Jump to ContentJump to Main Navigation
Scrambling, Remnant Movement, and Restructuring in West Germanic$

Roland Hinterhölzl

Print publication date: 2006

Print ISBN-13: 9780195308211

Published to Oxford Scholarship Online: January 2010

DOI: 10.1093/acprof:oso/9780195308211.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 22 September 2018

Summary and Conclusions

Summary and Conclusions

(p.213) 8 Summary and Conclusions
Scrambling, Remnant Movement, and Restructuring in West Germanic

Roland Hinterhölzl (Contributor Webpage)

Oxford University Press

Abstract and Keywords

This book has explained three salient syntactic phenomena in West Germanic, namely scrambling, remnant movement, and restructuring, and their interdependence was investigated in detail. It was argued that licensing movement not only affects the constituents of the VP but also applies to the major constituents of the clause, that is, to AspP and TP, which are taken to move into specific licensing positions in the C-domain. Moreover, restructuring involves movement of the infinitival AspP and TP into dedicated licensing positions in the matrix clause. It was considered that all movement operations involved in the complex derivations that lead up to remnant topicalization can be given a coherent account within a phase-based minimalist framework that only uses leftward movement into unique Specifiers of dedicated functional positions that are motivated by feature-checking requirements of the main constituents of the clause.

Keywords:   salient syntactic phenomena, West Germanic, scrambling, remnant movement, restructuring, remnant topicalization, minimalist framework, Specifiers

In this book, I have discussed three salient syntactic phenomena in West Germanic, namely, scrambling, remnant movement, and restructuring, and investigated in detail their interdependence. It was shown that the original account of remnant topicalization in (1a) in terms of prior VP-evacuation via scrambling, as has been proposed by Den Besten and Webelhuth (1989) and as is illustrated in (1b), is mistaken. Scrambling does not feed remnant movement, neither within the clause, as in (1), nor across clauses in cases of restructuring. What has been topicalized in (2a) is not the infinitival clause the direct object of which has been scrambled out, as is illustrated in (2b), but only the infinitival VP: a remnant category that is created by licensing movement and restructuring operations.



gelesen hat Hans das Buch

read has Hans the book


[CP [VP tSCR gelesen] hat [IP Hans das BuchSCR t VP ]]



zu kaufen hat er das Buch versucht

to buy has he the book tried


[[t SCR zu verkaufen] hat er das Buch versucht]

I have shown that within the clause it is licensing movement (emptying the VP up to the verb) that constitutes the basis for remnant movement of verbal projections, while in cases of restructuring it is remnant movement that feeds apparent scrambling across clauses.

(p.214) Licensing movement out of the VP also provides the basis for the antisymmetric account to the syntax of the West Germanic OV languages that I have developed and argued for in this book. All syntactic phenomena are derived within a purely right-branching clause structure via leftward movement into dedicated Specifiers of functional heads, rendering superfluous the operations of rightward movement and adjunction as well as the assumption of multiple Specifiers.

It is argued that licensing movement not only affects the constituents of the VP but also applies to the major constituents of the clause, that is, to AspP and TP, which are taken to move into specific licensing positions in the C-domain. Furthermore, I have argued that restructuring involves movement of the infinitival AspP and TP into dedicated licensing positions in the matrix clause. In this approach, restructuring has the same rationale as subject movement has in cases of (subject) raising constructions: A constituent that fails to be licensed in the embedded clause undergoes further movement to be licensed in the matrix domain. Thus, restructuring falls out as a special case within a theory of generalized licensing that obtains when the licensing heads in the C-domain are defective.

Constituents are licensed by checking/validating their features in dedicated functional projections. Along these lines, I have tried to combine Kaynean assumptions about phrase structure with minimalist assumptions about movement and feature checking. In particular, I have adopted the assumption that movements are triggered and that derivations are phase based (Chomsky 2001).

The results of the empirical investigation of the three syntactic phenomena of West Germanic within this combined approach that have interesting implications for the construction of the theory are summarized as follows:

  1. 1. The restrictions on VP-topicalization follow from the Phase Impenetrability Condition if cyclic Spell-out is assumed (chapter 7).

  2. 2. Despite the massive use of remnant movement in this approach, interpretational effects in coherent to-infinitives show that head movement cannot be dispensed with and must be defined in terms of antisymmetric c-command (chapter 5).

