (1.3. Our mini-grammar generates the ungrammatical sentences in (1.5) and (1.6), and it fails to generate the grammatical sentence in (1.7). Thus, we say the grammar in (1.3 overgenerates (1.5 and (1.6), and it undergenerates (1.7).

(1.5) *Susan cried the apple.

(1.6) *The apple sees John.

(1.7) Susan eats the apple with a knife.

We will not attempt refine our mini-grammar here. For a contemporary, pedagogical approach to Standard Theory, see Larson (2009). Standard Theory involved numerous language-dependent transformations. By the late 1960s Chomsky had already laid out his idea of Universal Grammar. The next revolution in genera- tive grammar, Extended Standard Theory was introduced in Chomsky (1973). Here, Chomsky 10 CHAPTER 1. FOUNDATIONS OF GENERATIVE GRAMMAR proposed "Conditions on Transformations", whereby, as the name suggests, transformations were subject to strict conditions. The idea is that the child has a much narrower range of choices to consider during acquisition. It was during Extended Standard Theory that X-Bar Theory became standard (Jackendoff, 1977; Stowell, 1981; Chomsky, 1970). In Stardard Theory, nominalizations were argued to be derived from a common underlying clausal core (Lees, 1960). This argument was based on numerous parallels between clausal and nominal constructions.

(1.8) Parallels between clausal and nominal projections

(a) The enemy destroyed the city. (b) the enemy’s destruction of the city (c) The actor portrayed himself. (d) the actor’s portrayal of himself

Chomsky (1970) argued that while many nominals can be derived from underlying clausal con- structions, not all can, as in the following examples.

(1.9) (a) the author of the book (b) Mary’s car

Structurally, of course, the nominals in (1.9) and the nominals in (1.8) have the same structure; however, if the nominals in (1.8) are derived from underlying clauses, then we have no way to derive the nominals in (1.9). We don’t want to lose the fact that all nominals have the same basic structure, however. To overcome this conundrum, Chomsky proposed that lexical categories have the same universal phrase structure rules. The PSRs shown here are updated slightly. X stands for any lexical category.

(1.10) general phrase structure rules

(a)XP → SpecXP X (b)X → X ...

These PSRs underlie Stowell’s 1981 development of X-Bar Theory. Stowel argued that if a sentence consists of a subject, INFL, and a VP, then either INFL or V must be the head of S. Given the large number of subject-object asymmetries and the dissociation of the subject from VP (Kratzer, 1996), Stowel reasoned that INFL is the head of the sentence. He then applied the rules in (1.10) to give the following. 1.2. METHODOLOGICAL CONCERNS 11

(1.11) PSRs for INFL

(a)IP → Subject I (b)I → I VP

Stowell further went on to generalize the now well known X-Bar schema, dispensing with phrase structure rules.

(1.12) XP

Specifier X0

X

Revised Extended Standard Theory (REST) represented a further shift into general rules and away from language-specific transformations (Chomsky and Lasnik, 1977). Chomsky and Lasnik were concerned with the opposition between descriptive and explanatory adequacy. In order for a grammar to be explanatorily adequate, it must be acquirable by children; hence, it must be simple. In order for a grammar to be descriptively adequate, it must capture the range of descriptive generalizations found in the world’s languages; hence, it must be rich enough to account for this variation. They explored some of these concerns in their 1977 paper. It was during REST that the now familiar T model began to take its now familiar form, Figure (1.1). Finally, the Government and (GB) approach took hold in the early 1980s (Chomsky, 1986, 1981a; Lasnik and Saito, 1992). The GB approach saw the creation of modules of grammar, including X-Bar Theory, the Control Module, the Binding Module, and so forth. GB Theory is characterized by an ever increasingly complex network of theoretical proposals. Part of the increase in complexity is due to the greater empirical breadth of the field. Syntacticians were now immersed in a much wider cross-linguistic base targeting languages in all four corners of the earth.

1.2 Methodological Concerns

Chomsky proposed three levels of adequacy that any linguistic theory must adhere to (Chomsky, 1964). It is the first of these that is of primary concern in this section.

(1.13) Levels of Adequacy

(a) Observational Adequacy - Does the theory cover an accurate description of the grammatical and ungrammatical sentences involved? (b) Descriptive Adequacy - Does the theory make any meaningful or insightful generalizations on the data? 12 CHAPTER 1. FOUNDATIONS OF GENERATIVE GRAMMAR

Base Lexicon Lexical insertion rules X-Bar Theory subcategorization

Deep Structure

Transformations

Surface Structure

Universal

Figure 1.1: REST Model 1.2. METHODOLOGICAL CONCERNS 13

(c) Explanatory Adequacy - Does the theory account for how children can acquire the proposed mechanism?

The concept of observational adequacy seems simple, but can be the source of methodological problems in linguistic research. First, let’s look at what we mean by this term. In simple terms, it means we have an accurate description of the set of grammatical and ungrammatical sentences under discussion. As an example, most English speakers generally agree on the judgements in the following paradigm.

(1.14) (a) Who do you think John met? (b) Who do you think that John met? (c) Who do you think met John? (d) *Who do you think that met John?

This phenomenon is the well known that-trace effect (Perlmutter, 1971; Chomsky, 1986). Observational adequacy is simply getting the data right. This is not always an easy task. Often when we work on an unfamiliar language, we may get the facts wrong; however, this can happen in one’s native language, too. This calls for sound methodological practices when conducting linguistic research. There are at least two issues with presenting a test sentence alone in written form. First, natural language is spoken or signed. Written language is an abstraction from natural human language. One aspect that this abstraction is missing is intonation. Since Pierrehumbert and Hirschberg (1990) it is acknowledged that intonational contours contribute to the meaning of a sentence. Consider the following English examples.

(1.15) (a) What did John eat? John ate an apple. (b) Who ate an apple? John ate an apple.

In the question/answer pairs above, the intonational contours of the two answers are quite different. Here’s a Korean example.

(1.16) (a) 밥 먹었어? pap mek-ess-e? rice eat-pst-informal ‘Did you eat?’ (b) 응, 법 먹었어 ung, pap mek-ess-e yeah rice eat-pst-informal ‘Yeah, I ate.’ 14 CHAPTER 1. FOUNDATIONS OF GENERATIVE GRAMMAR

Again, the intonational contours differ between the two sentences. Such intonational differences can give rise to incorrect or misleading results if not properly controlled for. Consider the following exchange.

(1.17) Snippet of a Field Research Session

EXPERIMENTER Can you say, "John pushed the boy with a stick."? CONSULTANT Yes EXPERIMENTER Does it mean that John used the stick to push the boy or that it was the boy with a stick that John pushed. CONSULTANT John used the stick to push the boy EXPERIMENTER Can it mean that the boy was holding the stick? CONSULTANT No.

The sentence John pushed the boy with a stick is indeed ambiguous in its written form, but the two meaning have different intonational contours (Jun, 2003; Carlson et al., 2001). If the experi- menter uses only one such contour during elicitation the consultant will not necessarily have the other possible meaning in mind. also plays an important role in sentence comprehension (Matthewson, 2004; Bochnak and Matthewson, 2020; Tonhauser and Matthewson, 2015; Deal, 2015). Sentences are often accept- able only in certain contexts. If the experimenter and the consultant have different contexts in mind, the consultant may give responses that do not correspond to what the experimenter is testing. Thus, a clear context should always be provided. Here is an example from Cable (2012) testing reflexivity in French.

(1.18) Reflexive and Reciprocal Scenarios

1 Reflexive scenario: Each boy slapped himself. Dave slapped himself. Tom slapped himself. Bill slapped himself.

2 Reciprocal scenario: Each boy slapped some other boy. Dave slapped Tom. Tom slapped Bill. Bill slapped Dave.

The experimenter describes both contexts (‘scenario’ in Cable’s terms) to the consultant and asks if the test sentences can be used to truthfully describe either of these contexts.

1 Judgment: Can truthfully describe both scenarios above. Les étudiants se sont frappés. the students refl aux hit ‘The students slapped themselves.’ 1.3. SYNTACTIC THEORIZING 15

2 Judgment: Can truthfully describe only the second scenario above. Les étudiants se sont frappés l’un l’autre. the students refl aux hit the one the other ‘The students slapped themselves.’

Tonhauser and Matthewson (2015) gives the following enlightening example of how the lack of a context can give rise to unreliable results. They note that Moltmann (2013) gives the following example with no context, along with the explanation that it "doesn’t sound right" and ?? as the grammaticality judgment.

(1.19) ??Socrates is a man. (Moltmann, 2013, p.36)

As Matthewson and Tonhauser explain, no context is given for the judgment of this sentence. We are left to assume that we’re talking about the famous Greek philosopher. We don’t know whether the sentence was uttered by someone recently (2020 at the time of writing) or by someone when Socrates was alive. It turns out English does have "life time effects" on sentences. That is, present tense statements about a person can typically only be made if that person is alive. Matthewson and Tonhauser then give the following Gitksan (Tsimshianic, ISO git) example (gloss slightly simplified). In this case, they merely asked the Gitksan speaker to judge whether the sentence was acceptable or not.

(1.20) Gyat=t Saklatiis man=det Socrates ‘Socrates is a man.’

The Gitksan speaker judged this sentence to be acceptable. It is not clear, now, what gives rise to the difference between the English and the Gitksan examples. The difference could be due to an actual difference in the grammars of these languages, that is, perhaps Gitksan doesn’t exhibit life-time effects on statements about people. Or, it could simply be that the Gitksan speaker had a different context in mind when judging the sentence.

1.3 Syntactic Theorizing

We now move on to the other two levels of adequacy developed by Chomsky. Descriptive adequacy making an accurate and meaningful generalization about the data. Considering the data in (1.14) again, this paradigm represents the now well accepted generalization that the sequence of an overt followed by a trace is ungrammatical. The generalization is often stated as follows.

(1.21) * ... that t ... 16 CHAPTER 1. FOUNDATIONS OF GENERATIVE GRAMMAR

Making accurate generalizations requires gathering large amounts of data to make sure your observation is not just an idiosyncratic fact about a small number of cases. Finally, explanatory adequacy involves proposing an analysis for a descriptive generalization that can be acquired by child-learners. Thus, if the linguist proposes a or a parameter setting in an analysis, there must be clues in the signal available to the child. 1.3. SYNTACTIC THEORIZING 17

Further Reading

• Chomsky (1965)

• Bochnak and Matthewson (2020) - an indispensable guide to linguistic methodology

• Newmeyer (2014) - A brief history of linguistics in the US, starting with its beginnings in Europe

• Harris (1995) - This monograph is a more frank look at disagreements between generative linguistics and other approaches to linguistics. 18 CHAPTER 1. FOUNDATIONS OF GENERATIVE GRAMMAR Chapter 2

The Lexicon, Features, and Phrase Structure

2.1 The Lexicon

The concept of a lexicon surely goes back to Antiquity. The introduction of Contini-Morava and Tobin (2000) gives a good historical overview of the concept of the lexicon. The lexicon is the repository of all the form-meaning matches of a language. The lexicon contains all the unpredictable information about words. Consider the following list of words.

(2.1) cat, cats, dog, dogs, foot, feet

Plurals in English are largely predictable, so the mental lexicon for English does not need to list cats and dogs separately. However, foot has an unpredictable , feet, so it must appear in the lexicon. Based on these data, we can deduce the following mini-lexicon for English.

(2.2) Mini-lexicon of English

(a) cat (b) dog (c) foot (d) feet (e) -s

The lexicon must also encode properties of the individual lexical items that are unpredictable. These properties are necessary to know how to use the lexical items in language. For instance, cat must indicate that it is a . It may encode other properties such as [animate] as is grammatically encoded in many languages. Here is a more detailed mini-lexicon for English.

19 20 CHAPTER 2. THE LEXICON, FEATURES, AND PHRASE STRUCTURE

(2.3) Mini-lexicon of English

(a) cat, N (b) dog, N (c) FOOT, N, {foot/singular; feet/plural} (d) ox, N (e) CHILD, N {child/singular; childr/plural} (f) pl, NUM, {/FOOT; -en/OX,CHILD; -s/elsewhere}

We won’t worry about the at this stage as we will sharpen it later. Words in capital letters indicate abstract forms of the root. Specific forms are indicated in curly brackets with the environment specified after the slash sign. For example, the abstract form FOOT has two realiza- tions: foot in the singular and feet in the plural. The plural marker, pl, has three (allomorphs or realizations. It is null for FOOT (also for sheep, and fish and many other words). It is -en for OX and CHILD. It is -s elsewhere. The derivation of children is shown below.

(2.4) Derivation of children

underlying form CHILD-pl select realization for CHILD childr-pl select allomorph for pl childr-en surface form children

To derive the word children the underlying forms CHILD and pl are selected. The form childr is chosen in response to the presence of plural number. The appropriate allomorph for plural number is chosen. According to the mini-lexicon, the appropriate plural number form for CHILD is -en. This is the basic structure of the lexicon - a collection of unpredictable lexical and grammatical information. In the next section we will see how the properties of the lexical items, the features are encoded.

2.2 Features

Lexical items have properties that have morphosyntactic manifestations throughout the sentence. Consider the following examples.

(2.5) Gender in French (fra)

(a) La pomme est vert-e the.fem apple is green-fem ‘The apple is green.’ 2.2. FEATURES 21

(b) Le crayon est vert the.masc pencil is green.masc ‘The pencil is green.’

Observe that the form of the and the form of the change depending on which lexical noun is present. Consider also the following data from the Busan dialect of Korean (Barrie and Lee, 2017). Here, the form of the complementizer varies with respect to the type of (polarity question or content question) and the type of ( or (연결동사/BE 동사)).

(2.6) Categorial dependencies in Busan Korean

(a) 니가 책을 읽나? ni-ka chayk-ul ilk-na? you-nom book-acc read-comp ‘Are you reading a book?’ (b) 니가 뭐를 읽노? ni-ka mwe-lul ilk-no? you-nom book-ACC read-COMP ‘What are you reading?’ (c) 저 사람이 수민이가? Ce salam-i Swumin-i-ka? that person-nom Swumin-cop-comp ‘Is that person Swumin?’ (d) 저 사람이 누고? Ce salam-i nwu-Ø-ko? that person-nom who-cop-comp ‘Who is that person?’

Lexical items are thought to have features, which indicate its special properties that give rise to the morpho-syntactic reflexes observed. Let’s take the French example first. The noun pomme (‘apple’) has the gender [fem] (feminine). That is, the lexical entry for pomme includes the fact that it has this feature. Likewise, the noun crayon (‘pencil’) has the gender feature [masc] (masculine). There are two singular definite in French. They both have roughly the same meaning as English the; however, the form la appears with feminine and the form le appears with masculine nouns. That is, the form of the determiner varies with the gender of the noun it modifies. Also, the form of the adjective agrees with the gender of the noun it predicates.

2.2.1 Types of Features

The notion of features in morphosyntax was borrowed from phonology. Specifically, Sound Pat- terns of English (Chomsky and Halle, 1968) made use of binary features (also called equipollent 22 CHAPTER 2. THE LEXICON, FEATURES, AND PHRASE STRUCTURE features). Each feature has a positive or negative value. A particularly well known use of binary features for lexical categories was proposed in Chomsky (1970). Two features were proposed [±N] and [±V].

(2.7) Nominal and Verbal Features - Binary

Nouns [+N, -V] [-N, +V] Prepositions [+N, +V] [-N, -V]

Although we understand lexical categories differently today, this early representation illustrates how binary features are used to classify lexical categories. A revision in our understanding of features again originally came from phonology. Linda Lom- bardi proposed privative features - features that are either present or absent rather than binary (Lombardi, 1995, 1991; Steriade, 1995). Thus, a feature is not positively valued, it is either present or absent. Here are the same nominal and verbal features from above with privative specifications.

(2.8) Nominal and Verbal Features - Privative

Nouns [N] Verbs [V] Prepositions [N, V] Adjectives []

These two types of feature systems are not merely notational variants. They make different predictions about how lexical categories behave. The following natural classes are predicted by the two feature systems.

(2.9) Predicted Natural Classes of Lexical Categories

Binary Features Privative Features Nouns and Prepositions both are [+N] Nouns and Prepositions both are [N] Verbs and Prepositions both are [+V] Verbs and Prepositions both are [V] Verbs and Adjectives both are [-N] Nouns and Adjectives both are [-V]

In general, privative features predict far fewer natural classes than binary features. It has been noted, however, that any random grouping of lexical categories can produce some kind of generaliza- tion (Déchaine, 1993; Stuurman, 1985). As a result the feature-based approach to lexical categories is no longer pursued. We come back to lexical categories in later chapters. 2.2. FEATURES 23

Number features have figured prominently in recent discussions on morphosyntactic features, particularly in relation to dual number. Here is an example from the Inuttut dialect of Inuktitut (iku), an Eskimo-Aleut language spoken in the northeast of Canada. We will come back to dual number after introducing feature geometries.

(2.10) Number in Inuttut (Gillon, 2015)

(a) nanuk ‘(one) polar bear’ (b) nanok ‘two polar bears’ (c) nanuit ‘three or more polar bears’

A feature geometry is another concept borrowed from phonology (Avery and Rice, 1989; Clements, 1985; Clements and Hume, 1995). Harley and Ritter (2002) argued that morphosyntactic features are also arranged in a feature geometry. Here is the universal geometry they propose. In the feature geometry Small Caps indicates an organizing node, and underlining indicates a default feature. An organizing node with no dependents receives a default interpretation. Thus, if there is no dependent of the Participant node, it is interpreted as the speaker (1st person).

(2.11) Referring Expression

Participant Individuation

Speaker Addressee Group Minimal Class

Augmented Animate Inanimate/Neuter

Feminine Masculine

The Participant node indicates either first or second person. Following Benveniste (1966, 1971) and Forchheimer (1953) Harley and Ritter propose that third person is not a and is represented by the lack of the Participant node (though see Nevins (2007b) for an opposing view). They note numerous asymmetries between 1st and 2nd person on the one hand with 3rd person on the other. To take one example, many languages have 1st and 2nd person , but lack 3rd person pronouns or use in lieu of 3rd person pronouns. Korean is such a languages. Many languages make different grammatical distinctions for 3rd person compared to 1st and 2nd person. They give the following Maltese (mlt) and Lyélé (lee) data. Maltese makes gender distinctions for 3rd person only. Lyélé makes formality distinctions in 1st and 2nd person only, while 3rd person makes a wide range of other distinctions. 24 CHAPTER 2. THE LEXICON, FEATURES, AND PHRASE STRUCTURE

(2.12) Nominative Pronouns of Maltese (Sutcliffe, 1936)

singular plural masculine feminine 1st jien(a) aèna 2nd int(i) intom 3rd hu(wa) hi(ja) huma

(2.13) Nominative Pronouns of Lyélé (Showalter, 1986)

singular plural 1st á né 1st formal né 2nd `n á 2nd formal á 3rd mass me 3rd human ń be 3rd augmentative o re 3rd diminutive e se 3rd elsewhere re se

The individuation node contains information about number and . The feature Group encodes a non-atomic interpretation. That is, it excludes a set of only one member. It typically gives rise to a plural interpretation. The default Minimal feature encodes the meaning of smallest possible number of members. In most cases it gives rise to a singular interpretation. The feature Augmented encodes a paucal interpretation for languages with paucal number. Let’s illustrate the system with four examples here. The first illustration is from Daga (DGZ), a Trans-New Guinea language from Papua New-Guinea, and illustrates a straightforward system with three persons and two numbers.

(2.14) Daga Pronouns (Murane, 1974)

singular plural 1st ne nu 2nd ge e 3rd me mu Here are the feature geometries that Harley and Ritter propose for the Daga pronouns.

(2.15) (a) 1st person RP RP

Part Indv Part Indv

Speaker Min Speaker Group 2.2. FEATURES 25

(b) 2nd person RP RP

Part Indv Part Indv

Addressee Min Addressee Group

(c) 3rd person RP RP

Indv Indv

Min Group

The Kahlina (car) language (a Cariban language spoken in the northern part of the Atlantic coast of South America, also known as ‘Carib’) encodes a distinction in the first person. We present the whole paradigm for , but give the feature geometries for the 1st person pronouns only.

(2.16) Kalihna pronouns (Hoff, 1968)

singular plural 1st.excl au aPna 1st.incl kïxko kïxka:ro 2nd amo:ro amïiyaro 3rd moxko moxka:ro The first person, inclusive, singular is understood as dual since an inclusive referent must contain at least two people. Here are the feature geometries for the first person pronouns. Note that the feature Minimal accurately captures the fact that a Participant node containing both the speaker and the addressee minimally contains two people.

(2.17) (a) 1st exclusive RP RP

Part Indv Part Indv

Speaker Min Speaker Group

(b) 1st inclusive RP RP

Part Indv Part Indv

Speaker Addressee Min Speaker Addressee Group

Tonkawa (tqw) is an extinct once spoken in the US. It has dual number as the following pronouns show. 26 CHAPTER 2. THE LEXICON, FEATURES, AND PHRASE STRUCTURE

(2.18) Tonkawa nominative pronouns (Hoijer, 1933)

singular dual plural 1st ca;ya geuca;ya geuca;ga 2nd na;ya wena;ya wena;ga 3rd ’aye;la ’awe;la ’awe;ga

Observe in the following feature geometries that dual number is formed by a combination of the features Minimal and Group. The feature group means that the set size is greater than one, and the feature minimal means that the set must be as small as possible. The smallest number greater than one is two, so this gives rise to a dual interpretation.

(2.19) singular dual plural

(a) 1st person RP RP RP

Part Indv Part Indv Part Indv

Speaker Min Speaker Min Group Speaker Group

(b) 2nd person RP RP RP

Part Indv Part Indv Part Indv

Addressee Min Addressee Min Group Addressee Group

(c) 3rd person RP RP RP

Indv Indv Indv

Min Min Group Group

Harley and Ritter acknowledge that this representation of dual number is unconventional. They present the following Hopi (hop) data as evidence for the geometries in (2.19).

(2.20) Hopi (Harley and Ritter, 2002, ex.19, citing Ken Hale, p.c.)

(a) Pam taaqa wari that man ran.sg ‘That man ran’ (b) Puma taP-taq-t yuPtu those pl-man-grp ran.pl ‘Those men ran. (three or more)’ 2.2. FEATURES 27

(c) Puma taaqa-t wari those man-grp ran.sg ‘Those two men ran.’

First, observe that there is a suppletive form of the verb that distinguishes singular from plural. Note crucially that the dual form is composed of a group morpheme and the singular form of the verb. Harley and Ritter take this as evidence for their analysis as the Group feature triggers group morphology on the noun and the Minimal feature triggers singular morphology on the verb. has played a large role in morpho-syntax since the feature-geometric analysis was first proposed (Cowper, 2005; Nevins, 2007a; Harbour, 2011). Finally, polyvalent features have been proposed. A polyvalent feature is a feature that can have one of several values. Case is often represented as a polyvalent feature (Adger, 2003, 2010).

(2.21) Examples of Case Features

(a) [case:nom] (b) [case:acc] (c) [case:dat]

Crucially, we want a feature system that’s flexible enough to account for the range of properties found in human language, but restrictive enough so as not to over-generate languages that don’t exist. Additionally, we want to use the simplest feature system possible. Thus, we adopt a more complicated feature system only when the data forces us to. In general for the discussion here, we will use privative features, but may appeal to other feature systems when the need arises.

2.2.2 Subcategorization

Lexical items have various restrictions on what kind of arguments they can take. The first type of restriction we deal with concerns restrictions on the categorial type of the argument (Grimshaw, 1979). For instance, the verb push can’t exist by itself. You have to push something (push the car, push the couch, push a box). So we say the verb push takes a noun phrase as a complement. We refer to this type of restriction as a subcategorizational restriction. For convenience, we also refer to this type of selection as c-selection, where c stands for category. For example, the lexical item know subcategorizes for or c-selects either a noun phrase (NP) or a clause (S), whereas the lexical item wonder subcategorizes for or c-selects only a clause.

(2.22) (a) Mary knows generative syntax. (b) Mary knows if John likes syntax. (c) *Mary wonders generative syntax. 28 CHAPTER 2. THE LEXICON, FEATURES, AND PHRASE STRUCTURE

(d) Mary wonders if John likes syntax.

Such selectional restrictions are pervasive throughout the world’s languages. In Korean, 싶다 takes only a clausal complement while 원하다 takes only a nominal complement.

(2.23) (a) 민수가 사과를 먹고 싶어한다. (b)* 민수가 사과를 싶어한다. (c)* 민수가 사과를 먹고 원하다. (d) 민수가 사과를 원하다.

For convenience we encode c-selectional restrictions with categorial features (Chomsky, 1995; Adger, 2010). Here are some examples. The first feature [V] indicates that the lexical item is a verb. The italicized u indicates that the feature is uninterpretable. For now, let’s think of this as a requirement for some features.

(2.24) (a) think [V, {uPabout/uClause}] (b) know [V, {uN/uClause}] (c) wonder [V, uClause] (d) 싶다 [V, uClause] (e) 원하다 [V, uN]

Let’s consider the verb wonder. It has a [uClause] feature. This is a notational shorthand to mean that it requires a clause as a complement. The verb 원하다 (‘want’) requires an NP as a complement.

2.2.3 Agree

Agreement is the co-variation of an element in a sentence with another phrase. In many languages, the verb agrees with the subject. Consider the following Spanish (spa) data.

(2.25) Spanish subject agreement

(a) yo com-o I eat-1sg.prs ‘I eat.’ (b) te com-es you.sg eat-2sg.prs ‘You eat.’ 2.2. FEATURES 29

(c) María com-e Maria eat-3sg.prs ‘Maria eats.’ (d) nosotros com-emos we eat-1pl.prs ‘We eat.’ (e) vosotros com-éis you.pl eat-2pl.prs ‘You all eat.’ (f) María y Juan com-en Maria and Juan eat-3pl.prs ‘Maria and Juan eat.’

The verb agrees in person and number with the subject. Patterns of agreement vary considerably among the world’s languages, and much current research is involved in determining what patterns of agreement are and are not found in natural language. Anticipating the discussion below on clausal structure, let’s assume that T is the locus of subject agreement. We encode the fact that T seeks an argument to agree with as [uφ:]. The u refers to an unvalued feature. In this case, T is looking for φ features (person, number, and gender). We say that the unvalued feature on T is the probe and the XP that contains the feature that values the probe is the goal.

(2.26) TP

KP T0

yoφ:1sg T VP

[uφ :] como

In this configuration, T enters into an agreement relation with the subject KP and is valued as first person singular. Swahili has both subject and object agreement. Consider the following example (Deen, 2005).

(2.27) Swahili agreement (Deen, 2005, p.33f)

(a) Juma a-na-m-pend-a Mariam Juma 3sg.sbj-prs-3sg.obj-like-indic Mariam ‘Juma likes Mariam.’ (b) ni-na-m-pend-a Mariam 1sg.sbj-prs-3sg.obj-like-indic Mariam ‘I like Mariam.’ 30 CHAPTER 2. THE LEXICON, FEATURES, AND PHRASE STRUCTURE

Consider, now, a more complicated system. Nishnaabemwin exhibits a complex agreement sys- tem that both the subject and the object. Consider the following data, paying close attention to the agreement prefix. Note that the morphemes labelled inv and dflt do not play a role here, but see Beck and Johnson (2004) for more details and a full analysis.

(2.28) Nishnaabemwin agreement (Béjar and Rezac, 2009)

(a) g-waabm-in 2-see-inv ‘I see you.’ (b) g-waabm-i 2-see-dflt ‘You see me.’ (c) n-waabm-ig 1-see-inv ‘He sees me.’ (d) g-waabm-ig 2-see-inv ‘He sees you.’

If one argument is third person and the other argument is either first or second person, then the verb agrees with the non-third person argument. If one argument is first person and the other argument is second person, then the verb agrees with second person. Note that these facts hold regardless of which argument is the subject and which is the object. The solution to this puzzling agreement pattern lies in feature geometries introduced above. Specifically, the probe is not a bare unvalued φ feature, but is a feature geometry Béjar (2003); Béjar and Rezac (2009); Preminger (2014). For Nishnaabemwin they propose that the probe for agreement is on v and consists of a small feature geometry.

(2.29) uπ

uParticipant

uAddressee

The probe search down for a corresponding feature. If it fails to find one, it can search in SpecvP–a mechanism they call cyclic agree. Consider the following hypothetical tree, with an external argument (EA) and an internal argument (IA). 2.3. BARE PHRASE STRUCTURE 31

(2.30) vP

KP v 0

EA v VP

uπ V KP

uParticipant IA

uAddressee

First the probe on v searches down to match its feature geometry. Specifically, it is looking for a second person feature. In both (2.28 a) and (2.28 d) it finds a second person, values the probe, and second person agreement appears on the verb. If it fails to find a second person argument inside the VP, it continues to search in SpecvP. If it finds one as in (2.28 b), then again second person agreement is found on the verb. Finally, if it fails to find second person at all, then it matches the best it can. This is the case in (2.28 c). Let’s consider these four cases in chart form. For convenience the hierarchy φ > Participant > Addressee is re-written as 3 > 1 > 2 to make the chart easier to read. The features of the probe that value the goal are in boldface. The unvalued feature on the target where agree fails is crossed out.

(2.31) Results of probing in Nishnaabemwin agreement

2.28 a 2.28 b 2.28 c 2.28 d EA v IA EA v IA EA v IA EA v IA [3] [u3] [3] [3] [u3] [3] [3] [u3] [3] [3] [u3] [3] [1] [u1] [1] [1] [u1] [1] [u1] [1] [u1] [1] [u2] [2] [2] [u2] [u2] [u2] [2]

2.3 Bare Phrase Structure

In order to build phrases and sentences into a coherent structure, a theory of phrase structure is needed. Bare Phrase Structure (BPS) is a minimalist theory of phrase structure, which grew out of X-Bar Theory (Jackendoff, 1977; Stowell, 1981). BPS consists of a single operation, Merge, as follows (Chomsky, 1994).

(2.32) Merge (a, b) → {a,{a,b}}

Bare Phrase Structure employs many of the same concepts as X-Bar Theory, so familiarity with the discussion above is useful. In Bare Phrase Structure, phrases are built up by an operation called 32 CHAPTER 2. THE LEXICON, FEATURES, AND PHRASE STRUCTURE

Merge, which takes two objects and merges them together into a single object. In the following example, X and YP are merged together. We have to give the output of Merge a label. That is, once we merge X and YP together, we have to determine what kind of phrase this is. We will get into these details shortly, but for now, let’s call the label XP. What this means is that the XP has the fundamental properties of X. So, a noun phrase, NP, has the fundamental properties of a noun, N, for example. Determining the label has given rise to much discussion since BPS was born. Labelling Theory, as it is known, deals with situations such as the following. How exactly do we determine the label of the root note in example (2.33)?

(2.33) ?

X YP

...Y...

The following algorithm has been proposed to determine the label after merge (Chomsky, 2013).

(2.34) (a) Merge (H, XP) → H is the label (b) Merge (YP, XP) → • If YP and XP share a common criterial feature, that feature is the label. • If YP and XP share no such feature, there is no label (c) Merge (H1, H2) → One of H1 or H2 is a root, the other is the label.

2.4 Clausal Structure

Let’s consider the basic structure of a sentence within Extended Standard Theory (Chomsky, 1965).

(2.35) S’

COMP S

NP AUX VP

subject V NP

object 2.4. CLAUSAL STRUCTURE 33

The nodes S and S’ are not compliant with Bare Phrase Structure. This in and of itself is not a reason to reject the nodes S and S’. It is possible that Bare Phrase Structure is not on the right track. Let us consider the structure in (2.35) in more detail, though. ? argued on the basis of formal parsimony that S should follow the X-Bar schema. That is, to keep the grammar simple, all projections should follow the schema in (1.12). Recall also the PSRs in (1.11). Travis (1984) also considered different kinds of word order in light of the structure in (2.35). S is composed of NP and VP in this structure. Aux is an optional element for overt auxiliaries. Consider now the following data.

