• Dutch
  • Frisian
  • Afrikaans
Show full table of contents
11.3.7Parasitic gaps

Wh-questions normally exhibit a one-to-one correspondence between wh-moved phrases and their traces, subsection I below will show that in prototypical cases such as (506a) traces are bound by a unique wh-moved phrase; the wh-phrase welk e boek en functions as the antecedent of the object gap indicated by the trace t. An example such as (506b) is an (apparent) exception to this otherwise robust generalization: the wh-phrase seems to function as the antecedent of both the object gap in the main clause and the object gap in the adverbial clause zonder te lezen. The formal linguistic literature refers to the interpretative gap in the adverbial clause as parasitic gap (pg) for reasons that will become clear in subsection II.

Example 506
a. Welke boeken heeft Jan ti opgeborgen?
  which books  has  Jan  prt.-filed
  'Which books has Jan filed?'
b. Welke boeken heeft Jan [zonder pgi te lezen] ti opgeborgen?
  which books  has  Jan  without  to read  prt.-filed
  'Which book has Jan filed without reading?'

As parasitic gap constructions have been studied on the basis of English data especially, Subsection II introduces the notion of parasitic gap on the basis of a small number of English examples. This will result in a set of five restrictions that are commonly assumed to be applicable to them. These restrictions will be taken as the starting point of our discussion of Dutch parasitic gap constructions in Subsection III.

[+]  I.  The bijection principle

One of the hallmarks of w h-movement is that wh-phrases in clause-initial position are associated with a more deeply embedded interpretative gap, as indicated by the structures in (507a&b). Such structures can be used as input for the semantic component of the grammar and be translated into semantic representations with a question operator and a variable, as in the primed examples.

Example 507
a. Wiei heeft Peter/hij ti vandaag bezocht?
  who  has  Peter/he  today  visited
  'Who did Peter/he visit today?'
a'. ?x (Peter/he visited x today)
b. Wiei heeft ti Jan/hem vandaag bezocht?
  who  has  Jan/him  today  visited
  'Who visited Jan/him today?'
b'. ?x (x visited Jan/him today)

There are several conditions on operator-variable representations in natural language that are not assumed for their counterparts in formal-logical systems. For example, while formal-logical systems allow vacuous quantifiers, that is, quantifiers that do not bind a variable, natural language does not. This can be seen as the result of a more general economy condition on natural language which prohibits superfluous elements in a representation: sentence (508a) is unacceptable despite the fact that a semanticist may consider its formal semantic counterpart in (508b) impeccable; cf. Chierchia & McConell-Ginet (1992:110).

Example 508
a. * Wie heeft Peter/hij Jan/hem vandaag bezocht?
  who has  Peter/he  Jan/him  today  visited
b. ?x (Peter/he visited Jan/him today)

Since a variable must be bound by an operator in order to form an interpretable sentence, the fact that the examples in (509) are uninterpretable does not come as a surprise; we will ignore the fact here that we do find constructions like (509a) in certain (e.g., generic) contexts that allow an implied theme argument and with pseudo-intransitive verbs, that is, verbs that take a cognate object.

Example 509
a. * Peter/hij heeft [e]object vandaag bezocht.
  Peter/he  has  today  visited
b. * Vandaag heeft [e]subject Jan/hem bezocht.
  today  has  Jan/him  visited

W h-moved phrases further differ from semantic operators in that they can normally bind a single interpretative gap at most: a sentence like $Wie heeft onderzocht?'Who has examined?' cannot be assigned the meaning indicated by the well-formed semantic representation in (510b); the only way to express this meaning is by using a reflexive pronoun: Wie heeft zichzelf onderzocht?'Who has examined himself?'.

Example 510
a. * Wiei heeft ti [e]object onderzocht?
  who  has  examined
b. ?x (x has examined x)

Koopman & Sportiche (1982) account for the observations above by postulating that natural language is subject to the bijection principle in (511); the specific phrasing of the principle is taken from Webelhuth (1992:143).

Example 511
Bijection principle
a. Every syntactic operator binds exactly one syntactic variable.
b. Every syntactic variable is bound by exactly one syntactic operator.
[+]  II.  Some characteristic properties of parasitic gaps

This subsection discusses an (apparent) problem for clause (511a) of the bijection principle in the sense that a single wh-phrase is related to more than one interpretative gap. Such cases have been studied intensively for English since Engdahl's (1983) seminal paper on this issue, but has received less attention in other languages. We will therefore introduce the notion of parasitic gap gap by using English examples. The results can then be used as a starting point for our description of Dutch in Subsection III. The discussion below is based on the more extensive review found in Culicover (2001).
      A standard example of a parasitic gap construction from English is (512a); given that the two interpretative object gaps are translated as variables bound by the same question operator in the informal semantic representation in (512b), this example seems to violate clause (511a) of the bijection principle.

Example 512
a. Which articlesi did John file ti [without reading pgi]?
b. ?x (x:articles) (Jan filed x without reading x)

The use of a trace in the object position of the main clause in (512a) is motivated by the fact that it can be independently established that wh-movement is possible from this position; cf. Which articlesi did John file ti? The reason for usingthe notion parasitic gap (pg) for the interpretative gap in the adverbial phrase is twofold. First, example (513a) shows that it cannot be a trace left by wh-movement of who, as adverbial clauses are islands for wh-extraction. Second, example (513b) shows that it cannot occur if the direct object of the main clause occurs in its base-position; the gap is thus "parasitic" on wh-movement of this phrase.

