Advanced Syntax: Lecture Notes

Citation preview

Anagnostopoulou/fox

Page 1

Advanced Syntax ; Spring 2007

Logical Form and Covert Movement 1. Cross Linguistic Variation in the form of wh-questions (1)

a. English (One wh-phrase moves overtly): Who do you think t gave what to whom? b. Bulgarian (All wh-phrases move overtly):

Koj kakvo na kogo dade?

Who what to whom gave?

'Who gave what to whom?'

c. Japanese (No wh-phrase moves overtly):

John-ga Mary-ni nani-o John-NOM Mary-DAT what-ACC 'What did John give to Mary?'

ageta no?

gave Q

How should we characterize thes cross-linguistic variation? Possibilities: 1. Same LFs; different phonological realizations. 2. Different LFs and different phonological realizations. If 1, what might be universally true about the LF of a wh-question? 2. The Study of Covert Movement Preliminary motivation A. Provides us with a relatively simple semantic treatment of quantifiers and wh-words B. Accounts for Inverse Scope C. Allows us to make sense of certain cross-linguistic variations Potential Predictions (structure diagnoses) A. B. C. D.

Island Conditions Conditions on Ellipsis (Antecedent Contained Deletion) Binding Theory Parasitic Gap licensing

In the ideal world, the type of movement that we will need to postulate for semantics will be diagnosed by A-D, which will also correlate with each other. For example, whenever Island Conditions will show us that there is no covert movement, the same will hold for Binding Theory and Antecedent Contained Deletion.

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Anagnostopoulou/fox

Page 2

Advanced Syntax ; Spring 2007

3. Some Semantic Background (2)

a. Some girl is tall. Is true iff ∃x (x∈girl)&(x∈tall) b. Every girl is tall.

Is true iff

∀x (x∈girl) Æ (x∈tall)

(3)

a. Some is a relationship between a set A and B that holds iff A∩B≠φ. b. every is a relationship between a set A and B that holds iff A⊆B.

(4)

How are the arguments A and B of a quantifier, Q, determined based on the syntactic structure?

In the cases in (2) the A is the sister of Q, and B is the sister of the QNP, QA. (5) Q NP (is) tall (6)

Simplest Answer: The arguments of Q are always its sister, NP, and the sister of QNP. (An argument of a function f is always its sister.)

Problem: (7)

I climbed every tree. Is true iff ∀x (x∈tree) Æ (x∈{y: I climbed y})

(7)’ I climbed every tree (8)

How is the second argument of every in (7)’ (the argument of every tree) determined?

The second argument of every seems to refer to the set {y: I climbed y}, but how is that set going to be the meaning of a linguistic constituent?

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Anagnostopoulou/fox

Page 3

Advanced Syntax ; Spring 2007

Our answer to (8) should also give us an account of the ambiguity of the following: (9)

A boy climbed every tree. a. Is true iff

∃x(x∈boy) & (x∈{climbed every tree})

b. Is true iff

∀x(x∈tree) Æ (x∈{a boy climbed x})

(10) a. How does a QNP find its argument when its sister is not a one place predicate (i.e., a set, e.g., when the QNP is generated in object position)? b. How are arguments determined in constructions that involve multiple quantification so as to account for scopal ambiguities such as the one exemplified in (9)?

4. The Relevance of Movement (11)

a. I decided to talk only to John. b. Only to John did I decide to talk ___________.

‘Only John is an individual x such that I decided to talk to x’

(12)

a. I demanded that you read not a single book. b. Not a single book did I demand that you read______________. ‘There is not a single book, x, such that I demanded that you read x.

(13) a. John talked to who Æ Who did John talk to___? ‘which is the person, x, such that John talked to x.’ b. John knows to talk to who Æ 1. John knows who to talk to __? ‘John knows who is the person, x, such that he ought to talk to x?’ 2. Who does John know to talk to __? ‘Who is the person, x, such that John knows he ought to talk to x?’ Predicate Abstraction: If A domiantes α1, and is merged with α1, A is interpreted as the appropriate set: Very informally: Merge(α1, A) is interpreted as [[α]]({x:A[x/1]}). 5. Covert Movement, if it were possible, would answers our questions in (10) (14)

I demanded that you read not a single book. LF1. LF2.

I demanded [that [Not a single book]1 [you read t1]]

[Not a single book]1 [I demanded that you read t1]

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Anagnostopoulou/fox

(9)

Page 4

Advanced Syntax ; Spring 2007

A boy climbed every tree. LF1

[every tree]1 [a boy climbed t1]

LF2

[A boy]2 [[every tree]1 [t2 climbed t1]]

The relevance of the VP internal Subject Hypothesis: Allows Surface Scope to be derived by fiewer instances of movement

(9)'

a. Surface Scope [IP [A boy]2 [VP [every tree]1 [VP t2 climbed t1]]] b. Inverse Scope [IP [every tree]1 [IP [A boy]2 [VP t2 climbed t1]]]

5. Evidence for QR Preliminary motivation A. Provides us with a relatively simple semantic treatment of quantifiers and wh-words B. Accounts for Inverse Scope C. Allows us to make sense of certain cross-linguistic variations Potential Predictions (structure diagnoses) A. B. C. D.

Island Conditions Conditions on Ellipsis (Antecedent Contained Deletion) Binding Theory Parasitic Gap licensing

6. Properties of movement (Islands) (15)

a. A (#different) student [[likes every professor] and [hates the dean]] (∃ > ∀)

b. c.

*(∀ > ∃)

*Guess who John likes and hates the dean? *Guess who likes who and hates the dean?

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Anagnostopoulou/fox

Page 5

Advanced Syntax ; Spring 2007

7. Structure Diagnostics Ellipsis Parallelism: An elided VP, VPE must be identical to an antecedent VPA at LF. (16) First I saw the man with the binoculars and then you did . Ellipsis as a diagnostic for structure; Antecedent Contained Deletion (ACD) (17)

a. I [VPA read every book you did ] b. I [VPA wanted to read every book you < VPE wanted to read>].

AT SS VPE and VPA are not identical, but at LF they could be if there is cover movement. (18) a. [every book you did

] I [VPA read t] b. [every book you < VPE wanted to read t>]

I [VPA wanted to read t].

The Sag/Larson and May Generalization: The QP that contains the elided VP has scope outside the antecedent VP. (19) a. I want a book. b. I want a book you do. (20) a. John’s mother wants what he wants. b. John’s mother wants what he does. (21) a. John refused to read EVERY book that we thought he HAD . He was willing to read only some of them. b. John refused to read EVERY book that we thought he should . #He was willing to read only some of them. 8. Structure Diagnostics: Binding Theory and Parasitic Gaps 8.1. Binding Theory Condition C (22) ??/*Someone introduced himi to every friend of John'si. Æ [every friend of John'si] Someone introduced him to t. Condition A (23) ??John and BIll said that [Mary bought every picture of each other/themselves]. (24) a. John knows which picture of himself Mary bought.

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Anagnostopoulou/fox

Page 6

Advanced Syntax ; Spring 2007

b. ??John knows which girl bought which picture of himself. 8. 2 Parasitic Gaps (25)

a. Which book did you file without reading? a. *Who filed which book without reading?

The Conclusion of the 80s: Binding Theory and the conditions on PG licensing apply at SS. Dissending : Williams Two types of Questions: 1. Chomsky’s question: Wouldn’t a theory without SS be a better theory. 2. Methodological question: From our perspective (that of trying to figure out wherher or not covert movement exists), isn’t the conclusion of the 80s, somewhat of an easy way out.

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Anagnostopoulou/fox

Page 7

Advanced Syntax ; Spring 2007

9. Possible Alternative for Condition C Lebeaux and Condition C (26)

a. Which argument that John1 made did he1 believe? b. ?? Which argument that John1 made a mistake did he1 accept?

Lebeaux’s account: A derivation is terminated the moment a representation (i.e. a step in the derivation) violates Condition C. Adjuncts can be merged with the item they modify counter-cyclically. Complements cannot (projection principle). Chomsky’s (1993) interpretation Lebeaux was right, but we can claim now that Binding Theory applies at LF, if we accept

the copy theory of movement.

Some more stuff that we didn’t get to:

But, don’t we loose the account of ACD (Hornstein, Schmitt, Fox)

Elided VP

(27)John [VP likes every boy Mary does

].

Anecedent VP

(28)

LF-traces: [every boy Mary does John

]

[VP-A likes t].

LF-copy-theory: [every boy Mary does John

] [VP-A likes [every boy Mary does

]].

A closer look (Fox 1995, 2002, drawing heavily on Fiengo and May 1994) (29)

a. ??/*Someone introduced himi to every friend of John'si.

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Anagnostopoulou/fox

Page 8

Advanced Syntax ; Spring 2007

b. ?Someone introduced himi to every person that Johni knows. c. I introduced himi to every person that Johni did. (30)

a. I wanted himi to meet every person that Johni thought I did . b. I wanted himi to meet every person that Johni thought I DID

What should we conclude? To be continued 9. Possible Alternative for Condition A We were wrong about the landing site of QR (31) a. ??The two rivals hoped that Bill would hurt (every one of) each-other’s operations. b. The two rivals hoped that someone would hurt (every one of) each-other’s operations.

(32)

*∃ > ∀

∀>∃

a. The two friends hoped that someone would buy each-other’s pictures of Mary. b. *The two friends hoped that someonei would buy each-other’s pictures of himselfi. c. *The two friends hoped that someonei would buy each-other’s pictures of hisi mother.

For covert wh-movement, we had a homework assignment, but here are the relevant facts (corrected for the problem of logophoricity as discussed on Friday 3/23): (a)

a. I told Mary1 which picture of herself1 John was looking at? b. * I told Mary1 that John was looking at a picture of herself1? c. * I told Mary1 which man was looking at which picture of herself1?

(b)

a. Who1 did you tell that Mary was looking at which picture of himself1? b. * Who1 did you tell that Mary was looking at a picture of himself1?

(c)

a. Which person did you tell that Mary1 bought which picture of herself1? b. * Which person did you tell that Mary1 bought a picture of herself1?

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

1

Reconstruction 1. Movement can have effects on Interpretation (Scope, Variable Binding, BT) Katz Postal hypothesis: Interpretation is determined at D-Structure However, beginning with Syntactic Structures there has been accumulating evidence against this hypothesis (we’ve already seen some evidence). 1.1 Variable Binding and Scope (1)

a. John seems to a (#different) teacher [ t to be likely to solve every one of these (∃ > ∀) ∗(∀ > ∃)1 problems]. b. [Every one of these problems] seems to a (different) teacher [ t to be likely t to be solved t by John. (∃ > ∀) (∀ > ∃)

(2)

a. ??his mother loves every boy. b. Every boy is loved by his mother.

(3)

a. *It is expected by his mother that every boy would be home on time. b. Every boy is expected by his mother t to be home on time. b. *This problem seems to his mother t to be likely to be solved by every boy. c. Every boy seems to his mother t to be likely to solve this problem.

1.2. Binding Theory Condition A: (4)

a. *It seems to himself that John would solve the problem. b. *The problem seems to himself t to have been solved by John. c. John seems to himself to have solved the problem.

1The

impossibility of wide scope for the universal quantifier can be seen by the ungrammaticality of (1a) when different receives a bound interpretation as in a different guard is standing on top of every building. We can further demonstrate the impossibility of the (∀ > ∃) scope relation by considering cases in which the alternative scope relation results in an interpretation which is cognitively anomalous, e.g.: # This soldier seems to someone to be likely to die in every battle. or #The ball seems to a boy to be under every shell. (c.f. Every shell seems to a (different) boy to be over the ball .)

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

2

(5)

a. *John expects Bill to praise himself. b. John expects himself to be praised.

(6)

a. I told John whether/that Mary bought a picture of himself. b. I told John which picture of himself Mary bought.

Condition B: (7)

a. John wants it to seem to everyone that he has solved the problem. b. John wants the problem to seem to everyone to have been solved by him. c. *John wants him to seem to everyone to have solved the problem.

Condition C: (8)

a. *It is expected by him that a picture of John would be on sale. d. A picture of John is expected by him t to be on sale.

(9)

a. *He bought a picture that John saw. b. Which picture that John saw did he buy?

A Possible Conclusion: the base position of an element is where a theta role is determined but other (interpretive) properties are determined at the landing site. (Scope, Binding Theory) A certain semantic procedure suggests itself, along with the postulation of QR, to which we will return. (10)

YP…t… YP(λx…x…)

2. Movement need not have effects on interpretation 2.1.

Scope (and Variable Binding) Reconstruction

A-Movement (11) a. Someone from New York is very likely t to win the lottery. b. Someone from New York seems t to be very likely t' to win the lottery. c. Many soldiers seem t to be very likely t' to die in the battle.

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

3

The trapping effect (Lebeaux, Hornstein) (12)

a. [At least one soldier]1 seems (to Napoleon) [t1 to be likely to die in every battle]. b. [At least one soldier]1 seems to himself1 [t1 to be likely to die in every battle]. c. [At least one soldier]1 seems to his1 commanders [t1 to be likely to die in every battle].

(13)

a. One soldier is expected (by Napoleon) [t to die in every battle]. b. One soldier1 is expected by his1 commander [t1 to die in every battle].

Variable Binding (Obviation of WCO, Engdahl) (14)

a. Which of his1 students did every professor1 talk to t? b. Which student of his1 did no professor1 talk to t? c. Which student of his1 did you think every professor1 talked to t? d. Which of his1 students did you think no professor1 talked to t?

(15)

a. *Which of his1 students t talked to every professor1? b. *Which student of his1 t talked to no professor1? c. *Which student of his1 did you think t talked to every professor1? d. *Which of his1 students did you think t talked to no professor1?

2.1.1. BT Reconstruction Condition A: (16)

a. Pictures of himself seem to John [t to be available] b. Friends of each other are expected by John and Mary [t to arrive on time] c. ??[Friends of each other] promised John and Mary [PRO to arrive on time]

(17)

a. Which picture of himself did Mary say that John likes t? b. Which of each others friends did Mary tell you that John and Fred like t? b. *Which of each others friends did Mary tell t that John and Fred like you?

Condition C:

Riemsdijk and Williams, Freidin, Lebeaux:

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

4

(18) a. [Which argument that John1 made] did he1 believe t? b. ??[Which argument that John1 is a genius] did he1 believe t? c. ?? [Whose evaluation of John1] is he1 proud of t (cf. Whose evaluation of him is John proud of)

The facts with Condition C are interesting since (in contrast to anything else we've seen up to now) they involve obligatory reconstruction. Something to think about: Are there cases of reconstruction for condition B. If not, why not? 3. The Relationship between Scope Reconstruction and BT Reconstruction Scope Reconstruction seems to be possible in many cases. The same is true of BT reconstruction. The question is, do they correlate?

3.1.

Condition C and Scope Reconstruction

De we get the following correlation? (7)

[ QP ...r-expression1...]2......pronoun1....t2

(8)

Scope Reconstruction feeds BT(C): Scope Reconstruction should be impossible in the structural configuration in (7).

A Movement (Fox, Romero, Sportiche) These judgments are reported in the literature, but the judgment in class did not confirm the prediction: (19)

(20)

a.

[A student of his1] seems to David1 [t to be in the other room]. (∃>seem) (seem >∃)

b.

[A student of David’s1] seems to him1 [t to be in the other room]. (∃>seem) ??(seem >∃)

a. For these issues to be clarified,

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

5

[Many more/new papers about his1 philosophy] seem to Quine1 [t to be needed]. a. #For these issues to be clarified, [Many more/new papers about Quine’s1 philosophy] seem to him1 [t to be needed]. A-bar Movement Lebeaux: (21) a. b.

[The papers that hei gave to Ms. Brownj] every studenti hoped [CP t' that shej will read t]. *[The papers that hei gave to Ms. Brownj]

shej hoped [CP t' that every studenti will revise t].

A variation on Lebeaux which relies on (argues for) VP adjunction (Fox): (36) a. [Which (of the) paper(s) that hei wrote for Ms. Brownj] did every studenti get herj * to grade? b. *[Which (of the) paper(s) that hei wrote for Ms. Brownj]

did shej * get every studenti * to revise?

c. [Which (of the) paper(s) that hei wrote for herj]

did Ms. Brownj * get every studenti to revise?

(22) [How many NP]1 φ(t1) How n: n many NP λt φ(t) Heycock: (18)

(18)

a. [How many people from his1 class] is John1 likely to meet? b. [How many people from John's1 class] is he1 likely to meet?



a. #[How many papers that John1 writes] does he1 think t will be published? b. [How many papers that John1 wrote]

does he1 think t will be published?

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

6

(21)

a. *How many people from Diana's1 neighborhood does she1 think there are t at the party? b. How many people from Diana's1 neighborhood does she1 think t are at the party? c. How many people from her1 neighborhood does Diana1 think there are t at the party?

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Anagnostopoulou / Fox

Advanced Syntax, Spring 2007

WH-MOVEMENT: ISLANDS, BARRIERS AND SUCCESSIVE-CYCLICITY PART 1. Subjacency [The handout incorporates a lot of material included in David Pesetsky’s previous handouts on the topic.] 1. BACKGROUND. ISLANDS Unlike A movement which cannot cross a finite sentence boundary, wh-movement can do so. Wh-movement is said to be unbounded: (1) (2)

a. b. a. b. c.

