Taalportaal - the digital language portal

Dutch
Frisian
Saterfrisian
Afrikaans

Show all

5.2.1.3.The implied subject PRO in om + te-infinitivals

quickinfo

This section is concerned with the implied PRO-subject in argumental om + te-infinitivals. We will begin in Subsection I with a more general discussion of the motivation to postulate a phonetically empty subject in (a specific subset) of infinitival clauses, subsection II continues by showing that the implied PRO-subject must be assigned a thematic role, just like any other nominal subject, subsection III concludes with a comprehensive discussion of the interpretation of the implied PRO-subject. The main topic in this discussion is the question as to whether subject/object control in examples like (370a&b) should be considered a locally restricted syntactic dependency. Our conclusion will be that this is not the case and that the factors determining the interpretation of the PRO-subject are instead determined by our knowledge of the world; cf. Van Haaften (1991:ch.4).

370

a.	Jani	beloofde	Mariej	[(om) PROi/*j	dat boek	te lezen].	subject control
	Jan	promised	Marie	comp	that book	to read
	'Jan promised Marie to read that book.'

b.	Jani	verzocht	Mariej	[(om) PROj/*i	dat boek	te lezen].	object control
	Jan	asked	Marie	comp	that book	to read
	'Jan asked Marie to read that book.'

c.	Jan keurt	het	af	[(om) PROarb	te vloeken].	generic interpretation
	Jan disapproves	it	prt.	comp	to curse
	'Jan disapproves of cursing.'

I. Why assume a phonetically empty PRO-subject?
II. Semantic restrictions on the implied PRO-subject and its controller
III. Control of the implied PRO-subject

[+] I. Why assume a phonetically empty PRO-subject?

Finite and infinitival object clauses like those in (371) differ in that the former have an overtly expressed subject (here the pronoun hij'he'), whereas the latter have a semantically implied subject. That the subject is semantically implied is clear from the fact that the two examples express the same number of thematic relations; in the two examples the matrix main verb beloven takes three arguments, the subject Jan, the direct object clause and the indirect object Peter, and the main verb lachen in the embedded clause takes one argument, which is expressed by the subject pronoun hij in the finite but remains unexpressed in the infinitival clause. This subsection shows that there are reasons for assuming that the semantically implied subject is actually syntactically present in the form of a phonetically empty noun phrase PRO; see Koster and May (1982), Paardekooper (1985/1986), Van Haaften (1991), and many others for similar arguments.

371

a.	Jan beloofde	Peter	[dat	hij/*PRO	niet	zou	lachen].
	Jan promised	Peter	that	he	not	would	laugh
	'Jan promised Peter that he wouldnʼt laugh.'

b.	Jan beloofde	Peter	[(om) PRO/*hij	niet	te zullen	lachen].
	Jan promised	Peter	comp	not	to will	laugh
	'Jan promised Peter not to laugh.'

We begin by showing that the postulated subject PRO in (371b) has specific interpretative properties; it is just like the pronoun hij'he' in (371a) in that it can be interpreted as coreferential with the subject, but not with the object of the matrix clause; we have made the interpretative restriction explicit in (372a) by means of indices. Example (372b) shows that these interpretational restrictions on PRO are not rigid, but depend on the matrix verb used: while the verb beloven'to promise' in (372a) triggers a so-called subject control reading, the verb verzoeken'to request' triggers an object control reading.

372

a.	Jani	beloofde	Peterj	[(om) PROi/*j	niet	te lachen].	subject control
	Jan	promised	Peter	comp	not	to laugh
	'Jan promised Peter not to laugh.'

b.	Jani	verzocht	Peterj	[(om) PROj/*i	niet	te lachen].	object control
	Jan	requested	Peter	comp	not	to laugh
	'Jan asked Peter not to laugh.'

As such, the interpretational restrictions do not seem to require the postulation of a syntactic element PRO, as we may simply account for these facts by attributing them to the semantics of the two verbs involved, which seems inevitable anyway. The postulation of PRO does help, however, to solve another problem concerning the interpretation of referential and reflexive personal pronouns. First consider the examples in (373) that show that referential pronouns like hem and reflexive pronouns like zichzelf are normally in complementary distribution; whereas the reflexive zichzelf must be bound by (= interpreted as coreferential with) the subject of its own clause, the referential pronoun hem cannot, and while the referential pronoun can (optionally) be bound by some element external to its own clause, the reflexive cannot.

