Elective Course: Control in English & Romanian: The Infinitive/Subjunctive divide

Course 4: Expanding Control: Control in Romanian

Last time: Minimalist perspectives on Control in English: Landau (1999 211! " comprehensive picture o# $C (EC%&C! ' ($C
1) &utting things in perspective
- all the contexts weve looked at so far: infinitives in English
*1: What is their e+uivalent in Romanian? ,: the su-.unctive (/con.unctiv0!
(1) a. John managed% tried%started [to nderstand !ontrol" (E! ver#s)
a. $on a reu1it%2ncercat%a 2nceput [s% &n'eleag% !ontroll."
#. John 3ants%hopes%3onders ho3 [to nderstand !ontrol (heor)" (*! ver#s)
#. $on vrea%sper4%se 2ntrea-4 cum [s% &n'eleag% teoria controlli"
- as +art of the ,-alkan .+rach#nd/ 0o1anian (geneticall) a 0o1ance langage) has significantl) lost its infinitives as
co1+le1ents to ver#s (while still retaining +rodctive infinitive strctres as co1+le1ents of ad2ectives and nons or
infinitival te1+oral adver#ials
1
*2: 3re s#2nctive for1s in 0o1anian finite or non-finite?
*5: What are the 1arks of the s#2nctive?
(4) a. 5aria vrea [s4 6tie 1ai 1lte des+re con2nctiv"
M& 7 8*67su-.8 9.
#. 5aria vrea [s% 6tie 6i $on: 6i elevii ei 1ai 1lte des+re9."
#. 5aria vrea ;elevii ei s% 6tie 99: ;s% elevii ei 6tie 9..
c. 5aria vrea [ca elevii ei s% 6tie 1ai 1lte 99"
[ca des+re con2nctiv s% 6tie 1ai 1lte elevii ei"
C& < dis+laced e1#edded 1aterial 7 M& " 9&
d. ?? 5aria vrea ca s4 6tie 9.

- strict ad.acenc: -et3een the modal particle s 7 v-

- the s#2nctive co1+le1enti=er a++ears onl) as long as ite1s fro1 the e1#edded clase are dis+laced to the
co1+le1ents left +eri+her) (4c)
- in standard langage: no ad2acenc) #etween ca < s% (ca s > 1ark of s+oken:collo?ial langage)(4d)
;; *4: Can Control <heor: -e adapted to accommodate such #inite contexts= > Can Control <heor: -e
adopted #or Romanian su-.unctives= ?# :es@ how=
Let us remem-er:
in English:
A Control (initiall: proposed #or English!: a relation -et3een an antecedent and &R$@ the empt: su-.ect o# an
in#initive (nonA#inite;! complement
- &R$ in complementar: distri-ution 3ith lexical (&s (i.e./ *0@ onl) a++ears in contexts which disallow the
a++earance of a lexical A*/ cf. John tried:1anaged:dared [*0@ to leave" vs. ;John tried:1anaged 5ar) to leave)
A needs a local antecedent (an antecedent in the 5ain !lase/ A@( farther)
7 *B: ?s the empt: su-.ect o# Romanian su-.unctive complements in control contexts a! &R$ or -!
pro=
; Control environments: Landau0s classes o# control ver-s:
- as+ectal < 1odal < i1+licatives (E!)
- desiderative < interrogative < 9 (*!)
a! i# &R$ no lexical su-.ect should -e allo3ed thereC the empt: su-.ect should re+uire a local antecedent
(one in the main clause!
-! i# pro (dis.oint! lexical su-.ects should -e allo3ed 7 the empt: su-.ect should allo3 a remote antecedent
(one that is not necessaril: contained in the main clause@ -ut in a higher clause!