Sam was likely _ to win *Sam was likely that _ won Who do you believe [that Mary said [that Sam will visit _]] The person [which you believe [that Mary said [that Sam will visit_]]] That man [I believe [that Mary said [that Sam will visit __ ]]]

Wh-movement is not entirely free though. For example, it cannot take place across sentential subjects and complex NPs. Contexts which do not allow wh-movement are called islands (Ross 1967).

2. (SOME) OF ROSS’S ISLANDS TO WH-MOVEMENT AND SUBJACENCY 2.1. Ross’s Islands The sentential subject constraint (3) subject CPs *Who did [[that Sue spoke to __ ] surprise me] The Complex NP-Constraint (4) case 1: relative clause modifiers of DPs. a. ??The Minimalist Program, which I'd love to meet [the person [who wrote __]] b. *The Minimalist Program, which I'd love to meet [the person [who you talked to about __] (5) case 2: CP complements to N.

1



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

a. ??John, who I regret [the fact [that I must invite __ to the party]] b. ??Pinochet, who the TV reported [a rumor [that the British would soon release __]] c. ??The Minimalist Program, which Mary offered [a proof [that Francis Bacon had written __]] 2.2. Subjacency and successive cyclic movement Chomsky (1973, 1977): Ross’s islands can be given a uniform account under Subjacency:1 (6)

Subjacency No rule may move an element from the position Y to the position X ……..X……[α…….[β……Y……]…..X….. where α and β are bounding nodes

Bounding nodes in English: DP, CP (in Chomsky 1977). Apparent unbounded movement proceeds through Comp. (7)

Escape hatches Wh-movement targets Spec,CP. There is only one Spec,CP.

(8)

[Who do[TP you believe [CP_ that [TP Mary said [CP _ that [TP Sam will visit _]]

Movement of who never violates Subjacency, since it proceeds through C. A note on ‘bridges’ Long distance wh-movement is allowed only under, so called, ‘bridge verbs’. Complements of non-bridge verbs are islands to extraction:2 (9) case 1: "manner of speaking" verbs ??Tom, who [the king whispered [that we should behead __] (10) case 2: factive verbs (worst with it) a. ?Mary, who Sue regretted [that she had to talk to __] b. *Mary, who Sue regretted it [that she had to talk to __]

1

Chomsky (1977) assumes that subjacency is a property of cyclic rules, i.e. part of the definition of the

Cycle. He suggests that no cyclic rule can move Y to X, where α and β are cyclic nodes.

2 Bridge verbs (Erteschik-Shir 1973) typically include verba sentiendi et dicendi and verbs expressing

speech (e.g. hear, believe, know, say, claim).

Bridge verbs strongly correlate to class of predicates allowing matrix clause phenomena (complementizer

deletion, topicalization, negative preposing) discussed, among others, in Holmberg & Platzack 1995.

.The class of verbs permitting embedded V2 in Mainland Scandinavian languages is much smaller than the

class allowing extraction (cf. Vikner 1995: 70).

2



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

In Chomsky (1977) it is suggested that rules of interpretation have to take care of the bridge condition (p.85) . 2.3. Subjacency and Ross’s islands If movement proceeds successive cyclically and if Subjacency holds, then Ross’s islands can be given a uniform explanation: The Complex NP-Constraint (11) case 1: relative clause modifiers of DPs. a. ??The Minimalist Program, which [TPI'd love to meet [DPthe person [CPwho [TP wrote __]] Movement crosses CP, DP. (12) case 2: CP complements to N. a. John, who [TPI regret [DP the fact [[CP _ that I must invite __ to the party]] Movement crosses CP and DP. Sentential subjects (13)

*Who did [DP [CP

__

[ that Sue spoke to __ ] surprise me]

-Movement crosses a CP and a DP.

Crucial assumption: sentential subjects are dominated by a DP node.

Apparent unbounded movement never crosses two bounding nodes because it proceeds successive cyclically. Only one CP is crossed in every step (14)

a. b. c.

[CPWho do you believe [CP _ that Mary said [CP _that Sam will visit _]] The person [CP which you believe [CP _ that Mary said [CP _ that Sam will visit_]]] That man [I believe [CP_ that Mary said [CP _that Sam will visit __ ]]]

2.4. Evidence for successive cyclicity [see Jason Merchant’s handout “Evidence for successive cyclic movement”] 1) West Ulster English (McCloskey 2000) Wh-movement cam strand all in the trace position: 3 Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(15)

a. What all did you give __ to the kids? b. What did you give __ all to the kids?

(16)

a. Who all did you send __ to the shops? b. Who did you send __ all to the shops?

(17)

a. Tell me what all you got __ for Christmas. b. Tell me what you got __ all for Christmas.

It is important to make sure that the all is not simply free to occur anywhere. It really does seem to stand next to the trace of the wh-word. It cannot occur in random places: (18)

*Who did he tell __ he was going to resign all.

(19) a. What did you do __ all after school the day? b. *What did you do __ after school the day all? c. *What did you do __ after school all the day. All stranding in C: (20) a. What all did he say [CP __ (that) he wanted__]? b. What did he say [CP __ (that) he wanted all __]? c. What did he say [CP all __ (that) he wanted __]? (21) a. What were you trying [CP all __to say __]? b. What did you mean [CP all __ for me to do __]? 2. Successive Inversion Aux inversion in Belfast English (Henry 1995) (22) a. Who did John hope [ would he see __]? b. What did Mary claim [did they steal __?] c. I wonder what did John think would he get __? d. Who did John say [did Mary claim [had John feared [would Bill attack __]?

4



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Obligatory Inversion in Spanish (Torrego 1984, Pesetsky & Torrego 2001) (23) a. A quién prestó Juan el diccionario? to whom lent John the dictionary b. Con quién podrá Juan ir a Nueva York? with whom will-be-able J. to go to NY c. Qué pensaba Juan [que le había dicho Pedro [que había publicado la revista]]]. what thought John that him had told Peter that had published the journal 'What did John think that Peter had told him that the journal had published?' Stylistic Inversion in French (Kayne and Pollock 1979) (24)

a.

Avec qui a prétendu Marie que sortirait Jean ? with whom claimed Mary that would leave Jean? ‘Who did Mary claim that Jean would leave with?’

b.

Sur qui a prétendu Marie que tirerait Jean ? on whom claimed Mary that would shoot Jean? ‘Who did Mary claim that Jean would shoot at?’

compare to: (25)

a. b.

Comment sait Marie que Luc est mort ? How knows Marie that Luc is dead ‘How is it that Mary knows that Luc is dead’? *Comment sait Marie qu’est mort Luc?

[inversion in the embedded clause is not licensed when the wh-word is extracted from the main clause] 3. Languages with wh-agreement (Haik 1990; Chung 1998; McCloskey 1990, 2002 among others). Irish Irish has three complementizers: aL, aN, go (L = triggers 'lenition' on following word; N = triggers 'nasalization' on following word) (26) Deir siad [CP gur [TP ghoid na síogaí í ]]. say they go-PAST stole the fairies her ‘They say that the fairies stole her away.’ (27) an ghirseach [CP a [TP ghoid na síogaí __ ]] the girl aL stole the fairies 5



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

‘the girl that the fairies stole away’ (28) an ghirseach [CP ar [TP ghoid na síogaí í]] the girl aN stole the fairies her ‘the girl that the fairies stole away’ aL occurs in structures in which wh-movement has occurred. Relative clauses: [see (27)] Questions: (29) Cá fhad a bhí siad fá Bhaile Átha Cliath t ? WH length aL be[PAST] they around Dublin ‘How long were they in Dublin?’ Clefts: (30) Ba i nDoire a dúradh a fuarthas é t. COP[PAST] in Derry aL was-said aL was-found it ‘It was in Derry that it was said it was found.’ It occurs in every position through which wh-movement has taken place: (31) Cén fear aL thiteann go talamh which man C falls to earth 'Which man falls to earth?' (32) Cé [aL bhuail tú] ? who C struck you 'Who did you hit?' (33) Cé aL mheas tú aL chonaic tú? who C thought you C saw you ‘Who did you think that you saw?’ (34) Cén t-úrscéal aL mheas mé aL dúirt sé aL thuig sé. which novel C thought I C said he C understood he ‘Which novel did I think he said he understood?’ (35) Níl a fhios agam cén fear a thiteann go talamh I don't know which man C falls to earth 'I don't know which man falls to earth.' (36) Níl a fhios agam cé [a bhuail tú] I don't know who C struck you 'I don't know who you hit.' (37) Níl a fhios agam cén t-úrscéal aL mheas mé aL dúirt sé aL thuig sé. I don't know which novel C thought I C said he C understood he ‘I don't know which novel I thought he said he understood.' 4. Intermediate copies Child English (Crain and Lillo-Martin 1999) 6



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(38) a. What do you think what Cookie Monster eats? (age 5;0) b. Who do you think who Grover wants to hug? (age 4;9) c. What do you think what's in that box? (3;3) Dutch dialects (Barbiers, Koeneman & Lekakou 2007): (39)

Wie denk je wie ik gezien heb? WHo think you who I seen have ‘Who do you think I have seen?’

5. Partial wh-movement (McDaniels 1989, Horvath 1997) (but see Horvath 1997 for arguments that partial wh-movement constructions do not result from movement). German (40) Was glaubst du, mit wem er gesprochen hat? what think you with whom he spoken has 'With whom do you think that he spoke?' Hungarian (41) Mit gondolsz, hogy kit látott János? what.ACC you.think that who.ACC saw.3sg J.NOM ‘Who do you think that Janos saw?’ 6. Reconstruction (was extensively discussed in previous classes) Fox 1999: (42) a. [Which paper that he1 wrote for Ms. Brown2] did every boy1 hope that she2 'd regrade? (Example answer: His worst one.) b. * [Which book that he1 asked Ms. Brown2 for] did she2 say every boy1 had to return by Monday? 2.5. Some more islands Subject DPs (“The Subject Condition” Chomsky 1973) (43)

*Who did [DP comments about __ ] surprise me]

7



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Wh-islands (the wh-island condition Ross 1967) (44) ?The Minimalist Program, which Bill asked [CP who [ we had spoken to __ about __] In its present form, Subjacency can’t account for these islands. In (43) only a DP is

crossed. In (44) only a CP is crossed.

Question: Can we account for these islands in terms of Subjacency?

Answer: Yes, if we include TP in the bounding nodes.

Then, wh-movement crosses a DP and a TP in (43):

(43’) *Who [TPdid [DP comments about __ ] surprise me]

And wh-movement crosses two TPs and a CP in (44):

(44’) ?The Minimalist Program, [which [TPBill asked [CP who [TP we had spoken to __

about __] If C allows only one specifier, movement is impossible from a clause whose Spec,CP is occupied by a distinct wh-phrase. Some complications …With respect to the Subject Condition: movement across a DP is not ill-formed in English: (45)

a. b.

Who did [TP he find [DP a picture of _ ] ? What books did [TP he write [DP reviews of _] ?

Chomsky (1977: 114) proposes that these involve extraposition of the PP: (46)

a. b.

He saw [DP a picture [PP of John ]] He saw [DP a picture t] [PP of John ]

Wh-movement targets the extraposed PP, and Subjacency is not violated. …With respect to the Wh-island Condition An alternative derivation has to be ruled out: If in (44) (i) which moves to embedded COMP, then (ii) which moves to matrix COMP and then (iii) who moves to embedded COMP, then Subjacency is not violated.

8



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Step (iii) violates the Strict Cycle Condition. (47) Strict Cycle Condition (Chomsky 1973: 243) No rule can apply to a domain dominated by a cyclic node A in such a way as to affect solely a proper subdomain of A dominated by a node B which is also cyclic node Assumption: CP and DP are cyclic nodes. Step (iii) violates the Strict Cycle Condition.

9



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

"On Wh-Movement" 1. Wh-movement

• In the literature at the time: perhaps the gap arises from deletion. • But comparatives show the stigmata of (AP) wh-movement.

Some ideas circa 1976: (1)

SSC/PIC No movement rule may involve X and Y in ... X. . . [ α. . . Y. . . ] ... X. . .

where α contains a [subject that c-commands Y] or is "propositional". (2)

the COMP escape hatch ... where Y is not in COMP [of α].

(3)

Free deletion in COMP: wh-phrase becomes null that becomes null for becomes null

(4)

Doubly-Filled COMP filter Only one of wh or C may occupy COMP.

(5) Two Movement Rules a. Move NP. b. Move wh-phrase. (6)

(7)

2. Comparatives

[ϕ-features!] [wh-features!]

The rule of wh-movement has the following general characteristics: a. it leaves a gap b. where there is a bridge, there is an apparent violation of subjacency, PIC, and SSC c. it observes CNPC d. it observes wh-island constraints [(49)] Goal of the paper: "Where we find the configuration [(6)] in some system of data, can we explain it on the assumption that the configuration results from wh-movement?"

(8) Over wh-word may show up a. John is taller than (what) Mary is. b. John is taller than (what) Mary told us that Bill is.

[(51)]

(9) Shows bridge/non-bridge and other island contrasts a. Mary isn 't the same as [ she was five __ years ago ] b. Mary isn't the same as [ John believes [ that Bill claimed [ that

she was __ five years ago] ] 1

c. *Mary isn't the same as [John believes [ Bill's claim [ that she was __ five years ago ] ] ] d. *Mary isn 't the same as [ I wonder [ whether she was __ five years ago] ] [(52)] (10) Strong crossover in comparatives (Bresnan 1975) a. More studentsi flunked than __ thought theyi would flunk. b. *More studentsi flunked than theyi thought __ would flunk. (11) Analysis: More students flunked than [wh-many students [ .... gap...]] Note: the examples are complicated by 1. extraposition of the comparative clause; and 2. non-obvious semantics

3. Topicalization • Is Topicalization just like left-dislocation? (12) Left-dislocation a. This book, I think you should read it. b. As for this book, I think you should read it.

• No! Left-dislocation does not look like movement, but topicalization does. (13) Topicalization shows bridge/non-bridge and other island contrasts a. This book, I really like. b. This book, I asked Bill to get his students to read. c. *This book, I accept the argument that John should read. d. *This book, I wonder who read.

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-2 (14) ...unlike Left dislocation c. As for this book, I accept the argument that John should read it. d. This book, I wonder who read it.

4. Indirect Questions and Relative Clauses: finite and infinitival

(15)

Questions

Some Phrase Structure rules a. S'' -> TOP S' b. S' -> COMP S c. S' -> COMP S"

Consequences: • Topic recursion allowed by rules (a) and (c). • The S' introduced by rule (a) could be a wh-clause. This yields the Topicalization construction if we propose that wh-deletion is obligatory here. • When it is not a wh-clause, it is Left Dislocation.

• The point: S' comes in both finite and non-finite flavors

(20)

Finite indirect questions show bridge/non-bridge and other island contrasts a. I wonder [who John saw]. b. I wonder [who John believed [that Mary would claim [that Bill would visit t] ]. c. *I wonder [ who John believed [the claim [ that Bill would visit ] ] ]. d. *Who2 did you wonder [whol t1 saw t2 ].

(21)

Infinitival indirect questions show bridge/non-bridge and other island contrasts a. I wonder [who to see].

b1. I wonder [who to order Mary [ to promise [to visit] ]].

b2. I wonder [ who to persuade Mary [ that she should promise [to visit] ] ]

c. *I wonder [who to insist on [the principle [ that Bill should visit] ] ].

d. *Who2 do you wonder [what1 to give t1 to t2]

*What2 do you wonder [ [ to whom]1 to give t2 t1 ]

Tacit Assumption: S' but not S" may be subcategorized. (16)

(17)

Cleft sentences show bridge/non-bridge and other island contrasts a. It is this book that I really like. b. It is this book that I asked Bill to get his students to read. c. *It is this book that I accept the argument that John should read d. *It is this book that I wonder who read Analysis of clefts: It is S".

Stipulation:

(i) the S' must show wh-movement; (ii) COMP (for some speakers) must not become "terminally null"

(18)

Pseudo-clefts a. This book is what I really like. b. This book is what I asked Bill to get his students to read. etc.

(19)

Analysis of pseudo-clefts NP is S'

• Note: The impossibility of an overt subject for the infinitives in (21) was a mystery for 1976 syntax. Case theory lay 2-3 years in the future.

