CAS LX 522 Syntax I (6) S trace (marks where

Fall 2000 September 18, 2000 movement took

Paul Hagstrom Week 2: Movement place from)
NP NP Aux VP
Movement 3 1 can 3
Det N I V t[NP]
this problem solve
Last time, we talked about subcategorization.
(7) a. [PP to John], Mary gave the book t[PP].
(1) a. I can solve this problem. b. [VP fix the car], I wonder whether she will t[VP].
b. This problem, I can solve.
(8) a. [S [NP this problem], I believe [S′ that [S I can solve t[NP] ] ] ].
(2) solve: [— NP] b. [S I believe [S′ that [S [NP this problem] I can solve t[NP] ] ] ].

Trace convention
Two problems: • solve has no object.
Movement transformations leave a trace behind.
• This problem has no place in the structure
(can’t be generated)
(9) a. I want to read this novel.
b. I wanna read this novel.
We could complicate the rules, or we could derive (1b) from (1a).
(10) wanna contraction: want to
wanna.
(3) Topicalization
Move XP and attach it as the leftmost constituent of S. (11) a. This novel, I want to read.
b. This novel, I want to be considered for a prize.
(4) S
qgp (12) a. This novel, I wanna read.
NP Aux VP b. * This novel, I wanna be considered for a prize.
1 can 3
I V NP (13) a. [NP this novel], I want to read t[NP].
solve 3 b. * [NP this novel], I want t[NP] to be considered for a prize.
Det N
this problem
Idea: The rule in (10) can see the trace; hence want is not next to to, *wanna.
(5) S The trace isn’t pronounced, but it is nevertheless there.

NP NP Aux VP
3 1 can 3
Det N I V NP
this problem solve 3
" Det N
z--------------- this problem
Some history… Wh-movement
Traces weren’t always a part of syntactic theory.
This novel is the thing read: This novel I want to read—It is still interpreted as the object. (16) Q. Which problem did you solve?
• Meaning could be read off deep structure (from PS rules, prior to movement). A. I solved number 18.

One problem: Two sentences with the same deep structure can have different meanings. (17) solve [— NP].

(14) Some student read every chapter. (18) [which problem]i did you solve ti ?
‘There is a student x such that for each chapter x, x read y.’
‘For every chapter y, there is a student x such that x read y.’ Dodging various complications for a moment…
Preview: These meaning differences seem to come about by movement. (19) I wonder [which problem will solve ti ].
(15) a. Some student read every chapter.
b. [NP Every chapter], some student read t[NP]. (20) I wonder… S′
qp
Comp S
But if something “moves” why can’t we hear it in a different place? which qgp
problem NP Aux VP
The structure of the grammar: DS  phrase structure rules 1 will 3
1 N V t[NP]
1 movement rules Mary solve
surface structure (abstract)
SS
3 more movement rules Comp = complementizer. E.g., that.
“phonetic form”
PF LF  “logical form” (meaning)
Wait— didn’t topicalization attach to S? Why move wh-phrases to Comp instead of
The idea: The movement of every chapter in (14) on the meaning of (15b) happens attaching them to S too?
after surface structure (no access to pronunciation).
We have real evidence that topicalization attaches to S (in particular, after Comp):
You probably shouldn’t be convinced yet, but we’ll add evidence for this view later…
(21) I know [S′ that [S [this problem]i I can solve ti. ] ]
On traces: If interpretation (meaning) is based on the representation at LF, the trace
in the object position also helps explain how this problem is understood
We don’t have such evidence here; you can’t use that with questions:
as the object of solve even after this problem has been moved away.
(22) a. * I wonder [which problem]i that Mary solved ti.
Note: Syntacticians often talk about things happening “before” and “after”—but at this b. * I wonder that [which problem]i Mary solved ti.
point, we aren’t really proposing that people do this in real-time (construct a DS, apply
movement rules to construct a SS, pronounce it, apply more movement rules to construct Assertion: There is good evidence for moving wh-phrases higher, but we need to talk
a LF, interpret the resulting representation). W e are still in the business of characterizing about some more things first.
knowledge, and this system (with its before–after structure) predicts native-speaker
intuitions to a certain degree of success. It remains an open question how much (if any) So, for now: Assume that and the target of wh-phrase movement is the same, Comp:
of the procedure we use to arrive at our prediction reflects how people actually produce a
sentence.
Wh-movement
Move wh-XP to Comp. So, sentences are either [+Q] or [–Q], and wh-movement only works for [+Q] sentences.

