Taalportaal - the digital language portal

Dutch1
Frisian
Saterfrisian
Afrikaans

Show all

13.2. A-Scrambling: nominal argument shift

quickinfo

Dutch allows a wide variety of word orders in the middle field of the clause. This subsection discusses the relative order of nominal arguments and clausal adverbs such as waarschijnlijk'probably'. All nominal arguments of the main verb may either precede or follow such adverbs, which is illustrated in (33) by means of a direct object and a subject. We will see that the word order variation in (33) is not free but restricted by information-structural considerations, namely the division between presupposition (discourse-old information) and focus (discourse-new information); cf. Van den Berg (1978), De Haan (1979) and Verhagen (1979/1986).

a.	Marie	wil	<het boek>	waarschijnlijk <het boek>	kopen.
	Marie	wants	the book	probably	buy
	'Marie probably wants to buy the book.'

b.	Morgen	zal	<die vrouw>	waarschijnlijk <die vrouw >	het boek kopen.
	tomorrow	will	that woman	probably	the book buy
	'Tomorrow that woman will probably buy the book.'

There are various analyses available for the word order variations in (33); see the reviews in the introduction to Corver & Van Riemsdijk (1994) and Broekhuis (2007/2008: Section 2.1). It has been claimed, for instance, that the orders in (33a) are not related to movement of the object. One version of this claim can be found in Neeleman (1994a/1994b), where it is claimed that both structures in (33) can be base-generated. We will refer to this as the flexible base-generation approach.

Marie wil [V' waarschijnlijk [V' dat boek kopen]]

Marie wil [V' dat boek [V' waarschijnlijk kopen]]

Another slightly more complex version of this claim is found in Vanden Wyngaerd (1988/1989), where it is claimed that the object obligatorily moves into a designated accusative case position, which is indicated in (35) as the specifier of XP. The word order variation is accounted for by assuming that the clausal adverb can be generated in different base-positions: it can be adjoined either to VP or to XP. We will refer to this as the flexible modification approach; see Booij (1974) for an earlier proposal with similar properties.

Marie wil [XP waarschijnlijk [XP het boeki X [VPti kopen]]]

Marie wil [XP het boeki X [VP waarschijnlijk [VPti kopen]]]

This section will opt for a movement analysis: we assume that the nominal arguments are generated to the right of the clausal adverb within the lexical domain of the clause but that they shift under certain conditions into a more leftward position in the functional domain to the left of the clausal adverbs.

Marie wil [VP waarschijnlijk [VP dat boek kopen]]

Marie wil [XP dat boeki X [VP waarschijnlijk [VPti kopen]]]

The details of this analysis, which we will refer to as the flexible movement approach, will be fleshed out in more detail in Subsection I; this subsection will also show that there are empirical reasons for preferring the flexible movement approach to the two alternative approaches. Subsection II discusses a concomitant effect of nominal argument shift on the intonation pattern of the clause: while non-shifted arguments can be assigned sentence accent, shifted arguments cannot. We will argue that this can also be used as an argument in favor of the flexible movement approach. Having thus firmly established that nominal argument shift is derived by movement, Subsection will argue that this movement is of the same type as found in, e.g., passive constructions: we are dealing with A-movement.

I. A flexible movement approach to nominal argument shift
II. Nominal argument shift and the location of sentence accent
III. Nominal argument shift is A-movement

[+] I. A flexible movement approach to nominal argument shift

This subsection provides a number of empirical arguments in favor of a flexible movement approach to nominal argument shift. Subsection A starts by arguing that object shift involves leftward movement: objects move into some landing site that is located higher than (that is, to the left of) the base-position of the subject; subjects move into the regular subject position right-adjacent to the complementizer/finite verb in second position (the specifier of TP). Subsection B continues by showing that the movement is restricted by the information structure of the clause: nominal argument shift only applies if the argument is part of the presupposition (discourse-old information) of the clause. Subsection C concludes by discussing a word order restriction on the output structures of nominal argument shift. Some of the issues addressed in the following subsections are discussed more extensively in Sections N8.1.3 and N8.1.4, but are briefly repeated here for convenience.