373

a.	Jani	vermoedt	[dat	Peterj	over zichzelfj/*i	praat].
	Jan	suspects	that	Peter	about himself	talks
	'Jan suspects that Peter is talking about himself.'

b.	Jani	vermoedt	[dat Peterj	over hemi/*j	praat].
	Jan	suspects	that Peter	about him	talks
	'Jan suspects that Peter is talking about him.'

All of this was extensively discussed in Section N5.2.1.5, where it was accounted for by assuming that reflexives must be bound in a specific local anaphoric domain, while referential pronouns must be free (= not bound) in that domain. We repeat the two relevant binding conditions in (374), and refer to N5.2.1.5, sub III, for a more detailed and more careful discussion of the notions of binding and local domain; it suffices for our present purposes to simply state that in examples such as (373) the relevant local domain is the embedded clause.

374

Reflexive and reciprocal personal pronouns are bound in their local domain.

Referential personal pronouns are free (= not bound) in their local domain.

Now consider the examples in (375). Although the referential and the reflexive personal pronoun are in complementary distribution in these examples, the conditions in (374) seem to be violated: if we assume that the entire sentence is the local domain of the pronouns, the binding of the referential pronoun in example (375b) would violate condition (374b); alternatively, if the infinitival clause is assumed to be the local domain, the binding of the reflexive in example (375a) would violate condition (374a).

375

a.	Jani	beloofde	Peterj	(om)	over zichzelfi/*hemi	te praten.
	Jan	promised	Peter	comp	about himself/him	to talk
	'Jan promised Peter to talk about himself.'

b.	Jani	beloofde	Peterj	(om)	over hemj/*zichzelfj	te praten.
	Jan	promised	Peter	comp	about him/himself	to talk
	'Jan promised Peter to talk about him.'

Now, also consider the examples in (376). Assuming that the examples in (375) and (376) have the same syntactic structure, they go against the otherwise robust generalization that referential and reflexive pronouns are normally in complementary distribution: The (a)-examples show that, depending on the matrix verb, the reflexive can in principle be bound by the subject or the object of the matrix verb, and the (b)-examples show that the same thing holds for the pronoun.

376

a.	Jani	verzocht	Peterj	(om)	over zichzelfj/*hemj	te praten.
	Jan	requested	Peter	comp	about himself/him	to talk
	'Jan requested Peter to talk about himself.'

b.	Jani	verzocht	Peterj	(om)	over hemi/*zichzelfi	te praten.
	Jan	requested	Peter	comp	about him/himself	to talk
	'Jan requested Peter to talk about him.'

The advantage of postulating the implied subject PRO is that it solves the two problems discussed above and enables us to maintain the two conditions in (374) with no further ado. Consider the structures that should be assigned to the examples in (375), given in (377). Since the verb beloven'to promise' triggers subject control, the implied subject PRO must be coindexed with the matrix subject Jan. As a result, the reflexive pronoun zichzelf in (377a) is bound and the referential pronoun hem in (377b) is free in its infinitival clause.

377

a.	Jani	beloofde	Peterj [local domain	(om) PROi	over zichzelfi/*hemi	te praten].
	Jan	promised	Peter	comp	about himself/him	to talk
	'Jan promised Peter to talk about himself.'

b.	Jani	beloofde	Peterj [local domain	(om) PROi	over hemj/*zichzelfj	te praten].
	Jan	promised	Peter	comp	about him/himself	to talk
	'Jan promised Peter to talk about him.'

If we conclude from this that infinitival clauses are just like finite clauses in that they constitute a local domain for the pronouns they contain, all facts will follow. First, the subject of the matrix clause must be interpreted as coreferential with the reflexive pronoun, whereas the indirect object cannot. If the reflexive pronoun is interpreted as coreferential with the subject of the matrix clause, it will also be correctly bound in its local domain by the implied subject PRO; however, if it is bound by the indirect object of the matrix clause, it would be incorrectly free in its local domain. Second, the referential pronoun can be interpreted as coreferential with the indirect object but not with the subject of the clause: if the pronoun is interpreted by the indirect object, it is still free in its local domain, as required, but if it is coreferential with the subject, it will also be incorrectly bound by the implied subject PRO within its local domain.
Next, consider the structures in (378) that should be assigned to the examples in (376). Since the verb verzoeken'to request' triggers object control, the implied subject PRO must be coindexed with the indirect object Peter of the matrix clause.