1
(he se of the s#2nctive in control environ1ents is a1ong the 1ost i1+ortant characteristics that 0o1anian shares with the other -alkan
langages. .ee also the *hB thesis of 5adalina *ralea: she classifies the -alkan s+rach#nd acc. to the extent to which the) have lost or still retain
their infinitives: DE)o reorganiFare a spa iului -alcanic 2n trei Fone: a! primei Fone@ 2i apar in lim-ile neogreac4 i macedonean4@ lim-i 2n care
dispari ia in#initivului cu valoare ver-al4 din sistemul gramatical este complet4C -! celei deAa doua Fone@ 2i apar in lim-ile -ulgar4 i
dialectele sGr-e ti de sudAest@ lim-i 2n care dispari ia in#initivului este aproape complet4@ el #iind 2nregistrat iFolatC c! celei deAa treia Fone@ 2i
apar in sGr-a@ croata standard i romGna@ lim-i 2n care in#initivul 2 i mani#est4 2nc4 vitalitatea)H
1
A Evidence seems to #avour the a! variant (i)e)@ &R$!) More speci#icall:@ in (exhaustive! control environments@ the
empt: su-.ect o# the su-.unctive complements seems to displa: the characteristics o# &R$@ not pro: it needs a
local antecedent and it disallo3s alternation 3ith a lexical (& em-edded su-.ect
(C) a. $on n a putut s% +lece. :: ; $on a +tt [s% +leci:s% +lec%1:s% +leca'i"
a.;$on n a +tt s% +lece 5aria. (no (dis2oint) lexical s#2ect allowed)
a. 5ateii 6i $on2 a fost invita'i la +etrecere/ &ns% $on2 n +oate [s% vin% e2:;i" (local antecedent)
"" ?oni nu a putut 6&R$i s4 plece8
#. 5aria nu reu1e1te s% a2ng% la +etrecere. :: ;5aria n re6e6te [s% a2ng:s% a2ngi:s% a2nge19"
#. ; 5aria n re6e6te s% a2ng% 3na la +etrecere. (no (dis2oint) lexical s#2ect allowed)
#. 5ariai 6i 3na2 a fost invitate la +etrecere dar 5ariai n va re6i [s% a2ng% ei:;2" (local antecedent)
"" Mariai nu reu1e1te 6&R$i s4 a.ung4 la petrecere8
c. 5atei a 2nceput s% cDnte. :: ;5atei a &nce+t [s% cant:s% cDn'i:s% cDnt%1"
c. ;5atei a &nce+t s% cDnte 5aria
c. 3nai toc1ai a sosit 6i 5atei2 a &nce+t [s% cDnte e2:;i" 77 Mateii 6&R$i a 2nceput s4 cGnte8
vs)
(E) a. $on vrea s% +lece. (+ossi#le co-reference reading #c. sa1e featres on the two ver#s)
#. $on vrea [s% +lec:s% +leci:s% +lec%1:s% +leca'i"
c. $on vrea s% +lece 5ihai:to'i 1safirii (dis2oint s#2ects (overt or nll) allowed)
d. 5ateii toc1ai a sosit 6i $on2 vrea [s% +lece ei:2" FF a1#igos
1. 9.6i $on2 vrea [s% +lece e2" 77 and $on wants to leave
4. 9..6i $on2 vrea [s% +lece ei" 77 and $on wants hi1 (5atei) to leave. 77 ?oni vrea 6s4 plece proi%.8
F(his conclsion (i.e./ that *0@ is the e1+t) s#2ect of #oth non-finite (infinitives in En.) G finite (0o1anian
s#2nctives) co1+le1ents 9. rather nsettling 77 1ore shold #e said a#ot this/ #t first:
2) Irie#l: on <raditional%&revious anal:ses o# Control in Romanian
A co11on idea: in spite of agreement features (i.e., finiteness), lack of temporal independence of subjunctive
complements in control situations renders these complements non-finite/ sch that the) are na#le to check
Ao1inative (in s+ite of overt +erson 1arking on the 8). &R$ as the empt: su-.ect
; lacJ o# temporal independence K the ina-ilit: o# su-.unctive complements selected -: (EC! predicates to encode
a temporal re#erence di##erent #rom that o# the matrix ver-:
(H) a. 31 &nce+t (ieri) s% citesc ;ai/!acum" ::; Incerc s% fi venit.