XP because… Wh-movement (statement two)

“Pied-piping” Move wh-XP to Comp when Comp is [+Q].
(23) a. I wonder [S′ [AP how difficult ]i [S the problem was ti ] ] .
b. I wonder [S′ [AdvP when ]i [S Mary solved the problem ti ] ]. Echo questions: actually [–Q]:
c. I wonder [S′ [PP to whom ]i [S Mary gave the book ti ] ]. (30) a. Mary solved which problem?
d. I wonder [S′ [NP who(m) ]i [S Mary gave the book [PP to ti ]] ]. b. You believe Mary solved which problem?
c. * You wonder Mary solved which problem?

“preposition stranding” Questioning out of questions: The wh-island condition

Notice that wh-phrases are regular phrases (NP, AP, PP, AdvP) (31) a. Howi do you think [S′ (that) [S Mary solved the problem ti ] ] ?
but with the property that they are question words. “[+wh]”, or a wh-XP. b. * Howi do you wonder [S′ whether [S Mary solved the problem ti ] ] ?
(Ans.: I wonder whether Mary solved the problem with her Palm Pilot.)
With respect to (23) and wh-XPs: who(m) is an NP, like which problem: b′. ? Howi do you wonder ti [S′ whether [S Mary solved the problem ] ] ?
(Ans.: I wonder desperately whether Mary solved the problem.)
(24) a. NP b. NP
3 1 (32) a. [Which way]i do you think [S′ (that) [S John went ti (to fix his car)] ] ?
Det N N b. * [Which way]i do you wonder [S′ whyj [S John went ti tj ] ] ?
which problem who (Ans: I wonder why John went north.)

(33) a. [Which way]i do you think John went ti ?

Conditions on wh-movement b. * [Which way]i do you wonder [S′ if [S John went ti ] ] ?
(Ans: I wonder if John went north.)
Not all S-bars are created equal:
A wh-island is a [+Q] S′; an island to wh-movement (things inside are “stranded”).
(25) a. I wonder which problem Mary solved.
b. I wonder if Mary solved this problem. Wh-movement (statement three)
c. * I believe which problem Mary solved. Move wh-XP to Comp when:
d. * I believe if Mary solved this problem. (i) Comp is [+Q]
(ii) the wh-XP is not inside a [+Q] S′ (except the one containing the Comp).
(26) a. * I wonder that Mary solved this problem.
b. I believe that Mary solved this problem. Complex Noun Phrase Condition (CNPC)…

Complex noun phrase: the claim that John went north, the hat I gave to John.
(27) wonder [ — S′:[+Q] ]
believe [ — S′:[–Q] ] (34) a. You believed [NP the claim that John went north].
b. * [Which way]i did you believe [NP the claim that John went ti ] ?
(28) a. I know that Mary solved the problem.
b. I know if Mary solved the problem. (35) a. Bill stole [NP the hat I gave to John ].
c. I know which problem Mary solved. b. * [To who(m)]i did Bill steal [NP the hat I gave ti ] ?