Finite Relative clauses (22)

(23)

Finite relative clauses: bare wh a. *the person [whom that I met __] b. the person [whom I met __] c. the person [that I met __] d. the person [ I met __] Finite relative clauses: PP pied-piping a. *the person [with whom that Mary spoke __ at the party] b. the person [with whom Mary spoke __ at the party] c. *the person [that Mary spoke __ at the party] d. *the person [Mary spoke __ at the party]

*wh-that ok wh ok that ok zero

*wh-that ok wh *that *zero

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-3 Non-Finite Relative clauses

5. Tough-constructions

(24)

(27)

Infinitival indirect questions show bridge/non-bridge and other island contrasts a. 1found a book for you to read t b. I found a book for you to arrange for Mary to tell Bill to give t to Tom c. I found a book for you to insist that Bill should read t d. I found a book for you to insist that Bill tell Mary that Tom should read t e. *I found a book for you to insist on the principle that Tom should read t [f. *Who did he find a book t to read]

• Stipulation:

•

Infinitival relative clauses: bare wh (oblg. deletion of wh) a. *a person [whom for Mary to invite __ to the party] *a person [whom for PRO to invite __ to the party] b. * a person [who Mary to invite __ to the party] *a person [who PRO to invite __ to the party] c. a person [for Mary to invite __ to the party] *a person [for PRO to invite __ to the party] d. *a person [Mary to invite __ to the party] a person [ PRO to invite __ to the party]

(26)

Infinitival relative clauses: pied-piped PP (wh ok) a. *a person [with whom for Mary to speak __ to the party] *a person [with whom for PRO to speak __ to the party] b. * a person [with whom Mary to speak __ to the party] a person [with whom PRO to speak __ to the party] c. *a person [for Mary to speak __ to the party] *a person [for PRO to speak __ to the party] d. *a person [Mary to speak __ to the party] *a person [ PRO to speak __ to the party]

*wh-for

Movement or deletion? Where does the subject get its θ-role from?

(28)

θ-role from downstairs? *John is easy to please Mary.

(29)

Movement? a. %Close tabs are easy (for us) to keep on Bill. b. %Headway is easy (for us) to make in this car.

Deletion of wh is obligatory in infinitival relatives, except when deletion is non recoverable [p. 98] (25)

a. It is easy (for us) to please John b. John is easy (for us) to please.

but: c. *There is easy (for us) to believe __ are multiple solutions to the problem. compare: d. %The planet is easy (for us) to believe __ exists. (30)

* wh ok for

ok zero

Tough constructions show bridge/non-bridge and other island contrasts a. John is easy (for us) to please t b. (i) John is easy (for us) to convince Bill to do business with t (ii) John is easy (for us) to convince Bill to arrange for Mary to meet t c. John is easy (for us) to convince Bill that he should meet t d. John is easy (for us) to convince Bill to tell Mary that Tom should meet t e. (i) *John is easy (for us) to convince Bill of the need for him to meet (ii) *John is easy (for us) to describe to Bill a plan to assassinate t f. (i) *what2 is John fun (for us) [(who1 ) to give t2 to t1 ] (ii) *who2 are the presents fun (for us) [(which1} to give t1 to t2] (iii) *[ to whom]2 are the presents fun (for us) [ (which1) to give t1 t2]

*wh-for

ok wh

"In short, the basic properties of easy-to-please constructions follow directly from the assumptions we have already made, assuming that here too wh-movement is crucially involved. The latter assumption is particularly natural in this case, since we have analogous forms in which the wh-phrase may directly appear...

* for

(31)

John is an easy person to please.

(32)

a. this is an easy violin on which to play sonatas b. this is a pleasant room in which to work

* zero

"Whatever the correct analysis of these strucutres may be, it seems clear that they involve, at some level, a phrase such as (33), as an adjectival modifier: (33)

a. easy - on which to play sonatas ( violin) b. pleasant - in which to work ( room)

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Anagnostopoulou / Fox

Advanced Syntax, Spring 2007

WH-MOVEMENT: ISLANDS, BARRIERS AND SUCCESSIVE-CYCLICITY Part II. CED, ECP [The handout incorporates a lot of material included in David Pesetsky’s previous handouts on the topic.] 1. HUANG’S (1982) CONDITION ON EXTRACTION DOMAINS (1)

CED A phrase A may be extracted out of a domain B only if B is properly governed

properly governed: governed by a lexical head extraction out of complements allowed / out of adjuncts and subjects not Facts subsumed under CED (Huang 1982 drawing on Kayne 1981): Subject-Object asymmetries in extraction Adjunct – Object asymmetries in extraction Generalizing: Non-complement (subject, adjunct) – complement (object) asymmetries in extraction. Illustrating: Subjects: Extraction from subjects illformed / extraction from objects wellformed: case 1: subject DPs (2)

a. b.

*Who did [DP pictures of _] please you? Who did you see [DP pictures of _]

case 2: subject CPs (3) *John, who [ [that Sue spoke to __ ] surprises me] case 3: subject PPs (in locative inversion) (4) *the room, which [ [in the middle of __ ] sat a frog]

1



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Adjuncts case 1: adjuncts in CP/TP/VP (5) a. ??The Minimalist Program, which [ Mary left [before she could finish __] b. ??The Minimalist Program, which [ Mary bought [so that she could read __ on the plane] c. *What did John arrive yesterday [sad about _]? d. * Who did John come back [before I had a chance to talk to _]? case 2: adjuncts in DPs (6) a. *The Minimalist Program, which I regret [your departure [before you could finish __]] b. *The Minimalist Program, which we witnessed [a purchase [so that someone could read on the plane]] c. *Which table do you like [the books on _]? vs. (7)

Which city did you witness [the destruction of _]?

case 3: relative clause modifiers of DPs. (8) a. ??The Minimalist Program, which I'd love to meet [the person [who wrote __]] b. *The Minimalist Program, which I'd love to meet [the person [who you talked to about __] c. *Who do you like [books [that criticize _ ]? Notes: This does not include the wh-island condition. This presumes that extraction from a complement CP does not require an intermediate landing site in Spec,CP. 2. SUBJACENCY ([CP OR TP] + DP) VS. CED: 1. The subjacency idea: a. Do not cross two among CP/TP or DP in a single move. T herefore, LD movement must stop in available landing sites. Landing site for A-bar movement is SPEC,CP. b. This does not get complement/non-complement asymmetry. 2. The CED idea: a. Do not cross a non-complement node in a move. 2



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

b. This does not get the wh-island condition. 3. Overlaps: a. Rel clauses: b. subject DPs:

subjacency and CED subjacency and CED

4. Non-overlaps: a. subject sentences: only CED (unless it is stipulated that CP is dominated by DP) b. adjunct CPs: only CED c. fact-that islands: only subjacency d. wh islands:

only subjacency

3. A COMMON VIEW IN THE EIGHTIES: SUBJACENCY AND CED AFFECT ONLY OVERT MOVEMENT 1. Evidence that wh-islands do not apply at LF (Baker 1970, discussed in Huang 1982: 495): (9)

a. b.

*What do you remember where I bought _? Who remembered where we bought what?

2. Evidence that the Sentential Subject Constraint does not apply at LF (Huang 1982: 495-496): (10)

a. b.

*Who did he say that [for Bill to marry _] was a surprise? ?Who said that [for Bill to marry who] was a surprise?

Note: No significant contrast between the following: (11)

a. b.

*Who did [that Bill married _] surprise you? *Who said that [that Bill married who] surprised you?

The ungrammaticality of (11b) is linked to the ungrammaticality of (12), which does not display movement: (12)

*He said that that Bill married Ann was a surprise

3. Evidence that the Subject Condition does not apply at LF (Huang 1982: 497): (13)

a. b.

*Who do you think that [pictures of _] would please John? Who thinks that [pictures of who] would please John? 3



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

4. Evidence that the bridge vs. non-bridge distinction does not apply at LF (p.497): (14)

a. b.

??Who did John whisper that he saw _? Who whispered that he saw who?

5. Evidence that the Adjunct Condition does not apply at LF (p. 497, 499, 503): (15) (16) (17) (18)

a. b. a. b. a. b. a. b.

*Who did John come back [before I had a chance to talk to_]? Who came back [before I had a chance to talk to whom]? *Which class did you fall asleep [during _]? Who fell asleep [during which class]? *Which table do you like [the books on_]? Who likes [the books on which table _]? *Who did Mary cry [after John hit _]? Who cried [after John hit who]?

6. Evidence that the Complex NP Constraint does not apply at LF (p. 492): (19) a. b.

(20)

a. b.

*In order to foil this plot, we must find out which senator the agent has [DP buts [CP that are trained to kill _] In order to foil this plot, we must find out which agent has [DP buts [CP that are trained to kill which senator] *Who do you like [DP books [CP that criticize _]? Who likes [DP books [CP that criticize who]?

4. The ECP: ECP phenomena 1. That-t effects Original motivation for the ECP (empty category principle): “that-trace” effects: (21) (22)

a. b. a. b.

*Who do you think [that [_ saw John]? Who do you think [that [John saw _] ? *Who do you wonder [how [_ bought the book]? ??What did you wonder [how [he bought t]

Subjacency had nothing to say about this asymmetry (the number of nodes crossed in subject movement and object movement are exactly the same). (23)

ECP A (non-pronominal) empty category must be properly governed

4



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Working definition of the ECP (in Hornstein and Weinberg 1995: 246) (24) An empty category must be: (a) Lexically/head governed: governed by a lexical X0, or (b) Antecedent governed: bound by (coindexed with and c-commanded by) a category that governs it Definition of government (Aoun and Sportiche 1981): (25) A governs B iff for all X, X a maximal projection, X dominates A iff X dominates B The that-t effect explained:

-Object trace is governed by V, a lexical governor

-Subject trace is not lexically governed, hence it must be antecedent governed.

The intermediate trace does not c-command the subject trace (no generalized X-bar for COMP); the trace is properly contained in COMP: *[whoi [do you think [S’ [COMP ti that ] [ti saw John]]]]

c-command is satisfied when the complementizer is empty because the entire COMP is

identified with ti which it exclusively dominates.

Cases that have been brought under the ECP 2. Personne in French Kayne (1981): Subject-object asymmetries also hold at LF: (26) a. b.

*Je n’ai exige que personne soit arrête I didn’t require that anybody arrested Je n’ai exige que la police arête personne

I didn’t require that the police arrest anybody

Kayne: at LF personne raises to the clause where ne indicates scope leaving a trace. Subject trace is not properly governed. Object trace is: (27)

a . b.

* [TP personne [TP je n’ai exige [CP _ que [TP t soit arrête]]]] [TP personne [TP je n’ai exige [CP _ que [TP la police arrête t]]]]

3. Superiority (Aoun, Hornstein and Sportiche (1981), Chomsky (1981) Kayne (1983): (28) a. b.

*Who remembers why who bought the book? Who remembers why we bought what? 5



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(29)

a. b.

(30)

a. b. c. d. e. f.

*Whoj whoi [ti remembers [whyk [ tj bought the book tk] problem: tj not properly governed Whatj whoi [ti remembers [whyk [ we [bought tj tk ]] ok: tj properly governed I don’t recall who bought what *I don’t recall what who bought *I don’t recall who left why *I don’t recall why who left *I don’t recall what Bill sang why I don’t recall why Bill sand what

The idea is that S-structure movement of adjuncts and subjects brings them to a position where they can properly govern their trace (they can c-command it) but LF movement of adjuncts and subjects leads them to a position where they can’t antecedent govern their trace (they can’t c-command it). On the other hand, LF movement of object wh-phrases is ok because their trace is properly governed (lexically head governed). 4. Huang’s observation: Adjunct extraction out of wh-islands leads to severe violations. They can fall under the ECP, i.e. they behave similarly to extractions of subjects out of wh-islands: The subject cases (see Huang, p. 562): (31)

a. b. c. d.

*Whoi did you wonder whyj ti came tj *Whoi did you wonder howj ti came tj *Whoi did you wonder wherej ti worked tj *Whoi did you wonder whenj ti will come tj

The adjunct cases (see Huang, p. 537): (32)

a. b. c. d.

*Whyi did you wonder whatj I bought tj ti *Howi did you wonder whatj I bought tj ti *Wherei did you wonder whatj I bought tj ti *Wheni did you wonder whatj I bought tj ti

vs.

e.

??Whati did you wonder whyj I bought ti tj

5. Wh-in-situ. Multiple Wh-questions involving LF movement of an adjunct or a subject are ill-formed, while when the object undergoes movement the sentence is wellformed (cases discussed in Hornstein and Weinberg, p. 248): (34)

a. b. c.

Who believes that John dropped what? *Who believes that what fell? *Who believes that John dropped the ball why?

6



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Note: the data are controversial (see fn 6 in Hornstein and Weinberg. Lasnik and Saito claim that (34b) is good with who instead of what. For alternative judgments see Aoun, Hornstein and Sportiche (1981), Kayne (1983), Aoun, Hornstein, Lightfoot and Weinberg (1987). 6. Wh-in-situ in French. Discussed in Hornstein and Weinberg (1995: 249). But data are controversial: (35)

a. b.

Jean Jean *Jean Jean

a dit said a dit said

que that que that

Pierre Pierre qui who

a vu qui saw who est venu came

7. Subjacency, CED, ECP Conceptually, CED and ECP are close, as they require reference to proper government. CED and ECP differ in that CED does not apply at LF, unlike ECP. In this respect, CED patterns with Subjacency.

7



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Anagnostopoulou / Fox

Advanced Syntax, Spring 2007

WH-MOVEMENT: ISLANDS, BARRIERS AND SUCCESSIVE-CYCLICITY Part III. Barriers 1. INTRODUCTION • Goal: an attempt to unify bounding (Subjacency a subcase of which is CED) and government (ECP) through the notion of barrier. • General idea: certain categories in certain configurations are barriers to government and barriers to movement. One barrier blocks government (ECP). More than one barriers block movement (Subjacency, CED). 2. ASSUMPTIONS 2. 1. Phrase structure X’ theory is generalized to non-lexical categories, i.e. to CP and IP 2. 2. Movement case 1. substitution: There is no movement to complement position Only heads can move to head positions Only XPs can move to specifier positions Only heads and XPs are visible to Move α NB: he takes head movement to be substitution. He discusses V-movement to C (p. 6 in V-2 environments) as an instance of head movement as substitution. Later on, he discusses V-to-I movement. He says that V “amalgamates with” I as a result of V-to-I movement. case 2. adjunction: Three crucial assumptions: (a) No adjunction to IP 1 Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced

Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

p. 5: “Another possibility is that wh-phrases have clausal scope and cannot be adjoined to IP; hence, they must move to the position of specifier of CP[…]. There is some reason to believe that this stronger condition is required, and I will henceforth assume it.1As a result, although the rule of Quantifier raising QR may involve adjunction to IP, this operation is barred for operators of the wh-type.” (b) No adjunction to argument NPs and CPs (1) Adjunction is possible only to a maximal projection that is a non-argument.

(p.6, principle 6).

Consequence: Adjunction to VP is allowed.

(c) An adjoined category is not dominated by the XP to which it adjoins: p. 7: (2)

[ β α [β ….]]

(3)

α is dominated by β only if it is dominated by every segment of β

2.3. Government (4)

α governs β iff α m-commands β and there is no γ, γ a barrier for β, such that γ excludes α

(5)

α m-commands β iff α does not dominate β and every γ (XP) that dominatesα dominates β

(6)

α excludes β if no segment of α dominates β

3. THE SYSTEM 3.1. Two notions of barrier: a) XP barriers: (7)

… α…[γ….β…..]

γ is an XP with the relevant properties. This type of barrier is relevant for movement (Subjacency/ CED) and for government (ECP). 1

In fn. 6 he says that the reason is the ungrammaticality of “*who thinks that who, I like” which according to Lasnik (work in progress back then, Lasnik and Saito) involves adjunction to IP.

2



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

b) Minimality barriers: (8)

….α….[γ’…..γ0…..β….]

γ’ is a barrier even though it is not an XP because it dominates a head γ0 which is a closer governor of β than α. This type of barrier (which reduces ambiguity of government) is relevant only for government (ECP). In what follows, I will concentrate on XP-barriers returning to Minimality Barriers later on. 3.2. XP-barriers CP and IP Neither CP alone nor IP alone are barriers for government.

IP is not a barrier: ECM

CP is not a barrier: No (e.g. Complementizer deletion depends on government from

outside, Stowell 1981).

Together, CP and IP are barriers, e.g. PRO in spec,IP is ungoverned in a structure like

[CP [IP PRO…..]]

Barrier (9)

γ is a barrier for β iff (a) or (b) a. γ immediately dominates δ, δ a BC for β b. γ is a BC for β, γ ≠ IP

Blocking Category (BC) (10)

γ is a BC for β iff γ is not L-marked and γ dominates β

L-marking (11)

α L-marks β iff α is a lexical category that θ-governs β

θ-government (12)

α θ-governs β iff α is a zero-level category that θ-marks β, and α, β are sisters

3



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

p. 13 θ-marking: only the notion of “direct θ-marking” (i.e. a zero level category α directly θ marks β only if β is the complement of α in the sense of X-bar theory) is relevant to Lmarking. Indirect θ-marking of the subject by the verb mediated by the VP is not relevant for L-marking. NB. On page 20 he discusses the relation of I to VP. He assumes that I, not being a lexical category, does not L-mark VP. Therefore VP is a BC and a barrier. He considers the following case that could suggest that I does L-mark VP: (13)

fix the car, I wonder whether he will

The fact that extraction across a wh-island is ok, seems to suggest that the VP trace is properly governed. Nevertheless, Chomsky assumes that I does not L-mark VP [this creates complications for head movement and NP movement]. 3.3. ECP (14)

A non-pronominal empty category must be properly governed

(15)

α properly governs β iff α θ-governs or antecedent-governs β

Mechanism (from Lasnik and Saito 1984): γ-marking. If α properly governs β it assigns the feature [+γ] to it. If α does not properly govern β it assigns the feature [-γ] to it. γ marking is permanent. γ-marking takes place at S-structure for elements of chains terminating in an A position and at LF for elements in chains terminating in an A’ position. 3.4. Subjacency (16)

If (αi, αi+1) is a link of a chain, then αi+1 is subjacent to αi,

“subjacent”= 1-subjacency.

best case: 0-subjacency

well-formedness: 1-subjacency

ungrammaticality: if two or more barriers are crossed.