Example 513
a. * Whoi did John file the articles [without consulting ti]?
b. John filed a bunch of articles [without reading them/*pg)].

Parasitic gap constructions are not limited to wh-questions but also occur in other constructions derived by wh-movement. This is illustrated in example (514a) for a relative clause; examples (514b&c) show that wh-movement of the phonetically empty relative pronoun OP is possible from the object position of the relative clause but not from the object position of the adverbial clause. Note in passing that Engdahl assigns (514c) a question mark, while we use an asterisk: this is because Culicover (2001) simply calls this example ungrammatical.

Example 514
a. Here is the paperi [OPi that John read ti [before filing pgi]].
b. Here is the paperi [OPi that John read ti [before filing his mail]].
c. * Here is the paperi [OPi that John read his mail [before filing ti]].

Culicover (2001) provides a number of properties of parasitic gap constructions that are generally accepted, while noting that these claims have all been challenged in the literature at some point. An adapted version of his list is given as (515).

Example 515
Restrictions on English parasitic gap constructions
a. Landing-site restriction: antecedents of parasitic gaps are in an A'-position.
b. Overt-movement restriction: antecedents of parasitic gaps are overtly moved.
c. Anti-c-command restriction: the trace of the antecedent of the parasitic gap and the parasitic gap do not c-command each other.
d. Categorial restriction: antecedents of parasitic gaps are noun phrases.
e. Multiple-island restriction: parasitic gaps and their antecedents cannot be separated by more than one island boundary.

Landing-site restriction (515a) refers to the fact that parasitic gap constructions typically occur in constructions derived by wh-movement; the English examples given above illustrate this point. This has led to the claim that the antecedent of the trace and the parasitic gap cannot be in an A-positions (that is, argument positions to which thematic roles, agreement features and/or case are assigned) but must be in an A'-position, which may account for the fact that parasitic gaps may also occur in, e.g., English heavy NP-shift constructions. We will see, however, that this claim is not generally accepted for Dutch parasitic gap constructions.
      The overt-movement restriction in (515b) is based on the standard generative assumption from the 1980's that wh-elements in situ undergo covert movement, that is, movement after the structure has been transferred to the phonological component of the grammar. Although this claim is no longer accepted by many generative linguists, the empirical issue still remains, which is that parasitic gaps cannot be licensed by wh-phrases occupying their base position; the wh-phrase which article in multiple question (516) does not license a parasitic gap. For convenience we will maintain the notion of overt-movement restriction without implying a specific stance on the issue of covert movement.

Example 516
* Who filed which articles [without reading pgi]?

The anti-c-command restriction in (515c) on the relation between the wh-trace and the parasitic gap can be derived from binding condition C, which forbids referential expressions to be A-bound, that is, to take a c-commanding antecedent in an argument position. This is done by extending to parasitic gaps the standard claim that wh-traces of nominal arguments exhibit the same binding behavior as referential expressions. The anti-c-command restriction can be used to account for the fact that subject traces block parasitic gaps more deeply embedded in their own clause, as illustrated by (517a&b), while traces left by wh-extraction of a subject from an embedded clause do not block parasitic gaps in matrix clauses, as illustrated by (517c). The examples are taken from Engdahl (1983) and Chomsky (1986); we will discuss a problem for the claim that wh-traces and parasitic gaps are subject to binding condition C in Subsection IIID, but we will accept this claim as a working hypothesis for what follows.

Example 517
a. * Which articlesi [ti got filed by John [without him reading pgi]]?
b. * Whoi [ti met you [before you recognized pgi]]?
c. Which papersi did John decide [before reading pgi] to tell his secretary [twere unavailable]?

The anti-c-command restriction also predicts the acceptability of examples like (518a&b), which are again taken from Engdahl (1983) and Chomsky (1986). It also accounts for the fact illustrated in the primed examples that substituting a simple gap for the complex noun phrase a picture of pg i is impossible: because the two gaps are both A'-bound by the wh-phrase in clause-initial position and the first gap c-commands the second, the second gap is incorrectly A-bound by the first gap. Note that on the assumption that nominal wh-traces and parasitic gaps are both subject to binding condition C, this result follows regardless of whether the first or the second gap is considered to be parasitic on wh-movement; we therefore did not specify the nature of the gaps in the primed examples.

Example 518
a. Which girli did you show [a picture of pgi] to ti?
a'. * Which girli did you show [ei] to [ei]?
b. Whoi would [a picture of pgi] surprise ti?
b'. * Whoi would [ei] surprise [ei]?

It should be noted that we can only maintain the anti-c-command restriction if we assume that the direct objects in (512)-(514) do not c-command the adjuncts containing the parasitic gaps. This assumption is consistent with the fact that complements are generated as the immediate sister of the selecting verb, but inconsistent with the c-command hierarchy that we introduced in Section N5.2.1.5, sub III; we refer the reader to the discussion of this issue in Contreras (1984), Koster (1987: Section 6.4) and Safir (1987), and to Lasnik (1999:ch.6) for a specific approach to English objects that may solve this problem.
      Categorial restriction (515d), according to which the wh-moved phrase must be nominal, has been claimed not to be cross-linguistically valid but can at least be seen as a strong tendency in English: wh-movement of APs or PPs normally does not license parasitic gaps. Two examples adapted from Cinque (1990:115) are given in (519); see Koster (1987:156-7) for more examples.

