Transcript PowerPoint

CAS LX 522
Syntax I
Week 12. Articulating the tree,
and some Applied Syntax
Reminder: Motivating AgrOP



+Fin aux/verb:
V Adv, V neg
Moves to T.
–Fin aux:
(V) Adv (V), (V) neg (V)
(Opt.) Moves to T.
–Fin verb:
(V) Adv (V), neg V
Moves over adv not neg??

We need there to be a head here in the
tree for the verb to move to…

That means we need to insert a whole
phrase (heads always head
something)…
TP
DS
T
T
NegP
pas Neg
Neg VP
ne
DPk
V
AdvP V
souvent
V
PP
TP
A new FP




+Fin aux/verb:
V Adv, V neg
Moves to (F, then to) T.
–Fin aux:
(V) Adv (V), (V) neg (V)
(Opt.) Moves to (F, then to) T.
–Fin verb:
(V) Adv (V), neg V
(Opt.) Moves to F
DS
T
T
Now we have a place for nonfinite main
verbs to move, past adverbs but under
negation. They can move to F.
NegP
pas Neg
Neg
ne
FP
F
F
VP
DPk
V
AdvP V
souvent
V
PP
AgrOP




AgrOP, Object agreement
phrase.
Don’t forget, this is French
we’re talking about here. In
French, V moves to T.
As the verb moves up to T, it
has to stop off in AgrOP (the
Head Movement Constraint
requires it), forming
successively more complex
heads.
 V
 AgrO+V
 T+[AgrO+V]
And why does the object have
to move to SpecAgrOP? To
get accusative Case.
CP
DPi
C
SS
C
TP
DPk
T
T
AgrOP
t i
AgrO
AgrO
VP
tk
V
V
ti
Morphology on French verbs




Past, varying persons: je mange-ai-s
‘eat’
tu mange-ai-s
il mange-ai-t
Fut, varying persons: je mange-er-ai
‘eat’
tu mange-er-as
il mange-er-a
Tense morphology is inside and separate
from subject agreement morphology.
Kind of looks like after tense, another,
subject-agreeing morpheme is attached…
C
AgrSP?




AgrOP, Object agreement
phrase.
AgrSP, Subject
agreement phrase?
Pleasingly symmetrical!
Complex heads:
 V
 AgrO+V
 T+[AgrO+V]
 AgrS+[T+[AgrO+V]]
C
AgrSP
DPk AgrS
AgrS
TP
T
ti
T
AgrOP
DPk AgrO
AgrO
VP
tk
V
V
ti
C
Split-INFL


C
AgrSP
The assumption of this structure
AgrS
is sometimes referred to as the
“Split-INFL” hypothesis; the
AgrS TP
INFLectional nodes have been
“split” into subject agreement,
T
tense, and object agreement.
Reminder: AgrSP+TP+AgrOP in
the era before the “Split-INFL”
was often referred to as INFLP
or IP. We’ve called it “TP”
(perhaps confusingly, we were
really talking about IP before).
T
AgrOP
AgrO
AgrO
VP
V
DP
V
DP
The EPP
& NOM





C
C
AgrSP
AgrS
We said before the T needs a specifier
(at SS), that’s the essential content of
the EPP. Plus, we said before that this is AgrS TP
where NOM is assigned.
T
Now there is AgrSP as well.
AgrOP is responsible for ACC.
T AgrOP
In a symmetrical world, seems like
AgrSP should be responsible for NOM.
AgrO
So, now that (kind of mysterious) double
motivation for moving to SpecTP has
been clarified: The subject has to move
AgrO VP
to both SpecTP and SpecAgrSP, but
each movement happens for a different
DP
reason. T for EPP, AgrSP for NOM.
V
V
DP
Adopting the Split-INFL
hypothesis

Lots of good syntax has been done both adopting the
Split-INFL hypothesis (trees contain AgrSP, TP, AgrOP)
or not (trees contain only TP, a.k.a. IP).

For many things, it doesn’t matter which you choose—
analyses can be directly translated into a Split-INFL tree
or vice-versa.