(17)

β is n-subjacent to α iff there are fewer than n+1 barriers for β that exclude α

4



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

4. ACCOUNTING FOR WELL-FORMED MOVEMENT AND FOR ISLAND VIOLATIONS 4.1. Well-formed movement: Arguments Short distance: (18)

who did [John see _] ?

w.r.t. ECP: the object trace is θ-governed by see, hence it is properly governed. w.r.t. Subjacency: Movement to [Spec,CP] proceeds through adjunction to VP: (18’) who did [IP John [VP t’ [VP see t ]]] Movement from t to t’ does not cross the category VP (it only crosses one segment of VP). Movement from t’ to the landing site does not cross any barrier. Only one segment of VP is crossed, and IP (which is a BC but not a barrier). Long distance: (19)

Who do you believe [that Mary said [that Sam visited _]]

Movement proceeds through adjunction to VP and Spec,CP:

(19’) Who do you [VP t5 [VP believe [ CP t4 that [Mary [VP t3 [VP said [CP t2 that

[Sam [VP t1 [VP visited t ]] 1. Movement from t to t1 crosses only one segment of VP 2. Movement from t1 to t2 crosses one segment of VP and IP (not a barrier). 3. Movement from t2 to t3 crosses CP (L-marked hence not a BC, hence not a barrier) And so on….. Adjuncts In order for the ECP not to be violated not a single barrier must be crossed (Subjacency is satisfied every time ECP is satisfied). (20)

how did you fix the car _ ?

One possible analysis: 5



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Adjunct external to IP (page 19):

(20’) how did [IP you [VP fix the car ] t]

Possible configuration (see Manzini 1992: for spelling this out):

CP 3 how C’ 3 did IP 9 you I’ how 3 I VP 5 fix the car Only one BC includes the trace and not the antecedent, IP, and IP is not a barrier. ECP

and Subjacency satisfied.

Alternative analysis:

VP-internal (page 29):

suggested by VP-topicalization cases like John wanted to fix the fender with a crowbar,

and fix it that way, he did. 4.2. Islands 1. CED a) Subject Islands (22)

a. b. c.

*the man who [IP [NP pictures of _ ] are on the table ] * the book that [IP [NP reading _ ] would be fun ] who does [IP [CP _ that Mary likes _] surprise you ]

The Subject XP is not L-marked and hence a BC and a barrier. IP inherits barrierhood from the Subject since it immediately dominates it. Two barriers are crossed. b) Adjunct Islands (23)

a. b.

*to whom did [IP they leave [before speaking t]] *who did [IP they leave [before speaking to _]]

6



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

The adjunct is not L-marked, hence a barrier. IP inherits barrierhood (tree below from Manzini 1992: 10) CP 3 what C’ 3 was IP 9 Mary I’ CP 23 I VP t’ C’ # 3 bothered because IP 6 Peter explained t 3. Wh-islands Cases he considers: Argument movement out of a tensed and an infinitival clause: (24)

a. b. c. d.

whati did you wonder [CP to whomj John gave ti tj ] to whomi did you wonder [CP whatj John gave tj ti ] whati did you wonder [CP to whomj to give ti tj ] to whomi did you wonder [CP whatj to give tj ti ]

A weak Subjacency violation arises from crossing one CP which inherits barrierhood from IP. Adjunct extraction (25)

a. b.

*How did John tell you [when to fix the car t] How did John know [which car to fix t]

ECP: one barrier is crossed. Tense effects -To account for the [+tense] restriction (many people find (24c,d) more acceptable than (24a,b), and with non-argument wh-phrases things improve even more: (26)

a. b.

which car did John tell you how to fix which car did he wonder whether to fix 7



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Suggestion: “tensed IP is an inherent barrier (possibly weak) to wh-movement, over and above the system just outlined, this effect being restricted to the most deeply embedded tensed IP…” (p.37). Cases with p-stranding: (27)

*who did you wonder [what John gave t to t]

They are more severe because of a ban of double adjunction of NP to VP.

8



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

ADVANCED SYNTAX 2007 ANAGNOSTOPOULOU/FOX

1

Superiority, Nesting and Crossing 1. ECP is not enough We have learned an account of the contrast in (1) (1) a. ??[Which book]1 did you ask who bought t1? b. *[Which person]1 did you what t1 bought t? We have provided an account of this contrast in a system that has two constraints against non-local movement (Subjacency which applies to all movement operations, and ECP, which only restricts the movement of subjects and adjuncts). However, there seems to be something that blocks (1)b independently of the ECP (as pointed out by Omer last week). (2) a. ??[Which book]1 did you ask who2 Mary told t2 [PRO to present t1]? b. *[Which person]1 did you ask what2 Mary told t1 [PRO to present t2]? (3) a. This is the violin wh1 that I wonder which sonatas2 to play t2 on t1. b. *These are the sonatas wh1 that I wonder which violin to play t2 on __. 2. Constraint on Crossing Dependencies (Kuno and Robinson) The Constraint on Crossing Dependencies (CCD): a. Two wh-dependencies cannot cross. b. Two dependencies (chains) C and C' are called crossing dependencies if the head of C c-commands the head of C' and the tail of C c-commands the tail of C': C…C'…C…C' c. Two dependencies (chains) C and C' are called nested dependencies if the head of C c-commands the head of C' and the tail of C' c-commands the tail of C: C…C'…C'…C Frazier and Fodor (1978): The CCD follows from the nature of the parsing mechanisms that enable “fillers” to be associated with “gaps”. Fillers are stored in memory by a “last in-first-out” device (a “stack”). 3. Superiority in English Problem #1 (Superiority): We seem to be loosing a generalization We might want to relate the contrast in (2) and (3) to the contrast in (4) (4) a. You asked who1 Mary told t1 [PRO to present what]. b. *You asked what1 Mary told who [PRO to present t1]. Pesetsky (1982): (4)b involves an LF violation of the CCD (which Pesetsky generalized and called the path containment condition PCC)

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

ADVANCED SYNTAX 2007 ANAGNOSTOPOULOU/FOX

2

(4') LFs of the sentences in (4): a. You asked what2 who1 Mary told t1 [PRO to present t2]. b. * You asked who1 what2 Mary told t1 [PRO to present t2]. These LFs are predicted by the Extension Condition, which is needed on independent grounds, hence provide a very interesting unified account for (2), (3) and (4). Conversely, the facts in (2), (3), and (4) provide independent evidence for covert wh movement. [To use the terminology of our class on covert movement, the CCD serves as a structure detector which indicates that there is covert movement.] Question: What would one need to say in order to apply the Frazier and Fodor idea to account for an LF constraint against crossing dependencies? 4. Superiority in Bulgarian Problem #2: Our generalization is wrong There is evidence from Bulgarian against the CCD: (5)

a. Koj1 kakvo2 t1 vižda t2? who what sees

cf. Who sees what?

Moreover, in Bulgarian crossing dependencies are preferred to nested dependencies: (6)

Superiority Effect in Bulgarian (Rudin 1988) The leftmost wh-phrase in a Bulgarian multiple question is the wh-phrase that moves overtly in the corresponding English multiple question.

(7)

a. Koj kakvo vižda? who what sees

cf. Who sees what?

b.*Kakvo koj vižda?

what who sees

cf. *What does who see?

(8)

a. Koj k´de udari Ivan who where hit Ivan

cf. Who hit Ivan where?

b.*K´de koj udari Ivan

cf. *Where did who hit Ivan?

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

ADVANCED SYNTAX 2007 ANAGNOSTOPOULOU/FOX

3

5. Richards’s Proposal Three components: 1. A derivational Theory of the effects we’ve seen in English: Attract Closest (Kitahara 1994, 1997, building on Kuno and Robinson 1972, Chomsky 1973, 1993, 1995) 2. Elimination of the strict cycle condition in favor of “featural cyclicity” (Chomsky 1995) 3. Tucking in (shortest move)1 5.1. Kuno and Robinson on Superiority in English (9)

An early statement of superiority A wh word cannot be preposed crossing over another wh. [Kuno and Robinson 1972]

This explains (4), but not (2-3): (4) a. You asked who1 Mary told t1 [PRO to present what]. b. *You asked what1 Mary told who [PRO to present t1]. (2) a. ??[Which book]1 did you ask who2 Mary told t2 [PRO to present t1]? b. *[Which person]1 did you ask what2 Mary told t1 [PRO to present t2]? (3) a. This is the violin wh1 that I wonder which sonatas2 to play t2 on t1. b. *These are the sonatas wh1 that I wonder which violin to play t2 on __. 5.2. Kitahara Chomsky’s account of superiority (4) Attract closest: Every instance of wh-movement to C must be movement of the highest wh-phrase in the c-command domain of C. Kitahara: this can also account for the PCC (2-3), if modified as follows: Every instance of wh-movement to C must involve movement of the closest moveable wh-phrase.2

5.3. Strict Cycle, the Extension Condition or Feature Cyclicity Island conditions require a principle of cyclicity. (10) Extension Condition: every instance of merge (internal, or external) must extend the structure. 1 2

With a proposed unification with shortest move, which we will skip.

This is slightly different from Kitahara’s actual proposal, but will do for our purposes.

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

ADVANCED SYNTAX 2007 ANAGNOSTOPOULOU/FOX

4

This condition would yield Bulgarian structures with the opposite order than that attested. (11) Feature Cyclicity: If a head H needs to attract an XP, attraction must take place before any other operation. Possible motivation: Late Merger. 5.4. Shortest Move Consider a stage of the derivation of a multiple wh-question just before wh-movement takes place: (12)

C +wh…Wh-phrase1 vižda Wh-phrase2?

At this point two things can happen: either wh-phrase1 or wh-phrase2 does. Attract closest determines that wh-phrase1 moves before wh-phrase2. This is the Chomsky-Kitahara explanation for English Superiority effects: (13)

Wh-phrase1 C+wh …t1 vižda Wh-phrase2?

If the Extension condition were postulated, we would get the wrong prediction for Bulgarian. However, if tucking-in derivations are allowed, the Bulgarian structure in (14) would be possible. Shortest move, insures that it is the only possible structure. (14)

Wh-phrase1 Wh-phrase2 C+wh …t1 vižda t2?

5.5. New Prediction: A preference for crossing dependencies in Bulgarian. Consider in greater detail the way Kitahara derives nested dependencies in English (15)

C +wh…Wh-phrase1 vižda Wh-phrase2?

Shortest move determines that wh-phrase1 moves to [Spec,CP]. Now another CP is constructed: (16)

C +wh …Wh-phrase1 C+wh…t1 vižda Wh-phrase2?

At this point there is only one wh-phrase that can be moved. Movement results in a minor violation of subjacency (wh-island). The only way to derive a crossing dependency would involve a violation an early violation of attract closest. However, that if Wh-phrase2 were able to move to become a specifier of CP, we would predict the following (given shortest move):

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

ADVANCED SYNTAX 2007 ANAGNOSTOPOULOU/FOX

(16')

C+wh …Wh-phrase1 Wh-phrase2 C+wh…t1 vižda t2?

Which given attract closest would be transformed as follows to a crossing dependency

(17)

Wh-phrase1 C+wh … Wh-phrase2 t1 C+wh…t1 vižda t2?

Richards (2001) discovered that this is the attested pattern. 5.6. Evidence that the higher wh-phrase moves first (PMC) Principle of Minimal Compliance: Only the first element that is the specifier of a X is subject to subjacency, shortest move, and attract closest. Spell-out the predictions 5.7. Other constructions that show Bulgarain-type Superiority

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

5

Anagnostopoulou / Fox

Advanced Syntax, Spring 2007

WEAK ISLANDS AND RELATIVIZED MINIMALITY

1. BACKGROUND. ECP EFFECTS, XP-BARRIERS AND MINIMALITY BARRIERS Recall two types of ECP effects: (i) Adjunct-Object asymmetries (1)

a. b. c. d.

*Whyi did you wonder whether I bought the book ti *Howi did you wonder whether I bought the book ti *Wherei did you wonder whether I bought the book ti *Wheni did you wonder whether I bought the book ti

vs.

e.

??Whati did you wonder whether I bought ti

(ii) Subject-Object asymmetries (2)

a. b.

*Who do you think [that [_ saw John]? Who do you think [that [John saw _] ?

(3)

a. b.

*Who do you wonder [how/whether [_ bought the book]? ??What did you wonder [how/whether [he bought t]

ECP: (4) (5)

A non-pronominal empty category must be properly governed α properly governs β iff α θ-governs or antecedent-governs β

-Barriers treatment of (1), (3) in terms of XP-barriers: adjunct and subject trace are not θ-governed, hence must be antecedent governed. The embedded CP is a barrier by inheritance, blocking antecedent government of the adjunct and the subject trace. Hence, an ECP violation. -Analysis of the that-trace effect in (2) in terms of the (Rigid) Minimality notion of barrier: (6)

….α….[γ’…..γ0…..β….]

1

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

γ’ is a barrier even though it is not an XP because it dominates a head γ0 which is a closer governor of β than α. This type of barrier (which reduces ambiguity of government) is relevant only for government (ECP). C’ is a minimality barrier for the intermediate trace, which cannot antecedent govern the subject trace resulting in an ECP violation: (7)

a.

*Who do you think [t’ [C’…that [ t saw John]?

The version without “that” is ok because a zero Comp does not have enough content to

make C’ into a barrier.

The minimality analysis also applies to the CNPC of the Noun-Complement type:

(8)

how did John announce [NP a plan [CP t2 to [t1 fix the car t]]

Even if they are no barriers for movement N’ is a minimality barrier for government of t2.

2. TWO MAIN MOTIVATIONS BEHIND RELATIVIZED MINIMALITY …..with respect to XP-barrierhood (a) The analysis of adjunct / object asymmetries in terms of XP-barrierhood does not generalize to two other cases where similar effects are found. It is desirable to provide a uniform treatment for these cases. Obenauer’s (1984) pseudo-opacity effects In French, the wh-quantifier combien (how much/ many) can optionally undergo leftbranch extraction: (9)

a. b.

[Combien de livres] How many of the books Combien a-t-il How many did he

a-t-il consultés t did he consult t consultés [t de livres] consult of the books



Certain VP-adverbials like beaucoup block combien-extraction while they do not interfere with movement of the whole object: (10) a. b.

[Combien de livres] How many of the books *Combien a-t-il How many did he

a-t-il beaucoup consultés t did he a lot consult t beaucoup consultés [t de livres] a lot consult of books

2

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

How can we make sure that combien-extraction is similar to adjunct extraction out of wh-islands? Two considerations support the view that combien-extraction is subject to the same locality conditions as adjunct extraction: -Extraction of combien across a wh-island gives rise to a strong (ECP) violation: (11) a. b.

?Combien de problèmes sais-tu [comment [PRO résoudre t t ]] How many of problems know-you how solve ‘How many of the problems do you know how to solve’ ?Combien sais-tu [comment [PRO résoudre [t de problèmes] t ]] How many know-you how solve of problems ‘How many of the problems do you know how to solve’

-Extraction of a VP-adjunct across beaucoup is ungrammatical (ECP): (12) a. b.

Comment a-t-il résolu [beaucoup de problèmes] t How has-he solved [many of problems] ‘How did he solve many of the problems?” *Comment a-t-il beaucoup résolu [t de problèmes] t

Combien extraction across beaucoup cannot be dealt with in the Barriers system because VP is not a barrier (and there is no place for accommodating the effect caused by the VPadverb).1 Ross’s inner islands Negation blocks extraction of adjuncts but does not affect extraction of arguments: (13)

a. b.

Bill is here, which they (don’t) know Bill is here, as they (*don’t) know

(14)

a.

It is for this reason that I believe that John was fired [ambiguous-reason adverbial modifies the main clause or the embedded clause] It is for this reason that I don’t believe that John was fired [unambiguous: only the main clause is modified]

b. (15) a. b.

How strongly do you believe that inflation will rebound [ambiguous-degree adverbial modifies the main clause (strength of belief) or the embedded clause (strength of inflation’s rebound] How strongly do you not believe that inflation will rebound

[only about strength of disbelief]

1

Well, that is not quite true. If the presence of a specifier determines the presence of V’ (see below) one could, perhaps, appeal to V’ as a minimality barrier. But wh-islands and beaucoup intervention are very similar, and yet the accounts that can be given for them in the barriers framework very different.

3



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(16) a. b.

What do you believe he weighed last week [ambiguous: “potatoes” or “200 pounds”] What do you not believe he weighed last week

[unambiguous: “potatoes” ]

Inner islands are created by other operators as well (the ones licensing NPIs, i.e ‘affective’ operators, Klima 1964): (17)

a. b.

