Transcript vincent - University of Essex

LFG Winter School 2004
Control & Complementation revisited
Nigel Vincent
University of Manchester
Issues
•
•
•
•
•
•
the semantic vs syntactic basis of control
the unity of control and raising
diachrony
partial vs exhaustive control
tense in control complements
backwards control
Syntax vs semantics:
two fallacies
i)
a semantic account = a lexical account
ii)
Structuralism vs eclecticism: if part of the
answer is syntactic, it is better if all of the
answer is syntactic
semantic ≠ lexical
[Culicover & Jackendoff provide] ‘a list of
controllers coded by thematic role: some
verbs are agent control verbs, others patient
control … and so on. This reduces the
theory of control to a lexical catalogue.’
[Hornstein & Boeckx 2003: 270]
eclecticism
The theory of control involves a number of
different factors: structural configurations,
intrinsic properties of verbs, other semantic
and pragmatic considerations.
[Chomsky 1981: 78-79]
Structuralism vs eclecticism:
a false dichotomy
“All agree that grammatical structure is
part of any adequate approach to control.
What distinguishes structuralists from
eclectics is whether this information
exhausts what is needed. All things being
equal then, structuralism is preferable if
attainable.”
Hornstein (2003: 26)
Semantics and control:
two real issues

Does a controlled complement correspond
to a property or a proposition?

How does the property/proposition contrast
relate to the distinction between COMP and
XCOMP?
Proposition vs Property
Dalrymple
Higginbotham
Hornstein
Landau
Pollard & Sag
Rosenbaum
Zec
Asudeh
Chierchia
Culicover & Jackendoff
Dowty
Jacobson
Montague
Theory independence
Property
Proposition
LFG
Asudeh
Dalrymple
Minimalism
Hornstein
Landau
The Chierchia argument
a)
b)
c)
d)
Nando tries whatever Ezio tries
Ezio tries to jog at sunrise
ERGO: Nando tries to jog at sunrise
Entailment fails if complement of (b) is
understood as the proposition Ezio jogs at sunrise
rather than the property jog at sunrise
Dalrymple on Chierchia

Sloppy vs strict identity ambiguities are not
always susceptible to this solution
Nando does whatever Ezio does
E broke his (=E’s) arm playing football
N broke his (=N’s) arm playing football
Arguments against
property-based analysis

Reflexive/reciprocal binding (cf above)

Wide scope/de re vs narrow scope/de dicto
Raising: both wide (de re)
and narrow (de dicto) possible
A goblin seemed to pinch Gonzo
= i)
(x seemed to be a goblin) &
(x pinched Gonzo)
= ii)
(x is a goblin) &
(x seemed to pinch Gonzo)
Equi: only wide scope (de re)
A goblin tried to pinch Gonzo
≠ i)
(x tried to be a goblin) &
(x pinched Gonzo)
= ii)
(x is a goblin) &
(x tried to pinch Gonzo)
Asudeh’s account


seem and try both take XCOMP at f-structure
Glue language distinguishes between the way the
semantic resources are consumed:
try :
(try, leave) John
seem:
(leave, John) seem
NB ‘Structure sharing is not necessarily at odds with
resource-sensitivity.’
[Asudeh 2002: 18]
Scope & Asudeh’s account

Equi semantics only gives wide scope

Raising semantics allows both scopes
The paradox of seem

Semantically seem must take a proposition

Syntactically all agree seem takes XCOMP

Yet XCOMP intuitively maps to a property
Proposition/property &
COMP/XCOMP
COMP
XCOMP
Property
look like
try
Proposition
say
seem
Serbo-Croat (Zec 1987)
Petar
je
pokusao
Peter
be.3SG
try.PSTPRT
da
dodje
COMP come.3SG.PRES
‘Peter tried to come’
Two arguments for the unity of
equi & raising verbs

Diachrony

Cross-linguistic differences
Diachrony (Barron 2001)

Equi verbs become raising verbs by a
gradual process (grammaticalization)
i) English promise, threaten
ii) ‘want’ in many langs > Future marker
Icelandic case
preservation & raising
a)
Drengina
vantar
mat
boys.def.acc lack.3sgpres food.acc
‘The boys lack food’
b)
Drengina
vir∂ist
boys.def.acc seem.3sgpres
vanta
mat
lack.inf
food.acc
‘The boys seem to lack food’
Icelandic equi
c)
Eg
vonasttil
I.nom hope
efni í
material in
a∂
to
vanta ekki
lack not
ritger∂ina
thesis.def
‘I hope not to lack material for my thesis’
Control in Tagalog
(Kroeger 1993)




