Contents

I The theory
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3
5 6 6 8 9 11 11 12 12 13 17 17 19 20 21 22 26 27 27 27 29 33 33 36 36 38

1 Hoopmans Syntactic analysis introduced. 1.1 Constituency tests. . . . . . . . . . . . . . . 1.1.1 Substitution . . . . . . . . . . . . . . 1.1.2 Ellipsis . . . . . . . . . . . . . . . . 1.1.3 Coordination . . . . . . . . . . . . . 1.1.4 Movement . . . . . . . . . . . . . . . 1.1.5 Summary, Chapter III of Hoopmans: 1.1.6 Exercise. . . . . . . . . . . . . . . . . 1.2 Class notes . . . . . . . . . . . . . . . . . . 1.2.1 Constituency tests . . . . . . . . . . 2 X theory. 2.1 General Structure . . . . . . . . 2.1.1 NPs . . . . . . . . . . . 2.1.2 VPs . . . . . . . . . . . 2.1.3 PP, AdjP, AdvP . . . . 2.1.4 Speciers: DP . . . . . 2.1.5 The structure of TP . . 2.1.6 CP . . . . . . . . . . . . 2.1.7 Clauses . . . . . . . . . 2.1.8 Finiteness . . . . . . . . 2.1.9 Movement of auxiliaries 3 Binding Theory 3.1 Principle A . . . 3.2 Principle B . . . 3.2.1 Pronouns 3.3 Principle C . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

II

Lecture notes and readings

3.4 3.5 Chapter 10. The forms of Sentences. Howard Lasnik. . . . . . . . Phrase Structure Naoki Fukui . . . . . . . . . . . . . . . . . . . . 1

41
43 49

CONTENTS 3.5.1 What is X-Theory . . . . . . . . . . . . . . . . . . . . . . 3.5.2 Class notes. . . . . . . . . . . . . . . . . . . . . . . . . . . Chomsky 1995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 52 53 59 59 59 60 67

3.6

4 A-Movements. Mark Baltin 4.1 Reading notes . . . . . . . . . 4.1.1 A-movement vs. A-bar 4.1.2 Binding Theory . . . 4.2 Head Movement Pollock . . . Appendices Appendix A Sample Trees Appendix B Theta Roles Appendix C Tables of concepts

. . . . . . . movement . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

71 73 75

Appendix D Paper ideas 77 D.1 OP-Ed vs. News: Where does Congressional debate stand? . . . 77 Appendix E Grammatical categories Appendix F Documentation F.1 Invocation . . . . . . . . F.2 How to convert a tree to F.3 Usage and features . . . F.4 Tree placement . . . . . F.5 Advanced features . . . F.6 How do I . . . ? . . . . . . F.7 Troubleshooting . . . . . F.8 Inspiration . . . . . . . for Qtree . . . . . . . brackets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 81 82 82 83 85 86 87 88 89

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

Lecture notes on Syntax I

Marlene Guraieb September 26, 2013

2 TP I T [past] T vP v v saw VP V V (saw) nP a man

CONTENTS

Part I

The theory

Chapter 1

Hoopmans Syntactic analysis introduced.

Denition 1. Consitutent. A Constituent is a string that speakers can manipulate as a single chunk. Note. If a string of words or morphemes is a constitutent, we will represent this constituency by grouping all the words or morphemes as daughters of a single mother node in a tree representation:

in the bedroom Or alternatively:

D N

N N

in the

bed room Denition 2. For each length n, the number of well-formed expressions with n morphemes is called the density of a language. Denition 3. A language is slender if there is some xed number k , such that for every number n, the number of well formed expressions of length n is no more than k. 5

CHAPTER 1. HOOPMANS SYNTACTIC ANALYSIS INTRODUCED.

Denition 4. A morpheme is the smallest grammatical unit in a language. Denition 5. A free morpheme can function independently as a word. Example town, dog which can appear with other lexemes such as town hall or dog house. A bound morpheme can appear only as part of a word, always in conjunction with a root or other bound morphemes. Denition 6. A Derivational morpheme is a bound morpheme which, when combined with a root, changes the semantic meaning or the part of speech of the aected word. Denition 7. An Inectional morpheme is a bound morpheme that modies a verbs tense or a nouns number without aecting the words meaning or class. Denition 8. A Parsing algorithm is a program that nds the syntactic structures of a sequence of morphemes. Denition 9. A constituent is a word or a group of words that functions as a single unit within a hierarchical structure. It does not have to be minimal. put a picture of Bill on your desk is a constituent, but so is a picture of Bill. Denition 10. Constituency tests are diagnostics employed to identify the constituent structure of sentences. Denition 11. A truth value is the attribute assigned to a proposition in respect of its truth or falsehood, which in classical logic has only two possible values (true or false). Denition 12. The polarity of a statement consists of its character as a positive or a negative.

1.1
1.1.1

Constituency tests.
Substitution

Denition 13. The substitution test involves replacing the test constituent with the appropriate pro-form (e.g. pronoun). Substitution normally involves using a denite pro-form like it, he, there, here, etc. in place of a phrase or a clause. If such a change yields a grammatical sentence where the general structure has not been altered (i.e. the truth value is preserved), then the test sequence is a constituent. Denition 14. In phonetics and phonology, a minimal pair is a set of words that dier in only one regard, frequently in a single particular phoneme. However, minimal pairs can also dier in tone, accent, or other features. Additionally, the term minimal pair has been extended to syntax to describe sentences or clauses that dier in a single word or branching structure.

1.1. CONSTITUENCY TESTS.

Note. Substitution should preserve truth values across contexts of evaluations (that is, non-accidentally). If the sentence is true without the substitution it should remain true after and if it is false then it should remain false. Tree: sentence W D this N P NP D A N T V N DP P Name Bill P Z T N P T

will put D

before tomorrow

a picture of

on your desk

girl in the red coat Note. The substitution test does not unambiguously tell us that anything that can be replaced by a single word will necessarily be a constituent. Whether this is the correct interpretation can only be determined a posteriori. Theorem 1. Transitivity? The transitivity of identity is the logical principle that, if a=b, and b=c, then a=c. Note. Swapping of constituents of the same category is possible when you have independently established that they ARE in fact constituents of the same category. It is not true that swappability by itself entails that the swapped strings are constituents and constituents of the same kind. Denition 15. A complimentary distribution entails two words where when one occurs the other cannot. Summary of substitution tests: The substitutions we are interested in replace a substring by a word, where that word plays the same kind of role in the sentence as the original string did, as we saw when we considered example (16). To try to guarantee this sameness of role, we require that the substitution be structured so as to preserve both well formedness and the truth value of the original sentence. In addition, we require that the substitution be general enough. Strings that can be manipulated as chunks under such substitution are constituents The kind of word that can be substituted for a string indicates its category) substitution by a pronoun indicates that the constituent is a DP substitution by one or ones indicates that the constituent is a NP substitution by there in its locative sense can indicate that the constituent is a PP

CHAPTER 1. HOOPMANS SYNTACTIC ANALYSIS INTRODUCED. String substitution failure is not a test for non-constituency

Substitution by a string longer than one word is not necessarily one that preserves constituency

1.1.2

Ellipsis

Denition 16. An ellipsis test is a test of constituency tat treats strings of words as chunks and attempts to substitute them by the null string. The test requires that the pear of sentences with or without ellipsis be true or false in the same situations (that is, have the same truth value).

Note. One condition that we impose is that ellipsis be done in those discourse contexts in which an antecedent sentence is present. Remember necessarily understood and understood in the same way. Ellipsis only seems to be able to aect a continuous string, that is a string of words or morphemes that is not interrupted by another string. On the morphological tree structure, this means that branches are not allowed to cross, which has the eect that elements that are part of a constitutent cannot be separated by elements that are not part of a constituent. elements in a constituent must form a continuous string.

Denition 17. A yes/no question is a question that can be answered by yes or no. It is in these cases that ellipsis is typically most ecient.

Denition 18. Tag question is a grammatical structure in which a declarative statement or an imperative is turned into a question by adding an interrogative fragment (the tag). I.e, in Youre John, arent you?, the statement Youre John is turned into a question by the tag arent you.

A tree:

1.1. CONSTITUENCY TESTS.

sentence DP D this N girl P in D NP PP DP A N T will V put D VP DP N P N P Q PP DP N T VP P Q

before tomorrow

the red coat

a picture of Bill on

your desk Denition 19. A do so replacement is a good constituency test that substitutes by two words: do so.

1.1.3

Coordination

Denition 20. Coordinated constituents are the two strings that can be coordinated by a conjunction to form a coordinated sentence. Coordination rule. We can coordinate two elements if: 1. we can say each of the two sentences independently 2. these two sentences have identical parts and dissimilar parts 3. we can substitute on dissimilar part for the other, preserving acceptability 1 . Corollary 1. Coordination test: If we have two acceptable sentences of the form A B D and A C D - where A, B, C and D represent (possible null) substrings and the string A B and C D is acceptable with the same meaning as A B D and A C D, this is evidence that B and C are both constituents, and constituents of the same kind. Note. Coordination is perhaps the only test in which failure is straightforwardly meaningful.
1 this indicates that if one of the two dissimilar parts is a constituent, both are constituents of the same kind (the substitution test)

10

CHAPTER 1. HOOPMANS SYNTACTIC ANALYSIS INTRODUCED.

Corollary 2. Interpreting Coordination Test Failure. If we have two acceptable sentences of the form A B D and A C D where none of A, B, C and D are bound morphemes, and the string A B and C D is not acceptable (even after we have xed agreement), then it is not true that: B and C are both constituents and constituents of the same kind. That is, coordination failure means that one or more of the following is true: 1. B is not a constituent, or 2. C is not a constituent, or 3. B and C are not of the same kind A tree: DP DP C DP

John and Mary DP C DP DP C DP DP

and Sue

John and Mary DP DP C DP DP C DP

John and

Mary and Sue DP DP C DP C DP

John and Mary and Sue

1.1. CONSTITUENCY TESTS. Right node raising

11

I do not understand very well. It is a problem with the coordination test where two constituents pass the test, but are not of the same kind: 1. They play unusual music. 2. I listen to unusual music. 3. They play and I listen to unusual music. 4. *Hence, They play and I listen are the same type of constitutents. Thats all I understand about this, ask.

1.1.4

Movement

Topicalization Note. Topicalization can only aect constituents.

1.1.5

Summary, Chapter III of Hoopmans:

Substitution : if a string S can be replaced by a single word, this is evidence that S is a constituent. In particular: Pronominalization : if a string S can be replaced by a pronoun, this is evidence that S is a DP One substitution : if a string S can be replaced by one, this is evidence that S is an NP Do so substitution : if a string S can be replaced by do so, this is evidence that S is a VP Ellipsis: if a string S can be deleted, this is evidence that S is a constituent Coordination: if a string S can be coordinated with a word (or phrase), this is evidence that S is a constituent, a constituent of the same category as the word (or phrase) it is coordinated with Movement : if a string S can be moved to another position (typically, all the way to the right or to the left), this is evidence that S is a constituent. In particular: Topicalization : DPs, PPs, VPs (VP preposing) Clefting : DPs, PPs PseudoClefting : VPs, APs, DPs, . . . Wh-movement : DPs, PPs Right node raising : . . .

12

CHAPTER 1. HOOPMANS SYNTACTIC ANALYSIS INTRODUCED. Heavy shift : DP Important caveat: When an experiment does not successfully apply to S, the reasons for failure could be extremely varied. As a result, it does not show that S is not a constituent.

