Danilo Reggiani
A short introduction to Control Theory
(doi: 10.1418/34542)
Licenza d’uso
L’articolo è messo a disposizione dell’utente in licenza per uso esclusivamente privato e personale, senza
scopo di lucro e senza fini direttamente o indirettamente commerciali. Salvo quanto espressamente
previsto dalla licenza d’uso Rivisteweb, è fatto divieto di riprodurre, trasmettere, distribuire o altrimenti
utilizzare l’articolo, per qualsiasi scopo o fine. Tutti i diritti sono riservati.
A SHORT INTRODUCTION TO CONTROL THEORY
A SHORT INTRODUCTION
TO CONTROL THEORY
DANILO REGGIANI
1. INTRODUCTION1
Control theory was born decades ago2 to explain some interface phenomena
between syntax and semantics. Control theory tried to generate a
mechanism able to explain how infinitives are assigned a subject that is not
phonologically present. Two main interpretations of the infinitive clauses
and their subject gap were recognized, an antecedent-referring interpretation
(1a) and a generic interpretation (1b).
(1) a. The fox told the crow [Δ to sing].
b. [Δ Singing for the fox] can be dangerous.
The antecedent-referring interpretation can be split in two different
categories: a local antecedent interpretation and a non-local (long distance,
henceforth LD) antecedent interpretation. The local antecedent is necessarily
placed in the higher clause that selects the infinitive clause as an argument
(2a). The LD antecedent can be higher (2b) or lower (2c) than the infinite,
but it can never be an argument of the verb that selects the infinitive clause
as a complement (Manzini, 1983; Landau, 1999). The subject of the infinite
can also be inferred by the context (2d) (Bresnan, 1982).
(2) a. The old mani said that the foxj told the crow [Δ*i/j to sing].
b. The crowi doesn’t know that [Δi singing hisi song] can be dangerous.
c. [Δi Singing a song with a piece of cheese in hisi mouth] convinced every
reader that the crowi must be very stupid.
1
I gratefully acknowledge two anonymous reviewers of Lingue e Linguaggio for insightful
remarks and comments. List of abbreviations: ACC: accusative; COMP: complementizer; DAT:
dative; DFLT: default; HAB: habitual; IRR: irrealis; NOM: nominative; PL: plural; SG: singular.
2
The term “control” however did not appear until the late Seventies. Rosenbaum (1965b)
used the term “Identity Erasure Transformation”.
133
d. The foxj knows what she wants. [Δj Challenging the crow] was a smart move.
It is clear that LD control and generic control are characterized by a
common property: they do not need a local antecedent. The definition of
local antecedent for control has been disputed (Landau, 1999, 2001). It is
traditionally assumed that a local antecedent is a c-commanding argument
of the verb that selects the infinitive (controlled) clause (Williams, 1980).
Locality and interpretation are thus matching into two types of control
predicates: obligatory control (OC) and non obligatory control (NOC). OC is
local and is not able to license arbitrary control or LD control. Sentences like
(2a) show also that OC contexts tend to block the emergence of NOC. OC is
also characterized by the impossibility of strict reading, while NOC allows
such a reading. OC (3a) allows only sloppy reading under VP-ellipsis, while
NOC (3b) allows both sloppy reading and strict reading under VP-ellipsis.
(3) a. The hare wants [Δ to run faster], and the tortoise does too.
b. The dove thinks that [Δ helping her friend] was a good idea, and the ant
does too.
Finally, OC allows only the DE SE reading, while NOC allows both DE
RE and DE SE reading.
(4) a. The crow Δi said that [Δi singing that song] was very stupid.
b. The crow Δi remembered [Δi being stupid].
In a DE SE reading the crow has the property of being a stupid animal
and being fully aware of that. In a DE RE reading the crow has the property
of being a stupid animal while not being aware of that.
134
A SHORT INTRODUCTION TO CONTROL THEORY
The verb meet is a collective predicate, a verb that requires a plural subject
(6d) or at least a semantically plural3 subject like board (6c). Every DP that is
not at least semantically plural cannot be the subject of a collective predicate
(6e). What happens then in (6a) and (6b)? A verb like want allows the
controlled argument to be semantically plural, even if the controlling antecedent
is purely singular. A case like (6a) is “perfectly natural when the speaker has
some salient group in mind” (Landau, 2004: 833), like the guys in (6a).
A verb like manage, on the contrary, does not allow the controlled
argument to be semantically plural, given a purely singular antecedent.
Landau (1999) recognizes two major properties of partial control.