(29) know [ — S′ ]
Wh-movement (statement four) Wh-movement (statement six)
Move wh-XP to Comp when: Move wh-XP to Comp when:
(i) Comp is [+Q] (i) Comp is [+Q] if wh-word does not move further.
(ii) the wh-XP is not inside a [+Q] S′ (except the one containing the Comp). (ii) the wh-XP is not inside a [+Q] S′ (except the one containing the Comp).
(iii) the wh-XP is not inside a complex NP. (iii) the wh-XP is not inside a complex NP.
(iv) there is no nearer Comp.
Successive-cyclicity
Relative clauses
(36) a. [which problem]i did John say [S′ Mary heard [S′ Robin solved ti ] ]?
b. Which problem did John say Mary believed Robin heard Susan solved ti ? (40) Bill heard [NP the speechi [S′ whichi [S Mary made ti ]]].
Wh-movement looks like it can go unboundedly far. " 1
z--------m
But there are good reasons to believe (we’ll hear about them later) that this is an illusion.
Relative clauses have a structure very similar to wh-questions, attached inside an NP.
(37) a. [S′ [S John knows [S′ [which problem]i [S Mary solved ti ] ] ] ].
b. [S′ [which problem]i does [S John know [S′ [S Mary solved ti ] ] ] ] ? Restrictive relatives restrict the reference of the head noun.
(sort of like an adjective: The moving speech, The speech which Mary made)
We see that there’s a “place to stop” below know: intermediate trace
The head noun and the wh-phrase are co-indexed (meaning they share the same referent).
(38) [S′ [which problem]i does [S John know [S′ ti′ [S Mary solved ti ] ] ] ] ?
Appositive relatives don’t restrict the reference, but provide additional information
" 1" 1
z--------------mz--------m (41) a. Mary, who you met yesterday, just bought a house.
b. Maryi, [S′ who [S you met ti yesterday ] ],
The Transformational Cycle — each S′ is self-contained.
You apply all the rules inside S′ before moving on to the higher S′. Free relatives involve -ever and don’t modify a head noun
Another way to think about this: wh-movement can only go to the closest Comp. (42) a. I will buy [whateveri you sell ti ].
(A locality condition…) b. Whoever just arrived unplugged my lamp.
Wh-movement (statement five)
Move wh-XP to Comp when: (43) a. Bill heard the speech [which Mary made].
(i) Comp is [+Q] b. Bill heard the speech [that Mary made].
(ii) the wh-XP is not inside a [+Q] S′ (except the one containing the Comp). c. Bill heard the speech [Mary made].
(iii) the wh-XP is not inside a complex NP.
(iv) there is no nearer Comp. (44) made: [ — NP ] (in the sense of made a speech).
This doesn’t quite work. The first movement should be ruled out by condition (i). The structure of (43a–c) is the same except for the relative clause.
But made needs an NP argument—where is it in (43b–c)?
[–Q] It must be there. So it must be silent, but basically means the same thing as which.
A null wh-phrase; we’ll call it Op (for Operator) for reasons that will become clear later.
(39) [S′ [which problem]i does [S John believe [S′ ti′ [S Mary solved ti ] ] ] ] ?
" 1" 1 (45) Bill heard [NP the speechi [S′ Opi [S Mary made ti ]]]. =(43c)
z--------------mz--------m " 1
z--------m
The analysis of (43b) is that Op moves to attach to Comp, which already has that. NP-movement
(46) a. …the speech that Mary made. Passive.
b. …[NP the speechi [S′ Opi that [S Mary made ti ]]].
c. (52) a. Mary solved the problem.
Comp b. The problem was solved (by Mary).
3
Op that The problem, still the thing solved (the logical object), is in subject position [NP,S].
Some support for this analysis: If Op (or indeed which) is really wh-movement in (53) solve [— NP]
relative clauses, it should be sensitive to the
conditions we discussed before. We could introduce a new lexical item, solvepassive, but we’d have to do it for every verb.
Instead, let’s derive (52b) from (52a).
(47) a. * I know the way Opi (that) John wonders [wh-island why Bill went ti].
b. * I know the way Opi that John made [CNP the claim that Bill went ti]. (54) S
qgp
(48) a. * I know the way whichi John wonders [wh-island why Bill went ti]. q 1 p
b. * I know the way whichi John made [CNP the claim that Bill went ti]. NPi Aux VP
# tense qgp
So: • If relative clauses can have a null wh-phrase, why can’t wh-questions? the problem V NP PP
• Why can you have that with Op but not with which? # 1 #
*I heard the speechi whichi that Mary made ti . was solved ti by Mary
Recoverability Condition
The content of a null category must be recoverable (from a co-indexed overt category in
the sentence). NP-movement
(49) a. When did Mary buy the book? Move NP to an empty subject position
b. Where did Mary buy the book?
c. How did Mary buy the book? (Where we have to allow our PS rules to generate an empty subject position.)
d. * Opi did Mary buy the book?
Raising constructions.
(50) a. * Opi did Mary buy ti?
b. * Opi did Mary give a book ti? (55) a. It seems (that) Mary has solved the problem.
b. Mary seems to have solved the problem.
(51) Bill heard the speechi [S′ Opi that [S Mary made ti]].

Doubly Filled Comp Filter These are synonymous; Mary is the logical subject of solve,
*[Comp wh-XP that], if wh-XP is overt (non-null). but it’s the structural subject of seems, the matrix (top level) clause.