1.1.6

Exercise.

Draw the tree structure for the following sentence: The boy will bring my mother Bills most recent book about global warming when he can. sentence DP P N T will VP. V bring P DP N NP P Bills A AP A NP N P A DP PP DP N VP

The boy

my mother

book about

global warming

most recent

1.2

Class notes

Where is syntax in the overall architecture of the grammar? Theres the lexicon which stores elements (words morphemes). Through the operation select we pick some words. The main operations we need are merge (External merge) and move (internal merge). Syntax is the bridge between sound and meaning. Syntax puts it all together for you. Lingering questions Why doesnt the logical form precede the phonological form. what is grammar?

1.2. CLASS NOTES

13

Anaphors: reexives (things that end in self or selves) + reciprocal (one another). They are put together by binding theory (antecedent): Binding theory for anaphors: anaphors must be bound in some local domain (the binder has to segment the anaphor) The study of binding theory is a very good clue that theres a lot more going on in human language than just word order. Everyone and no one are quantiers.

1.2.1

Constituency tests

That tall student of Lingistics from Brussels will read the new novel quickly. Substitution. Pronomialization for DPs: That tall student of linguistics from Brussels - he the new novel - it There-substitution Brussels One substitution for NPs: tall student of linguistics - one student of linguistics - one tall student of linguistics from Brussels - one novel - quickly Do so substitution for VPs. read the new novel quickly - do so read the new novel - do so Ellipsis: quickly ... (AdvP adverbial phrase) new ... (AP adjectival phrase) tall ... read the new novel of Linguistics from Brussels Coordination: That tall and brilliant student of linguistics... of linguistics and yoga * student and scholar of linguistics. what youre doing here is right node raising. * John bakes pies and Mary cakes. This is gapping. You think Mary cakes is a constituent, but its really the gapping of bakes cakes that. Topicalization Moving things around and not Clefting Pseudo-clefting WH-movement Heavy (NP) Shift sentence Np T VP That tall student of Linguistics from Brussels will read the new novel quickly. My attempt:

14

CHAPTER 1. HOOPMANS SYNTACTIC ANALYSIS INTRODUCED.

S NP DP P that Adj tall N student P PP N P NP AdjP NP AdjP PP N M will V read D VP NP NP N AdvP Adv quickly

the Adj

new novel

of linguistics from Brussels

Class

1.2. CLASS NOTES

15

TP DP D That AP A tall N student P of NP PP DPNP N linguistics Syntactic categories: That D (determiner) tall A student N of preposition linguistics N will t (tense) Aux (Auxiliary) read V the D new A novel N quickly Adv. Phrasal category The dierent between a complement and an adjunct is the replacement. Theres an EPP (extended projection principle) is what makes the Tense phrase T, it means it is lacking something (an object). P from NP NP PP DPNP N Brussels t will V read D VP DP NP NP N novel T VP AdvP Adv quickly

the AdjP new

16

CHAPTER 1. HOOPMANS SYNTACTIC ANALYSIS INTRODUCED.

Chapter 2

X theory.
2.1 General Structure

Lets frigging kill X-bar theory once and for all!!! Consider the NP structure: NP: (D) (AdjP+) N (PP+) When we do constituency tests on this, we nd out that there is more embedded structure here than just the horizontal structure depicted here. We can break up this rule, then and we get: NP: (D) N N: AdjP N N: N PP N: N (PP) We also have VP horizontal structure as: VP: (AdvP+) V ({NP/CP}) (PP+) (AdvP+) Which now becomes: VP: V V: AdvP V V: V PP V: V AdvP 17

18 V: V ({NP/CP}) Now AdjP: AdjP: (AdvP) Adj Adj : Adj (PP) Or PP: PP: (AdjP) P P : P (PP) P: P DP

CHAPTER 2. X THEORY.

This leads us to infer that the rules all have the following form: XP:ZP X X : X (YP) X: (YP) X X: X (WP) This is the principle of phase structure in X theory: XP: (ZP) X A phrase (XP) consists of optionally another phrase and a bar level projection (X ) X : YP X or X: X YP A bar-level projection (X ) can consist of another X and another phrase (recursivity property) X: X (WP) A bar level projection (X) consists of a head of the same category (X) and optionally another phrase. Structurally this looks like this (with any number of X, since it is recursive): XP ZP YP X X X YP

X WP What are the names of the dierent parts of the structure?

2.1. GENERAL STRUCTURE

19

The phrase which is immediately dominated by XP (ZP in our tree) is the specier SPEC??. The phrase dominated by X and the sister of X is an adjunct. The phrase which is sister to X is the complement. Note the dierence between complements and adjuncts. There is only one complement, that which is sister to X, whereas there can be any number of adjuncts (sort of like a christmas tree, you can hang them all there if you want and keep getting X as the head). The complement of a head is more intimately related to it (think of book of poems vs. book on the table ). the complement of N in English is almost always introduced by the preposition of. X theory allows for only one complement and indeed in an NP we cannot have to of-PPs of this sort ([*]The book of poems of ction). An adjunct on the other hand, feels more optional. X theory allows for any number of adjuncts and they can be reordered freely. X theory also predicts that the complement PP of an NP must be rst, it cannot be re-ordered with respect to adjunct PPs. A useful test to dierentiate adjuncts and complements is coordination (Coordinating two elements of a given category gives an element of the same category). You cannot conjoin a complement and an adjunct PP: [*] The book of poems and with the red spine, but you can conjoin complements and adjuncts with themselves. Also one replacement is very useful: one -replacement test: One can stand in for an N, but not for an N. This predicts that you should not ever be able to get one followed by a complement PP; One should only be able to be followed by adjunct PPs. [*] The one of essays on the third shelf.

2.1.1

NPs

Adjuncts do not have to be on the right. Left-handed adjuncts to NP include AdjP: The big red boring book of poems. The problem is that complements do not always have to be on the right either. A complement is the phrase which is sister to the head, but either of these structures has a complement XP:

N XP XP N Think for example of N AdjP N There can only be one complement

linguistics book

20

CHAPTER 2. X THEORY.

here: [*] The linguistics book of essays or The boring linguistics book (boring adjunct; linguistics complement). An interesting ambiguity: The French Teacher can either mean The teacher of French (complement) or The teacher from France (adjunct).

NP N NP French N teacher AdjP

NP N N N teacher

French

How do we disambiguate? Complements have to be closest to the head: (1) (2) The French German teacher The German French teacher

One cannot strand the complement: (3) The French one

And conjuncts must be of the same category: (4) (5) The French and Math teacher The tall and German teacher

2.1.2

VPs

In X every phrase has the same basic structure. In particular, VPs have the same properties as NPs: Only one complement Adjuncts which can be of any number and are reorderable: In VPs the direct object is the complement: (6) The students ate the sandwiches

Other things (Advs, PPs) are adjuncts:

2.1. GENERAL STRUCTURE (7) (8) The students left at 7 oclock The students left swiftly.

21

X theory rules: There can be only one complement: (9) (10) * The students ate the sandwiches the pizza Cf. The students ate the sandwiches and the pizza

The complement must be closest to the head (11) (12) The students ate the pizza in record time. * The students ate in record time the pizza

Adjuncts may be re-ordered (13) (14) The students ate the pizza in record time on Thursday. The students ate the pizza on Thursday in record time.

Do so cant strand the complement. (15) (16) * John ate the pizza and Mary did so the sandwiches John ate the pizza in short order but Mary did so in record time.

Likes conjoin only with likes (17) (18) John ate the pizza quickly and with gusto. [??] Mary ate the pizza and with gusto.

Note: The reason example 18 doesnt sound so bad is that it is possible to interpret this as Mary ate the pizza and she did so with gusto leaving she did so as unpronounced. It is hard to get around this problem, so this test is not very reliable for VPs.

2.1.3

PP, AdjP, AdvP

It turns out to be more dicult to show parallels in PPs, AdjPs, and AdvPs, but we will still assume that they follow the same structural rules as VPs and NPs. So-replacement cant strand the complement (AdjP) (19) John was afraid of tigers; Mary was less so (* of lions).

There can be only one complement (AdjP) (20) * John was afraid of tigers of lions

There can be only one complement (PP) (21) * John fell o the roof the house.

22

CHAPTER 2. X THEORY.

2.1.4

Speciers: DP

There is a third position in X that we havent talked about (after adjuncts and complements): Specier (ZP here).

XP ZP YP X X WP The main example of a specier we have seen so far is the D in the NP (the in the books or this in this book ). This is a problem for X theory because D is a head, and speciers are suppose to be phrases. If we want to believe in X theory our structure for NP that has D in its specier cannot really be the structure. Speciers should have phrases (XPs), yet D is a head. If D is a head, we have an immediate conclusion we can draw from X. D heads a DP. X X YP

DP D D the In fact this is wrong. The relationship is stronger, the DP is not inside the NP at all! Rather, the NP is inside the DP. The NP is the complement to D.

2.1. GENERAL STRUCTURE

23

DP D D NP N N book This is the way things are in X, but what evidence do we have that this is right? Consider the genitive (possesive) s in English: Johns hat The students sandwich The man from Australias book The man on the hill by the trees binoculars Notes that the s attaches to the whole possessor phrase! In the last two examples, it isnt even attached to the head noun (its the mans book and binoculars, not Australias or the trees, after all). This is not a noun sux. It is more like a little word that signals possession, standing between the possessor and the possessee. It is also impossible to have both a s and a determiner. (22) (23) (24) * The buildings the roof Cf. The roof of the building * The tigers the eye

Determiners like the and the possession marker s seem to be in complementary distribution -if one appears, the other cannot. This would only make sense if both the and s are instances of the category D; DP can have only one head. This suggests a structure like this for possessor phrases. The possessor ([DP the student]) is in the specier of DP and of course, this can be as complex a DP as we like The very hungry linguistics student by the tree with the purple owers over there.

24

CHAPTER 2. X THEORY.

DP DP D D the NP N N D s D NP N N book

student Another interesting thing is the recursive property. Since the possessor is a DP in the specier position of a DP this can go on forever. DP DP DP DP D D the NP N N D s D NP N N mother D s D NP N N brother s] D D NP N N roommate

student Another thin is the question of what to do with apparently simple NPs like John or students. Are these also DPs? According to what we just said, the subject of the sentence is always a DP as is the object of a verb or of a preposition etc. and never just an NP. Think of how we draw pronouns. since a pronoun can be the subject of a sentence, a pronoun must be part of a DP.

2.1. GENERAL STRUCTURE

25

For pronouns, however, theres some reason to believe that they actually head the DP, that is, the pronoun is a D. Consider the following:

(25) (26) (27)

You politicians are all alike. We linguists need to stick together The media always mocks us academics

These seem to have a pronoun followed by a noun inside the DP; we can make sense of thi if the pronoun is a D which can optionally take an NP complement. So in the basic case, it looks like we should treat pronouns as being of category D:

DP D D we NP N N linguists

But then what do we do with bare nouns and proper names? Students or John? For students we want to believe that it is an instance of the N category (in order to make sense of the students or we students or Johns stuedents ). But if this N is contained in a DP (the complement of a D head), where is the D? In order to maintain consistency well suppose that in bare nouns, D is present but null (it has no phonological representation, we write this as ). So, for the bare noun students we have:

26

CHAPTER 2. X THEORY.