Where it doesn’t matter, it doesn’t matter, but sometimes
it matters. (On the final, for example)
Adopting the Split-INFL
hypothesis

The general program is that every dissociable piece
of the structure should get its own place in the
lexicon, its own functional head…





Subject agreement is basically common across verbs, an
independent piece.
Tense too is an independent piece.
And object agreement
And… plural marking… and progressive -ing, aspectual en, …
In Syntax II, we’ll spend a lot of the semester looking
at places in the tree where functional projections
need to be added.
Split-INFL

CP
C
In recent literature, almost
everything you read will
C
AgrSP
make this assumption, that
cross-linguistically, the
AgrS
clause is minimally
constructed of these
AgrS TP
projections, generally in
this order:
T





CP
AgrSP
TP
AgrOP
VP
T
AgrOP
AgrO
AgrO
VP
A word about the PP



Actually, there is one place where we haven’t unified
Case-assignment, namely the oblique Case that is
assigned to the complement of P.
For now (this whole semester), we’ll just leave it at that.
P can assign oblique Case to its sister, although V
cannot assign accusative Case to its sister.


So DPs that are objects of prepositions don’t need to go
anywhere, they’re fine where they are.
Sometimes the distinction is made between structural
and inherent Case:


Structural Case (NOM, ACC) is assigned by movement to
someplace in the structure (SpecAgrOP, SpecAgrSP).
Inherent Case is assigned in place, e.g., by P.
Moving away from English

Recall that the model of language we’re working
with is one in which languages are for the most
part the same, but differ in the settings of certain
parameters, such as order between object and
verb. What are possible parameter settings?
English
UG
Japanese
Moving away from English

We’ve seen a couple, but the only way to
discover what they are is to look at how
other languages differ.

Recall, for example, the V-to-T parameter that
differentiated French from English.
English
UG
Japanese
Japanese



Taroo-ga ano hon-o
kat-ta.
Taro-NOM that book-ACC buy-PAST
‘Taro bought that book.’
Taroo-ga ano hon-o
kat-ta
no?
Taro-NOM that book-ACC buy-PAST Q
‘Did Taro buy that book?’
Hanako-ga [Taroo-ga ano hon-o
kat-ta
to] omotteiru.
H.-NOM
T.-NOM that book-ACC buy-PAST that thinks
‘Hanako thinks that Taro bought that book.’
Japanese

CP
Japanese appears to be quite
strictly head-final. The head of
an XP comes after the
complement.
C
DPi

We can draw the structure of a
Japanese tree like this, using
the same system, only with
head-final structures.

(I drew this series of slides for a
previous lecture preceding the
introduction of the split-INFL
hypothesis—what do we think about
AgrSP and AgrOP in Japanese?)
C
to
TP
T
Taroo-ga VP
ti
T
-ta
V
DP
ano hon-o
V
kat
Japanese

As in English, nominative
case (ga) is assigned to
the DP in the specifier of
TP, accusative case (o) is
assigned to the sister of V.


(Or, in terms of Split-INFL?)
Does the verb move to T?
Tough question. Notice
that it doesn’t have to to
get the word order right.
CP
C
C
to
TP
DPi
T
Taroo-ga VP
ti
T
-ta
V
DP
ano hon-o
V
kat
Japanese



Taroo-ga ano hon-o
kat-ta.
Taro-NOM that book-ACC buy-PAST
‘Taro bought that book.’
ano hon-o
Taroo-ga kat-ta.
that book-ACC Taro-NOM buy-PAST
‘Taro bought that book.’
How might this come about?

The q-criterion dictates that the object ano hon-o starts out
as the sister of V. Like in What did I buy?
Japanese

CP
ano hon-o
Taroo-ga kat-ta.
that book-ACC Taro-NOM buy-PAST
‘Taro bought that book.’
C
DPi



This must start out the same
way as Taroo-ga ano hon-o
katta.
Thus ano hon-o must move
to where we see it.
Scrambling.
So where does it go?
C
TP
[-Q]
T
Taroo-ga VP
ti
T
-ta
V
DP
ano hon-o
V
kat
Japanese
CP
C

ano hon-o
Taroo-ga kat-ta.
that book-ACC Taro-NOM buy-PAST
‘Taro bought that book.’
C
TP
DPj
[-Q]
TP
ano hon-o DPi