Example 519
a. * How tiredi can one feel ti [without being pgi]?
b. * [The man [to whomi I went ti [without speaking pgi]]] is there.

The examples above have shown that parasitic gaps are typically found in islands for wh-extraction, such as the adjuncts in (512) and (514) or the subject in (518b). Kayne (1984:ch.8) and Contreras (1984) have noted, however, that parasitic gaps cannot be embedded in islands within an island, as stated by the multiple-island restriction in (515e). This is illustrated by the contrasts in acceptability indicated in (520) and (521), in which the two (b)-examples should be construed as alternative realizations of the adverbial clauses in the (a)-examples, and the abbreviation OP again indicates the phonetically empty relative pronoun.

Example 520
a. the person [OPi that John described ti [adjunct ...]]
b. ? [adjunct without examining [object any pictures of pgi]].
b'. * [adjunct without [subject any pictures of pgi] being on file].
Example 521
a. the paper [OPi that we should destroy ti [adjunct ...]]
b. ? [adjunct before someone steals [object a copy of pgi]].
b'. * [adjunct before [subject a copy of pgi]] gets stolen by someone].

      Kayne detects a "sharp contrast" between the two alternative realizations of the adjunct clauses and attributes this to the fact that the parasitic gaps are embedded in a single (adjunct) island in the primeless (b)-examples but in two islands in the primed examples, an adjunct island and an additional subject island.
      Now that we have briefly discussed the five restrictions in (515), we conclude our brief survey of English parasitic gaps by noting that Engdahl (1983) has found a great deal of variation in speakers' judgments on parasitic gap constructions. Furthermore, it seems that the acceptability of parasitic gap constructions depends on the phrases they are embedded in; parasitic gaps in non-finite clauses such as (520b) are more likely to be accepted by speakers than parasitic gaps in finite clauses such as (521b). Or, stated somewhat differently, speakers who accept parasitic gaps in finite adjunct clauses such as (521b) will also accept them in non-finite adjunct clauses such as (520b), while the inverse does not necessarily hold. Engdahl's hierarchy is given in a shorter and slightly adapted form as (522): it expresses that parasitic gaps are best in infinitival adjunct clauses, somewhat less favored in finite argument/adjunct clauses, and least favored in relative clauses.

Example 522
Accessibility hierarchy for occurrences of parasitic gaps (simplified): infinitival adjunct clause > finite argument clauses > finite adjunct clauses > relative clauses
[+]  III.  Parasitic gaps in Dutch

Since Dutch parasitic gap constructions have received relatively little attention and since it is sometimes quite difficult to extract acceptability judgments from non-linguistic speakers, some of the acceptability judgments on the data below rely on our own intuitions; moreover, the attested variation in judgments implies that not all Dutch speakers will accept the judgments given here or elsewhere in the literature. The main point is, however, that many speakers do have the indicated contrasts between the examples in each set of examples. The reader is therefore requested to interpret the judgments as statements about the relative acceptability of the examples in each given set (which actually also holds for all other judgments provided in this work). The following subsections deal with parasitic gap constructions we find or do not find in Dutch by means of a discussion of the five generalizations in (515).

[+]  A.  The landing-site and overt-movement restriction in (515a&b)

Dutch and German data have given rise to an ardent debate about the landing-site restriction in (515a); this is related to the fact that parasitic gaps are not only licensed by wh-moved but also by scrambled phrases. It should be noted, however, that the debate is not only about the landing-site restriction as such, as it is intertwined with a much broader debate about the nature of scrambling: is it A- or A'-movement, or is it something totally different? In order to separate the two issues, we start by discussing some core data on parasitic gaps; this discussion will also touch upon the overt-movement restriction in (515b). After that, we continue with a brief discussion on the nature of scrambling, an issue discussed more extensively in Chapter 13. We will then introduce a test, based on binding, that can be used for discriminating between A- and A'-movement, which will be used in a more detailed discussion of the problematic scrambling data. Since we will see that there is no decisive argument against it, we will provisionally conclude that the landing-site restriction also applies to Dutch parasitic gap constructions. This does not imply that there are no problems left for this restriction, which we will demonstrate on the basis of passivized parasitic gap constructions.

[+]  1.  Some data

Landing-site restriction (515a) correctly predicts that wh-moved phrases may serve as antecedents of parasitic gaps. This is illustrated in (523) for a wh-question, a topicalization construction, and a relative clause.

Example 523
a. Welke boekeni heeft Jan [zonder pgi te lezen] ti opgeborgen?
  which books  has  Jan  without  to read  prt.-filed
  'Which books has Jan filed without reading?'
b. Deze boekeni heeft Jan [zonder pgi te lezen] ti opgeborgen?
  these books  has  Jan  without  to read  prt.-filed
  'These books, Jan has filed without reading.'
c. [De boeken [diei Jan [zonder pgi te lezen] ti opgeborgen heeft]] zijn weg.
  the books  which  Jan without  to read  prt.-filed  has  are  gone
  'The books that Jan has filed without reading are missing.'

The overt-movement restriction in (515b), on the other hand, does not seem to hold for Dutch as the multiple wh-question in (524a) is fully acceptable. The situation is, however, more complex than it seems at first sight, as (524b) is unacceptable.