DP

NP N N students

Proper names are even more complicated. We will assume that they are more like the pronouns than like bare nouns -the proper noun is an instance of the category D. DP D D John We have now seen the rst instance of a specier: the possessor phrase in a DP. X theory allows for only one specier (only one complement and many adjuncts). As predicted, there can only be one possesoor phrase per DP: (28) (29) The students book * The student the professors book

The structure of X also predicts that the specier should be the element furthest away from the head, outside of all adjuncts and complements: (30) (31) (32) The students big red book of poems * Big the students red book of poems * Big red the students book of poems

2.1.5

The structure of TP

TP : DP T

2.1. GENERAL STRUCTURE T : T VP

27

The subject is in the speciier of TP (SpecTP for short). Thats like saying DP daughter of TP. TP DP T

T VP A simple sentence: I left. TP DP D D I T ed TP T VP V V leave

2.1.6

CP

CP: SpecCP C C : C TP

2.1.7

Clauses

Matrix clause: A simple clause (subject, predicate) which stands on its own. It is often called root clause or main clause. Embedded clauses or subordinate clauses (self explanatory). Often (always?) CPs.

2.1.8

Finiteness

There are several dierent kinds of clauses.

28

CHAPTER 2. X THEORY. innitive verbs use to and the bare stem of the word (tenseless and without agreement) The innitival form of the verbs are nonnite and when it has tense it is called nite All matrix clauses are nite Embedded clauses can be either nonnite or nite The hallmark of niteness is the presence of tense and agreement. In English we nd a lot of zero morphology in the tense and agreement system. You walk and to walk sound the same, but the rst one has agreement (2nd person), while the second one is innite. Present tense is zero morpheme. This can be represented by in the TP There is also agreement with the subject of the clause: I am, he is, we are, she walks. Finite verbs have tense and or agreement marking (even if its ) In English having an over (non-) tense sux generally takes priority over subject agreement: I walk; he walked. I walked; he walked. The only exception is the copula to be which shows both tens and subject agreement: I am; he is. I was; you were. However, the assumption is that they are both there abstractly. Finite verbs agree with the subject and have tense morphology. Because of the zero morphology, it isnt always obvious when a clause is nonnite. Although to is a good tip-o, its not always present in a non nite clause: (33) (34) (35) I told you to eat broccoli I saw you eat broccoli I know you eat broccoli

The rst one is clearly non nite, but so is one of the other ones. How do we ush them out? The problem is that the you form (2sg; 2pl) does not show overt subject agreement. Lets try to tell by changing the 2nd person to the third: (36) (37) I say him eat broccoli I know he eats broccoli

Booom! There we go. The rst one is non nite and the second one is not.

2.1. GENERAL STRUCTURE

29

Another point to notice is the form of the pronoun: In nite clauses the masculine 3sg pronoun is he, but in nonnite clauses it is him. The dierence between he and him is a dierence in Case- Case marks the position or role of a pronoun in the structure. A pronoun in a subject position of a nite clause has nominative (subject) case: I left, he left, she left... A pronoun in almost any other position (object position; subject of a nonnite clause) has accusative (object) case: He met me; He met him: he met us. Another way to tell wether a clause is nite is to look at the complementizer, if there is one. The complementizer that always introduces nite clauses, and the complementizer for always introduces nonnite clauses. (38) (39) Johns parents wish for him to succeed. Johns parents wish that he will succeed.

A clause, nite or nonnite, must have a T node, must have a TP. In a nonnite clause the T often is where we see to. In a nite clause, T is where we see modals like should would might shall... Note that these clauses do not show subject agreement, but they are nevertheless nite (and arguably show tense distinctions should vs. shall): (40) (41) He should leave I might leave

T is also where we seem to see auxiliary verbs, have and be. Auxiliary verbs are a special kind of verbs, but they are verbs (they arent modals), and it isnt clear that they should be classied in category T rather than V. Thus, we see auxiliary verbs in T because auxiliary verbs are verbs, the head of a VP, and then they move to T.

2.1.9

Movement of auxiliaries

John is happy. The verb be starts out (abstractly) as shown here, the head of the VP:

30

CHAPTER 2. X THEORY.

TP DP D D John T [+past] V be T VP V AdjP Adj Adj happy The verb be then moves (before we pronounce it) up to T, but not if there is a modal in T (John might not be happy.). Think of this as the opposite of what happens when [past] -ed hops down from T to V to form past tense verbs. TP DP D D V+T TP T VP V V AdjP Adj Adj happy Draw an arrow from to be+[past].

John be+[past]

2.1. GENERAL STRUCTURE John has written. TP DP D D V+T TP T VP V V VP V V written

31

John have+[past]

32

CHAPTER 2. X THEORY.

Chapter 3

Binding Theory
Now lets kill Binding from the Hoopmans once and for all! Anaphors: reexive pronouns (myself, yourself, ourselves, yourselves) and reciprocals (one another, each other) Pronouns: Personal pronouns (he him it I me)

3.1

Principle A

Denition 21. C-command Node X c-commands node Y if a sister of X dominates Y. Node X thus commands her sister and all her nieces. C is meant to evoke constituent. The relation c-command is not reexive, i.e. a node cannot c-command itself. Reexives A reexive must be coreferential with another DP, its antecedent in the same sentence A reexive must agree with its antecedent in person, number and gender (42) (43) the boyi likes himselfi * the boyi likes herselfi

The DP antecedent of a reexive must c-command the reexive (44) (45) Johni believes that Billj saw himselfj * Johni believes that Billj saw himselfk No antecedent. 33

34

CHAPTER 3. BINDING THEORY The reexive and its antecedent must all be in the same XPs which have a subject. (Think of it as the reexive and the antecedent being clausemates) 1 . (46) * Johni believes that Billj saw himselfi The reexive and its antecedent are not in the same TP.

This is a fundamental result of contemporary syntactic theory. We see systematic correlations between our conclusions regard- ing constituency based on the tests discussed in chapter 3, and the geometrical predictions made by c-command. Principle A can be stated more succinctly by: An anaphor must be bound in its domain. Reciprocals Reciprocals are also anaphors in english, and thus subject to Principle A above. (47) Ourrabbitandtheneighbor scat
i

like [each other]j Ourrabbitxandtheneighbor scaty

(48)

are such that [x likes y and y likes x] recirocalanteceden A reciprocal requires an antecedent taht is plural. This antcedent must c-command the reciprocal, and it must be close enough: within all the same XPs that have a subject, exactly like reexives: 1. (49) (50) 2. (51) (52) Johni heard theirj criticism of each otherj . Johni heard theirj criticism of themselvesj * Theyi heard Johnj s criticism of each otheri * Theyi heard Johnj s criticism of themselvesi

1 Remember that antecedents do not need to be DP subjects of TP. Since DPs can occur in other postitions too: Where can we nd DPs?

the complement of a verb or a PP or the subject of a DP (like in possesors)

3.1. PRINCIPLE A

35

Variable paraphrase We have studied how anaphors (reciprocals and reexives) relate to their antecedents in English. Note that we can paraphrase the following sentences: (53) (54) (55) (56) Johni likes himselfi ? Johni likes Johni The students boysk are proud of themselvesk The students boysk are proud of the students boysk .

However, there is a more complex kind of use where antecedents are quantied DPs. They allow us to express things that really cannot be expressed n any other way. (57) (58) Everyone herei likes himselfi # Everyone herei likes everyone herei

Or worse! (59) (60) No Italian spyi betrayed himselfi # No Italian spyi betrayed no Italian spyi

The correct way to paraphrase these quantied DPs is using a variable, because they stand for a list of statements. Thus, (61) (62) (63) (64) (65) (66) Everyone herei likes himselfi If x is a person here, x likes x. For every person x, x likes x. No Italian spyi betrayed himselfi if x is an Italian spy, x did not betray x For no Italian spy x, x betrayed x

Reexives are powerful! This may be one of the reasons why languages have pronouns and reexives. However, for such cases like these, it is a bit strange to talk about coreference as in referring to the same objects as between the reexive and its antecedent. This is particularly clear in the case of the sencence No italian spy betrayed himself. What does the DP No Italian spy refer to? This is a complicated question, but lets say for now that we need to get rid of the term coreference, we will change it for boundedness, by which we mean that the anaphor has to be c-commanded by and coindexed with its antecedent DP. With this we can restate the most important conclusions of Principle A: An anaphor must be bound in its domain.

36

CHAPTER 3. BINDING THEORY A DP is bound (by its antecedent) just in the case where there is a ccommanding DP which has the same index

The domain of a DP anaphor is the smallest XP which has a subject and which has a DP c-commanding the anaphor.

We can indicate the domain of a reexive by putting a box around the nodes that are in the same XPs with a subject as the reexive. This is called the domain of the reexive. It will always be the smallest XP with a subject that contains the reexive.

3.2
3.2.1

Principle B
Pronouns

If we collect a sample of the examples considered in the previous section and change the anaphors to simple accusative pronouns, in most cases, the good sentences become bad and the bad ones become good:

(67) (68)

Maryi likes herselfi * Maryi likes heri

Pronouns appear to have a complementary distribution with anaphors because generally replacing an anaphor by a pronoun with the same antecedent yields a deviant result (or vice versa, xing a deviant anaphor binding). This is a tentative approximation, but not entirely true. The domain of a DP pronoun is the smallest XP with a subject that contains a DP. Principle B: a pronoun cannot be bound in its domain (i.e. it cannot have a c-commanding antecedent in its domain). Lack of complementarity, two trees:

3.2. PRINCIPLE B

37

TP DP D D theyi T [pres] V like DP DP each otheri D s T V DP D NP N N books Variable paraphrase to know whats going on here: (69) (70) (71) They like each others books. For all x in the group dened as they: If y is a book owned by x, then x likes y For all x,y in the group they if x owns the book, then it follows that y likes the book

The second one is more accurate, I believe. Because they like each others books does not mean they like their own books necessarily, think of they like each others books, but they despise their own. The anaphor in this case is correctly bound by a subject that c-commands it, however it is not in the locality, hence we need to think about the domain clause more: The domain of an anaphor is the smallest XP which has a subject and which has a DP c-commanding the anaphor. So in this case the anaphor each other seeks an antecedent until it reaches the subject of the clause. These formulas are needed when the antecedent of the pronoun is a nonreferential expression. Denition 22. Non-referential expression Unlike proper names or denite descriptions (co called because they describe an object and are [+denite]) such as the president of the committee or this girl, or indenite descriptions like a man I met yesterday, which in context pick out a real or imaginary object in the world, non-referential expressions such as nobody, or many people, do not.

38

CHAPTER 3. BINDING THEORY

Among non-referential expressions we nd quantied expressions (typically DPs with a D such as no or every or each, such as several people, few books, no boy, any house, each day, everyone from Singapore, who, which book about ants, and how many sandwiches.