Filters. Exclude sentences which the PS rules and transformations otherwise generate. Like with passive, we will derive (55b) from a sentence with Mary as subject.
(instead of encoding the restriction in the transformations).
(56) [S [NP e ] seems [S [NP Mary ] to have solved the problem] ].
" 1
z--------m
Only certain verbs do this: seems, be likely, appear, …—raising verbs. Specified Subject Condition
These verbs also can take it as a subject, an expletive.
(61) a. Maryi appears [ti′ to be likely [ti to win]].
(57) a. It appears (that) Mary has solved the problem. b. * Maryi appears [it is likely [ti to win]].
b. Mary appears to have solved the problem.
c. It is likely that Mary will solve the problem. (62) a. John is believed [ti to have beaten Bill].
d. Mary is likely to solve the problem. b. * Bill is believed [John to have beaten ti].
(58) a. * It is confident (that) Mary will win. (61b) is bad by TSC, but (62) can’t be explained that way.
b. * Mary is confident to win.
c. Mary is confident (that) she will win.
(63) a. [S [NP e ] is believed [S [NP John ] to have beaten Bill] ].
Tensed S condition " 1
z----------m
The embedded sentences from which subject-to-subject raising applies are not, and b. * [S [NP e ] is believed [S John to have beaten [NP Bill ] ] ].
cannot be, tensed:
" 1
z--------------------m
(59) a. Mary seems [to have solved the problem].
b. * Mary seems [has solved the problem].
c. * Mary seems [will solve the problem]. There are many options here (e.g., you can only raise a subject, or you can’t skip a
subject position), but the explanation people adopted was the Specified Subject
Tensed clauses have [+Tense] in their Aux element, Condition.
Untensed clauses have [–Tense] in their Aux element,
pronounced as to (infinitive marker) NP-movement (third statement)
Move NP to an empty subject position, as long as
(60) S (i) NP is not contained in a Tensed S (other than the target S).
qgp (ii) NP is not separated from the targeted position by a specified subject.
NPi Aux p
# [+Tense] VP Extraposition and Heavy NP shift
Mary qp
V S
1 qgp (64) a. [Details of a secret plan to finance the rebels] have emerged.
seems NP Aux VP b. Details have emerged of a secret plan to finance the rebels.
ti [–Tense] rp
to V NP (65) [NP Details [PP ti ] ] [VP have emerged [PP of a secret plan to finance the rebels]].
# 3 1 "
have solved the problem z--------------m

NP-movement (second statement) Extraposition (a rightward movement rule—unlike the rest we’ve seen)
Move NP to an empty subject position, Move XP and attach it as the rightmost constituent of VP.
unless NP is contained in a Tensed S (other than the target S).
XP because it isn’t just PP…
(66) a. A man ti has come forward [S′ who claims to be the culprit ]i.
The extraposition rule wildly overgenerates—there are conditions on extraposition too, Consider:
but we’ll worry about that later.
(72) S
Heavy NP shift (a subcase of Extraposition) qgp
Move a ‘heavy’ NP and attach it as the rightmost constituent of VP. q 1 p
NP Aux VP
Mary 1 qp
(67) a. Mary read [NP all the books she had borrowed]. ti V NP
b. ? Mary returned [NP all the books she had borrowed] to the library. 3 #
c. Mary returned ti to the library [NP all the books she had borrowed]i. solve Tensei the problem
-ed
(68) a. Mary read Aspects.
b. Mary returned Aspects to the library.
c. * Mary returned ti to the library [NP Aspects ].

Movement of terminal categories (73) S

qgp
q 1 p
Affix hopping. NP Aux VP
Mary 3 3
(69) Aux → Tense (Modal) (Neg) Vi Tense ti NP
solve -ed #
(70) Mary solved the problem. the problem

(71) S
qgp
NP Aux VP We can decide between them on the basis of evidence from adverbs.
1 1 3 Adverbs are a constituent of VP, on the left or right
N 1 V NP (74) VP → (Adv) V … (Adv)
Mary Tensesolve #
[+Past] the problem (75) a. John [VP cleverly avoided Bill ].
b. John [VP rarely visited Bill ].
We have to get from (71) to (70). [+Past] ends up on V (solved). Well, movement. c. * John avoided [VP cleverly Bill ].
d. * John visited [VP rarely Bill ].
Two options: Move solve to attach to Aux.
Move Tense to attach to V. So the verb is inside the VP; (72) is the victor.

It matters which one we choose… Affix Hopping

Move Tense (from Aux) to V provided Aux does not dominate a modal.

(76) John might cleverly avoid Bill.

Do-support
V-raising
(77) John did not avoid Bill. Raise V to Aux, provided:
i) V has the feature specification [+AUX]
For some reason, avoid does not get the past tense, which is instead realized on did. ii) Aux does not dominate a modal.
Affix Hopping was blocked:
Affix Hopping
(78) John -ed not avoid Bill. Move Tense (from Aux to V), provided:
i) V has the feature specification [–AUX]
not intervenes between past tense -ed, and so Affix Hopping cannot merge them. ii) Aux does not dominate a modal or neg.