Example 524
a. Wie heeft welke boekeni [zonder pgi te lezen] opgeborgen?
  who  has  which books   without  to read  prt.-filed
b. * Wie heeft [zonder pgi te lezen] welke boekeni opgeborgen?
  who  has  without  to read  which books  prt.-filed

Since the position of the object in (524b) is taken to be its base position within the VP, we may assume that this is the construction that resembles the English multiple wh-question in (516) most closely. It seems that (524a) is derived from this structure by means of leftward movement of the object into some structurally higher position; more precise representations of the examples in (524) are thus as indicated in (525).

Example 525
a. Wie heeft welke boekeni [zonder pgi te lezen] [VPti opgeborgen]?
  who  has  which books   without  to read  prt.-filed
b. * Wie heeft [zonder pgi te lezen] [VP welke boekeni opgeborgen]?
  who  has  without  to read  which books  prt.-filed

The leftward movement of the object in (525a) is known as scrambling, and the non-interrogative counterparts of the examples in (525) given in (526) show that scrambling is indeed able to license parasitic gaps; cf. Bennis & Hoekstra (1984).

Example 526
a. Jan heeft die boekeni [zonder pgi te lezen] [VPti opgeborgen]?
  Jan has  those books   without  to read  prt.-filed
  'Jan has filed these books without reading them.'
b. * Jan heeft [zonder pgi te lezen] [VP die boekeni opgeborgen]?
  Jan has  without  to read  those books  prt.-filed

The contrast between the (a)- and (b)-examples of (525) and (526) would follow from the landing-site and overt-movement restriction in (515a&b) if scrambling were an instance of A'-movement. The following subsection will show, however, that this is not easy to determine and that much depends on the specific version of the overall theory adopted.
      The examples in (527) illustrate again that antecedents of parasitic gaps can be scrambled or wh-moved phrases. These examples also show that parasitic gaps easily alternate with overt referential personal pronouns if their antecedent is a scrambled phrase (see, e.g., Bennis & Hoekstra 1984 and Huybregts & Van Riemsdijk 1985), but that this is harder if the antecedent is interrogative; this holds especially if the wh-phrase is non-D-linked, which is the prototypical use of the interrogative pronoun wat'what (although it sometimes can get a D-linked reading in specific contexts which will be ignored here)'.

Example 527
a. Jan heeft het boeki/heti [zonder pgi/heti te bekijken] ti weggelegd.
  Jan has  the book/it  without  pg/it  to look.at  away-put
  'Jan has put the book/it away without looking at it.'
b. Welke boeki heeft Jan [zonder pgi/?heti te bekijken] ti weggelegd?
  which book  has  Jan  without  pg/it  to look.at  away-put
  'Which book has Jan put away without looking at (it)?'
b'. Wati heeft Jan [zonder pgi/*heti te bekijken] ti weggelegd?
  what  has  Jan without  pg/it  to look.at  away-put

To our knowledge the contrasts in acceptability between the three types of example in (527) has not been observed before. It seems plausible to relate the differences to the degree of referentiality of the antecedents of the parasitic gap; referential noun phrases and pronouns obviously have a high degree of referentiality, while D-linked wh-phrases like welke boeken'which books' and non-D-linked wh-pronouns like wat'what' have an intermediate and a low degree of referentiality, respectively.

[+]  2.  A theoretical intermezzo: scrambling and A- and A'-movement

The term scrambling refers to the fact that in certain languages the word order of constituents may vary, and for Dutch and German it is normally used to refer to certain changes in the word order of the middle field of the clause. The notion is somewhat misleading, however, as it suggests that it refers to a single operation with well-defined properties. Chapter 13 will show, however, that there are various types of operation with quite different properties that may affect the word order of the middle field of the clause: some have properties of A-movement while other have properties of A'-movement. But even if we restrict the notion of scrambling to leftward movement of nominal arguments (that is, subjects and objects), it is very difficult to determine definitively what type of movement we are dealing with, as this is closely related to the overall theory that one adopts. This subsection contains a brief theoretical digression in order to illustrate this.
      The notion argument position (A-position) denotes positions in the clause that can be occupied by arguments of the verb only. Such positions are characterized by the fact that they can be assigned specific syntactic features, the three main types of which are: thematic roles, structural case and nominal agreement features (person, number, and gender). Prototypical A-positions are the subject and the object position. The notion non-argument position (A'-position) denotes positions that can also be occupied by non-arguments (adverbial phrases, etc.). Such positions function as landing sites for elements with a specific logico-semantic role (such as operator or negation) or an information-structural function (topic, focus, etc.); a prototypical A'-position is the clause-initial position that can be filled by any clausal constituent as a result of wh-movement.
      The number of A- and A'-positions postulated in generative grammar has increased considerably over the years. As for A-positions for nominal arguments of verbs, there were only two positions available in the early 1980's: the object and the subject position in the simplified structure in (528a). The object position within VP is the position to which the thematic role of theme, accusative case and (for languages that exhibit object agreement) object agreement features can be assigned; the subject position is the position to which the thematic role of agent, nominative case and the subject agreement features can be assigned. Arguments can sometimes also pick up their features in different places; in the unaccusative construction in (528b) the subject John is base-generated in the object position, where it is assigned the thematic role theme, and subsequently moved into the subject position, where it is assigned nominative case and the subject agreement features.