#. ;A a1 +tt s% vin s%+t%1Dna viitoare. :: ; A voi +tea s% vin ieri la +etrecere.
c. ;A re6esc s% fi v%=t fil1l.
vs.
(J) a. #cum vrea s% +lec la !hicago $e %& a$rilie. (Karkas 1LME: CJ4)
#. 3c1 $on vrea ca 5ihai s% +lece 1Dine:+este o s%+t%1Dn%.
c. 3c1 trei =ile $on vroia s% 1erge1 1Dine la fil1 (dar &ntre ti1+ s-a r%=gDndit)
< 1an) athors clai1 that lack of te1+oral inde+endence (with @!:E! +redicates) correlates with the ina#ilit) of the
s#ordinate to allow the ca-co1+le1enti=er (vs. co1+le1ents of *! ver#s (volitionals)/ which allow #oth te1+oral
1is1atch as well as an overt co1+le1enti=er):
(N) a. ;$on a &ncercat ca +e 5ihai s%-l +ede+seasc%. (Bo#rovie .orin 1LLE: JO-J1)
#. ; 5aria a &nce+t ca +artitra s% o cDnte
vs.
(M) a. 5aria vrea [ca $on s% c1+ere tortl 1Dine"
#. 5aria s+er% [ca 1Dine s% c1+ere $on tortl"
2)1) Landau (24!
Claim: s#2nctive co1+le1ents in -alkan langages +arallel the #ehavior of infinitival co1+le1ents in English
a! s#2nctive co1+le1ents selected #) control ver#s (know (how), learn, manage, begin, stop, continue! evince
o-ligator: control constructions "" Controlled'subjunctives
-! s#2nctive co1+le1ents selected #) -elieve@ decide@ arrange@ persuade@ 3ish@ 3ant@ hope K partial control
constructions "" (ree'subjunctives
- the dividing line: the tense s+ecifications of the s#ordinate
4
77 !-s#2nctives descri#e a single eventP K-s#2nctives descri#e two (as+ectall) inde+endent) events
A 3orJing to3ards a uni#ied picture o# Control@ Landau (24! esta-lishes a correlation -et3een the 2 t:pes o#
IalJan su-.unctives and the t3o t:pes o# $C in Engl:
(L! a) 9 selects a CAsu-.unctive in the IalJan languages i## it selects an ECAin#initive in other languages
-) 9 selects an MAsu-.unctive in the IalJan languages i## it selects a &CAin#initive in other languages
(Landau 24: L5N!
CAsu-.unctives :
A the null su-.ect &R$ (it has to -e identical to the su-.ect o# the main clause ver-!
- descri-e a sigle event via the s#ordinating ver#
(L) a. *ot s% +lec (e > the leaving/ not the a#ilit))
#. 31 &nce+t s% citesc. (e > the readingP the 1atrix ver# is a 1ere inchoative)
c. (re#ie s% &nv%'. (e > the learning/ not the necessit) to do so)
MAsu-.unctives :
A do not impose an: restrictions on the t:pe o# (& su-.ect o# the su-ordinate (the null su-.ect pro, i)e)@ it can -e
interpreted as dis.oint in re#erence #rom the su-.ect o# the MC!
(1O) a. $on vrea s% +lece.