  3. 3. The principle of Attract Closest can only handle a subset of movement operations and must be replaced with a principle that guarantees strict cyclicity in a derivation (chapter 4).

  4. 4. The phenomenon of scrambling implies that if we want to exclude optional movement within a derivational framework, not only a small set of formal features can be taken to underlie a syntactic derivation, as is assumed in narrow syntax, but interface features must be imported into the syntax and be taken to drive derivations as well (chapter 2).

  5. 5. The flexibility of adjunction operations can be replaced with a system that allows for feature assignment in the course of the derivation (chapter 2).

The most important empirical results are discussed and summarized in the individual subsections dedicated to the different chapters that follow.

(p.215) 8.1 Scrambling and optionality

Chapter 2 provides a comprehensive discussion of scrambling operations in West Germanic. I first argue that two types of scrambling operations must be distinguished. I conclude that the scrambling operation in which the scrambled element is stressed has clear properties of A'-movement: It is not clause-bound and may affect arguments and predicates alike. Then, I focus on the properties of the type of scrambling operation in which the scrambled element does not receive a special stress or is completely unstressed.

I argue that this type of scrambling operation, called scrambling proper, is clause-bound and needs to be captured as an A-movement operation. I identify two types of triggers for scrambling proper, namely, specificity, defined in pragmatic terms as being familiar to speaker and hearer in the discourse situation, and scope, defined in terms of relational features. I argue that both types of features are checked by A-movement into the Specifiers of functional heads. While specificity is proposed to be checked in the Specifier positions of heads that license weak pronouns, the checking of the relational scope features, in the absence of free adjunction, requires an extension of the minimalist framework that allows for the introduction of non-lexical features in the course of the derivation.

Finally, I address the claim by H&R that scrambling is essentially optional and therefore defies any account as triggered movement operation. I argue that a trigger account is indeed feasible in a copy theory of movement in which both LF- and PF-conditions determine which copy is to be spelled out.

8.2 A VO-based account of verb raising and verb projection raising

Chapter 3 and chapter 4 introduce Verb Raising and Verb Projection Raising constructions in Dutch and West Flemish. First, I provide an empirical argument for a VO-based approach to the syntax of the West Germanic OV languages. With the help of the infinitival marker it is shown that all VP-internal material is moved out of the VP into dedicated licensing positions in the middle field. Then, I address the question of how VR and VPR are to be accounted for in a VO-based approach.

8.2.1 Verb raising

As is illustrated in (3), a typical case of VR in Dutch, nominal arguments of the infinitive and adverbs and adverbials that modify it precede the selecting verb, while the infinitive itself and a sentential complement of the infinitive (3c) follow the selecting verb. In (3), constituents that belong to the embedded clause are given in brackets. Within a VO-based approach, we have to assume that (3a) and (3c) are derived from an underlying structure of the type given in (3b).



dat Jan [Marie het boek morgen] wilde [geven]

that Jan Marie-DAT the book tomorrow wanted give

‘that Jan wanted to give Marie the book tomorrow’


dat Jan wilde [PRO Marie het boek morgen geven]

that Jan wanted Marie the book tomorrow give


dat Jan [Marie morgen] wilde [vertellen dat Piet ziek is]

that Jan Marie tomorrow wanted tell that Piet sick is

‘that Jan wanted to say to Marie tomorrow that Piet is sick’

(p.216) I present two arguments that show that the simplest possibility of relating (3a, c) with (3b), namely, by scrambling embedded material into the matrix clause, is untenable. (1) Verb-particles, small clause predicates, and idiomatic expressions cannot scramble, as has been shown in chapter 2, but precede the matrix verb in VR constructions. (2) Adverbs cannot undergo long-distance scrambling.

Furthermore, I argue that an alternative approach in which it is assumed that constituents of the embedded clause undergo licensing movement into the matrix domain, rather than scrambling—as has been proposed by Zwart (1993) and then recast in a more advanced minimalist system by Wurmbrand (2001)—is untenable as well. There are two arguments against such an approach to VR-constructions. (1) It would forestall a unified account of VR- and VPR-constructions and (2) Coherent infinitives comprise more than one licensing domain for their arguments and modifiers, as I argued in chapter 5.