[+] A. Two empirical arguments in favor of the flexible movement approach

This subsection provides a review of two classical empirical arguments in favor of a movement analysis to nominal argument shift: Wat voor split and VP-topicalization.

[+] 1. Wat voor split

The standard argument in favor of a movement analysis of nominal argument shift is that placement of the nominal argument in front of the clausal adverb gives rise to a freezing effect. We demonstrate this in (37) by means of the so-called wat voor split. Example (37a) first shows that the string wat voor een boek can be fronted as a whole and should therefore be considered a phrase; the full string functions as a direct object. This, in turn, strongly suggests that the split in (37b) is derived by wh-extraction of wat from the wat voor-phrase. The acceptability contrast between the two (b)-examples shows that the wat voor split requires the remnant of the direct object to follow the modal adverb waarschijnlijk'probably'; cf. Den Besten (1985). If the word order difference between the (b)-examples is indeed related by leftward movement of the direct object across the clausal adverb, the unacceptability of (37b') can be accounted for by appealing to freezing: the wh-element wat has been extracted from a moved phrase.

a.	Wat voor een boek	zal	Marie waarschijnlijk	kopen?
	what for a book	will	Marie probably	buy
	'What kind of book will Marie probably buy?'

b.		Wat	zal	Marie waarschijnlijk	voor een boek	kopen?
		what	will	Marie probably	for a book	buy

b'.	*	Wat	zal	Marie voor een boek	waarschijnlijk	kopen?
		what	will	Marie for a book	probably	buy

Den Besten (1985) also claims that the wat voor split is categorically excluded for subjects of transitive verbs but Reuland (1985), Broekhuis (1987/1992), De Hoop (1992) and Neeleman (1994a) have shown that the split is possible if the subject is not in the regular subject position but occupies a position more to the right; this is clear from the fact that the split is possible if the regular subject position in (38b) is filled by the expletiveer, but not if the expletive is absent.

a.	Wat voor vogels	zullen	(er)	je voedertafel	bezoeken?
	what for birds	will	there	your bird.table	visit
	'What kind of birds will visit your bird table?'

b.	Wat	zullen	??(er)	voor vogels	je voedertafel	bezoeken?
	what	will	there	for birds	your bird.table	visit
	'What kind of birds will visit your bird table?'

This suggests that the subject is moved into the regular subject position from a more deeply embedded (more rightward) base-position in the clause. The introduction to this chapter has shown that in current generative grammar it is generally assumed that this base-position is the specifier of the light verb v, as indicated in (39).

[CP ... C [TP ... T [XP ... X [vP Subject v [VP ... V ... ]]]]]

Example (33b) has further shown that subject shift may cross the clausal adverb waarschijnlijk'probably', for which reason we have assumed that such adverbs demarcate the left boundary of the lexical domain (which is now taken to be the vP). If so, the movement of the object in (33a) also targets a position in the functional domain of the clause. A currently more or less standard assumption is that both types of nominal argument shift are motivated by case assignment: the subject and the object (optionally) move into the specifier of some functional head that is responsible for structural case assignment: T for nominative case and some functional head X for accusative case. The many different proposals concerning the nature of X need not concern us here; we will therefore not digress on what X is and refer the reader to Broekhuis (2008: Section 3.1) for a review of a number of recent proposals (including proposals that dispense with the category X altogether).

For completeness’ sake, we want to conclude the discussion of the wat voor split by pointing out that it is not clear whether freezing should really be held responsible for the unacceptability of example (37b') and example (38b) without the expletive er. The reason for this is that interrogative wat voor-phrases are non-D-linked and this may simply block object/subject shift; that nominal argument shift of wat voor-phrases is indeed impossible is strongly suggested by the sharp acceptability contrast between the two multiple wh-questions in (41).

a.	Wie zal	waarschijnlijk	wat voor boek	kopen?
	who will	probably	what for book	buy
	'Who will probably buy what kind of book?'

b.	*	Wie zal	wat voor boek	waarschijnlijk	kopen?
		who will	what for book	probably	buy

Notwithstanding this, the absence of a freezing effect in (37b) and (38b) with the expletive still supports the claim that remnants of wat voor-phrases should be located within the lexical domain of the clause, and hence also the claim that the subject and the object are base-generated within vP.