(7) a. Partial control is an instance of OC.
b. Only some matrix verbs allow partial control in their complement clauses.
The second important property of partial control is that only some verbs
in the matrix clause allow partial control of the complement clause. The
whole class of complement controlling verbs (i.e. roughly the verbs that
select an infinitive clause as a complement in English) can be divided,
according to Landau (1999), in two distinct categories: partial control (PC)
verbs (matrix verbs able to yield partial control) and exhaustive control
(EC) verbs (matrix verbs unable to yield partial control).
(9) Exhaustive Control verbs
a. *John told Mary that hei forgot Δ*i+ to meet at six.
b. *The chairi dared Δ*i+ to convene during the strike.
c. *John told Mary that hei had Δ*i+ to separate before it was too late.
3
Landau (2004) calls “mereology” the feature that characterizes semantically plural nouns
like board: board is [+Mer] whereas John is [-Mer].
135
DANILO REGGIANI
136
A SHORT INTRODUCTION TO CONTROL THEORY
OC NOC
EC PC
Is subject to locality constraints ∨ ∨ ×
Allows LD control × × ∨
Allows arbitrary control × × ∨
Allows DE RE reading × × ∨
Allows strict reading × × ∨
Allows PC reading × ∨ ∨
TABLE 1. EC, PC AND NOC
137
DANILO REGGIANI
138
A SHORT INTRODUCTION TO CONTROL THEORY
139
DANILO REGGIANI
into complement clause, the argument of the matrix verb that is syntactically
nearest to the controlled argument is the controller (Rosenbaum, 1965a).
To account for cases of OC in adjunct control, Hornstein (2001, 2003)
argues that the MTC can be extended to control into adjuncts headed by
after, before, while, etc. (i.e. to control into OC adjuncts). In those adjuncts
the controlled argument is c-commanded and controlled by the subject of the
matrix verb. The problem with this generalization is that objects seem to be
able to bind and thus c-command into adjuncts (23b).
(23) a. Johni will drink no winej [before Δi/*j eating some food]
b. John will drink no winej [before itj is ready for drinking]
A theory of OC adjunct control based on direct A-movement across
object would violate the MLC. To solve this problem, Hornstein (2001,
2003) proposes that OC adjuncts contain a trace in subject position and
that this trace is part of a “sideward movement” chain. Sideward movement
moves a constituent from one subtree to another and allows consequently the
extraction of an argument out of an adjunct. Sideward movement extracts the
highest (i.e. closest to the edge) argument of an adjunct and copies it into an
argument position into the matrix clause. Hornstein (2001), following Nunes
(1995), argues that sideward movement is involved also in parasitic gaps.
(24) a. Johni read a book before ti eating a pineapple.
b. Which booki did you read ti before Fred reviewed ti?
Sideward movement allows John (which is, being the subject, the
leftmost/highest argument) in (24a) to move out of the adjunct in the
subject position of the matrix clause. In (24b) the same sideward movement
allows the object of the adjunct clause to move from the adjunct clause
subtree to the matrix clause subtree, in the position of the object.
Hornstein claims that the best feature of the MTC is the elimination of the
“control module”. The control module is basically that theoretical mechanism
that, given the distribution of PRO, provides the interpretation of PRO in different
environments (OC vs. NOC). The control module was in fact assumed by early
theories of control (Chomsky, 1981): “very little is known about the control
module. Its powers are generally invoked when there is no principled or insightful
alternative at hand” (Boeckx & Hornstein, 2004: 433, footnote 2). The MTC
provides a principled account of the old theory of the double nature of PRO,
pronominal and anaphoric, by adopting t instead of PRO in OC and pro instead
of PRO in NOC. However, PRO was in fact a good way of accounting for the
distribution of control phenomena, allowing the same infinitival complement
in English to receive a different interpretation in a different environment.
How is it possible to license control phenomena and avoid wrong interpretations
in the MTC? How is it possible, in other words, to let A-movement stick to OC
environments and let pro stick to NOC environments?
140
A SHORT INTRODUCTION TO CONTROL THEORY
141
DANILO REGGIANI
Only the insertion of a trace, by the way, would satisfy the principle (27a).
Hence only a trace can be in the subject position of the infinitive clause in (30).
142
A SHORT INTRODUCTION TO CONTROL THEORY
4
Henceforth I will call simply ec the empty category that syntactically represents null object
and null dative.