I then describe in detail my account of VR in terms of remnant movement of the main phases of the infinitival clause into dedicated licensing positions in the matrix clause that dispenses with rightward movement as well as with unmotivated scrambling operations. The approach is based on the generalizations about the basic clausal structure outlined in section 4.2 (cf. [19]) and on the assumption that coherent infinitives are CP-complements, rather than reduced clauses. In this approach, arguments and adjuncts are not moved individually into the matrix clause but are pied-piped by the movement of a larger constituent. This larger constituent is argued to be the infinitival TP, which is moved to a position below all adverbs in the matrix clause. The basic tenets of restructuring are given in the following section.

8.2.2 Restructuring

  1. 1. movement of the embedded Aspect Phrase into [Spec,StatP] of the embedded clause

  2. 2. movement of the remaining TP of the embedded clause into [Spec,PredP] of the matrix verb via [Spec,MoodP] in the embedded clause

  3. 3. movement of the infinitival Tense-head to the functional head that licenses the controller of PRO to ensure the identification of the infinitival subject (as is discussed in detail in chapter 5)

  4. 4. additional XP-movement of AspP into a Specifier of the selecting verb to account for the formation of verb clusters and the appearance of the IPP-effect (see later discussion and chapter 6 for details)

The complex derivation of a VR-construction in Dutch is illustrated in (4). Given my assumptions, (4a) is derived from the source structure in (4b). In the first step, the arguments leave the VP to be licensed in the embedded clause, as is illustrated in (4c). (p.217) In the next step, AspP that has been emptied up to the verb is moved into [Spec,StatP] of the infinitival (and then moved on into a position that precedes the matrix verb but is spelled out in the infinitival clause. The result of this operation is shown in (4d).

In the following step, the remaining TP of the infinitival is moved via [Spec,MoodP] of the embedded clause to [Spec,PredP] of the matrix verb. The resulting structure is given in (4e). The reason for these licensing movements into the matrix domain is a deficient complementizer that cannot fully value the licensing projections of AspP and TP in the infinitival C-domain. In particular, it is assumed that StatP fails to check the subcategorization of the matrix verb; thus the infinitival AspP moves into a checking position in the V-domain of the selecting verb, giving rise to the formation of verb clusters. The MoodP is unable to temporally link the infinitival TP, which, not denoting an event-token, fails to qualify as an argument of matrix verb and thus “restructures” as a predicate by moving into the PredP of the matrix verb.

In the final step, both the matrix subject and the embedded direct object scramble to positions above the sentential adverb vaak. This last step is optional. Hence both dat vaak Jan het boek lang wil lezen and dat Jan vaak het boek lang wil lezen are fine sentences in Dutch. If we replace the adverb often with sentence negation in (4), then scrambling of both arguments becomes obligatory (in the absence of any contrastive focus) as we expect (cf. the discussion of scrambling in chapter 2). Scrambling of embedded arguments is enabled by extraction of the TP out of the embedded CP. Here we see one effect of the differentiation between licensing movement and scrambling. The arguments of the coherent infinitive are licensed in the embedded clause but can undergo scrambling according to their referential or quantificational properties in the domain of the matrix clause.



dat Jan het boek vaak lang wil lezen

that Jan the book often long wants read

‘that Jan often wants to read the book for a long time’


[dat vaak [VP Jan wil [CP … [lang [AspP 0 [VP PRO lezen het boek]]]]]]

Step 1: licensing movement in the embedded clause


[dat vaak Jan [VP wil [CP [TP PRO het boek lang [AspP [VP lezen]]]]]]

Step 2: AspP moves into [Spec,StatP] in the embedded clause


[dat vaak Jan wil [CP [AspP lezen] [TP PRO het boek lang tAspP]]]

Step 3: TP moves into [Spec,PredP] in the matrix clause


[dat vaak Jan [PredP [TP PRO het boek lang t AspP ] wil [CP lezen tTP ]]]

Step 4: scrambling of the matrix subject and the embedded object


[dat Jani het boekj vaak ti [PredP [TP PRO tj lang] wil lezen]]