[+] 2. VP-topicalization

Another classic argument in favor of a movement analysis of nominal argument shift involves VP-topicalization; see De Haan (1979) and Webelhuth & Den Besten (1987/1990). Since nominal argument shift is optional, the analysis in (40) correctly predicts that VP-topicalization may either pied pipe or strand the direct object.

a.	Marie wil	<het boek>	waarschijnlijk <het boek>	kopen.
	Marie wants	the book	probably	buy
	'Marie probably wants to buy the book.'

b.		[VP	Het boek	kopen]	wil	Marie waarschijnlijk tVP.
		[VP	the book	buy	wants	Marie probably

b'.		[VPti	Kopen]	wil	Marie het boeki	waarschijnlijk.
		[VPti	buy	wants	Marie the book	probably

The analysis in (40) further accounts for the fact illustrated in (43) that VP-topicalization cannot strand the object in a position following the clause adverb, as there simply is no landing site for the object there; Section 13.3.2 will return to the fact that (43) is acceptable if the object is contrastively accented.

	*	[VPti	Kopen]	wil	Marie waarschijnlijk	het boekitVP.
	*	*[VPti	buy	wants	Marie probably	the book

It should be noted that the acceptability contrast between of (42b') and (43) is a problem for the flexible modification approach in (35), repeated here as (44), according to which the object is obligatorily moved into its case position, as this would allow us to derive both (42b') and (43) by means of VP-topicalization: the former can be derived from (44b) and the latter from (44a).

Marie wil [XP waarschijnlijk [XP het boeki X [VPti kopen]]]

Marie wil [XP het boeki X [VP waarschijnlijk [VPti kopen]]]

The acceptability contrast between (42b') and (43) also poses a serious problem for the flexible base-generation approach in (45) because topicalization is often claimed to involve maximal projections only; if so, (42b') and (43) are both predicted to be ungrammatical, as they can only be derived by movement of the verbal head in isolation. If we do allow V-topicalization, there still is a problem because we then wrongly predict both (42b') and (43) to be acceptable as there would be no a priori reason for assuming that (43) cannot be derived from (45a) by V-topicalization.

Marie wil [V' waarschijnlijk [V' het boek kopen]]

Marie wil [V' het boek [V' waarschijnlijk kopen]]

The flexible movement approach in (40) can also easily account for the fact that it is not possible to pied pipe clausal adverbs by pointing to the fact that these are not included in the lexical projection of the verb (that is, vP); cf. Section 8.4.

[Waarschijnlijk het boek kopen] wil Marie.

[Het boek waarschijnlijk kopen] wil Marie.

[Waarschijnlijk kopen] wil Marie het boek.

The flexible modification approach cannot account for the unacceptability of the examples in (46). The reason is that this approach can only account for the acceptability of the examples in (42b&b') by assuming that VP-topicalization can affect either XP or VP in (44). Consequently, it should be possible to derive example (46a) from (44a) by topicalization of the higher segment of XP, example (46b) from (44b) by topicalization of XP, and (46c) from (44b) by topicalization of the higher segment of VP. Even if we assumed that only the lower segments of XP and VP can be topicalized, the unacceptability of example (46b) would remain a problem. Similar problems arise for the flexible base-generation approach, as it should be possible to derive the examples (46a) and (46b) from, respectively, (45a) and (45b) by topicalization of the higher segments of V', and (46c) from (45b) by topicalization of the lower segment of V'. Even if we assume that only the lower segments of V' can be topicalized, an option that should be allowed in order to make it possible to derive example (42b) from (45a), the unacceptability of (46c) would remain a problem. We conclude from this that the flexible modification and the flexible base-generation approach can only account for the unacceptability of the examples in (46) by appealing to ad hoc restrictions on what can or cannot be topicalized; the VP-topicalization data thus favor the flexible movement approach.