(2.36) (a) John ate an apple. (b) 영희가 사과를 먹었다 Yenghuy-ka sakwa-lul mek-ess-ta Younghui-nom apple-acc eat-pst-decl ‘Younghui ate an apple’

In order to account for these two examples, the subject NP and the VP must be adjacent, as shown in (2.35). Furthermore, the order of V and the object NP must be set for SVO languages such as English and for SOV languages such as Korean. This is accomplished nicely by the structure in (2.35). Consider, however, the following data.1

(2.37) Ich bin ins Kino gegangen. German, (Travis, 1984, ex.126) I am in the cinema gone ‘I went to the cinema’

(2.38) ...daß ich ins Kino gegangen bin. German, (Travis, 1984, ex.128) ...that I in the cinema gone am ‘...that I went to the cinema.’

(2.39) Y mae Sion yn mynd adref. Welsh, (Travis, 1984, ex.127)) prt is John prt go home ‘John is going home.’

(2.40) Ich gehe ins Kino. I go to the cinema ‘I’m going to the cinema.’

1Travis takes the Welsh example from Sproat (1983) 34 CHAPTER 2. THE LEXICON, FEATURES, AND PHRASE STRUCTURE

In the first German example and the Welsh example the component of the verb are split in two discontinuous units. Also, the second German example suggests an underlying SOV order. This, however, would not allow us to account for the word order in German in matrix clauses as in the fourth example above. Finally, VOS order, as in Malagasy, cannot be accounted for at all by the structure in (2.35). Based on facts like these, Travis argued that S must follow the X-Bar schema as follows.

(2.41) IP

NP I0

subject I VP

V NP

object

Specifically, Travis argued that the head position, INFL, provides a way for separating the verbal inflection from the , and for providing alternative word orders by head movement. Next, S is the location of COMP. Again, from an X-Bar theoretic point of view, this structure should be re-written as follows.

(2.42) CP

XP C0

specifier C IP

Stowell argues that COMP must be the head of S based on selectional restrictions. Certain verbs require an interrogative clause as a complement and certain verbs require an indicative clause as a complement. Consider the following examples.

(2.43) I think that you ate an apple.

(2.44) *I think what you ate.

(2.45) I wonder what you ate.

(2.46) *I wonder that you ate an apple.

Illocutionary force is encoded on COMP, so in order for the verb to select the complement clause of the correct type, COMP must be the head of the category the verb selects. This is accomplished in the model in (2.42). 2.5. CONFIGURATIONALITY 35

2.5 Configurationality

Configurationality refers to the tendency for language to impose rigid word order on the con- situents in the sentence based on grammatical function. Such languages can be described as SOV, SVO, VOS, etc. Configurational languages are associated with a distinct VP based on subject-object asymmetries, such as idioms and VP proforms, among others.

(2.47) (a) kick the bucket, chew the fat, etc. (b) Mary read the book and so did John.

Here, we consider the differences between configurational and so-called non-configurational lan- guages. Hale (1983) identified three properties of non-configurational languages.

1 free word order

2 extensive -drop

3 discontinuous constituency

Since Hale’s first proposal, our understanding of free word order has deepened, and it is now un- derstood that word order depends on properties such as topic and . Non-configurationality is now better known as discourse configurationality. Let’s consider Hale’s original proposal. He noted that Warlpiri (WBP), a Pama-Nyugen language spoken in Australia, has free word order.

(2.48) Warlpiri Free Word Order (Hale, 1983)

(a) Ngarrka-ngku ka wawirri panti-rni man-erg aux kangaroo spear-non.pst ‘The man is spearing the kangaroo.’ (b) Wawirri ka panti-rni ngarrka-ngku kangaroo aux spear-non.pst man-erg ‘The man is spearing the kangaroo.’ (c) panti-rni ka ngarrka-ngku wawirri spear-non.pst aux man-erg kangaroo ‘The man is spearing the kangaroo.’

Extensive pro-drop refers to the observation that arguments may be freely dropped as long as they are recoverable from the discourse. The sentences in (2.48 can be uttered with just the verb as long as the subject and the object are recoverable from the discourse. Consider the following examples.

(2.49) Warlpiri Pro-Drop (Hale, 1983) 36 CHAPTER 2. THE LEXICON, FEATURES, AND PHRASE STRUCTURE

(a) Ngarrka-ngku ka panti-rni man-erg aux spear-non.pst ‘The man is spearing him/her/it.’ (b) Wawirri ka panti-rni kangaroo aux spear-non.pst ‘He/she is spearing the kangaroo’ (c) Panti-rni ka spear-non.pst aux ‘He/she is spearing him/her/it.’

Finally, discontinuous constituency is illustrated in the following examples. In the first sentence, the object is a constituent formed by the head noun and the . In the second sentence the head noun and the demonstrative are split between two positions.

(2.50) Warlpiri Discontinuous Constituency (Hale, 1983)

(a) [Warirri yalumpu] kapi-rni panti-rni [kangaroo dem] aux spear-non.pst ‘I will spear that kangaroo.’ (b) [Warirri] kapi-rni panti-rni yalumpu [kangaroo] aux spear-non.past dem ‘I will spear that kangaroo.’

Hale originally proposed a flat structure for discourse configurational languages to account for the free word order and the relative lack of subject-object asymmetries. Hale also proposed that additional nominals could freely be added or missing in discourse configurational languages. Thus (2.50 b) is accounted for by assuming two NPs (one containing the noun and one containing the demonstrative) that are linked to the object . We will not get into the precise details of Hale’s proposal here.

(2.51) S

NP V NP

Jelinek (1984) proposes an amendment to Hale’s analysis. Based on the difference between sub- ject and object agreement, she proposes that the agreement markers are the actual arguments, occupying argument positions as in a configurational language. The overt nominals are adjoined clause-externally, akin to topics in configurational languages. This simplifies the grammar by re- ducing the number of possible constructions while still accounting for the properties of discourse configurational languages. Let’s consider a Warlpiri example (Jelinek, 1984, ex.12). 2.5. CONFIGURATIONALITY 37

(2.52) ngajulu-rlu ka-rna-ngku nyuntu-Ø nya-nyi [Warlpiri] 1-erg aux-1sg.nom-2sg.acc you-abs see-non.pst ‘I see you.’

Under the Pronominal Argument Hypothesis, this sentence has the following structure. Both Hale and Jelinek assumed that the location of the second position clitic (the auxiliary) is handled by phonology. For convenience, we have assumed the VP-Internal Subject Hypothesis for the subject agreement marker, see page 39.

(2.53) S’

NP1 S’

ngajulu-rlu NP2 S

nyuntu-Ø Aux VP

1 2 0 ka-rna i -ngku j ti V

V tj

nya-nyi 38 CHAPTER 2. THE LEXICON, FEATURES, AND PHRASE STRUCTURE

Further Reading

• Baker (1996) - a fairly recent treatment of discourse configurationality and polysynthesis

• Kiss (1995) - a monograph dealing with discourse configurationality in language

• Ritter and Wiltschko (2014) - This paper gives an extensive cross-linguistic discussion of the category INFL. Crucially it addresses languages that do not have tense as in well-known European languages.

• Preminger (2014) - This monograph present a detailed theory of agreement. It proposes that agreement can fail without the derivation crashing. Chapter 3

Ditransitives and the vP Layer

In this section we will examine what has become known as the vP layer in more detail.

3.1 VP-Internal Subject Hypothesis

Williams (1980) distinguishes between external arguments (외재논항) and internal arguments (내재논항). These two notions are related to the structural differences between subjects and objects. Consider the following old-fashioned structure of a basic .

(3.1) S

NP VP

Mary V NP

read the book

These terms indicate how the argument is structurally related to the verb. The external argument is external to the VP, and the internal argument is internal to the VP. Note, however, that the subject is not always an external argument. If the subject is derived, as in a passive, then the subject is an internal argument. We will discuss passives more in Chapter 5. Consider the following more updated structure.

39 40 CHAPTER 3. DITRANSITIVES AND THE VP LAYER

(3.2) TP

0 NPi T

The book T VP

was V ti

read

Although the book is the subject of the sentence, it originated as the thematic object of the verb. Thus, the terms internal argument and external argument are related to the of the argument to the verb. The grammatical subject, regardless of whether the sentence is active or passive, appears in SpecTP and receives nominative Case. Consider the following data.

(3.3) (a) All the children have eaten lunch. (b) The children have all eaten lunch.

We can assume a structure for the first sentence approximately as follows.

(3.4) TP

QP T0

Q NP T VP

all the children have V NP

eaten lunch

Example (3.3 b) challenges this structure. In this example ‘all’ is a floated quantifier that appears displaced from the NP it modifies (?). Before we consider how to deal with (3.3 b) let’s consider some additional data. Since Ross (1967) it has been known that extraction from one member of a conjunct results in ungrammaticality. Example (3.5 a) violates the Coordinate Structure Constraint. Extraction from both members of a conjunct, however, is fine, (3.5 b). This kind of movement is known as across-the-board movement, or ATB movement for short.

(3.5) (a) *What did Sally eat and Mary buy an orange? (b) What did Sally eat and Mary buy?

These are illustrated below. 3.1. VP-INTERNAL SUBJECT HYPOTHESIS 41

(3.6) (a) *CP

0 NPi C

what C TP

did TP and TP

Sally eat ti Mary buy an orange

(b) CP

0 NPi C

what C TP

did TP and TP

Sally eat ti Mary buy ti

Observe that in (3.6 b) what appears to move from both conjuncts to the left edge of the clause. No, VPs can be coordinated, too, and that’s where the picture gets tricky. First, we look at some straightforward cases.

(3.7) (a) The children ate lunch and played outside. (b) The candies were bought at the store and given to the children.

In the active sentence there is no movement – only two coordinated VPs, (3.8 a). In the second sentence there are two coordinated VPs in a passive sentence. As the tree in (3.8 b) shows ATB movement of the subject.

(3.8) (a) TP

NP T0

the children T VP

VP and VP

ate lunch played outside 42 CHAPTER 3. DITRANSITIVES AND THE VP LAYER

(b) TP

0 NPi T

the candies T VP

were VP and VP

bought ti at the store given ti to the children

Burton and Grimshaw (1992) and McNally (1992) both observed that one conjunct can be active and one can be passive.

(3.9) Sam will win the race and be awarded a prize.

The structure for this under our current assumptions would look as follows.

(3.10) TP

0 NPi T

Sam T VP

will VP and VP

win the race be awarded ti a prize

The problem is that (3.10 appears to violate the Coordinate Structure Constraint. The solution to this problem and the problem with the floated quantifier is to generate the subject inside the VP and raise it to SpecTP. This is called the VP Internal Subject Hypothesis. Here is the proposed derivation for Sam ate an apple.

(3.11) TP

0 NPi T

Sam T VP

0 ti V

V NP

ate an apple 3.1. VP-INTERNAL SUBJECT HYPOTHESIS 43

Example (3.9), then, appears as follows. Observe that there is no Coordinate Structure Con- straint violation as the subject has raised to SpecTP by ATB movement.

(3.12) TP

0 NPi T

Sam T VP

will VP and VP

t i win the race be awarded ti a prize

The floated quantifier can now be derived as follows. When the subject raises from VP to SpecTP, the entire QP can raise, or just the NP can raise, stranding the quantifier. Both of these are shown below.

(3.13) (a) TP

0 QPi T

Q NP T VP

0 all the children have ti V

V NP

eaten lunch

(b) TP

0 NPi T

the children T VP

have QP V0

Q ti V NP

all eaten lunch

One final advantage of the VP-Internal Subject Hypothesis is that we now have a way to ensure the external argument receives a theta-role. Consider the following examples. 44 CHAPTER 3. DITRANSITIVES AND THE VP LAYER

(3.14) (a) Mary ate an apple. (b) Mary fears monsters.

The verb eat assigns an theta-role to its subject, while the verb fear assigns an theta-role to its subject. It is assumed that theta-roles are assigned by lexical heads. If we assume the VP-Internal Subject Hypothesis, then the verb can assign a theta-role to the external argument before it raises to SpecTP. In the next section, however, we will see that it is necessary to revisit the notion of theta-roles and external arguments.

3.2 Severing the External Argument

Consider, now, the following examples. While the internal arguments have a wide variety of thematic relations, the external argument is either an or an .

(3.15) (a) Sam ate an apple. (b) Sam moved the table. (c) Sam walked from home to work. (d) Sam attended that university. (e) Sam likes apples. (f) Sam laughed.

Often, the external argument is not necessary to create a well-formed sentence.

(3.16) (a) Susan boiled the water. (b) The water boiled.

(3.17) (a) Fred broke the window. (b) The window broke.

These kinds of alternations are pervasive around the world. Often; however, in some languages there is a morphological change on the verb when this alternation happens. Here’s an example from Korean.

(3.18) (a) 민수가 문을 열었다. (b) 문이 열렸다.

Marantz (1984) offers the following data showing that the verb and an internal argument can give rise to an idiomatic meaning; however, a verb and an external argument rarely, if ever, give rise to idiomatic meanings. 3.2. SEVERING THE EXTERNAL ARGUMENT 45

(3.19) (a) throw a baseball (b) throw support behind a candidate (c) throw a boxing match (i.e., lose on purpose (paraphrase mine)) (d) throw a party (e) throw a fit

These facts lead us to propose that the external argument is thematically "severed" from the verb. That is, the internal arguments receive a theta-role from the verb, but the external argument receives a theta-role receives a theta-role from somewhere else. Kratzer (1996), picks up on an idea proposed in Larson (1988), discussed below, using Malagasy data such as the following (Hung, 1988). She notes Hung’s observation that distinct morphology appears whenever there is an external argument.

(3.20) (a) m-an-sasa ny lamba Rasoa prs-act-wash the clothes Rasoa ‘Rasoa washes the clothes.’ (b) m-an-ala prs-act-take.out ‘takes (something) out’

The Halkomelem data in Table (3.1) show the same property (Galloway, 1993).

Halkomelem English Halkomelem English q’óy ‘die’ q’ót-t ‘kill’ íkw’ ‘be lost’ íkw’-et ‘throw away/discard’ qw’és ‘fall into water qw’s-et ‘push into water’

Table 3.1: Halkomelem Alternations

Kratzer proposed a new functional head, now called v ("little v") that introduces the external argument and assigns a theta-role to it. Let’s see some sentence from before using v to introduce the external argument.

(3.21) (a) Sam won the race. (b) The children have all eaten lunch. 46 CHAPTER 3. DITRANSITIVES AND THE VP LAYER

(3.22) (a) TP

0 NPi T

Sam T vP

0 ti v

v VP

V NP

won the race

(b) TP

0 NPi T

the children T vP

have QP v 0

Q ti v VP

all V NP

eaten lunch

We will come back to this idea when we discuss unaccusativity below; however, we can see that the alternations in (3.16 – (3.18) can be handled by assuming that v is either present or absent. Thus we have the following partial structures for John boiled the water and The water boiled.

(3.23) (a) vP

NP v 0

John v VP

V NP

boil the water 3.3. DITRANSITIVES 47

(b) VP

V NP

boil the water

3.3 Ditransitives

Ditransitive verbs such as give were challenging for syntactic theory as there are two internal argu- ments rather than just one. In Revised Extended Standard Theory until Government and Binding Theory it was simply assumed that the two objects were both sisters to the verb.

(3.24) (a) give Mary a book (b) 영희에게 책을 주다

(3.25) (a) VP

V NP NP

give Mary a book

(b) VP

PP NP V

NP P 책을 주다

영희 에게

Barss and Lasnik (1986) note various asymmetries between the first NP and the second NP in English. Namely, the first NP appears to c-command the second NP, but not vice versa. This observation is not captured in the tree in (3.25 a) as both NPs c-command each other.

(3.26) (a) I showed John himself (in the mirror). (b) *I showed himself John (in the mirror).

Our current understanding of the structure of ditransitives started with Larson (1988, 1990). In addition to the data on reflexives above, Larson also explored bound variable pronouns. A variable is any element whose value changes with its context. A variable must be bound by some element that defines its value. We will discuss the notion of binding in much more detail later, but for now we will just use a simple definition. We say that the quantifier must bind the pronoun in order to get the reading that matches up each student with his or her missing book. In order for a quantifier to bind a pronoun, the quantifier must c-command it. Consider the following examples. 48 CHAPTER 3. DITRANSITIVES AND THE VP LAYER

(3.27) Every student forgot their book.

In (3.27), the interpretation of their varies with every student, which contains the quantifier every. A more formal representation of this relationship is as follows.

(3.28) ∀x, x a student, x forgot x’s book. (∀ = ‘for all’)

In this representation, the relationship between the quantifier and the variable is spelled out more clearly. Consider, now, the following data.1

(3.29) (a) I denied each worker their paycheque. (b) *I denied its recipient each paycheque. (c) I showed every trainer their lion. (d) *I showed its trainer every lion.

Observe that one of the objects is a quantified phrase (each X ) and the other object contains a variable. Again, the quantifier in the first object successfully binds the variable in the second object, but not vice versa. This observation can be explained if we assume the first object c-commands the second object, but the second object does not c-command the first object. In other words, the first object asymmetrically c-commands the second object. The next set of data we will look at makes use of negative polarity items (NPIs), such as anyone, in the upcoming example. NPIs have the interesting property that they must be c- commanded by or a yes/no question operator.

(3.30) (a) I didn’t see anyone. (b) Did you see anyone? (c) *I saw anyone. (≠ I saw no one/someone)

When we apply these facts to ditransitive constructions we observe the following:

(3.31) (a) I gave no one anything. (b) I sent no presents to any of the children. (c) *I gave anyone nothing. (d) *I sent any of the presents to none of the children.

1Larson’s original data set did not use singular they. However, many speakers use only they as a bound variable, reserving he and she for definite descriptions. 3.3. DITRANSITIVES 49

Observe that in the first two examples the NPIs anything and any are acceptable as shown; however, in the last two examples they are not. Again, this leads us to conclude that the first object asymmetrically c-commands the second object. From these three sets of data we see that the first object must c-command the second object, but not vice versa. The ternary structure fails to predict this asymmetry. What we need is a structure in which the first object c-commands the second object, but not vice-versa. Larson (1988) proposed a double VP projection, the higher of which is now known as vP. The two objects appear in the VP as follows.

(3.32) VP

NP V0

a book V PP

give to Mary

When v merges with the VP the verb raises to v.

(3.33) vP

Subject v 0

v VP

0 Vi v NP V

give a book ti PP

to Mary

Many verbs in English (especially Germanic verbs) can appear in the double object construc- tion, illustrated below.

(3.34) (a) Mary gave a book to John. - ditransitive construction (b) Mary gave John a book. - double object construction

The analysis of ditransitives and double object constructions has been a matter of considerable interest (Harley and Jung, 2015; Harley, 2002; Hallman, 2015; Bruening, 2010; Basilico, 2008; Beck and Johnson, 2004; Kayne, 1984). Beck and Johnson (2004) used the interaction between the again and ditransitives to elucidate how the structures in (3.34) are related. One crucial question is 50 CHAPTER 3. DITRANSITIVES AND THE VP LAYER whether these two sentences have the same underlying structure or not. Crucially, their argument relies on the of the adverb again.

(3.35) Sally opened the door again.

(a) Sally opened the door, and that had happened before. (repetitive) (b) Sally opened the door, and the door had been open before (restitutive)

To be clear, under the repetitive reading Sally opens the door. Somehow it becomes closed, and then Sally opens it again. Under the restitutive reading, the door is in an open state. Somehow it becomes closed, and then Sally returns it to an open state. They adopt Von Stechow (1996) and assume a uniform of again. The different readings are due to again attaching to different locations in the tree. This requires a more articulated structure. In the following structure, the lexical item open refers only to the state. It combines with an abstract verb meaning "become". The different adjunction sites give rise to the different meanings as shown.

(3.36) vP

vP againrepetitive

Thilo v 0

v VP

V* VP

BECOME AdjP

AdjP againrestitutive

Adj NP

open the door

Now we consider these facts in light of ditransitive verbs. The following sentence is ambiguous as shown.

(3.37) Thilo gave Satoshi the map again.

(a) Thilo gave Satoshi the map, and that had happened before. (repetitive) (b) Thilo gave Satoshi the map, and Satoshi had had the map before. (restitutive) 3.3. DITRANSITIVES 51

However, when we consider benefactive constructions, the meanings start to diverge. Compare the following two examples.

(3.38) Thilo sewed a flag again for Satoshi.

(a) Thilo sewed a flag for Satoshi, and that had happened before. (repetitive) (b) Thilo sewed a flag for Satoshi, and there had been a flag before. (restitutive)

(3.39) Thilo sewed Satoshi a flag again.

(a) Thilo had sewed Satoshi a flag, and that had happened before. (repetitive) (b) Thilo sewed Satoshi a flag, and Satoshi had had one before. (restitutive)

Specifically, in (3.39), it must be the case that Satoshi comes to have of the flag; however, in (3.38) such a meaning is not obligatory. Based on this difference, two different structures are proposed for the give. Benefactive constructions are considered later. Following Beck and Johnson (2004) and Harley (2002) the structures for (3.34 a) and (3.34 b), respectively, are as follows. Beck and Johnson argue that the difference is meaning is detectable in the sew examples, but not in the give examples.

(3.40) (a) vP

vP againrepetitive

Thilo v 0

v PP

CAUSE PP againrestitutive

NP P0

a book Ploc PP

to Mary 52 CHAPTER 3. DITRANSITIVES AND THE VP LAYER

(3.41) (a) vP

vP againrepetitive

Thilo v 0

v PP

CAUSE PP againrestitutive

PP P0

to Mary Phave NP

a book

We will use these more complex structured for the discussion only if they are needed. For general discussions, we will represent give as V.

3.4 vP Refined

The traditional analysis of passives involves deletion of the vP projection. Although we discuss passives in more detail below, we note that v encodes accusative Case and the external theta-role. v is also considered to be the locus of verbalizing morphology.

(3.42) verbalizing morphology in English

(a) clarify, liquify, glorify, etc. (b) pluralize, accessorize, brutalize, etc. (c) formulate, abbreviate, stipulate, etc.

Harley (2007) notes, however, that the verbalizing morphology remains in passives, making this straightforward analysis problematic.

(3.43) (a) The butter was clarified. (b) The noun was pluralized. (c) The conditions were clearly stipulated in the contract.

Harley (2013) adduces evidence from Hiaki (an Uto-Aztecan language spoken in Mexico and the US) to argue for a further split as follows. 3.4. VP REFINED 53

(3.44) VoiceP > vP > VP (or √P)

First she notes the existence of verbalizing morphology in Hiaki, akin to the English facts in (3.42). A simple example follows.

(3.45) Hiaki verbalizing morphology

(a) siki ‘red’; sikisi ‘to redden’ (b) heewi ‘yes!’; heewite ‘to agree’ (c) kari ‘house’; karite ‘to build a house’

(3.46) Santos kari-te [Hiaki (Harley, 2013, ex.15)] Santos house-vlzr ‘Santos builds a house.’

The straightforwardly translates to the following structure.

(3.47) vP

NP v 0

Santos √ v

kari -te

Hiaki also has productive applicatives. Harley gives the following example.

(3.48) Inepo Jose-ta pueta-ta eta-ria-k [Hiaki (Harley, 2013, ex.12b)] I Jose-acc door-acc open-appl-prfv ‘I opened the door for Jose.’

Harley provides extensive evidence that the applied object (Jose) is c-commanded by the external argument. This translates to the following partial structure. 54 CHAPTER 3. DITRANSITIVES AND THE VP LAYER

(3.49) vP

NP v 0

Inepo ApplP v

NP Appl0

Jose-ta √P Appl

NP √ -ria

pueta-ta ela

A problem arises when we consider applicatives with the verbalizing morphology above. Consider the following example.

(3.50) Santos Maria-ta kari-te-ria [Hiaki (Harley, 2013, ex.17)] Santos Maria-acc house-vlzr-appl ‘Santos is building a house for Maria.’

According to the structure we’ve built so far, we would expect the order root-appl-vlzr. Harley shows that if VoiceP, rather than vP is responsible for introducing the external argument the facts above fall into place.

(3.51) VoiceP

NP Voice0

Santos ApplP Voice

NP Appl0

Maria-ta vP Appl

√ v -ria

kari -te 3.4. VP REFINED 55

Further Reading

• Pylkkänen (2008)

• Legate (2014)

• Alexiadou et al. (2015) - These three works provide extensive discussions and analyses of VoiceP, vP, and applicatives. 56 CHAPTER 3. DITRANSITIVES AND THE VP LAYER Chapter 4

Nominal Syntax

In this chapter, we will refine many of the concepts of phrase structure that we learned in the previous chapter. The goal of this chapter is to expand our theory of phrase structure to be able to account for a broader range of data. In the first few sections of the chapter, we will re-evaluate the core structure of the clauses and see that additional functional categories are required to account for the properties of clauses we look at. In the second half of the chapter, much of the discussion will centre around ditransitive verbs, which are difficult to account for given the state of phrase structure that we left off with in the previous chapter. In this chapter we will see how to use various diagnostics to argue for a particular structure.

4.1 DP Theory

Prior to X-Bar Theory and Government and Binding theory the structure of a noun phrase was represented as an NP, the specifier of which contains the determiner. The structure of a noun phrase such as the cow was represented as in (4.1).

(4.1) NP

Det N

the cow

The determiner can be a single word such as the, a, this, that etc. Since the determiner is a single word, it looks more like a head than a phrasal projection. Building on the parallel between deter- miners and verbal inflection, Abney (1987) proposed the following structure for nominal projections. Here, the D head takes an N as a complement.

57 58 CHAPTER 4. NOMINAL SYNTAX

(4.2) DP

D N

the cow

Let’s review several lines of evidence in favour of this structure. Consider the following paradigm. Observe that the morphemes the and ’s (the marker) are in complementary distribution. That is, only one can be present in the same DP. Let’s assume, for the moment, that both the and ’s are D heads. If there is only one D head per DP, then only one of these lexical items can be present in a given DP.

(4.3) (a) the hat (b) John’s hat (c) *John’s the hat

Before we continue, let us take a brief digression on . We note that the possessor can be quite complex. In the following example, the possessor is the man I saw yesterday. This suggests that the possive marker ’s is not an affix on the noun but is rather a separate syntactic head that appears to the right of the possessor DP. Let us assume that the possessor appears in the specifier of DP.

(4.4) the man I saw yesterday’s hat

Compare the trees for John’s hat and the man I saw yesterday’s hat

(4.5) DP DP

DP D’ DP D’

John D N The man I saw yesterday D N

’s hat ’s hat

Furthermore, note that both D heads require a complement. To wit, the by itself is not an acceptable utterance. Below we will see that John’s is acceptable, but only if the complement is understood from the discourse. In an out-of-the-blue context it is not acceptable. We assume, then, the following lexical entries.

the D, c-selects __ NP ’s D, c-selects __NP, __ DP 4.2. NOMINALIZERS AND ROOTS 59

Some arguments for DP are rooted in the parallelism between nominals and clauses. One such parallelism is subject agreement. In many languages, subjects trigger agreement with a verb or an auxiliary. Consider the following Hungarian data. Abney argues that the parallelism between the possessor agreement and the subject agreement is evidence that the possessor functions as the "subject" of the noun phrase. Just as clausal subjects appear in SpecTP and trigger agreement with the T head, nominal possessors also appear in SpecDP and trigger agreement on the D head.

(4.6) Hungarian (Bernstein, 2008)

(a) az én-ø vendég-e-m the I-nom guest-poss-1sg ‘my guest’ (b) a te-ø vendég-e-d the you-nom guest-poss-2sg ‘your guest’ (c) Mari-ø vendég-e-ø (the) Mary-nom guest-poss-3sg ‘Mary’s guest’ (d) Mari-ø alud-t-ø Mary-nom sleep-pst-3sg ‘Mary slept.’

Longobardi (1996) accounts for the following observation in Italian by assuming head movement from N to D. Consider the following data.

(4.7) N to D raising in Italian

(a) L’antica Roma era una città potente. the ancient Rome was a city powerful ‘Ancient Rome was a powerful city.’ (b) Roma antica era una città potente Rome ancient was a city powerful ‘Ancient Rome was a powerful city.’ (c) *Antica Roma era una città potente. ancient Rome was a city powerful ‘(Ancient Rome was a power city.)’

4.2 Nominalizers and Roots

Lexical words are quite different from grammatical words and morphemes. The knowledge of whether to use a nominative case marker or an accusative case marker depends on morphosyntactic knowl- edge. The choice, in Korean, of whether to use -이 or -가 for nominative case depends on phonological 60 CHAPTER 4. NOMINAL SYNTAX knowledge. The choice of whether to use the or a in English depends on syntactic and semantic knowledge. These choices all demonstrate the kind of subconscious linguistic knowledge speakers possess. Lexical words are quite different. The choice of whether to use mussel or oyster does not depend on any kind of linguistic knowledge. It depends on the speaker’s knowledge of shellfish. The choice of whether to use 이불 or 요 depends of the speaker’s knowledge of Korean sleeping customs. Lexical roots, then are not quite linguistic in the same way grammatical morphemes are. Although proposals differ from one researcher to the next, it seems that roots are merely sound meaning correspondences devoid of grammatical information. Consider also the fact that many lexical roots are free to appear as different lexical categories. Consider the following examples.

(4.8) (a) My favourite colour is red. (b) The red car is in the driveway. (c) You can redden the icing with this food colouring.

Here, red is being used as a noun, adjective, and verb, respectively. To be used as a verb, it required the affix -en. Many roots in Korean can be used either as nouns or as verbs, with the appropriate morphology.

(4.9) (a) 깊이 깊다 (b) 놀이 놀다

Next are some Lushootseed (Salish) examples illustrating the same point. Observe that the lexical items for go and coyote can be used either as nouns or as verbs. Based on similar evidence from many other langauges around the world we conclude that roots are not specified for lexical category.

(4.10) Lushootseed (Eijk and Hess, 1986, p.324)

(a) P´uxw ti sbiàw ˙ go det coyote ‘The coyote goes.’ (b) sbiàw ti P´uxw ˙ coyote det go ‘The one who goes is a coyote.’

Consider, next, the traditional distinction between inflectional and derivational morphology. While inflectional morphology is fully productive and paradigmatic, derivational morphology is not. Any verb can take past tense morphology, for example, but the verbalizer -ize, for example, attaches only to some bases (modernize, verbalize,*coolize (make cool)). 4.2. NOMINALIZERS AND ROOTS 61

Marantz (1997, 2001) objects to this claim and argues that in many cases derivational mor- phology is productive. In English, for example, -ation productively attaches to -ize (modernization, verbalization, etc.). In Korean -(으)로 productively attaches to 적 (일반적으로, 무심적으로, etc.). Of course, some derivational morphology is lexically restricted. In English, -th can attach only to a small number of bases (width, length, warmth, etc.). In Korean, -가락 attaches to very few bases, too (손가락, 숟가락, etc.). To capture this different behaviour in derivational morphology, and to capture the fact from above that lexical roots are typically free to appear as any lexical category, Marantz proposed that roots are category neutral and are categorized by light category heads. The heads include n for nouns, a for adjectives, and v for verbs, which we saw in the previous chapter. The following example shows the derivation for the three forms of red in example (4.8).

(4.11) Categorization of the root red nP aP vP

n √ a √ v √

red red -en red

We assume that as a suffix -en triggers head movement to get the surface form redden. Under this approach, affixes that attach directly to the root will be sensitive to lexical idiosyn- crasies, while affixes that attach outside the root will be sensitive to the outermost affix. Recall that -ation productively attaches to -ize. Consider the following example.