8.2.3 Verb projection raising

In cases of VPR, the verb cluster may contain arguments and adverbs that belong to the embedded infinitival, as is illustrated in (5). What is peculiar about this partition of arguments is the following observation: A scopal element that is outside of the verb cluster can take its scope inside or outside of the verb cluster, but an element (p.218) that is part of the verb cluster may take its scope only inside of the verb cluster. It is important to note that an element outside of the verb cluster in its narrow scope reading always takes wide scope with respect to material that has undergone “VPR.” So, for instance, (6a) cannot mean that Jan made Valere three times read two books (possibly different ones at each occasion).



da Marie Jan nen boek tR wilt [geven]R

that Maria Jan a book wants give


da Marie Jan tR wilt [nen boek geven]R

that Marie Jan wants a book give


da Marie tR wilt [Jan nen boek geven]R

that Marie wants Jan a book give

‘that Marie wants to give Jan a book’

In my approach, the VPR-structure in (6a) differs from a typical VR-structure only in the amount of structure that is moved by the Aspect Phrase into the C-domain of the infinitival. The interpretational effect described in (6a) then simply follows from the mechanics of the system. In the infinitival clause, arguments cannot only undergo licensing movement but can also undergo scrambling, for instance, in order to take scope over an adjunct. In chapter 2, we have seen that a DP that is scrambled across an adverb has necessarily wide scope with respect to such an adverb but has necessarily narrow scope with respect to such an adverb if it fails to scramble. If we assume that it is possible to pied-pipe the phrase that corresponds to the domain that contains these adverbs with AspP-movement into the C-domain, then it follows without stipulation that the DP in (6a) necessarily has wide scope over the adverb in the VPR-complement: In more simple terms, in order for an element to move into the domain of the selecting verb via TP-movement, it has to take scope over the elements that stay behind and are pied-piped by movement of AspP, as is illustrated in (6b). In (6b), PPD stands for pied-piped domain, that is, the domain that has been pied-piped by movement of AspP.



da Jan Valere twee boeken deeg drie keers lezen

that Jan Valere two books made three times read


Summary and Conclusions

8.2.4 An account of sentential complementation

Finally, I propose an account of sentential complementation that allows us to derive the movement operations argued for earlier. The basic idea is that the local C-domain is responsible for licensing the AspP and TP in the clause, with the complementizer acting as a placeholder for the selectional requirements of the selecting verb. In this approach, movement of AspP and TP, which we argued to make up restructuring, occur in every clause.

Following recent work on the split C-domain (cf. Rizzi [1997]), I assume that the C-domain is made up of various functional projections. I assume that the complementizer (p.219) is inserted in StatP, where it licenses the finiteness of the clause and moves through MoodP, where the tense of the clause is linked to the speaking time with matrix clauses and to the matrix event time with embedded clauses, to Force, constituting the highest head in the C-domain. This is illustrated in (7), where it is assumed that the traditional CP corresponds to ForceP (CP = ForceP).

  1. (7)

    Summary and Conclusions

In an embedded clause with a non-deficient complementizer, the complementizer will value the heads Stat0 and Mood0 such that the embedded AspP and TP can be licensed in the respective Specifier, as is illustrated in (8). In this approach, the finite verb in matrix clauses only undergoes local movement within the C-domain: After the Aspect Phrase that contains the finite verb has been moved into [Spec,StatP], the latter extracts and moves via the head of MoodP to the highest head in the C-domain, as is illustrated in (9).

In the case of a restructuring verb, the complementizer is deficient and fails to license the embedded AspP and TP, which move on into dedicated licensing positions in the matrix clause, as is illustrated in (10).

  1. (8) licensing movements in an embedded clause

    Summary and Conclusions

  2. (9) licensing movements in a main clause

    Summary and Conclusions

  3. (p.220)
  4. (10) licensing movements in a coherent infinitive

    Summary and Conclusions

Assuming that a biclausal analysis is appropriate, let me take up at this point the question of whether the embedded clause should be a full CP, as is proposed earlier, or something smaller. In other words, the question arises of how we can distinguish between an analysis of coherent infinitives with a defective C-layer or a missing C-layer. If we assume that coherent infinitives are TPs in the latter approach, the same kind of licensing movements into the matrix clause without the intermediate steps in the C-domain in the embedded clause could be assumed.

I would like to discuss this point again in order to render some assumptions that I make in different parts of this book more explicit. On the empirical side, I argue for the presence of MoodP and StatP in restructuring infinitives in order to be able to handle extraposition. In the absence of rightward movement, TP- and AspP-movement must be taken to apply in every clause in coherent infinitives (cf. the discussion of the data in [8] in chapter 7).