3.3

Principle C

Non-pronominal expressions Proper names, descriptions (like the man on the corner, my brother) or epithets (like the idiot) are not pronomials. They are called R-expressions, where the R suggests independently referential. Non-pronomial expressions cannot be bound in the way anaphors and pronouns can. However, they may be coindexed with preceding pronouns: (72) (73) (74) * Hei said that Peteri took the car After you spoke to himi , Peteri took the car The builder of hisi house visited Peteri

In the rst example, He c-commands Peter, but not in the second and third examples. We can describe these observations by stating that non-pronomials cannot be c-commanded by a coindexed pronoun, i.e. non-pronomials cannot be bound. Principle C: An R-expression cannot be bound. Remember: Denition 23. Boundedness The reference of an expression is determined by a c-commanding antecedent. Denition 24. Domain Of an anaphor:The smallest XP containing a DP c-commanding the anaphor which has a subject. Of a pronoun: The smalles XP containing a pronoun which has a subject. More generally, the domain of a DP is the set of nodes included in the smallest XP with a subject that contains the DP. Theorem 2. Chomsky (1981) Principles of binding theory: Principle A. An anaphor must be bound in its domain Principle B. A pronoun must be free (= not bound) in its domain Principle C. An R-expression cannot be bound. Lemma 3. Agreement Pronouns and reexives agree with their antecedent in person, number and gender. Note: How about these examples:

3.3. PRINCIPLE C (75) (76) (77) Thati is [a bird]i Thati s [the truth]i [BobDylan]i is [RobertZimmerman]i

39

That is, two expressions can be co-referential when they are related to a verb (e.g. be) that requires it, even if there is no binding. This implies that these strings are good if there is no coindexing. Again, what we want to say here is that there is no binding, but just a kind of accidental coreference: Two expressions end up being be co-referential just because of the circumstances, even when it is not intended or assumed by the speaker. In this case, clearly, it is not a result of binding theory that the expressions are coreferential.

40

CHAPTER 3. BINDING THEORY

Part II

Lecture notes and readings

41

3.4. CHAPTER 10. THE FORMS OF SENTENCES. HOWARD LASNIK. 43

3.4

Chapter 10. The forms of Sentences. Howard Lasnik.

Some preeliminary denitions of lexical categories :2 Denition 25. A phrase structure rule is a part of the system of knowledge underlying the procedures for the analysis of structures and the production of structures. I.e. S rightarrow NP VP. represents the traditional subjectpredicate division in English. or, NP rightarrow (det 3 ) N [Parenthesis indicate an optional item.]
2 A pro-form is a type of function word or expression that stands in for (expresses the same content as) another word, phrase, clause or sentence where the meaning is recoverable from the context. They are used either to avoid repetitive expressions or in quantication (limiting the variables of a proposition). Pro-forms are divided into several categories, according to which part of speech they substitute:

A pronoun substitutes a noun or a noun phrase, with or without a determiner: it, this. A pro-adjective substitutes an adjective or a phrase that functions as an adjective: so as in It is less so than we had expected. A pro-adverb substitutes an adverb or a phrase that functions as an adverb: how or this way. A pro-verb substitutes a verb or a verb phrase: do. A pro-sentence substitutes an entire sentence or subsentence: Yes or (some have argued[1] ) that as in That is true. An interrogative pro-form is a pro-form that denotes the (unknown) item in question and may itself fall into any of the above categories. determiner is a word, phrase or ax that occurs together with a noun or noun phrase and serves to express the reference of that noun or noun phrase in the context. That is, a determiner may indicate whether the noun is referring to a denite or indenite element of a class, to a closer or more distant element, to an element belonging to a specied person or thing, to a particular number or quantity, etc. Common kinds of determiners include denite and indenite articles (like the English the and a[n]), demonstratives (like this and that), possessive determiners (like my and their), and quantiers (like many, few and several). In English: Denite determiners, which imply that the referent of the resulting noun phrase is dened specically: The denite article the. The demonstratives this and that, with respective plural forms these and those. Possessives, including those corresponding to pronouns my, your, his, her, its, our, their, whose and the Saxon genitives formed from other nouns, pronouns and noun phrases (ones, everybodys, Marys, a boys, the man we saw yesterdays ). These can be made more emphatic with the addition of own or very own. Interrogatives which, what (these can be followed by -ever for emphasis). Relative determiners: which (quite formal and archaic, as in He acquired two dogs and three cats, which animals were then...); also whichever and whatever (which are of the type that form clauses with no antecedent: Ill take whatever money theyve got). Indenite determiners:
3A

44

The indenite article a or an (the latter is used when followed by a vowel sound). The word some, pronounced [s()m] (see Weak and strong forms in English), used as an equivalent of the indenite article with plural and non-count nouns (a partitive). The strong form of some, pronounced [sm], as in Some people prefer dry wine; this can also be used with singular count nouns (Theres some man at the door). For words such as certain and other see below. The word any, often used in negative and interrogative contexts in place of the article-equivalent some (and sometimes also with singular count nouns). It can also be used to express alternative (see below). Quantiers, which quantify a noun: Basic words indicating a large or small quantity: much/many, little/few, and their comparative and superlative forms more, most, less/fewer, least/fewest. Where two forms are given, the rst is used with non-count nouns and the second with count nouns (although in colloquial English less and least are frequently also used with count nouns). The basic forms can be modied with adverbs, especially very, too and so (and not can also be added). Note that unmodied much is quite rarely used in armative statements in colloquial English. Phrases expressing similar meanings to the above: a lot of, lots of, plenty of, a great deal of, tons of, etc. Many such phrases can alternatively be analyzed as nouns followed by a preposition, but their treatment as phrasal determiners is supported by the fact that the resulting noun phrase takes the number of the following noun, not the noun in the phrase (a lot of people would take a plural verb, even though lot is singular). Words and phrases expressing some unspecied or probably quite small amount: a few/a little (learners often confuse these with few/little), several, a couple of, a bit of, a number of etc. Cardinal numbers: zero (quite rare as determiner), one, two, etc. In some analyses these may not be treated as determiners. Other phrases expressing precise quantity: a pair of, ve litres of, etc. Words and phrases expressing multiples or fractions: half, half of, double, twice, three times, twice as much, etc. Those like double and half (without of) are generally used in combination with denite determiners (see Combinations of determiners below). Words expressing maximum, sucient or zero quantity: all, both, enough, sucient, no. Note that many of these quantiers can be modied by adverbs and adverbial phrases such as almost, over, more than, less than, when the meaning is appropriate. Words that enumerate over a group or class, or indicate alternatives: each, every (note that every can be modied by adverbs such as almost and practically, whereas each generally cannot) any (as in any dream will do; see also under indenite determiners above), either, neither Personal determiners: The words you and we/us, in phrases like we teachers; you guys can be analyzed as determiners. Examples: As all we teachers know . . . Us girls must stick together. (informal) These examples can be contrasted with a similar but dierent use of pronouns in

3.4. CHAPTER 10. THE FORMS OF SENTENCES. HOWARD LASNIK. 45 The structure that the two examples determine for The man left can be represented as follows: S NP det N VP V

The man left VPs can also contain NPs, as in The man solved the problem. So then we have another phrase structure indicated by VP rightarrow V (NP) [again, the parenthesis indicates that it is optional]. We can also embed one sentence inside another: The child thinks the man solved the problem. An additional VP phrase structure rule immediately allows forthis possibility: VP rightarrow V S [S stands for sentence.] This process can be continued indenitely 4 . Denition 26. A phrase structure rule that can be applied indenitely are called recursive. By providing nite means for generating an unlimited number
an appositional construction, where the use of other pronouns is also permitted but the pronouns cannot be preceded by the (pre-) determiner all. Examples: I/we, the undersigned, . . . , We, the undersigned, . . . , but not All we, the undersigned, . . . Other cases: The words such and exclamative what (these are followed by an indenite article when used with a singular noun, as in such a treat, what a disaster!) Noun phrases used as determiners, such as this colour and what size (as in I like this colour furniture; What size shoes do you take?) Words such as same, other, certain, dierent, only, which serve a determining function, but are grammatically more likely to be classed simply as adjectives, in that they generally require another determiner to complete the phrase (although they still come before other adjectives). Note that the indenite article in combination with other is written as the single word another.
4 Note that not all verbs can be substituted for think or solve. This requries that the large lexical category V is divided into smaller lexical subcategories, each with its own special privileges of occurrence. We can express the properties of the subcategories as: sleep <> solve <N P > think <S . Which means that sleep can be inserted only into a VP with no direct object (intransitive verb), solve requires a direct object to ll out its VP (solve what, hence a transitive verb) or think (also a transitive verb, but it takes an S not an NP as a complement)

46 of sentences, recursive rules supply a crucial part of the answer to the fundamental question of creative use of language (nite capacity and innite number of grammatical sentences). Denition 27. In transformational grammar we think that each senctence has two representations: The rst one is the deep structure of the sentence in which the thematic relations and the closely connected subcategorization properties of the sentence are captured; and the second one, the surface structure in which the order of the parts are actually pronounced. Example: This problem, John solved. Is grammatical, however, it does not conform to the rules of solve as a transitive verb that needs a noun phrase following it. However, This problem is actually solveds NP, only it does not follow the word, but is displaced in the sentence. Hence, the deep structure of This problem, John solved. is still: S NP N V VP NP N

John solved det

this problem But its surface structure is now: S NP det N NP N V S VP t

this problem

John solved Where the t, called trace is meant to indicate the position from which movement took place. Denition 28. An operation relating one phrase structure representation to another phrase structure representation is known as a transformation. Transformational analysis provides a means for capturing the felt relatedness among sets of sentences.

3.4. CHAPTER 10. THE FORMS OF SENTENCES. HOWARD LASNIK. 47 Denition 29. A sequence of transformationally related phrase markers beginning with a deep structure and culminating in a surface structure is called a transformational derivation. Denition 30. The transformation that relates the structure in both phrase structure trees above is called topicalization. It displaces an item and attaches it to another constituent by creating a higher instance of the target category from which both the moved item and the target category hang. This particular sort of attachment is called adjunction. Another example of transformational derivation is wh-movement. As in Which problem will John solve? Where which problem serves as the direct object of solve but precedes it. The deep structure of Which problem will John solve? would be then: [.S [.N John ] [.T will ] [.VP [.V solve ] [.NP [.det this ] [.N problem ] ] ] ] However, not that in the deep structure the order of John and the auxiliary verb will is inverted. This is called interrogative inversion. Examples of interrogative inversion Susan must leave. Must Susan leave? Harry can swim. Can Harry swim? 5 Bill is sleeping. Is Bill sleeping? However there is no structurally independent rule to make this transformations that always works. In fact, it relies on structured hierarchical organization. Theorem 3. Interrogative inversion process-structure dependent rule: Beginning with a declarative deep structure, move the rst auxiliary verb following the subject to the front to construct an interrogative.
5 The

verbs customarily classed as modals in English have the following properties:

They do not inect, except insofar as some of them come in presentpast (presentpreterite) pairs. They do not add the ending -(e)s in the third-person singular (the present-tense modals therefore follow the preterite-present paradigm). They are defective: they are not used as innitives or participles (except occasionally in non-standard English; see Double modals below), nor as imperatives, nor (in the standard way) as subjunctives. They function as auxiliary verbs: they modify the meaning of another verb, which they govern. This verb generally appears as a bare innitive, although in some denitions a modal verb can also govern the to-innitive (as in the case of ought). They have the syntactic properties associated with auxiliary verbs in English, principally that they can undergo subjectauxiliary inversion (in questions, for example) and can be negated by the appending of not after the verb. The following verbs have all of the above properties, and can be classed as the principal modal verbs of English. They are listed here in presentpreterite pairs where applicable: can and could may and might shall and should will and would must

48 Declarative: S NP The S VP can swim

man who is here Interrogative: (By thm. above.) S M Can NP NP S t V swim

the man who is here However, there is still another thing here that is left unresolved. What happens to sentences like: Mary sleep.6 ? To make the interrogative inversion you need the verb do Does Mary sleep?7 . There must be a way to formalize the intuition that this is the same thing that is happening in the other examples i.e. The man who is here can swim and Mary has read the book. To do this, think of the sentence Mary will sleep. As NPs are headed by nouns and VPs are headed by verbs, this sentence is headed by will, a modal auxiliary verb. From this perspective the type of S=Mary will sleep. is really AuxP. And Aux is a constituent intermediate between Aux and the maximal projection of Aux, AuxP. Then we have a new rule: AuxP NP Aux Aux Aux VP And according to this rule, the deep structure of Mary will sleep. is:

6 The verb sleep here is accompanied by the inectional ending s that carries the tense (present) and the subject (third person singular). this is called agreement information. 7 Here the inectional ending of sleeps is realized as the inectional ending of the dummy verb do.