[+] B. Information-structural restrictions on nominal argument shift

Example (33a), repeated here as (47a), shows that the direct object het boek'the book' may either precede or follow the clausal adverb waarschijnlijk'probably'. Although this suggests that object shift is optional, the examples in (47b&c) show that this is not always correct: indefinite direct objects must follow while definite object pronouns must precede the clausal adverb.

a.	dat	Marie	<het boek>	waarschijnlijk <het boek>	koopt.
	that	Marie	the book	probably	buys
	'that Marie will probably buy the book.'

b.	dat	Marie	<*een boek>	waarschijnlijk <een boek>	koopt.
	that	Marie	a book	probably	buys
	'that Marie will probably buy a book.'

c.	dat	Marie	<het>	waarschijnlijk <*het >	koopt.
	that	Marie	it	probably	buys
	'that Marie will probably buy it.'

In fact, the two orders in (47a) are not always equally felicitous either. The order in which the direct object precedes the clausal adverb is normally used if the referent of the noun phrase is already part of the domain of discourse; cf. Verhagen (1986). This is illustrated by the question-answer pair in (48): due to the fact that the direct object was already introduced as a discourse topic in question (48a), it precedes the adverb in answer (48b). Note that we abstract away from the fact that there is an even better way of answering question (48a): by substituting the pronoun het'it' for the noun phrase het boek'the book'.

a.	Wat	doet	Marie	met het boek?	question
	what	does	Marie	with the book
	'What is Marie doing with the book?'

b.	Ik denk	dat	ze	<het boek>	waarschijnlijk <#het boek>	koopt.	answer
	I think	that	she	the book	probably	buys
	'I think that sheʼll probably buy the book.'

When uttered out-of-the-blue, a question such as (49a) requires an answer in which the direct object provides new information and follows the clausal adverb; the order in which the object precedes the adverb is possible only if the referent of the direct object is already part of the domain of discourse, for example, when the speaker and the addressee are discussing Jan’s wish list, which includes a specific book title.

a.	Wat	koopt	Marie voor Jan?	question
	what	buys	Marie for Jan
	'What will Marie buy for Jan?'

b.	Ik denk	dat	ze	<#het boek>	waarschijnlijk <het boek>	koopt.	answer
	I think	that	she	the book	probably	buys
	'I think that sheʼll probably buy the book.'

The discussion above shows that direct objects preceding the clausal adverb refer to discourse-old information, whereas direct objects following the clausal adverb refer to discourse-new information. Since definite pronouns and indefinite noun phrases typically refer to, respectively, discourse-old and discourse-new information, their placement relative to the clausal adverb in the examples in (47) follows naturally. Another fact that follows naturally from this information-structural restriction on argument placement is that epithets always precede clausal adverbs; they always refer to an active discourse topic.

dat	Jan <de etter>	waarschijnlijk <*de etter>	haat.
that	Jan the son.of.a.bitch	probably	hates
'that Jan probably hates the son of a bitch.'

It should be noted, however, that the notion of discourse-new information should be taken quite broadly in that it is not confined to the referential properties of the noun phrase. An example illustrating this, inspired by Verhagen (1986:106ff.), is given in (51). Although the referent of the noun phrase de verkeerde'the wrong person' is clearly identifiable for both participants, the neutral continuation of the discourse is as given in (51b): this is due to the fact that Peter is now characterized as "the wrong person to give the relevant information to". Note in passing that example (51b') is possible with a contrastive accent on the noun phrase, in which case this utterance is likely to be followed by another one revealing the identity of the person that should have been informed.

a.	Ik	heb	het	aan Peter	verteld.	speaker A
	I	have	it	to Peter	told
	'I have told it to Peter.'

b.	Dan	heb	je	waarschijnlijk	de verkeerde	ingelicht.	speaker B
	then	have	you	probably	the wrong.one	prt.-informed
	'Then you have probably informed the wrong person.'

b'.	*	Dan	heb	je	de verkeerde	waarschijnlijk	ingelicht.	speaker B
		then	have	you	the wrong.one	probably	prt.-informed