378

a.	Jani	verzocht	Peterj [local domain	(om) PROj	over zichzelfj/*hemj	te praten].
	Jan	requested	Peter	comp	about himself/him	to talk
	'Jan requested Peter to talk about himself.'

b.	Jani	verzocht	Peterj [local domain	(om) PROj	over hemi/*zichzelfi	te praten].
	Jan	requested	Peter	comp	about him/himself	to talk
	'Jan requested Peter to talk about him.'

If we maintain the earlier conclusion that the infinitival clause constitutes a local domain for the pronouns it contains, the facts again follow. First, the indirect object of the matrix clause must be interpreted as coreferential with the reflexive pronoun, whereas the subject cannot. If the reflexive pronoun is interpreted as coreferential with the indirect object, it will also be correctly bound in its local domain by the implied subject PRO; however, if it is bound by the subject, it would be incorrectly free in its local domain. Second, the referential pronoun can be interpreted as coreferential with the subject but not with the indirect object of the matrix clause: if the pronoun is interpreted as coreferential with the subject, it is still free in its local domain, as required, but if it is coreferential with the indirect object, it will also be incorrectly bound by the implied subject PRO within its local domain.
A similar argument can be based on the behavior of the reciprocal personal pronoun elkaar'each other', which is subject to the same binding condition as reflexive pronouns. In addition, the reciprocal is bound by a plural antecedent: see the contrast between Jan en Marie groetten elkaar'Jan and Marie greeted each other' and *Jan groette elkaar'*Jan greeted each other'. For our present purpose it is also important to note that the plurality requirement cannot be evaded by assuming that the reciprocal takes a "split" antecedent; an example such as (379a) is unacceptable and the intended assertion can only be expressed by the more complex construction in (379b), in which elkaar does have a plural antecedent.

379

a.	*	Jani	stelt	Peterj	aan elkaari&j	voor.
		Jan	introduces	Peter	to each.other	prt.

b.	[Jan en Peter]i	stellen	zichi	aan elkaari	voor.
	Jan and Peter	introduce	refl	to each.other	prt.
	'Jan and Peter introduce themselves to each other.'

The crucial observation is that the ban on split antecedents seemingly breaks down exactly in those cases in which the implied subject PRO is able to take a split antecedent. The verb voorstellen'to propose' in (380a), for example, does allow an interpretation according to which Jan proposes that Marie and he himself will build a tree house; this reading can be forced by adding the modifier samen'together'. Example (380b) shows that the verb voorstellen ostensibly forces a split-antecedent reading on the reciprocal. However, given that the true antecedent is the implied subject PRO of the infinitival clause, this should not be seen as a violation of the ban on split antecedents for reciprocals.

380

a.	Jani stelde	Elsj	voor	[(om) PROi&j	(samen)	een boomhut	te bouwen].
	Jan proposed	Els	prt.	comp	together	a tree.house	to build
	'Jan proposed to Els to build a tree house together.'

b.	Jani stelde	Elsj	voor	[(om) PROi&j	elkaari&j	te helpen].
	Jan proposed	Els	prt.	comp	each.other	to help
	'Jan proposed to Els to help each other.'

To sum up, this subsection has shown that the postulation of an implicit PRO-subject in infinitival clauses is motivated by the fact that it enables us to maintain in full force a number of robust generalizations concerning binding of referential, reflexive and reciprocal personal pronouns. Without the postulation of PRO the formulation of a descriptive generalization concerning the distribution of these pronouns will become much more complex or even require special stipulations to handle cases of the kind discussed in this section.

[+] II. Semantic restrictions on the implied PRO-subject and its controller

The claim in Subsection I that the PRO-subject of the infinitival clause is semantically implied is tantamount to stating that it is assigned a thematic role by the infinitival verb. The examples in (381a-d) show that this thematic role can be agent if the infinitive is an (in)transitive, theme if the infinitive is an unaccusative, and goal if the infinitive is an undative verb. The implied subject PRO can also be the subject (external argument) of a complementive like aardig'kind' in (381e).