a. $oni vrea s% +lece +roi . )Ioni vrea s $lece Ioni)
a. $oni vrea s% +lece +ro2. ($oni vrea s% +lece 1safirii2)
- descri-e t3o separate events (aspectuall:! de#ine their o3n temporal domain
(11) a. 36 vrea s% +lec (e1> the wishingP e4 > the leaving)
#. .+er s% re6esc. (e1 > the ho+eP e4 > the scceeding)
c. 31 decis s% +lec%1 devre1e. (e1 > the decisionP e4 > the leaving)
!-s#2nctives K-s#2nctives fc. of te1+oral s+ecification G the t)+e of (nll) s#2ect
2)2) Oordan (21! (on the model o# Landau 24!: EC su-.unctives vs) &C su-.unctives
in EC su-.unctives (after 1odals:i1+licatives:as+ectal 8s) the . is *0@ #ecase
a! lexical . (and hence a nll $ro) are disallowed
(14) ;5ara1 a &ncercat s% scrii tu2%pro2 o scrisoare. (4O1O: 1CH)
-! the nll s#2ect lacks inde+endent reference 7 re?ires a c-co11anding antecedent
(1C) Mara1 a &ncercat e1 s% scrie ea 2ns41i1 o scrisoare (4O1O: 1CJ)
c! QB! is disallowed
(1E) 0ad1 stie c% Mara2 a &ncercat e2 s% scrie ea 2ns41i2 :Pel 2nsu1i1 scrisoarea
in &C su-.unctives (after volitionals) the controller is se1anticall) sg./ *0@ is se1anticall) +l. de to the collective
1odifier together (cf. Qanda 1LLL/ 4OOE)
(1H) 0ad1 crede [c% 5ara4 s+er% s% re=olve +ro#le1a *m$reun"
2)5) Miniteness ' control: <he /calculus o# control0 (Landau 211!
Claim: $C is not restrained to in#initive (i)e)@ nonA#inite! complements@ it is possi-le in #inite complements as 3ell)
- starting fro1 the ass1+tion that Rthe +heno1enon of finite control is crosslingisticall) ro#stS (4O11: 1O1)/ Qanda
sets ot to isolate the #initeness determinants o# $CP the) are 4:
1) semantic tense/ detecta#le #) the possi-ilit: o# tense mismatch -t3) the main clause and the su-ordinate and
4) morphological agreement/ that is the +resence of overt ver# 1or+holog) (+hi-featres (nr/ gender))
Qanda (4O11: 1OC) for1lates the following generali=ation on the finiteness rle for @!:
(1J) The finiteness rule for O
?n a #ull: speci#ied complement clause (i)e)@ the ? head carries slots #or -oth 6<8 and 6,gr8!:
a) ?# ? carries -oth semantic tense and agreement (67<@ 7,gr8!@ (C o-tains)
-) Else3here@ $C o-tains)
C
.o1e clarifications:
- Qanda (4O11) discsses the distinction -et3een $C and Dno controlH ((C!
- A! co1+le1ents are those which host lexical B*s or $ro as s#2ect
- the rle in (1J) is Ran elsewhere rleS (C al3a:s o-tains in complements speci#ied as 67 <@ 7 ,gr8 (which
ensres lexical B*s:+ro as s#2ects (as well as se1antic tense))P $C is the else3here case o# (C@ o-taining in
environments 3here either o# the t3o heads or -oth are negativel: speci#ied)
$C is predicted to o-tain in the #ollo3ing three t:pes o# contexts:
1) [- (/ - 3gr" (e.g./ English (ntensed) infinitives 7 E!)
(1N) a. 5ar)i re1e1#ered:forgot [*0@i:;2:;-ill to lock the door". (4O11: 1OE)
#. 5ar) didnt 1anage [*0@i:;2:;-ill to lock the door"
4) 6A<@ 7 ,gr8 (e.g./ -alkan s#2nctives/ so1e inflected infinitives) ;;;
C) [< (/ -3gr" (e.g./ English (tensed) infinitives 7 *!)
(1M) 5ar)i +lanned:hated [*0@i:;2:;-ill to lock the door"
F o# interest #or us: the situation in 2!: in s+ite of their te1+oral +ro+erties/ 0o1anian s#2nctives are alwa)s s+ecified
[< 3gr" #) virte of their overt agree1ent 1arkers (finiteness)
control in Romanian 3ill ultimatel: depend on the temporal speci#ication o# the complement:
- i# the complement is 6A<8 "" control%$C
- i# the complement is 67<8 "" no control%(C
-for 0o1./ E!:@! ver#s are ass1ed to select [-(" co1+le1ents 77 engender @! G *0@ as the e1+t) s#2ect:
(1M) 5ariai va &ncerca [*0@i s% n i se faca dor de -charesti". (Qanda 4O11: 14O-141)
5) Qome pro-lems #or previous anal:ses (#or those anal:ses that argue in #avour o# &R$ as the
empt: su-.ect o# Romanian su-.unctives in control environments!