The standard analysis of this
is that scrambled arguments
move to adjoin to TP—like
quantifiers do. Same kind of
movement as QR.
T
Taroo-ga VP
ti
T
-ta
V
tj
V
kat
Japanese
CP
C



ano hon-o
Taroo-ga kat-ta.
that book-ACC Taro-NOM buy-PAST
‘Taro bought that book.’
So languages also differ in
whether or not they allow
scrambling.
A large majority of the
scrambling languages are
also SOV languages,
although why that would be
remains unclear.
C
TP
DPj
[-Q]
TP
ano hon-o DPi
T
Taroo-ga VP
ti
T
-ta
V
tj
V
kat
Korean

Korean is in many respects structurally very
similar to Japanese; strictly head-final, allows
scrambling, has Case marking.

Chelswu-ka ku chayk-ul ilk-ess-ta.
Chelswu-NOM that book-ACC read-PAST-DECL
‘Chelswu read that book.’
Korean negation



Chelswu-ka ku chayk-ul ilk-ess-ta.
Chelswu-NOM that book-ACC read-PAST-DECL
‘Chelswu read that book.’
Chelswu-ka ku chayk-ul an-ilk-ess-ta.
Chelswu-NOM that book-ACC NEG-read-PAST-DECL
‘Chelswu didn’t read that book.’
Chelswu-ka ku chayk-ul ilk-ci
anh-ess-ta.
Chelswu-NOM that book-ACC read-CI NEG.do-PAST-DECL
‘Chelswu didn’t read that book.’
CP
Korean negation


DS
C
C
-ta
TP
Chelswu-ka ku chayk-ul an-ilk-ess-ta.
Chelswu-NOM that book-ACC NEG-read-PAST-DECL
‘Chelswu didn’t read that book.’
T
NegP T
-ess
If this is the DS for the Korean “short
negation,” how do we get the right word
order?
Neg
Neg
an
VP
DP
C.-ka DP
V
V
ilk
ku chayk-ul
CP
Korean negation





Chelswu-ka ku chayk-ul ilk-ci
anh-ess-ta.
Chelswu-NOM that book-ACC read-CI NEG.do-PAST-DECL
‘Chelswu didn’t read that book.’
C
C
-ta
TP
Chelswu-ka ku chayk-ul an-ilk-ess-ta.
Chelswu-NOM that book-ACC NEG-read-PAST-DECL
‘Chelswu didn’t read that book.’
If this is the DS for the Korean “short
negation,” how do we get the right word
order?
We could head-move the verb up the tree to
Neg.
So what’s happening in “long negation”?
SS
DPj
T
C.-ka NegP T
-ess
Neg
Neg+Vi
an-ilk
VP
tj
V
DP
ti
ku chayk-ul
CP
Korean negation


DS
C
T
NegP T
-ess
Suppose that the DS is the same,
except that we now have a special
“nominalized” form of the verb (cf.
reading, perhaps).
Neg
Neg
an
VP

Suppose that ilk-ci doesn’t move to
Neg.
DP
C.-ka DP

See how we might analyze this?
C
-ta
TP
Chelswu-ka ku chayk-ul ilk-ci
anh-ess-ta.
Chelswu-NOM that book-ACC read-CI NEG.do-PAST-DECL
‘Chelswu didn’t read that book.’
V
V
ilk-ci
ku chayk-ul
CP
Korean negation



But since ilk-ci didn’t move to an (and
in fact probably doesn’t count as a verb
anymore, but as a noun), there is no
verb in the area.
C
C
-ta
TP
Chelswu-ka ku chayk-ul ilk-ci
anh-ess-ta.
Chelswu-NOM that book-ACC read-CI NEG.do-PAST-DECL
‘Chelswu didn’t read that book.’
At SS, we have a tense morpheme (ess) which needs to attach to a verb.
SS
DPj
T
C.-ka NegP T
-ess
Neg
Neg
an
VP
tj
V
DP
V
ilk-ci
ku chayk-ul
CP
Korean negation




Actually, in more literary Korean it is
also possible to find an uncontracted
form that looks like …ilk-ci ani ha-essta.
Just like English…
C
C
-ta
TP
Chelswu-ka ku chayk-ul ilk-ci
anh-ess-ta.
Chelswu-NOM that book-ACC read-CI NEG.do-PAST-DECL
‘Chelswu didn’t read that book.’
Hence we insert do (in Korean, ha),
which gets contracted as anh.
PF
DPj
T
C.-ka NegP T
ha-ess
Neg
Neg
an
VP
tj
V
DP
V
ilk-ci
ku chayk-ul
Dutch

Let’s shift gears a bit and look at some Germanic
languages. Like Dutch (this works for German too).
What’s the word order? Is it head-initial? Head-final?