143
DANILO REGGIANI
144
A SHORT INTRODUCTION TO CONTROL THEORY
145
DANILO REGGIANI
146
A SHORT INTRODUCTION TO CONTROL THEORY
147
DANILO REGGIANI
148
A SHORT INTRODUCTION TO CONTROL THEORY
These features are the key to understand the distinction between EC and
PC. These features are also necessary to understand the distribution of PRO
and controlled predicates cross-linguistically. The feature [±Agr] marks the
presence of overt phi-agreement on the head I: English infinitives and gerunds
are characterized by the feature [-Agr] whereas Balkan subjunctives (that are
able to license control) are characterized by the feature [+Agr]. The feature
[±T], on the heads C and I, distinguishes tensed [+T] and untensed [-T]
complements: tensed complements allow tense mismatch and partial control;
untensed complements, on the contrary, disallow both partial control and
tense mismatch. The feature [±R], as assumed by Reinhart & Reuland (1993),
distinguishes DPs that are referential expressions [+R] (and thus can be used
to directly select an entity in the discourse) and elements like anaphors and
logophors that are referentially defective [-R] and need to have a linguistic
antecedent in the sentence. PRO, being [-R], needs a sentential antecedent
in order to be interpreted in OC contexts. Landau assumes that PRO “freely
picks phi-features before entering the derivation. In the case of PRO those
features are anaphoric, but valued (like the phi-features of himself)” (Landau,
1999: 17). The feature [±R] on PRO, also, enters the derivation as valued.
Recall now Landau’s generalization on PC and EC, reported in section
1.1. Landau assumes that PC and EC are characterized also by different
syntactic mechanisms that yield the possibility of semantic number
mismatch between controller and controllee (PC) or the impossibility of
such a semantic number mismatch (EC).
(51) Exhaustive control
[CP … F DP … [ CP C[-T] [IP PRO[-R] [ I’ I[-T, -Agr, -R] [VP t[PRO] …]]]]]
149
DANILO REGGIANI
150
A SHORT INTRODUCTION TO CONTROL THEORY
in the first agreement between I and PRO: PRO entered the derivation with
an unvalued feature [Mer], according to the assumption (53a).
The mismatch in semantic number between the controlling DP and PRO is
then possible: PRO, having no value for feature [Mer], can be the subject of
a collective predicate without committing feature clash. On the contrary, PRO
that has inherited from F the feature [-Mer] (or a simple singular DP marked
with the feature [-Mer]) cannot be the subject of a collective predicate.
The double route of agreement and the relation between tense and licensing
of partial control is roughly summarized in (55).
(55) a. TENSED COMPLEMENT (PC)
i. tense mismatch is tolerated
ii. I-to-C movement blocks direct agreement between F and PRO
the unvalued feature [Mer] remains unvalued
iii. partial control is possible
b. UNTENSED COMPLEMENT (EC)
i. tense mismatch is NOT tolerated
ii. direct agreement between F and PRO
the unvalued feature [Mer] is valued by F
iii. partial control is IMpossible
THE DISTRIBUTION OF CONTROL PHENOMENA. Landau (2004) deals with the
distribution of the category PRO crosslinguistically. PRO is licensed in
clauses, headed by a C head, that are characterized by abstract agreement
and/or anaphoric tense (the one which characterizes untensed complements).
Landau (2004) stipulates that the presence of PRO is regulated by an
“R-assignment Rule”. The R-assignment Rule is a “honest stipulation”
(Landau, 2004) that allows or blocks the presence of the feature [±R] (and
hence the capability of agreeing with PRO) on the heads C and I of every
clause. The aim of the R-assignment Rule is to relate the licensing of PRO
to the features (the type of agreement and tense) of the controlled clauses.
(56) R-assignment Rule
a. tense mismatch is tolerated
b. Each head (C, I) bears the feature [-R] iff it is marked [-Agr, +T],
[+Agr, -T] or [-Agr, -T]
A direct consequence of the R-assignment Rule is that if a head lacks
[±Agr] or [±T] then no feature [±R] is assigned.7
The features [Agr] and [T] are specified on the head I. The head I bears
the feature [+Agr] if there is overt subject-verb agreement and the feature
[-Agr] if there is abstract (i.e. covert) subject-verb agreement. Speaking of
7
The R-assignment rule is the core of the mechanism that licenses control phenomena and is
what Landau also calls “the calculus of control”.
151
DANILO REGGIANI
the head C, Landau assumes that it bears the same type of [T] features that
is present on the head I, unless the clause is characterized by independent
tense: in that case the head C has no feature [T].
