Introduction to Transformational Grammar, LINGUIST 601

November 17, 2008

Infinitival Complementation 1: Control

So far we have looked more or less exclusively at finite clauses. We now turn our attention to
infinitival clauses.

We will look at the following three cases:

(1) a. Control : the subject of the infinitival is a null pronoun called PRO:
David tried [PRO to dance].
[PRO to dance with David] is fun.
b. Raising: the subject of the infinitival moves to a higher subject position:
Makotoi appears [ti to be happy].
Joeyi seems [ti to be exhausted].
c. ECM : the subject of the infinitival is an overt NP:
I believe [Angela to be innocent].
Minjoo wants [him to stay].

1 PRO and Control

(2) a. Dave tried [to write a paper about causatives].
b. Junko decided [to visit UMass in December].
c. Bernhard plans [to teach a seminar in the Fall].

• The embedding predicate is a two place predicate - someone who tries/decides/plans some-
thing and something that is tried/decided/planned.

• The infinitival clause typically cannot have an overt subject.

(3) a. *Dave tried [Dave to write a paper about causatives].

b. *Junko decided [Junko to visit UMass in December].
c. *Bernhard plans [Bernhard to teach a seminar in the Fall].

PRO is typically taken to be in complementary distribution with overt pronouns and pro.
Potential Counterexample: predicates like want

1.1 Some motivation for PRO

Given our assumptions about θ-roles and the EPP, we need to assume that there is a null subject
in the infinitival clauses under discussion. What could this subject be?
• An important theoretical consideration:

(4) θ-criterion: An argument DP must receive a θ-role, and may receive only one θ-role.
The only one θ-role requirement rules out movement and hence DP-traces/copies of DPs in the
infinitival subject in the infinitival subject position.

pro could be a possibility (see Borer (1989)), but pro as generally conceptualized can only appear
in positions where it can get case and where overt DPs may also appear.

If we keep to the only one θ-role part of the θ-criterion, we need to postulate a new kind of entity
- a null pronoun, called PRO, which can satisfy the EPP requirement, receive a θ-role, and which
does not need case.

(5) Some null DPs:

a. DP-trace/copies of DPs that we do not pronounce: products of movement
b. pro - a null pronoun that (i) alternates with overt pronouns/ (ii) needs case
c. PRO - a null pronoun that (i) does not alternate with overt pronouns/ (ii) does not
need case

(6) a. Junkoi decided [PROi to visit UMass in December].

b. Fredi promised Alex [PROi to finish his paper by Monday].
c. Fred persuaded Alexi [PROi to finish her paper by Monday].

(The co-indexing indicates what DP controls the PRO. If the proposal that this relationship
is essentially a semantic one is correct, then these indices do not need to be represented in
the syntax.)

(7) PRO can only be the subject of a non-finite clause.

a. *Pam believes PRO solved the problem.
b. *Sarah saw PRO.
c. *Sarah saw pictures of PRO.
Potential Counterexample: finite control in Persian (Darzi (2008))

As always, once we postulate a null entity we have to make sure that it appears only where we
want it to appear. Here are some major proposals.

(8) Case Resistance (Bouchard (1984))

a. PRO is case-resistant.
b. PRO is in complementary distribution with overt pronouns and pro because the latter
need to be assigned (or check) a Case.

(9) PRO Theorem (Chomsky (1981))

a. PRO is [+anaphoric,+pronominal].
b. The only way for it to survive is to not have a governing category i.e. not have a
governor.
c. Non-finite T0 is not a governor.

2
 d. Case-licensing positions are always governed, hence PRO cannot appear there.

(10) Null Case (Chomsky and Lasnik (1993), Martin (2001))

a. PRO has null case, a special case that only it can bear.
b. Non finite T0 licenses Null Case.
i. Naomi tried to solve the problem.
ii. It was difficult for Naomi to solve the problem.
iii. Naomi believes *(her) to have solved the problem.
iv. *It seems to Naomi to have solved the problem.
v. She seems to Kim to have solved the problem.
Martin (2001)’s refinement: raising T0 does not license case, control T0 licenses Null
Case.