(4.12) nP

n vP

-ation v √

-ize modern

After head movement we get the following structure.

(4.13) nP

n vP

v j n tj ti

√i v -ation

modern -ize 62 CHAPTER 4. NOMINAL SYNTAX

Now, the derivation in (4.13) is rather complex and cumbersome. When such a detailed structure is not necessary to the discussion many linguists in practice use the representation in (4.14) for short. Note that while this example appears to violate the requirement under BPS that all branching be binary, this real structure is as in (4.13), which respects binary branching.

(4.14) NP

N

modernization

Harley (2005) suggest that n allows us to solve a problem with one-replacement. Consider the following data.

(4.15) (a) The student with long hair and that one with short hair. (understood as “that [student] with short hair”) (b) This student with long hair and that one. (understood as “that [student with long hair]”) (c) *This student with long hair and one with short hair (understood as “[this student] with short hair”.) (d) *Every student with long hair and one with short hair. (understood as “[Every student] with short hair”)

(4.16) (a) This writer of novels with long hair and that one with short hair. (understood as “That [writer of novels] with short hair”) (b) This writer of novels with long hair and that one. (understood as “That [writer of novels with long hair]”) (c) *?This writer of novels with long hair and that one of poems with short hair. (understood as “That [writer] of poems with long hair”)

The proform one can target a noun and adjoined adjectives. It must target the complement of the noun if present (4.16 c), but cannot target demonstratives or quantifiers (4.15 c–4.15 d). Jack- endoff (1977) accounted for these facts with the following structure.1 Jackendoff proposed that one-replacement targets any available N node. This will necessarily include the complement to the noun (if present) and exclude determiners.

1Harley assumed of -insertion for the complement to the noun. This issue does not affect the discussion. 4.2. NOMINALIZERS AND ROOTS 63

(4.17) NP

Det N0

the AdjP N0

smart N0 PP

N PP with long hair

student of chemistry

If we update this structure to comply with DP Theory and Bare Phrase Structure, we need merely say that one-replacement targets NP rather than N.

(4.18) DP

D NP

the AdjP NP

smart NP PP

N PP with long hair

student of chemistry

We run into a problem, however, if there is no complement. Crucially, while we derive the fact that of chemistry cannot be stranded in (4.18) we now falsely predict that from Spain cannot be stranded in (4.19).

(4.19) DP

D NP

the AdjP NP

smart NP PP

N PP with long hair

student from Spain 64 CHAPTER 4. NOMINAL SYNTAX

Harley argues that assuming an nP projection can handle this problem. We now provide the final trees for these two DPs. If we assume one-replacement targets nPs, then all the facts above fall into place.

(4.20) (a) DP

D nP

the AdjP nP

smart nP PP

n √P with long hair

√i n ti PP

stud -ent of chemistry

(b) DP

D nP

the AdjP nP

smart nP PP

nP PP with long hair

n ti from Spain

√i n

stud -ent

Finally, it has been suggested the n is the locus of nominal classification Wiltschko (2014); Kramer (2015). Many languages classify nouns according to some feature such as gender or animacy. Let’s take a clear example from Spanish.

(4.21) Classification of Spanish Nouns 4.3. NUMBER 65

root feminine form masculine form amig amiga (‘female friend’) amigo (‘male friend’) ti tia (‘aunt’) tio (‘uncle’) alumn alumna (‘female student’) alumno (‘male student’) cantor cantora (‘female singer’) cantor (‘male singer’)

From the data we can see that the feminine marker is -a and the masculine marker is either -o or is null. There are other gender markers in Spanish, but this is sufficient for the current discussion. For additional discussion on gender and how it is encoded in the nominal hierarchy the reader should consult Picallo (1991), Ritter (1992), and Kramer (2015). To recapitulate, then, the structure we have so far for a noun phrase such as the singer is as follows. The trees both before and after movement are shown.

From the data we can see that the feminine marker is -a and the masculine marker is either -o or is null. There are other gender markers in Spanish, but this is sufficient for the current discussion. For additional discussion on gender and how it is encoded in the nominal hierarchy the reader should consult Picallo (1991), Ritter (1992), and Kramer (2015). To recapitulate, then, the structure we have so far for a noun phrase such as the singer is as follows. The trees both before and after movement are shown.

(4.22) the singer DP DP

D nP D nP

the n √ the n ti

-er sing √i n

sing -er

4.3 Number

Beginning in the late 1980s the relationship between morphological structure and syntactic structure began to be questioned (Baker, 1988; Pollock, 1989). The idea that morphology as an independent module of grammar could be dispensed with began to gain widespread appeal as this would signifi- cantly simply human grammar. Of course debates on the independence of morphology still persist, and many linguists still assume the lexicalist hypothesis - the idea that syntax manipulates only fully formed words. Within this context Elizabeth Ritter challenged the idea that number is merely a morphological reflex on the noun but rather has its own syntactic representation (Ritter, 1992, 66 CHAPTER 4. NOMINAL SYNTAX

1993). She proposed that NumP appears between DP and NP (nP had not yet been proposed at this time).

(4.23) DP

D NumP

Num NP

noun

The proposal seems quite reasonable when one compares English data with, say, Niuean data. In Niuean, the number marker is a free-standing word (Massam, 2020). At the time Ritter proposed NumP, however, facts such as those in Niuean were not well known among generative linguists.

(4.24) Examples of Number

(a) the book-s, the dog-s (b) he falu a tau fuata Niue erg some lnk pl youth Niue ‘some Nieuan youths’

Ritter presents data from Hebrew possessive constructions to argue for the presence of NumP. The first construction is called the . It is characterized by having a bare subject genitive and noun-subject-object order.

(4.25) Hebrew Construct State (Ritter, 1992, ex.1)

(a) ahava-t dan et iSt-o love-gen Dan acc wife-his ‘Dan’s love of his wife’ (b) DP

D NP

Ni DGEN DP N’

ahava Dan ti DP

et iSt-o

The construct state genitive can be derived easily by head movement from N to D, as shown in the tree above. Consider, however, the free genitive. Consider the following example, with the assumed underlying structure. 4.3. NUMBER 67

(4.26) Hebrew free state genitive, (Ritter, 1992)

(a) ha-ahava Sel Dan et iSt-o the-love of Dan acc wife-his ‘Dan’s love of his wife’ (b) DP

D NP

ha DP N0

Sel Dan N DP

ahava et iSt-o

There is no space in (4.26 b) for the noun to raise to. Ritter 1992; ? proposed an additional funtional projection, a Number Phrase (NumP), that appears between DP and NP. Here are the derivations she proposed for the construct state genitive and the free state genitive, respectively.

(4.27) (a) DP

D NumP

ha Num NP

Ni Num NP N’

ahava Sel Dan ti DP

et iSt-o

(b) DP

D NumP

Numi DGEN Dank Num’

Nj Num tj NP

ahava tk N’

ti DP

et iSt-o 68 CHAPTER 4. NOMINAL SYNTAX

As was common at the time Ritter assumed the Weak Lexicalist Hypothesis, whereby inflectional morphology is syntactic and derivational morphology is added to the base in the Lexicon. Shortly earlier Pollock (1989) had argued for a split TP, in which tense and agreement morphology is added syntactically by head movement. Likewise, Ritter argued that number, which is inflectional in Hebrew, is also added by head movement. Ritter (1992) also considered data from Hiatian Creole in support of a number phrase. She notes that Hiatian Creole has a singular determiner and a plural determiner.

(4.28) Hiatian Creole determiners (Ritter, 1992, ex.18)

(a) liv la book the.Sg ‘the book’ (b) liv yo book the.Pl ‘the books’

Ritter proposes that yo is actually a Num head, as in the following tree.

(4.29) DP

NumP D

NP Num +def

liv yo

Lefebvre (1982) observes that la yo (in that order) is possible in some dialects of , suggesting head movement and right-adjunction.

(4.30) DP

NumP D

NP ti D Numi

liv la yo

4.4 Case

Case (격) indicates the grammatical relationship of a noun phrase with the sentence. The gram- matical subject in many languages appears with nominative case (주격) and the direct object appears with accusative case (대격). Here are some examples. 4.4. CASE 69

(4.31) Nominative and Accusative Case

(a) I saw him. (I - nominative; him - accusative) (b) 영희가 책을 읽었다. (Yenghuy-ka - nominative; chayk-ul - accusative) (c) mater alit pull-os mother.nom nourishes children-acc ‘Mother nourishes the young ones.’ [Latin]

The Korean example clearly shows the nominative case suffix (-ka/-가) and the accusative case suffix (-ul/-을). Not all languages have an overt suffix for both nominative and accusative. In the Latin example we see an accusative suffix (-os), but there is no overt nominative suffix. Finally, Modern English has no overt case suffixes; however, pronouns inflect for case. For example, the first person singular nominative pronoun is I, and the first person singular accusative pronoun is me. Dative case (여격) is used to indicate the indirect object. Consider the following Lithuanian example (Mathiassen, 1996).

(4.32) Mergait-˙e dav˙e vaik-ui obuol-i˛ girl-nom gave boy-dat apple-acc ‘The girl gave the boy an apple.’

Case is often marked by bound morphemes as shown in the examples above. Case can also be marked by free morphemes. This is common in . Here is an example from Tahitian (Polinsky and Potsdam). Note that there is no marker for nominative case. Accusative case (acc) is marked by the particle ‘i. (‘ represents a glottal stop.)

(4.33) Tahitian

(a) te ma’ue nei te mau manu. asp fly asp det pl bird ‘The birds are flying.’ (b) ‘ua ‘ite te tamaiti ‘i te mau manu perf see det child acc det pl bird ‘The child saw (the) birds.’

The Tahitian facts clearly show that case and determiners are treated distinctly. The structure we will argue for is the following. 70 CHAPTER 4. NOMINAL SYNTAX

(4.34) KP

K DP

D nP

n NP

noun

? noticed several similarities between and case markers. Specifically, they no- ticed that case drop in Japanese and complementizer drop in English were sensitive to the same conditions. In English, the complementizer can be dropped if the CP is adjacent to the verb. Here are their examples for complementizer drop in English (?, p.159).

(4.35) (a) John believes (that) Mary will win. (b) John believes wholeheartedly ?*(that) Mary will win.

When the verb is adjacent to the CP, the complementizer can be dropped. Likewise, in Japanese accusative Case on the object can be dropped only when it is adjacent to the verb. Here are the Japanese facts they discussed (?, p.158), citing ?.

(4.36) Japanese (?)

(a) John-ga dare-(o) nagutta no? John-nom who-(acc) hit q ‘Who did John hit?’ (b) Dare-*(o) John-ga nagutta no?

Although the theoretical machinery that Travis and LaMontagne used no longer holds, the similarity is suggestive of a parallel structure. More telling is the selectional restriction between verbs and objects. Recall that certain verbs place selectional restrictions on the kinds of clauses they take, as in the following examples. The verb believe c-selects a declarative clause only. The verb know c-selects either a declarative or an interrogative clause. The verb wonder c-selects an interrogative clause only.

(4.37) (a) John believes that Mary likes apples. (b) *John believes what Mary likes. (c) John knows that Mary likes apples. (d) John knows what Mary likes. 4.4. CASE 71

(e) *John wonders that Mary likes apples. (f) John wonders what Mary likes.

In many languages, certain verbs c-select the case of the direct object. Consider the following German and Finnish examples. In German, most verbs select an accusative object; however, the verb help selects a dative object. In Finnish, the verb like selects an elative object and the verb trust selects an illative object.

(4.38) German

(a) Ich helfe meinem Freund. I.nom help my.dat friend ‘I help my friend.’ (b) Ich sehe meinen Freund. I.nom help my.acc friend ‘I see my friend.’

(4.39) Finnish ?

(a) Ville pitää Marja-sta jopa humalaise-na. Ville likes Maria-ela even drunk-ess ‘Ville likes Maria, even when s/he is drunk.’ (b) Marja luottaa Ville-en jopa alastoma-na. Maria trusts Ville-ill even naked-ess ‘Maria trusts Ville, even when s/he is naked.’

Let’s now see how we derive a simple sentence such as John ate the apple. 72 CHAPTER 4. NOMINAL SYNTAX

(4.40) CP

C TP

0 KPi T

K DP T vP

0 D nP ti v

n tj v VP

Nj n V v tk KP

John ate K DP

D nP

an n tl

N n

apple 4.4. CASE 73

Further Reading

• ? - This deals with the structure of pronouns in great detail using data from a variety of languages.

• Kramer (2015) - This monograph deals with several aspects of gender in noun phrases.

• Bernstein (2008) - This review article provides an up to date understanding of the DP Hypothesis

• ? - This review article summarizes recent advances in the functional structure of nominals.

• ? - This chapter reviews recent advances in nominal structure in Germanic, Romance, and Balkan languages

• ? - This article gives an in depth analysis of nominal structure in Japanese, including classifiers.

• ? - This article analyzes classifiers in Chinese.

• ??? - A series of papers in which Bošković proposes the DP parameter: Some languages project DP, other languages have a flat NP. 74 CHAPTER 4. NOMINAL SYNTAX Chapter 5

The Architecture of Syntax

5.1 Phrasal Movement and Head Movement

We have seen that elements in the syntax tree can move. ? proposes that UG should be as simple as possible. One specific proposal was that Move and Merge should be considered a single operation. For convenience, Merge is sometimes referred to as external merge, highlighting the fact that a new syntactic object is added to the derivation. Recall that a phrase or a head is always merged to the root node and never anywhere inside the tree. Move, then, is sometimes referred to as internal merge. A syntactic object from lower inside the tree is merged to the root node. The terms ‘external merge’ and ‘internal merge’ are for convenience only. They are both instances of Merge. We have seen some instances of phrasal movement in previous chapters. Let’s consider an abstract example using both a syntagmatic tree diagram and set-theoretic notation.

(5.1) (a) XP

0 YPi X

X ZP

Z ti (b) {X,{X,{Z,{Z,YP}}}} - before movement (c) {X,{YP,{X,{X,{Z,{Z,YP}}}}}} - after movement

We were already introduced to head movement in Chapter 2. Let’s look at a hypothetical example. In the following example Y undergoes head movement to X. Until now we have not been specific about the mechanics of head movement. Observe in the following example that when Y raises it adjoins to X. X now forms a complex head consisting of Y and X.

75 76 CHAPTER 5. THE ARCHITECTURE OF SYNTAX

(5.2) XP

X YP

Yi X ti ZP

Consider the following data.

(5.3) (a) John will have eaten lunch.

(b) Willi John ti have eaten lunch?

(c) *Havei John will ti eaten lunch?

(5.4) German

(a) Liesel hat den Apfel essen können. Liesel has the.ACC apple eat.inf be.able.inf ‘Liesel was able to eat the apple.’ (b) Hat Liesel den Apfel essen können? has Liesel the.ACC apple eat.inf be.able.inf ‘Was Liesel able to eat the apple?’ (c) *Können Liesel hat den Apfel essen? be.able.inf Liesel has the.ACC apple eat.inf ‘(Liesel was able to eat the apple.)’

Travis also noted Baker’s analysis of noun incorporation in her analysis of head movement (Baker, 1988). The following Onondaga (Iroquoian) example illustrates the relevant point.

(5.5) (a) waPhahninúP neP oyękwaP waP-ha-hninu-P neP o-yękw-aP fact-3sg.m.ag-buy-punc det npref-tobacco-nfs ‘He bought tobacco.’ (b) waPhayękwahninúP waP-ha-yękw-a-hninu-P fact-3sg.m.ag-tobacco-epen-buy-punc ‘He bought tobacco.’

In the first example the verb takes a full DP as a complement. (We ignore KP here to keep the structure simple. Also, the nominal prefix is shown as Agr(eement)P(hrase). The precise details do not matter for this discussion.) Note that the verb 5.1. PHRASAL MOVEMENT AND HEAD MOVEMENT 77

(5.6) VP

V DP

hninu D AgrP

neP Agr NumP

o- Num nP

n ti

Ni n

yękw aP

The noun root, N, undergoes head movement to n, the immediately c-commanding head, and gives rise to the observed morpheme order. Note that in the noun incorporation example in (5.5 b) only the noun root is found. None of the other nominal morphology is present, and there is no determiner. Baker takes this as evidence that the verb takes a bare NP as a complement.1

(5.7) VP

V ti

Ni V

yękw hninu

Again, what we observe is the head movement targets the immediately c-commanding head. The fact that noun incorporation typically involves impoverished nominal structure suggests that a head movement account is involved, though see ? for an alternative view. Travis used these facts suggest to argue that head movement must target the immediately c- commanding head. This is known as the Head Movement Constraint (Travis, 1984). The mechanism of Merge investigated so far allows us to introduce phrases into the structure that have been selected. Let’s take the following example. The ditransitive verb give selects two arguments–the direct and the indirect object. The indirect object contains a preposition that selects its object, John. TP requires something in its specifier, so moves (or internally merges) the subject.

(5.8) (a) Mary gave a book to John.

1Note that I have represented the complement to the verb as N rather than as NP. This is because an NP structure would entail unary branching, which we wish to avoid. In the 1980s, there was no theoretical ban on unary branching 78 CHAPTER 5. THE ARCHITECTURE OF SYNTAX

(b) CP

C TP

0 KPi T

Mary T vP

0 ti v

v VP

0 Vj v KP V

gave a book tj PP

P KP

to John

Let’s examine the set-theoretic representation of some of the projections. The PP, VP and TP are represented as follows. The subscripts are used make the structures more readable. In reality the full structure are of course much larger.

(5.9) (a) {to, {to, KPJohn}}

(b) {give, {KPa book, {give, {give, PPto John}}}}

(c) {T, {KPMary, T’}}

Traditionally non-selected material is said to be adjoined to phrases. Let’s take the following example.

(5.10) Pat sang the song loudly.

(5.11) vP

KP v 0

Pat v VP

Vi v VP AdvP

sang ti KP loudly

the song 5.2. TRANSFER AND THE INTERFACES 79

Lexicon

Overt Movement Spell-Out

MovementCovert

Morphology

PF LF

Figure 5.1: T-Model of Syntax

5.2 Transfer and the Interfaces

You may be familiar with the notions of Deep Structure (DS) and Surface Structure (SS) from Government and Binding Theory. Again, in an effort to simplify UG Chomsky proposed that DS and SS could be eliminated (Chomsky and Lasnik, 1995). Syntactic representations still need to be interpreted for meaning and production. The conceptual-intentional interface is responsible for interfacing with the cognitive properties of the mind/brain. The articulatory-perceptual in- terface is responsible for converting the lingusitic representation into a series of commands for the speech apparatus (or the manual apparatus in the case of signed languages). Figure (5.1 shows the levels of representation that are thought to be necessary for language to be usable. The syntactic representation is converted into phonetic form (or PF) to be sent to the articulatory-perceptual interface. Likewise, the syntactic representation is also converted into logical form (or LF) to be sent the conceptual-intentional interface. When syntactic structure is built it reaches a point called Spell-Out, and the derivation is sent to the two interface level, PF and LF. Recall from above that movement may take place. So far, we have seen only instances of overt movement. However, covert movement may take place after Spell-Out on the branch of the derivation that leads to LF. This movement will not be detected by PF, so it will appear to the speaker/listener that no movement has taken place. On the branch of the derivation that leads to PF, morphological and phonological changes may take place, but these will not give rise to changes in meaning, as these changes are not detectable by LF. 80 CHAPTER 5. THE ARCHITECTURE OF SYNTAX

The earliest application of covert movement was quantifier raising (QR), discussed in depth by May (1985). Consider the following example. This sentence is ambiguous and has the two readings indicated.

(5.12) Someone read every book

(a) There is one person such that that person read every book. ∃ > ∀ (b) For every book someone or other read it. ∀ > ∃

The quantifier some, as in someone, something, some book, and so forth is an existential quan- tifier. It presupposes the existence of some entity. In our example it presupposes the existence of a person. The quantifier every is a universal quantifier. It presupposes, roughly speaking, a universal set of entities. In our example, it presupposes a universal set of books. Of course, it is restricted to a contextually salient situation. Thus, this sentence does not refer to every book in the universe, but rather to every book in some situation, such as a particular list of books under discussion, or the set of books on someone’s bookshelf, for example. In the first reading, in (5.12 a, we say that the existential quantifier takes scope over the universal quantifier. In the reading in (5.12 b, we say that the universal quantifier takes scope over the existential quantifier. This is also called the "wide scope reading". May (1985) proposed that scopal relationships are calculated at LF. In order for a quantifier X to take scope over a quantifier Y, X must c-command Y at LF. For the reading in (5.12 a nothing needs to happen. For the reading in (5.12 b, however, the phrase every book must c-command someone at LF. There is no overt movement, so May proposed that the phrase every book undergoes QR at LF to give rise to the reading in (5.12 b. For clarity, covert movement is often represented as a dashed line.

(5.13) Structure at Spell-Out 5.2. TRANSFER AND THE INTERFACES 81

CP

C TP

KPi VoiceP

0 someone ti Voice

Voice vP

v VP

Vj v tj KP

read every book

(5.14) Structure at LF, after covert movement

CP

C TP

KPk TP

every book KPi VoiceP

0 someone ti Voice

Voice vP

v VP

Vj v tj tk

read

Covert movement has also been implicated in wh-in-situ. We will discuss wh-movement in detail 82 CHAPTER 5. THE ARCHITECTURE OF SYNTAX later, but we can illustrate the phenomenon here. Huang (1982) discussed wh-in-situ in Mandarin and proposed LF movement or covert movement handle the similarities between wh-constructions in languages with and without wh-movement. Consider the following data.

(5.15) (a) I wonder what John ate. (b) I think that John ate an apple. (c) Wˇo xiˇang-zh¯ıdào L¯ıs¯ı ch¯ı-le shénme I wonder Lisi eat-pvf what ‘I wonder what Lisi ate.’ (d) Wˇo rènwéi L¯ıs¯ı ch¯ı-le píngguˇo I think Lisi eat-pvf apple ‘I think Lisi ate an apple.’

Like English, wonder in Mandarin selects an interrogative clausal complement and know selects a declarative clausal complement. In order to maintain a uniform analysis of selection Huang proposed that the wh-phrase in Mandarin move to SpecCP covertly. PF movement is, of course, a theoretical possibility. Within Distributed Morphology PF operations are thought to play a role in morphosyntax (Halle and Marantz, 1993; Embick and Noyer, 2007, 2001). Certain clitics, such as second position clitics are also thought to be subject to PF movement (Halpern, 1995). One aspect of grammar that has garnered much discussion is head movement. Recall from above that movement targets the root projection. Head movement is different, however. Observe that it does not target the root, but rather the head of the highest projection.

target of XP movement

XP

X YP

...

target of head movement

As a result, it has been suggested that head movement takes place at PF, since it does not obey standard syntactic constraints (Boeckx and Stjepanović, 2001). 5.3. PHASE THEORY 83

5.3 Phase Theory

Although we take up wh-movement in much more detail in Chapter 7, we will discuss some aspects of it here, as it has played a significant role in the history of syntactic theorizing. Wh-phrases are capable of undergoing long-distance movement. Consider the following example.

(5.16) (a) Which book does John think that Mary said Fred would read? (b) John thinks that Mary said Fred would read Sam Dae.

Which book is the thematic object of read as can be ascertained from the answer in the following example. There is evidence that the wh-phrase does not move in one fell swoop. Rather, it moves in steps. Consider the following German data.

(5.17) (a) Was denkst du, was wir essen sollen? what think you what we eat should ‘What do you think we should eat?’ (b) Wo sagt er, wo das Buch liegt? Where said he where the book lies ‘Where did he say the book is?’

The wh-words was (‘what’) and wo (‘where’) have undergone long-distance movement; however, they have left a copy at the left edge of the lower clause. It is assumed that the wh-phrases move through the specifier of each CP. Again, in chapter 7 we will cover more examples of long-distance movement. In the following tree vP has been removed to save space. 84 CHAPTER 5. THE ARCHITECTURE OF SYNTAX

CP

0 KPi C

which book C TP

does KP T0

John T VP

V CP

0 think ti C

C TP

that KP T0

Mary T VP

V CP

0 said ti C

C TP

KP T0

Fred T VP

would V ti

read

From the German examples we see that a copy of the wh-phrase is left in each specifier, a phenomenon known as wh-copy constructions (McDaniel, 1986, 1989; van Riemsdijk, 1982). The 5.3. PHASE THEORY 85 wh-copy construction provides evidence for the copy theory of movement described above. This step-by-step movement, called successive cyclicity, has had a long history in generative syntax (Chomsky, 1973, 1981b; Rizzi, 1982; Huang, 1982). These cycles are called phases and are identified by phase heads Chomsky (2001, 2000); Svenonius (2004).2 If vP is split into VoiceP > vP, then Voice is typically taken to be the phase head. Chomsky (2000) assumes that only v that introduces an external argument (that is, transitive and unergative v) is a phase head. We will come back to this point below. Likewise, if K is assumed then K, rather than D would be assumed to be the phase head, although research in this domain is scant, and the phase properties inside nominals requires further research.

(5.18)C, v*, D, P

The phase head defines the Spell-Out domain as follows. Once the phase head has satisfied all its requirements the sister to the phase head undergoes Spell-Out and is sent to the interfaces. Let’s consider the lower C head in the following example.

(5.19) I wonder which book Mary read.

Let’s consider the stage when the embedded C is being built.

(5.20) CP

C TP

0 KPi T

Mary T vP

0 ti v

v VP

Vj v tj KP

read which book

The KP which book moves to SpecCP as this is an interrogative clause. The phase is now complete, and the sister to the phase head, TP, undergoes Spell-Out and is sent to the interfaces.

2It is an open-ended question whether n is also a phase head; however, Marantz (2001) proposes that all categorizers (n, v, and a) are phase heads. 86 CHAPTER 5. THE ARCHITECTURE OF SYNTAX

CP

0 KPk C Phase

which book C TP

0 KPi T

Mary T vP

0 ti v

v VP

Vj v tj tk

(5.21) read

The TP shown is sent to the interfaces for interpretation at PF and LF. The CP remains and is selected by the verb wonder. The Spelled-Out portion is in grey italics.

(5.22) VP

V CP

0 wonder KPk C

which book C TP

You may have noticed that the vP phase was bypassed in this example. According to Phase Theory, the wh-phrase stops in SpecvP as follows.

(5.23) (a) vP

KP v 0

Mary v VP

Vj v tj KP

read which book 5.3. PHASE THEORY 87

(b) vP

KPj vP

which book KP v 0

Mary v VP

Vi v ti tj

read (c) vP

KPj vP

which book KP v 0

Mary v VP

Vi v

read

The wh-phrase is first merged as the object of the verb and then moves to SpecvP. The sister of v is Spelled-Out and is sent to the interfaces. Evidence for successive-cyclic movement of wh-phrases through intermediate SpecCP positions has a long history. It will be addressed again in Chapter 7. Evidence for successive-cyclic movement through SpecvP is less forthcoming, however. Let’s consider some evidence from Tagalog. Tagalog, like many related Philippine languages, has a complex voice system in which one argument is singled out as prominent. In Tagalog specifically, that argument is signaled with ang and is often referred to as the subject. Consider the following examples (Kroeger, 1993, p.13-14). Note that infixes are indicated inside . Thus, in the first example, theverb root is bili, and the infix -um- indicates perfective aspect and that the nominative argument is the subject. Note also that non-nominative arguments are typically marked with genitive case, gen.

(5.24) Tagalog Voice, (Kroeger, 1993, ex.13)

(a) b-um-ili ang lalake ng isda sa tindahan buy nom man gen fish dat store ‘The man bought the fish at the store.’ (b) b-in-ili- ng lalake ang isda sa tindahan buy-obj gen man nom fish dat store 88 CHAPTER 5. THE ARCHITECTURE OF SYNTAX

‘The man bought the fish at the store.’ (c) b-in-ilh-an ng lalake ng isda ang tindahan buy-loc gen man gen fish nom store ‘The man bought the fish at the store.’ (d) i-b-in-ili ng lalake ng isda ang bata ben-buy gen man gen fish nom child ‘The man bought the fish for the child.’

In the first example, the subject is marked with ang (nominative), and subject morphology appears on the verb (subj). In the second example, the object is marked with ang, and object morphology appears on the verb (obj). In the third example the location is marked with ang, and locative morphology appears on the verb (loc). Another property of Tagalog is that only the ang-marked phrase can engage in wh-movement. In Tagalog, the wh-phrase moves to the left edge of the clause. The wh-phrase is followed by a morpheme labelled ang. Consider the following data.

(5.25) wh-movement in Tagalog, (Rackowski and Richards, 2005, ex.2)

(a) Sino ang bigy-an ng lalaki ng bulaklak? who ang give-dat gen man gen flower ‘Who did the man give the flower to?’ (b) *Sino ang nag-bigay ang lalaki ng bulaklak? who ang prfv.subj-give nom man gen flower ‘(Who did the man give the flower to?)’

(5.26) Wh-movement in Tagalgo (Kroeger, 1993, p.62f)

(a) Ano ang kain mo? what ang eat you.gen ‘What did you eat?’ (b) Sino ang gawa ng sapatos na iyon? who ang make gen shoes link dem ‘Who made those shoes?’

Observe that in (5.25 a) the indirect object has undergone wh-movement, and the verb agrees with the indirect object (dat). If the verb agrees with a different argument, as in (5.25 b), then the indirect object cannot undergo wh-movement. Example (5.26) has more examples that illustrate the same point. Only the ang-marked phrase can undergo wh-movement, and it is the ang-marked phrase that agrees with the verb. Rackowski and Richards (2005) take these facts to mean that the ang-marked phrase, the subject, must pass through SpecvP to trigger agreement on the verb before moving to its final landing space.3 3See Kroeger (1993) for arguments that the ang-marked phrase is the grammatical subject in Tagalog. 5.3. PHASE THEORY 89

Chamorro (cha) also has wh-agreement on its verbs, suggesting that the wh-phrase has moved through SpecvP (Chung, 1998).

(5.27) Hafa sinangani hao ni chi’lu-mu malago’-ña? what wh.was.told you obl sibling-agr wh.want-agr ‘What did your sister tell you that she wants?’

? gives semantic evidence that SpecvP is a reconstruction site. Consider the following exam- ples.

(5.28) (a) [Which of the papers that he1 gave Mary2 ] did every student1 ok ask her2 to read * carefully? (b) *[Which of the papers that he1 gave Mary2 ] did she2 * ask every student1 to revise * ?

In both of these examples he1 is a bound variable that must be c-commanded by the operator every student. At the same time, Mary cannot be c-commanded by a co-referential pronoun. In both cases, a Principle C violation is incurred if the phrase in square brackets is interpreted in (or ‘reconstructs to’) its base position at LF, indicated by the underlined asterisk. However, if the bracketed phrase moves through SpecvP as indicated, then Mary can escape a Principle C violation, but the variable can still be bound by its operator. Fox thus reasoned that SpecvP must be an escape hatch for movement, hence a phase. ? showed, however, that the same reasoning can be applied to unaccusative and passive v, suggesting that v is general is a phase.

(5.29) (a) [At which of the parties that he1 invited Mary2 to] was every man1 ok introduced to her2 * ? (b) *[At which of the parties that he1 invited Mary2 to] was she2 ok introduced to every man1 * ? (c) [At which ceremony he1 invited Mary2 to] did every winner1’s name ok escape her2 * ?

In light of the arguments raised by Legate and the issues of the split vP raised in the previous chapter, the phase properties of the vP layer still require further research. Phases constitute cyclic domains for Spell-Out (Chomsky, 2000, 2001; Fox and Pesetsky, 2005; Uriagereka, 2012). At Spell-Out, the derivation is transferred to the interfaces: LF and PF. Once a chunk of the derivation has been transferred to the interfaces, it is no longer available for overt syntactic processes. This is called the Phase Impenetrability Condition. There are three com- peting models for the timing of Transfer, which we show here. The first we label here as ‘immediate 90 CHAPTER 5. THE ARCHITECTURE OF SYNTAX transfer’. As soon as all the requirements of the phase head are met, the sister to the phase head is spelled out.