Wim koopt het boek.
Wim buys the book
‘Wim buys the book.’
…dat Wim het boek koopt.
…that Wim the book buys
‘…that Wim buys the book’

Dutch

Dutch main clause sentences are not
SVO…

Dat boek kocht Wim gisteren.
that book bought Wim yesterday
‘That book Wim bought yesterday.’

Gisteren kocht Wim dat boek.
yesterday bought Wim that book
‘Yesterday Wim bought that book.’
Dutch V2




When there is an auxiliary, the auxiliary goes
second, and the verb goes last.
Gisteren heeft Karel dat boek gekocht
Yesterday has Karel that book bought
‘Yesterday Karel bought that book.’
and when embedded, they both go at the
end…
…dat Karel gisteren dat boek gekocht heeft.
…that Karel yesterday that book bought has
‘…that Karel bought that book yesterday.’
Dutch V2




XP V S O
XP Aux S O V
…C S O V
…C S O V Aux
What’s happening here?
Compare:
Has Bill gone to the movies?
I wonder if Bill has gone to the movies.
Dutch V2




XP V S O
XP Aux S O V
…C S O V
…C S O V Aux
It appears that in main clauses the tensed
verb moves to C; in embedded clauses it
doesn’t.
Like in English questions…
Dutch V2




XP V S O
XP Aux S O V
…C S O V
…C S O V Aux
So, is Dutch head-initial or head-final?
By now we should be able to tell what VP,
TP, and CP look like.
Dutch V2
CP
DPj
Wim
C

C+[Vi+T]k TP
koopt
t j

T
tk
VP
tj

V
DP
het boek
ti
Dutch appears to be
head-final in VP and TP,
but head-initial in CP.
The (finite) verb moves
from V to T and then to
C in matrix clauses
Then something moves
into SpecCP. It could be
the subject…
Dutch V2
CP
DPk
C
het boek

C+[Vi+T]k TP
koopt
DPj
Wim

T
tk
VP
tj

V
tk
ti
Dutch appears to be
head-final in VP and TP,
but head-initial in CP.
The (finite) verb moves
from V to T and then to
C in matrix clauses
Then something moves
into SpecCP. It could be
the object…
Dutch V2
CP
C

TP
C
dat
DPj
Wim
T
T+Vi
VP
tj
koopt
V
DP
het boek
ti
When C is filled (in
an embedded
clause, with dat),
the verb moves
only to T and
nothing moves to
SpecCP.
V2


So another parameter of variation between
languages seems to be whether V moves to C
and requires SpecCP to be filled (“V2”).
English has a little bit of what appears to be
“residual V2” with negatives.



Never had I seen such a thing.
Under no circumstances will I buy that book.
There are complications with treating this like V2 in
German and Dutch (can you think of them?) which
will be addressed in Syntax II.
VSO: Reminder about Irish

Irish: VSO, Aux SVO.







Phóg Máire an lucharachán.
kissed Mary the leprechaun
‘Mary kissed the leprechaun.’
Tá Máire ag-pógáil an lucharachán.
Is Mary ing-kiss the leprechaun
‘Mary is kissing the leprechaun.’
We might have also analyzed this as V-to-T-to-C (like
German but without the filled SpecCP), but for…
VSO order in Irish

There seem to be cases when C is filled and
the order is still VSO—so the verb doesn’t
move to C.