It is by lethal injection that many people believe that John was executed *It is by lethal injection that few people believe that John was executed

(18)

a. b.

Few people did anything *A few people did anything

(19) a. b.

Why do few people think that Bill was fired (no long distance construal for why) Why do a few people think that Bill was fired (ambiguous)

A unified account between these cases and wh-islands is desirable. Wh-islands, pseudo-opacity islands and inner islands are called weak islands (they weakly or not at all influence argument extraction while they block adjunct extraction). ….with respect to Minimality-barrierhood: Rigid minimality barriers are problematic for various reasons: (a) Minimality barriers are never created by V0 and I0. In Barriers this is explained as follows: (i) it is stipulated that V’ is not projected because the VP has no specifier. (ii) It is stipulated that I is intrinsically defective. (b) Adjunct extraction is never blocked by an intervening head, unlike subject extraction: (20) How do you think [t’ that [Bill solved it t ]] In order to account for this, Chomsky 1986 stipulates that that is deleted at LF, the level where γ-marking takes place for adjuncts. Chomsky does not discuss cases where C is realized as an inflected auxiliary (e.g. How did Mary solve the problem), but these can be taken care of if it is assumed that I-to-C movement takes place at PF. (c) A conceptual issue: There is an asymmetry in the definition of minimality barriers. There are two types of government: head government and antecedent government. The asymmetry lies in the fact that an intervening potential head governor blocks both head government and antecedent government, while an intervening potential antecedent governor does not block any type of government. Rizzi proposes to explore the 4

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

consequences of a symmetric approach to minimality, according to which an intervening potential head governor blocks head government and an intervening potential antecedent governor blocks antecedent government. At an intuitive level, this immediately unifies wh-islands, pseudo-opacity islands and inner islands. In all cases, an intermediate potential antecedent governor (the wh-phrase, the VP-adverb beaucoup and negation) blocks the relationship between the (no-theta governed) antecedent (adjunct wh-phrase or combien) and its trace, resulting in an ECP violation. 3. LOCALITY AND RELATIVIZED MINIMALITY: GOALS AND DEFINITIONS Rizzi’s Locality Theory Starting point: There are two fundamentally different types of movement. Local movement: NP movement, head movement, clitic movement, adjunct A’ movement. Non Local (long) movement: A’ movement of arguments subject to weaker locality. Assumption: for the different types of local movement a uniform statement of locality conditions can be given. Task of locality theory: (i) A principled theory of locality conditions on local movement. (ii) A principled account for long A’ movement of arguments. Chapter 1. Strict locality: antecedent government (XP-barriers + relativized minimality).2 Chapter 2. No rigid minimality. That-t effects are not locality effects. They result from the Proper Head Government Condition on traces. Chapter 3. Long A’ movement of arguments. Arguments carrying a referential theta-role have a referential index which permits them to enter a binding relationship (no strict locality which restricts the (antecedent) government relation between operators and variables). Two main points of criticism (Frampton 1991): (a) Very little discussion of Subjacency and the conditions under which it obtains (b) No principled partition of positions into A and A’ positions. Rizzi’s Locality System a) He remains vague w.r.t. Subjacency. b) ECP (initial version, p.4): (21) 2

A nonpronominal empty category must be

The view that strict locality results from antecedent government is in agreement with Barriers.

5



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(i) Theta governed or (ii) antecedent-governed c) He assumes that there are two types of government defined as follows: (22) Head Government:

X head-governs Y iff (i) X ∈ {A, N, P, V, Agr, T } (ii) X m-commands Y (iii) no barrier intervenes (iv) Relativized Minimality is respected

(23) Antecedent Government:

X antecedent-governs Y iff (i) X and Y are coindexed (ii) X c-commands Y (iii) no barrier intervenes (iv) Relativized Minimality is respected

d) He is vague as to what definition of barrier he adopts. If the definition by Chomsky (1986), then there is some redundancy between barrierhood and RM. The notion of barrier he tentatively adopts (p. 6) is that XPs that are not directly selected by [+V] elements are inviolable barriers for government. This subsumes, the adjunct CED, the subject CED, the CNPC of the RC type, the CNPC of the complememt type, i.e. the, so called, strong islands (even though the CNPC of the complement type looks rather like a weak island). In fn. 6 he mentions that RM does not apply even to weak islands, such as the complements of factive verbs (he adopts Kiparsky & Kiparsky’s analysis according to which factive complements are dominated by a DP node): (24) *How do you regret that he solved the problem e) Relativized Minimality (25) Relativized Minimality X α-governs Y only if there is no Z such that (i) Z is a typical potential α-governor for Y (ii) Z c-commands Y and does not c-command X Intervener for head government: (26) Z is a typical head governor for Y= Z is a head m-commanding Y Interveners for antecedent government Partition:

A-chains A’-chains X0-chain

(NP-movement) (wh-movement) (head-movement)

6



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(27) a. b. c.

Z is a typical potential antecedent governor for Y, Y in an A-chain = Z is an A specifier c-commanding Y Z is a typical potential antecedent governor for Y, Y in an A’-chain = Z is an A specifier c-commanding Y Z is a typical potential antecedent governor for Y, Y in an X0-chain = Z is a head c-commanding Y

The class of elements that block antecedent government depends upon the type of chain that is formed (A, A-bar, head). ‘relativized: relative to the type of movement.

4. HOW IT WORKS 4.1. A’ Chains Wh-Islands (28)

*How do you wonder [which problem [PRO to solve t t’]]

Even if adjunction to IP is allowed, RM is violated: (29) How do you [t’ [ wonder [which problem [t’’ [PRO to [t’’’ [solve t t’’’’]]]]]] t’ cannot antecedent govern t’’ because the A’ specifier which problem intervenes. Desirable consequence: the stipulation that there can be no adjunction to IP can be dispensed with. Absence of that-trace effects (30) How do you think [t’ that [Bill solved it t’’]] No RM effect, i.e. no A’ specifier blocking antecedent government of any of the traces in

the chain.

Desirable consequence: no need to resort to Rigid Minimality and “that-deletion” at LF.

Pseudo-opacity Recall: (10) a. b.

[Combien de livres] How many of the books *Combien a-t-il How many did he

a-t-il beaucoup consultés t did he a lot consult t beaucoup consultés [t de livres] a lot consult of books 7



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(12) a. b.

Comment a-t-il résolu [beaucoup de problèmes] t How has-he solved [many of problems] ‘How did he solve many of the problems?” *Comment a-t-il beaucoup résolu [t de problèmes] t

Analysis: beaucoup is an A’specifier of VP blocking antecedent government of the trace of combien and comment by e.g. the intermediate trace adjoined to VP. Compare (12) to (31): (31) Pourquoi Why

a-t-il beaucoup has-he many

résolu [t de problèmes] t solved of the problems

This is grammatical because pourquoi is a VP-external adverb (unlike manner adverbs). Improvement under en cliticization: (32) Combien How many

en of them

a-t-il beaucoup has he a lot

consultés consulted

Analysis: [combien en] moves to AgrO above beaucoup and then combien-extraction takes place. Evidence: en-cliticization triggers participle agreement: (33)

a. b.

Combien How many Combient How many

a-t-il conduit did he drive en a-t-il of them did he

de voitures of cars conduites

drive+Agr

Inner islands Recall the two subcases: Sentential negation: (13)

a. b.

Bill is here, which they (don’t) know Bill is here, as they (*don’t) know

(14)

a.

It is for this reason that I believe that John was fired [ambiguous-reason adverbial modifies the main clause or the embedded clause] It is for this reason that I don’t believe that John was fired [unambiguous: only the main clause is modified]

b.

Affective operators: (17)

a.

It is by lethal injection that many people believe that John was executed 8

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(34)

b.

*It is by lethal injection that few people believe that John was executed

a.

It is for this reason that everyone believes that Bill was fired (both readings) It is for this reason that no one believes that Bill was fired (only high construal)

b.

For sentential negation: pas and not are A’specifiers (of T or Neg) blocking antecedent government of adjunct traces. For affective operators: Affective operators undergo overt (with I-to-C movement) or covert movement to Spec,CP (non.affective ones undergo adjunction to IP), hence they occupy an A’ position and they block antecedent government (when the spec,CP position is filled and the affective operator is a subject as in What did no one say, Spec,IP counts as an A’ position (LF movement of the subject to C could not take place because it would create a Superiority violation as in What did who say). 4.2. A-Chains Super-raising: (35)

*John seems that it is likely [t to win]

RM: “it” blocks Antecedent Government between “John” and its trace. Necessary ingredient: arguments in A-movement constructions must be antecedent governed. In Barriers Chomsky says that for raising on the basis of the assumption that subjects are not theta-governed. However, the need for Antecedent Government also holds for passives: (36)

a. b.

It seems that Bill was told that… *Bill seems that it was told that…

Rizzi (1990: 93): the superraising cases fall under the Theta Criterion: (37)

(i) Each Theta position belongs to a chain containing exactly one argument (ii) Each argument belongs to a chain containing exactly one Theta position

“…if the Theta Criterion is defined in terms of chain, and chain is defined in terms of antecedent government [..], then in (36b) (Rizzi’s (41c)) the object theta role cannot be assigned to the appropriate formal object (a chain containing the argument Bill…)…”. Rizzi (2001) discusses further these cases. He points out that one could link the superraising cases to the impossibility of extracting out of a tensed clause: 9

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(38)

*John seems that t will win

(35) and (36b) would then not fall under RM. He points out, though, that there are languages (dialects of Turkish, Moore 2003) that permit raising across a tensed clause but not when a subject is present in support of RM. SSC effects with clitics in causatives? In causative constructions, a dative clitic cannot appear in the main clause in the presence of a subject in the embedded clause: (39) a.

b.

c.

Jean a laissé Pierre parler John has let Peter speak 'John has let Peter speak to Mary' Jean l' a laissé parler John him has let speak 'John has let him speak toMary'

*Jean lui a laissé Pierre Jean to her has let Peter 'Jean has let Peter to speak to her'

a to

Marie (French) Mary

a to

Marie

Mary

parler

speak

4.3. Head Chains Main facts falling under the Head Movement Constraint (Travis 1984): (40)

a. b. c.

They could have left Could they have left? *Have they could left?

In RM terms: in (40c) “have” does not antecedent govern its trace because “could” intervenes. 4.4. Head Government Captures the fact that ECM cannot take place across a C: (41)

a. b.

*John tried [CP C [IP Bill to win]] *John wonders [how C [IP Bill to win]]

Desirable consequence: the inheritance clause of Barriers not needed (but note that Rigid Minimality would also work). 5. THAT-TRACE EFFECTS AND PROPER HEAD GOVERNMENT Core facts of the that-.trace effect: (42) a.

Who do you think [0 [t read the book]] ? 10

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

b. c.

*Who do you think [that [t read the book]] ? *Who did you pray [for [t to do it]]?

RM radically departs from previous approaches which take “that” / “for” to block the relationship between the subject trace and the intermediate trace. For RM there is no locality violation in (43): (43)

[CP t’ that [IP t….

Rizzi proposes that the “that-t” effect follows from the Proper Head Government Condition: (44)

A non-pronominal empty element must be properly head governed

What does ‘proper’ mean? Not ‘lexical’. Infl is a proper governor (the VP-topicalization cases discussed also by Chomsky): (45)

and go home I wonder whether he did

To capture the complement /specifier asymmetry (Infl governs VP but not the subject) Rizzi proposes that proper government is ‘government within the immediate projection of a head’, i.e. within X’. (He also considers a linear definition but dismisses it for several reasons.) He incorporates the PHGC into ECP, a conjunctive definition: (46)

ECP: a non-pronominal empty category must be (i) properly head governed (Formal Licensing) (ii) theta-governed or antecedent governed (Identification)

Evidence for the Proper Head Government Requirement: -Heavy NP Shift: *[ t are intelligent] all the students who can solve the problem This should be ok because it is antecedent governed. But it is not because it violates the PHGR. vs.

I would like to introduce t to Mary all the students who can solve this problem

-Extraction of Measure Phrase from AP 11

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

*How is he [t tall]? This can be derived from the PHGR if the adjective does not properly head governed the trace. In languages where this is good (e.g. Italian), the measure phrase can be to the right of A as an adjunct to AP (the adjoined phrase can be properly head governed by a higher head, e.g. T). How the facts are derived: 1. That-t effect: t is not properly head governed. Infl does not govern the subject trace within its immediate projection and “that” is inert for government. 2. *Who did t see Mary? t is not properly head governed by I in C because in (47) I does not govern t within its immediate projection (its immediate projection is C’) and C is inert for government: (47)

CP 3 C’ 3 C IP ! 3 I t

3. V-2 languages: (48) Wer hat sie gesagt [CP t’ ist [t gekommen ]? Who has she said is come ‘Who did she say came?’ In V-2 languages C carries morphosyntactic features which attract V and make C an appropriate proper head governor. that-t effect obtains in at least some German dialects: (49)

a. b.

? Was glaubt Hans dass Fritz gestohlen hat? What believes Hans that Fritz stolen has ‘What does Hans believe that Fritz stole?’ *Wer glaubt Hans dass das Auto gestohlen hat? Who believes Hans that the car stolen has ‘*Who does Hans believe that stole the car’

4. Adjuncts 12

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(50)

a. b.

*Who do you think that t came? *Who did t come?

a. b. c.

How do you think that he came t? How did he come? Why did John leave?

vs. (51)

The adjunct trace is properly head governed by V or T. Sentential adverbs like why can be directly base generated in Comp and hence they leave no trace (evidence: they can’t be left in situ, they do not trigger stylistic inversion). If adjuncts are adjoined to VP, they can be properly head governed by T across the one VP segment. If they are dominated by VP, as adjunct Aps predicated of the object, they are not extractable (T cannot govern across the VP, see page 50): (52)

a. b.

*How angry did you meet Bill? *How raw did he eat the meat?

5. Agreement in Comp (53)

a. b.

Who do you think [0 [t left]]? Who [0 [ t left]]?

When a zero Comp is in spec,head Agreement with a wh-operator or a trace, spec-head agreement makes Comp into an appropriate head governor. C in English is either ‘that’ or Agr. Agr must be licensed via coindexation with a specifier. The Agr/spec-head agreement in Comp analysis is supported by phenomena of agreement in Comp: (a) In Kinande, agreement in class. (b) aL in Irish (c) qui = que + Agr in French (when a trace is in C). (d) da becomes die in West Flemish A case of an agreeing +wh C. Norwegian som present only when the local subject moves: (54) a. b.

Vi wet [hvem *(som) [t snakker med Marit ]] We know who that talks with Mary Vi wet [hvem (*som) [Marit snakker med t]]

We know who that Mary talks with

6. Strategies of subject extraction: (i) Agreement in Comp turns C into a governor 13

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(ii) The trace is spelled out as a resumptive pronoun (in Swedish) (iii) Extraction takes place from postverbal position (the postverbal subject is taken to be in a VP-adjoined position properly governed by INFL, just as adjuncts are). 7. Relative clauses Relative clauses show the reverse that-t effect (anti that-t effect): (55)

a. b.

The thing that happened is terrible *The thing happened is terrible

a. b.

What do you believe that happened What do you believe happened

vs. (56)

In languages like French, subject relative clauses use the same form of complemetizer as subject questions: (57) La chose qui t est arrivee est terrible The thing that has happened is terrible (58) Que crois-tu qui est arrive? What do you think that has happened For English, the idea is to propose that the agreeing form of Comp is null in declaratives and ‘that’ in relatives. Support: there are languages that use a special complementizer in relatives distinct from declaratives (e.g. wo in Swiss German). He proposes that there are two features characterizing Comp: +/-wh, +/-pred(idactive) +wh –pred: +wh +pred: -wh +pred: -wh –pred:

I wonder what 0 you saw The thing which 0 you saw The thing Op that you saw I know that you saw it

(59) *The thing [Op Agr [t happened]] is terrible This is ruled out either (i) because a null operator cannot enter agreement due to the fact that it is an anaphor and anaphors are incompatible with agreement or (ii) because null operators must delete at LF and therefore they are not allowed to have syntactic features and license an agreeing element. No problem when the operator is overt (either because overt operators are anaphoric or because they have features that license agreement): (60) The thing [which 0 [t happened ]] is terrible 14



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Analysis of (61): (61)

The thing that happened is terrible

Standard analysis: operator transmits its index to ‘that’ turning it to an appropriate governor for the subject trace. For Rizzi: ‘that’ must be turned into a governor and it can’t agree with the Operator because of (59). Solution: ‘that’ is in agreement with the head of the relative. This is A agreement (agreement with an element in A position). It makes C an appropriate governor for the subject position.