Confirms validity of semantic approach to
control à la Sag & Pollard (1991)
Same verb can trigger both f-control and acontrol
f-control constructions defined over
syntactic relations
a-control constructions defined over
semantic relations
Partial vs exhaustive control
l
l
l
l
l
a)
the
The chair managed [PRO to gather
committee at 6]
PRO = the chair
The chair preferred [PRO to gather
b)
at 6]
PRO = the chair + the committee
‘… one can already see how damaging the
very existence of partial control is to the
thesis “control is raising”. Simply put: there
is no partial raising.’
[Landau 2003: 493, emphasis his]
Landau’s map of control
Obligatory Control
Non-oblig Control
Restricted to complement
In subject & adjunct
(VP-internal) infinitives (VP-external) infinitives
Exhaustive
Partial
Long
Arbitrary
distance
Tenseless
Tensed
PRO is a
PRO is
logophor
generic
Tense and control
‘an infinitival complement belongs to the
PC class iff it is tensed’
[Landau 2000: 6]
PC vs EC verbs
[– tense]; exhaustive
aspectual
(begin, continue, …)
modal
(need, be able, …)
implicative
(dare, manage, …)
[+ tense]; partial
desiderative
(want, prefer, …)
factive
(hate, regret, …)
propositional
(claim, believe, …)
interrogative
(wonder, ask, …)
PC and LFG



Partial control is a sub-case of anaphoric
control
Separate TENSE features in main and
embedded clause
An account (still to be developed) of the
interaction of the two TENSE features
Bill said Sally would arrive late
PRED

SUBJ
TENSE



COMP




' say < (SUBJ) (COMP) >'

'Bill'


present
PRED
' arrive < (SUBJ) >'


' Sally'
SUBJ

TENSE future



' late'
ADJ


Bill persuaded Sally on Tuesday
to leave on Thursday
PRED

SUBJ
OBJ

TENSE
ADJ



COMP



' persuade < (SUBJ) (OBJ) (COMP) 

'Bill'


' Sally'

past


' on Tuesday'

PRED
' leave < (SUBJ) >'



' pro'

SUBJ


TENSE future




' on Thursday'

ADJ


Culicover & Jackendoff



free control (= non-obligatory control)
• arbitrary control
• long distance control
nearly free control (= discourse control)
unique control (= obligatory control)
Backward control (BC)
‘BC is a biclausal control configuration in which
the lower coindexed subject is expressed and the
thematic subject in the higher clause is
unpronounced.’ [Polinsky & Potsdam 2002: 261]
Languages showing BC




Tsez, Bezhta, Tsaxur (Nakh-Daghestanian)
Malagasy (Austronesian)
Japanese
Jacaltec (Mayan)
(and perhaps a few others)
Properties of BC languages
Languages with BC tend to have some/all the
following properties:
 verb at clause edge (VOS or SOV)
 BC occurs with aspectual verbs
 such verbs show control/raising ambiguity
 the effect is lexically specific
CONTROL
[kid-ba
girl.II-ERG
RAISING
kid
girl.II-ABS
ziya
b-isr-a]
y-oq-si
cow.III.ABS III-feed-INF II-beginPAST.EVID
[ziya
b-isr-a]
y-oq-si
cow.III.ABS III-feed-INF II-beginPAST.EVID
‘The girl began to feed the cow’
Evidence for BC



Clitic buy/yuy is restricted to 2nd position in
the main clause
Raising structure:
girl.ABS yuy [cow feed] begin
Control structure:
[girl.ERG cow feed] yuy begin
Theoretical implications of BC
‘… if our description of Tsez is on the right track,
then it argues for a syntactic theory that permits
BC. We suggest that a minimalist architecture in
which movement may take place overtly or
covertly in conjunction with a movement analysis
of control … successfully accounts for BC.’
[Polinsky & Potsdam 2002: 277]
BC: movement vs nonmovement



Structure sharing (cf reconstruction)
f-command vs c-command
The two verbs ‘begin’ in Tsez
begin 1 (raising): < (XCOMP)> (SUBJ)
begin 2 (equi):
??
begin 2
Two possibilities:
< (SUBJ) (XCOMP)> :
< (SUBJ) (COMP)>:
but (SUBJ) is in
the wrong place
but violates fcommand
[girl
SUBJ
cow feed]
OBJ XCOMP
begin
<SUBJ, XCOMP>
VP
(XCOMP) =
S

NP
(SUBJ) = 
S

S
V
= =
Quantifiers in Tsez BC
(Cormack & Smith 2004)

*[Each boy.ERG book read] begin

Each boy.ABS [book read] begin
i.e. BC is incompatible with wide scope
readings
A further problem:
control in Balinese
English
*To take the medicine was tried by me

Excluded because Adjunct cannot be a
controller (Bresnan 1982)
What is the function of to take the medicine?
A further problem:
control in Balinese
Balinese
[Ø-naar ubad ento]SUBJ tegarang
AV.eat
medicine that OV.try

tiang
1PSG
Arka & Simpson (1998) assume a level of
syntactic a-structure where tiang is a term.
This circumvents the problem of how to get
control into SUBJ.
Conclusions

Desirability of maintaining a unified account of
obligatory control and raising

The LFG account extends naturally to cover new
sorts of data, e.g backwards control

The LFG account offers insights not otherwise
easily captured, e.g. property based view

BUT partial control effects suggest redrawing the
boundary between f-control & a-control