Affix Hopping (second statement) (82) a. * Mary be worked at home.

Move Tense (from Aux) to V provided Aux does not dominate a Modal or Neg. b. * Mary [Aux [Tense] ] be work at home.

Do-support Ouhalla adds to Affix Hopping

Insert do to support stranded Tense. iii) VP does not dominate a V with the feature specification [+AUX].

Both Affix Hopping and Do-support provide “morphological support” for Tense. This is getting complicated. V-raising and Affix Hopping have almost complementary
Tense is an affix, it can’t stand alone. conditions. We can simplify this a little bit:

Neither is needed if there is a modal in Aux: Goal: Support Tense.

(79) a. John should not avoid Bill. Preference: Prefer V-raising to Affix Hopping, Affix Hopping to Do-support.
b. John cannot eat bacon.
• If Tense is already supported (if there is a modal in Aux), goal is met.
Auxiliary verbs have and be: • If Tense is still unsupported and V-raising can be applied (there is a [+AUX] V),
apply it; goal is met.
(80) a. John is not happy. • If Tense is still unsupported and Affix Hopping can be applied (Aux doesn’t dominate
b. John has not visited Bill. Neg), apply it; goal is met.
c. * John did not be happy. • If Tense is still unsupported, apply Do-support (as a “last resort”).
d. * John did not have visited Bill.
V-Raising
Have and be are unique among English verbs, using strategy (73) instead;
the verb raises to Aux to merge with Tense. Raise V to Aux provided V has the feature specification [+AUX].

To distinguish have and be from other verbs, we will say they are [+AUX], Affix Hopping
whereas other verbs are [–AUX]. Move Tense (from Aux to V), provided Aux does not dominate Neg.

If Tense needs support, a [+AUX] verb can raise—but won’t if not needed. (83) a. * John did avoid Bill.
b. John did avoid Bill.
(81) a. John must be happy.
b. John might have visited Bill.
(84) a. Mary will solve the problem. (89) a. Was John at home?
b. Will Mary solve the problem?
b. DS: [S′ [+Q] [S John [Aux Tense ] [VP be at home ]]]
(85) a. You can drive my car.
b. Can you drive my car? c. VR: [S′ [+Q] [S John [Aux [V be ]i Tense ] [VP ti at home ]]]
" 1
To make a question, the Aux element is moved over the subject—to Comp. z------m

(86) S′ d. AR: [S′ [Aux [V be ]i Tense ]j [S John tj [VP ti at home ]]]

qp " 1
Comp S z----m
[+Q] qgp
1 NP Aux VP
Auxi 1 1 3 (90) a. Did John solve the problem?
can you ti V NP
" 1 drive # b. DS: [S′ [+Q] [S John [Aux Tense] [VP solve the problem ]]]
z----------m my car
c. AR: [S′ [Aux Tense ]i [S John ti [VP solve the problem ]]]
Aux-raising " 1
Move Aux to Comp, when Comp has the feature specification [+Q]. z----m

d. do: [S′ [Aux Tense [do] ]i [S John ti [VP solve the problem ]]]
This happens in wh-questions too… (we ignored this part earlier)
Note: Affix Hopping didn’t apply here, but there was no Neg.
(87) a. Which problem can Bill solve?

b. DS: [S′ [+Q] [S Bill [Aux can ] [VP solve which problem]]] Affix Hopping
Move Tense (from Aux) to V provided
c. SS: [S′ [NP which problem]i [Aux can]j [S Bill tj [VP solve ti ]]] i) Aux does not dominate a Modal or Neg
" " 1 1 ii) V has the feature specification [–AUX]
1 z----m 1 iii) VP does not dominate a V with the feature specification [+AUX]
z------------------m iv) Aux is adjacent to VP which dominates the targeted V

But, Aux-raising doesn’t apply in embedded questions: or

Affix Hopping
(88) a. I wonder which problem Bill can solve. Move Tense (from Aux to V), provided
b. I wonder if Bill can solve this problem. i) Aux does not dominate Neg.
ii) Aux is adjacent to VP which dominates the targeted V
Aux-raising (second statement)
Move Aux to Comp, provided We could even conflate these two if Neg counts as making Aux non-adjacent to VP.
i) Comp has the feature specification [+Q] More soon…
ii) Comp is situated in a root clause.

Interactions…