In chapter 4, I argue that the C-domain, though defective, is not completely inert. The complementizer in coherent constructions is not fully valued or underspecified. I argue that it selects for a nonfinite complement, thereby excluding finite clauses from restructuring in German, and for a subjunctive complement, thereby excluding factive complements from restructuring contexts.

On more conceptual grounds the rationale of this approach is that these verbs simply s-select for propositions and that the canonical syntactic representation of a proposition is uniformly taken to be a CP. In other words, the difference between restructuring and non-restructuring verbs is not that the latter take propositions and the former something smaller, say, event-descriptions or something, but that they differ solely in the way they license their propositional complements.

The basic question about coherent infinitives has always been the issue of whether these infinitives are full or reduced clauses. I have no new insight to offer that could decide this question. My analysis solely provides a technical solution that allows us to treat coherent infinitives as full CPs, provided that we accept the notion of deficiency as it is understood earlier.

(p.221) 8.3 Coherent infinitives in German and the issue of monoclausality

In chapter 5, I provide empirical arguments to show that coherent bare infinitives in German are biclausal. The empirical arguments come from two different domains, the order of adjuncts and the binding properties in coherent infinitives, but point into the same direction, namely, that more than one licensing domain is present in coherent infinitives.

8.3.1 The order of adjuncts

Adjuncts in coherent infinitives with a selecting modal verb can appear in an order that is impossible in monoclausal structures. As is illustrated in (11), an aspectual adverb that modifies the matrix verb precedes a temporal adverb that modifies the infinitive in a coherent infinitival construction. Since temporal adjuncts occur in a structurally higher position than aspectual adjuncts (cf. Cinque [1999]), this order is ungrammatical in a single clause. These data speak against Cinque’s (2001) analysis of modal verbs as functional restructuring verbs as well as Wurmbrand’s (2001) monoclausal analysis of modal verbs as lexical restructuring verbs. In my approach these data are unproblematic, since coherent infinitives are taken to comprise a separate licensing domain for each verb and adjuncts that modify the infinitive are moved, via TP-movement to PredP, into a position below all adjuncts that modify the matrix verb.



weil Peter mich schon lange heute besuchen wollte

since Peter me already for-a-long-time visit wanted

‘already for a long time Peter has wanted to visit me today’


*weil mich Peter schon lange heute besucht hat

since me Peter already for-a-long-time today visited has


*weil das Peter schon lange heute wollte

since that Peter already for-a-long-time today wanted


weil mich Peter heute schon lange besucht hat

since me Peter today already for-a-long-time visited has

8.3.2 The binding properties of Ecm-infinitives

The binding properties of coherent infinitives embedded under ECM-verbs reveal that these infinitivals comprise two (distinct) binding domains. Assuming that the binding domain within the clause is constituted by the TP by default, I take these data as indicating that coherent infinitives must be at least as big as TPs. As is indicated in (12), while in a single clause both objects must be disjoint from the subject, the embedded object in a coherent infinitive can be coreferent with the matrix subject. Since the negation in (12) can have matrix scope (the most natural reading), the embedded subject and object must have moved into the matrix TP, after restructuring has moved the embedded TP into [Spec,PredP] of the matrix verb, as is illustrated in (12d).




weil eri ihn*i/j sie nicht waschen liess

since he him her not wash let-Past


weil eri sie ihni/j nicht waschen liess

since he her him not wash let-Past


weil der Hansi ihn*i/j ihm*i/k vorstellte

since the Hans him-ACC him-Dat introduced


[CP weil [TP1 er siei ihnj [nicht [PredP [TP2 ti tj] waschen liess/sah]]]]

Assuming that the embedded subject is Case-licensed in the matrix clause and the embedded object is Case-licensed in the embedded clause, the binding properties of ECM-infinitives follow from the principle in (13). These data therefore also provide evidence against the accounts of Wurmbrand (2001) and Zwart (1993), who assume that in VR-constructions the arguments of the infinitive must be licensed in the domain of the matrix verb.

  1. (13) Pronouns and anaphors are interpreted in their Case-licensing positions (i.e., they must be reconstructed into their Case-licensing position before the Binding Theory applies at LF).