Example 528
a. [S John T(ense) [VP buys the book]].
b. [S Johni T(ense) [VPti leaves]].

Given that the object and subject positions exhaust the A-positions postulated it is a virtual necessity to assume that scrambling is A'-movement targetings some A'-position in the middle field of the clause. It is therefore not surprising that an early article such as Bennis and Hoekstra (1984) arrives at this conclusion.
      The fact illustrated in (528b) that the syntactic features of a certain argument can be scattered over more than one position within the clause has ultimately given rise to the hypothesis that there is a one-to-one relationship between features and positions. For example, instead of assuming that all features for the direct object are generated in a single position, it is now generally assumed that these are assigned by different functional heads like those indicated by capitals in (529) to their complement or specifier: the main verb assigns the role theme, the AGR-head assigns the agreement features and the CASE-head assigns accusative case. Something similar is assumed for subjects. Note that the names used in (529) for these functional heads are just randomly chosen. given that a large number of implementations of the main idea can be found in the literature since Pollock's (1989) seminal paper on this issue.

Example 529
[XP [accusative] CASE [AGRP [person, number, gender] AGR [VP V theme]]]

Since all A-positions in (529) are potential landing sites for the theme argument, it will be clear that the number of potential A-movements in the derivation of sentences has vastly increased compared to the earlier proposal in (528); the same holds in fact for verb movement, as all functional heads in (529) are assumed to be potential landings sites for the verb. This makes it possible to analyze scrambling of nominal arguments as A-movement, the position taken in Broekhuis (2008/2011), who argues that the theme position in (529) is cross-linguistically the base position of the object, that the agreement features are located in the object position preceding the verb in clause-final position (which in earlier versions of the theory was considered to be the base position of the object in Dutch), and that scrambling of the object targets the accusative position.
      Since the seminal work by Haegeman & Zanuttini (1991), Haegeman (1995) and Rizzi (1996/1997), there has also been a proliferation of A'-positions; while in the early 1980's there was just one clearly defined A'-position, the landing site of wh-movement, more recent research claims to have identified a large number of additional A'-positions in structurally lower positions, which can be targeted by negative, focused, topical, quantified phrases, etc. Again, this makes it possible to analyze certain forms of scrambling (including those involving leftward movement of nominal arguments) as A'-movement. All of this implies that we cannot simply appeal to theory-internal considerations, but must develop empirical tests for supporting claims on the A- or A'-status of a specific form of scrambling.

[+]  3.  Test for determining A- and A'-movement: Binding

We will use binding as a diagnostic tool in order to establish whether the object movement found in the scrambling variant of the parasitic gap construction should be considered A- or A'-movement, as these movement types can be shown to differ in whether of not they affect binding relations. We illustrate this by using English data in order not to bias our discussion of Dutch beforehand.
      A'-movement does not alter binding options, as is clear from the examples in (530): the (a)-examples show that topicalization of the reflexive pronoun does not change its binding potential and the (b)-examples that topicalization of a potential antecedent does not create new binding posisibilities. We refer the reader to Section 11.3.6 on reconstruction for a more extensive discussion as well as the relevant Dutch data.

Example 530
a. John admires himself the most.
a'. Himselfi John admires ti the most.
b. *Ibelieve himself to admire Bill the most.
b'. * Billi, I believe himself to admire ti the most.

A-movement, on the other hand, does affect binding, as is clear from the subject raising examples in (531), taken from Den Dikken (1995); see Section for an introduction to subject raising. The traces indicate the current standard analysis of examples of this sort: in (531a) the expletive there is raised from the subject position of the infinitival clause into the subject position of the matrix clause; in (531b), it is the noun phrase some applicant s that is ultimately raised into the subject position of the clause. The crucial thing is that in (531a) the noun phrase some applicant s is clearly located in the infinitival clause and therefore does not c-command the complement of the to-PP, the reciprocal each other, while in (531b) the noun phrase some applicant s is moved into the subject position of the matrix clause and does c-command the reciprocal each other from this position. The acceptability contrast between the two examples thus shows that A-movement differs form A'-movement in that it does affect binding.

Example 531
a. * Therei seem to each other [ti to be some applicantsi eligible to the job].
b. Some applicantsi seem to each other [t'i to ti be eligible to the job].

The examples in (532) show essentially the same for the bound variable reading of referential pronouns: the quantifier in (532a) is embedded in the infinitival clause and therefore does not c-command the pronoun embedded in the complement of the to-PP, while the quantifier in (532b) is in the subject position of the matrix clause, from which it does c-command the pronoun. This accounts for the fact that the bound variable reading is only available in the latter case.

Example 532
a. * Therei seems to his mother [ti to be someone eligible for the job].
b. Someone seems to his mother [t'i to be ti eligible for the job].
[+]  4.  Empirical problems for the landing-site restriction: Webelhuth's paradox

The contrast between A- and A'-movement with respect to binding discussed in the previous subsection has played a major role in the discussion of the question as to whether scrambling of nominal arguments should be seen as A- or A'-movement, or perhaps even does not involve movement at all; a representative sample of these approaches can be found in Corver & Van Riemsdijk (1994).
      Webelhuth (1989/1992) has argued that Dutch/German object scrambling exhibits properties of both A- and A'-movement in that object scrambling not only licenses parasitic gaps, but also feeds binding, a fact known as Webelhuth's Paradox. That object scrambling may license parasitic gaps was already illustrated in (526), and that it may also feed anaphor binding is illustrated in (533); cf. Vanden Wyngaerd (1988/1989). Note in passing that example (533a) seems to improve somewhat if the adverbial phrase namens elkaar'on behalf of each other' is assigned contrastive accent; we will ignore this effect here, which may indicate that (533a) is derived from (533b) by means of reconstructible focus movement.