(5.30) Immediate Transfer (Chomsky, 2000)

vP

KP v 0

EA v VP

V KP

IA

Icelandic has quirky case constructions in which the subject has dative case and the object has nominative case. In this situation, T agrees with the nominative object. In the model in (5.30) the nominative object is transferred before T is merged, and Agree will be impossible. Consider the following example (?, ex.3). Observe that the verb agrees with the object.

(5.31) Henni leiddust strákarnir her.dat found.3pl boys.nom.pl ‘She found the boys boring’ [Icelandic]

Thus, the ‘delayed transfer’ model was proposed. Under this model the sister to a phase head is transferred as soon as the next highest phase head is merged in. Observe in the following tree that T can agree with the object inside the VP since VP is not sent to the interfaces until C is merged in the derivation.

(5.32) Delayed Transfer (Chomsky, 2001) 5.3. PHASE THEORY 91

CP

C TP

Spec T0

T vP

KP v 0

EA v VP

V KP

IA

The third model we can call the ‘whole transfer’ model. This model is often assumed in works on prosody, as it is often recognized that phases often correspond to prosodic units (Newell and Scheer, 2017; Kratzer and Selkirk, 2007; Compton and Pittman, 2010).

(5.33) Whole Transfer

CP

C TP

Spec T0

T vP

KP v 0

EA v VP

V KP

IA

In the whole transfer model, once the higher phase head is merged in, the entire phase, including the phase head and it specifier(s) is transferred to the interfaces. 92 CHAPTER 5. THE ARCHITECTURE OF SYNTAX

Further Reading

• ? - This review article covers the basic properties of Merge and BPS and discusses ideas for cross-linguistic (especially with Japanese) and for future directions in research.

• Citko (2014) - This pedagogical monograph describes the background on phases, discusses current research, and suggests areas for future research. Chapter 6

Case and A-Movement

The distinction between A-movement and A-bar-movement (also A-movement) was first discussed in ?. A-movement is movement into an A-position (argument position). A-bar movement is move- ment into an A-bar-position (something other than an argument position). The terminology can be understood in reference to the Extended Standard Theory representation of a sentence.

(6.1) S0

XP COMP S

A-bar position XP INFL VP

A-position ...V...

The specifier of S is an A-bar position, and the specifier of S is an A-position. Updating to a contemporary structure as presented in previous chapters, an A-position is typically identified as an position where a theta-role is assigned or where Case is assigned. In this chapter, we will examine Case and A-movement.

6.1 Case Assignment

Anyone with a Classical background will be familiar with the notion of from studying Latin or Ancient Greek. The academic study of Case goes back both to Ancient Greece, where Thrax undertook a massive study of the grammar of Greek, and earlier to ancient India, where P¯an. ini studied Sanskrit grammar. For more details, see ??. Let’s start by looking at the following Latin sentence. Notice that the subject is marked with nominative case (nom) and the direct object is marked with accusative case (acc).

93 94 CHAPTER 6. CASE AND A-MOVEMENT

(6.2) amic-us puell-am amat. friend-nom girl-acc loves ‘The friend loves the girl.’

Notice the suffixes on the noun phrases. We call these case suffixes. If you are familiar with Latin, you will recall that there are six cases. For now, let’s just worry about the two that we see in (6.2. We note that the subject appears with nominative case and the direct object appears with accusative case. For now, we will not worry about other cases you may know such as dative, ablative, and so forth. Although case marking is extremely limited in modern English, it still appears on pronouns. Compare I saw him and He saw me. The pronouns I and he are used as subjects only, and are thus assumed to be marked with nominative case, while the pronouns me and him are used as objects only are marked with accusative case. Case markings are not always visible, however. For example, in the English translation of the Latin sentence above there is no visible case marking on the noun. We still say that the nouns have been assigned Case, though. Notice the capital ‘C’ on the word Case here. We use the term ‘case’ to refer to the various morphological markings as in Latin, Russian, Finnish and other languages with overt case morphology. The term ‘Case’, with a capital ‘C’ is used to refer to the identification of the structural position of the noun – that is, is the noun in subject position or object position. This type of Case is more theoretical and will be built up in this chapter. While the presence of case varies from language to language, it is assumed that Case is present in all languages. For our purposes, we will start with the assumption that Case and case are identical for nominate and accusative, though we will see below that we will modify this assumption slightly. Thus, if the form of a pronoun is he or she, for example, we can conclude that this argument has nominative Case. If the form of a pronoun is him or her, we can conclude it may have accusative Case. Note that the forms him and her also appear as complements to prepositions. We will discuss this difference later, however. First we will cover nominative Case.

6.1.1 Nominative Case

Consider the following sentences. Pay close attention to the subject of the verb eat in all three sentences.

(6.3) Distribution of Nominative and Accusative Pronouns

(a) She ate the spinach. (b) For her to eat the spinach would be surprising. (c) I would really like for her to eat the spinach.

In example (6.3 a), the verb is tensed. In particular, it is marked with past tense. Furthermore, the subject of this sentence appears with nominative Case, as we have discussed above. In examples 6.1. CASE ASSIGNMENT 95

(6.3 b) and (6.3 c), however, we still understand her to be the logical subject of the sentence – that is, her gets the external θ-role (the subject θ-role) from eat. However, her is marked with accusative Case. Notice further that the verb eat in the latter two sentences are infinitives. That is, they are not have any tense. This correlation led researchers to conclude that nominative Case and tense are intimately intertwined such that finite T is responsible for assigning nominative Case (??). If there is no thematic subject, but a tensed T is still present, then an expletive appears in SpecTP.

(6.4) (a) It’s raining. (b) It seems that John left early. (c) John seems to have left early.

Looking first at (6.4 a), observe that there is no meaningful subject. There is no ‘it’ that is doing the raining. In fact, in all these sentences the word it does not contribute any meaning to the sentence – it does not refer to anything. Since the expletive does not refer to anything, it does not get a θ-role – only actual participants in an event or state get a θ-role. They are also sometimes referred to as pleonastic pronouns. As we saw above, Icelandic shows the relationship between tensed T and nominative Case more clearly. Consider the following examples (?).

(6.5) (a) Henni leiddust þeir 3sg.f.dat was.bored.3pl 3pl.nom ‘She was bored with them.’ (b) Mér líkar bókin 1sg.dat like.3sg book.nom ‘I like the book’ (c) *Mér líka bókin 1sg.dat like.1sg book.nom ‘I like the book’

Observe that the verb does not agree with the grammatical subject in these examples. Rather agreement holds between the nominative nominal and T.1 Let’s consider the relationship between tense and agreement more closely before we relate this to nominative Case. Tensed T bears both tense and agreement morphology, giving rise to the idea that the two are related (Chomsky, 1982; Rizzi, 1982). A quick glance at a number of languages shows the correlation between tense and agreement. The English data show that 3rd person singular agreement is obligatory with the present tense form and unavailable with the to-infinitival form.

(6.6) English

1The agreement facts in Icelandic are more complicated than made out to be here. Consult ??? for more details. 96 CHAPTER 6. CASE AND A-MOVEMENT

(a) He eat*(s) apples. (b) For him to eat(*s) apples...

The Italian data show that the present tense form of mangiare (‘eat’) bears subject agreement and that the infinitival form has no agreement.

(6.7) Italian

(a) Io mangi-o una mela. I eat-1sg an apple ‘I’m eating an apple.’ (b) Vogli-o mangi-are una mela. want-1sg eat-inf an apple ‘I want to eat an apple’

The following Hungarian data show the same correlation. The form for ‘like’ bears both present tense and agreement, while the form for ‘read’ bears neither tense nor agreement.

(6.8) Hungarian, (Rounds, 2001)

(a) Szeret-ek olvas-ni. like.prs-1sg read-inf ‘I like to read.’ (b) Szeret-sz olvas-ni. like.prs-2sg read-inf ‘You like to read.’

Returning to the issue of nominative Case, we ask whether it is tense or agreement that is responsible for the assignment of said Case. Given the data above it is impossible to determine. Ra- poso (1987) discusses inflected infinitives in European Portuguese. Crucially, inflected infinitives bear agreement, but not tense.2 Observe that the subjects bear nominative case.

(6.9) Portuguese Inflected Infinitives

(a) Será dificil eles aprovar-em a proposta will.be difficult they.nom approve.inf-3pl the proposal ‘It will be difficult for them to approve the proposal.’ (b) É melhor ficar-es em casa is better star.inf-2sg in house ‘It’s better for you to stay at home.’ 2? proposes that has tense but lacks agreement. Assuming that Chinese has case (??) it is possible that the licensing of nominative Case is parameterized. See also ? for a discussion on agreement in . 6.1. CASE ASSIGNMENT 97

Other Romance varieties also possess inflected infinitives. They are also found in Welsh and Hungarian. Inflected infinitives strongly suggest that it is agreement on T that is responsible for the ability of T to assign nominative Case, a position which is still generally accepted (Chomsky, 2000, 2001).

6.1.2 Accusative Case

Since the time of Aristotle and P¯an. ini it was noted that the accusative case (대격) appears on the direct object (직접 목적어). Let’s look at some examples first.

(6.10) Accusative Case

(a) Yenghuy-ka sakwa-lul mek-ess-ta 영희가 사과를 먹었다. Younghui-nom apple-acc eat-pst-decl ‘Younghui ate an apple.’ (b) Hans isst ein-en Apfel Hans eat.3sg.prs an-masc.acc apple ‘Hans is eating an apple.’ [German]

In traditional grammar it was assumed that certain verbs assign accusative Case. For instance, the verb laugh does not assign accusative Case, but the verb kick does. Furthermore, it was thought that Case is obligatorily assigned. Thus, in the following examples, laugh cannot take a direct object, while kick must take a direct object.

(6.11) (a) John laughed (*the joke). (b) John kicked *(the fence).

Note, however, the counter-examples in (6.12). The last six of these are taken from ?. In reality, there is a great deal of flexibility in the argument structure of individual lexical items. While many traditional grammars still contend that individual verbs are lexically specified as being transitive or intransitive, the idea that such strict specifications simply do not exist in grammar is gaining ground.3 This alternative view holds that a verb such as kick tends to be transitive, but can be intransitive.

(6.12) (a) John laughed himself silly. (b) The baby kicked. (often said by pregnant women)

3It is important to remember here that many grammars that hold that verbs are either transitive or intransitive are doing so from either a pedagogical or a descriptive perspective and not from the perspective of explaining human grammar within a generative framework. 98 CHAPTER 6. CASE AND A-MOVEMENT

(c) John ate the apple. (d) John ate at the apple. (e) The sea ate into the coastline. (f) John ate me out of house and home. (g) John ate. (h) John ate his way into history.

As an introduction to the ideas that are about to follow, consider the following examples of active and passive sentences.

(6.13) (a) I saw him. (b) He was seen. (c) cip-ul namwu-lo mandul-ess-ta. 집을 나무로 만들었다 house-acc wood-instr make-pst-decl ‘(Someone) made the house out of wood.’ (d) cip-i namwu-lo mandul-e-ci-ess-ta. 집을 나무로 만등러졌다 house-nom wood-instr make-inf-pass-pst-decl ‘The house was made out of wood.’

In the passive examples the internal argument does not bear accusative case. Rather, it bears nominative case. Luigi Burzio is credited with bringing the importance of this phenomenon to light, although David Perlmutter’s research was a precursor to Burzio’s discussion (?Perlmutter, 1978). The phenomenon in question is known as Burzio’s Generalization and is stated as follows.

(6.14) All and only the verbs that can assign a theta-role to the subject can assign Accusative Case to an object.

With the proposal of v (?Chomsky and Lasnik, 1995), Chomsky proposed that v was responsible for assigning accusative Case. Given that passives lack accusative Case and an external argument, it was reasonable to assume that passives lack the vP projection.

(6.15) (a) vP

KP v 0

Mary v VP

V KP

ate the apple 6.1. CASE ASSIGNMENT 99

(b) VP

V KP

ate the apple

Under this view, the internal argument in (6.17 a) receives accusative Case from v. In (6.17 b), however, v is absent, so the internal argument receives nominative Case from T. Likewise, -inchoative pairs can receive the same treatment. Consider the following pair of sentences.

(6.16) (a) The enemy sank the ship. (b) The ship sank.

In the causative variant in (6.16 a) the internal argument receives accusative Case from v, and in the inchoative variant in (6.16 b) the internal argument receives nominative Case from T. The relevant portions of the tree are shown again.

(6.17) (a) vP

KP v 0

the enemy v VP

V KP

sank the ship

(b) VP

V KP

sank the ship

Under this view inchoatives and passives receive the same analysis; however, their properties differ. As discussed by ?, passives retain passives retain the theta-role of the external argument, even though the external argument is not projected. Inchoatives, on the other hand, do not. Compare the following sets of data.

(6.18) (a) The enemy sank the ship on purpose. (b) The ship was sunk on purpose. (c) The ship sank (*on purpose). 100 CHAPTER 6. CASE AND A-MOVEMENT

(6.19) (a) The fraudster sank the ship to collect the insurance money. (b) The ship was sunk to collect the insurance money. (c) The ship sank (*to collect the insurance money).

Notice also that the agent can appear optionally in a by-phrase in passives. The same is not true for inchoatives.

(6.20) (a) The ship was sunk (by the enemy). (b) The ship sank (*by the enemy).

Alexiadou et al. (2015) propose that syntactic transitivity and semantic transitivity are in- troduced by separate features on the Voice head. Here is their proposal, updating the D feature to a K feature. Note that there is a tremendous amount of cross-linguistic variation in passives and causative-inchoative alternations. This is just one proposal. Alexiadou et al. acknowledge that tradtional tests for agentivity fail for inchoatives (as shown above); however, they contend that an external causer is still possible with inchoatives. Consider the following paradigm, where the external argument is a causer rather than an agent.

(6.21) (a) The sun melted the ice. (b) The ice was melted (by the sun). (c) The ice melted (from the sun).

They also discuss the by-itself diagnostic. The phrase by-itself can mean ‘alone’ or ‘of no par- ticular cause’. It is this second meaning that Alexiadou et al. focus on. They give the following example.

(6.22) A: But you didn’t take care of the trampoline. B: That’s not true! I did everything I was supposed to, but the trampoline blew away by itself. B’: ??That’s not true! I did everything I was supposed, but the trampoline was blown away by itself.

They propose, then, that inchoatives have the same structure as transitives, but that the features of the Voice head differ among different kinds of detransitivized predicates. Here is the typology they propose.

(6.23) Typology of Voice

(a) Voice[λx, uK] - active voice 6.1. CASE ASSIGNMENT 101

(b) Voice[λx] - passive voice (c) Voice[uK] - SE reflexives4 (d) Voice has no features - lexically marked inchoative verbs (found in Greek) (e) Voice absent - unmarked inchoatives (found in English)

The string λx is a semantic notation to introduce an agent variable. This means there is a semantic agent present. In the active voice it is satisfied by the argument introduced by the [uK] feature.5 In passives it is satisfied by , a semantic operation whereby a variable is satisfied by merely stating that an entity exists that satisfies it, but is not mentioned. Here, then, are the proposed structures for the sentences in (6.18).

(6.24) (a) VoiceP

KP Voice0

the enemy Voice vP

λx,uK v VP

V KP

sank the ship

(b) VoiceP

Voice vP

λx v VP

V KP

sank the ship

(c) vP

v VP

V KP

sank the ship

4SE reflexives are found in Romance languages, among others. See Alexiadou et al. (2015) for more details on their proposal. 5The full semantic is, in simplified terms, as follows: λxλP.x is the agent of P 102 CHAPTER 6. CASE AND A-MOVEMENT

Note in these trees that we have now lost the correlation between Voice/v and accusative Case. Note, however, with nominative case above we assumed that non-finite T lacked the ability to assign nominative Case. Crucially, it is not the absence of the T head that fails to give rise to nominative, but rather a different kind of T head. Likewise, we can assume that it is a different kind of Voice head that fails to give rise to accusative Case assignment. Chomsky (2000, 2001) assumed without argumentation that accusative Case assignment is also dependent on object agreement. Most of the languages worked on within the framework of Mini- malism, however, do not have overt object agreement, so to investigate the hypothesis a language with overt object agreement is needed. ? examined data from Amharic (amh), which has overt case marking and object agreement. Consider the following data (?, ex.(3-4)), see also ?.

(6.25) Accusative Case in Amharic

(a) l@mma w1SSa j-aj-al Lemma dog 3sg.nom-see-aux ‘Lemma sees a dog.’ (b) *l@mma w1SSa j-aj-@w-al Lemma dog 3sg.nom-see-3sg.acc-aux ‘Lemma sees a dog.’ (c) l@mma w1SSa-w-1n j-aj-@w-al Lemma dog-def-acc 3sg.nom-see-3sg.acc-aux ‘Lemma sees a dog.’ (d) *l@mma w1SSa-w j-aj-@w-al Lemma dog-def 3sg.nom-see-3sg.acc-aux ‘Lemma sees a dog.’ (e) l@mma w1SSa-w-1n j-aj-al Lemma dogdef-acc 3sg.nom-see-aux ‘Lemma sees a dog.’

Examples (6.25 a) and (6.25 c) suggest that there is a correlation between object agreement and accusative Case. In (6.25 a) the indefinite object lacks accusative case, and object agreement is absent. (6.25 b) shows that object agreement is impossible when the object is not case marked. When the object is definite, it bears accusative case morphology, and the verb appears with object agreement, (6.25 c). Note that the case suffix is obligatory when the object is definite, (6.25 d). The first four examples suggest a tidy correlation between object agreement and accusative Case assign- ment. However, example (6.25 e) shows that when the direct object is definite, object agreement is optional. Baker suggests, then, that the correlation between Case and agreement is parameterized. We will revisit the issue of Case assignment at the end of this chapter. For now, we assume that Voice may assign accusative Case to the direct object. 6.2. A-MOVEMENT 103

6.2 A-Movement

As we have seen above, SpecTP in English must be filled.

(6.26) (a) The children have all t eaten lunch. (b) *(There) is a unicorn in the garden. (c) The apple was eaten. (d) *was eaten the apple.

The same is not true for all languages, however. Consider the following Italian examples.

(6.27) Passives in Italian

(a) La mela è stata mangiato t the apple is been eaten ‘The apple was eaten.’ (b) È stata mangiato la mela was been eaten the apple ‘The apple was eaten.’

Explaining this difference has been notoriously challenging. Chomsky (1982) proposed the Ex- tended Projection Principle (EPP), which states that a tensed clause must contain a subject in SpecTP. The EPP is little more than a diacritic to indicate that in certain languages SpecTP must be filled.6 Later, Chomsky proposed the notion of feature strength to describe movement (Chom- sky and Lasnik, 1995). Again, this is acknowledged to be little more than a diacritic.7 The notion of feature strength holds that strong features trigger merge of a phrase in the specifier position of the head that possesses the strong feature. Specifically for the observation that SpecTP in a tensed claused must be filled in English, Chomsky proposed that in English T has a strong D feature. We can say a strong K feature, assuming a KP. This is notated as [uK*]. The asterisk indicates that the feature is strong and must be satisfied by merging something with a K feature SpecTP. As it stands, we have little more than a diacritic EPP feature to encode the observation that the highest argument raises to SpecTP. Nevertheless, A-movement is a well-attested phenomenon across many languages. Passives discussed above are a well-known example of this. We suggested above that passives result from a defective Voice head that is incapable of assigning accusative Case

6See Alexiadou and Anagnostopoulou (1998) for a suggestion that EPP in languages like English is satisfied by KP movement (DP movement in their terms) while EPP in languages like Italian is satisfied by V movement to T. In a similar vein, Massam (2001) proposes that the V feature on T is satisfied by VP movement rather than by V movement. 7For a recent attempt that reduces overt movement to prosodic properties of language, see Richards (2016) 104 CHAPTER 6. CASE AND A-MOVEMENT

or introducing an external argument. For convenience, we will notate this as VoiPASS. Here is an example.8

(6.28) (a) The apple was eaten. (b) CP

C TP

0 KPi T

the apple T VoiPASSP

was VoiPASS vP

v VP

Vj v tj ti

eaten

A-movement is also exhibited by raising to subject and raising to object constructions.9

(6.29) (a) Mary seems to enjoy syntax. - raising to subject (b) Mary expects Bill to enjoy syntax - raising to object

Raising to subject constructions are relatively uncontroversial as the word order clearly indicates that movement has taken place. Raising to object constructions, however, have given rise to con- siderable discussion in the literature (Postal, 1974; Chomsky and Lasnik, 1977; Lasnik and Saito, 1991b). Crucially, the debates centered on whether the object raises to the matrix clause (raising to object) or remains in-situ (ECM). The two possibilities are shown below.

(6.30) (a) Mary expects Billi[TPtitoenjoysyntax]. −

raisingtoobjectanalysisMaryexpects[TPBilltoenjoysyntax]. − ECManalysis

Part of the answer to this question requires us to look at particle verbs in English, as in the following examples (Johnson, 1991).

(b)(6.31) (a) Pat took the garbage out.

8We have not dealt here with how auxiliaries are handled in English (or in other languages). For a more traditional approach, see Adger (2003). For a more innovative approach that tackles problems not addressed by Adger, see Bjorkman (2011) 9Raising to object is also known as ECM and will be discussed in more detail later. 6.2. A-MOVEMENT 105

(b) Pat took out the garbage.

Updating Johnson’s terminology and structures with work by Basilico (2008), we can represent the analysis as follows.10 Note that this structure requires to a novel conception of head movement and head formation. In particular, the verb movement out of the complex V+PRT head has the appearance of excorporation (Roberts, 1991). We will set these issues aside as the main goal here is to examine object movement.

(6.32) VoiP

KP Voi0

Pat Voi vP

0 v i Voi KPk v

Vj v the garbage ti VP

took V tk

tj Prt

out

To account for the order in which the object appears to the right of the particle, Basilico argued that the object could be merged into the derivation later, into a position where movement is not obligatory.

(6.33) VoiP

KP Voi0

Pat Voi vP

v i Voi v KP

Vj v ti V the garbage

took tj Prt

out

10See also Koizumi (1995) for additional arguments for object movement. 106 CHAPTER 6. CASE AND A-MOVEMENT

Johnson, then, argued that the traditional notion in which the object and the verb are adjacent is not tenable in light of particle verb constructions. Basilico then gave a more updated analysis of the same kinds of constructions. Note, though, that the structures given here deviate slightly from Basilico’s original proposal. The reader is advised to consult both Johnson and Basilico for further details. The issue of object movement is even more apparent in ECM constructions. Consider the fol- lowing data. Note that (6.34 c) is possible only for some English speakers.

(6.34) (a) John made out that Bill is a fool. (b) John made Bill out to be a fool. (c) %John made out Bill to be a fool.

The particle verb make out means to make something appear to be true. It can take a finite clausal complement or an ECM complement as in the two examples above. Since the particle out is part of the matrix clause and Bill is part of the embedded clause, it is clear that the embedded subject has raised to an object position in the matrix clause. Lasnik and Saito (1992) present the following data, which also shows that the object raises to a higher position.

(6.35) (a) The DA proved the defendants1 to be guilty during each other1’s trials. (b) The DA made the defendants1 out to be guilty during each other1’s trials. (c) *?The DA made out the defendants1 to be guilty during each other1’s trials.

The during each other’s trials modifies the matrix predicate, so must appear in the matrix clause. Since it contains an reciprocal anaphor, it must be c-commanded by its antecedent, strongly suggesting that the object the defendants must raise overtly to the matrix clause. Further- more, when the particle verb make out is used with the adjunct containing the anaphor, then it is no longer an option for speakers who find (6.34 c) grammatical to leave the object in the lower clause. It remains to be investigated how the facts discussed by Lasnik and Saito can be reconciled with the structures proposed by Basilico.

6.3 Unaccusativity

Since Perlmutter (1978) it , is generally accepted that there are two types of intransitive verbs, as in the following examples.

(6.36) (a) The baby cried. (b) The baby grew. 6.3. UNACCUSATIVITY 107

In (6.36 a) the single argument is an external argument, and in (6.36 b) the single argument is an internal argument. These constructions are unaccusative and unergative, respectively. They have roughly the following representations.

(6.37) (a) TP

0 DPi T

the baby T vP

0 ti v

v VP

V

cried (b) TP

0 DPi T

the baby T VP

V ti

grew

The first illustration of unaccusativity is from Perlmutter’s 1978 discussion on semantic cor- relates of unaccusativity. There, he notes that unergatives are overwhelmingly agentive and that unaccusatives typically lack agentivity. Consider the following Dutch data that Perlmutter dis- cusses. Impersonal passives are possible with unergative verbs, but not with unaccusative verbs. In the framework that Perlmutter assumed (Role and Reference Grammar) he argued that impersonal passives are universally impossible with unaccusative verbs; however, he mentions that the claim requires further study.11

(6.38) Impersonal Passives with Unergative Verbs in Dutch

(a) Er wordt hier door de jonge lui veel gedanst. it becomes here by the young people much danced ‘Young people dance here a lot.’ 11He also notes that many pedagogical and grammars of Dutch (and other languages) claim that impersonal passives are possible with some intransitive verbs, give a few examples, but do not go into further details. He goes on to suggest that his discussion on the correlation between impersonal passives and the unergative/unaccusative distinction gives us a more concrete way to talk about this phenomenon. 108 CHAPTER 6. CASE AND A-MOVEMENT

(b) Er wordt in deze kamer vaak geslapen. It becomes in this room often slept ‘People sleep in this room often.’ (c) Er wordt hier veel geskied. It becomes here a lot skied ‘People ski a lot here.’

(6.39) Impersonal Passives with Unaccusative Verbs in Dutch

(a) *Er werd door het water binnen een kwartier verdampt. it became by the water within a quarter hour vanished ‘(The water evaporated within a quarter hour.)’ (b) *In dit weeshuis wordt er door de kinderen erg snel gegroeid. in this orphanage becomes it by the children very fast grown ‘(Children grow very fast in this orphanage.)’ (c) *Er werd door het water uit de kraan gegutst. it became by the water from the tap gushed ‘(Water gushed from the tap.)’

Perlmutter gives an extensive list of examples, only a few of which are shown here, to illustrate the difference between unergatives and unaccusatives. Although Perlmutter did acknowledge some cross-linguistic variation in the range of unaccusatives and unergatives, he did propose a universal semantic correlate - agentivity. Consider the following paradigm.

(6.40) Varying Adicity of the English verb roll

(a) Mary rolled the log down the hill. (b) The log rolled down the hill. (c) Mary rolled down the hill (to make herself dizzy). (d) Mary rolled down the hill (after she fell asleep and lost muscle tension).

The first example is clearly transitive and the second example is clearly unaccusative, as the underlying object now appears as the grammatical subject. The last two examples show that roll is ambiguous between an unergative and an unaccusative reading, following Perlmutter’s arguments. Before reviewing the analysis of unaccusatives let’s examine Burzio’s ? discussion of ne-cliticization in Italian. Consider the following data. Observe that the partitive clitic ne can target the object, but not the subject.

(6.41) ne cliticization in Italian transitive verbs 6.3. UNACCUSATIVITY 109

(a) Giovanni inviterà molti esperti Giovanni will invite many experts ‘Giovanni will invite many experts’ (b) Giovanni ne inviterà molti Giovanni ne will invite many ‘Giovanni will invite many of them.’ (c) Esamineranno il caso molti esperti will examine the case many experts ‘Many experts will examine the case’ (d) *Ne esamineranno il caso molti ne will examine the case many ‘(Many of them will examine the case.)’

Given the observation with transitive verbs that ne can target the object, but not the subject, the following examples with the verb telephone are unsurprising. However, observe that the single argument of the verb arrive can also be targetted by ne.

(6.42) ne cliticization in Italian intransitive verbs

(a) Molti esperti telefoneranno many experts will telephone ‘Many experts will telephone.’ (b) *Ne telefoneranno molti ne will telephone many ‘(Many of them will call.)’ (c) Molti esperti arriveranno many experts will arrive ‘Many experts will arrive.’ (d) Ne arriverano molti ne will arrive many ‘Many of them will arrive.’

Numerous other diagnostics of unaccusativity have been proposed including noun incorporation (?), there-insertion, and nuclear scope (?). The tripartite analysis of the VP layer (VoiceP > vP > VP) is fairly well accepted, even if the full structure is not always shown. Recall that nominative Case is absent in non-finite clauses (for him to leave...). Recall also the standardly accepted analysis in which T is not missing, but rather lacks tense and agreement features. In a similar vein, with passives and unaccusatives, neither vP nor VoiceP is missing, rather, they lack the feature responsible for assigning accusative Case. Whether that is object agreement or not is unclear given Baker’s discussion above. Given Burzio’s 110 CHAPTER 6. CASE AND A-MOVEMENT

Generalization, it is typically assumed that the feature responsible for introducing the external argument is also responsible for (or is parasitized by) accusative Case. Thus, we can assume a structure as follows for a typical unaccusative.12

(6.43) CP

C TP

0 KPi T

the children T VoiP

Voi vP

v VP

V ti

6.4 Ergativity and the Dependent Theory of Case

Recall that the grammatical object is assigned nominative Case and that the grammatical object is accusative Case. This is referred to as nominative-accusative case alignment. To be more precise, we distinguish between the subject of a transitive verb and the subject of an . The subject of a transitive verb is referred to as the AGENT (A). The subject of an intransitive verb is referred to as the SUBJECT (S). The object of a transitive verb is referred to as the (P). In a language with ergative-absolutive case alignment the AGENT bears ergative case, while the SUBJECT and PATIENT bear absolutive case. The following Basque (eus) example illustrates this type of alignment.

(6.44) Ergativity in Basque (Rezac et al., 2014, ex.6)

(a) Nekane-k Miren eta Jon ikusi ditu. Nekane-erg Miren.abs and Jon.abs seen aux.3pl.abs.3sg.erg ‘Nekane saw Miren and Jon’ (b) Miren eta Jon etorri dira. Miren.abs and Jon.abs come aux.3pl.abs ‘Miren and Jon came.’

The analysis of ergative-absolutive languages has given rise to a huge output of research (Bittner and Hale, 1996; Wiltschko, 2006; Johns, 1992; Legate, 2008; ?; Coon, 2013; Baker and Bobaljik,

12Note, however, that Alexiadou et al. (2015) do propose that unmarked inchoatives lack the VoiP projection. 6.5. PROPERTIES OF A-MOVEMENT 111

nominative-accusative ergative-absolutive

S S NOM ABS ACC ERG

A P A P

Figure 6.1: Case Alignment Systems

2017; Coon and Preminger, 2017; Marantz, 1984). For a typological overview see Dixon (1994). For a survey of generative approaches to ergativity see Aldridge (2008). The difficulty in analyzing ergativity resides in the fact that the Case of the highest argument varies. Assuming the subject appears or receives Case from T, in transitive verbs, T must assign ergative Case, and in intransitive verbs, T must assign absolutive Case. This is nothing more than a stipulation of the facts, and such a proposal has never been seriously entertained. Marantz (1984, 1991) proposed a novel and significantly different theory of Case assignment, which has gained much ground. Specifically, the Dependent Theory of Case works as follows. Languages parametrically choose between how dependent Case is assigned. In nominative-accusative languages, dependent Case is assigned to the lower argument. In ergative-absolutive languages, dependent Case is assigned to the higher argument.