8.3.3 Coherent to-infinitives

Coherent to-infinitives in German have figured prominently in the debate of whether coherent infinitives are to be analyzed as monoclausal or biclausal structures. Coherent to-infinitives allow for the so-called long passive and both Haider (1991) and Wurmbrand (2001) take this fact as conclusive evidence for the monoclausality of these infinitives. As is illustrated in (14), in a long passive what would be the object in the embedded clause is realized as the Nominative subject of the matrix verb.


Der Zaun wurde zu reparieren versprochen

the fence was to repair promised

‘Someone promised to repair the fence’

According to Haider (1991), this indicates that the object of the infinitive is Case-licensed by the matrix verb (with the object receiving Nominative Case if the matrix verb is passivized). That an embedded argument is Case-licensed by the matrix verb is familiar to us from ECM-verbs. In an ECM-construction, the subject as the highest argument can undergo Case-licensing movement into the matrix domain. But the same analysis cannot be applied to coherent to-infinitives, since the embedded object cannot be taken to undergo Case-licensing movement into the matrix domain across the embedded subject, that is, PRO. Therefore, a biclausal analysis of coherent to-infinitives seems to be impossible.

However, such an analysis is possible in my account of restructuring that is based on remnant movement of parts of the infinitival clause. The derivation proceeds in (p.223) the following way: The embedded object pronoun contained within the Agreement Phrase is pied-piped by movement of the AspP via the C-domain of the infinitival into [Spec,AspP] of the matrix verb, while PRO contained within the embedded TP is moved into [Spec,PredP] of the matrix verb. Note that in this first step the embedded object does not move across PRO. In the second step, the embedded object moves out of AspP to its Case-licensing position in the matrix clause. Movement of the embedded object is again unhampered by PRO, since PRO in the matrix clause is contained in a larger phrase, namely, the embedded TP. The derivation for a typical case like weil Hans ihr ihn zu waschen empfahl (‘since Hans recommended to her to wash him’) is given in (15).


[CP Hans ihr empfahl [CP [TP PRO [AgrP ihn zu waschen]]]]

Hans her recommended                               him/self to wash

Step 1: AgrP (the extendend AspP) moves via [Spec,StatP] into [Spec,CP]

a) [CP Hans ihr empfahl [CP [AgrP ihn zu waschen] [TP PRO tAgrP]]]

Step 2: AgrP moves to [Spec,AspP] in the matrix clause

b) [CP Hans ihr [AspP [AgrP ihn zu waschen] empfahl [CP tAgrP [TP PRO tAgrP]]]

Step 3: TP moves to [Spec,PredP] in the matrix clause

c) [CP Hans ihr [PredP [TP PRO tAgrP] [AspP [AgrP ihn zu waschen] empfahl [CP tAgrP]]]]

Step 4: The embedded direct object moves into its Case-position without crossing PRO

d) [CP Hans ihr [AgrO ihnk [PredP [TP PRO tAgrP][AspP [AgrP tk zu waschen] empfahl]]]]

This shows that a biclausal analysis of coherent to-infinitives is possible, if an independent explanation for why the embedded object cannot be Case-licensed within the infinitival clause is provided. I argue that coherent to-infinitives are nominalized infinitives (also called gerunds in this book), in which the categorial status of the infinitival marker is responsible for blocking assignment of structural Case to the object. Then I provide two arguments that indicate that a biclausal analysis of coherent to-infinitives is not only possible but also necessary. First, I show that the binding properties of anaphors in coherent to-infinitives, illustrated in (15e) later, are problematic for a monoclausal approach, while they follow naturally from the tenets of my biclausal approach. Second, I take Haider’s (1991) tests and criteria for monoclausal infinitives (derived from the properties of coherent to-infinitives) and show that these properties also hold of the other coherent infinitives, including those that clearly have biclausal properties. Since a monoclausal analysis of coherent ECM-infinitives and of coherent modal infinitives is impossible, the monoclausal analysis of coherent to-infinitives is to be discarded on grounds of parsimony.



*weil sichi/j ihrj der Hansi zu waschen empfahl

since herself/himself her the Hans to wash recommended

‘since Hans recommended to her to wash herself/himself’

(p.224) 8.4 The IPP-effect and the unified account of verb clusters in West Germanic

In chapter 6, I propose an account of the IPP-effect and provide a uniform format for the analysis of left- and right-branching verb clusters in German, Dutch, and West Flemish.