3.5. PHRASE STRUCTURE NAOKI FUKUI

49

AuxP NP N Mary Aux Aux will VP V

sleep To derive the interrogative from this structure, will, the head of AuxP is fronted. Note that this new rule applies also to the sentence Mary sleeps. In this case, the head is the inection 8 . Thats why we use the dummy do, so that the inection can be fronted just as the case with will. AuxP NP N In Aux VP V sleep Denition 31. This transformation is called Ax hopping. This is analyzed as an instance of adjunction.

Mary present 3rd sing

3.5

Phrase Structure Naoki Fukui

Denition 32. Phrase markers Where is a single symbol and , , and are strings of symbols (of which only is necessarily non-null): . The phrase markers generated by phrase structure rules express three kinds of information about syntactic representations: 1. the hierarchical grouping of the constitutents of the structure (Dominance) 2. the ype of each constituent (labeling) 3. the left-to-right order (linear order) of the constituents (precedence)
8 since

the inection cannot stand alone here, it is called an ax.

50 For example: Phrase structure rules: 1. S NP VP 2. VP V NP 3. NP D N 4. D the 5. N boy 6. N problem 7. V solved Phrase marker (tree): S NP D N V D VP NP N

the boy solved

the problem Now all this was abolished in favor of: Denition 33. Context free phrase structure rules: . Notice now that and are necessarily null. But it became apparent that certain important generalizations about the phrase structure of human language cannot be stated in terms of phrase structure rules alone. This led to the development of X-theory (a general theory of phrase structure), which is the main topic of this paper. Denition 34. Transformations Formal operations applying to linguistic representations constructed in accordance with the general principles of phrase structure. Note that a substantive change in the theory of phrase structure necessarily has important implications for the theory of transformations.

3.5.1

What is X-Theory

Two considerations motivated it: 1. the notion of possible phrase structure rules; 2. cross categorical generalizations.

3.5. PHRASE STRUCTURE NAOKI FUKUI

51

As in (1 above) General rule: an XP cannot be a phrase of X if there is no X. That is, a Noun Phrase cannot be an NP without a noun, and a Verb phrase without a verb. This means that phrase structure in human language is endocentric in the sense that it is constructed besed on a certain element called the head of a phrase. A head determines the essential properties of the phrase. Obviously, cannot capture this, because nowhere in this rule is the specication that there is a relationship between them. We thus need some other mechanism that captures this main insight of Xtheory: the endocentricity of phrase structure. As in (2 above) we are concerned with some observed parallelisms that exist across dierent categories. Chomsky (1970), however, refutes this Transformationalist Hypothesis, and argues that the theory of grammar should not allow a nominalization transformation (or any other transformation with similar expressive power) because it performs various operations that are never observed in any other well-argued cases of transformations. 9 Denition 35. Lexicalist Hypothesis A hypothesis which entails that syntactic transformations operate on syntactic constituents only, and can only insert or delete designated elements. This means that transformations cannot be used to insert, delete, permute, or substitute subparts of words. The lexicalist hypothesis comes in two versions: (a) a weak version which says that transformations cannot be used in derivational morphology (= Weak Lexicalist Hypothesis), and (b) a strong version which says that transformations can also not be used in the domain of inection (= Strong Lexicalist Hypothesis) Chomskys explanation of X-Theory 1. X X . . . 2. X [Spec,X]X The X in 1 is a variable ranging of the class of lexical categories N(ouns), V(erbs), A(djectives), P(repositions). The symbol X called Xbar) stands for a constituent (phrase) containing X as its head as well as those elements appearing in the place indicated by ..., these elements are called the complement of X. The schema in 2 above introduces a still larger phrase X (X double bar) containing X and pre-head elements associated with X called the specier (Spec) of X (hence the notation [Spec,X]). Examples of speciers include determiners as [Spec,N], auxiliary elements as [Spec,V], comparative structures like very as [Spec,A] etc. X and X, which share the same basic properties of the head X are called projections of X with the latter (X) referred to as the maximal projection of X (since it does not project any further). The basic claims of Chomskys preeliminary presentation of X-theory a. Every phrase is headed, i.e., has an endocentric structure, with the head X
9 What does it mean to not allow nominalizations? Should they not be allowed in speech or just stricken o the list of theoretically sound transformations?

52 projecting to larger phrases. b. Heads (categories) are not atomic elements; rather, they are feature complexes, consisting of the primitive features [ N] and [ V]. c. UG 10 provides the general X-schemata of the following sort, which govern the mode of projection of a head: 1. XX . . . 2. X + [Spec, X] X

3.5.2

Class notes.

Items of the lexicon are divided into two major subtypes: lexical categories and functional categories. Problems for phrase structure (PS rules): they overgenerate in two directions. i.e. John gave on the table. (There are certain types of phrases that do not let you delete arguments.) theyre redundant (theyre in the lexicon). I.e. put is stored in the lexicon with its c-selection properties (categorical selections rquieres Dp1,DP2 and a prepositional phrase - john put the book on the table); (Projection principle) Phrase structure rules did not capture endocentricity. Syntactic parallelism across categories. Theorem 4. Projection principle All lexical information of a predicate must be syntactically encoded. X-Theory: All phrases in human language have the same structure. All phrases are headed, period. Doesnt matter if youre lexical or functional. The features of the head dictate when it becomes a maximal category. In the early days you had multiple X levels and now we think that we can have multiple XP levels. XP YP specier X 0 X ZP

Complement Chomsky(1970) talks about syntactic parallelisms. (a) John is refusing the oer. (The verb assigns accusative case to the object)
10 Universal

Grammar

3.6. CHOMSKY 1995

53

(b) The enemy is destroying the city. (The verb assigns accusative case to the object) Gerundive nominals: (a) Johns refusing the oer. (Prepositions can also assign case, but we dond need it here, so refusing is still kind of acting like a verb because it is assigning case.) (b) The enemys destroying the city. Derived Nominals (a) Johns refusal of the oer (b) The enemys destruction of the city. Look for reasons for saying that gerundives are more verby and derived nominals are more nouny. roles?

3.6

Chomsky 1995

X-Theory was an eort to resolve the tension between explanatory and descriptive adequacy. Separating the lexicon from the computations removed a redundancy between phrase structure rules and lexical properties. Chomsky discusses three levels of adequacy of a grammatical description: 1. Observational adequacy: Characterizes the behaviors somehow allowed by the grammar [e.g. set of sentences produced under perhaps idealized conditions; related concept: E-language]. The theory achieves an exhaustive and discrete enumeration of the data points. There is a pigeonhole for every observation. 2. Descriptive adequacy: Characterizes the structural descriptions allowed by the grammar [related concept: I-language] the theory formally species rules accounting for all observed arrangements of the data The rules produce all an only the well formed constructs and relations of the protocol space 3. Explanatory adequacy: Related to language acquisition. Selects among descriptively adequate theories of grammar those that derive the set of structural descriptions as:

54 (a) properties of the initial state of the child S0 (Language Acquisition Device), modied by (b) individual experience (Primary Linguistic Data) which selects among the options allowed by UG. The theory provides a principled choice between competing descriptions. It deals with the uttermost underlying structure. It has predictive power. Denition 36. Logical Form (LF) a mental representation of a linguistic expression, derived solely from Surface Structure. In the words of Chomsky, LF captures those aspects of semantic representation that are strictly determined by grammar, abstracted from other cognitive systems. Denition 37. Computational system of human language The biolinguistic project views languages as systems consisting of a computational system (Cs) and a lexicon. A computational system includes the syntactic and the semantic systems that together provide the expressive power of language. The conceptual system (lexicon) is what provides common sense understanding of the world. The Computational system of human language CHL is a concept by Chomsky coined to accompany his theory that the principles and operations of the human computational system are universal. Parameters in the system relate to the lexicon, but the operations are the same for every language. Note. Only maximal projections are relevant to LF interpretation. CHL should acces only minimal and maximal projections 11 Minimal and maximal projections are not identied by any special marking, so they must be termined from the structure in which they appear. Denition 38. Maximal Projection A category Xthat does not project any further is a maximal projection XP and one that is not a projection at all is a minimal projection. CHL may select an item from N (at the lexicon) or perform some permitted operation on the syntactic objects already formed. One such operation is necessary on conceptual grounds alone: merge. Denition 39. Merge The operation merge is one of the basic operations in the Minimalist Program, a leading approach to generative syntax, when two syntactic objects are combined to form a new syntactic unit (a set). Merge also has the property of recursion in that it may apply to its own output: the objects combined by Merge are either lexical items or sets that were themselves formed by Merge. This recursive property of Merge has been claimed to be a fundamental characteristic that distinguishes language from other cognitive faculties.
11 question:

does this have to do with constituency?

3.6. CHOMSKY 1995

55

Merge is an indispensable operation of a recursive system ... which takes two syntactic objects and and forms the new object K = {{, }}. The simples object constructed from and is the set , , so we take K to involve at least this set where and are the constituents of K . However, and are not sucient to interpret K , so K must be at least of the form K = {{, }}, where identies the type to which K belongs. Call the label of K . Then the syntactic objects that we have until now are: (SO.a) lexical items (SO.b) K = {, {, }} where and are objects and is the label of K . Objects of type (SO.a) are complexes of features, listed in the lexicon. The recursive step is (SO.b). Suppose a derivation has reached state = {, , i , ..., n }. Then application of an operation that forms K as in (SO.b) converts to = {K, i , ..., n }, including K , but not and . Note. Great example from Wikipedia Merge is a function that takes two objects (say and ) and merges them into an unordered set with a label (either or , in this case ). The label identies the properties of the phrase. Merge (, ) {, {, }} For example, Merge can operate on the lexical items drink and water to give drink water. Note that the phrase drink water behaves more like the verb drink than like the noun water. That is, wherever we can put the verb drink we can usually put the phrase drink water: I like to (drink)/(drink water). (Drinking/Drinking water) is fun. Furthermore, we typically cant put the phrase drink water in places where we can put the noun water: We can say Theres some water on the table, but not Theres some drink water on the table. So, we identify the phrase with a label. In the case of drink water, the label is drink since the phrase acts as a verb. For simplicity, we call this phrase a verb phrase or VP. Now if we were to Merge cold and water to get cold water, then we would have a noun phrase or NP with the label water. The reader can verify that the phrase cold water can appear in the same environments as the noun water in the three test sentences above. So, for drink water we have the following: Merge (drink, water) {drink, {drink, water}} Notation: Denition 40. A terminal element LI, to be an item selected from the numeration (of the lexicon?), with no parts relevant to CHL . Denition 41. A category X min is a terminal element with no categorical parts. X is a category variable that can be substituted by any of the categories.