These approaches all share a premise, namely that PRO either does not have Case or the Case that
it has is unique to PRO. However, in every case where we can test this premise, it has turned out
to be false: see Sigurdsson (1991) for Icelandic, Cecchetto and Oniga (2004) for Latin and Italian,
and Landau (2004) and Landau (2006) for a theory that constrains the distribution of PRO given
that PRO has normal case.

1.2 Control as Raising

Control and Raising are generally taken to involve different modules of the grammar - control
theory and movement, respectively.

Control Theory, however, remains poorly understood and there has been an active line of work
originating in Hornstein (1999), which tries to derive Control via movement. To derive Control
via movement, the only one θ-role part of the θ-criterion must be given up.

Note that deriving Control via Movement is not the same thing as saying that Raising and Con-
trol are the same. There is no denying that there are many many properties on which control
constructions differ from raising constructions.

Landau (2003) suggests that since control constructions are very different from raising construc-
tions, we want to keep the derivation of the two distinct. In their reply to Landau (2003), Boeckx
and Hornstein (2004) point out that there are in fact certain parallels between control and raising
and claim that the non-parallels can be derived from independent differences between raising
and control.

1.3 Different Kinds of Control

Control construction can be subdivided along several dimensions. Some of these subclasses seem
more amenable to a movement analysis than others.

3
1.3.1 Subject vs. Object Control

(11) Ditransitive control predicates:

a. subject control: Fred promised Alex to finish his paper by Monday.
b. object control: Fred persuaded Alex to finish her paper by Monday.

There are many object control predicates but very few subject control predicates. This has led
some researchers to propose a Minimal Distance Principle, which forces object control (cf. Rosen-
baum (1967)). MDP-violating subject control is taken to involve a special structure where the
Minimal Distance Principle is in fact respected (cf. Larson (1991)).

The conclusion (most?) other researchers have reached is that while the availability of a control
relationship is a property of the syntax, the exact identity of the controller (subject vs. object)
follows from the semantics of the embedding predicate (cf. Dowty (1985), Culicover and Jack-
endoff (2001), Jackendoff and Culicover (2003) among others) The intuition expressed by Dowty
is that there could not be a verb that had the same meaning as promise, but which was object
control (and vice versa). Facts from acquisition add an additional twist to this discussion - MDP
violating subject control verbs like promise seem to be acquired much later (see Boeckx and Horn-
stein (2003)).

1.3.2 Obligatory, Arbitrary, and Optional Control

Most of the cases of control seen so far involve obligatory control i.e. the subject of the infinitival
clause can only be interpreted as dependent on an argument of the embedding predicate for its
interpretation.

(12) Obligatory Control:

a. Infinitival (non-wh)-Complements:
i. Angela tried [to disinvite him].
ii. Fred promised Alex [to finish his paper by Monday].
iii. Fred persuaded Alex [to finish her paper by Monday].
b. Infinitival Adjuncts:
i. Andre read Rushdie’s article about Coetzee [to make a presentation in his class].
ii. Roumi went to Tromsoe [to talk to Sylvia].

Not all instances of control are obligatory - in some cases the PRO seems to lack an obvious
controller1 and takes on a generic/arbitrary interpretation - these cases are referred to as PROarb .
Arbitrary control is diagnosed by its ability to bind oneself and the availability of a paraphrase
that involves the pronoun one.

(13) Arbitrary Control:

a. - unique argument of embedding predicate:
i. [PROarb to walk along Paradise Pond in the Fall] is fun.
ii. [PROarb to behave oneself in public] is important.
1 But see Epstein (1984) and Bhatt and Izvorski (1997) who argue that even in these cases, there is an implicit controller.

4
 iii. It is not allowed [PROarb to perjure oneself].
b. - part of a wh-CP:
i. Minjoo knows [how PROarb to behave oneself in public].
ii. Tim wonders [how PROarb to protect oneself from creditors].

In some cases, such as when the infinitival clause is embedded in a wh-CP, arbitrary control is not
the only option:

(14) optional control:

a. Minjooi knows [how PROi to behave herself in public].
b. Timi wonders [how PROi to protect himself from creditors].