381

a.	Jani	probeert	[(om) PROi	te slapen].	agent
	Jan	tries	comp	to sleep
	'Jan is trying to sleep.'

b.	Jani	probeert	[(om) PROi	Marie	te helpen].	agent
	Jan	tries	comp	Marie	to help
	'Jan is trying to help Marie.'

c.	Jani	probeert	[(om) PROi	niet	te vallen].	theme
	Jan	tries	comp	not	to fall
	'Jan is trying not to fall.'

d.	Jani	probeert	[(om) PROi	het boek voor niets	te krijgen].	goal
	Jan	tries	comp	the book for free	to get
	'Jan is trying to get the book for free.'

e.	Jani	probeert	[(om) PROi	aardig	te zijn].	subject of complementive
	Jan	tries	comp	kind	to be
	'Jan is trying to be kind.'

Of course, there are a number of additional conditions that must be satisfied due to the semantic properties of the matrix verb. For example, the verb proberen'to try' implies that the PRO-subject is able to control or at least consciously affect the eventuality expressed by the infinitival argument clause. For this reason, sentences such as (382) are unacceptable or minimally trigger a stage context reading, that is, a context in which the event denoted by the verb is intentional (like falling in a training session) or involves pretense (like dying in a stage play).

382

a.	$	Jani	probeert	[(om) PROi	te vallen].	theme
		Jan	tries	comp	to fall
		'Jan is trying to fall.'

b.	$	Jani	probeert	[(om) PROi	te sterven].	theme
		Jan	tries	comp	to die
		'Jan is trying to die.'

Furthermore, the controller of PRO should ideally be able to perform the eventuality denoted by the infinitival construction. The subject of the matrix clause in examples such as (383), for example, should not only satisfy the selection restrictions of the matrix verb proberen'to try', but also those of the infinitival verb—it cannot refer to a single individual as this would not satisfy the selection restriction imposed by the infinitival verbs zich verspreiden'to spread' and omsingelen'to surround' that their subjects refer to larger sets of individuals (if headed by a count noun).

383

a.	De soldateni	proberen	[(om) PROi	zich	te verspreiden].
	the soldiers	try	comp	refl	to spread
	'The soldiers are trying to disperse.'

a'.	$	De soldaati	probeert	[(om) PROi	zich	te verspreiden].
		the soldier	tries	comp	refl	to spread

b.	De soldateni	proberen	[(om) PROi	het gebouw	te omsingelen].
	the soldiers	try	comp	the building	to surround
	'The soldiers are trying to surround the building.'

b'.	$	De soldaati	probeert	[(om) PROi	het gebouw	te omsingelen].
		the soldier	tries	comp	the building	to surround

The fact established earlier that the implied PRO-subject may be assigned the thematic role of theme predicts that om + te-infinitivals can be passivized. Sentences of this form do not seem to be very frequent and are perhaps slightly formal, but an example such as (384b) shows that this prediction is indeed correct.

384

a.		Marie werd	gekozen	tot voorzitter.
		Marie was	elected	as chairman

b.	Marie	probeerde	[(om) PROi	gekozen	te worden	tot voorzitter].
	Marie	tried	comp	chosen	to be	as chairman
	'Mary tried to be elected Chair.'

It is important to note that although impersonal passivization is fully acceptable in Dutch, this is never possible with infinitival clauses. The contrast between (385a) and (385b) suggests that infinitival clauses differ from finite clauses in that they cannot be impersonal but must have a PRO-subject. Of course, one might want to explore the possibility that there is a PRO-subject in (385b) with a thematic role similar to that of the expletiveer in (385a) and claim that the unacceptability of (385b) is due to the fact that subject control would lead to an incoherent interpretation with Marie functioning as the subject of the impersonal passive. However, this would lead us to expect impersonal passivization of the matrix clause to improve the acceptability of the utterance, and example (385c) shows that this is not borne out. We therefore conclude that om + te-infinitivals must have a PRO-subject and that (385b) is unacceptable because it fails to meet this condition.