(57) Specifying [T] on embedded I/C
Anaphoric tense → [-T] on I and C
Dependent tense → [+T] on I and C
Independent tense → [+T] on I and Ø on C
He further stipulates that the feature [Agr] on C is usually not present
on C, since “agreement is rarely espressed on C” (Landau, 2004: 871). If
[Agr] is present on C, then it must be [+Agr]. [-Agr] is unavailable on C.
Recall that data from English and Balkan languages provide us two
case studies. The first case study is the case of control into infinitival
complements that can be found in English or Italian. English infinitives do
not license a DP/pro (i.e. an R-expression or a pronoun) in subject position.
(58) EC-infinitive
*[CP … F DP …[ CP
C[-T] [IP [I’ I[-T, -Agr, *-R] [VP DP / pro[+R] …]]]]]
[CP … F DP …[ CP
C[-T] [IP PRO[-R] [ I’
I[-T, +Agr, -R] [VP t[PRO] …]]]]]
152
A SHORT INTRODUCTION TO CONTROL THEORY
153
DANILO REGGIANI
3. CASE TRANSMISSION
Landau (2003) argues that the change of case in Icelandic between the
controlling argument and the controlled argument (the case is marked on a
154
A SHORT INTRODUCTION TO CONTROL THEORY
155
DANILO REGGIANI
Icelandic that are problematic for the hypothesis presented by Boeckx &
Hornstein (2006a). Quirky case preservation is persistent in A-movement
but is impossible in control.
(64) ECM constructions
a. Ég tel strákana (hafa verið) kitlaða.
I believe the.boy.PL.ACC to.have been tickled.ACC
‘I believe the boys to have been tickled’
b. Ég tel strákunum (hafa verið) bjargað.
I believe the.boy.PL.ACC to.have been rescued.DFLT
‘I believe the boys to have been rescued’ (Bobaljik & Landau, 2009)
(65) Passivized ECM constructions
a. Strákarnir eru taldir (hafa verið) kitlaðir.
the.boy.PL.NOM are believed.PL to.have been tickled.NOM
‘The boys are believed to have been tickled’
b. Strákunum er talið (hafa verið) bjargað.
the.boy.PL.DAT are believed.DFLT to.have been rescued.DFLT
‘The boys are believed to have been rescued’ (Bobaljik & Landau, 2009)
Examples (64) and (65) show that A-moved quirky case remains quirky and
is not overwritten by a structural case. This clearly rules out the overwriting
hypothesis by Boeckx & Hornstein, i.e. the hypothesis that the subject of a
quirky case assigning verb changes its case after undergoing A-movement.
Bobaljik & Landau then observe that if, according to the MTC, control
is (almost) reduced to raising, it is impossible to explain the difference
between (66a) and (66b).
(66) a. Ég tel Maríu hafa verið tekna/*tekin af lögreglunni.
I believed Maria.ACC to.have been taken.ACC/*NOM by the.police
‘I believed Maria to have been taken by the police’
b. Ég bað Maríu að vera *tekna/tekin af lögreglunni.
I asked Maria.ACC to be taken.*ACC/NOM by the.police
‘I asked Maria to be taken by the police’ (Thrainsson, 1979)
The hypothesis by Boeckx & Hornstein is unable to explain why there is
a nominative in the lower clause in (66b), where an accusative is expected.
According to Bobaljik & Landau (2009), there is no case transmission
between the complement clause and the matrix clause. Each clause hosts an
independent case assignment. That is how the controlled argument in (66b)
receives nominative case. Boeckx, Hornstein & Nunes (2010) argue that a
little assumption can solve the problem discovered by Bobaljik & Landau
(2009). They propose that the assignment of an additional theta-role to a
DP removes its quirky case. This proposal solves the problem exemplified
in (64) and (65): not every A-moved DP changes its quirky case, but only
those DPs that have received a new thematic role, i.e. only DPs involved
156
A SHORT INTRODUCTION TO CONTROL THEORY
in a control relation. In (64) and (65) the quirky DPs conserve their quirky
case when they undergo passivization and raising because passivization and
raising do not involve the assignment of new thematic roles.
The proposal by Boeckx, Hornstein & Nunes (2010) then solves the
problem exemplified in (64) and (65), but what about (66)? Boeckx,
Hornstein & Nunes observe that, according to the literature, the accusative
case transmission in control structures varies strongly across the speakers
(Sigurðsson, 2008). In other words, the acceptability judgments in (66b) are
not correct since some speakers allow case transmission across such control
constructions. (67) shows more correct acceptability judgments.