56 XP can be NP
12

, VP, PP

13

, DP

14

, CP

15

, IP

16

, AP

17

Denition 42. A head is necessarily a terminal element. Denition 43. An X o (zero level category) is a head or a category formed by adjunction to the head X, which projects. Denition 44. The head of the projection K is H (K ). Theorem 5. If H = H (K ) and K is maximal, then K = HP . Example if N = H ({coldwater}) then N = N P . Note. We are also interested in the maximal zero-level projection of the head H (say the T head of TP with V and perhaps more adjoined). We refer to this object as H 0max . Chomsky (1970) remarks Read Chomsky 1. (a.) Johns refusing the oer (b.) the enemys destroying the city 2. (a.) Johns refusal of the oer (b.) The enemys destruction of the city 3. (a.) John is refusing the oer (b.) The enemy is destroying the city. In case (1) whats happening is you start with:
12 A phrase headed by a noun. Noun heads can take CP or PP complements, DP complements are excluded since nouns are not Case assigners. The specier position of an NP is occupied by what are generally called post-determiners. NPs are complements of DPs. 13 a phrase headed by a preposition. It usually takes a DP complement but certain types of CPs can also appear in the complement position of PPs. PPs themselves can be complements of dierent constituents such as verbs, nouns and adjectives. 14 the head of a Determiner Phrase, a closed class item taking an NP complement dening its deniteness. Feature composition: [F, N, +V ] 15 Complementiser phrase a phrase headed by one of the three complementisers that, if or for (in structures like It is important [for Jim to pass this exam] where for is used not as a preposition but as a prepositional complementiser.) The complement of a CP is an IP, the specier position is occupied by moved wh-elements or whether. 16 in traditional grammars the Inectional Phase is a phrase headed by an inectional element which can be a modal auxiliary (e.g. may, should, will), innitival to or the bound morphemes expressing tense ( ed, s) the latter undergoing Ax Lowering to form a unit with the verb. In the present approach, however, it has been argued that the head position of the IP contains only the modal auxiliaries and the (in English) invisible agreement morpheme, information about Tense can be found in an independent vP hosting innitival to, and the bound morphemes -ed and -s also appear here. The specier position of an IP is occupied by the subject (see canonical subject position), the complement of an I is usually a VP or vP (but see small clauses for an exception). IPs are complements of CPs or ECM verbs. 17 a phrase headed by an adjective. In the complement position we can nd PPs and nite and non-nite CPs. DPs and exceptional clauses are excluded since adjectives are not Case assigners. APs are complements of DegPs.

3.6. CHOMSKY 1995

57

VP DP John V V DP

refusing the oer Then you layer a DP on top and insert a determiner: So John moves up from the the subject to a determiner phrase. To nominalize refusing. Dierences between gerundives assign case. Therefore they maintain some of the properties of verbs. Things that can modify verbs vs. things that can modify nouns. The question (The lexicalist hypothesis) is where do you form these words (gerundives)? Are they done in the lexicon or in the syntax. The DM proposal is that it is formed in the syntax. Roots arent really born being verbs or nouns they do that next. The main thing about this example is that the sentences up there are very similar to the noun phrases, so Chomsky posited that they should be related syntactically and share the same properties. This is one of the things that led to X-Theory. Refuse moves to the empty spot. Possesors and external arguments share something in common. This is referred to the VP internal Subject Hypothesis : we should think of external arguments as subjects merging against VPs and falling out. X-Bar theory comes along and every syntactic phrase was the same: Anti-symmetry proposal. Substitution vs. Adjunction. Phrasal adjunction vs. Head adjunction vs. substitution operations. With adjunction you just put something there, with substitution you think that there was something there. Head movement. John baked a cake. predicate baked two arguments internal argument (a cake) external argument (John) [.VP [.DP [.D ] [.NP [.N John ] ] ] [.V [.V bake ] [.DP [. a ] [.NP [.N cake ] ]]]] For the tense what happens with tense is: [.TP [.T [.T 0 ed ] [.VP [.DP [.D ] [.NP [.N John ] ] ] [.V [.V bake ] [.DP [.D a ] [.NP [.N cake ] ] ] ] ] ] ] Heads move to heads, speciers move to speciers. Most important parts of UG: merge (external merge)

58 move (internal merge) project (what is the label thats going to project.) agree (what happens between a head and an object below. sometimes that induces movement sometimes it is not.) Denition 45. In morphology and syntax, a clitic is a morpheme that has syntactic characteristics of a word, but depends phonologically on another word or phrase.[1] The term is derived from the Greek for leaning. It is pronounced like an ax, but plays a syntactic role at the phrase level. In other words, clitics have the form of axes, but the distribution of function words. For example, the English possessives is a clitic in the phrase the king of Englands horse: It looks like a sux, but its position at the end of the king of England rather than on king is like that of a separate word.

Chapter 4

A-Movements. Mark Baltin

4.1
4.1.1

Reading notes
A-movement vs. A-bar movement

Argument movement (A-movement) displaces a phrase into a position where a xed grammatical function is assigned, such as in movement of the object to the subject position in passives: 1. Fred read the book. 2. The book was read [. . . ] (by Fred). - A-movement Non-argument movement (A-bar movement or A-movement), in contrast, displaces a phrase into a position where a xed grammatical function is not assigned, such as movement of a subject or object NP to a pre-verbal position in interrogatives: 1. You think Fred loves Mary. 2. Who do you think [. . . ] loves Mary? - A-bar movement 1. You think Fred loves Mary. 2. Who do you think Fred loves [. . . ]? - A-bar movement The A- vs. A-bar distinction is a reference to the theoretical status of syntax with respect to the lexicon. The distinction elevates the role of syntax, locating the theory of voice (active vs. passive) almost entirely in syntax (as opposed to in the lexicon). A theory of syntax that locates the active-passive distinction in the lexicon - i.e. the passive is not derived via transformations from the active - will reject the distinction entirely. 59

60

CHAPTER 4. A-MOVEMENTS. MARK BALTIN

4.1.2

Binding Theory

Here an important concept is c-command : Node X c-commands node Y if a sister of X dominates Y. This relation can be illustrated graphically with a simple tree. It is the relation between X and anything under a sister of X, here Z, which dominates everything under the triangle below. e.g. Y: i X Z ......Y ...... Also, we need the notion of strong crossover. Crossover eects are restrictions on possible binding or coreference found between certain phrases and pronouns. Crossover eects are divided into weak crossover or strong crossover. Strong crossover (SCO) refers to the restriction on possible binding or coreference found between certain phrases and pronouns, as below: 1. Whoi does hei love ti ? 2. Hei loves each boyi . In (3), the object wh-phrase who is said to have crossed over the subject pronoun he; in this conguration, the individuals picked out by he cannot covary with those selected by who (that is, he cannot be bound by the quantier associated with who). The same set of data found with quantiers, as in (4), has often been claimed to be due to a covert movement of the quantier phrase at a more abstract level of syntactic representation (such as Logical Form (LF)). There is also evidence that Focus constituents exhibit crossover eects. Binding domain : The clause containing a pronominal, anaphor or R-expression. In the tradition of generative syntax, it is argued that all kinds of nouns can be classied by the combination of two features, [anaphor] and [pronominal], features which are binary. The binding characteristics of a noun (or its corresponding empty category) are determined by the values of these features, either plus or minus. Thus, a noun that is [-anaphor, -pronominal] is a referentialexpression R-expr, such as a common noun or proper name. One that is [anaphor, +pronominal] is a pronoun, such as English he or they. [+anaphor, -pronominal] is a reexive, such as himself or themselves. The actual requirements on where a type of noun can occur are given in three conditions A, B, and C (also referred to as principles). Condition A states that an anaphor (reexive) must have a local (nearby) antecedent (expression that refers to the same entity). Thus, John washed himself obeys Condition A: the antecedent of himself, which is John, is nearby, and both refer to the person John. In contrast, *John asked Mary to wash himself is unacceptable, because the reexive and its antecedent are too far away from each other.

4.1. READING NOTES

61

Condition B states that a pronoun can have an antecedent, as long as the antecedent is not local (i.e. far away) or doesnt c-command the pronoun. Thus, John asked Mary to wash him obeys Condition B: John is the antecedent of him, and him is suciently far away; on the other hand, *John washed him, where John is intended to be the antecedent of him, is unacceptable. Condition C states that an R-expression cannot have an antecedent that ccommands it. Thus, *He asked Mary to wash John, with the interpretation that He is the antecedent of John, is unacceptable. What is a trace? A trace is an empty (phonologically null) category that occupies a position in the syntactic structure. Empirical evidence pointing to the existence of traces, independently of all theory-specic considerations, has also been presented in the literature. For example, for many English speakers, the contraction of want to to wanna is possible in some contexts, but not in others: Who does Vicky want to vote for t? Who does Vicky wanna vote for? Who does Vicky want to win? Who does Vicky wanna win? Apparently the trace comes from the transformation of the declarative to the question. Think of the dierence: Vicky wants to vote for Peter. Who does Vicky want to vote for t ? Vicky wants Peter to win. Who does Vicky want t to win? The existence of the trace blocks the possibility of contracting want to to wanna. A-movement is movement of an element to what is known as as an argument position. An argument position must be where we merge our arguments. VP DP John V V DP

likes cake

62

CHAPTER 4. A-MOVEMENTS. MARK BALTIN

TP DP John T DP John V T VP V DP

likes cake A-movement is movement to satisfy morphosyntactic features: agreement case EPP feature on T There is also A movement (do not confuse this). Look it up. They do dierent things and distinguish movement to satisfy this features than movement to satisfy other features. It could be XP movement (phrasal movement driven by these kinds of features (see above)). Think about Bindin Theroy Again BT says we have three types of DPs in the language: 1. Anaphors (self each other) [Principle A of binding theory: gotta be bound locally.] 2. Pronominales (I, you, she, we) [Principle B of binding theory: pronouns have to be free locally ] 3. R-expressions [Principle C of binding theory: gotta be free everywhere, nothing can bind me.] Baltin says that the traces left by A-movement are anaphors. And since they are anaphors they should obey principle A of binding theory: Anaphors must be bound locally by something in an A position. Spec TP. Subjects have to raise outside of the VP. John might not have liked cake. might not have are coming between the verb like and its subject argument which is evidence that the subject must be raised to a SPec TP position. Burzios generalization. Perlmutter if you have a single external agent You can passivize an unergative vs. unaccusative

4.1. READING NOTES

63

Its an intransitive predicate is an intransitive predicate whose sole argument is an external argument. Unaccusative theme patient Internal ARgument. You can passivize as long as you have an external argument. If you only looked in english you may think that you have to have two external arguments to passivize, but it is true that you can passivize in other languages when you have a single external argument. If youre going to passivize you need to have an external argument. Unergatives: Intransitive which take a single argument. Smile or dance or sing. XP X DP Determiner phrase Unnacusativity: intransitive predicate whose sole argument is an internal argument. In terms of roles a theme or a patient. i.e. The lamp broke. The subject is not an agent!! Internal arguments merge internally to the VP. Classically it is argument that thematic arguments merge as a complement to the verb. I broke the vase. VP V broke Theme DP X YP

the vase Passive movement is unacusativity again. If you look at all these diagnostics for unacusativity cross lingusitically they should hold with the passives, and in fact they do. In English you can passivized and use the resultative construction. A lot of arguments have been proposed to distinguish between unergatives and unaccusatives. RAising structure: won V do I have any internal arguments? no then Won is a v external argument? yes, John so John DP do I have any modals or auxiliary verbs? yes, you have have (a perfect construction) Then we have a tense head, (innitival to) A TP needs a subject though, so John has to raise. But John has no case, then we need our VP Seems then he have nothing else but our nite tense head,

64

CHAPTER 4. A-MOVEMENTS. MARK BALTIN

so we take our tense head and raise John again.

There are three copies of John in this sentence one.