385

a.	Er	werd	gelachen	in de zaal.
	there	was	laughed	in the hall
	'There was laughter in the hall.'

b.	*	Mariei	probeerde	[(om)	gelachen	te worden].
		Marie	tried	comp	laughed	to be

c.	*	Er	werd	geprobeerd	[(om)	gelachen	te worden].
		there	was	tried	comp	laughed	to be

[+] III. Control of the implied PRO-subject

The implied PRO-subjects of argumental om + te-infinitivals are normally controlled by the subject or the object of the verbs selecting them, although there are cases in which the PRO-subject takes a split antecedent or receives a generic interpretation. One of the important questions in this subsection is whether these cases should be considered as instances of so-called obligatory and non-obligatory control. This question has received a wide variety of answers in the literature depending on the definition of these notions. Our point of departure will be the operational definition in (386), which will be more extensively discussed in Subsection A on the basis of a number of standard English examples.

386

Obligatory control requires the antecedent of PRO to:

be overtly realized in the sentence containing PRO;

be local (a co-argument of the infinitival clause containing PRO);

be a c-commanding nominal argument (subject or object);

be unique (cannot be "split").

Object and subject control are illustrated in example in (387). Such examples are often considered as cases of obligatory control, subsections B and C will investigate these control constructions in more detail and argue that we are dealing with obligatory control in the sense of (386) only in appearance.

387

a.	Jani	beloofde	Elsj	[(om) PROi/*j	dat boek	te lezen].	subject control
	Jan	promised	Els	comp	that book	to read
	'Jan promised Els to read that book.'

b.	Jani	verzocht	Elsj	[(om) PROj/*i	dat boek	te lezen].	object control
	Jan	requested	Els	comp	that book	to read
	'Jan requested Els to read that book.'

According to the definition in (386), the examples in (388) are straightforward cases of non-obligatory control constructions: the PRO-subject in (388a) does not take a unique but a so-called split antecedent, which is constituted by both the subject and the object of the main clause, and in (388b) the antecedent does not have to be overtly realized, in which case PRO receives an arbitrary/generic interpretation. Cases like these will be discussed in Subsection D.

388

a.	Jani	stelde	Elsj	voor	[(om) PROi+j	samen	te werken].	split antecedent
	Jan	proposed	Els	prt.	comp	together	to work
	'Jan proposed to Els to collaborate.'

b.	Jani	keurt	het	af	[(om) PROarb	te vloeken].	arbitrary interpretation
	Jan	disapproves	it	prt.	comp	to curse
	'Jan disapproves of cursing.'

Our conclusion that we are not dealing with obligatory (that is, syntactically regulated) control in the examples in (388) raises the question as to what determines the type of control relation in om+ te-infinitivals; this question will be the main topic of Subsection E.
Before we start our discussion, we want to point out that the definition of obligatory control in (386) is not uncontroversial; since the distinction between obligatory and non-obligatory control was introduced in Williams (1980), it has given rise to a great deal of theoretical discussion and individual researchers have drawn the dividing line at different places; Bennis & Hoekstra (1989a), for example, claim that (386a-c) are not decisive for establishing obligatory control (and they in fact claim the same for anaphor binding but their judgments leading to this conclusion are not shared by all speakers; cf. Van Haaften 1991 and Petter 1998).
We also wish to point out that the extensive lists of control verbs (that is, verbs taking an infinitival complement with a PRO-subject) in the following discussion are based on those found in Van Haaften (1991) and Petter (1998), but adapted to the classification of verbs in Table 1, which was proposed in Section 1.2.2, sub II, and Chapter 2.

Table 1: Classification of verbs according to the type of nominal arguments they take

	name used in this grammar	external argument	internal argument(s)
no internal argument	intransitive: snurken'to snore'	nominative (agent)	—
	impersonal: sneeuwen'to snow'	—	—
one internal argument	transitive: kopen'to buy'	nominative (agent)	accusative (theme)
	unaccusative; arriveren'to arrive'	—	nominative (theme)
two internal arguments	ditransitive: aanbieden'to offer'	nominative (agent)	dative (goal) accusative (theme)
	nom-dat: bevallen'to please'	—	dative (experiencer) nominative (theme)
	undative: krijgen'to get'	—	nominative (goal) accusative (theme)

[+] A. Obligatory versus non-obligatory control

Obligatory control is normally assigned an operational definition; in order to be able to speak of obligatory control, the antecedent of PRO must at least satisfy the four restrictions in (386), repeated here as (389).