For obvious reasons, these cases are referred to as involving optional control.

Landau (2000)/Landau (2003) makes a further distinction noting that the PRO subject of initial
adjuncts can be interpreted as non-arbitrary and yet not controlled by the matrix subject.

(15) Non-Obligatory Control (NOC)

a. Maryi was baffled. [Even after PROi revealing her innermost feelings], John remained
untouched.
b. Maryi lost track of Johnj , because, [PROi,j having been angry at each otheri,j ], he had
gone one way and she another.
c. [Having PROi just arrived in town], the main hotel seemed to Billi to be the best place
to stay.

He argues that the wh-infinitival cases of arbitrary control are really special cases of partial control
and should not be mixed with cases of NOC like the above. In the wh-infinitival cases, even the
arbitrary PRO must include the subject in its reference. True disjoint reference is not allowed.

(16) a. Johni wondered [who PRO to introduce hisi fiancee/*himi to].

b. Johni asked [how PRO to talk to Mary/*himi about oneself].

1.3.3 Partial vs. Exhaustive Control

There is also a class of cases where the matrix predicate provides only part of the reference of the
subject of the infinitival clause (see Landau (2000) for details).

(17) Partial Control:

We thought that ....
a. Hei wanted [PROi+ to meet in the lobby/do the dishes together].
b. The chairi preferred [PROi+ to gather at 6].
c. Billi regretted [PROi+ meeting without a concrete agenda].
d. Maryi wondered [whether PROi+ to apply together].

Other predicates do not allow partial control.

(18) a. *Bevi began [PROi+ to do the dishes together].

b. *The chairi managed [PROi+ to gather at 6].

5
1.3.4 Implicit Control

The controller of PRO can be an implicit argument - i.e. an argument that does not seem to be
syntactically projected. Languages differ in the extent to which they allow for implicit arguments
to control a PRO.2

(19) Unaccusative vs. Passives

a. No implicit argument, No Control:
* The shipi sank [PROi to collect the insurance].
b. Implicit agent, Control:
The shipi was sunk [PROi to collect the insurance].

(20) Implicit accusatives vs. Implicit datives:

a. Implicit Accusatives:
This leads *(onei ) [PROi to draw the following conclusion].
b. Implicit Datives:
John said/shouted (to the visitorsi ) [PROi to return later].
(Unlike English, both are good in Italian. See Rizzi (1986).)

1.4 Some Properties of Control Constructions

(Setting aside cases of NOC. For that see Landau (2001))

• The controller can never be an expletive.

(21) a. *Therei tried [PROi to annoy David].

b. *Iti hopes [PROi to win]. (with expletive it)

This is a definitional property of control.

• The PRO is always a subject.

(22) a. Hei tried [PROi to annoy David].

b. *Hei tried [David to annoy PROi ].
c. Hei tried [PROi to be annoyed at David].

• The controller of PRO needs to be an argument of the predicate to which the infinitival clause is
attached.

(23) a. Hei thinks [that Ij tried [PRO∗i/j to annoy David]].

b. Hei thinks [that Ij persuaded Mildredk [PRO∗i/∗j/k to leave]].

(24) c-command (follows from argument requirement and the fact that the clausal complement
is the innermost argument)
a. [Hisi parents]j tried [PRO∗i/j to annoy David].
2 There is a generalization called Bach’s Generalization according to which object controllers may not be omitted. Bach’s

Generalization holds as long we restrict ourself to accusative object controllers. Dative object controllers may be implicit.

6
 b. Ii persuaded [Mildredj ’s mother]k [PRO∗i/∗j/k to leave].

• PRO cannot be a real expletive.

(25) a. [For Mary to dance] would be fun.

b. [PRO to dance] is fun.
c. [For there to be a party tonight] would be fun.
d. *[PROexpl to be a party tonight] would be fun.
e. [For it to seem that Mary is a non-smoker], she will have to get new rugs.
f. *[PROexpl to seem that Mary is a non-smoker], she will have to get new rugs.