TP DP John Tnomepp V seems DP John Tepp to PerfP Perf have DP John VP V V won T VP TP T

Note on this tree: EPP means [Epp] ? means < John > every DP has a roof (silent D head) and the arrows are missing.

When we have our raising predicates we dene scope in terms of c-command relationships.

we do not see this with th

So the john at the right moves to the left twice, but he needs to move twice because of the EPP feature.

4.1. READING NOTES

65

TP DP Someone Tnomepp V hopes DP PRO Tepp to PerfP Perf have VP DP PRO V V T VP TP T

won See handout Denition Clitic : In morphology and syntax, a clitic is a morpheme that has syntactic characteristics of a word, but depends phonologically on another word or phrase. The term is derived from the Greek for leaning.[2] It is pronounced like an ax, but plays a syntactic role at the phrase level. In other words, clitics have the form of axes, but the distribution of function words. For example, the English possessive s is a clitic in the phrase the king of Englands horse: It looks like a sux, but its position at the end of the king of England rather than on king is like that of a separate word. Resultative. In linguistics, a resultative is either an adjectival phrase indicating the state of a noun resulting from the completion of the event denoted by the verb, or a verbal construction denoting the result state of an event. This verbal construction type of resultative is usually considered part of the eld of aspect. Mary laughed hoarse. - it could be depictive (she was laughing while she was hoarse) It is ne to modify a subject of transitive or an unnergative, only subjects of unnacusatives. Pollock (dont pay too much attention) Roberts Correo de Canada. Track #: EM022736300CA Well formed trees: the output of (repeated) Merge operations, arise as a result of the interaction between the following properties:

66

CHAPTER 4. A-MOVEMENTS. MARK BALTIN 1. Lexical properties of individual atoms determine the environments in which they can occur in accordance with the Projection principle 1 . 2. Locality of selection : If an atom selects an element, it acts as a head. This head must have the selected element or a molecule headed by the element as its complement or its subject, or as the subject of its TP. Selection is local in the sense that there is a maximal distance between a selector and what it selects. 3. n-ary branching : in syntactic trees, a mother node can have many daughters because Merge is an n-ary function: it can mege several trees into a single new one. In practice, there are no more than three or four daughters to a single mother. 4. X-bar theory : The hierarchical organization of syntactic trees is conjectured to be constrained by X-bar theory: across categories, a head X takes zero or more complements, zero or one subjects, and zero or more adjuncts. The head X forms a constituent with its complements: X. X in turn forms a consitutent with its subject XP. XP can form XP consituents with its adjuncts. C-command : A B H J D C E F G A c-commands nothing (it dominates everything) B c-commands C, D, E, F, G but not A C c-commands only B, H, J D c-commands only E, F, G E c-commands only D F and G c-command (only) each other H and J c-command (only) each other
1 Properties

of lexical items must be satised.

4.2. HEAD MOVEMENT POLLOCK Practice: Maryi s pictures of herselfi surprised Bill. TP DP DP Mary D s D NP N N pictures T ed V V surprised T VP NP Bill

67

4.2

Head Movement Pollock

AgrP Agr

DP

Agr0 VP

V0 l Jean Why does the negative particle not block adverb movement in English but not in other languages? How can auxiliaries raise if English has a head? But what is an AgrP? Its a functional phrase that is involved in agreement and case (case and agreement are completely syntactic, ) Because this head (AgrP) is not doing anything lexical for us, lets say that they dont exist. Denition 46. M-command If X and Y are two nodes in a syntactic tree, X m-commands Y if and only if: -X does not dominate Y, -Y does not dominate X, and -the maximal projection of X dominates Y. Phrases can move a lot longer. This is movement of a maximal projection

68

CHAPTER 4. A-MOVEMENTS. MARK BALTIN

to a specier. This is Amovement. Were not violating minimality because it is the frist phrase that can have [+WH]. If youre a Head movement why NPI is a semantic notion. Here is a place where head movement has an eect (semantic eect). Not every boy can make the basketball team. Elements: Negation Not quantier every can possibility modal CP DP Who T Perf T has C C DP john T Perf T Perf has has DP John v kissed V VP DP PerfP vP TP

kissed who

Appendices

69

Appendix A

Sample Trees
S NP subji V verb VP NP ti

71

72

APPENDIX A. SAMPLE TREES

Appendix B

Theta Roles
Thematic roles It is often convenient to identify arguments of (Fregean) predicates in terms of the following thematic roles, which are illustrated below. agent instrument cause experiencer recipient path location measure theme Agents are arguments that bring about a state of aairs. The line between agents, on the one hand, and instruments or causes, on the other, can be fuzzy, but agents are (or are perceived to be) conscious or sentient, in a way that instruments or causes arent. Some examples are given in (1)-(3). (1) a. Agent: The lions devoured the wildebeest. b. The boys caught some sh. c. My mother wrote me a letter. (2) a. Instrument: This key opens the door to the main oce. b. They must have used indelible ink. (3) a. Cause: Hurricane-force winds demolished much of the town. b. An epidemic killed o all of the tomatoes. c. An economic downturn put thousands of workers out of work. Experiencers are arguments that undergo a sensory, cognitive, or emotional experience. (4) a. Experiencer: The rhesus monkey had never seen snow before. b. Many people fear snakes. c. Their resourcefulness struck her as admirable. Recipients are arguments that receive something (whether good or bad) in a situation. (5) a. Recipient: They gave the workers a raise. b. He spared me his usual sob story. c. I paid my landlord the rent. Recipients can be the endpoints of paths. (6) a. Path: Id like to send this package to my sister. (my sister = recipient) b. Lucky raced across the lawn to the edge of the forest. c. We drove the scenic route. Locations are simply places; like recipients, they can serve as endpoints of paths. (Chapter 7 contains some discussion concerning the dierence between recipients and locations). (7) a. Location: We put the book on the shelf. b. Id like to send this package to France. (to + France = path) c. Lucky raced across the lawn to the edge of the forest. (to + the edge of the forest = path) 73

74

APPENDIX B. THETA ROLES

Measure or amount arguments express extension along some dimension (length, duration, cost, and so on). (8) a. Measure: They rowed for three days. b. The book costs ten dollars. Finally, the thematic role of theme is something of a catch-all. According to one denition, theme refers to an argument undergoing motion of some sort, including motion in a metaphorical sense, such as a change of state. As is usual in the syntactic literature, we will also use the term for arguments that are most aected in a situation or for the content of an experience. (9) a. Theme: The lions devoured the wildebeest. b. This key opens the door to the main oce. c. Hurricane-force winds demolished much of the town. d. They gave the workers a raise. e. Id like to send this package to France. f. Many people fear snakes.

Appendix C

Tables of concepts
Category Open class Noun Verb Adjective Adverb Closed class categories Preposition Determiner Numerals Complementizers Auxiliaries Modals Coordinators Negation/armation that: if: for: whether: C C C C free free, free free Abv. N V A Adv P D Num C V v or M Coord Neg/A on, of by through into from for to with the a this some every one two three that if whether for have be do will would can couls may might shall should and or but no not too so Examples

[+tense] [+tense] [+tense]

[-q] selects nite +tense TP complement [+q], selects nite +tense TP complement [-q] selects nonnite tense TP complement [+q] selects TP complement

75

76

APPENDIX C. TABLES OF CONCEPTS

Appendix D

Paper ideas
D.1 OP-Ed vs. News: Where does Congressional debate stand?

Take the corpus of newspaper articles. Use the topics of political news and opeds. Now, try to identify any syntactic regularity that is in the OP-ED section that is not in the news section (or vice-versa? Im not sure if this works). Now look at congressional debate, which one is it more alike to? Is it more like Op-Ed writing or like political reporting? Perhaps something having to do with minimal pairs? Caveats: 1. You have to take at face value that op-eds are meant to convince and political news are meant to inform. Your conclusion will only be as good as this claim. 2. There may not be any syntactic regularity. You have read one chapter of a syntax book, and already you know that regularity is friggin hard to nd. George Lako Livia Polanyi

77

78

APPENDIX D. PAPER IDEAS

Appendix E

Grammatical categories
Category Aspect Denition For verbs. Expresses a temporal view of the event or state expressed by the verb.

Types Imperfective, per

Tuesday

9C

19C

Wednesday

10C

21C

79

80

APPENDIX E. GRAMMATICAL CATEGORIES

Appendix F

Documentation for Qtree

A Documentation for qtree, a L TEX tree package1 by Jerey Mark Siskind, with a front end by Alexis Dimitriadis

The qtree package consists of QobiTree, a package of tree-drawing macros written by Je Siskind, and a front end that allows trees to be specied in bracket notation, using whitespace to separate tokens. Tree nodes, which can have labels of any size or complexity, are automatically arranged on the page, usually with quite good results. Provisions exist for ne-tuning the default layout. The front end also centers trees (by default) and provides some other nice features. A simple tree may look like this, \Tree [.S This [.VP [.V is ] \qroof{a simple tree}.NP ] ] which produces: S This V is VP NP a simple tree

The node labels in trees may be quite complicated; they may contain font changes and math-mode text, line breaks introduced with \\ (which produce
1 Thanks to Je Siskind for permission to distribute the QobiTree code. Please direct comments to Alexis Dimitriadis, [email protected].

81

82

APPENDIX F. DOCUMENTATION FOR QTREE

centered lines), etc. The trees produced have a maximum depth of 20, with a maximum of ve branches at any one node. Unlike many other tree macros, qtree automatically adjusts for the width and height of tree labels, and is pretty good at arranging nodes on the page. Trees are dened using a version of the bracket notation familiar to linguists. Tree elements are delimited by white space; braces can be used to create multiword labels. Qtree does not rely on \catcode changes for its operation, allowing trees to be included in footnotes and other moving environments without problems.

F.1

Invocation

A Qtree.sty is a L TEX package designed to be installed in a directory of style A les, and included with the L TEX 2 command \usepackage{qtree}. A Postscript specials Qtree relies on L TEXs picture environment to draw the trees. Because this environment is rather limited, the lines used to draw trees look better if qtree is used with the package eepic.sty, which provides enhancements to the picture environment. This version of qtree will automatically include eepic.sty if it can nd it, but automatic inclusion can be suppressed by use of the package option [noeepic]. This may be necessary if the le will be processed with a driver that does not understand PostScript specials (e.g., pdatex ). In that case qtree will be included as follows:

\usepackage[noeepic]{qtree} Tree centering Trees are centered by default, but you can turn centering o with the command \qtreecenterfalse. Normally this would be invoked in the preamble, but it is possible to turn tree centering o and on (with the corresponding \qtreecentertrue) at any point. These commands obey normal scoping rules. If used inside, say, an example environment, their eect will only apply until the end of that environment. There is no provision for automatically right-adjusted trees.

F.2

How to convert a tree to brackets

Reading or writing a complex tree in bracket notation is not terribly easy for humans; it helps to have an editor that can show matching braces as they are typed in. The procedure described here should allow you to easily convert a tree to bracket notation. If you dont have any diculty with this, just skip this section and do it any way you want! 1. Draw the tree you want to enter on a piece of paper, so you can look at it.