389

be overtly realized in the sentence containing PRO;

be local (a co-argument of the infinitival clause containing PRO);

be a c-commanding nominal argument (subject or object);

be unique (cannot be "split").

These properties of obligatory control will be illustrated by means of the English examples in (390) to (392). The examples in (390) show that the antecedent must be overtly realized in the sentence containing PRO; cf. Bresnan (1982) and Manzini (1983). Example (390a') shows that passivization, and the concomitant demotion of the subject, is impossible in subject control constructions, while example (390b'') indicates that omission of the nominal object is impossible in object control structures. We use the index "?" to indicate that this is due to there being no suitable controller available in the syntactic structure.

390

Johni promised Billj [PROi/*j to shave himselfi].

subject control

a'.

Billj was promised [PRO? to shave himself?].

a''.

Johni promised [PROi to shave himself].

Johni asked Billj [PROj/*i to shave himselfj].

object control

b'.

Billj was asked [PROj to shave himself].

b''.

Johni asked [PRO? to shave himself?].

That the antecedent of PRO must be a co-argument of the infinitival clause containing PRO can be illustrated by means of the examples in (391), which show that the unacceptable examples in (390) cannot be saved by embedding them in a larger sentence that does have a potential antecedent of PRO; since the antecedent must be within the clause headed by the subject/object control verbs to promise and to ask, the subjects of the main clauses headed by to think cannot function as such.

391

Johni thinks [that Billj was promised [PROi to shave himselfi]].

Billj thinks [that Johni asked [PROj to shave himselfj]].

That the antecedent of PRO must be a c-commanding nominal argument (subject or indirect object) is clear from the fact that the passive counterpart of (390a') does not improve when we add an agentive by-phrase: *Billj was promised by Johni [PROi to shave himselfi] is unacceptable because the antecedent of PRO is not a nominal argument of the matrix verb but part of the adverbial agentive by-phrase. Finally, the unacceptability of the examples in (392) shows that the antecedent of PRO must be unique in the sense that PRO cannot have a split antecedent.

392

Johni promised Billj [PROi+j to leave together].

Johni asked Billj [PROi+j to leave together].

It is normally assumed that obligatory control requires all four restrictions to be satisfied. The theoretical motivation is that obligatory control is comparable to binding of reflexive pronouns and NP-movement to subject position in passive, unaccusative and raising constructions. All of these exhibit properties of locally restricted syntactic dependencies are characterized by being obligatory (which derives property (389a)), local (which derives property (389b)), involve c-command (which derives property (389c)), and unique (which derives property (389d)); see Koster (1984a/1984b) for a more extensive discussion. Consequently, it is sufficient to show for just one of the restrictions in (389) that it does not hold in order to establish that we are dealing with non-obligatory control.

[+] B. Subject Control

By definition, subject control verbs must be minimally dyadic: they must have an infinitival argument clause as well as a subject that functions as the antecedent of the implied PRO-subject. This is consistent with the fact that subject control verbs are normally transitive or ditransitive verbs, or verbs taking a prepositional object clause. In (393), we give a small sample of transitive subject control verbs.

393

Transitive verbs: aandurven'to dare', aankunnen'to be up to', afzweren'to renounce', begeren'to desire', beogen'to aim at', bestaan'to have the nerve', doorzetten'to go ahead with', leren'to learn', durven'to dare', pogen'to try', nalaten'to refrain', ontwennen'to break oneʼs habit', overwegen'to consider', proberen'to try', popelen'to be eager', pretenderen'to pretend', schuwen'to shun', trachten'to try', vermijden'to avoid', verzuimen'to fail', wagen'to dare', weigeren'to refuse', uitproberen'to test', uitstellen'to postpone', verafschuwen'to abhor', verdienen to deserve', verdragen'to endure', verdommen'to flatly refuse', vergeten'to forget', verleren'to lose the hang of', vermijden'to avoid', vertikken'to refuse', verzaken/verzuimen'to neglect oneʼs duty', wagen'to dare', weigeren'to refuse'Transitive verbs: aandurven'to dare', aankunnen'to be up to', afzweren'to renounce', begeren'to desire', beogen'to aim at', bestaan'to have the nerve', doorzetten'to go ahead with', leren'to learn', durven'to dare', pogen'to try'</