Example 533
a. * Hij heeft namens elkaar dejongens bezocht.
  he  has  on behalf of each other  the boys  visited
b. Hij heeft dejongensi namens elkaar ti bezocht.
  he  has  the boys  on behalf of each other  visited
  'He visited the boys on behalf of each other.'

Webelhuth's crucial observation, illustrated by the German example in (534), is that scrambling can simultaneously feed binding and license a parasitic gap. The structure indicated is the one assigned by Webelhuth: the scrambled quantified direct/accusative object jeden gast binds the possessive pronoun embedded in the indirect/dative object seinem Nachbarn'his neighbor', which licenses a bound variable reading, while it simultaneously licenses a parasitic gap. Such examples cannot be reproduced in Dutch because it does not easily allow inversion of indirect and direct objects in double object constructions.

Example 534
Peter hat jeden gasti [ohne pgi anzuschauen] seinem Nachbarn vorgesteld.
  Peter has  each guest  without  to.look-at  his neighbor introduced
'Peter introduced each guest to his neighbor without looking at him (each guest).'

Webelhuth assigns examples such as (534) a question mark, noting that they are "as good or as bad as" other parasitic gap constructions. He concludes from these examples that the dichotomy between A- and A'-positions is too coarse, and that we have to postulate a third, Janus-faced position that exhibits properties of both A- and A'-positions. This reasoning was sound at the time of Webelhuth's publication, but the increase of A- and A'-positions that followed in the 1990's allows a somewhat different view on examples of this kind: instead of assuming that the scrambled phrase is moved into its surface position in one fell swoop, we can now claim that it arrives there in a step-by-step fashion; see Mahajan (1990/1994) for early suggestions of this sort. This results in structures such as given in (535) with an additional trace t' added: if the first movement step is A-movement, the added trace is in an A-position and thus able to bind the reciprocal/possessive pronoun; if the second step is A'-movement, the scrambled phrase ends up in an A'-position, from which it can license the parasitic gap.

Example 535
Peter hat jeden gasti [ohne pgi anzuschauen] t'i seinem Nachbarn ti vorgesteld.
  Peter has each guest without  to.look-at  his neighbor  introduced
'Peter introduced each guest to his neighbor without looking at him (each guest).'

Since it has generally been assumed since Chomsky (1986) that A'-movement cannot precede A-movement, a restriction which has become known as the ban on improper movement, the proposed solution for Webelhuth's paradox makes a very strong prediction: the phrase containing the parasitic gap must be in a structurally higher position than the phrase containing the A-bound pronoun. This does not seem easy to test, however. At first sight, the German example in (536a), taken from Mahajan (1990:60), seems to confirm this prediction: since the direct object binds a parasitic gap, it must be in an A'-position and therefore cannot bind the possessive pronoun.

Example 536
a. *? Peter hat jeden gasti seinem Nachbarn [ohne pgi anzuschauen] ti vorgesteld.
  Peter has each guest  his neighbor  without  to look-at  introduced
b. *? Peter hat jeden gasti der Maria [ohne pgi anzuschauen] ti vorgesteld.
  Peter  had  each guest  the Marie  without  to look-at  introduced

It should be noted, however, that Müller & Sternefeld (1994) and Lee & Santorini (1994) claim that replacement of the indirect object seinem Nachbarn by an indirect object without a pronoun, such as der Maria in (536b), does not improve the result. This suggests that example (536a) is excluded for independent reasons and therefore does not bear on the issue under discussion. We cannot replicate the German data for Dutch double object constructions because indirect objects normally precede direct objects. But perhaps the examples in (537), in which the bound pronoun and the parasitic gap are both embedded in an adjunct, can be used to illustrate the same thing; note that the (b)-examples should be read as continuations of the (a)-example.

Example 537
a. dat Jan de rivaleni ...
  that  Jan the rivals
b. [zonder pgi aan te kijken] t'i namens elkaar ti feliciteert.
  without  prt. to look  on.behalf.of each.other  congratulates
b'. ?? namens elkaar t'i [zonder pgi aan te kijken] ti feliciteert.
  on.behalf.of each.other  without  prt. to look  congratulates

The judgments on these examples are somewhat problematic, however. First, we should note that Neeleman (1994a) gives the continuation in (537b') as acceptable, which means that the judgment given here is not uncontroversial. Second, we tend to think that this continuation only leads to a marginally acceptable result if the adverbial PP namens elkaar is followed by a brief intonation break. If so, the infinitival clause may be epenthetic and this would much complicate the analysis because it is often assumed that epenthetic phrases are clause-external. This means that the status of the continuation in (537b') is simply insufficiently clear, so that we cannot base any firm conclusion on this case. We therefore provisionally assume that the predictions that follow from the ban on improper movement are essentially correct until more conclusive counterevidence is provided.
      Note that Neeleman provides example (537b') in order to argue that scrambling is in fact not a movement operation; he argues instead that scrambled phrases are base-generated in their surface position, as indicated in representation (538a): if true, this would imply that the landing-site and the overt-movement restriction should both be rejected. Neeleman claims that nominalizations such as (538b) also support the hypothesis that parasitic gaps can be licensed by noun phrases occupying their base-position: the noun phrase boeken is able to license the parasitic gap despite the fact that is base-generated as the complement of the preposition van.