6. REFERENTIAL INDICES AND THE ARGUMENT-ADJUNCT ASYMMETRY System so far: ECP : A nonpronominal empty category must be (i) properly head governed (ii) Theta-governed or antecedent governed Facts that can be accounted for (62) (63)

a. *Which student do you wonder [how [t could solve the problem t ]] b. ?Which problem do you wonder [how [PRO to solve t]] c. How do you wonder [which problem [PRO to solve t]] a. *Which student do you think [t that [t could solve the problem]] b. Which problem do you think [t that [Bill could solve t]] c. How do you think [t that [Bill could solve the problem]]

In Null Subject Languages the subject cases above are ok (correlating with free inversion). Analysis: (62a) violates clause (i) and clause (ii) of the ECP (62b) conforms with both clauses of the ECP (62c) violates clause (ii) of the ECP (63a) violates clause (i) of the ECP (63b,c) conform with both clauses of the ECP Problems Conceptual Redundancy: both clauses of the ECP refer to a kind of head government 15

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Disjunction: disjunctive formulation of the second clause is a problem: we are grouping together things without understanding why. Empirical Measure phrases -There are lexically selected adverbials which should satisfy the theta-government clause of the ECP, which are ruled out when extracted out of wh-islands: (64)

a. b.

John weighted apples John weighted 200lbs

(65)

?What did John wonder how to weigh?

Only the first reading survives. Idioms -Same with nominal parts of idioms: (66)

a. b.

What headway do you think you can make on this project *What headway do you wonder how to make on this project

-Long distance subject extraction

-Is predicted to be as bad as adjunct extraction but it is not:

(67)

a. ?*Who do you wonder whether we believe[t’ [ t can help us]? b. ??Who do you wonder whether we believe [ t’ [we can help t] c. *How do you wonder whether we believe [ t’ [ we can help Bill t]?

(65a) is not as good as (65b) but considerably better than (65c). The problem is that if we assume that intermediate traces are present only when necessary, then (67a) should be as bad as (67c). In both cases the t’ trace is needed to license the t trace and in both cases the t’ trace is not antecedent governed. Rizzi’s solution To eliminate the second clause of the ECP. 1. (68) ECP (final version) : A nonpronominal empty category must be properly head governed 2. Referential theta roles license referential indices 16

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

3. Arguments with referential theta-roles may enter binding (long-distance): (69)

X binds Y iff (i) X c-commands Y (ii) X and Y have the same referntial index

If no referential index is available operators must be connected to the variable via government. Government is local. Operators are connected to variables through a Chain. In (67)

a. ?*Who do you wonder whether we believe[t’ [ t can help us]? b. ??Who do you wonder whether we believe [ t’ [we can help t] c. *How do you wonder whether we believe [ t’ [ we can help Bill t]?

subject extraction is better than adjunct extraction because the operator can be connected to the subject trace t (the t’is.needed to satisfy ECP, the PHGC) via binding. The reason why subject extraction is worse than object extraction is that there is an intermediate trace, and intermediate traces lead to deviance. An interpretive asymmetry When an operator is connected to the variable via the chain scope reconstruction is possible: (52)

a.

b.

c. d.

Tell me what everyone gave t to Bill on his birthday Ambiguous: all gave the same thing, or there a potentially different thing Tell me what you think that everyone should give to Bill (still ambiguous, the ambiguity results from reconstruction because “Tell me who thinks that everyone left” cannot have scope of everyone over what, i.e. there is no LF raising to the higher Comp) Tell me what you don’t think that everyone should give to Bill (unambiguous) Tell me what you wonder why everyone gave to Bill (unambiguous).

7. SOME ISSUES RAISED BY FRAMPTON (1991) 1. Eliminating Antecedent Government, the Minimal Link Condition and Subjacency. Relativized Minimality can be seen as a constraint on movement rather than antecedent government. According to this interpretation, Move α is local if the movement does not 17 Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

skip any potential landing sites. This will permit elimination of the notion of antecedent government (it is not clear why it antecedent government and head government constitute a natural class / the disjunction behind the ECP is unsatisfactory).3 Subjacency is still needed. Hence: …an instance of Move α is local if it satisfies Subjacency and does not skip over intermediate landing sites…” (Frampton 1991: 5). 2. No principled explanation for the A vs. A’ position distribution of positions. Long discussion of the pseudo-opacity facts. Evidence that indeed “beaucoup” must be in spec,VP. Nothing can intervene between “beaucoup” and the verb. Even manner adverbs must be to the left of the QP-adverb: ()

a. b.

Il a soigneusement beaucoup replié de feuilles He has carefully many folded of paper-sheets ‘He has carefully folded many sheets of paper’ *Il a beaucoup soigneusement replié de feuilles He has many carefully folded of paper-sheets

Two QP adverbs cannot cooccur: ()

*Il a beaucoup beaucoup aimé de femmes He has many much loved of women ‘He has loved many women a lot’

The order of VP peripheral elements are Floated Q>Manner Adverbs >QP-adverbs (compatible with Koopman & Sportiche’s version of the VP-internal subject hypothesis). Why would the spec,VP position qualify as an A’ position? Potential answer: it is the QPadverb that makes it an A’ position. It is the nature of the intervening elements, not the position itself, which blocks chain formation. 3. If movement respects RM, and if the object moves to AgrO, then how is the object moved across the subject and the subject moved across the raised object without violating RM? [an issue extensively discussed in the Minimalist literature, especially the literature on Equidistance following Chomsky 1993, 1995]. 4. Rizzi’s execution of inner islands (negation is not a head but an XP in spec of TP, affective operators move to spec,CP at LF, spec,IP can be an A’position) has many problems.

3

Note that The Minimal Link Condition of later research is understood as a principle according to which movement cannot skip potential landing sites, as suggested by Frampton.

18

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-Spec,IP an A’ position as a Last Resort: No principled explanation for the A / A’ partition. No Superiority expected by what did who see (‘who’ could be interpreted in Spec,IP). -Sentential negation as an XP: But not blocks head movement operations (e.g. Infl lowering resulting in do-support ), unlike never. Alternative: NegP is present when sentential negation and affective operators are present. An Op is hosted in spec, NegP inducing RM effects.

19

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Anagnostopoulou / Fox

Advanced Syntax, Spring 2007

MINIMALIST INQUIRIES AND DERIVATION BY PHASE – MAIN POINTS [This handout contains material from Winfried Lechner’s handout “Economy in the Minimalist Program” David Pesetsky’s handouts “Minimalist Inquiries”, “Derivation by Phase”] Computational Operations a. Select (I) Select [F] from the universal feature set {F} (II) Select LEX, assembling features from [F] (III) Select LA (lexical array) from LEX (IV) Map LA to EXP, with no recourse to [F] for narrow syntax b. Merge: "takes two syntactic objects (a, b) and forms K(a, b) from them. c. Agree: "establishes a relation (agreement, Case-checking) between an LI a and a feature F in some restricted search space (its domain)." d. Move: combining Merge and Agree. [A-movement if motivated by a phi-feature; Abar if motivated by a P ["peripheral"]-feature] “The operation Move establishes agreement between a and F and merges P(F) to aP, where P(F) is a phrase determined by F (perhaps but not necessarily its maximal projection) and aP is a projection headed by a. P(F) becomes the Spec of a ([Spec,a]).

Notes on Operations: 1) Lexical Array: A collection of Lexical Items. Lexical Array is a Numeration if some lexical items are selected more than once (DbP: p. 11) Recall: The NUMERATION (N) contains a subset of the lexicon (and a subscript indicating the number of occurrences of each item) N = {boy1, girl1, the2, see1} (1)

a. The numeration N contains indexed lexical items. b. For each lexical item, the index marks the number of times this lexical item is 1



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

used c. Each time the derivation accesses the numeration, it selects a single lexical item. d. Each time the derivation selects a lexical item, its index is reduced by 1. e. The derivation stops if the numeration has been exhausted and all indices equal 0. 2) “Chains are determined by identity, with no need for indices or some similar device to distinguish chains from repetitions, also violating the Inclusiveness Condition….” (DbP: p. 11)

Occurrences -Move creates two occurences of a single a, where an "occurence of a" is the full context of a. - "Chain" is a set of occurences. If occurences are "full contexts" we don't need to say that a chain is asequence, since there will be a containment relation between the contexts that allows us to reconstruct whatever we might needed the ordering property of a sequence for.

Deriving Procrastinate Move is more complex than its subcomponents.

Move is more complex than even its subcomponents together -- since it involves the extra

step of determining pied piping.

Consequently:

(2)

Merge or Agree "preempt" Move.

"This yields most of the empirical basis for Procrastinate", p. 102

Selectional properties of functional heads: Lexical items fall in two types, substantive and functional Core Functional Categories (CFC) a. C b. T c. v SELECTING PROPERTIES OF CFCs 2

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

A) S-SELECTION selected: -C can be unselected (root).

-T and v must be selected.

-C is selected only by substantive categories.

-T is selected by either C or V. If it is selected by C it has the complete set of φ-features.

If it is selected by V it is defective for φ-features (Tdef).

-v is only selected by T.

selecting: -C selects T.

-T selects verbal elements.

-v also selects verbal elements but in addition it may select a nominal element in the

specifier position, the external argument.

B) EPP-SELECTION, MOVE AND PURE MERGE Each functional head my have an EPP-requirement, i.e. allows an extra Spec beyond its s-selection. For v, this is the case in object shift constructions. For T this is the EPP. For C, in overt wh-movement. Move is triggered to satisfy an EPP requirement of a head. Pure Merge (with an expletive) is also possible, but only in T. Not in v and C. The ban on expletive merge in v is attributed to (3): (3)

Pure merge in theta position is required of (and restricted to) arguments

(3) also forbids arguments from merging directly with T or C instead of v.

Phases Two motivations for phases:

a) Conceptual

Complexity Considerations:

(i) Simple Operations preempt more complex ones (ii) Search space is limited (locality) 3

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(iii) Access to the feature set F is restricted by the procedure repeated from above: (I) Select [F] from the universal feature set {F} (II) Select LEX, assembling features from [F] (III) Select LA (lexical array) from LEX (IV) Map LA to EXP, with no recourse to [F] for narrow syntax (iv) Computation is locally determined (no look-ahead).

b) Empirical

Introduced to resolve the problem of sentences with expletives like:

(4) (5)

It is fun PRO to t go to the beach It is about time PRO to leave t It was decided PRO to be executed t at dawn There was assumed to be a reason why a man is in the garden

where the expletive is merged after Move takes place.

Recall the cases that motivated the Merge over Move condition:

(6a) and (6b) start with the same numeration, but differ in the way the derivation

proceeds.

(6)

a. There seems to be someone here

b. *There seems someone to be here

(7) N0 = {there1, T2, seems1, to1, be1, someone1, here1} In (6a), the expletive is merged into the lower SpecTP position: (8) Therei seems ti to be someone here a. N0 = {there1, T2, seems1, to1, be1, someone1, here1} b. [TP T[EPP] to be someone here] Nk = {there1, T1, seems1, to0, be0, someone0, here0}

L c. [TP there T[EPP] to be someone here] Nl = {there0, T1, seems1, to0, be0, someone0, here0}

d. [TP there T[EPP] to be someone here] Nm = {there0, T1, seems1, to0, be0, someone0, here0}

e. [TP T[EPP] [VP seems [TP there T[EPP] to be someone here]]] Nn = {there0, T0, seems0, to0, be0, someone0, here0} f. [TP therei T[EPP] [VP seems [TP ti T[EPP] to be someone here]]] g. [TP therei T[EPP] [VP seems [TP ti T[EPP] to be someone here]]]

Merge T°

Merge expletive there

Check EPP

Merge T° Move expletive Check EPP

4

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

In (6b), the subject someone is moved to the lower SpecTP position, and the expletive merged into the higher SpecTP: (9) *There seems someonei to be ti here a. N0 = {there1, T2, seems1, to1, be1, someone1, here1} b. [TP T[EPP] to be someone here] Nk = {there1, T1, seems1, to0, be0, someone0, here0} c. [TP someonei T[EPP] to be ti here] Nl = {there1, T1, seems1, to0, be0, someone0, here0} d. [TP someonei T[EPP] to bei here] Nm = {there1, T1, seems1, to0, be0, someone0, here0} e. [TP T[EPP] [VP seems [TP someonei T[EPP] to be ti here]]] Nn = {there1, T0, seems0, to0, be0, someone0, here0} f. [TP therei T[EPP] [VP seems [TP someonei T[EPP] to be ti here]]] g. [TP therei T[EPP] [VP seems [TP someonei T[EPP] to be ti here]]]

Merge T° Move someone Check EPP Merge T° Merge expletive there Check EPP

Given the Merge over Move analysis of (6), it is not clear why raising is ever possible. In examples like (4) & (5), the expletive could merge to satisfy the EPP property of T and movement of PRO or a man should be blocked. Chomsky’s answer: expletives are not always available. If only a subset of LA (Lexical Array) is available to derivation, then the expletive in (4)&(5) is not in the subarray and hence not available. Thus, EPP motivates Move. Chomsky (1998: 19) proposes that at each stage of the derivation a subset LAi is extracted, placed in active memory and submitted to the derivational procedure. Once LAi is exhausted, the computation may proceed or it may return to LA and extract LAj proceeding as before. The proposal then is that in the above sentences, the expletive is not contained in the subarray LAi and therefore it cannot be accessed. Characterization of subarrays: they determine a natural syntactic object which is taken to be the closest counterpart to a proposition: either a verb phrase with all theta-roles assigned, or a full clause including tense and force. Definition of Phases: phases of a derivation are syntactic objects which are derived by choice of a subarray LAi. CPs and vPs constitute phases while TPs not. vPs constitute phases only when they are transitive, not when they are passive or unaccusative (Chomsky defines phases as being propositional rather than convergent because of complexity considerations). Merge over Move: when the expletive is in the same phase as the element that could undergo raising.

5

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Phases and Cyclicity 1) Strong Cyclicity Condition Chomsky proposes that phases must satisfy the strong cyclicity condition in (10): (10) The head of a phase is "inert" after the phase is completed triggering no further operations. According to (10), a phase head cannot trigger Merge or Attract is a latter phase. This means that derivations proceed phase by phase. This derives the effects of ‘strength’ in Minimalism-Stage 1 and of the Strict Cycle Condition (in earlier frameworks). 2) Phase Impenetrability Condition (PIC) strong version Strict version (Minimalist Inquiries; see below for a weaker definition) (11) In phase α with head H, the domain of H is not accessible to operations outside α, but only H and its edge This forces Move of a phrase inside the phase to Spec,H in order for this phrase to be accessible to operations outside α, i.e. a strong form of Subjacency. Consequences -wh-movement must proceed through the edge of C, i.e. successive cyclic movement through Spec,CP. -an object that undergoes wh-movement undergoes first (covert) object shift—in order to be accessible for wh-movement. (recall the adjunction to VP-extraction strategy in barriers) -To implement this without having "look ahead" problems, Chomsky (2000) proposes that that an EPP or P (force-topic-focus etc.) feature may be optionally assigned to the head of a phase [unlike the EPP-feature of T which might be Universal, EPP or P features on v or C are properties of phases rather than head). A long discussion on the assignment of EPP-features on v in connection with the [±Object Shift] parameter is included in Chomsky (2001). 3) Cyclic Spell-Out In Chomsky (1995, footnote 50) it is noted that there are -interpretable features that may have a PF reflex, for example φ-features on T, v or Case on DPs. This is problematic if 6

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

the checking relation is overt: the -Interpretable feature is checked and erased before Spell-Out but its phonetic matrix remains. To resolve the problem, Chomsky (2000) proposes that instead of single Spell-Out there is cyclic Spell-Out: deleted features are erased but only after they are sent to the phonological component—possibly at the phase level. Spell-out applies cyclically in the course of a derivation. Cyclic Spell-Out is contingent on feature checking operations. In the previous model, there where two cycles, the cycle of overt and the cycle of covert derivations. With cyclic Spell-Out, there is only one cycle and all operations are cyclic. The issue is stated more clearly in DbP (p. 5): “The operation Spell-Out removes LF-uninterpretable material from the syntactic object K and transfers K to the phonological component. It must therefore be able to determine which syntactic features are uninterpretable, hence to be removed. Prior to application of Agree, these are distinguished from interpretable features by lack of specification of a value [see below for details on Agree]. After application of Agree the distinction is lost. To operate without reconstructing the derivation, Spell-Out must therefore apply shortly after the uninterpretable features have been assigned values […]. Spell-Out must be strongly cyclic (as assumed in MI) […]. In contrast to Extended Standard Theory-based systems, this system has no overt-covert distinction with two independent cycles; rather it has a single narrow-syntactic cycle. Furthermore, the phonological cycle is not a third independent cycle, but proceeds essentially in parallel”. 4) Phase Impenetrability Condition (PIC): weaker version In DpP he modifies the PIC in order to capture the fact that the edge of a lower phase is Spelled-Out in the next higher phase. General Guiding Principle: (12) Ph1 is interpreted/ evaluated at the next relevant phase Ph2 PIC (modified version) given a configuration [ZP Z….[HP a [H YP]]] (13) The domain of H is not accessible to operations at ZP, only H and its edge are accessible to such operations A consequence: under the modified version, T can access the domain of v. Under the strict version it cannot. He writes (p.14): “But T in the domain of Z can agree with an element within its complement, for example, with the in-situ quirky nominative object of its v*P complement…”

7

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Comment: ….but the issue probably does not arise since nominative objects occur with an unaccusative v (a weak phase), unless the PIC also holds for weak phases….