An bhfaca tú an madra?
Q See you the dog
‘Did you see the dog?’
Duirt mé gur phóg Máire an lucharachán.
Said I that kissed Mary the leprechaun
‘I said that Mary kissed the leprechaun.’
VSO order in Irish




We had suggested that Irish leaves
the subject in VP-internal position.
In essence, then, Irish seems to be a
CP
SS
V-to-T type language—but without the
EPP.
C
Note: Not everyone likes saying that
a language can choose not to obey
TP
C
the EPP. However, if the alternative
has EPP universal and some
T
languages can use proexp to satisfy it,
the two alternatives are not different.
T+Vi VP
We now have an alternative way to
DP
analyze this… one that doesn’t
V
require either suspension of the EPP
or proexp. What is it?
ti
…
VSO in Std. Arabic


Standard Arabic seems to be VSO like Irish, but
can provide clearer evidence for this idea that
VSO leaves the subject lower than the standard
(English-type) subject position.
Std. Arabic: Allows both VSO and SVO orders.


ra?a-a l-?awlaad-u Zayd-an
saw-3S the-boys-NOM Zayd-ACC
‘The boys saw Zayd.’ (VSO)
l-?awlaad-u ra?a-w Zayd-an
the-boys-NOM saw-3PL Zayd-ACC
‘The boys saw Zayd.’ (SVO)
VSO in Std. Arabic
ra?a-a l-?awlaad-u Zayd-an
saw-3S the-boys-NOM Zayd-ACC
‘The boys saw Zayd.’ (VSO)
 l-?awlaad-u
ra?a-w Zayd-an
the-boys-NOM saw-3PL Zayd-ACC
‘The boys saw Zayd.’ (SVO)


Notice that the verb agrees with the
subject in the SVO order—in the VSO
order the verb just carries 3sg agreement.
SVO/VSO order in Std. Arabic

C
CP
SS
C
VSO

TP
T

T+Vi VP
DP
ti
V
…
That is, there is
agreement marking
where the subject is in
CP
SS
SpecTP.
SVO
C
Where there is nothing
(or proexp) in SpecTP,
TP
C
the agreement comes
out as (a default) 3sg.
DPj
T
This looks like an
example of Spec-head
T+Vi VP
agreement. Features
are checked for identity.
tj
V
ti
…
SVO/VSO order in Std. Arabic

C
CP
SS
C
VSO
But what if we take our
AgrSP and AgrOP phrases
into account?

What is AgrSP after all?
TP

What is the EPP?
T

Do we need to say that
Arabic (or Irish) is a “nonEPP” language, or make
use of proexp?
T+Vi VP
DP
ti
C
CP
SS
C
SVO
TP
DPj
T
T+Vi
VP
tj
V
…
V
ti
…
SVO/VSO order in Std. Arabic
SS
VSO

AgrSP

AgrS
AgrS+Tj+Vi TP
DPk

T
Spec-head agreement is
usually considered to be
responsible for agreement
between subject and verb
AgrSP
(via participation of AgrS).
C has [+wh] feature which
DPk AgrS
needs to match with a
feature of its specifier.
Spec-head agreement is AgrS+T +V TP
j
i
often taken to be, broadly
speaking, a kind of “feature
tk’
sharing” configuration.
…
VP
tj
tk
SVO
T
…
VP
t i’
tk
V
ti
SS
…
V
ti
…
And back to little v…

John gave the book to Mary.

Recall that this is the structure that
we came up with to get the word
order right, and to comply with X-bar
theory.
We determined there must be a “little
v”, a light verb, to which the V moves
overtly. This little v assigns the Agent
q-role. So English has a v in its
lexicon that assigns the Agent q-role.


A somewhat radical idea occurs…
vP
SUB
v
v
VP
DO
V
V
IO
VP shells

Let’s go back and consider VP shells a bit in connection
with unaccusatives.
VP

The ice melted.
The boat sank.
The door closed.



V
V
melt
DP
the ice
The ice, the boat, the door are all Themes, suggesting
that the verbs are unaccusative—the argument starts in
object position.
VP shells
VP
V
V
melt



DP
the ice 

So far, so good.
Now, Bill melted the ice.
The ice is still Theme. The verb is still
melt.
Uniform Theta Assignment Hypothesis
(UTAH) (Baker 1988): Two arguments
which fulfill the same thematic function
with respect to a given predicate must
occupy the same underlying (DS)
position in the syntax.
So the ice must still be a complement of
the verb at DS.
VP shells
VP

V

V
melt
DP
the ice

In Bill melted the ice what have we
done?
We’ve added a causer, an agent.
Bill caused [the ice to melt].