Example 538
a. Jan bracht zijn boekeni [zonder pgi in te kijken] terug.
  Jan brought  his books  without  into  to look  back
  'Jan brought his books back without looking into them.'
b. het [zonder pgi in te kijken] terugbrengen van boekeni
  the   without  into to look  bring-back  of books

Although this argument might have been sound in the early 1990's, in more recent years it has been argued that there is much more movement within noun phrases than meets the eye; see Hoekstra (1999) for an analysis of this example that adopts the movement approach to parasitic gaps. It is therefore no longer evident that example (538b) provides evidence in favor of the base-generation approach to parasitic gaps; we will return to this approach in the next subsection, where it will be shown to have a serious empirical inadequacy.

[+]  5.  A final problem for the landing-site restriction: passive constructions

The previous subsection has shown that Webelhuth's paradox receives a more or less natural explanation in the more recent versions of generative grammar that make more clause-internal A- and A'-positions available. There is, however, still a serious problem for landing-site restriction (515a), as various linguists have claimed independently of each other that parasitic gaps can occur in Dutch passive constructions. Broekhuis (1987/1992) claims that the result is somewhat less acceptable than in other cases but attributes this to the fact that the implied PRO-subject of the infinitival clause requires a controller (cf. Van Haaften 1991), as is clear from the fact illustrated in (539a) that the construction is also marked if the parasitic gap is replaced by an overt pronoun. De Hoop & Kosmeijer (1995) and Neeleman (1994a) give their examples as straightforwardly acceptable, which may be related to the fact that they include an agentive door-phrase, which may help to identify the implied PRO-subject; example (539b) shows that adding a door-phase indeed improves the parasitic gap construction in (539a).

Example 539
a. ? dat het boeki [zonder PRO zei/pgi te bekijken] ti werd weggelegd.
  that  the book  without  them/pg  to look.at  was  away-put
  'that the book was put away without looking at it.'
b. dat het boeki door Jan [zonder PRO pgi te bekijken] ti werd weggelegd.
  that  the book  by Jan  without  to look.at  was  away-put
  'that the book was put away by John without looking at it.'

To our knowledge, the consequences of the relative acceptability of the passive constructions in (539) have not yet been fleshed out. Broekhuis (1987/1992) suggests that the subject position is in fact not an A- but an A'-position in Dutch, which he supports by claiming that subjects of subject raising constructions such as (540a) are not able to bind (into) an indirect object of the matrix clause; cf. the discussion of the English examples in (531) and (532). Much rests on his claim that examples such as (540) are ungrammatical but this may be an overstatement; the judgments may simply not be clear enough to draw any firm conclusions.

Example 540
a. ? Zij leken elkaar/zichzelf [TPti ziek te zijn].
  they  seems  each.other/themselves  ill  to be
  'They seemed to each other/themselves to be ill.'
b. ? Iedereeni leek zijn moeder [TPti de beste kandidaat te zijn].
  everyone  seemed  his mother  the best candidate  to be
  'Everyone seemed to his mother to be the best candidate.'

      Another possibility, which has not been explored so far, is that the nominative noun phrase die boeken does not occupy the subject position at all in examples like (539). This is a plausible option because definite noun phrases can easily be shown not to occupy the regular subject position if they are part of the new information focus of the clause. This is illustrated in (541a), which shows that the definite noun phrase need not be right-adjacent to the complementizer dat'that' but may also occur in a more rightward position. That information structure is involved is clear from the fact that (phonetically reduced) referential subject pronouns, which are intrinsically part of the presupposition of the clause, do not have this option; cf. Section 13.2.

Example 541
a. dat <de boeken> waarschijnlijk <de boeken> verkocht worden.
  that     the books  probably  sold  are
  'that the books probably are to be sold.'
b. dat <ze> waarschijnlijk <*ze> verkocht worden.
  that   they  probably  sold  are
  'that they probably are to be sold.'

This would predict that the examples in (539) would become unacceptable if we substitute a referential pronoun for the noun phrase die boeken'those books'. It is not clear to us whether this prediction turns out to be true; although the examples in (542) may indeed be somewhat harder to interpret, this may simply be a side effect of the fact that they are given without an appropriate context.

Example 542
a. ?? dat zei [zonder PRO pgi te lezen] ti werden opgeborgen.
  that  they  without  to read  were  prt.-filed
  'that they were filed without reading them.'
b. ? dat zei door Jan [zonder PRO pgi te lezen] ti werden opgeborgen.
  that  they  by Jan  without  to read  were  prt.-filed
  'that they were filed by Jan without reading them.'

De Hoop & Kosmeijer (1995) and Neeleman (1994a) claim that parasitic gaps can be licensed by an antecedent in an A-position, which amounts to saying that the landing-site restriction does not apply to Dutch. Their claim further implies that the standard assumption that parasitic gaps are subject to binding condition C should be replaced by the assumption that they are subject to binding condition A or B. The fact that the antecedent of a parasitic gap is external to the infinitival clause in (543) suggests that the parasitic gap is free in its local domain; it is therefore clear that parasitic gaps are not subject to binding condition A.