Agree, Probes, Goals, Incomplete Agreement, Activity, Freezing, Defective Intervention A functional head T or v has uninterpretable φ-features. The φ-feature set is a probe that seeks a goal, namely "matching" features that establish agreement. The relation of the probe of T to its goal is the T-associate relation. Locating the goal (the φ-features of a DP with structural Case), the probe erases under matching. Taking structural Case to be an uninterpretable feature, it too erases under matching with the probe. The erasure of uninterpretable features of probe and goal is the operation Agree. (14) T be elected an unpopular candidate • • • •

T has φ and EPP features. Probe: φ-features of T Goal: an unpopular candidate, which has matching features. P(G): "pied piping" of a phrase determined by the goal of T's probe

"...taking structural Case to be a reflex of an uninterpretable φ-set, it too erases under matching with the probe." Movement = • selection of P(G) • move of P(G) • feature-deletion under match (Agree)

Matching is a relation that holds of a probe P and a goal G. Not every matching pair induces Agree. To do so, G must be in the domain D(P) of P and satisfy locality conditions. General conditions on Agree: (15)

(I) Matching is feature Identity (II) D(P) is the sister of P (III) locality reduces to "closest c-command"

Feature Identity: identity in the choice of feature and not the value. Deletion: Deletion is taken to be a "one fell swoop" operation, dealing with the φ-set as a unit. Its features cannot selectively delete: either all or none. Chomsky claims that the φ-features of T do not agree with different NPs. He further assumes that only a probe with a full set of φ-features is capable of deleting the feature that activates the matched goal. 8

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Incomplete Agreement For example, a deficient T is assumed to be one lacking the complete set of φ-features. Thus it cannot check the Case of a DP (in successive cyclic raising) which is the feature activating the goal. Similarly, participle agreement does not yield Case checking because it is incomplete (no person agreement) Activity condition (16)

A goal must bear some uninterpretable feature [otherwise it is frozen in place].

Under the assumption that uninterpretable features serve to implement operations, it is structural Case that enables the closest goal G to select P(G) to satisfy EPP by Merge. -This is why structural case exists!

-The "character" of the Case (nominative, accusative)merely registers the identity of the

probe, so that "structural Case itself" is a single, undifferentiated feature. This is why

differently-cased DPs can interfere with each other.

For there-type expletives, the assumption is that they are active due to an uninterpretable person feature they have. This permits T to enter a Case checking relation with the Associate. For Quirky subjects, the assumption is that in addition to the inherent case feature, they also have a Structural Case feature that makes them active: «....Suppose quirky Case is (theta-related) inherent Case with a structural Case feature, as is often suggested in one or another form. Then it too is immobile once it reaches a Case-checking position. If the φ-features of T that check the structural Case of raised quirky subject themselves delete, we have default T; if they remain, we have remote agreement with some lower accessible nominative..." (Chomsky 1998: 43) [Note that there is a problem here. The whole thing is presented as if there is a choice. There isn't one though--or there shouldn't be one--since any T that is non-defective should erase its features immediately if possible. It seems that with respect to "freezing" quirky DPs behave as if they have structural Case--but with respect to "deletion of features of the goal" they do not. Note also the continuation: "In his detailed review of Icelandic agreement, Sigurðsson (1996) concludes that remote NOM allows number agreement but not first/second person agreement. That would follow if the [person] feature o T reduces to [3person] (the default choice) when it attracts quirky Case or EXPL in SPEC,T.." But note that these facts suggest that deletion is not a "one fell swoop" operation for quirky subjects and maybe expletives)

9

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

• Expletive there must have properties quite similar to Tdefective . • Since it moves around like a normal DP, it has some attractable feature, e.g. [person] -- call it G. [But since it is not phi-complete it does not delete features on T.] • But it does not delete the probe features, as witnessed by LD agreement. (17) there were declared guilty three men • When there raises to normal T the story is:

-The full set of f-features on T deletes the uninterpretable feature G of there.

-G on there is deleted by the f-features on T, so it stops raising.

-[T Agrees with its associate...]

Note that LD agreement is not specifically a property of expletive constructions, but of constructions where the specifier of TP does not have a full set of f-features. Thus dative subject constructions also show LD agreement. Conclusions so far: (i) Long-distance agreement is a T-associate (probe-goal) relation. (ii) EPP can be satisfied by: (a) Merge of expletive [T-associate agr.] (b) Merge of associate [your basic boring sentence] (c) Merge of a closer to T than the associate [dative subjects etc.]

Freezing and defective intervention effects If structural Case has been already been checked (deleted), the phrase P(G) is «frozen in place», unable to move further to satisfy EPP in a higher position. Uninterpretable features render a goal active, able to implement an operation: to select a phrase for Merge or to delete a probe. Merge and Agree require a goal that is both local and active. This means that an element that is inactive but still local may cause intervention effects. • Some further refinements on Agree and Merge in Chomsky (2001) TYPOLOGY OF ININTERPRETABLE FEATURES Uninterpretable features serve to implement Agree and Move. They are of three kinds: (i) Features that select a target α

φ-features on v, T (optionally C) 10



Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

(ii) Features that determine whether α offers a position for movement and if so what kind of category can move to that position (iii) Features that select the category β that is moved

EPP-features Case

Presence of uninterpretable features renders them active, so that matching leads to agreement. (33)

a. b.

Probe and goal must be both active for Agree to apply α must have a complete set of φ-features to delete uninterpretable features of the paired matching element β

Locality conditions yield an intervention effect if probe α matches inactive β which is closer to α than matching Γ, barring Agree(α,Γ). UNINTERPRETABLE FEATURES ARE UNVALUED. AGREE VALUES THEM. Chomsky (2001) takes uninterpretable features to be unvalued. They receive their values under Agree. Match is not identity but nondistinctness: same feature, independently of value. When an element is defective it is unable to inactivate a matched element by deleting its unvalued features. Illustration -Raising and ECM (34)

a. b.

there are likely to be awarded several prizes several prizes are likely to be awarded

(35)

a. b.

we expect there to be awarded several prizes we expect several prizes to be awarded

Spec,Tdef of α is filled either by Merge or Move. Tdef matches SUBJ in some of its features (to implement raising) but not all (to preclude inactivation). This feature is taken to be [person]. Expletives must also have the feature [person] since they raise. And pure expletives of the there-type have no other formal features. In a framework that dispenses with categorial features [person] plays the role formerly assigned to [D] or [N] features. When φ-complete, T values and deletes structural Case for N. The φ-set of N (which is always φ-complete) both values and deletes the φ-features of T (with and without movement). With defective probe, agreement is not manifested and Case of the matched goal is not assigned a value: raising T exhibits no agreement, and participles lack person; neither determines the Case of the matched N, which depends on a higher non-defective probe, T (in raising) and v (in ECM). 11

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Minimalist Inquiries (Chomsky 1998/2000) 1. How it all works (1)

How it works: Part 1 [p. 101] (I) Select [F] from the universal feature set {F} (II) Select LEX, assembling features from [F] (III) Select LA (lexical array) from LEX (IV) Map LA to EXP, with no recourse to [F] for narrow syntax

(2)

How it works: Part 2 a. Merge: "takes two syntactic objects (a, b) and forms K(a, b) from them. b. Agree: "establishes a relation (agreement, Case-checking) between an LI a and a feature F in some restricted search space (its domain)." c. Move: combining Merge and Agree. [A-movement if motivated by a j-feature; A-bar if motivated by a P ["peripheral"]-feature]

Occurrences • Move creates two occurences of a single a, where an "occurence of a" is the full context of a. • "Chain" is a set of occurences. If occurences are "full contexts" we don't need to say that a chain is a sequence, since there will be a containment relation between the contexts that allows us to reconstruct whatever we might needed the ordering property of a sequence for.] Prioritizing • Move is more complex than its subcomponents. • Move is more complex than even its subcomponents together -- since it involves the extra step of determining pied piping. • Consequently: (3) •

Merge or Agree "preempt" Move. "This yields most of the empirical basis for Procrastinate", p. 102

2. The problems (4)

• • • •

Core Functional Categories (CFCs) a. C b. T c. v All may bear uninterpretable j-features Only C may be unselected (i..e. be the root). T has a full set of j-features if selected by C, otherwise it is defective (ECM/Raising). v may take an external argument (EA) [Key: this is in addition to any other SPEC it gets.]

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-2­

(5)

Specifiers: each CFC gets one "beyond its s-selection1" [relevant to v] thanks to "EPP features" a. for C, a raised wh-phrase b. for T, the surface subject c. for v, the shifted object in Object Shift

(6)

[pp. 102-103] Some properties of these CFCs: a = [XP [(EA) H YP]] (i) How they get their specifiers: If H is v/C, XP [the outermost specifier] is not introduced by pure Merge [possible issues with C: whether? how come? Polish czy?] [T may have an expletive inserted as XP, so T is not mentioned] (ii)

Their social relations with the next highest T:

In the configuration [ß Tß...a], ß minimal,2

(a) if H (head of a) is C [or a lower T], Tß is independent of a [i.e. CP is a "closed system" -- no inbound or outbound agreement; anticipates the notion "phase"] (b) if H is v, Tß agrees with EA, which may raise to SPEC-Tß though XP [i.e. an accusative-marked object] cannot [Assumption: Object Shift position is higher than EA position because of (1) bottom-to-top tree building, and (2) Merge before Move. Observation: only the EA can raise and only EA triggers agreement with T. (c) if H is Tdefective, XP raises to SPEC-Tß if there is no closer candidate g for raising [This is raising to subject. I guess he forgot about ECM...]

(7)

Theta-theoretic principle Pure merge in a theta-position is required of and restricted to arguments. [Derives (6i) since v's XP position is not a theta-position and C has no theta-position. Also guarantees that no arguments are merged directly in Spec,TP.]

3. Phase (8)

Complexity considerations (i) Simple operations preempt more complex ones (ii) Search space is limited (locality) (iii) Access to the feature set [F] is restricted by (1). (iv) Computation is locally determined (no look-ahead)

1 "Semantic selection", here = q-role.

2 Easy to get confused here, the "a" mentioned here is intended to be the same a in (6).

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-3­

• •

Why is "raising" ever possible, given (i) and the availability of expletives to satisfy the EPP property of T? Answer: perhaps expletives are not always available. Perhaps only a subset of LA is available to derivation, so that if expletive is not in the subarray, it is not available. Thus, EPP motivates Move. [This Chomsky 1995's numeration.]

The chunk of derivation that has access to a given subarray is called a phase. Phases = vP and CP (categories that are "propositional") •

Solves a problem for numeration without phase (Marantz, Thursday class 1994):

(9)

There was assumed to be a reason why a man is in the garden.

Where availability of there upstairs should pre-empt movement of a man to the subject of be downstairs. If there is only one phase, i.e. the root phase, as in Chomsky 1995. (10)

Strong cyclicity condition The head of a phase is "inert" after the phase is completed, triggering no further operations.

[Phase ≠ convergent domain: because of successive-cyclic wh-movement -- assuming the wh-phrase has an uninterpretable feature like Case on nouns, only deleted in its final (specifier of interrogative C) position.] -­ and, of course, assuming that the CPs through which wh-movement passes are phases. (Alec's problem arises in these cases as well: At which bus top was there a reason to suppose that a linguist got off? So we know that phases don't work differently when wh-movement happens to happen.] Phases also provide a rationale for successive-cyclic movement if they are "impenetrable" except for their periphery. (11)

Phase impenetrability In phase a with head H, the domain of H is not accessible to operations outside a, but only H and its edge.

[This could not be stated if phase=convergent domain, since phrases move from the edge of a phase — on the assumption that if a category moves, there must be s omething non-convergent about it.] (12)

"Crash" in a world with phase impenetrability The derivation crashes if at the end of a phase a with head H, the domain of H contains an uninterpretable feature. [buried in the prose, bottom of p. 108]

This allows successive-cyclic movement, where movement is driven by the checking of a feature on some later phase. [Question: What is motivating movement to the phase edge in the case of successive-cyclic wh­ movement?? There is also discussion of QR? Is it possible that movement to the periphery is "free" in some sense?] Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-4­ (13)

•

EPP-features a. The head of phase PH (i.e. C and v) may be assigned an EPP- and P-feature. b. T bears an EPP-feature perhaps universally. [i.e. optionality of EPP is a property of the phase]

Assignment of optional EPP/P-feature is the last operation of a phase. [Navigation assistance: we are now on page 109]

4. Probes, Goals: No Agree unless Active Probes and Goals (14)

T be elected an unpopular candidate

• • •

T has uf and EPP features. Probe: f-features of T Goal: an unpopular candidate, which has matching features.

•

P(G): "pied piping" of a phrase determined by the goal of T's probe

"...taking structural Case to be a reflex of an uninterpretable f-set, it too erases under matching with the probe." Movement = • selection of P(G) • move of P(G) • feature-deletion under match (Agree) How probe-goal works: (I) matching is feature identity (II) D(P) ("domain of P") is the sister of P (III) locality reduces to "closest c-command" Closest: (15) Equidistance "Terms of the same minimal domain are 'equidistant' to probes." [not used until much later, to get the EA out of vP over an object-shifted object] (16)

Minimal Domain The minimal domain of a head H is the set of terms immediately contained in projections of H.

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-5­

Undifferentiated features (17) • •

Activity condition A goal must bear some uninterpretable feature [otherwise it is frozen in place].

This is why structural case exists! The "character" of the Case (nominative, accusative)merely registers the identity of the probe, so that "structural Case itself" is a single, undifferentiated feature. This is why differently-cased DPs can interfere with each other.

Agreement on T •

If Case is an undifferentiated feature on the goal in examples like Probes and Goals

• •

(14), then by parity of reasoning the f-features of the probe are not specified for values. Actual "agreement" is a result of the rule "Agree". i.e. uninterpretable -> value unspecified

• •

This yields "defective intervention constraints", where the closest bearer of the features sought by a probe is nonetheless inactive. The key point: Being active is not a requirement for Goalhood, but is a requirement for Agreement.

[We are now on page 123.]

5. Fullness of features •

If one f-feature on probe deletes, all delete. Evidence: no agreement in distinct features with distinct DPs.

•

Likewise, unless all f-features on goal delete, none of them delete. Evidence: participles that lack person features may attract a DP, but do not cause the f--features of the goal to delete. That is why you get participle agreement with passive and unaccusatives, alongside T-agreement with the same DP.

•

Similarly, Tdefective (to) can attract a DP if it has, say, just [person], and allow the DP to move on in a raising construction.

•

More generally: for a and b a probe and a goal, neither can delete f--features of the other unless it is f-complete.

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-6­

• • • •

Expletive there must have properties quite similar to Tdefective . Since it moves around like a normal DP, it has some attractable feature, e.g. [person] -- call it G. [But since it is not f-complete it does not delete features on T.] But it does not delete the probe features, as witnessed by LD agreement.

(18)

there were declared guilty three men

When there raises to normal T the story is: • The full set of f-features on T deletes the uninterpretable feature G of there. • G on there is deleted by the f-features on T, so it stops raising. • [T Agrees with its associate...] Note that LD agreement is not specifically a property of expletive constructions, but of constructions where the specifier of TP does not have a full set of f-features. Thus dative subjectconstructions also show LD agreement. (19)

Conclusions so far: (i) Long-distance agreement is a T-associate (probe-goal) relation. (ii) EPP can be satisfied by: (a) Merge of expletive [T-associate agr.] (b) Merge of associate [your basic boring sentence] (c) Merge of a closer to T than the associate

[dative subjects etc.]

6. Inertness again (20)

Time out for ontology (A) lexical items LI (B) modified lexical items MLI (C) sets K constructed from given elements a,ß.

"An MLI is an LI with uninterpretable features deleted." A note on Case • Recall that structural Case is there to make DPs "active". • This means that Case-checking requirements do not motivate movement, beyond allowing it to happen. • The action is in the f-features of T. • [Case only ever deletes because it's part of the f package.] Wh-movement is much the same • wh-phrases have uninterpretable wh amd interpretable Q, which matches uninterpretable probe uQ on C. • For successive-cyclic movement, C (and v) may have a non-specific P-feature which attracts wh­ phrases but does not delete their wh-feature. Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-7­

•

The wh-island condition arises because wh in an interrogative Q has its wh-feature deleted and thus is inert, while still bearing Q -- thus blocking access to lower wh-phrases.

(21) All the phrases marked with superscript "I" are inert: (i) *[John to seem [tI is intelligent]] (would be surprising) (ii) *(we hoped) [PRO to be decided [tI to be killed at dawn]] (iii) *[DO this book] seem [tDO to read [tDOI [never [[SU any students] tread]]]] (iv) *there seem [a [SU several people]I are [PRED friends of yours]]

7. It works (22)

Some features of these CFCs: [...] (ii)

a = [XP [(EA) H YP]]

Their social relations with the next highest T:

In the configuration [ß Tß...a], ß minimal,

(a) if H is C [or a lower T], Tß is independent of a [i.e. CP is a "closed system" -- no inbound or outbound agreement] (b) if H is v, Tß agrees with EA, which may raise to SPEC-Tß though XP cannot [Assumption: Object Shift position is higher than EA position because of (1) bottomto-top tree building, and (2) Merge before Move. Observation: only the EA can raise and only EA triggers agreement with T.] (c) if H is Tdefective, XP raises to SPEC-Tß if there is no closer candidate G for raising [Raising to subject; I guess this forgets ECM...]