5)1) all the a#ove exa1+les ass1e that the (ni?e) controller alwa)s a++ears: has to a++ear in the 1atrix clase -ut
there are e!amples which clearl" show that this uni#ue argument can appear in the subjunctive complement: (< the
ass1+tion that the canonical s#2ect +osition in 0o1anian is +ost-ver#al)
(1L) a. &ncearc% [s% cDnte 8ictor la tro1#on" (3l#oi 4OON: 1L4)
#. ;$on &ncerc% [s% cDnte 8ictor la tro1#on"
(4O) a. ei:;2 A a re6it s% vin% 8ictor2.
- these are cases of -acJ3ard control (the antecedent a++ears downstairs/ in the co1+le1ent and fro1 there it
,controls the e1+t) 1atrix s#2ect
- langages like Treek and Romanian allow #ree variation -et3een -acJ3ard and #or3ard control 3ith all
control ver-s. .ee (41) #elow/ with (41a) as an instance of forward control and (41#) #ackward control. (41c)
shows that the s#2ect can a++ear in #asicall) all the slots availa#le to it (in the 1atrix and the co1+le1ent)/ all
connected to varios degrees of +ro1inence (cf. 3l#oi 4OON):
(41) a. $on 6tie [s% danse=e tango".
#. Utie [s% danse=e tango $on".
c. [($on) 6tie ($on)" [s% danse=e ($on) tango ($on)"
A such examples pose pro-lems #or the &R$ anal:sis@ since the) show that lexical s#2ects are availa#le in the
co1+le1ent and *0@ cannot alternate with lexical B*s:+ro
5)2) Ro 3e reall: have &artial Control in Romanian= (as claimed -: Oordan 21@ see a-ove!
3l#oi (4OON) shows that nlike En. (41c)/ $omanian re#uires s"ntactic pluralit" of semanticall" plural predicates: in
(41a): eu cannot control the e1#edded s#2ect/ which is closer to a referential $ro than *0@
(41) a. ; Ei vrea [s% +lec xi *m$reun": ; 8rea [s% +lec e &1+ren%"
E
$ want s#2 leave-1sg together : want-1sg s#2 leave-1sg $ together
#. Ei vrea [s4 plec4m x *m$reun" (4OON: 1LC)
$ want s#2 leave-1+l together
c. Johni told 5ar)2 that hei wanted [*0@i<2 to leave together".
Romanian lacJs the &C e##ect mani#est in English) in Romanian@ the classical opposition is that -et3een
$(-ligator:! C(ontrol! (understood as EC! vs) ($C %((o! C(ontrol!)
4) Challenging the evidence #urther: ?s there s"ntactic control in Romanian= ?# not &R$@ 3hat
is the exact nature o# the null em-edded su-.ect o# Romanian su-.unctive complements (in
control environments!=
Claims: next ti1e
1) there is no s+ntactic control in 0o1anian there are no obligator+ control ver#s in 0o1anian (+roof (FF) fro1
s#2ect i1+licative ver#s (a *ncerca, a reu-i) G o#2ect control ver#s (a ruga, a obliga, a ordona, a *ncuraja, a
convinge)
4) in 0o1anian/ the ct-off +oint for @! is lower down Qandas @! hieracrch)
in Qanda: as+ectal 7 1odal 7 implicative 7 volitional 7 interrogative 7 9..

E! [-(" *! [<("
@r clai1: as+ectal 7 1odal 7 implicative 7 volitional 7 interrogative 7 9..

@! [-(" A@! [<("
C) a$$arent cases of @! in 0o1anian are linked to the lack of te1+oral inde+endence of the s#ordinate (i.e./ its
s+ecification as [-(") and the: are actuall: raising instances/ since the matri! verb ssuch as modals and
aspectuals displa" unaccusative behaviour
E) the few(er than +reviosl) ass1ed) control instances in 0o1anian are cases of raising 77 Romanian is a
raising rather than a control language ( En G V so1e other -alkan langages W e.g./ Treek)
H