(6.45)1 Nominative-Accusative Languages: (a) If there is one KP in a Case domain, it receives nominative Case (b) If there are two KPs in a Case domain, the lower one receives dependent Case: accusative 2 Ergative-Absolutive Languages: (a) If there is one KP in a Case domain, it receives absolutive Case (b) If there are two KPs in a Case domain, the higher one receives dependent Case: ergative

6.5 Properties of A-Movement

A-movement movement is typically characterized as being clause bound and lacking reconstruc- tion effects (Baltin, 2001; Boeckx, 2001). While wh-movement can rise any number of clauses, 112 CHAPTER 6. CASE AND A-MOVEMENT

A-movement (passive, raising to subject, raising to object) is either clause-bound, or can raise only out of non-finite clauses.

(6.46) (a) Which booki did John say that Mary thinks that Bill told Fred to read ti? (b) The book was read. (c) *The book was said that John read.

Chomsky (1993) and Lasnik (1999) argue that A-movement cannot undergo reconstruction based on the following data.

(6.47) (a) It seems everyone isn’t here, yet. (b) Everyone seems not to be here, yet.

In (6.47 a) the universal quantifier everyone can take high or low scope, while in (6.47 b) everyone can only be interpreted with high scope. See ?, however, for an in depth discussion regarding the possibility of reconstruction with A-movement. ?, for instance, observes that reconstruction is possible with some forms of A-movement.

(6.48) (a) Someone seems to be sick. (b) Someone seems sick.

(6.48 a) the existential quantifier someone can take high or low scope, while in (6.48 b) someone can take high scope only. 6.5. PROPERTIES OF A-MOVEMENT 113

Further Reading

• Alexiadou et al. (2015) - This monograph presents a contemporary analysis of unaccusativ- ity and the causative/inchoative alternation.

• ? - This monograph presents a bold alternative to the general view presented here. Ramchand proposes that verbs are composed of several functional projections that give rise to the various properties of verbs, only some of which have been described in this chapter.

• ? - This monograph gives a detailed descriptive account of Case, in terms of both morphology and syntax. 114 CHAPTER 6. CASE AND A-MOVEMENT Chapter 7

A-Bar Movement

7.1 Introduction

In the previous chapter we examine situations where a phrase was displaced from it usual position, as in the following pair.

(7.1) (a) Mary ate the apple. (b) The apple was eaten.

In the passive sentence the object the apple no longer appears to the right of the verb, where we usually expect to find objects in English. Rather, it appears in the usual subject position. This displacement is carried out by the operation Internal Merge (aka Move). In this chapter we examine cases of displacement over a much longer stretch. Consider the following examples.

(7.2) (a) Which book does John think that Mary read? (b) Even this book Susan thinks that Bill should read.

In both of these examples, the phrases which book and even this book are the thematic direct object of read. In both cases, the phrases appear significantly displaced from their thematic positions. This is a different kind of movement that we will examine in detail in this chapter. We will see that both of these examples arise by the same kind of movement, which we call A-Bar Movement. We begin with a discussion on the constraints of A-Bar movement.

7.2 Constraints on A-Bar Movement

It is well known that wh-movement is not entirely free, but is subject to various island constraints (Ross, 1967; Chomsky, 1977,?; Huang, 1982). It was Ross who also noted several kinds of movement (later known as A-bar movement were all subject to the same kinds of constraints. I present a

115 116 CHAPTER 7. A-BAR MOVEMENT partial list of constraints on wh-movement that have come to be known since the time of Ross, with illustrations from English below. Each constraint is illustrated with wh-movement, constructions, focus movement, and topicalization.

1 Adjunct Islands

2 wh-Islands

3 Coordinate Structure Constraint

4 Complex NP Constraint

5 Sentential Subject Constraint

6 Factive Islands

(7.3) Adjunct Islands

(a) John left because Bill ate the last piece of cake. (b) *What did John leave because Bill ate? (c) *the piece of cake that John left because Bill ate. (d) *Only the last piece of cake John left because Bill ate. (e) *The last piece of cake, John left because Bill ate.

In these examples there is an adjunct because John ate the last piece of cake. Any attempt to move something out of this adjunct results in ungrammaticality. The following examples all show the same pattern.

(7.4) wh-Islands

(a) Mary knows when Susan ate the apple.

(b) *Whati does Mary know [when Susan ate ti]? (c) *the apple that Mary knows when Susan ate (d) *Even this apple Mary knows when Susan ate. (e) *The apple, Mary knows when Susan ate.

(7.5) Coordinate Structure Constraint

(a) Pat ate an apple and an orange. (b) *What did Pat eat an apple and? (c) *the orange that Pat ate an apple and 7.3. THE MECHANISM OF A-BAR MOVEMENT 117

(d) *Even an orange Pat ate an apple and! (e) The orange, Pat ate an apple and.

(7.6) Complex NP Constraint

(a) Pat read the book that Mary wrote on her sabbatical.

(b) *Wheni did Pat read [the book that Mary wrote ti]? (c) *the sabbatical that Pat read the book that Mary wrote on (d) *Only on her sabbatical did Pat read [the book that Mary wrote t ] (e) *Her sabbatical Pat read the book that Mary wrote on t.

(7.7) Sentential Subject Constraint

(a) That Bill ate the last piece of cake angered John. (b) *What did that Bill ate anger John? (c) *the cake [that [that Bill ate t] angered John] (d) *Even the last piece of cake that that Bill ate angered John. (e) *The last piece of cake [that Bill ate t] angered John.

(7.8) Factive Islands

(a) Pat knows that Fred ate an apple. (b) *What does Pat know that Fred ate? (c) *the apple that Pat knows that Fred ate (d) ?Even an apple Pat knows that Fred ate. (e) ?That apple, Pat knows that Fred ate.

The important contribution from Ross’ work is that all forms of A-bar movement show rather uniform island effects.

7.3 The Mechanism of A-Bar Movement

Our current understanding of wh-movement and other related A-Bar movements takes its current shape from Chomsky (1977). Chomsky argued that wh-movement freely targeted any wh-phrase to move to SpecCP (then S-bar) as long as it obeyed certain conditions including the following.

(7.9) (a) Subjacency 118 CHAPTER 7. A-BAR MOVEMENT

(b) The Propositional Island Condition (c) The Specified Subject Condition

The Propositional Island Condition was meant to rule out the violations in (7.8), and the Spec- ified Subject Condition was meant to rule out the violations in (7.7). Subjacency was meant to explain the observation that A-bar movement proceeds in steps rather than in one fell swoop, as discussed in Chapter 5. The Subjacency principle states that movement cannot cross two bounding nodes at once.1

(7.10)*X i [α ... [β ... ti ... ] ... ], where α and β are bounding nodes

In Chomsky’s original proposal the bounding nodes were NP and S–KP and TP in updated terms. Let’s reconsider an example of long-distance wh-movement in light of Subjacency.

(7.11) (a) Whati did [TP Mary say [CP ti that [TP John bought ti ] ] ]?

(b) Whati did [TP Mary say [CP when [TP John bought ti ] ] ]?

Observe that in (7.11 a) the wh-phrase crosses only one TP at a time as it stops in SpecCP on its way to its surface position. In (7.11 b), however, the presence of the wh-phrase in the intermediate SpecCP forces the wh-phrase what to cross two TP nodes at once, in violation of Subjacency. Rizzi (1982) argues that the identity of the Subjacency nodes is subject to parametric variation. Specifically, based on the following Italian data he proposes that the binding nodes in Italian are NP and S-bar – KP and CP in updated terms.

(7.12) Tuo fratello, a cui mi domando che storie abbiano raccontato era your brother, to whom me I.ask what stories they.have told was molto preoccupato very worried ‘Your brother, who I wonder what stories they told to (him) was very worried.’

In this example, the wh-phrase a cui (‘to whom’) forms the operator of the relative clause that modifies fratello (‘brother’). Notice that it crosses an intermediate wh-phrase che storie (‘which stories’). Here is the structure of the relative clause.

(7.13) [CP a cuii [TP mi domando [CP che storie [TP abbiano raccontato ti ]]]]

Observe that the wh-phrase crosses only one CP, that is, crosses only one bounding node in this structure. Although Subjacency no longer plays a large role in syntactic theorizing today, the

1This is actually a simplication of Chomsky’s original proposal for Subjacency. He originally intended to account for a much broader range of phenomena beyond A-bar movement. 7.3. THE MECHANISM OF A-BAR MOVEMENT 119 original observations of the time still form the foundation of how we conceptualize wh-movement today. The analysis of wh-movement did not change much during during the GB era, except in response to changes in the overall architecture of grammar. Notable contributions include Condition on Extraction Domains (Huang, 1982) and the Clause Typing Hypothesis (Cheng, 1991). A general change in the understanding of movement took place in the transtion from GB to Minimalism. Specifically, movement in GB Theory was thought to be free, handled by a single rule Move α. Movement in Minimalism is thought to be costly, and takes place only to remove an uninterpretable feature. As the following trees show wh-movement takes place overtly to satisfy the uninterpretable [uwh] feature on the interrogative C head. Like T in English, we can assume that the interrogative C head has an EPP property that requires its specifier to be filled.

(7.14) (a) CP

Cuwh TP

... KP ...

Kwh DP

what

(b) CP

0 KPi C

Kwh DP Cuwh TP

what ... ti...

This creates a problem for long-distance wh-movement, however. While the matrix interrogative C has the EPP property that requires it to be filled, what forces movement through the intermediate SpecCPs?

(7.15) What does John think [CP t that Mary bought t?]

Recent discussions in Labelling Theory have addressed this question. Previously, representational constraints on the grammar required the existence of the intermediate traces at LF, hence requiring look ahead (Aoun and Li, 2003). In other words, the wh-phrase would have to know that later on at LF it needs to move through the intermediate SpecCP, so it does so in the overt syntax. All things 120 CHAPTER 7. A-BAR MOVEMENT being equal, we don’t wish to endow the grammar with the power to look ahead into what might happen later in the derivation (Collins, 1997b). Recall the mechanism of Labelling Theory from Chapter 5. If two XPs undergo Merge, the label is formed by the criterial feature that they share. When a wh-phrase merges with a C’ that has an [uwh] feature, they share a criterial feature, namely [wh]. This feature forms the label of the CP (though we continue to write CP for convenience).

(7.16) CP[wh]

0 KP[wh]i C[wh]

what Cuwh TP

... ti...

When the wh-phrase merges with the intermediate CP, it shares no criterial feature. Under the Labelling Algorithm, no label can be created.

(7.17) ∅

0 KP[wh]i C

what Cdecl TP

... ti...

Chomsky argues that an unlabelled structure cannot be tolerated at LF. Either KP or C’ must raise. In this case, KP raises, leaving C’ behind to label the resultant structure as CP.2 See Ott (2015); Moro (2000, 2009) for a development of the idea the unlabelled structures force movement. Note that we still have not addressed the reason for the initial movement of the wh-phrase out of its original position. We suggested above that the wh-phrase moves to satisfy the EPP property of the interrogative C head; however, that head has not yet been introduced into the derivation at the intermediate CP stage. We will come back to the impetus for wh-movement later, however. Until now, we have treated wh-movement uniformly. As far back as a distinction between strong and weak islands was made. Consider the following data from Huang (1982) and Rizzi (1990).

(7.18) (a) ??[Which problem]i do you wonder how John could solve ti?

(b) *[Which student]i do you wonder how ti could solve the problem?

(c) *Howi do you wonder which problem John could solve ti?

2Recall crucially that the nodes C’ and CP are notational conveniences to make the tree easier to read. The acutal label is simply C, or, more precisely, the feature set indicated by C. 7.4. WH-IN-SITU 121

(d) [Which problem]i do you wonder how to solve ti?

(e) *Howi do you wonder which problem to solve ti?

In addition to cyclicity effects afforded by phase theory, Rizzi (1990) argues that Minimality effects are responsible for many of the island effects discussed above (see also Manzini, 1992). Although the precise mechanism of minimality is beyond the scope of this discussion, the basic premise can be described here. Consider the following scenario.

(7.19) X ... Y ... Z

Minimality states that X cannot govern Z if there is a closer potential governor, Y, that could govern Z. Potential governors in the current context include the following.3

• wh-phrase

• theta-assigner

Consider examples (7.18 a) and (7.18 d). Notice that although these two sentences violate subja- cency, they do not violate minimality. In both cases, the trace of movement is c-commanded by the head that assigns it a theta-role, namely the V head. In (7.18 b), (7.18 c), and (7.18 e), on the other hand, the moved wh-phrases violate subjacency. Furthermore, they do not have a theta-role, so they also violate minimality. This notion of minimality is captured under the Prin- ciple (ECP) (Chomsky, 1981a). As it stands, there is no current explanation within Minimalism for the distinction between strong and weak islands, the ECP being the most recent explanation.4 For a pedagogical explanation of these facts within a GB framework, the reader is advised to consult Haegeman (1991). For a brief discussion of movement out of wh-infinitival clauses see Barrie (2007).

7.4 Wh-in-situ

Huang (1982) was one of the earliest treatments of wh in-situ phenomena. The following Mandarin example illustrates wh in-situ. Observe that the object wh-phrase appears in the same position as non-wh objects.

(7.20) Mandarin wh in-situ

(a) Zhangsan chi li pingguo Zhangsan eat perf apple ‘Zhangsan ate an apple.’ 3Note that in order for Y to govern Z, Y must c-command Z in addition to meeting other requirements as mentioned in the bullet points. The precise definition of government changed throughout the GB era before it was decided to abandon the notion. 4Thanks to Victor Pan for pointing this out to me. 122 CHAPTER 7. A-BAR MOVEMENT

(b) Zhangsan chi li shenme Zhangsan eat perf what ‘What did Zhangsan eat?’

Observe further that wh in-situ is also found in multiple wh-questions in English.

(7.21) Who bought what?

The core of the problem is illustrated by the following examples.

(7.22) (a) Who wonders who bought what? (b) ni xiang-zhidao [ wo weishenme mai shenme ]? you wonder I why buy what ‘What do you wonder why I bought (it)?’

First, note that (7.22 a) is ambiguous (Baker, 1968). This ambiguity has come to be known as Baker’s Ambiguity. It has the two readings as follows, along with a possible answer to illustrate the two readings.

(7.23) Baker’s Ambiguity - First Reading

(a) Who is the x such that x wonders who the y is and what the z is such that y bought z? (b) more simply: Who is the x such that x wonders who bought what? (c) answer: John wonders who bought what.

(7.24) Baker’s Ambiguity - Second Reading

(a) Who is the x and what is the y such that x wonders who the z is such that z bought y? (b) more simply: Who is the x and what is the y such that x wonders who bought y? (c) answer: John wonders who bought the apples; Mary wonders who bought the oranges; and Sally wonders who bought the cherries.

Observe that for the second reading the embedded what is interpreted in the matrix SpecCP at LF. Likewise for the Mandarin example, what is interpreted in the matrix clause. If we assume that the wh-phrase raises to the matrix SpecCP at LF it crosses another wh-phrase. These observations led the idea that only overt movement is sensitive to Subjacency. However, wh-in-situ does not behave uniformly with respect to island sensitivity. Note that only one reading is available for (7.22 b). Crucially, of the two logically possible readings, only the first one below is possible.

(7.25) (a) What do you wonder why I bought (it)? 7.4. WH-IN-SITU 123

(b) *Why do you wonder what I bought (for that reason)?

Thus, an argument/adjunct asymmetry seems to hold for wh-in-situ (see also Lin, 1992; Soh, 2005; Tsai, 1994). This required a more fine-grained approach to wh-in-situ. Specifically, it was proposed that wh-arguments are licensed by unselective binding (Pesetsky, 1987), and that wh-adjuncts undergo LF movement. Thus, LF movement is always subject to Subjacency. Wh- arguments in Chinese, then, do not move at LF, but are licensed by unselective binding. To understand unselective binding, consider the following English sentence.

(7.26) Mary only gave John a book.

The operator only can bind a focus variable in either John, gave, or in a book. Unselective binding was argued to account for at least some kinds of wh-in-situ. Consider the basic wh-in-situ question in Mandarin.

(7.27) Lisi chi-le shenme? Lisi eat-prfv what ‘What did Lisi eat?’

Let’s consider a very basic structure to understand the concept of unselective binding. In the following tree the dashed line indicates that the [uwh] feature on C looks down to unselectively bind a wh variable that c-commands.

(7.28) Unselective Binding in Mandarin Chinese

CP

Cuwh TP

KP T0

Lisi T VP

V KPwh

chi-le shenme 124 CHAPTER 7. A-BAR MOVEMENT

unselective binding LF movement Subjacency insensitive obeys English wh-in-situ – Mandarin wh-arguments wh-adjuncts

Table 7.1: Typlogy of wh-in-situ, to be revised

The idea is that unselective binding is not sensitive to Subjacency as no movement takes place. LF movement of wh-adjuncts, however, is sensitive to Subjacency (just as overt wh-movement is). We now have an explanation of the set of facts we have seen so far. Table 7.1 shows that in Chinese wh-arguments are licensed by unselective binding, so show no Subjacency effects. Chinese wh-adjuncts, however, undergo LF movement and so are subject to Subjacency. Note however that Japanese differs from Chinese in terms of island effects. Consider the following data (Watanabe, 2001). Observe that the in-situ wh-phrase inside a relative clause is fine, but that the wh-island is degraded (as in English). These facts will require a revision to the findings summarized in Table 7.1.

(7.29) wh-in-situ in Japanese

(a) kare-wa [dare-ga kaita] hon-o yonde-iru no he-top who-nom wrote book-acc read-prog q ‘Who is the x such that he is reading the book that x wrote?’ (b) ??[nani-o doko-de kata ka] oboete-iru no what-acc where-at bought q remember-prog q ‘(What do you remember where we bought (it)?)’

In Japanese, when a wh-phrase is contained in a relative clause, Watanabe (1992) proposed that the whole relative clause is pied-piped at LF, thus, no Subjacency violation takes hold. In the ungrammatical example (7.29 b) the wh-phrase nani-o (‘what’) cannot be interpreted in the matrix clause. Notice that we cannot relate this difference to an argument/adjunct asymmetry as we did for Chinese since in both cases we have only wh-arguments. Watanabe (1992) proposed that Japanese differs from Chinese the following way. Wh-movement in Japanese involves overt movement of a wh-operator.

(7.30) [CP Opi [CP [DP ti nani-o ] ] ]

Thus, Watanabe argued that only overt wh-movement is sensitive to Subjacency. His evidence is multiple wh-questions, as discussed for English above. Consider the following minimal pair. Partial trees showing Watanabe’s analysis are shown below. 7.4. WH-IN-SITU 125

(7.31) Multiple wh-questions in Japanese (Watanabe, 1992)

(a) ??John-wa [Mary-ga nani-o katta kadooka] Tom-ni tazuneta no John-top Mary-nom what-acc bought whether Tom-dat asked q ‘ (What did John ask Tom whether Mary bought (it)?)’ (b) John-wa [Mary-ga nani-o katta kadooka] dare-ni tazuneta no John-top Mary-nom what-acc bought whether who-dat asked q ‘Who did John ask whether Mary bought what?’

(7.32) Subjacency violating overt movement of operator

CP

0 Opi C

TP C

KP T0 no

John-wa VP T

CP V0

TP Cwh KP V

KP T0 kadooka Tom-ni tazuneta

Mary-ga VP T

KP V

ti KP

nani-o

In this structure the operator moves overtly to the matrix SpecCP crossing the intermediate wh C head. This operator crosses two bounding nodes (TP) giving rise to ungrammaticality. Consider now the structure for the multiple wh-question. 126 CHAPTER 7. A-BAR MOVEMENT

(7.33) Subjacency insensitive unselective binding

CP

Op C0

TP C

KP T0 no

John-wa VP T

CP V0

TP Cwh KP V

0 KP T kadooka ti KP tazuneta

Mary-ga VP T dare-ni

KP V

tj KP

nani-o

In this tree the higher operator moves overtly, but does not cross any intervening wh-phrase. The lower operator moves at LF to the matrix clause. Under Watanabe’s proposal LF movement is not subject to Subjacency, so the derivation converges. Watanabe argues that this mirrors the English facts, so the same analysis covers both the Japanese and the English data. What is left out, however, is the difference between wh-arguments and wh-adjuncts in Chinese; however, Watanabe’s analysis could be amended to account for it. Given these two lines of arguments, we now have to families of analyses to account for wh-in-situ. As shown in Table 7.2, the crucial issue is whether LF movement is subject to Subjacency or not. This issue is essentially left open to future research, which can be answered by studying a wider array wh-in-situ languages around the world. 7.5. MULTIPLE WH MOVEMENT 127

Watanabe Huang wh-in-situ sensitive to wh-islands Japanese overt Op movement LF movement Mandarin adjuncts wh-in-situ immune to wh-islands Japanese multiple wh LF movement unselective binding Mandarin arguments English multiple wh

Table 7.2: Current analyses of wh-in-situ

7.5 Multiple Wh Movement

In English a multiple wh-question requires movement of the highest wh-phrase to SpecCP.

(7.34) (a) Who bought what? (b) *What did who buy?

In some languages, multiple wh-questions involves movement of more than one wh-phrase to the sentence-initial position. This is referred to as multiple wh-movement. Here are some examples (Rudin, 1988).

(7.35) Multiple wh-movement

(a) Koj kogo vižda? who whom sees ‘Who sees whom?’ Bulgarian (b) Cine cu ce merge? who with what goes ‘Who goes by what (means of transportation)?’ Romanian (c) Ko koga vidi? who whom sees ‘Who sees whom?’ Serbo-Croatian

Rudin (1988) notes that there are two types of multiple wh-fronting languages, which she la- bels multiply-filled SpecCP [+MFS] languages (Bulgarian and Romanian), and non-multiply-filled SpecCP [-MFS] languages (Serbo-Croatian, Polish, and Czech). These two types of languages differ on a number of properties. The first property we discuss is multiple long-distance movement. In [+MFS] languages, all wh- phrases must move to the SpecCP where they take scope. Observe in the following example that 128 CHAPTER 7. A-BAR MOVEMENT both wh-phrases must appear in the matrix SpecCP. Note that Rudin shows a much fuller range of data to substantiate her arguments.

(7.36) Multiple long-distance wh-movement in Bulgarian (Rudin, 1988, p.450)

(a) Koji k˘udej misliš [če e otiš˘ul ti tj ]? who where think.2sg that has gone ‘Who do you think (that) went where?’

(b) *Koji misliš [če e otiš˘ul ti k˘ude ]? who think.2sg that has gone where ‘(Who do you think (that) went where?)’

In [-MFS] languages, however, only one wh-phrase may undergo long-distance wh-movement.5 Rudin shows additional data from Polish and Czech that show the same pattern.

(7.37) Lack of long-distance wh-movement in Serbo-Croatian (Rudin, 1988, p.453f )

(a) Koi želite [da vam šta kupi ti ]? who want.2sg to you what buy.3.sg ‘Who do you want to buy you what?’

(b) *Koi štaj želite [da vam tj kupi ti ]? who what want.2sg to you buy.3.sg ‘(Who do you want to buy you what?)’ wh Rudin considered other differences between [+MFS] and [-MFS] languages; however, we will consider only one further difference. In [+MFS] languages the order of the wh-phrases in multiple wh-questions is relatively fixed. This fixed order is referred to as superiority and is illustrated in the following English examples.6 Observe that when there are two wh-phrases, the higher of the two must raise.

(7.38) Superiority in English

(a) Who bought what? (b) *What did who buy?

Consider the following Bulgarian examples (Rudin, 1988, p.472f ). Observe that the wh-phrases obey superiority. Rudin notes that no other word order is possible for (7.39 c).7

5Rudin notes that some speakers do accept multiple wh-movement, but that all speakers accept leaving one wh-phrase in-situ, unlike in [+MFS] languages such as Bulgarian. 6Superiority can be violated in English D-linked wh-phrases: Compare Which student read which book? and Which book did which student read? 7Pesetsky (2000) and Richards (2001) show that there is some flexibility in the order of wh-phrases when there are three or more. In all cases, however, the highest wh-phrase must obey superiority. 7.5. MULTIPLE WH MOVEMENT 129

(7.39) Superiority obeyed in Bulgarian

(a) Koj kogo vižda? who whom sees ‘Who sees whom?’ (b) *Kogo koj vižda? whom who sees ‘(Who sees whom?)’ (c) Koj kogo na kogo e pokazal? who whom to whom has pointed out? ‘Who pointed out who to whom?’

Consider now the following Serbo-Croatian data (Rudin, 1988, p.473). Observe that the order of the wh-phrases is free.

(7.40) Superiority not obeyed in Serbo-Croatian

(a) Ko koga vidi? who whom sees ‘Who sees whom?’ (b) Koga ko vidi? whom who sees ‘Who sees whom?’ (c) Ko je što kome dao? who has what to whom given ‘Who gave what to whom?’ (d) Ko je kome što dao? who has to whom what given ‘Who gave what to whom?’ (e) Etc. All six orders possible.

Rudin proposed that wh-phrases in some languages move to SpecCP, but in other languages adjoin to TP. She proposes that in [+MFS] languages SpecCP can host more than one wh-phrase. The highest wh-phrase raises to SpecCP first, followed by the lower wh-phrases. This process came to be known as tucking-in (Freidin, 1999; Richards, 2001). Here is a simplified structure for (7.39 a). 130 CHAPTER 7. A-BAR MOVEMENT

(7.41) CP

KPi

0 koj KPj C

kogo C TP

0 ti T

T VP

V tj

vižda

Rudin’s structure for multiple wh-quesitons in [-MFS] languages involves only one wh-phrase in SpecCP. The other wh-phrases adjoin to SpecTP. There is, however, no question of superiority since the wh-phrases are raising to different positions. Either one may raise to SpecCP since they are not in competition. Here is a simplified structure for (7.40 a).

(7.42) CP

0 KPi C

ko C TP

KPj TP

koga ti VP

V tj

vidi

We can now explain the long-distance movement properties from above. Recall that only [+MFS] languages allow multiple long-distance movement. Since only [+MFS] languages allow multiple specifiers only [+MFS] languages will allow more than one wh-phrase to escape the clause. [-MFS] languages allow only one SpecCP. Under the assumption that a single specifier cannot be re-used by other wh-phrases, this means that the other wh-phrases in [-MFS] languages must remain in the lower clause. Here is a schematic showing the difference between the two language types. 7.6. PARTIAL MOVEMENT AND SCOPE MARKING CONSTRUCTIONS 131

[+MFS] languages [CP wh1 wh2 wh3 ... [CP t1 t2 t3 C[TP ... t1 t2 t3 ]]]

[-MFS] languages [CP wh1 ... [CP t1 [TP wh2 wh3 ... t1 t2 t3 ]]]

See the suggested reading section at the end for further developments of multiple wh-movement.

7.6 Partial Movement and Scope Marking Constructions

In the generative literature partial movement and scope marking constructions were first discussed by van Riemsdijk (1982) and McDaniel (1989). Here is a German example followed by a Romani example. The scope marker is typically the equivalent of the word ‘what’ and is glossed as what here for clarity.8

(7.43) Scope Marking Constructions (McDaniel, 1989, p.569)

(a) Mit wemi glaubt Hans ti Jakob jetzt ti spricht? with whom believes Hans Jakob now speaks ‘With whom does Hans believe Jakob is now speaking?’

(b) Was glaubt Hans mit wemi Jakob jetzt ti spricht? what believes Hans with whom Jakob now speaks ‘With whom does Hans believe Jakob is now speaking?’ [German]

(c) Kasi o Demìri mislinol ti so i Arìfa dikhˇva ti? whom det Demir thinks that det Arifa saw ‘Who does Demir think that Arifa saw?’

(d) So o Demìri mislinol Kasi i Arìfa dikhˇva ti? what det Demir thinks whom that det Arifa saw ‘Who does Demir think that Arifa saw?’ [Romani]

When there are more than two clauses separating the interrogative C and the wh-phrase, the wh-phrase can appear in any of the intermediate SpecCP positions (or in the SpecCP where it takes scope). Consider the following data.

(7.44) Long distance scope marking constructions in German (McDaniel, 1989, p.575)

(a) Mit wemi glaubst du [CP ti dass Hans meint [CP ti dass Jakob ti with whom believe you that Hans thinks that Jakob gesprochen hat]]? spoken has ‘With whom do you believe that Hans thinks that Jakob talked?’ 8As McDaniel notes, there is a great deal of dialect variation in German with respect to the availability of long- distance movement and scope marking constructions. 132 CHAPTER 7. A-BAR MOVEMENT

(b) Was glaubst du [CP mit wemi dass Hans meint [CP ti dass Jakob ti what believe you with whom that Hans thinks that Jakob gesprochen hat]]? spoken has ‘With whom do you believe that Hans thinks that Jakob talked?’

(c) Was glaubst du [CP was dass Hans meint [CP mit wemi dass Jakob what believe you what that Hans thinks with whom that Jakob ti gesprochen hat]]? spoken has ‘With whom do you believe that Hans thinks that Jakob talked?’

McDaniel also shows that scope marking constructions are sensitive to Subjacency. Consider the following data, which include wh-islands. Note crucially that the wh-phrase that is co-indexed with the scope marker has not moved out of the wh-island. Thus no movement has taken place that violates Subjacency.

(7.45) Scope marking constructions and wh-islands (McDaniel, 1989, p.576f )

1 1 (a) *Was fragt sie sich [CP warumj Hans tj glaubt [CP weni Jakob ti what asks she self why Hans thinks whom Jakob gelobt hat]]? praised has 1 1 ‘(Who does she wonder why Hans thinks Jakob praised (them )?)’ [German] 1 1 (b) *So na iane [CP sosqej o Demiri mislinol tj [CP kasi marjum what neg know why det Demir thinks whom I hit ti]]?

1 1 ‘(Who don’t you know why Demir thinks I hit (them )?)’ [Romani]

McDaniel discusses several other properties of scope marking constructions; however, we will consider only the data shown to this point. She argued that the wh-phrase undergoes usual A-bar movement to an intermediate SpecCP. This wh-phrase forms a chain with the scope marker, which is used to indicate where the wh-phrase is interpreted at LF. This is the Direct Dependency Analysis. McDaniel argues on the basis of data as in (7.45) that Subjacency is not a condition on movement, but is rather a condition on representations. Observe that the relation between the scope marker and the wh-phrase with which it is co-indexed crosses another wh-phrase (and hence crosses two bounding nodes). In modern terms, this has given rise to the Minimal Match Condition (Aoun and Li, 2003). We can represent McDaniel’s analysis as follows.

(7.46) Direct Dependency Analysis of scope marking constructions 7.6. PARTIAL MOVEMENT AND SCOPE MARKING CONSTRUCTIONS 133

CP

what1 C0

C TP

T VP

V CP

1 0 KP i C

wh-phrase C TP

T VP

ti

This discussion has barely scratched the surface on the cross-linguistic diversity and the range of analyses of scope marking constructions. Aside from the analysis presented here - the Direct Dependency Analysis, Dayal (1994) proposed the Indirect Dependency Analysis on the basis of scope marking constructions in , which we will not cover here. 134 CHAPTER 7. A-BAR MOVEMENT

Further Reading

• Huang (1982) - This dissertation is one of the earliest generative analyses of wh-movement in a wh in-situ language, Mandarin Chinese.

• Cable (2010) - an extended discussion of numerous cross-linguistic differences in wh-movement, including under studied languages

• Richards (2016) - a radically different idea suggesting that wh-movement takes place to satisfy prosodic constraints of the language

• Pesetsky (2000) - a complex look at wh-movement, which distinguishes LF XP movement from LF feature movement

• Richards (2001) - an in depth look at the differences between wh-in-situ and multiple wh- movement Chapter 8

Raising and Control

8.1 Introduction

In the complex sentences examined so far there is no obligatory co-reference between the matrix subject and the embedded subject. Consider the following examples.