We’ve already supposed that the light
verb v assigns the Agent q-role in
ditransitives…

It isn’t much of a jump to think of v as
actually having a contribution to the
meaning, something like CAUSE.
VP shells
vP
DP
Bill
v
v
VP
V
V
melt
DP
the ice

Bill melted the ice.

So, something like this, where the
main verb moves up to the light
verb (which we had evidence for
in ditransitives).


Later, Bill will move to SpecTP,
SpecAgrSP for Case and EPP
reasons.
Why does V move to v? We’ll
assume that it does this for a reason
analogous to why V moves to T (for
French verbs, say). Might be
universal, actually. “v needs a V to
move to it”.
VP shells



Note. Even though v may carry a “causative”
meaning, this does not mean that it is
synonymous with the English word “cause”.
There is a difference in the “directness” of the
causal connection. What it really seems closest
to is “Agent”.
The water boiled.
Bill boiled the water


Billi T ti v+boil the water
Bill caused the water to boil

Bill cause TP
Back to the radical idea…






So, we have v, which assigns an Agent q-role.
We have Agent q-roles in clauses other than Bill
sank the boat and Bill gave a boat to Edward.
We also have an Agent q-role in sentences like
Bill ate the sandwich.
Are there two ways to assign the Agent q-role?
What if v is the way the Agent q-role is
assigned?
What would Bill ate the sandwich look like?
Bill ate the sandwich
vP
DP
Bill

Well, we already saw essentially
what it would look like. It looks
just like Bill melted the ice.

v assigns Agent to Bill, V (eat)
assigns Theme to the sandwich.
v
v
VP
V
V
eat
DP
 Also note: The subject is still in
the
“SpecVP” except that we’ve
sandwich
sharpened our picture of what
“VP” is. A “VP” with an Agent is
really a vP and a VP.
Bill lied.

In fact, things get weirder…
Consider Bill lied.
That’s got an Agent, so it’s got a v.

So, it could look like this.

But lie is really (also?) a noun,
right? Is this a coincidence?

vP

DP
Bill
v
v
VP
V
V
lie

(How about Bill danced, Bill walked,
Bill sneezed, …)
Bill lied?
vP
DP
Bill

One proposal out there about this
kind of verb is that it really is built
from the noun.

That is, we have v+N, which
would come out to mean
something like ‘Bill was the agent
of a lie.’

If that’s right, it means v really is
its own thing, and moreover, it’s
responsible for giving these verbs
their verby nature.
v
v
NP
N
N
lie
AgrOP
The sandwich was eaten

Let’s think about passives.

What happens in a passive?
AgrO
AgrO vP

DP
Bill
v
v

VP
V
The Agent q-role is suppressed.
Accusative Case is no longer
available to the object.
What does that mean in these
terms, considering v to be the
V
DP
thing that assigns Agent and
eat
the
sandwich AgrOP to be the thing that gives
Case?

The sandwich was eaten

Sure, no vP, no AgrOP.

Everything else follows as before:




VP
V
V
eat
DP
the
sandwich

The sandwich needs Case.
SpecTP needs to be filled.
The sandwich moves to SpecTP.
The sandwich moves to SpecAgrSP.
Burzio’s generalization is now that
there is an AgrOP if and only if
there is a vP. They come and go
together.
The sandwich was eaten
VP

V
V
be
So, we end up with something like
this, where AspP is where vP used
to be.
AspP

Asp
Asp
-en
VP
V
V
eat
DP
the
sandwich
(Since passive is actually a different
sort of thing from aspectual have
eaten and be eating, sometimes
people call this VoiceP)
And back to ditransitives

In the split-INFL system, we have
something like this:

The V moves to v, and eventually to
AgrO.
AgrO assigns Case, and it should
only be compatible with transitive
verbs, so v needs to get close
enough to verify that they match (we
can think of this as AgrO “pulling up”
the v). (Perhaps source of BG?)
The object moves to SpecAgrOP to
get/check Case.
The subject moves up to TP and
SpecAgrSP.



AgrOP
AgrO
AgrO vP
SUB
v
v
VP
DO
V
V
IO