Case a: a = [XP [C TP]] • If T is non-defective, and the derivation didn't crash at a, then the f-set of T has been deleted. • No element within TP can still have a structural case feature undeleted, because the element in agreement with T creates a "defective intervention effect". • So a higher T can't interact with the contents of a=CP.

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-8­ Case b: a = [XP [EA [v YP]] • XP is inactive, since its Case-feature has been deleted by v's f-set. • But EA is equidistant with XP from the higher T, so it can be a goal of T's probe, Case c: [omitted for reasons of space]

8. Architectural questions • • • •

Deleted features enter PF, so spell-out is cyclic in some sense. Suggestion: by phase. So there is a single cycle, all operations are cyclic. Overt/covert operations are interspersed.

Finale: cyclicity, labels, why specifiers are higher than complements.

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

24.952 (2005)

Derivation by Phase1

currently: C selects Tcompl; V selects Tdef

1. The Rules So Far: Agree For a probe and a goal to Agree: • probe and goal must both be active (=have uninterpretable features) • the goal must be the closest instance of the Agreeing feature to the probe Having two constraints leaves open the possibility of defective intervention effects: a close, inactive Goal blocking a lower, active Goal from agreeing with a Probe.

• A feature is uninterpretable iff it is also unvalued. Advantage: The syntax can detect an uninterpretable feature by simple inspection of its value.

•

Alternative approach to (1) (with two independent parts?)

alternatively: C is -complete: T is -complete only when necessary. (p.8) “necessary”=in order to delete uninterpretable  on a selector. [So C has some uF satisfiable by the TP that it embeds.] Ccompl: Tcompl :: vcompl: Vcompl ---> Burzio’s Generalization [v* with EA has some uF satisfiable only by the VP that it embeds.] Selection (partly) reduces to conditions on deleting uninterpretable features and if C:T :: v:V, then T is maybe (like V) sort of substantive rather than functional? CP If EPP=-completeness -->no successive-cyclic movement to Spec of defective T, just Agree.

Agree deletes features  on
iff  is
-complete: Tdef is -incomplete, for example. Strong Phases: v*P (transitive vP)

Move • Move = Agree + Determine Pied-Piping + Merge: Some features (like EPP) aren’t satisfied by Agree alone, and force Move (1) [Susan Tcompl seems T Move (1):

Move (2):

(2)

[ __ Tdef to be sleepy __ ]] (1)

Tdef has -features, Susan has Case-features, both Active: they Agree. Susan's -features are complete, so Tdef’s -features delete. def’s features are incomplete, so Susan's Case feature remains EPP forces actual movement to Spec,TP mutual annihilation: Tcompl has complete -features, as does Susan, so both lose their uninterpretable features and become inactive.

maybe DP maybe PP? ="propositional" categories

• Phases are the functionally headed XP s (if T is a substantive, i.e. lexical (N-like), head. • Non-phases can’t be extracted stranding their functional head. • Phases furnish reconstruction sites for scope and binding.

2. Phases and Cyclic Spell-Out PIC: for HP a strong phase dominated by a strong phase ZP; the domain of H2 is not accessible to operations at ZP, but only H and its edge

[T has no Case feature, but Susan's can still be deleted, via Match of the -features. No, this doesn't follow from anything...]

Maximize matching effects: cf. Pesetsky’s (1989) Earliness, Chomsky (1995)’s Strength, Richards' (1997) Featural Cyclicity

1

2

This handout is Norvin’s, edited here and there, except for the last section.

domain of H = what H c-commands.

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-2-

(2)

[C

[

T be likely [there to arrive a man]]]

• T, there Agree in [person] [u person] deletes on there, but T remains untouched (because there’s
-set is defective) • EPP forces raising of there to Spec TP • T, a man Agree in -features

a man’s Case feature deletes, T’s
-features delete

But wait a minute! Didn't the -features of caught disappear under Match with fish. Why is caught an acceptable Goal for the u-features of T? Answer: (6) deletion/erasure distinction Features get deleted (marked for deletion) but don’t actually get erased until Spell-Out.

Why couldn’t a man satisfy the EPP-feature of T?

A general alternative: Frampton, Gutman, Legate, Yang -- an HPSG-like mechanism of unification ensures that whatever happens to the feature set of caught will affect the feature set of fish, and conversely.

(3)

4. Th/Ex Education4

*[ C

[

T be likely [there to arrive a man]]]

Answer: Violates “Maximize Matching”: there may Agree with T first, and may check EPP. By MM, since it may, it must. (EPP is a feature, sort of like any other, soft of, sort of...)

(7)

Why doesn’t there block Agree of T with a man? Because it moves to Spec TP, and: (4)

• Only the head of an A-chain is visible for the MLC

Puzzling word orders a. *There came several angry men into the room b. There came into the room several angry men c. *There was placed a large book on the table d. There was placed on the table a large book e. There was a large book placed on the table f. *There entered a strange man the room g. There entered the room a strange man

True in English, but the facts are the opposite in Icelandic, for instance...

3. And of course you've been wondering about...3 [related facts discussed by (among others) Anagnostopoulou and Alexiadou] (5)

[ C [ T seem [there to have been [caught several fish]]]] (8)

• In Icelandic (but not Romance, Mainland Scandinavian), caught agrees with several fish — in Case as well as in number and gender. No person agreement! • The u
-features of caught are matched with [and are expected to delete under] Agree with fish. Because caught is -incomplete (no [Person]), it does not delete Case on fish (much less value it), despite Agree. • Problem: What about the Case feature of caught? It can't delete under Match with Case on fish because fish's Case feature isn't valued yet! Yet morphologically it will show the same case as fish. • Idea! It only looks like caught and fish agree in Case. Actually, each is having Case valued independently by T. Step 1: T probes there, which is -incomplete, leaving T unsatisfied. Step 2: T probes caught, valuing its Case feature (nominative). But caught is  incomplete (lacks [Person]), so T is still unsatisfied. Step 3: T probes fish, valuing its Case feature (nominative). At last, T has found a -complete element, and lives happily ever after. 3

Norvin's joke.

a. Quand partira [vP ton ami ]? when will-leave your friend b.*Quand mangera [vP Marie sa pomme]?

when will-eat Marie her apple

--> In a transitive construction, something must escape vP (by magic, if necessary). "Thematicization/Extraction" (=Th/Ex) • has no effects on semantics (unlike Object Shift) --> "operation of the phonological component"

(but one that crucially has effects on the narrow syntax:)

4

My joke, not Norvin's.

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-3"At the relevant stage of the cycle, the syntactic object
so far constructed is transferred to the phonological component for application of Th/Ex. The narrow-syntactic computation then proceeds on course with
unchanged except that the trace of Th/Ex is phonologically empty even prior to the strong phase level, at which point the position would have become phonologically empty even if not subject to Th/Ex." [footnote: "Note that this amounts to highly limited access of narrow syntax to effects of the phonological component"]

(15)

*There seems a man to be in the room

5. Object Shift: "stopover" positions • Icelandic has Object Shift, English doesn't....

Th/Exed NPs can't wh-move, or be extracted from: (9)

a. *How many packages did there arrive in the mail? b. *What did there arrive in the mail several packages of __ ?

(16) *The Christmas Trolls ate the pudding not

• ...unless (on various assumptions: phase impenetrability, Cyclicity, etc...) the object moves further:

Not a property of there-associate relations generally: (10)

a. How many packages are there in the room? b. What are there [several packages of __] in the room?

(17)

What did the Christmas Trolls

not eat

?

But does hold for leftward Th/Ex as well as rightward:

[other instances of stopover positions: French participle agreement...]

(11)

• Why does English behave this way? Consider an OS configuration for English:

*What are there [books about __] being sold (in Boston these days)

(18)

It is accessible to Agree (obeys Case Filter).

• the book should block Agree(T, John) (i.e. there should be a defective intervention

(12)

a. Th/Ex is an operation of the phonological component b. Traces are inaccessible to Move (but accessible to some other operations)

Move=Agree + Pied-pipe + Merge. • PRO, pro can Merge • traces can Agree (Th/Exed NP obeys Case Filter)

--> trace cannot Pied-Pipe (PRO, pro are heads)

(13) (14)

a. a man [OP to talk to __ ] b. *a man [[to OP] to talk __ ]

[

T

[

the book John v [ read ___ ]

-->Th/Exed NP is inaccessible to Move

constraint). In fact, in Icelandic, a shifted NOM object can even agree with T.

• ...assume Equidistance is not the way to go: (19)

Equidistance Terms of the edge of HP are equidistant from probe P

change to: (20)

Edge condition The phonological edge of HP is accessible to probe P

[way countercyclic, dude]5

How ec's matter to narrow syntax a. Empty Categories disallow Pied-Piping b. inactive trace disallows Match (A-traces don't block MLC--see above)

• Move applies freely • MLC evaluated (representationally!) at strong phase level So why isn't Icelandic English?

So, output of Th/Ex can't participate in Move, but can Agree, etc. Rightward Th/Ex doesn't iterate (Right Roof Constraint) Neither does Leftward Th/Ex:

• T can probe further in Icelandic? • Icelandic has, in addition to OS, Disl (
Th/Ex in English)? 5

Norvin's joke once again. Not my style.

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-4-

Evidence for Disl: • OSed pronoun in MSc can't bind anaphors (Holmberg and Platzack (1995)) • OSed pronouns are above v*P, auxiliaries... • Icelandic: Jon-DAT find-PL the computers-NOM not ugly-PL

6. Holmberg's Generalization (sort of)

English/French: (22) does not hold Since (22) does not hold, both INT and INT' are available for an unshifted direct object (even when V moves out of VP), so by (21b), v* has no INT-related reason to be assigned an EPP feature. Therefore there is no OS, regardless of interpretation. OV languages ???

A paradox of Icelandic object shift 1. 2. 3.

Object shift is only available when V has moved out of VP. When object shift is available, its presence/absence is correlated with specificity/non-specificity. In particular, *non-shifted specific and *shifted nonspecific When object shift is unavailable, an unshifted object is compatible with specific or non-specific interpretation.

The motivating force of Chomsky's analysis of Icelandic (21)

a. The EPP position of v*P is assigned specific interpretation INT. b. But v* receives an EPP feature only if that's the only way to assign INT to some argument.

(21b) is meaningful because of (22): (22)

PARAMETER: At the phonological border of v*P, XP is assigned non-specific interpretation INT'.

[phonological border of HP = position not c-commanded by phonological material within HP] Icelandic: (22) holds ==> S uppose
is a direct object in a verb-initial VP, and the V does not move.
may freely be specific or non-specific, since it is not at the phonological border of v*P Since OS is not necessary in order to achieve INT, OS is impossible by (21b). ==>

S uppose
is a direct object in a verb-initial VP, and the V does move. If
does not undergo OS, it receives INT' and is non-specific (possibly conflicting with lexical properties, in the case of a definite DP or pronoun). If
does undergo OS, it will receive INT, and will be specific. OS is possible, since assignment of EPP to v* is the only way to achieve INT for
, given V-movement and (22).

Cite as: David Pesetsky, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Chomsky’s Phases (short introduction) 1. Initial evidence for phases (Marantz’s problem). (1) a. There seems tthere to be a man in the room. b. *There seems a man to be ta-man in the room.

If there is in the numeration, we will have competition between merge and move at the embedded IP, and move will win. Or, stated differently, Move is always delayed as much as possible (given the elements available in the numeration) Marantz’s problem: there can be inserted in a matrix CP in cases in which earlier insertion, in an embedded CP, would have delayed movement. (2) There seems tthere to be a man in the room in which a woman seems to be sitting. Cf: A man seems to be in the room in which there seems to be a woman sitting. Chomsky’s response: The system is allowed to access the lexicon again once a CP has been constructed. More specifically, a derivation can be thought of in the following way (3) Steps of a derivation 1. Access the numeration. 2. Apply merge and move as many times as you want. 3. (Optional): once CP is constructed, insert it into a new numeration along with other lexical items (or CPs). Chomsky’s terminology: A constituent XP, the construction of which is/can-be followed by lexical access is called a phase. Chomsky’s additional claim (see also Nissenbaum): phases define impenetrable domains to movement. (4) Phase Impenetrability Condition: If X is dominated by a complement of a phase YP, X cannot move out of YP. Additional Claim: vP/VP is also a phase. This claim is made in order to explain successive cyclic movement through vP/VP. But, is there evidence that the lexicon can be accessed again after vP is constructed. (For (1b) to be bad, there can’t be accessed after the VP headed by seems is constructed. See Legate’s LI paper on phases.) It’s possible that we have two phases CP, vP, but only CP construction can be followed by lexical access. Theory of Isalnds (to be worked out): Certain phases do not allow wh-movement to proceed through their Spec. These phases are the island. Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-2-

Cyclic Linearization (Fox and Pesetsky) 1. Introduction

• There is good reason to think that wh-movement involves intermediate steps, for example,

movement to the left edge of an embedded CP:

(5)

Wh-movement through left edge of CP I wonder [which book he thinks [CP ___ Mary read ___ ]] 2 1

Evidence: scope reconstruction, all-stranding (McCloskey 2000), agreement phenomena (Chung passim), and islands.

• Why does wh-movement proceed through the left edge of CP? • Logic of a common answer: Things would go wrong otherwise. Question: What would go wrong otherwise?

Common answer: There are syntax-internal structural conditions that require such movement, e.g. (6):

(6)

Phase Impenetrability Condition (Chomsky 1999/2000) In phase α with head H, the domain of H is not accessible to operations outside α, but only H and its edge.

Our alternative: An answer in terms of linearization, i.e. conditions on the syntax-phonology interface (7)

Properties of wh-movement in (5): a. Movement 1 "revises" word order with respect to elements in the lower CP; b. Movement 2 "revises" word order with respect to elements in the upper CP; but c. Movement 2 does not revise word order with respect to elements in the lower CP. i.e. the moved wh-phrase precedes all elements in the lower CP both before movement and after movement,

Our conjecture: Property (7c) is not a coincidence. If it did not hold, the sentence could not be linearized.

Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-3-

2. General proposal

• Derivations proceed "bottom-to-top". • Certain syntactic domains created in a derivation are Spell-out Domains (roughly: CP, DP and vP/VP; but see Sabbagh 2004). By this, we mean (for now) domains whose construction is immediately followed by linearization. These roughly correspond to Chomsky's notion of phase.

• Linearization adds new ordering statements to the set of statements established by the linearization of previous Spell-out Domains. (8)

Consequence: Linearization Preservation The linear ordering of syntactic units is affected by Merge and Move within a Spell-out Domain, but is fixed once and for all at the end of each Spell-out Domain.

• For example: if leftward movement of which book out of CP in (5) were to take place from a

position preceded by an overt element X within CP, the ordering "X precedes α" would have to be altered in the higher domain -- contrary to the proposal.

Predictions of other proposals: a. Movement only possible from the edge of a relevant domain. b. Successive-cyclic movement through the edges of relevant domains is required independent of linearization. Predictions of our proposal: a. Movement is possible from the non-edge of a relevant domain so long as the previously established linearization is not disrupted. b. When there is no need to linearize, successive-cyclic movement through the edges of relevant domains is not required. Summary of relevant evidence: a. Non-Edge Movement: Object Shift in Scandinavian (Holmberg 1999) is possible only when elements that preceded the object in VP still precede the object after it has shifted, as a consequence of other movement operations. [Cf. similar proposals by Müller 2001, Sells 2000, Williams 2002, among others.] Verb movement to C in Scandinavian is possible only when elements that preceded the verb in VP still precede the verb after V-to-C movement. Crucial evidence: the contrast between Quantifier movement (Rögnvaldsson 1989; Jónsson 1996, Svenonius 2000) and wh-movement in V2 environments. Main verb movement in English is possible only when the element that precedes it (the subject) moves to a position where it continues to precede the verb, hence no V-movement to C when the subject is in Spec,IP — but auxiliary verbs, externally merged outside vP, are subject to no such restriction. We can argue, contrary to tradition, that English main verbs move overtly to I. Subject scrambling from vP when object has scrambled to the edge of vP in Korean and Japanese (Ko 2004) is possible only if the object ultimately scrambles to a position higher than the subject, recreating the order of elements in vP. Cite as: Elena Anagnostopoulou and Danny Fox, course materials for 24.952 Advanced Syntax, Spring 2007. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

-4-

b. Non-successive-cyclic movement (NOT PART OF THIS TALK): When linearization is not necessary, non-edge movement is generally possible. Example: Ellipsis which shows the phenomenon of "Salvation by Deletion" (Ross 1967; Chomsky 1972; Lasnik 2001; Merchant 2002; Fox and Lasnik 2003). Ellipsis in this proposal has a capacity for evil as well as good. By eliminating ordering contradictions, it allows extraction from certain islands, but elimination of ordering statements can also leave remnant elements unordered, making pronunciation impossible. A locality condition on multiple sluicing is explained as a consequence of this "dark side" of ellipsis.

3. How it works

• Suppose a bottom-to-top derivation has created the syntactic domain D in (9), where D is also a Spell-out domain. Assume that the Spell-out operation establishes the ordering statements given in (9). (9)

[D X Y Z] Ordering:

X