(67) Hún bað Ólaf að fara bara einn/einan í veisluna.
she.NOM asked Olaf.ACC to go just alone.ACC/NOM to the.party
‘She asked Olaf to just go alone to the party’ (Sigurðsson, 2008)
Boeckx, Hornstein & Nunes propose that both case transmission and case
independence in (67) can be accounted for by the MTC. Since, according to
the MTC, non-finite I does not assign nominative case, the controller (Olaf
in (67)) receives case only after moving into the matrix clause. Which case
then surfaces on the lower copy of the control chain? Boeckx, Hornstein &
Nunes predict two possible outcomes: the structural case assigned to the
controller in the matrix clause or a default case (nominative in Icelandic)
which is applied to those traces that do not bear a case feature. The answer
given by Boeckx, Hornstein & Nunes, summed up in (68), tentatively
accounts for cases of case transmission and case independence in Icelandic.
(68) a. Quirky case is erased (and requires rewriting) when the DP is assigned a
new theta-role.
b. Caseless copies:
i. can inherit the case received by a higher copy.
ii. can receive default nominative case only if the higher copy that
receives case has a different set of theta-roles.
This answer given by Boeckx, Hornstein & Nunes is of course
incomplete since it does not offer a detailed explanation for the behaviour
of caseless copies. The theory proposed by the authors does not precisely
predict when caseless copies inherit case from the A-chain or receive default
case. However no theory, at the time being, seems to be able to make such a
prediction, since case transmission in Icelandic is still quite a murky subject.
The summary given in (68) is a kind of case transmission filter. It is
a set of generalizations that describes how case transmission works in
Icelandic. These generalizations, despite being able to address the issue
of case transmission, still show some problems. First of all, the filter
exemplified in (68) lacks elegance. Take for example (68b-ii). It is not clear
what should be the link between the surfacing of the default case and the
157
DANILO REGGIANI
4. CONCLUSIONS
This article has shown the main changes in control theory in the last
fifteen years. We have seen how different theories have provided different
interpretations for different infinitival complements. The challenge between
the theories will be won by the theory that is less assumptive and that is able
to explain most phenomena. The MTC (section 2.1) and the so-called calculus
of control (section 2.2) are the main contenders. They are however far from
158
A SHORT INTRODUCTION TO CONTROL THEORY
REFERENCES
Adler, A. N. (2006). Syntax and Discourse in the Acquisition of Adjunct Control. Ph.D.
dissertation. Cambridge, MA: Massachusetts Institute of Technology.
Alexiadou, A., Anagnostopoulou, E., Iordăchioaia, G. & Marchis, M. (2010). No
objections to backward control. In N. Hornstein & M. Polinsky (Eds), Movement
Theory of Control (pp. 89-118). Amsterdam: Benjamins.
Bobaljik, J. D. & Landau, I. (2009). Icelandic control is not A-movement: The case
from case. Linguistic Inquiry 40 (1), 113-132.
Boeckx, C. & Hornstein, N. (2004). Movement under control. Linguistic Inquiry 35 (3),
431-452.
Boeckx, C. & Hornstein, N. (2006a). Control in Icelandic and theories of control.
Linguistic Inquiry 37 (4), 591-606.
Boeckx, C. & Hornstein, N. (2006b). The virtues of control as movement. Syntax 9 (2),
118-130.
Boeckx, C. & Hornstein, N. (2007). On (non-)obligatory control. In W. D. Davies & S.
Dubinsky (Eds.) (pp. 251-262).
Boeckx, C., Hornstein, N. & Nunes, J. (2010). Icelandic control really is A-movement:
Reply to Bobaljik and Landau. Linguistic Inquiry 41 (1), 111-130.
Borer, H. (1989). Anaphoric AGR. In O. Jaeggli & K. J. Safir (Eds.), The Null Subject
Parameter (pp. 69-109). Dordrecht: Kluwer.
Bresnan, J. (1982). Control and complementation. Linguistic Inquiry 13 (3), 343-434.
Chierchia, G. (1984). Topics in the Syntax and Semantics of Infinitives and Gerunds.
Ph.D. dissertation. Amherst, MA: University of Massachussetts at Amherst.
Chomsky, C. (1969). The Acquisition of Syntax in Children from 5 to 10. Cambridge,
MA: MIT Press.
Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, N. (1993). A Minimalist Program for linguistic theory. In K. Hale & J.
Keyser (Eds.), The View from Building 20 (pp. 1-52). Cambridge, MA: MIT Press.