Example 543
Subjecti .... (door NPj) [zonder PROj .... pgi .... te Vinfinitive] ti ...

The claim that the antecedent can be in an A-position thus inevitably leads to the conclusion that parasitic gaps are subject to binding condition B. This, in its turn, predicts that the antecedent of the parasitic gap may be bound by the subject of an (in)transitive matrix clause in the representation in (544a). We have not been able to construct such cases but this might be related to Van Haaften's claim that the implicit PRO-subject of the infinitival adjunct clause is normally controlled by the subject: if the subject controls PRO and binds the parasitic gap, this results in a violation of binding condition B because the parasitic gap would then also be bound within its local domain by the PRO-subject. A concrete example that illustrates this point is given in (544b).

Example 544
a. * [Subjecti .... [zonder PROi .... pgi .... te Vinfinitive] ...]
b. Jani werkte [zonder PROi zichzelfi/*pgi rust to gunnen].
  Jan  worked  without  himself/pg  rest to allow
  'Jan worked without allowing himself any rest.'

The claim that parasitic gaps are subject to binding condition B also predicts, however, that they behave like referential personal pronouns in that they can be bound by a nominal argument in some higher clause, but this is at odds with the contrast found in (545), which shows that while the referential personal pronoun haar'her' can be bound by the subject of the highest clause, Els, the parasitic gap cannot; cf. Bennis (1986:55).

Example 545
Elsi zei [dat Janj [zonder PROj haari/*pgi te raadplegen] daartoe besloten had].
  Els  said   that  Jan  without  her/pg  to consult to.that decided  had
'Els said that Jan had decided that without consulting her.'

If we want to maintain that parasitic gaps are subject to binding condition B, we can only account for this contrast in a principled way by appealing to one of the other restrictions in (515). If we follow De Hoop & Kosmeijer (1995) in adopting the traditional claim that the gap of the infinitival clause is parasitic on some movement operation in the matrix clause, we can appeal to the anti-c-command restriction in (515c), which will be discussed in the next subsection. If we follow Neeleman's (1994a) base-generation approach, the overt movement and anti-c-command restriction are no longer applicable, while the categorial and island restriction are both satisfied; this approach therefore requires the introduction of some (yet unknown) ad hoc stipulation.
      This subsection has discussed a final problem for the landing-site restriction by showing that the subject of Dutch passives can function as the antecedent of a parasitic gap. We have shown that if the antecedent of parasitic gaps can indeed be located in an A-position, the movement approach should be considered superior to a base-generation approach. We may also consider the possibility, however, that Dutch parasitic gaps are not true parasitic gaps, as has been proposed on other grounds for Dutch by Huybregts & van Riemsdijk (1985) as well as for German (see Culicover 2001 for references), but this seems less attractive because Dutch parasitic gaps seem to be well-behaved with respect to the other restrictions in (515). Yet another possibility is that there is simply something special about the infinitival clauses in the passive constructions in (539), given that Van Haaften's (1991:108) comparable passive examples without a parasitic gap are all severely degraded regardless of the presence of a door-phrase; this is illustrated in (546b).

Example 546
a. De politiei arresteerde mij [zonder PROi zichi te legitimeren].
  the police  arrested  me  without  refl  to identify
  'The police arrested me without identifying themselves.'
b. * Ik werd (door de politiei) gearresteerd [zonder PROi zichi te legitimeren].
  was  by the police  arrested  without  refl  to identify

If the PRO-subject of an adverbial zonder-clause must indeed be controlled by the subject of the matrix clause, the examples in (539) are not only surprising because they violate the landing-site restriction, but also because they exhibit exceptional control behavior. This should make us cautious not to jump to far-reaching conclusions on the basis of these examples only.
      Our discussion of parasitic gaps in passive constructions has not resulted in any clear conclusion but ended with a list of possible routes one might take to approach such examples. Since we have no further insights to offer at the moment, we leave this issue to future research.

[+]  B.  The anti-c-command restriction in (515c)

This subsection investigates the anti-c-command restriction, according to which the parasitic gap and the trace of its antecedent are not allowed to c-command each other, subsection II has mentioned that Engdahl (1983) found that the acceptability of parasitic gap constructions depends on the nature of the clause embedding the parasitic gap, as expressed by the accessibilityhierarchy in (547). Our discussion in the following subsections will follow this hierarchy with one divergence related to the fact that Engdahl's hierarchy is restricted to clauses: it does not include English cases such as Whoi would [a picture of pgi] surprise ti? in which the parasitic gap is embedded in a noun phrase. We will discuss the Dutch counterpart of these examples before the discussion of parasitic gaps embedded in relative clauses.

Example 547
Accessibility hierarchy for occurrences of parasitic gaps (simplified): infinitival adjunct clause > finite argument clauses > finite adjunct clauses > relative clauses

The discussion in the following subsections is greatly indebted to Bennis (1986:ch.1), which in its turn is based on earlier work of his with Teun Hoekstra (1984); Subsection 1 will include a discussion of an important restriction on Dutch parasitic gap constructions related to preposition stranding that is taken from this work.

[+]  1.  Parasitic gaps embedded in infinitival adjunct clause

All Dutch examples so far involve par