(8.1) (a) John1 thinks that he1 should enter the race. (b) John thinks that Mary should enter the race. (c) 영희는1 pro1 통사론을 공부해야 되는지 궁긍하다. Yenghuy-nun1 pro1 thongsalon-ul kongbwu-ha-y-ya doy-nunci kwunkum-ha-ta Younghui-top pro syntax-acc study-do-inf-must become-q wonder-do-decl ‘Younghui wonders if she should study syntax.’ (d) 영희는 민수가 통사론을 공부해야 되는지 Yenghuy-nun Minswu-ka thongsalon-ul kongbwu-ha-y-ya doy-nunci Younghui-top Minsoo-nom syntax-acc study-do-inf-must become-q 궁긍하다. kwunkum-ha-ta wonder-do-decl ‘Younghui wonders if Minsoo should study syntax.’

Observe that while the embedded subject may be coreferential with the matrix subject, it does not have to be. Consider, now, the following English examples.

(8.2) (a) Mary decided to enter the race. (b) Fred seems to like chocolate.

In both of these examples we understand the matrix subject also to be the thematic or under- stood subject of the infinitive. We will come up with more precise definitions of raising and control below, but we can think of them as situations in which the matrix and embedded subjects must be coreferential.

135 136 CHAPTER 8. RAISING AND CONTROL

In English, virtually all raising and control constructions involve the use of non-finite verbs. Non-finite clauses require us to revise our understanding of clause structure. Let’s review some properties of non-finite clauses first. Observe that tense and agreement are both absent in non-finite clauses.

(8.3) (a) Petra managed to fix the toaster. (b) Travis tried to cook an omelette. (c) Gerry told Alice to wait for five minutes.

(8.4) (a) Petra fixed the toaster. (b) Travis cooked an omelette. (c) Gerry expected that Alice would wait for five minutes.

Looking just at the first example, we see that the verb manage appears with past tense mor- phology. The embedded verb appears with what we call an infinitival marker, to. We observe that infinitives cannot appear with tense morphology.

(8.5) *Petra managed to fixed the toaster.

Furthermore, agreement morphology cannot appear on the infinitive, either. Although English exhibits extremely impoverished verbal agreement, the third person singular form appears with the agreement marker –s, as shown on the matrix verb in the following example, but cannot appear on the embedded infinitive.

(8.6) *Petra manages to fixes five toasters a day.

What we observe is that tense and agreement features are in complementary distribution with the infinitive marker to. We suggest, then, that since to marks the lack of tense and agreement, it is the head of T and has the feature [-tense]. Recall also that Case assignment in non-finite clauses differs from Case assignment in finite clauses. The differences are morphologically visible.

(8.7) (a) She fixed the toaster. (b) I expect her to fix the toaster.

Let us consider some constituency tests to sharpen this idea further. Consider the following data.

(8.8) (a) John decided to [eat some spaghetti] and [drink some wine]. (b) Eat anchovies, John never would __, even though Ashleigh tried to __. 8.2. DIAGNOSTICS FOR RAISING AND CONTROL 137

The first example shows that eat some spaghetti and drink some wine are constituents, and in fact look like run-of-the-mill vP constituents. The second example confirms this. This is an example of vP fronting, which strands the auxiliary at the end of the sentence. In a non-finite clause, it strands the infinitival marker to. These facts fall into place under the assumption that to is a tenseless T head. Thus, we have the following partial structure so far. The subject position (specifier of vP) is left empty as it is the topic of the next discussion.

(8.9) TP

T vP

to Spec v 0

v VP

Vi v ti KP

eat some spagetti

We will now consider the structure of infinitivals in detail. If we examine the following two pairs of sentences, we may initially suspect that (??) and (8.10 c) are structurally isomorphic, as are (8.10 b) and (8.10 d). That is, we suspect that each pair has the same structure and that the only difference between them is the matrix verb in each case.

(8.10) (a) Mary wants John to enter the race. (b) Sally appears to live in a small apartment. (c) Mary told John to enter the race. (d) Sally decided to live in a small apartment.

8.2 Diagnostics for Raising and Control

One important similarity in each pair is that the understood subject of the infinitive is the same. In (8.10 a and (8.10 c, we understand John to be the one entering the race. In (8.10 b and (8.10 d, we understand Sally to be the one living in the small apartment. There are, however, several structural differences between these two pairs of sentences. Let’s examine these first, then try to derive the different properties structurally. The following diagnostics are rooted in research in the early 1970s (Perlmutter, 1978, 1970). 138 CHAPTER 8. RAISING AND CONTROL

8.2.1 Idiom Tests

Idioms are a useful diagnostic for many syntactic tests. It is important to note that idioms must appear as a unit upon initial merge. So, although they may be separated by a movement operation, they are still initially merged as a constituent to get the idiomatic reading. For our purposes, we require an idiom with a subject - these are not too common, but there are a few. The idiom test must be executed with care. What we are trying to see is how the understood subject is related to the infinitival clause, hence the need for idioms with subjects. Here are the steps to formulating a test sentence.

1 Identify the entity that is coreferential with the subject of the infinitival clause.

2 Identify the infinitival clause and the corresponding portion of an idiom. That is, the TP without the subject.

3 Change the verb of the idiom to an infinitive.

4 Replace the infinitival clause of the test sentence with the infinitival form of the idiom.

5 Replace the understood subject of the embedded infinitival clause with the subject of the idiom.

In the following examples, the highlighted portion illustrates the application of each step above.

(8.11) (a) Mary wants John to enter the race. (b) Mary wants John to enter the race. All hell broke loose. (c) ...to break loose (d) Mary wants John to enter the raceto break loose. (e) Mary wants Johnall hell to break loose. (f) Mary wants all hell to break loose.

To interpret the results of the test we see if the idiomatic reading is retained. Consider the following results. In (8), both the idiomatic and literal meanings are available. The first sentence, for example, could mean that Mary wants some secret to become known. In (9), however, the idiomatic meanings are not available.

(8.12) (a) Mary wants the cat to be out of the bag. (b) Mary wants the shit to hit the fan. (c) Mary wants all hell to break loose. (d) Mary wants tabs to be kept on John’s spending habits. 8.2. DIAGNOSTICS FOR RAISING AND CONTROL 139

(8.13) (a) *Mary told the cat to be out of the bag. (b) *Mary told the shit to hit the fan. (c) *Mary told all hell to break loose. (d) *Mary told tabs to be kept on John’s spending habits.

In order for the idiomatic reading to hold, the idiom must be merged as a constituent. This suggests that the cat is part of the embedded clause in (8.12 a, but part of the matrix clause in (8.13 a. More specifically, we propose that the DP that appears immediately after the verb want is merged as the subject of the infinitive, while the DP that appears immediately after the verb tell is merged in the matrix clause. This can be represented schematically as follows. (10)a. . . . want [DP to-infinitive] b. . . . told [DP] [to-infinitive] Before moving on to the next test, let’s test the other pair of sentences in (7). In these sentences, Sally is interpreted as the understood subject of the infinitive, so we replace Sally with the subject of the idiom. Here are the results. Note that a verb like appear usually requires an embedded state, so the infinitive has been changed to the perfect by the addition of have. Making changes in the aspect of the embedded clause does not affect the interpretation of the results.

(8.14) (a) The cat appears to be out of the bag. (b) The shit appears to (have) hit the fan. (c) All hell appears to have broken loose. (d) Tabs appear to have been kept on John’s spending habits.

(8.15) (a) *The cat decided to be out of the bag. (b) *The shit decided to hit the fan. (c) *All hell decided to break loose. (d) *Tabs decided to be kept on John’s spending habits.

These results are similar to the ones above. Thus, the subject of appear must originate in the embedded infinitival clause to ensure that the idiom is merged as a constituent, while the subject of decide is not.

(8.16) (a) appear [ DP to-infinitive] (b) DP decide [to-infinitive] 140 CHAPTER 8. RAISING AND CONTROL

8.2.2 Expletive Subjects

Before you read this section, you may wish to review the discussion on expletives on page 127. Recall crucially that expletives do not receive a θ-role. Consider, now, the following contrast. With want, the embedded subject can be an expletive, while with tell, it cannot.

(8.17) (a) John wants it to rain. (b) John wants there to be chocolate available during the break.

(8.18) (a) *John told it to rain. (b) *John told there to be chocolate available during the break.

Notice that an expletive is available as a subject in the infinitival clauses in the sentences in example (8.17, but not in (8.18, as long as the embedded predicate is consistent with an expletive subject. Thus, in the sentences in (8.17 it cannot be the case that the subject of the infinitive receives a θ-role from want, or the expletive would not be able to appear in this location. In (8.18, however, the fact that expletives are excluded from this position can be explained if we assume that the subject of the infinitive gets a θ-role from the matrix verb tell. Thus, we propose the following θ-role assignment relations for these two verbs.

(8.19) (a) ... want DP to-infinitive (DP receives external θ-role from infinitive) (b) ...tell DP to-infinitive (DP receives internal θ-role from tell

Again, we get similar results with appear and decide.

(8.20) (a) It appears to have rained. (b) There appears to be chocolate available during the break.

(8.21) (a) *It decided to rain. (b) *There decided to be chocolate available during the break.

8.2.3 Voice Transparency

This final test is based on the fact that active and passive sentences have the same set of truth conditions. Thus, the two sentences Mary ate the orange and The orange was eaten by Mary must either both be true or both be false. It is impossible that one is true while the other is false. With this background in mind, let us consider the following data.

(8.22) (a) The doctor examined the patient. = The patient was examined by the doctor. 8.2. DIAGNOSTICS FOR RAISING AND CONTROL 141

(b) Mary wants the doctor to examine the patient. = Mary wants the patient to be examined by the doctor. (c) Mary told the doctor to examine the patient. ≠ Mary told the patient to be examined by the doctor.

Example (8.22 a) is the test sentence that shows us that the active and passive sentences are synonymous. Under want in (8.22 b, the truth conditions do not change between the active and the passive versions; however, under tell in (8.22 c, the two sentences are not synonymous. This is also shown by the following data. The sentence in (8.23 a) is contradictory and the sentence in (8.23 b) means that Mary wants a contradictory situation to hold. The sentence in (8.23 c), however, is perfectly logical.

(8.23) (a) #The doctor examined the patient, but the patient wasn’t examined by the doctor. (b) #Mary wants the doctor to examine the patient, but Mary doesn’t want the patient to be examined by the doctor. (c) Mary told the doctor to examine the patient, but Mary didn’t tell the patient to be examined by the doctor.

Again, we similar results for appear and decide. There is a slight difference in meaning with appear; however, this contrasts with the stark difference in meaning with decide.

(8.24) (a) The doctor examined the patient. = The patient was examined by the doctor. (b) The doctor appears to have examined the patient. = The patient appears to have been examined by the doctor. (c) The doctor decided to examine the patient. ≠ The patient decided to be examined by the doctor.

We can make sense of these facts if the embedded subject under want is part of the embedded clause, but not the embedded subject under tell. In the next section, we will bring these facts together to understand the structure of non-finite clauses. For reasons that will become clear shortly, verbs such as decide and tell are called control predicates, and verbs such as want (as shown here) and appear are called ECM and raising predicates, respectively. The name ECM stands for Exceptional Case Marking, which is explained below.

8.2.4 Scope Reconstruction

(Landau, 2013, pp.12-13) offers the following scope contrasts to distinguish between raising and control. Consider the following examples. 142 CHAPTER 8. RAISING AND CONTROL

(8.25) (a) Seven civilians are likely to starve to death this weekend. (b) Seven civilians are afraid to starve to death this weekend. (c) Poirot proved at least two collaborators to have killed the duchess. (d) Poirot asked at least two collaborators to kill the duchess.

Landau observes that in (8.25 a) and (8.25 c) the quantified nominal can take scope either in the matrix clause or in the embedded clause. In (8.25 b) and (8.25 d), on the other hand, the quantified nominal can take scope only in the matrix clause. To be specific, the two readings of (8.25 a) are as follows.

(8.26) (a) There are seven civilians, and it is likely that they will starve to death. (7 > likely) (b) It is likely that seven civilians will starve to death. (likely > 7)

The same two readings can be found with (8.25 c).

8.3 The Analysis of Raising and Control

8.3.1 GB Analysis of Raising and Control

The traditional analysis of raising and control has undergone considerable changes since Rosen- baum (1967). The notion of PRO was introduced as early as Chomsky and Lasnik (1977), but was developed in Chomsky (1981a). PRO is a null anaphoric pronoun that takes an antecedent in the super-ordinate, finite clause. Typically, the controller is the closest c-commanding KP (first noted by Rosenbaum, 1967), but promise and similar verbs don’t conform to this generalization.1

(8.27) (a) Mary1 decided PRO1 to eat an apple. (b) Mary persuaded Bill1 PRO1 to eat an apple. (c) Mary1 promised Bill PRO1 to eat an apple.

Turning to the structure of the infinitival clause, the standard analysis for control involves a CP infinitival clause and a PRO subject. The standard analysis of raising involves a TP infinitival clause. There is some evidence for this bifurcation in wh-infinitivals. Some control verbs can take a wh-infinitival clause as a complement, meaning the infinitival clause must be a CP, (8.28 a-8.28 b). Raising verbs, however, never take a wh-infinitival complement clause, (8.28 c). Nevertheless, not all control verbs can take a wh-infinitival complement clause, (8.28 d). Consider the following examples.

(8.28) (a) John decided where to eat dinner.

1See Larson (1991) for an attempt at explaining the peculiar properties of promise. 8.3. THE ANALYSIS OF RAISING AND CONTROL 143

(b) John told Bill what to eat. (c) *John expects Bill what to eat. (d) *John wants what to eat.

Here, then, are the standard structures, which can be found in most syntax textbooks. Note that I have not shown the VoiceP layer or object movement as discussed above.

(8.29) Subject Control Construction

CP

C TP

1 0 KP i T

Mary T vP

0 ti v

v VP

Vj v tj CP

wants C TP

1 0 PRO k T

T vP

0 to tk v

v VP

Vl v tl KP

eat an apple 144 CHAPTER 8. RAISING AND CONTROL

(8.30) Subject Raising Construction

CP

C TP

0 KPi T

Mary T vP

v TP

0 Vj v ti T

seems T vP

0 to ti v

v VP

Vk v tk KP

like apples

8.3.2 Raising to Object

We discussed in Chapter 6 several reasons for assuming that the object in English moves overtly to a Case-licensing position. Several of these arguments are made within the context of raising-to-object constructions, so we consider more arguments here. During the GB era, the standard analysis of raising-to-object constructions was Exceptional Case Marking (ECM) (Chomsky, 1981a). At this time, it was assumed that Case was checked in a particular structural configuration. The relevant portion of the tree is as follows.

(8.31) (a) Mary believes Bill to like anchovies. 8.3. THE ANALYSIS OF RAISING AND CONTROL 145

(b) v 0

v VP

Vi v ti TP

0 believes KPj T

Bill to tjlike anchovies

In the GB era it was the verb (not v) that assigned accusative Case to the object. ECM referred to the particular configuration in (8.31 b), where the accusative-marked infinitival subject is in the specifier of the complement to the verb. Thus, one of the original arguments for the TP/CP distinction between raising and control constructions was theory-internal. Lasnik and Saito (1991a) argued that the infinitival subject raised to object position at LF based on evidence such as the following (as we saw in Chapter 6).

(8.32) (a) The DA proved [none of the defendants to be guilty] during any of the trials. (b) *The DA proved [that none of the defendants were guilty] during any of the trials.

The argument was that NPI licensing can be satisfied at LF, so the subject none of the defendants does not have to raise overtly. As Runner (1995) points out, however, there is evidence that the infinitival subject raises overtly to the matrix clause. Here is some of the data that he considered. Note that these facts have been mentioned previously in the literature (Postal, 1974; Johnson, 1991; Lasnik and Saito, 1992). In the first example the adverb incorrectly modifies the matrix verb. Thus Greg must have raised to the matrix clause. The second example with the particle verb make out was discussed above. In the third example the quantifier all has been stranded meaning the infinitival subject the boys has raised.

(8.33) (a) Mike expected Greg incorrectly to take out the trash. (b) She made Jerry out to be famous. (c) John expects the boys all to have left.

All of these facts suggest that (as discussed in the previous chapter) the object raises overtly in English. As Runner (2006) points out, however, there is no agreed upon analysis for how this takes place. Given the tripartite split VP layer, we can make a tentative proposal; nevertheless, this point requires future research. Here is an example.2 2One avenue of research to explore is the articulation of aspect, which we have not looked at here. Travis (2010) explores several ideas related to VP structure and aspect in a wide variety of languages, which also discusses overt object movement in English. 146 CHAPTER 8. RAISING AND CONTROL

(8.34) (a) Mary expects Bill to win the race. (b) CP

C T

0 KP1 T

Mary T VoiP

0 ti Voi

Voi vP

0 v j Voi KPl v

Vk v Bill tj VP

expects tk TP

0 tk T

T VoiP

0 to tk Voi

Voi vP

0 v m Voi KPo v

Vn v the race tm VP

win tn to

Before moving on to an updated analysis of control, we discuss a recent attempt to reduce control to movement. This move, although bold and not without its problems, would greatly simplify the grammar by removing PRO and its associated mechanisms from UG.

8.3.3 Control as Movement

In a series of papers, Norbert Hornstein and Cedric Boeckx attempted to reduce all forms on obligatory control to raising (Hornstein, 1999; Boeckx and Hornstein, 2003, 2004, 2006). A debate ensued with several researchers arguing in defense of distinguishing control and raising (Culicover 8.3. THE ANALYSIS OF RAISING AND CONTROL 147 and Jackendoff, 2001, 2006; Landau, 2003, 2007). In short, Hornstein argued that Control involves movement rather than PRO. In the following example Mary receives a theta-role from the embedded clause and then raises to the matrix clause to receive another theta-role there. Although this is a violation of the , Hornstein argues that it can be dispensed with as it is a hold-over from the GB era without much support from a Minimalist perspective.

(8.35) Maryi decided ti to ti read a book.

Hornstein noted that movement and control both obey Rosenbaum’s 1967 Minimal Distance Principle (MDP) alluded to above. Specifically, PRO is controlled by the closest c-commanding KP. Furthermore, both raising and control can only target the subject in the embedded infinitival clause, that is, the highest argument. This mirrors raising constructions. Consider the three schematized constructions below.

(8.36) (a) [TP KPi V[CP/TP ...ti/PRO... ] ]

(b) [TP KP V KPi [CP/TP ...ti/PRO... ] ]

(c)* [TP KPi V KP [CP/TP ...ti/PRO... ] ]

To remind ourselves, here are examples of each.

(8.37) Raising

(a) Maryi seems ti to like syntax.

(b) Mary expects Johni ti to pass the exam.

(8.38) Control

(a) Mary1 decided PRO1 to go to graduate school. (b) Mary persuaded John1 PRO1 to study harder for the exam.

It is this similarity that most strongly suggests that raising and control should be analyzed by the same mechanism. In separate lines of arguments, Landau on the one hand, and Culicover and Jackendoff on the other raise various arguments against conflating raising and control to movement (see the references cited above). One such argument concerns the MDP. While the examples shown above corroborate Hornstein’s approach to conflating raising and control to movement, not all instances of control fit neatly into this pattern. Consider the following examples.

(8.39) (a) Pat1 promised Alex PRO1 to wash the dishes. (b) The guard asked the prisoner1 PRO1 to leave the room. 148 CHAPTER 8. RAISING AND CONTROL

(c) The pupil1 asked the teacher PRO1 to leave early.

The reader should note that there is tremendous speaker variation in these forms.3 Nevertheless they are indicative of the speech for many English speakers. To the best of the author’s knowledge, however, no detailed variationist or socio-syntactic study has been undertaken on this variation. The second two examples illustrate a phenomenon known as control shift. Namely the identity of the controller can shift between the subject and the object depending on context. Promise is a clear violation of the MDP. Based partly on these facts, the authors above have argued that control cannot be reduced to movement. Another property that argues against such a reduction is partial control, discussed in the next section.

8.4 Partial Control

A crucial distinction between types of infinitives concerns the distinction between finiteness and tense. Consider the following example. The verb meet is a collective predicate and requires a plural subject.

(8.40) (a) We met at 6am. (b) *I met at 6am. (c) The chair decided to meet at 6am.

Here, PRO cannot be strictly coreferential with the controller, or we would expect this sentence to be ungrammatical, contrary to observation. This phenomenon is known as partial control (Wilkinson, 1971; Williams, 1980; Landau, 2000). We can represent it as follows.

(8.41) The chair1 decided PRO1+ to meet at 6am.

As this representation suggests the controller must be a subset of the set denoted by PRO1+. That is, (8.41) can be paraphrased as follows.

(8.42) The chair1 decided that she1 and others would meet at 6am.

By contrast exhaustive control is the obligatory identical co-reference between the controller and PRO. Laudau provides the following example.

(8.43) (a) The chair managed to arrive by 6am. (b) *The chair managed to meet at 6am.

3The author, for instance, finds (8.39 c) slightly odd, and prefers The pulil asked the teacher to be allowed to leave early much better. Control shift is often dependent on passivization in the embedded clause (see also Wurmbrand, 2001). 8.4. PARTIAL CONTROL 149

The matrix subject the chair must be identical in reference to PRO under the verb manage. Thus, the second example is ungrammatical since meet requires a plural subject. To be precise (Landau, 2001, p.3) proposes the following typology of control.

(8.44) Control

Obligatory Control Non-obligatory Control

Exhaustive Control Partial Control Long-Distance Control Arbitrary Control

Landau (2001) provides the following classification for exhaustive control and partial control verbs and provides examples as shown below.

(8.45) (a) exhaustive: implicative, aspectual, modal (b) partial: factive, propositional, desiderative, interrogative

(8.46) Exhaustive Control

(a) John managed to solve the problem. implicative (b) John began to solve the problem. aspectual (c) John had to solve the probleml. modal

(8.47) Partial Control

(a) John hated to solve the problem. factive (b) John claimed to have solved the problem. propositional (c) John hoped to solve the problem. desiderative (d) John wondered how to solve the problem. interrogative

Bresnan (1982) observed while infinitives are non-finite, they do not entirely lack tense. They typically describe an unrealized event. Stowell (1982) observed that infinitives can have either in- dependent tense or dependent tense. Consider the following examples.

(8.48) Tense in infinitives (Stowell, 1982, p.563)

(a) Jenny remembered to bring the wine. (b) Jenny tried to lock the door.

Landau showed that this difference can be illustrated with temporal . Consider the following examples. 150 CHAPTER 8. RAISING AND CONTROL

(8.49) Tense in infinitives (Landau, 2001, p.6)

(a) *Yesterday, John managed to solve the problem tomorrow. (b) Yesterday, John wanted to solve the problem tomorrow.

Landau showed that the set of verbs that allow only exhaustive control take infinitives that lack tense, while the set of verbs that allow for partial control take infinitives that have independent tense. This difference turns out to be crucial in his analysis of partial control. His analysis departs from the recognition that syntactic and semantic plurality are distinguished. Consider the following data, contrasting British English with American English.

(8.50) Semantic Plurality, British English

(a) The committee gathered before the vote. (b) The committee consulted each other before the vote. (c) The chair preferred to gather before the vote. (d) The chair preferred to consult each other before the vote.

(8.51) Semantic Plurality, American English

(a) The committee gathered before the vote. (b) *The committee consulted each other before the vote. (c) The chair preferred to gather before the vote. (d) *The chair preferred to consult each other before the vote.

What Landau shows through these examples is that semantic plurality is enough to license the use of collective predicates (such as gather) generally in English. Where the two dialects differ, however, is in the licensing of reciprocals. In British English, semantic plurality can license the recip- rocal, but in American English, it cannot. What’s crucial is that partial control behaves uniformly in both dialects. He proposes a feature [±SP] (semantic plurality), which is assigned to nominal elements. Crucially, as assumes there is a [+SP] PRO and a [-SP] PRO. Finally, Landau assumes that PRO is anaphoric and requires an antecedent. The feature he uses is [-R] ([-Referential]). Any element that is marked [-R] must enter into an Agree relationship with something that is [+R] in order to obtain a referent.

(8.52) (a) committee [+SP] (b) chair [-SP] (c) PRO [+SP] 8.4. PARTIAL CONTROL 151

(d) PRO [-SP]

To explain partial control, we need to make use of the fact that the infinitive of a partial control construction is tensed. Landau represents this with the feature [±T]. Both C and T bear the same value for this feature.4 We interpret this here via an agree relation shown by a dashed line. Note further that the type of CP complement ([-T] for exhaustive control complements and [+T] for partial control complements) depends on the matrix verb. Thus, the matrix verb must select a CP with the relevant feature matrix in C. We are now ready to consider his analysis (Landau, 2004). To keep the tree simple we will use only a bipartite vP structure. Also, certain aspects of his analysis have been simplified to facilitate discussion. Let’s consider first an example of exhaustive control.

(8.53) The chair managed to enter the conference room.

4This is similar to Chomsky’s (2008) notion of feature inheritance, in which the features of T are copied from C. However, in Landau’s analysis independent clauses, which do not exhibit control, lack the feature [±T]. It is left to future research how to reconcile Chomsky’s and Landau’s approaches. 152 CHAPTER 8. RAISING AND CONTROL

CP

C TP

0 KPi T

the chair T vP [+T, +Agr] 0 ti v

v VP

Vj v tj CP

managed C TP [-T] 0 PROk T [-R,-SP] T vP

0 to tk v [-T] v VP

Vl v ti KP

enter the conference room

Landau assumes that T and C enter into an Agree relationship (see footnote 4). PRO needs to find a referent since it is [-R], so enters into an Agree relation with KP in the matrix clause.5 As a result, PRO and the controller are identical in reference, and only exhaustive control is possible. To be clear, both the controller and PRO must have the same specification for [±SP].

(8.54) The chair preferred to gather in the conference room.

5Landau (2004) argues that PRO enters into an Agree relationship with the matrix T (or v in the case of ob- ject control), which then agrees with the controller. This fact becomes important for his analysis of control into subjunctives. 8.4. PARTIAL CONTROL 153

CP

C TP

0 KPi T

the chair T vP [+T, +Agr] 0 ti v

v VP

Vj v tj CP

preferred C TP [+T, +Agr] 0 PROk T

T vP

0 to tk v [+T, -Agr] v VP

Vl v ti PP

gather in the conference room Landau suggests partial control arises as follows. Note first that PRO is in an Agree relation with the embedded T. In contrast with the exhaustive control structure above, partial control crucially relies on the embedded C entering into an Agree relation with the matrix T. In accomplish this, Landau suggests that [+T, +Agr] on C or T triggers the addition of an uninterpretable [+R] on the relevant node. In a finite clause the uninterpretable [+R] is checked off the by subject. PRO, however, cannot do this in the embedded clause, hence the agree relation between the embedded C and the matrix T. Now, there is an indirect Agree chain between PRO and the matrix T, establishing a control relation and identifying the matrix subject as the controller of PRO. Note, crucially, there is no direct agree relation between PRO and the matrix subject. The final assumption that Landau has to make is that the feature [±SP] is found only on nominals and on functional heads that agree with nominals (namely T and v). Crucially, C cannot bear this feature. Since PRO and the matrix T are joined by a chain which contains C, it is possible for the matrix subject and PRO to have different values of [±SP]. 154 CHAPTER 8. RAISING AND CONTROL

8.5 Control into Subjunctive Clauses

Control has been observed in finite clauses in a number of languages including Hebrew (Lan- dau, 2004; Hagit Borer, 1986), Persian (Ghomeshi, 2001), Balkan languages, Malagasy, and Dogrib (Saxon, 1984), and Kannada. Consider the following Hebrew examples (Landau, 2004).

(8.55) Control in Hebrew (Landau, 2004, p.816f )

(a) Gil hizkir le-Rina1 še- prol/*2 tin’al et ha-delet. Gil reminded to-Rina that- pro will-lock.3sg.f acc the-door ‘Gil reminded Rina to lock the door.’ (b) Gil hizkir le-Rina1 se-*(hil/2) na’ala et ha-delet. Gil reminded to-Rina that-*(she) locked.3sg.f acc the-door ‘Gil reminded Rina that she had locked the door.’

Landau shows that the availability of control depends on the tense of the subjunctive clause. Only future subjunctive clauses support control, as the examples above show. As is known with English, control depends on the matrix verb. Some verbs select a control complement and some do not. Landau makes the same observation with Hebrew.

(8.56) (a) Rina himlica le-Gil1, še- pro1/*2 ya’avod yoter kaše. Rina recommended to-Gil that- pro will-work.3sg.M more hard ‘Rina recommended to Gil to work harder.’ (b) Rina hisbira le-Gil1 še-*(hul) ya’avod yoter kaše. Rina explained to-Gil that-*(he) will-work.3sg.M more hard ‘Rina explained to Gil that he would work harder.’

8.6 Non-obligatory Control

Landau (2000) defines non-obligatory control (NOC) as an infinitival clause with a controlled PRO, whose controller need not be a c-commanding DP in the superordinate clause. Consider the following example (Manzini, 1983).

(8.57) Mary1 knew that it would help Bill PRO1 to behave herself in public.

Other examples of NOC include the following.

(8.58) (a) John thinks it would be fun to go to Jeju. (b) It’s a good idea to brush your teeth every day.

(Landau, 2000, p.33) offers the following four criteria for distinguishing obligatory control (OC) from non-obligatory control (NOC). 8.6. NON-OBLIGATORY CONTROL 155

1 Arbitrary control possible with NOC, not with OC

2 Long-distance control possible with NOC, not with OC

3 Strict reading of PRO possible with NOC, not with OC

4 De re reading of PRO possible with NOC, not with OC - OC can only do

First, let’s consider arbitrary control. Landau offers the following examples. Observe that arbi- trary control is impossible with obligatory control, in the first two examples.

(8.59) Arbitrary control impossible with obligatory control

(a) *John tried PROarb to be quiet.

(b) *John remembered PROarb not to smoke around the babies.

(8.60) Arbitrary control possible with non-obligatory control

(a) It is dangerous for babies PROarb to smoke around them.

(b) PROarb to behave oneself in public would help John.

Long-distance control is possible only with non-obligatory control. Again, Landau offers the following examples.

(8.61) Long-distance control impossible with obligatory control

(a) *Mary1 knew that John dared PRO1 to perjure herself.

(8.62) Long-distance control possible with non-obligatory control

(a) John1 said that Mary thought that PRO1 shaving himself would bother Sue. (Chierchia and Jacobson, 1986)

Observe that in the first example, obligatory control, long-distance control is impossible. In the second example, long-distance control is possible. The next diagnostic requires us to understand the difference between strict and sloppy readings. Consider the following example. VP ellipsis in English allows for either a strict or sloppy reading on the elided pronoun. In the strict reading, the referent of the pronoun does not change. In both clauses his refers to Bill. In the sloppy reading, the referent of the pronoun changes freely. In this case, the referent of his varies with the subject, so in the sloppy reading, the elided pronoun his refers to Fred.

(8.63) (a) Bill talked to his mother, and so did Fred. 156 CHAPTER 8. RAISING AND CONTROL

(b) strict reading: Bill1 talked to his1 mother, and Fred talked to his1 mother, too. (c) sloppy reading: Bill talked to his mother, and Fred2 talked to his2 mother, too.

Consider now the following examples Landau gives.

(8.64) Strict and sloppy readings with control

(a) John tried PRO to leave, and Bill did, too. - OC, sloppy reading only (b) John thinks that PRO feeding himself will be difficult, and Bill does, too. - NOC, strict or sloppy reading

In (8.64 a) the PRO in the elided clause can refer to Bill only. Thus, this sentence can mean only that John tried for John to leave and that Bill tried for Bill to leave. In (8.64 b) however, either a strict or sloppy reading is available. The elided portion can mean either that Bill thinks that John feeding himself will be difficult or that Bill feeding himself will be difficult. The difference between the de re and de se readings requires us to formulate a context. In the following scenario, Mary has won a gold medal for skiing at the Olympics. Unfortunately, she later has a stroke and loses her memory. She is later watching a TV show about past skiers at the Olympics. She is unaware that the current skier is actually her. She recognizes that the skier’s performance was excellent and says, "I bet she’s gonna win a gold medal." She is still unaware that she is talking about herself. Now, in this context consider the following examples.