The examples in (52) show that subjects behave in essentially the same way as the objects in (47); cf. Van den Berg (1978). This is slightly obscured, however, by a complicating factor, namely that indefinite subjects may precede the clausal adverb if they are interpreted as specific (known to the speaker but not to the addressee) or if they are part of a generic sentence. We will ignore this here but return to the distinction between specific and non-specific indefinite subjects in Subsection C.

a.	dat	<die vrouw>	waarschijnlijk <die vrouw>	het boek koopt.
	that	that woman	probably	the book buys
	'that that woman will probably buy the book.'

b.	dat	<#een vrouw>	waarschijnlijk <een vrouw>	het boek koopt.
	that	a woman	probably	the book buys
	'that a woman will probably buy the book.'

c.	dat	<ze>	waarschijnlijk <*ze>	het boek koopt.
	that	she	probably	the book buys
	'that sheʼll probably buy the book.'

The discussion above has shown that the relative order of the object/subject and the clausal adverb is sensitive to the information-structural function of the object/subject. This favors an approach in which the restriction on word order is formulated in terms of properties of the subject/object and thereby again disfavors the flexible modification approach in (35), according to which the word order variation is due to alternative placements of the adverb. The flexible modification approach also runs up against a contradiction concerning the placement of clausal adverbs relative to the regular subject position, the specifier of TP. Consider the expletive constructions in (53). If we adopt the standard assumption that the expletive er occupies the regular subject position, which is corroborated by the fact that it is right-adjacent to the complementizer dat, the acceptability contrast between the two examples in (53) shows that clausal adverbs must follow this subject position.

a.	dat	er	waarschijnlijk	een man	op straat	loopt.
	that	there	probably	a man	in.the.street	walks
	'that there is probably a man walking in the street.'

dat waarschijnlijk er een man op straat loopt

The conclusion that clausal adverbs cannot be located in front of the regular subject position makes it very unlikely that the order variation in (52a) can be accounted for by assuming variable base-positions for the modal adverb, as suggested by the line of reasoning found in Vanden Wyngaerd (1989): if the subject is to occupy the regular subject position in order to receive nominative case, the order in an example such as dat waarschijnlijk die man op straat loopt'that that man is probably walking in the street' would imply that the clausal adverb can precede the regular subject position, contrary to fact, as shown by (53b). The resulting contradiction does not arise if we assume subject shift; see Broekhuis (2009b) and Vanden Wyngaerd (2009) for more discussion.

[+] C. Interaction of different types of argument shift

If we adopt the claim that nominal argument shift targets a position in the functional domain of the clause where the subject/object can be assigned case, we can summarize the findings from Subsection B as in (54); see Broekhuis (2008:ch.3), De Hoop (1992:ch.3) and Delfitto & Corver (1998) for somewhat different implementations of the same idea.

Nominal arguments expressing discourse-new information stay within the lexical domain.

Nominal arguments expressing discourse-old information move into their case position in the functional domain of the clause.

Now consider again the derivation suggested in (40), repeated here as (55). This derivation, in tandem with the two generalizations in (54), predicts that an object expressing discourse-old information will cross a subject that expresses discourse-new information.

Although this prediction is more or less accurate for languages like German, it is clearly wrong for Standard Dutch, since in the middle field of the clause the subject normally precedes the direct object, as stated in the restriction on linear word order in (56): see, e.g., De Haan (1979:ch.4), Haegeman (1993a/1995), Williams (2003) and Müller (2000/2001) for extensive discussion of this restriction.

Ordering restriction on nominal argument shift in Standard Dutch: nominal argument shift does not affect the unmarked order of the nominal arguments (agent > goal > theme).

The word order restriction in (56) can only operate in full force if one of the generalizations in (54) is violated. The discussion in the following subsections will show that this is indeed what we find; cf. Broekhuis (2008/2009).