In chapter 4, I argue that the infinitival TP and AspP cannot be licensed in the embedded C-domain and move into dedicated licensing positions in the matrix clause. While the infinitival TP is licensed in [Spec,PredP] of the matrix clause, the infinitival AspP that contains the dependent verb moves into a licensing position below PredP that needs to be identified. The purpose of the final step in the licensing movement of the dependent verb is twofold: (1) the subcategorization of the selecting verb needs to be checked. Following Bech (1955/1985), I assume that a verb selects for the status of its nonfinite complement. That is, it determines whether the dependent nonfinite verb is a participle, a bare infinitive, or a to-infinitive. (2) Following recent work on Tense that requires verbs to be temporally anchored, I propose that dependent verbs must be linked to the matrix event time. Nominal categories, including nominalized infinitives, are exempted from this formal licensing requirement.

Chapter 6 discusses various types of evidence for determining which functional positions in the V-domain serve which licensing function. Example (16) illustrates the functional positions in the V-domain that are argued for in this book.


Skeleton of functional positions in the V-domain


(zu) [F2P


[VP … ]]]]

8.4.1 Accounting for the IPP-effect

In section 6.2, I argue that IPP-infinitives are hidden participles and that the IPP-effect reduces to a structural incompatibility between the participial prefix and the infinitive dependent on the restructuring verb, on account of the fact that the languages and dialects in which the participle is formed without a prefix, namely, Frisian and Low German, do not display an IPP-effect.

In the West Germanic languages that display the IPP-effect, the participle is formed by affixation of the prefix ge and the suffix t/d. I follow Halle and Marantz (1993) in assuming that inflected forms are (partially) derived in the syntax. More specifically, I propose that the participial prefix ge is inserted in [Spec,F2P] of the participial phrase. The verb in the participial phrase will first move to F2, to check its prefix, and then up to the Aspect-head to merge with the suffix that contains the temporal interpretation. In the final step the prefix left-adjoins to the complex of verb and suffix, to form the participle before Spell-out. This is illustrated in (17).

  1. (17) [AspP -t [F2P [ge] [F2 [VP V]]]]

If the verb in the participle phrase is a restructuring verb, then the dependent infinitive will move into [Spec,F2P] for licensing purposes. It follows that a verb in (p.225) participial form and a bare infinitive selected by such a verb rule each other out. In this case the participial prefix is blocked by the dependent infinitive, that is to say, it cannot be inserted. Since the prefix is selected by the participial suffix, the latter is dropped and the verb remains in F2 and is spelled out with the default morphology of a bare infinitive. Instead a zero-morpheme is inserted in the head of AspP that contains the formal feature [+ participle] and a semantic feature [+ past] to guarantee the correct interpretation of the hidden participle phrase. This is illustrated in (18). IPP-infinitives in German are obligatorily right-branching, since movement of the zero-morpheme into the Aspect Phrase of the selecting auxiliary strands the IPP-infinitive in [F2P] below.

  1. (18) [AspP 0 [F2P [dependent infinitive] IPP-infinitivei [VP ti ]]]

8.4.2 A unified analysis of verb clusters in West Germanic

While the surface data (in German) suggest that dependent verbs in restructuring contexts are licensed in [Spec,AspP] of the selecting verb, I provide evidence from the syntax of IPP-complements in West Flemish and Afrikaans that participles and verbal infinitives move through the Specifier of F2P below AspP (and sometimes remain there). With the help of Frisian data, I establish that the West Germanic dialects have two types of infinitives, one being directly licensed in AspP, the other, like participles, moving through [Spec,F2P] below AspP, thereby giving rise to the IPP-effect. Linking this observation with the fact that coherent to-infinitives, which I proposed to analyze as nominal infinitives in chapter 5, never give rise to an IPP-effect, I propose that F2 is responsible for temporal linking of dependent verbs, while the subcategorization of the matrix verb can be checked either in [Spec,AspP] or in [Spec,F2P]. The latter choice seems to depend on the amount of head movement of the verb in the V-domain in a language. While German nonfinite verbs always move into the highest head in the V-domain, that is, AspP, nonfinite verbs in Dutch and West Flemish only move as far as F2. This has important consequences for the analysis of verb clusters in the three languages. A dependent nonfinite verb in a left-branching verb cluster in German must be analyzed as occupying [Spec,AspP] preceding the selecting verb in Asp0. A dependent nonfinite verb in a right-branching verb cluster in Dutch or West Flemish must be analyzed as occupying [Spec,F3P] following the selecting verb in F2.