(26) (from Lasnik (1992):244)

a. [*(There) having been a robbery], there was an investigation.
b. There was a crime without [*(there) being a victim].
c. [PROi having witnessed the robbery], Johni aided the investigators.
d. Harryi was a witness without [PROi being a victim].

But PRO can function as weather/pseudo-ambient it:

(27) a. It can hail [without it snowing].

b. It can hail [without PROit snowing].

References
Bhatt, R., and R. Izvorski (1997) “Genericity, Implicit Arguments, and Control,” paper presented
at SCIL VII; available at ftp://ling.upenn.edu/studentpapers/bhatt/PROarb.ps.
Boeckx, C., and N. Hornstein (2003) “Reply to ‘Control is not Movement’,” Linguistic Inquiry 34:2,
269–280.
Boeckx, C., and N. Hornstein (2004) “Movement under Control,” Linguistic Inquiry 35:3, 431–452.
Borer, H. (1989) “Anaphoric AGR,” in O. Jaeggli and K. Safir, eds., The Null Subject Parameter,
Kluwer, Dordrecht, 69–109.
Bouchard, D. (1984) On the Content of Empty Categories, Foris, Dordrecht.
Cecchetto, C., and R. Oniga (2004) “A Challenge to Null Case Theory,” Linguistic Inquiry 35:1,
141–149.
Chomsky, N. (1981) Lectures on Government and Binding, Foris, Dordrecht.
Chomsky, N., and H. Lasnik (1993) “The theory of principles and parameters,” in J. Jacobs,
A. von Stechow, W. Sternefeld, and T. Vennemann, eds., Syntax: An international handbook of
contemporary research, de Gruyter, Berlin, 506–569.
Culicover, P. W., and R. Jackendoff (2001) “Control is not Movement,” Linguistic Inquiry 32:3,
493–512.
Darzi, A. (2008) “On the vP analysis of Persian Finite Control Constructions,” Linguistic Inquiry
39:1, 103–116.

7
Dowty, D. (1985) “On recent analyses of the semantics of control,” Linguistics and Philosophy 8,
291–331.
Epstein, S. D. (1984) “Quantifier-pro and the LF Representation of PROarb ,” Linguistic Inquiry 15:3,
499–505.
Hornstein, N. (1999) “Movement and Control,” Linguistic Inquiry 30:1, 69–96.
Jackendoff, R., and P. W. Culicover (2003) “The Semantic Basis of Control in English,” Language
79:3, 517–556.
Landau, I. (2000) Elements of Control: Structure and Meaning in Infinitival Constructions, Kluwer,
Dordrecht.
Landau, I. (2001) “Control and Extraposition: The Case of Super-Equi,” Natural Language and
Linguistic Theory 19:1, 109–152.
Landau, I. (2003) “Movement out of control,” Linguistic Inquiry 34:3, 471–498.
Landau, I. (2004) “The Scale of Finiteness and the Calculus of Control.” Natural Language and
Linguistic Theory 22, 811–877.
Landau, I. (2006) “Severing the Distribution of PRO from Case,” Syntax 9, 153–170.
Larson, R. (1991) “Promise and the theory of control,” Linguistic Inquiry 22:1, 103–139.
Lasnik, H. (1992) “Two Notes on Control and Binding,” in R. K. Larson, S. Iatridou, U. Lahiri,
and J. Higginbotham, eds., Control and Grammar, Studies in Linguistics and Philosophy 48,
Kluwer Academic Publishers, Dordrecht, Reidel, 235–252.
Martin, R. (2001) “Null Case and the Distribution of PRO,” Linguistic Inquiry 32:1, 141–166.
Rizzi, L. (1986) “Null Objects in Italian and the Theory of pro,” Linguistic Inquiry 17:3, 501–558.
Rosenbaum, P. (1967) The grammar of English predicate complement constructions, MIT Press, Cam-
bridge, MA.
Sigurdsson, H. A. (1991) “Icelandic Case-Marked Pro and the Licensing of Lexical Argument,”
Natural Language and Linguistic Theory 9:2, 327–363.