F.3. USAGE AND FEATURES

83

2. Imagine that the tree is a large peninsula, and your pencil is a boat sailing around it. Starting just to the left of the root node, move downwards, following the outline of the tree until you come back to the root node (on the right side, having moved counterclockwise around the tree), without crossing any of the trees lines. 3. (a) Every time you are at the left side of a non-terminal node, type a left bracket, and the label for that node. (b) Every time you are at a leaf node, type in the contents of that node. (c) Finally, every time you are on the right of a non-terminal node, type a right bracket (and the node name again, if you want to help keep them straight). Its dicult to show all this without including a picture, but consider the following tree; in the variant on the right, the numbered subscripts show the order in which the brackets and labels are written. A B C D E [1 A ]13 [2 B ]9 [10 E ]12 three11

three [3 C ]5 [6 D ]8 one4 two7

one two

Accordingly, we would generate the tree by typing the following: \Tree [.A [.B [.C one ] [.D two ] ].B [.E three ] ].A

F.3

Usage and features

Syntax The qtree front end reads a tree description written in the familiar (to linguists) bracket notation. Tree labels are delimited by whitespace. To make a multi-word node label, enclose it in braces; note also that TEX discards the spaces immediately after control sequences (commands whose name consists of a backslash followed by letters), hence if a node label ends with a control sequence, like \ldots in the following example, you need to enclose it in braces too. CP Spec(CP) which car ...

84

APPENDIX F. DOCUMENTATION FOR QTREE

\Tree [.CP [.{\sc Spec}(CP) {which car} ] {\ldots} ] Label matching For convenience, a label for a non-terminal node can be written either after the left bracket or after the right bracket corresponding to that node. Thus the following are equivalent: \Tree [.S when [.NP the cat ] sleeps ] \Tree [.S when [ the cat ].NP sleeps ] To help keep braces matched when editing large trees, the front end allows the option of writing a label after both the left and the right bracket of the same node, as shown for the node NP below. In this case the two labels provided must be identical, token for token. \Tree [.S when [.NP the cat ].NP sleeps ] No labels Sometimes we want to draw an abbreviated tree without a label at every intermediate node. Qtree now draws such trees properly, as in the following example. CP Spec(CP) C0 Comp(CP) \Tree [.CP Spec(CP) [ C^0 Comp(CP) ] ] Roofs It is possible to draw a triangular roof above a phrase that is treated as a unit. (See example on page 85). This is done with the command \qroof, which can appear anywhere a leaf can appear. The slope of the roof is equal to the ratio \qroofy / \qroofx (these counters may be reset to any pair of integers between zero and six; the default is 1/3). To create a roof labeled NP over the phrase the book, write \qroof{the book}.NP If the phrase contains line breaks introduced with \\, the resulting lines are ush left, not centered. Again, it is possible for the phrase to be a construction of arbitrary complexity; but the roof is implemented as a leaf node (it is not part of the original QobiTree), and so the syntax of \qroof does not allow further branches of the tree to appear under the roof.

F.4. TREE PLACEMENT

85

Subscripts, superscripts and primes Trees are constructed in a special environment in which things like NP_i, N^0, automatically format their subscripts or superscripts in math mode, giving NPi and N0 , respectively. The command that arranges this is called \automath, and can be enabled outside the tree environment, if desired. (It is turned o with \noautomath). This feature relies on \catcode changes for its operation; in trees that appear in footnotes or oats, all sub- or superscripts must be explicitly placed in math mode, as you would ordinarily do. As a further convenience, constructions like X$$, producing X, can be abbreviated X\1. (If you simply type X you get X, with an apostrophe rather than a prime). There is also X\2, producing X, and X\0, producing X0 . These commands also arrange for subtle improvements in the centering of labels that use them. Here is an example using some of these features: IP NPi Roses I0 are ti V0 going I VP V PP out of style

\Tree [.IP [ Roses ].NP_i [.I\1 [ are ].I\0 [.VP t_i [ [ going ].V\0 \qroof{out of style}.PP ].V\1 ].VP ].I\1 ] Granted, by the time the examples get this big, the bracketed format isnt all that readable, but its certainly no worse than any other tree format, and you can add white space to make it a little better.

F.4

Tree placement

Numbered examples etc. A tree generated with qtree can be placed in a numbered example environment, in \parboxes, inside math formulas, tables, pictures, etc. The tree nodes can also contain arbitrarily complex material although, unfortunately, it is not possible to embed a recursive call to qtree.

86

APPENDIX F. DOCUMENTATION FOR QTREE

For hard-to-explain reasons, trees often appear farther to the right than is visually appealing; but not to worry, you can move them sideways by hand. (Note the \hskip in the next example, which moves the tree 0.5 inches to the left). Side by side trees Multiple trees, or text and trees, can be arranged side by side. This can generally be done by just arranging commands one after another; it usually helps to turn o tree centering. If necessary the positioning can be adjusted with \hskip. (2) a. S NP el he V VP V NP lo it NP a-mi me A b. S T B cc note D

hizo decir made say

\begin{enumerate} \qtreecenterfalse \item[(2)] a. \hskip -0.5in\Tree [.S [.NP \el\\he ] [.VP [.V hizo\\made ] [.V decir\\say ] [.NP lo\\it ] [.NP a-mi\\me ] ].VP ] b. \Tree[ A [.T {B\\ \em note} cc ].T D ].S \end{enumerate}

F.5

Advanced features

Escaping the parser There is provision for sneaking directives past the front end. If a word begins with an exclamation mark, the next word (i.e., up to the next space) will be passed through unchanged, except for stripping o the !. (Braces should be used to pass through larger groups). This is mainly useful for the manual width-adjustment directives \faketreewidth and \qsetw, described below. Note that \qroof should not be preceded by an exclamation mark. Fine tuning The command \qsetw{<length>} (where <length> might be 0.5in, 36pt, etc.) tells QobiTree to override its default calculation of the width of the just-nished node (thats the node ending just to the left of where the directive was issued), and instead consider that width to be <length>. Similarly, \faketreewidth{<text>} sets the width of the last node to be equal to the width of <text> (which again can contain \\ commands etc.) <text> is not

F.6. HOW DO I . . . ?

87

actually typeset but is used just to compute the fake width of the node on the top of the stack. For example, the default placement rules would produce tree (a) below. By setting the width of the subtree headed by B to 1cm, we get tree (b). a. A B a b c d C B a b c d b. A ! C

\begin{center} \qtreecenterfalse a. \Tree [.A [ a b c d ].B C ] \hfil b. \Tree [.A [ a b c d ].B !\qsetw{1cm} C ] \end{center} When you use \qsetw or \faketreewidth you are on your own. They can either shrink or enlarge the space taken by the node and may result in trees with overlapping labels. The low-level interface The guts of qtree are the tree macros written by Je Siskind, named QobiTree. Using the original interface (which is still accessible with this package) the example tree shown on page 86 would be written like this: \begin{center} \leaf{A} S \leaf{B\\ \em note} \leaf{cc} \branch{2}{T} \leaf{D} A T D \branch{3}{S} B cc \qobitree note \end{center} These macros operate like a stack machine. You push TEX boxes onto the stack of tree nodes, then you pop them o to make branching nodes which get pushed back on the stack.

F.6

How do I . . . ?

Make my tree t in the page? Try one or more of the following: reduce the surrounding font size with \small or another size command, before you begin the tree; reduce the amount of space between subtrees with \qsetw or \faketreewidth; consider placing your tree sideways in the page, with one of the packages that provide rotation commands.

88

APPENDIX F. DOCUMENTATION FOR QTREE

Draw movement arrows from one node of a tree to another? Use Emma Peases tree-dvips package. Despite its name it is not a very convenient tool for creating trees, but its many line- and arrow-drawing commands can be used to decorate trees drawn with qtree. Use qtree with pdatex ? Pdatex does not support the PostScript specials generated by the package eepic.sty, which qtree loads automatically. At present, you have the following non-ideal options: 1. Do not use pdatex; generate a PDF le by using a PostScript-to-PDF converter. The disadvantage of this solution is that the slideshow and hypertext capabilities of PDF are not available with the resulting les. 2. Suppress the automatic inclusion of eepic, by using the package option [noeepic]. This unfortunately results in lower-quality graphics, but is probably your best option if you need to use pdatex. (You will also have to do without tree-dvips if you adopt this option). The ideal solution would be to develop a PDF driver for eepic, or for some other extention to the picture environment. Please let me know if you know of such a thing. Line up the text from all the leaf nodes on one horizontal line? As far as I can tell, qtree s design is incompatible with this style of tree. Id love it if there was an easy way to give qtree this capability (or the next one), but if there is, I havent gured it out. Draw dashed or dotted branches between certain nodes? Again, I cant see any way to incorporate this functionality into qtree, given the syntax of the front end. You can fake it to some extent, by creating lines that are part of a node as far as qtree is concerned, but which look like branch lines.

F.7

Troubleshooting

Disclaimer: This package is distributed in the belief that it is useful in its present form. I welcome any comments or reports of other problems or desirable features. But as usual, no guarantees, promises, etc. can be made about the present or future state of this code. The following problems are not really the fault of qtree, but fortunately they have easy solutions. Some very short lines are not drawn This problem appears to be caused A by the limited inventory of line slopes in the L TEX picture environment. For example, the tree fragment [.X a b ] will produce invisible branch lines from X to a and b, but the lines will reappear if the labels are made wider. Install the picture enhancement styles (eepic.sty), and the problem will go away.

F.8. INSPIRATION

89

Qtree will not work with journal style X Any number of things could be going wrong, of course, but start by checking if the journals style redenes the tabular environment. Qtree makes internal calls to tabular, so this is a frequent source of problems. Usually the styles writer has saved the original denition of \tabular under a dierent name, so all you need to do is arrange for the original denition to be restored for the calls to \Tree. There is now a hook to make this easier: If you dene a command named \qtreebugfixhook, it will be implicitly called by \Tree, with local scope (so that any redenitions it causes are automatically cancelled at the end of the call to \Tree). For example, the JNLE style (nle.sty) saves the commands to begin and end a table as \oldtabular and \endoldtabular, respectively, and the replacement macros result in r e a l l y w i d e trees. The following will restore the original denitions for calls to \Tree only. \def\qtreebugfixhook{\let\tabular=\oldtabular \let\endtabular=\endoldtabular} Kluwers house style saves the original denitions as \klu@tabular and \klu@endtabular, so to use qtree with it, do the following. (You need the \makeatletter call to use commands that contain an @-sign). \makeatletter \def\qtreebugfixhook{\let\tabular=\klu@tabular \let\endtabular=\klu@endtabular} \makeatother

F.8

Inspiration

Here is a last demo, illustrating some of the things you can do with qtree. (This example, and parts of the above exposition, were adapted from the original documentation for QobiTree). The cup slid from John to Mary. GO(cup, [Path FROM(John), TO(Mary)]) IP Fracture The cup cup NP Fracture cup The cup Nspec N
\def\CUP{{\bf cup}}

slid from John to Mary GO(x, [Path FROM(John), TO(Mary)]) IP Fracture . ! . . . . .

90

APPENDIX F. DOCUMENTATION FOR QTREE

\def\Nspec{N$_{\mbox{\sc spec}}$} \Tree [.{{\em The cup slid from John to Mary.}\\ {GO(\CUP, $[_{\rm Path}$ FROM({\bf John}), TO({\bf Mary})])}\\IP} [.\fbox{Fracture} [.{{\em The cup}\\\CUP\\NP} [ {{\em The}\\$\bot$\\\Nspec} {{\em cup}\\\CUP\\N} ].\fbox{Fracture} ] [.{{\em slid from John to Mary}\\ {GO({\it x}, $[_{\rm Path}$ FROM({\bf John}), TO({\bf Mary})])}\\IP} [ $\vdots$ $\vdots$ !\faketreewidth{WWW} ].\fbox{Fracture} ] ].\fbox{Fracture} % repeated label ]