(8.65) (a) Mary1 expects that she1 will win a gold medal. (b) Mary1 thinks that PRO1 getting a gold medal will be exciting. (c) Mary1 expects PRO1 to get a gold medal.

In the first example, although she is coreferential with Mary, Mary does not have to be aware of this - only the speaker of (8.65 a) does. This is referred to as the de re reading, where Mary’s beliefs are not taken into account for evaluating the reference of she. Moving to (8.65 c) next, we see that only a de se reading is possible. Crucially, PRO is coreferential with Mary, but Mary must have this belief about herself (hence de se) to make the sentence felicitous. Since Mary is unaware the skier on TV is her, sentence (8.65 c) is false. Now, for the non-obligatory control sentence in (8.65 b), this sentence is true even considering the fact that Mary is unaware that the skier is her. We end here with a brief note on the structural differences between obligatory and non-obligatory control. As Landau (2000) notes, obligatory control is found in VP-internal arguments and non- obligatory control is found elsewhere. 8.6. NON-OBLIGATORY CONTROL 157

Further Reading

• Davies and Dubinsky (2004) - This is an advanced textbook that chronicles this history of raising and control from its origins to the dawn of the .

• Landau (2013) - This monograph is an advanced, cross-linguistic treatment of control within a contemporary generative framework. It also provides numerous diagnostics to distinguish between raising and control.

• Madigan (2008) - This dissertation provides an in depth discussion and analysis of control in Korean

• Wurmbrand (2001) - This monograph deals with control in German and other languages. It touches on many similar phenomena with English.

• Polinsky (2013) - This review article gives an overview of the issues surrounding the syntax of raising and control 158 CHAPTER 8. RAISING AND CONTROL Chapter 9

Linearization

Note - large portions of this discussion are taken word-for-word from Barrie (2012).

9.1 Introduction

Linearization is the process of converting a syntagmatic structure into a linear string of morphemes. Until the Government and Binding era, it was assumed that Phrase Structure Rules simply stip- ulated both constituency and linear order (Chomsky, 1955, 1957, 1965). Consider the following example.

(9.1) (a)S → NP VP (b)VP → V NP [English, French, Mandarin] (c)VP → NP V [Korean, Japanese, Lakhota]

The rule in (9.1 a stipulates that S consists of an NP and a VP and that the NP is to the left of the VP. Likewise, the rule in (9.1 b stipulates that VP consists of V and NP and the V is to the left of NP. The relationship "to the left of" is interpreted as "temporally prior to". With the advent of typological studies on word order (Greenberg, 1963), X-Bar Theory (Jack- endoff, 1977; Stowell, 1981), and Principle and Parameters (Chomsky, 1981b) the Headedness Pa- rameter was proposed to capture the difference between languages such as English on the one hand and Korean on the other. We turn to a discussion of the Headedness Parameter next.

9.2 The Headedness Parameter

The Headedness Paramater or Directionality Parameter holds that the order between a head and its complement is parametrically set (Chomsky, 1981b; Stowell, 1981; Travis, 1984). The idea

159 160 CHAPTER 9. LINEARIZATION

property OV VO adposition postposition preposition N and Relative Clause RelC N N RelC question formation Subject-Aux/V inversion sentence-final question particles

Table 9.1: Correlates of OV and VO order

is that all phrases are set to one of the following two settings.1

(9.2) XP XP

X ZP ZP X

The Headedness Parameter was, of course, originally formulated within X-Bar theory; however, we have updated the formalism here for BPS. Many researchers still assume that human language has some form of the Headedness Parameter. The Greenbergian correlates for OV and VO order are shown in Table (9.1). A major consequence of this approach was that clusters of word order properties (postpositions, post-nominal determiners, and OV order versus prepositions, pre-nominal determiners, and VO order; see Greenberg 1963) could be easily accounted for with one parametrically determined setting for headedness of all XPs. To be sure, the Headedness Parameter was conceived as a category-neutral specification of the linear order of the specifier, head and complementizer in a given language (Chomsky 1981; Stowell 1981). Let’s consider a Korean example to illustrate some of the properties in Table (9.1.)

(9.3) 나는 민수가 산 케이크를 포크로 먹었다. na-nun Minswu-ka sa-n kheykhu-lul phokhu-lo mek-ess-ta I-top Minsoo-nom buy-pst cake-acc fork-with eat-pst-decl ‘I ate the cake that Minsoo bought with a fork.’

In example there is a postposition (-lo ‘with’) and a pre-nominal relative clause (Minswu-ka sa-n ‘that Minsoo bought’), both textbook properties of SOV languages. Although it is not shown here, Korean is also wh-in-situ, another property correlated with SOV order. While the Greenbergian correlations are very strong, they are not absolute. There are a num- ber of disharmonic languages, that is, languages which do not fit entirely into the VO or the OV category. Consider the following Mandarin example. Mandarin is an SVO language, but has obligatory prenominal relative clauses. Again, it is now shown here, but Mandarin is wh-in-situ.

1It has rarely been proposed that the order between the specifier and X’ is parameterized. Instead, it seems to be uniformly accepted that the specifier is always on the left (Oishi, 2003; Richards, 2008). Here, the terms Headedness Parameter and Directionality Parameter diverge. While the Headedness Parameter is only about the head and its complement, the Directionality Parameter is concerned also with the order between the specifier and its sister. 9.3. ANTISYMMETRY 161

Chinese languages in general are textbook examples of disharmonic languages in the SOV/SVO binary macroparameter.

(9.4) Wˇo ch¯ı nˇı mˇai de dàng¯ao. I eat you buy de cake ‘I’m eating the cake that you bought.’

As with many macro-parameters, exceptions kept piling up, threatening the viability of this ap- proach.2 Well known examples include prepositions in SOV languages (German, Persian) and SOV languages with post-nominal relative clauses (Hiaqui, Kurdish). A quick look through http://wals.info/WALS provides numerous examples. These disharmonic languages led some researchers to pursue micro- parametric approaches to word order and others to abandon a parametric approach altogether. Despite the problems mentioned above, the Headedness Parameter was the dominant view of linearization until Kayne (1994) proposed his theory of Antisymmetry, which is the topic of the next section.

9.3 Antisymmetry

The Headedness Parameter in the previous section makes strong cross-linguistic predictions, many of which hold up. Indeed, most verb-final languages have postpositions rather than prepositions, for example. There are, however, a number of disharmonic properties found in many languages around the world, as discussed in the previous section. Richard Kayne presents a number of asymmetries that would be predicted if we admit Direc- tionality Parameters (Kayne, 2013).

1 dislocations and hanging topics

2 clitics and pronouns

3 agreement

4 relative clauses

5 serial verbs

6 coordination

7 pronominalization

2See Baker (2008) for a discussion of micro-parameters and macro-parameters. 162 CHAPTER 9. LINEARIZATION

Many languages have left dislocations and hanging topics. In some languages, they are exclusively on the left, such as Haitian Creole and Gungbe. In Italian and Catalan, such dislocations can be on the left or the right, but there are distinct asymmetries between the two. So far as we know, there is no language that exclusively has to the right only. These differences would not be expected under a symmetric view of syntax. Kayne notes one of Greenberg’s universals, which states that if object pronouns are post-verbal, full objects are, too. Kayne re-interprets this to indicate that pronouns can move further left than full noun phrases, but not vice versa. The following English data illustrate this.

(9.5) (a) I read all the books. (b) *I read the books all. (c) ?I read all them. (d) I read them all.

Likewise, the object clitic in French moves further to the left than a full KP object. Consider the following examples.

(9.6) Clitic Movement in French

(a) Pierre voit le chien. Peter sees the dog ‘Peter sees the dog.’ (b) Pierre le voit. Peter it sees ‘Peter sees it.’

Again, there is an asymmetry in which the pronoun or clitic moves to a higher position, which is always to the left. This asymmetric behaviour would not be expected with directionality parameters. There is no variety of English in which (9.5 b) is grammatical, but (9.5 d) is not. Again, under a symmetric view of syntax this asymmetry would not be expected. It is well known that agreement in English existentials depends on word order. Consider the following paradigm, which is consistent with many (but not all) dialects of English.

(9.7) (a) Some books *is/are on the table. (b) A book and a magazine *is/are on the table. (c) There is/are some books on the table. (d) There is/*are a book and a magazine on the table. 9.3. ANTISYMMETRY 163

The preverbal subject obligatorily agrees with the verb or auxiliary. Postverbal subjects either have default 3rd person singular agreement or are restricted to first conjunct agreement.34 Kayne notes the following asymmetries for relative clauses. Prenominal relative clauses tend to lack complementizers (such as English that) and relative pronouns, while postnominal relative clauses tend to one or both of these. Prenominal relative clauses tend to be non-finite, while postnom- inal relative can be finite. There is no reason to expect the asymmetry if language had directionality parameters. Kayne notes that the order between a verb and an auxiliary can vary cross-linguistically, as the following English and German examples show.

(9.8) (a) We believe that John has telephoned. (b) Wir glauben, daß Hans telefoniert hat. we believe that Hans telephoned has ‘We believe that Hans has telephoned.’

He notes, following Carstens (2002), that the order of serial verbs across language is universal.5 That is, for head-final languages, we do not see the mirror order for serial verb constructions (SVC). Consider first Sranan Tongo [srn] (cited as Sranan in Carstens).

(9.9) (a) seni wan boskopu gi tigri send a message give Tiger ‘Send Tiger a message.’ (b) a teki a fisi seri s/he take the fish sell ‘S/he sold the fish.’ (c) no teki baskita tyari watra neg take basket carry water ‘Don’t carry water with a basket.’

The order between the verb and the object in all these examples show that Sranan Tongo is a left-headed language. Assume the standard analysis for serial verb constructions in which one VP takes another VP as a complement Campbell (1989); Larson and LeFebvre (1991); Collins (1997a). An approximate structure is as follows.

(9.10) Partial structure of SVC in Sranan Tongo

3For more discussion on first conjunct agreement see Munn (1999), Benmamoun (1998), Aoun et al. (1994), and Miyagawa (2009). 4Kayne also discusses asymmetries with agreement in Romance languages. 5Carstens cites Muysken (1988) for the original observation. 164 CHAPTER 9. LINEARIZATION

VP

V VP

take KP V0

basket V KP

carry water

Consider, now, the following examples from I.jo. (ijo) a group of languages and dialects also called . Observe that the order between the verb and its objects shows that it’s a right headed language.

(9.11) (a) dúma tun-nì a-pí.rí. song sing-prt her-give ‘Sing her a song.’

(b) ayá bara-ki. àki. dúma tun new way-emph take song sing ‘Sing a song in a new way.’

(c) áràú. zu ye àki. buru tèri-mí s/he basket take yam cover-pst ‘She covered the yam with a basket.’

Given the observation that I.jo. is right-headed, we would expect one of the following structures. In the first example we see the putative structure assuming the possibility of the specifier projecting on the right.

(9.12) Putative structure for I.jo. SVC assuming right specifiers

VP

VP V

V0 KP take

KP V basket

yam cover

As Carstens notes, however, since the subject is on the left, it is reasonable to assume the specifier is also on the left. She suggests, then, the following putative structure. 9.3. ANTISYMMETRY 165

(9.13) Putative structure for I.jo. SVC assuming left specifiers

VP

VP V

KP V0 take

basket KP V

yam cover

Neither of these structures gives the correct word order for I.jo.. Carstens notes that Muysken made this general observation for SVCs in both VO and OV languages. Crucially, the word order in SVCs seems to be reflecting some underlying SVO order, according to Kayne. The difference between Sranan Tongo and I.jo., then, is that in Sranan Tongo the object remains in-situ, while in I.jo. the object moves up past the verb. Kayne notes that of the following possible orders only the first one is ever found in natural language.

(9.14) (a) KP and KP (b) *and KP KP (c) *KP KP and

This observation falls out naturally if we assume coordination is handled by a functional pro- jection with a specifier and a complement with Spec-Head-Complement order. &P

KP &0

John & KP

and Mary

Finally, Kayne notes that there is an asymmetry with backwards pronominalization ( in traditional grammar). Consider the following examples with coreference between John and he.

(9.15) (a) When John arrived home he fell asleep. (b) When he arrived home John fell asleep.

While English allows backwards pronominalization somewhat freely,6 other languages have very 6To my ears it has a slight literary feel. 166 CHAPTER 9. LINEARIZATION little or no backwards pronominalization at all. Kayne does not comment on the structural mani- festations of this observation, but notes that it is not expected under a symmetric view of syntax. We can call Classical Antisymmetry the original proposal in Kayne (1994). The technical implementation goes as follows. Word order is captured by the Linear Correspondence Axiom (LCA). The LCA is formulated as follows (Kayne, 1994, p. 6).

(9.16) d(A) is a linear ordering of T.

A is the complete set of ordered pairs , , . . . where

Xi and Yi are any two non-terminal syntactic nodes such that Xi asymmetrically c-commands Yi . d(A) is the mapping from A to the set of ordered pairs of terminals such that ai is dominated by

Xi and bi is dominated by Yi . T is the set of terminal nodes. Thus, d() is the Cartesian product of d(X) and d(Y). Alternatively, one can think of the LCA as follows. When a syntactic node X asymmetrically c-commands a node Y, all the terminal nodes dominated by X precede all the terminal nodes dominated by Y. The following example gives the definition of c-command adopted in Kayne (1994).

(9.17) X c-commands Y iff X and Y are categories and X excludes Y and every category that dominates X dominates Y. (Kayne, 1994: 18).

The following example illustrates how the LCA is implemented. In these examples, a, b, and c are terminals; A, B, and C are minimal projections dominating only terminals, and AP, BP, CP are maximal projections. Both minimal and maximal projections count as categories with respect to the definition in (9.17). Consider now the following example.

(9.18)

BP

AP BP

A B CP

a b C

c The set A, upon which the LCA is evaluated, consists of the set of ordered pairs in (9.19), which gives rise to d(A) in (9.20), the mapping from A to the set of pairs of ordered terminals as explained above.

(9.19) A = {, , , } 9.3. ANTISYMMETRY 167

(9.20) d(A) = {, , , }

In order for a derivation to be linearized, every terminal node must be ordered with respect to every other terminal node, either directly, or by transitivity. In other words, linear ordering must be total. Furthermore, linear ordering cannot be contradictory. That is, if A asymmetrically c-commands something inside B, then B cannot asymmetrically c-command anything inside A. Such a configuration would result in contradictory linearizations, and the derivation would crash at the PF interface. As a result of the LCA, tight restrictions are placed on phrase structure. Each maximal projection can contain only one specifier or one adjunct (the distinction between the two essentially being dissolved). Multiple adjunction or adjunction to a maximal projection that already has a specifier is prohibited. This is because the two adjuncts, or an adjunct and a specifier to the same maximal projection cannot be linearized with respect to each other.7 This is shown in the following example.

(9.21)

BP

AP BP

A DP BP

a D B CP

d b C

c Here BP has either two adjuncts/specifiers (the distinction between the two having evaporated). There is no asymmetric c-command relation between AP and DP in this structure under Kayne’s definition of c-command. The reason for this is that the three instances of BP in this example constitute a single category. Thus, B is dominated by BP, but AP and DP are not. They are thus not dominated by any category in this structure. In order for a category β to dominate an element α every segment of β must dominate α. Since the lowest segment of BP does not dominate either AP or DP, the category BP does not dominate either of these elements. As a consequence, this structure introduces a contradictory ordering. Observe that AP asymmetrically c-commands D, giving rise to the order ; however, DP also asymmetrically c-commands A, giving rise to the contradictory order . 7Guimarães (2008)argues that some constructions with multiple specifiers are actually compliant with the LCA. This is true only if one of the specifiers is a head and if there is no higher head that selects the XP with multiple specifiers as a complement. As we will we will see below, multiple specifiers are permitted in Dynamic Antisymmetry (as long as any point of symmetry that is formed is subsequently removed). 168 CHAPTER 9. LINEARIZATION

Such a simple and tight proposal has several consequences for phrase structure, expounded upon in Kayne (1994) and Bianchi (1999). One major consequence is the shape of phrase structure. The traditional distinction between specifiers and adjuncts is neutralized, and only one specifier may appear in a given phrase (i.e., multiple specifiers are banned). The order of the elements in a phrase is universally Spec-Head-Complement. Furthermore, right-adjunction is impossible.

9.4 Dynamic Antisymmetry

Kayne (1994) formulates Antisymmetry as a condition on representations. A major shift from GB Theory to Minimalism is the shift away from conditions on representations to conditions on deriva- tions. It is with this shift towards conditions on derivations that Moro proposed Dynamic Antisym- metry (Moro, 2000). Recall that Italian does not require an overt subject. Consider the following pair of examples.

(9.22) (a) Gli studenti sono arrivati. the students are arrived ‘The students arrived.’ (b) Sono arrivati gli studenti. are arrived the students ‘The students arrived.’

Although the subject may raise to the left of the auxiliary, it may remain in situ. In light of this fact Moro (2009) asks us to consider the following data.

(9.23) (a) [una foto del muro] è [ t [la causa della rivolta]] a picture of the wall is the cause of the revolution ‘A picture of the wall is the cause of the revolution.’ (b) [la causa della rivolta] è [[una foto del muro] t ] the cause of the revolution is a picture of the wall ‘The cause of the revolution is a picture of the wall.’ (c) *è [[una foto del muro] [la causa della rivolta]] is a picture of the wall the cause of the revolution

Adopting Bare Phrase Structure and assuming that linearization is a PF constraint, Moro (2000) develops a version of Antisymmetry that can tolerate violations of the LCA, as long as they’re repaired by LF. He takes as a starting point the following points of symmetry.

(9.24) (a) XP

X Y 9.4. DYNAMIC ANTISYMMETRY 169

(b) XP

ZP XP

YP XP (c) XP

ZP YP

In these three examples, the elements in the ovals c-command each other. Moro proposes that this symmetry acts as a trigger for movement. In fact, Moro strengthens this and argues that symmetry, not morphology (i.e., uninterpretable features and the like), serves as the sole trigger for movement. When two elements are in a configuration of symmetric c-command, one of them must raise to a higher position to break this point of symmetry. Moro illustrates this with equative constructions such as in the following pair. Recall the Italian examples from above, too.

(9.25) (a) The winner of the race is the man with the green hat. (b) The man with the green hat is the winner of the race.

Under Moro’s analysis, the two DPs the winner of the race and the man with the green hat c-command each other forcing one to raise. In this case, either DP can raise. In other situations, certain properties of the grammar will dictate that one of the two elements must raise. Dynamic Antisymmetry has made a more modest impact on current syntactic research than the general Antisymmetric program, but its influences certainly are not absent (Baauw, 1998; Barrie, 2011; Guasti and Moro, 2001; Hsieh and Sybesma, 2007). 170 CHAPTER 9. LINEARIZATION

Further Reading

• Kayne (1994) - This is Kayne’s original monograph where he proposes Antisymmetry

• Kayne (2013) - Although Kayne’s theory of Antisymmetry is difficult to reconcile with SOV languages, this paper discusses why we may want to do so.

• Barrie (2012) - This review article discusses Antisymmetry and Dynamic Antisymmetry Alphabetical Index

I.jo., 164 external argument, 39, 44 wh-in-situ, 82, 160 feature inheritance, 151 Across-The-Board Movement, 40, 41, 43 features agreement, 20 binary, 22 Amharic, 102 person, 23 ATB, see Across-The-Board Movement privative, 22 Finnish, 71 Baker’s Ambiguity, 122 first conjunct agreement, 163 Basque, 110 floated quantifiers, see quantifier float Bulgarian, 128 French, 20, 159, 162 Burzio’s Generalization, 98, 110 Burzio, Luigi, 98 gender, 64 German, 33, 71, 76, 83, 97, 131, 132, 163 c-selection, 27 Greenberg, Joseph, 160 Chamorro, 89 Chao Yuen Ren, 8 Hale, Ken, 35 Chinese, see also Mandarin, Cantonese, etc. Halkomelem, 45 Chomsky, Noam, 8, 79 Harley, Heidi, 52, 62 clitic, 37, 82, 108, 162 Head Movement Constraint, 77 clusivity, 25 Hebrew, 66 complementizer, 21 Hiatian Creole, 68 control shift, 148 Hopi, 26 Hornstein, Norbert, 147 Daga, 24 Hungarian, 59, 96 discourse configurationality, 35 Icelandic, 90, 95 Empty Category Principle, 121 idioms, 35, 138 Exceptional Case Marking, 104, 141, 144 Infinitive, 104 Extended Standard Theory, 32 inflected, 96

171 172 ALPHABETICAL INDEX infinitive, 136 negative , 48 Inflected infinitive, 96 Nishnaabemwin, 30 internal argument, 39 Niuean, 66 Inuktitut noun incorporation, 76 Inuttut, 23 number, 23 Inuttut, 23 dual, 23, 26 island, 116 Ojibwe Italian, 59, 96, 103, 108, 118 Nishnaabemwin, 30 Jackendoff, Ray, 62 Onondaga, 76 Japanese, 70, 124, 159 Perlmutter, David, 98 Johnson, Kyle, 106 phonetic form, 79 Kahlina, 25 Phrase Structure Rules, 159 Kayne, Richard, 161 Portuguese, 96

Korean, 8, 28, 60, 159, 160 quantifier float, 40, 43 Busan, 21 Case, 68, 97 Raising to object, 104 morphology, 61 Raising to subject, 104 passive, 98 Ritter, Elizabeth, 23, 65, 66 Romani, 131, 132 label, 120 Romanian, 127 Lakhota, 159 Scope Marking Construction, 131 Latin, 68, 93 Serbo-Croatian, 128, 129 Linear Correspondence Axiom, 166 , 163 Lithuanian, 69 Spanish, 28, 64 logical form, 79 Sranan Tongo, 163 Lushootseed, 60 Structuralism, 7 Lyélé, 23 Subjacency, 117, 124, 125, 132 Malagasy, 45 superiority, 128 Maltese, 23 Swahili, 29 Mandarin, 82, 121, 159, 160 Tagalog, 87 Marantz, Alec, 61 Tahitian, 69 Massam, Diane, 66, 103 Tonkawa, 25 May, Robert, 80 Travis, Lisa, 33, 34, 70, 76, 77, 145 Merge, 31 tucking in, 129 Minimal Distance Principle, 147, 148 minimality, 121 variable binding, 47 ALPHABETICAL INDEX 173

VP-Internal Subject Hypothesis, 37, 39 Warlpiri, 35 Welsh, 33 174 ALPHABETICAL INDEX Bibliography

Abney, Stephen (1987). The English Noun Phrase and its Sentential Aspect, Ph.D. Dissertation, MIT, Cambridge, MA.

Adger, David (2003). Core Syntax, Oxford: Oxford University Press.

Adger, David (2010). ‘A minimalist theory of feature structure’, in Anna Kibork and Greville G. Corbett (eds.), ‘Features: Perspectives on a Key Notion in Linguistics’, 185–218, Oxford: Oxford University Press.

Aldridge, Edith (2008). ‘Generative Approaches to Ergativity’, Language and Linguistics Compass 2 (5): 966–995, URL http://dx.doi.org/10.1111/j.1749-818X.2008.00075.x.

Alexiadou, Artemis and Elena Anagnostopoulou (1998). ‘Parametrizing AGR: Word Order, V- Movement and EPP-Checking’, Natural Language & Linguistic Theory 16 (3): 491–539.

Alexiadou, Artemis, Elena Anagnostopoulou, and Florian Schäfer (2015). External Arguments in Transitivity Alternations, Oxford: Oxford University Press.

Aoun, Joseph, Elabbas Benmamoun, and Dominique Sportiche (1994). ‘Agreement, Word Or- der, and in Some Varieties of ’, Linguistic Inquiry 25 (2): 195–220, URL http://www.jstor.org/stable/4178858. Publisher: The MIT Press.

Aoun, Joseph and Yen-hui Audrey Li (2003). Essays on the Derivational and Representational Nature of Grammar: The Diversity of Wh-Constructions, Cambridge, MA: MIT Press.

Avery, Peter and Keren Rice (1989). ‘Segment structure and coronal underspecification’, Phonology 6 (2): 179–200.

Baauw, Sergio (1998). ‘Subject-Verb Inversion in Spanish: Movement as Symmetry Breaker’, Lin- guistics in the Netherlands 15: 1–12.

Baker, Carl Leroy (1968). Indirect questions in English, Ph.D. Dissertation, University of Illinois.

175 176 BIBLIOGRAPHY

Baker, Mark and Jonathan David Bobaljik (2017). ‘On inherent and dependent theories of ergative case’, in Jessica Coon, Diane Massam, and Lisa Travis (eds.), ‘The Oxford handbook of ergativity’, 111–134, Oxford: Oxford University Press. Publisher: Oxford University Press Oxford.

Baker, Mark C. (1988). Incorporation: A Theory of Grammatical Function Changing, Chicago, IL: University of Chicago Press.

Baker, Mark C. (1996). The Polysynthesis Parameter, Oxford: Oxford University Press.

Baker, Mark C. (2008). ‘The Macroparameter in a Microparametric World’, in Theresa Biberauer (ed.), ‘The Limits of Syntactic Variation’, 351–374, Amsterdam: John Benjamins.

Baltin, Mark (2001). ‘Movement to the Higher V is Remnant Movement’, Linguistic Inquiry 33 (4): 653–659.

Barrie, Michael (2007). ‘Control and Wh-Infinitivals’, in William D. Davies and Stanley Dubinsky (eds.), ‘New Horizons in the Analysis of Control and Raising’, 263–279, Dordrecht: Springer.

Barrie, Michael (2011). Dynamic Antisymmetry and the Syntax of Noun Incorporation, Dordrecht: Springer.

Barrie, Michael (2012). ‘Antisymmetry and Dynamic Antisymmetry’, Language and Linguistics Compass 6 (1): 21–35, URL https://onlinelibrary.wiley.com/doi/abs/10.1002/lnc3.319. _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/lnc3.319.

Barrie, Michael and Hyunjung Lee (2017). ‘Complementizer allomorphy in Busan Korean’, in Michael Yoshitaka Erlewine (ed.), ‘Proceedings of GLOW in Asia XI’, 15–27, National University of Singapore: MIT Department of Linguistics and Philosophy, MITWPL.

Barss, Andrew and Howard Lasnik (1986). ‘A Note on and Double Objects’, Linguistic Inquiry 17: 347–354.

Basilico, David (2008). ‘Particle verbs and benefactive double objects in English: high and low attachments’, Natural Language & Linguistic Theory 26 (4): 731–773, URL https://doi.org/10.1007/s11049-008-9057-x.

Beck, Sigrid and Kyle Johnson (2004). ‘Double Objects Again’, Linguistic Inquiry 35 (1): 97–123, URL http://www.mitpressjournals.org/doi/abs/10.1162/002438904322793356.

Benmamoun, Elabbas (1998). ‘Agreement Asymmetries and the PF Interface’, .

Benveniste, Emile (1971). ‘The Nature of Pronouns’, in ‘Problems in General Linguistics’, 217–222, Coral Gables, FL: University of Miami Press. BIBLIOGRAPHY 177

Benveniste, Émile (1966). ‘La Nature des Pronoms’, in ‘Problèmes de Linguistique Générale’, 251– 257, Paris: Gallimard.

Bernstein, Judy (2008). ‘Reformulating the Hypothesis’, Language and Linguis- tics Compass 2 (6): 1246–1270.

Bianchi, Valentina (1999). Consequences of Antisymmetry: Headed relative clauses, Berlin: Mouton de Gruyter.

Bittner, Maria and Ken Hale (1996). ‘Ergativity: Toward a Theory of a Heterogenous Class’, Lin- guistic Inquiry 27 (4): 531–604.

Bjorkman, Bronwyn (2011). BE-ing default : the morphosyntax of auxiliaries, Ph.D. Dissertation, Massachusetts Institute of Technology.

Bochnak, M. Ryan and Lisa Matthewson (2020). ‘Techniques in Complex Se- mantic Fieldwork’, Annual Review of Linguistics 6 (1): 261–283, URL https://doi.org/10.1146/annurev-linguistics-011619-030452. _eprint: https://doi.org/10.1146/annurev-linguistics-011619-030452.

Boeckx, Cedric (2001). ‘Scope Reconstruction and A-Movement’, Natural Language & Linguistic Theory 19 (4): 503–548.

Boeckx, Cedric and Norbert Hornstein (2003). ‘Reply to "Control Is Not Movement"’, Linguistic Inquiry 34 (2): 269–280.

Boeckx, Cedric and Norbert Hornstein (2004). ‘Movement under Control’, Linguistic Inquiry 35 (3): 431–452.

Boeckx, Cedric and Norbert Hornstein (2006). ‘The Virtues of Control as Movement’, Syntax 9 (2): 118–130, URL http://dx.doi.org/10.1111/j.1467-9612.2006.00086.x.

Boeckx, Cedric and Sandra Stjepanović (2001). ‘Head-ing toward PF’, Linguistic Inquiry 32 (2): 345–355.

Bresnan, Joan (1982). Theory of Complementation in English Syntax, Ph.D. Dissertation, MIT, Cambridge, MA.

Bruening, Benjamin (2010). ‘Double Object Constructions Disguised as Prepositional Datives’, Linguistic Inquiry 41 (2): 287–305, URL http://www.mitpressjournals.org/doi/abs/10.1162/ling.2010.41.2.287. 178 BIBLIOGRAPHY

Burton, Strang and Jane Grimshaw (1992). ‘Coordination and VP-Internal Subjects’, Linguistic Inquiry 23 (2): 305–313, URL http://www.jstor.org/stable/4178769. Publisher: The MIT Press.

Béjar, Susana (2003). Phi-Syntax: A Theory of Agreement, Dissertation, University of Toronto, Toronto.

Béjar, Susana and Milan Rezac (2009). ‘Cyclic Agree’, Linguistic Inquiry 40 (1): 35–73.

Cable, Seth (2010). The Grammar of Q: Q-Particles, Wh-Movement and Pied-Piping, Oxford: Ox- ford University Press.

Cable, Seth (2012). ‘Reflexives, Reciprocals and Contrast’, Journal of Semantics 31 (1): 1–41, URL https://doi.org/10.1093/jos/ffs020. _eprint: https://academic.oup.com/jos/article- pdf/31/1/1/4610841/ffs020.pdf.

Campbell, Richard (1989). The Grammatical Structure of Verbal Predicates, Ph.D. Dissertation, University of California, Los Angeles.

Carlson, Katy, Charles Clifton, and Lyn Frazier (2001). ‘Prosodic Boundaries in Adjunct Attachment’, Journal of Memory and Language 45 (1): 58 – 81, URL http://www.sciencedirect.com/science/article/pii/S0749596X00927621.

Carstens, Vicki (2002). ‘Antisymmetry and Word Order in Serial Constructions’, Language 78 (1): 3–50.

Cheng, Lisa Lai Shen (1991). On the Typology of Wh-Questions, Ph.D. dissertation, MIT.

Chierchia, Gennaro and Pauline Jacobson (1986). ‘Local and Long Distance Control’, in Stephen Berman, Joyce McDonough, and Jae-Woong Choe (eds.), ‘Proceedings of NELS 16’, 57–74, Amherst, MA: Graduate Linguistics Student Association, University of Massachusetts.

Chomsky, Noam (1955). The Logical Structure of Linguistic Theory, Chicago, IL: University of Chicago Press.