Chomsky, N. & Lasnik, H. (1993). The theory of Principles and Parameters. In J. Jacobs,
A. von Stechow, W. Sternfeld & T. Vennemann (Eds.), Syntax: An International
Handbook of Contemporary Research, Vol. 1. (pp. 506-569). Berlin: Walter de Gruyter.
Costantini, F. (2008). Movement or control? Some hints from floating quantification.
University of Venice Working Papers in Linguistics 18, 131-144.
Davies, W. D. & Dubinsky, S. (Eds.) (2007). New Horizons in the Analysis of Control
and Raising. Dordrecht: Springer.
Hornstein, N. (1999). Movement and control. Linguistic Inquiry 30 (1), 69-96.
Hornstein, N. (2001). Move! A Minimalist Theory of Construal. Malden/Oxford: Blackwell.
Hornstein, N. (2003). On control. In R. Hendrick (Ed.), Minimalist Syntax (pp. 6-81).
159
DANILO REGGIANI
Malden/Oxford: Blackwell.
Hornstein, N. (2006). A short note on obligatory control. UMWPiL 14, 39-46.
Kuno, S. (1975). Super Equi-NP deletion is a pseudo-transformation. NELS 5, 29-44.
Landau, I. (1999). Elements of Control. Ph.D. dissertation. Cambridge, MA:
Massachusetts Institute of Technology.
Landau, I. (2001). Control and extraposition: the case of super-equi. Natural Language
& Linguistic Theory 19 (1), 109-152.
Landau, I. (2003). Movement out of control. Linguistic Inquiry 34, 471-498.
Landau, I. (2004). The scale of finiteness and the calculus of control. Natural Language
& Linguistic Theory 22 (4), 811-877.
Landau, I. (2007). Movement resistant aspects of control. In W. D. Davies & S.
Dubinsky (Eds.) (pp. 293-325).
Landau, I. (2008). Two routes of control: evidence from case transmission in Russian.
Natural Language & Linguistic Theory 26 (4), 877-924.
Lebeaux, D. (1984). Anaphoric binding and the definition of PRO. NELS 14, 253-274.
Lee, F. (2003). Anaphoric R-expressions as bound variables. Syntax 6 (1), 84-114.
Manzini, M. R. (1983). On control and control theory. Linguistic Inquiry 14 (3), 421-446.
Monahan, P. (2003). Backward object control in Korean. In G. Garding & M. Tsujimura
(Eds.), WCCFL 22 Proceedings (pp. 356-369). Somerville, MA: Cascadilla Press.
Nunes, J. (1995). The copy theory of movement and linearization of chains in the
Minimalist Program. Ph.D. dissertation. University of Maryland.
Polinsky, M. & Potsdam, E. (2006). Expanding the scope of control and raising. Syntax
9 (2), 171-192.
Reggiani, D. (2010). Dyslexia and the Acquisition of Syntax: Passive and Control.
Ph.D. dissertation. University of Verona.
Reinhart, T. & Reuland, E. (1993). Reflexivity. Linguistic Inquiry 24 (4), 657-720.
Rizzi, L. (1986). Null objects in Italian and the theory of pro. Linguistic Inquiry 17 (3),
501-557.
Rodrigues, C. (2007). Agreement and flotation in part and inverse partial control
configurations. In W. D. Davies & S. Dubinsky (Eds.) (pp. 213-229).
Rosenbaum, P. S. (1965a). A principle governing deletion in English sentential
complementation. Technical report. International Business Machine Corporation.
Rosenbaum, P. S. (1965b). The grammar of English predicate complement
constructions. Ph.D. dissertation. Massachusetts Institute of Technology.
Sigurðsson, H. Á. (2008). The case of PRO. Natural Language & Linguistic Theory 26
(2), 403-450.
Terzi, A. (1997). PRO in finite clauses. The Linguistic Review 14 (4), 335-360.
Thrainsson, H. (1979). On Complementation in Icelandic. New York: Garland Press.
Torrego, E. (1996). On quantifier float in control clauses. Linguistic Inquiry 27 (1), 111-126.
Varlokosta, S. (1993). Control in Modern Greek. UMWPiL 1, 144-163.
Williams, E. (1980). Predication. Linguistic Inquiry 11 (1), 203-238.
Danilo Reggiani
Università degli Studi di Verona
Facoltà di Medicina e Chirurgia
P.le Scuro 10, 37134 Verona
Italy
e-mail: danilo.reggiani@gmail.com
160