[+] 1. Direct object and subject shift

Example (57a) shows again that definite subjects may be located to the right of clausal adverbs like waarschijnlijk if they are part of the focus of the clause, that is, refer to discourse-new information. The examples in (57b&c) show that the subject and the object can both shift to the left of the clausal adverb provided they are part of the presupposition of the clause. The effect of the ordering restriction on nominal argument shift in (56) is illustrated by (57d); this example shows that a presuppositional object cannot shift across the subject if the latter is part of the focus of the clause and thus has to follow the modal adverb. This means that example (57a) is information-structurally ambiguous in that it also allows the direct object to be part of the presupposition of the clause; since the discourse-old object occupies a position within the lexical domain, this results in a violation of restriction (54b).

a.	dat	waarschijnlijk	de jongens	dit boek	gelezen	hebben.
	that	probably	the boys	this book	read	have
	'that the boys have probably read this book.'

dat de jongens waarschijnlijk dit boek gelezen hebben.

dat de jongens dit boek waarschijnlijk gelezen hebben.

dat dit boek waarschijnlijk de jongens gelezen hebben.

The results are different if we replace the direct object dit boek'this book' by the pronoun het'it'. Example (58a) first shows that the object pronoun differs from non-pronominal objects in that it cannot remain within the lexical domain of the clause if the subject is part of the focus of the clause. Example (58b) shows that it behaves like non-pronominal objects in that it cannot cross the subject, but (58c) shows that it differs from non-pronominal objects in that it is able to push the subject up into the regular subject position of the clause. This means that the subject in (58c) can be interpreted as referring to discourse-new information in violation of the restriction in (54a), as is clear from the fact that this example can be used as an answer to the question Wie hebben het boek gelezen?'Who have read the book?'.

a.	*	dat	waarschijnlijk	de jongens	het	gelezen	hebben.
		that	probably	the boys	it	read	have

b.	*	dat	het	waarschijnlijk	de jongens	gelezen	hebben.
		that	it	probably	the boys	read	have

c.	dat	de jongens	het	waarschijnlijk	gelezen	hebben.
	that	the boys	it	probably	read	have
	'that the boys probably have read it.'

It should further be noted that examples such as (58a) become fully acceptable if the subject is given contrastive stress; this shows that in such cases the subject may block object shift of the pronominal object in violation of the information-structural restrictions in (54b); we refer the reader to Section 13.3 for a discussion of the placement of contrastively focused phrases.

dat	waarschijnlijk	de jongens	het	gelezen	hebben.
that	probably	the boys	it	read	have
'that the boys have probably read it.'

We find more or less the same pattern with indefinite subjects. The situation is somewhat complicated, however, by the fact that, depending on its placement with respect to the clausal adverb, the subject can receive a non-specific interpretation (unknown to speaker and hearer) or a specific interpretation (known to the speaker but unknown to the hearer); if the indefinite subject twee jongens in (60) follows the clausal adverb waarschijnlijk'probably', it is preferably interpreted as non-specific, while twee jongens can only be interpreted as specific if it precedes waarschijnlijk (see also N8.1.4, sub I).

a.	dat	waarschijnlijk	twee jongens	dit boek	gelezen	hebben.
	that	probably	two boys	this book	read	have
	Ambiguous: 'that two (of the) boys have probably read this book.'

b.	dat	twee jongens	waarschijnlijk	dit boek	gelezen	hebben.
	that	two of the boys	probably	this book	read	have
	Specific only: 'that two of the boys have probably read this book.'

The result changes again if we replace the direct object het boek'the book' by the pronoun het'it'. Placement of the indefinite subject after the clausal adverb, as in (61a), again requires the subject to be assigned contrastive stress; in case of a more neutral intonation pattern the pronoun pushes the subject up into the regular subject position right-adjacent to the complementizer, as in (61b). The fact that the subject in (61b) may provide discourse-new information again violates the information-structural restriction in (54a), and the fact that the contractively stressed subject in (61a) is able to block object shift of the pronoun violates the restriction in (54b).

a.	dat	waarschijnlijk	twee jongens/*jongens	het	gelezen	hebben.
	that	probably	two boys	it	read	have
	'that two boys (not girls) have probably read it.'