Chomsky, Noam (1957). Syntactic Structures, Janua Linguarum, vol. 4, The Hague: Mouton.

Chomsky, Noam (1959). ‘Review of Skinner (1957) Verbal Behaviour’, Language 35 (1): 26–58.

Chomsky, Noam (1964). ‘Current Issues in Linguistic Theory’, in ‘Fodor, Jerry A.(ed.); Katz, Jerrold J.(ed.). The Structure of Language: Readings in the . Englewood Cliffs, NJ : Prentice Hall, 1964.’, 50–118.

Chomsky, Noam (1965). Aspects of the Theory of Syntax, Cambridge, MA: MIT Press. BIBLIOGRAPHY 179

Chomsky, Noam (1970). ‘Remarks on Nominalization’, in Roderick Jacobs and Peter Rosenbaum (eds.), ‘Readings in English ’, 184–221, Washington: Georgetown UP.

Chomsky, Noam (1973). ‘Conditions on Transformations’, in Stephen-R. Anderson and Paul Kiparsky (eds.), ‘A Festschrift for Morris Halle.’, 232–86, New York: Holt, Rinehart and Winston.

Chomsky, Noam (1977). ‘On wh-movement’, in Peter W. Culicover, Thomas Wasow, and Adrian Akmajian (eds.), ‘Formal Syntax’, 71–132, New York, NY: Academic Press, Inc.

Chomsky, Noam (1981a). Lectures on Government and Binding, Dordrecht: Foris.

Chomsky, Noam (1981b). ‘Principles and Parameters in Syntactic Theory’, in Norbert Hornstein (ed.), ‘Explanation in Linguistics: The Logical Problem of Language Acquisition’, Longman Ling. Lib.. London; 25, 32–75, London: Longman.

Chomsky, Noam (1982). Some Concepts and Consequences of the Theory of Government and Bind- ing, Cambridge, MA: MIT Press.

Chomsky, Noam (1986). Barriers, Cambridge, MA: MIT Press.

Chomsky, Noam (1993). ‘A Minimalist Program for Linguistic Theory’, in Ken Hale and Samuel J. Keyser (eds.), ‘The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger’, 1–52, Cambridge, MA: MIT Press.

Chomsky, Noam (1994). Bare Phrase Structure, MIT Occasional Papers in Linguistics, vol. 5, Cam- bridge, MA: MIT Department of Linguistics and Philosophy, MITWPL.

Chomsky, Noam (1995). The Minimalist Program, Cambridge, MA: MIT Press.

Chomsky, Noam (2000). ‘Minimalist Inquiries: The Framework’, in Roger Martin, D. Michaels, and Juan Uriagereka (eds.), ‘Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik’, 89–156, Cambridge, MA: MIT Press.

Chomsky, Noam (2001). ‘Derivation by Phase’, in Michael Kenstowicz (ed.), ‘Ken Hale: A Life in Language’, 1–52, Cambridge, MA: MIT Press.

Chomsky, Noam (2008). ‘On Phases’, in P. Oltero (ed.), ‘Foundational Issues in Linguistic Theory’, 133 – 166, Cambridge, MA: MIT Press.

Chomsky, Noam (2013). ‘Problems of projection’, Lingua 130 (1): 33–49.

Chomsky, Noam & and Howard Lasnik (1977). ‘Filters and Control’, Linguistic Inquiry 8: 425–504.

Chomsky, Noam and Morris Halle (1968). The Sound Pattern of English, New York, NY: Harper & Row. 180 BIBLIOGRAPHY

Chomsky, Noam and Howard Lasnik (1995). The Minimalist Program, Current Studies in Linguis- tics. (CSLing). Cambridge, MA; 28, Cambridge, MA: MIT Press.

Chung, Sandra (1998). The Design of Agreement: Evidence from Chamorro, Chicago: University of Chicago Press.

Citko, Barbara (2014). Phase Theory: An Introduction, Cambridge studies in linguistics, Cambridge: Cambridge University Press.

Clements, G. N. (1985). ‘The geometry of phonological features’, Phonology Yearbook 2 (1): 225–252. Publisher: Cambridge University Press.

Clements, Nick and Elizabeth Hume (1995). ‘The Internal Organization of Speech Sounds’, in John Goldsmith (ed.), ‘The Handbook of Contemporary Phonlogical Theory’, 245–306, Cambridge: Blackwell.

Collins, Chris (1997a). ‘Argument Sharing in Serial Verb Constructions’, Linguistic Inquiry 28 (3): 461–497, URL http://www.jstor.org/stable/4178987.

Collins, Chris (1997b). Local Economy, Cambridge: MIT Press.

Compton, Richard and Christine Pittman (2010). ‘Word formation by phase in Inuit’, Lingua 120 (9): 2167–2192, URL http://ling.auf.net/lingBuzz/000480.

Contini-Morava, Ellen and Yishai Tobin (2000). Between Grammar and Lexicon, Current Issues in Linguistic Theory, Amsterdam: John Benjamins Publishing Company.

Coon, Jessica (2013). Aspects of Split Ergativity, Oxford Studies in Comparative Syntax, Oxford: Oxford University Press.

Coon, Jessica and Omer Preminger (2017). ‘Split Ergativity is not about Ergativity’, in Jessica Coon, Diane Massam, and Lisa Travis (eds.), ‘The Oxford handbook of ergativity’, 226–252, Oxford: Oxford University Press. Publisher: Oxford University Press Oxford.

Cowper, Elizabeth (2005). ‘A Note on Number’, Linguistic Inquiry 36 (3): 441–455.

Culicover, Peter and Ray Jackendoff (2001). ‘Control is not Movement’, Linguistic Inquiry 32 (3): 493–512.

Culicover, Peter W. and Ray Jackendoff (2006). ‘Turn Over Control to the Semantics!’, Syntax 9 (2): 131–152, URL http://www.blackwell-synergy.com/doi/abs/10.1111/j.1467-9612.2006.00085.x. BIBLIOGRAPHY 181

Davies, William D. and Stanley Dubinsky (2004). The Grammar of Raising and Control, Malden, MA: Blackwell.

Dayal, Veneeta (1994). ‘Scope Marking as Indirect Wh Dependency’, Natural Language Semantics 2 (2): 137–170.

Deal, Amy Rose (2015). ‘Reasoning about Equivalence in Semantic Fieldwork’, in M. Ryan Bochnak and Lisa Matthewson (eds.), ‘Methodologies in Semantic Fieldwork’, 157–174, Oxford: Oxford University Press.

Deen, Kamil Ud (2005). The Acquisition of Swahili, Language Acquisition and Language Disorders, Amsterdam: John Benjamins Publishing Company.

Dixon, R.M.W. (1994). Ergativity, Cambridge: Cambridge University Press.

Déchaine, Rose-Marie (1993). Predicates across categories: Towards a category-neutral syntax, Ph.D. Dissertation, University of Massachussetts, Amherst.

Eijk, Jan P. [Van and Thom Hess (1986). ‘Noun and verb in salish’, Lingua 69 (4): 319 – 331.

Embick, David and Robert Rolf Noyer (2007). ‘Distributed morphology and the Syntax/Morphology Interface’, in Gillian Ramchand and Charles Reiss (eds.), ‘The Oxford Handbook of Linguistic Interfaces’, 289–324, Oxford: Oxford University Press.

Embick, David and Rolf Noyer (2001). ‘Movement Operations after Syntax’, Linguistic Inquiry 32 (4): 555–595, URL https://doi.org/10.1162/002438901753373005. _eprint: https://doi.org/10.1162/002438901753373005.

Forchheimer, Paul (1953). The Category of Person in Language, Berlin: WalterDe Gruyter.

Fox, Danny and David Pesetsky (2005). ‘Cyclic Linearization and its interaction with other aspects of grammar: a reply’, Theoretical Linguistics 31 (1-2): 235–262.

Freidin, Robert (1999). ‘Cyclicity and Minimalism’, in Samuel David Epstein and Norbert Hornstein (eds.), ‘Working Minimalism’, 95–126, Cambridge, Mass.: MIT Press.

Galloway, Brent (1993). A Grammar of Upriver Halkomelem, Berkeley, Los Angeles, London: Uni- versity of California Press.

Ghomeshi, Jila (2001). ‘Control and Thematic Agreement’, Canadian Journal of Linguistics 46 (1/2): 9–40.

Gillon, Carrie (2015). ‘Investigating D in Languages With and Without Articles’, in Lisa Matthew- son and Ryan Bochnak (eds.), ‘Methodologies in Semantic Fieldwork’, 175–203, Oxford: Oxford University Press. 182 BIBLIOGRAPHY

Greenberg, Joseph (1963). Universals of Language, Cambridge, MA: MIT Press.

Grimshaw, Jane (1979). ‘Complement Selection and the Lexicon’, Linguistic Inquiry 10 (2): 279– 326.

Guasti, Maria-Teresa and Andrea Moro (2001). ‘Romance and dynamic antisymmetry: Essays Offered to Lorenzo Renzi’, in Guglielmo Cinque and Giampaolo Salvi (eds.), ‘Current Studies in Italian Syntax’, 173–188, Amsterdam: North Holland.

Guimarães, Maximiliano (2008). ‘A note on the strong generative capacity of standard Antisymme- try Theory’, Snippets 18: 5–7.

Haegeman, Liliane (1991). Introduction to Government and Binding Theory, Oxford: Blackwell.

Hagit Borer (1986). ‘I-Subjects’, Linguistic Inquiry 17 (3): 375–416, URL http://www.jstor.org/stable/4178498. Publisher: The MIT Press.

Hale, Ken (1983). ‘Warlpiri and the Grammar of Non-Configurational Languages’, Natural Language and Linguistic Theory 1 (1): 5–47.

Halle, Morris and Alec Marantz (1993). ‘Distributed Morphology and the Pieces of Inflection’, in Kenneth Hale and Samuel J. Keyser (eds.), ‘The View from Building 20: Essays in Linguistic in Honor of Sylvain Bromberger’, 111–176, Cambridge, MA: MIT Press.

Hallman, Peter (2015). ‘Syntactic Neutralization in Double Object Constructions’, Linguistic In-

quiry 46 (3): 389–424, URL http://www.mitpressjournals.org/doi/abs/10.1162/LINGa00187.

Halpern, Aaron (1995). On the Placement and Morphology of Clitics, Stanford, CA: CSLI Publica- tions.

Harbour, Daniel (2011). ‘Valence and Atomic Number’, Linguistic Inquiry 42 (4): 561–594, URL

https://doi.org/10.1162/LINGa00061._eprint : https : //doi.org/10.1162/LING_a_00061.

Harley, Heidi (2002). ‘Possession and the Double Object Construction’, Yearbook of Linguistic Vari- ation 2 (1): 31–70.

Harley, Heidi (2005). ‘One-replacement, unaccusativity, acategorial roots and Bare Phrase Struc- ture’, in Slava Gorbachov and Andrew Nevins (eds.), ‘Harvard Working Papers in Linguistics, volume 11’, 59–78.

Harley, Heidi (2007). ‘External Arguments: On the independence of Voice° and v°’, files, CD6: GLOW XXX. BIBLIOGRAPHY 183

Harley, Heidi (2013). ‘External arguments and the Mirror Principle: On the distinctness of Voice and v’, Lingua 125 (1): 34–57.

Harley, Heidi and Hyun Kyoung Jung (2015). ‘In Support of the PHAVE Analysis of the Double Object Construction’, Linguistic Inquiry 46 (4): 703–730.

Harley, Heidi and Elizabeth Ritter (2002). ‘Person and number in pronouns: a feature-geometric analysis’, Language 78 (3): 482–526.

Harris, Randy Allen (1995). The Linguistics Wars, Oxford: Oxford University Press.

Hoff, Berend Jacob (1968). The Carib language: Phonology, morphophonology, morphology, texts and word index, The Hague, Paris: Mouton.

Hoijer, Harry (1933). Tonkawa: An Indian language of Texas, New York: Columbia University Press.

Hornstein, Norbert (1999). ‘Movement and Control’, Linguistic Inquiry 30 (1): 69–96.

Hsieh, Feng-fan and Rint Sybesma (2007). ‘On the Linearization of Chinese Sentence Final Particles: Max Spell Out and Why CP Moves’, .

Huang, C.-T. James (1982). Logical Relations in Chinese and the Theory of Grammar, Ph.D. Dis- sertation, MIT, Cambridge, MA.

Hung, Henrietta J. (1988). ‘The Structure of Derived Nouns and Verbs in Malagasy: A Syntactic Account’, .

Jackendoff, Ray (1977). X-bar Syntax: A Study of Phrase Structure, Linguistic Inquiry Monographs, Cambridge, MA: MIT Press.

Jelinek, Eloise (1984). ‘Empty categories, case, and configurationality’, Natural Language & Lin- guistic Theory 2 (1): 39–76.

Johns, Alana (1992). ‘Deriving Ergativity’, Linguistic Inquiry 23 (1): 57–87.

Johnson, Kyle (1991). ‘Object Positions’, Natural Language & Linguistic Theory 9 (4): 577–636.

Jun, Sun-Ah (2003). ‘Prosodic phrasing and attachment preferences.’, Journal of Psycholinguistic Research 32 (2): 219–249. Place: Germany Publisher: Springer.

Kayne, Richard (1984). Connectedness and Binary Branching, Dordrecht: Foris.

Kayne, Richard (1994). The Antisymmetry of Syntax, Cambridge, MA: MIT Press. 184 BIBLIOGRAPHY

Kayne, Richard (2013). ‘Why are there no directionality parameters?’, in Theresa Biberauer and Michelle Sheehan (eds.), ‘Theoretical Approaches to Disharmonic Word Order’, 219–244, Oxford: Oxford University Press.

Kiss, Katalin É. (ed.) (1995). Discourse Configurational Languages, Oxford Studies in Comparative Syntax, New York, Oxford: Oxford University Press.

Koizumi, Masatoshi (1995). Phrase Structure in Minimalist Syntax, Ph.D. Dissertation, MIT, Cam- bridge, MA.

Kramer, Ruth (2015). The Morphysyntax of Gender, Oxford: Oxford University Press.

Kratzer, Angelika (1996). ‘Severing the External Argument from its Verb’, in Johan Rooryck and Laurie Zaring (eds.), ‘Phrase Structure and the Lexicon’, 109–138, Dordrecht: Kluwer.

Kratzer, Angelika and Elisabeth Selkirk (2007). ‘Phase theory and prosodic spellout: The case of verbs’, The Linguistic Review 24 (1): 93–135.

Kroeger, Paul (1993). Phrase Structure and Grammatical Relations in Tagalog, Standford, CA: CSLI Publications.

Landau, Idan (2000). Elements of Control: Structure and meaning in infinitival constructions, Dor- drecht: Kluwer Academic Publishers.

Landau, Idan (2001). ‘Control and Extraposition: The Case of Super-Equi’, Natural Language & Linguistic Theory 19 (1): 109–152.

Landau, Idan (2003). ‘Movement out of Control’, Linguistic Inquiry 34 (4): 471–498.

Landau, Idan (2004). ‘The Scale of Finiteness and the Calculus of Control’, Natural Language & Linguistic Theory 22 (4): 811–877.

Landau, Idan (2007). ‘Movement-Resistant Aspects of Control’, in William D. Davies and Stanley Dubinsky (eds.), ‘New Horizons in the Analysis of Control and Raising’, No. 71 in Studies in Natural Language and Linguistic Theory, 293–325, Dordrecht: Springer.

Landau, Idan (2013). Control in Generative Grammar: A Research Companion, Cambridge: Cam- bridge University Press.

Larson, Richard (2009). Grammar as Science, Cambridge, MA: MIT Press.

Larson, Richard and Claire LeFebvre (1991). ‘Some Issues in Verb Serialization’, in ‘Serial Verbs: Grammatical, Comparative and Cognitive Approaches’, 185–208, Amsterdam: John Benjamins Publishing Company. BIBLIOGRAPHY 185

Larson, Richard K. (1988). ‘On the Double Object Construction’, Linguistic Inquiry 19 (3): 335– 391.

Larson, Richard K. (1990). ‘Double Objects Revisited: Reply to Jackendoff’, Linguistic Inquiry 21 (4): 589–632.

Larson, Richard K. (1991). ‘Promise and the Theory of Control’, Linguistic Inquiry 22 (1): 103–139.

Lasnik, Howard (1999). ‘Chains of Arguments’, in Samuel Epstein and Norbert Hornstein (eds.), ‘Working Minimalism’, 189–215, Cambridge, MA: MIT Press.

Lasnik, Howard and Mamoru Saito (1991a). ‘On the Subject of Infinitives’, vol. 27, 324–343, Chicago Linguistic Society.

Lasnik, Howard and Mamoru Saito (1991b). ‘Weakest Crossover’, Linguistic Inquiry 22 (4): 687– 720.

Lasnik, Howard and Mamoru Saito (1992). Move Alpha: Conditions on its Application and Output, vol. MIT Press, Cambridge, MA.

Lees, Robert (1960). The Grammar of English Nominalizations, The Hague: Mouton.

Lefebvre, Claire (1982). ‘L’expansion d’une catégorie grammaticale: Le determinant la’, in Lefebvre, Claire, Magloire-Holly, Hélène, and Nanie Piou (eds.), ‘Syntaxe de l’haïtien’, 21–63, Ann Arbor, MI: Karoma.

Legate, Julie Anne (2008). ‘Morphological and Abstract Case’, Linguistic Inquiry 39 (1): 55–101, URL http://www.mitpressjournals.org/doi/abs/10.1162/ling.2008.39.1.55.

Legate, Julie Anne (2014). Voice and v: Lessons from Acehnese, Cambridge, MA: MIT Press.

Lin, Jo-wang (1992). ‘The Syntax of Zenmeyang ‘how’ and Weishenme ‘why’ in Mandarin Chinese’, Journal of East Asian Linguistics 1: 293–331.

Lombardi, Linda (1991). Laryngeal features and laryngeal neutralisation, University of Mas- sachusetts at Amherst, PhD Dissertation, University of Massachussetts, Amherst.

Lombardi, Linda (1995). ‘Laryngeal features and privativity’, The Linguistic Review 12 (1): 35– 59, URL https://www.degruyter.com/view/journals/tlir/12/1/article-p35.xml. Place: Berlin, Boston Publisher: De Gruyter Mouton.

Longobardi, Giuseppe (1996). ‘The Syntax of N-raising: a minimalist theory’, OTS Working Papers, Utrecht 1–151. 186 BIBLIOGRAPHY

Madigan, Sean (2008). Control Constructions in Korean, Ph.D. Dissertation, University of Delaware.

Manzini, Maria Rita (1983). ‘On Control and Control Theory’, Linguistic Inquiry 14 (3): 421–446, URL http://www.jstor.org/stable/4178338. Publisher: MIT Press.

Manzini, Maria Rita (1992). : A Theory and Some of Its Empirical Consequences, Cam- bridge, MA: MIT Press.

Marantz, Alec (1984). On the Nature of Grammatical Relations, Cambridge, MA: MIT Press.

Marantz, Alec (1991). ‘Case and Licensing’, in Germán F. Westphal, Benjamin Ao, and Hee-Rahk Chae (eds.), ‘ESCOL 91: Proceedings of the Eighth Eastern States Conference on Linguistics’, 234–253.

Marantz, Alec (1997). ‘No Escape from Syntax: Don’t Try Morphological Analysis in the Privacy of Your Own Lexicon’, University of Pennsylvania Working Papers in Linguistics 4 (2): 201–225.

Marantz, Alec (2001). ‘Words’, West Coast Conference on Formal Linguistics .

Massam, Diane (2001). ‘Pseudo Noun Incorporation in Niuean’, Natural Language & Linguistic Theory 19 (1): 153–197.

Massam, Diane (2020). Niuean: Predicates and Arguments in an Isolating Language, Oxford Studies of Endangered Languages, Oxford: Oxford University Press.

Mathiassen, Terje (1996). A Short Grammar of Lithuanian, Slavica Publishers, URL https://books.google.co.kr/books?id=jqtiAAAAMAAJ.

Matthewson, Lisa (2004). ‘On the Methodology of Semantic Fieldwork’, International Journal of American Linguistics 70 (4): 369–415.

May, Robert (1985). Logical Form: Its Structure and Derivation, Cambridge, MA: MIT Press.

McDaniel, Dana (1986). Conditions on wh-chains, Ph. D. Dissertation, City University of New York.

McDaniel, Dana (1989). ‘Partial and Multiple Wh-Movement’, Natural Language & Linguistic The- ory 7 (4): 565–604.

McNally, Louise (1992). ‘VP Coordination and the VP-Internal Subject Hypothesis’, Linguistic Inquiry 23 (2): 336–341, URL http://www.jstor.org/stable/4178772. Publisher: The MIT Press.

Miyagawa, Shigeru (2009). Why Agree? Why Move?, Cambridge, MA: MIT Press. BIBLIOGRAPHY 187

Moltmann, Friederike (2013). ‘The semantics of existence’, Linguistics and Philosophy 36 (1): 31–63, URL https://doi.org/10.1007/s10988-013-9127-3.

Moro, Andrea (2000). Dynamic Antisymmetry, Cambridge, MA: MIT Press.

Moro, Andrea (2009). ‘Rethinking symmetry: a note on labelling and the EPP’, Snippets 19: 17–18.

Munn, Alan (1999). ‘First Conjunct Agreement: Against a Clausal Analysis’, Linguistic Inquiry 30 (4): 643–668.

Murane, Elizabeth (1974). Daga grammar: From morpheme to discourse, Norman, OK: Summer Institute of Linguistics.

Muysken, Pieter (1988). ‘Parameters for serial verbs’, in Victor Manfredi (ed.), ‘Niger-Congo syntax and semantics 1’, 65–75, Boston, MA: Boston University African Studies Center.

Nevins, Andrew (2007a). ‘Dual is Still More Marked Than Plural’, Harvard Working Papers in Linguistics 12: 6.

Nevins, Andrew (2007b). ‘The representation of third person and its consequences for the person- case constraint’, Natural Language and Linguistic Theory 25 (2): 273–313.

Newell, Heather and Tobias Scheer (2017). ‘Prosodic Domains: overthrowing the hierarchy’, CD12.

Newmeyer, Frederick (2014). Linguistic Theory in America, San Diego, CA: Academic Press.

Oishi, Masayuki (2003). ‘When Linearity Meets Bare Phrase Structure’, Current Issues in English Linguistics 2: 18–41.

Ott, Dennis (2015). ‘Symmetric Merge and local instability: Evidence from split topics’, Syntax 18 (2): 157–200.

Perlmutter, David M. (1970). ‘The two verbs "begin"’, in Roderick Jacobs and Peter Rosenbaum (eds.), ‘Readings in Transformational Grammar’, vol. 107-119, Waltham, MA: Ginn & Co.

Perlmutter, David M. (1971). Deep and Surface Structure Constraints in Syntax, American Problem Studies, New York City, NY: Holt, Rinehart and Winston.

Perlmutter, David M. (1978). ‘Impersonal Passives and the Unaccusative Hypothesis’, 157–189, files: BLS.

Pesetsky, David (1987). ‘WH-in-situ: Movement and Unselective Binding’, in Eric Reuland and Alice Ter Meulen (eds.), ‘The Linguistic Representation of (In)definiteness’, 98–129, Cambridge, MA: MIT Press. 188 BIBLIOGRAPHY

Pesetsky, David (2000). Phrasal Movement and its Kin, Cambridge, MA: MIT Press.

Picallo, M. Carme (1991). ‘Nominals and nominalization in Catalan’, Probus 3: 279–316.

Pierrehumbert, Janet and Julia Hirschberg (1990). ‘The Meaning of Intonational Contours in the Interpretation of Discourse’, in P. R. Cohen, J. Morgan, and M. E. Pollack (eds.), ‘Intentions in Communication’, 271–311, Cambridge, MA: MIT Press.

Polinsky, Maria (2013). ‘Raising and control’, in Marcel Den Dikken (ed.), ‘The Cambridge Hand- book of Generative Syntax: Grammar and Syntax’, 577–606, Cambridge: Cambidge University Press.

Polinsky, Maria and Eric Potsdam (). ‘Austronesian syntax’, in ‘Oceania’, Berlin: Mouton. Backup Publisher: Mouton.

Pollock, Jean-Yves (1989). ‘Verb Movement, Universal Grammar, and the Structure of IP’, Linguistic Inquiry 20 (3): 356–424.

Postal, Paul (1974). On Raising, Cambridge, MA: MIT Press.

Preminger, Omer (2014). Agreement and Its Failures, Cambridge, MA: MIT Press.

Pylkkänen, Liina (2008). Introducing Arguments, Cambridge, MA: MIT Press.

Rackowski, Andrea and Norvin Richards (2005). ‘Phase Edge and Extraction: A Tagalog Case Study’, Linguistic Inquiry 36 (4): 565–599.

Raposo, Eduardo (1987). ‘Case Theory and Infl-to-Comp: The Inflected Infinitive in European Portuguese’, Linguistic Inquiry 18 (1): 85–109.

Rezac, Milan, Pablo Albizu, and Ricardo Etxepare (2014). ‘The structural ergative of Basque and the theory of Case’, Natural Language & Linguistic Theory 32 (4): 1273–1330, URL http://www.jstor.org/stable/43697756. Publisher: Springer.

Richards, Marc D. (2008). ‘Desymmetrization: Parametric Variation at the PF Interface’, Canadian Journal of Linguistics 53 (2/3): 275–300.

Richards, Norvin (2001). Movement in Language: Interactions and architectures, Oxford: Oxford University Press.

Richards, Norvin (2016). Contiguity Theory, Cambridge, MA: MIT Press. van Riemsdijk, Henk (1982). ‘Correspondence Effects and the Empty Category Principle’, in Yukio Otsu, Henk Van Riemsdijk, K. Inoue, A. Kamio, and N. Kawasaki (eds.), ‘Studies in Genera- tive Grammar and Language Acquisition: A report on recent trends in linguistics’, 5–16, Tokyo: Editorial Committee. BIBLIOGRAPHY 189

Ritter, Elizabeth (1992). ‘Cross-Linguistic Evidence for Number Phrase’, Canadian Journal of Lin- guistics 37 (2): 197–218.

Ritter, Elizabeth (1993). ‘Where’s Gender?’, Linguistic Inquiry 24 (4): 795–803.

Ritter, Elizabeth and Martina Wiltschko (2014). ‘The composition of INFL’, Natural Language & Linguistic Theory 32 (4): 1331–1386, URL https://doi.org/10.1007/s11049-014-9248-6.

Rizzi, Luigi (1982). Issues in Italian Syntax, Dordrecht: Foris Publications.

Rizzi, Luigi (1990). Relativized Minimality, Cambridge, MA: MIT Press.

Roberts, Ian (1991). ‘Excorporation and Minimality’, Linguistic Inquiry 22 (1): 209–218.

Rosenbaum, Peter S. (1967). The Grammar of English Predicate Complement Constructions, Cam- bridge, MA: MIT Press.

Ross, John (1967). Constraints on Variables in Syntax, Ph.D. Dissertation, MIT, Cambridge, MA.

Rounds, Carol (2001). Hungarian: An Essential Grammar, Essential grammar, London and New York: Routledge.

Rudin, Catherine (1988). ‘On multiple questions and multiple wh-fronting’, Natural Language and Linguistic Theory 6: 445–501.

Runner, Jeffrey T. (2006). ‘Lingering Challenges to the Raising-to- object and Object-control Constructions’, Syntax 9 (2): 193–213, URL http://dx.doi.org/10.1111/j.1467-9612.2006.00088.x.

Runner, Jeffrey Thomas (1995). Noun Phrase Licensing and Interpretation, Doctoral Dissertation, Amherst. 11.

Saxon, Leslie (1984). ‘Control and Agreement in Dogrib’, in Gloria Alvarez, Belinda Brodie, and Terry McCoy (eds.), ‘Proceedings of the Annual Eastern States Conference on Linguistics’, 128– 139, Columbus, OH: Department of Linguistics, Ohio State University.

Showalter, Catherine (1986). ‘Pronouns in Lyele’, in Ursula Wiesemann (ed.), ‘Pronominal Systems’, 205–216, Tübingen: Gunter Narr.

Soh, Hooi Ling (2005). ‘Wh-in situ in Mandarin Chinese’, Linguistic Inquiry 36 (1): 143–155.

Sproat, Richard (1983). ‘VSO Languages and Welsh Configurationality’, in Isabelle Haik and Diane Massam (eds.), ‘Papers in Grammaticality Theory’, MIT Working Papers in Linguistics, vol. 5, Cambridge, MA: MIT Department of Linguistics and Philosophy, MITWPL. 190 BIBLIOGRAPHY

Steriade, Donca (1995). ‘Underspecification and ’, in John Goldsmith (ed.), ‘Handbook of Contemporary Phonology’, Malden, MA: Blackwell.

Stowell, Tim (1981). Origins of Phrase Structure, Ph.D. Dissertation, MIT, Cambridge, MA.

Stowell, Tim (1982). ‘The Tense of Infinitives’, Linguistic Inquiry 13: 561–570.

Stuurman, Frits (1985). Phrase structure theory in generative grammar, Publications in Language Sciences, Dordrecht: Foris Publications.

Sutcliffe, Edmund F. (1936). A grammar of the Maltese language, London: Oxford University Press.

Svenonius, Peter (2004). ‘On the Edge’, in David Adger, Cécile De Cat, and George Tsoulas (eds.), ‘Peripheries: Syntactic Edges and their Effects’, 259–287, Dordrecht: Kluwer.

Tonhauser, Judith and Lisa Matthewson (2015). ‘Empirical evidence in research on meaning’, .

Travis, Lisa de Mena (1984). Parameters and Effects of Word Order Variation, Ph.D. Dissertation, Massachussets Institute of Techonology, Cambridge, MA.

Travis, Lisa de Mena (2010). Inner Aspect: The Articulation of VP, Amsterdam: John Benjamins.

Tsai, Dylan (1994). On Economizing the Theory of A-Bar Dependencies, Ph.D. Dissertation, MIT, Cambridge, MA.

Uriagereka, Juan (2012). Spell-Out and the Minimalist Program, Oxford linguistics, Oxford: Oxford University Press.

Von Stechow, Arnim (1996). ‘The different readings of wieder "again": A structural account’, Journal of Semantics 13 (1): 87–138.

Wang, William S-Y. and 王士元 (1989). ‘Language in China: A Chapter in the His- tory of Linguistics / 言展的史回’, Journal of Chinese Linguistics 17 (2): 183–222, URL www.jstor.org/stable/23757139. Publisher: [Chinese University Press, Project on Linguistic Analysis].

Watanabe, Akira (1992). ‘Subjacency and S-structure movement of wh-in-situ’, Journal of East Asian Linguistics 1 (3): 255–291.

Watanabe, Akira (2001). ‘Wh-in-situ Languages’, in Mark Baltin and Chris Collins (eds.), ‘The Handbook of Contemporary Syntactic Theory’, 201–225, Malden, MA: Blackwell.

Wilkinson, Robert (1971). ‘Complement Subject Deletion and Subset Relations’, Linguistic Inquiry 2 (4): 575–584, URL http://www.jstor.org/stable/4177670. Publisher: The MIT Press. BIBLIOGRAPHY 191

Williams, Edwin (1980). ‘Predication’, Linguistic Inquiry 11 (1): 203–238.

Wiltschko, Martina (2006). ‘On ’Ergativity’ in Halkomelem Salish’, in Alana Johns, Diane Massam, and Juvénal Ndayiragije (eds.), ‘Ergativity: Emerging Issues’, 197–227, Dordrecht: Kluwer.

Wiltschko, Martina (2014). The Universal Structure of Categories: Towards a Formal Typology, Cambridge: Cambridge University Press.

Wurmbrand, Susi (2001). Infinitives: restructuring and clause structure, Berlin, New York: Mouton de Gruyter.