Right-branching verb clusters in Dutch and Westflemish:

temporal linking and subcategorization checking in [Spec,F2P], the selecting

verb remains in F2

Spell-out in [Spec,F3P], the left edge of the V-domain remains empty

[AspP (te) [F2P [V2] V1 [F3P [V2] [VP … ]]]]


Left-branching verb clusters in German:

temporal linking in [Spec,F2P], subcategorization checking in [Spec,AspP]

Spell-out in [Spec,AspP], the selecting verb moves to the highest V-position

[AspP [V2] V1 [F2P [V2] [F3P [VP … ]]]]


Right-branching verb clusters in German

same as above but the dependent infinitive is spelled out in [Spec,F2P]

the left edge of the V-domain is occupied

[AspP [V2]V1 [F2P [V2] [F3P [VP … ]]]]


Right-branching verb cluster projected by an IPP-infinitive in German

[AspP Aux [F2P [V3 V2] [F3P [VP … ]]]]

the zero-morpheme adjoins to the auxiliary in the highest head position in the

V-domain; thus [Spec,AspP] remains an available escape hatch

(p.226) 8.4.3 Extraction from the V-domain and the Phase Impenetrability Condition

Based on the preceding analysis of verb clusters in West Germanic, I show that the different status of topicalized right-branching verb clusters in German and Dutch/West Flemish follows from the prosodic condition in (21) and the Phase Impenetrability Condition (PIC). While right-branching verb clusters in Dutch can be topicalized, topicalization of right-branching verb clusters, with the exception of clusters that comprise IPP-infinitives, leads to ungrammaticality in German, as is illustrated in (20).



?*[müssen lesen können] wird er den Text

must read can will he the text


?[haben lesen wollen] wird er den Text

have read want-IPP will he the text

‘he will have wanted to read the text’


[moeten kunnen lezen] zal hij het boek

must can read will he the book

‘he will have to be able to read the book’

  1. (21) A right-headed phonological phrase in a verb cluster must sit on a right branch with respect to the non-head.

Extraction out of a verb cluster must proceed via the left edge of the V-domain, that is, [Spec,AspP]. Extraction of a right-branching verb cluster will thus lead to a violation of the prosodic constraint in (21). Violation of this constraint will lead to ungrammaticality if there is another Spell-out option as is the case in (20a), which could have been spelled out as lesen können müssen, but only to a marked grammatical result if there is no other Spell-out option, as is the case in (20b), since IPP-infinitives are obligatorily right-branching.

This account presupposes that Spell-out options in the V-domain are fixed before the derivation reaches the C-domain, since no such prosodic condition is at work in the C-domain in German. To sum up, the account of the rather subtle differences in (20) is based on three assumptions: (1) The Aspect Phrase (not the VP) constitutes a phase, (2) Spell-out is cyclic (rather than ensuing at the end of the entire derivation), and (3) there are interface constraints (like the mapping rule between syntactic (p.227) structure and prosodic structure in [21]), whose violation leads to ungrammaticality under certain conditions.

8.5 Extraposition, VP-topicalization, and the status of gerunds

Chapter 7 addresses a number of open questions. The first issue pertains to extraposition, a notoriously difficult topic within antisymmetric approaches. Without trying to devise a comprehensive account of extraposition, section 7.1 addresses a technical problem that is brought about by the account of sentential complementation developed in chapter 4. Because of their licensing requirements, CP-complements become part of the verb cluster (they are licensed in [Spec,F3P] of the selecting verb and due to the PIC become inaccessible for further computation. Clearly, this result is unwanted since CP-complements (1) can be topicalized and (2) must be “extraposed” from left-branching verb clusters.

8.5.1 Extraposition from verb clusters

I provide a technical solution to this problem, which treats extraposition as leftward movement into a high Specifier in the clause that follows from the condition in (22), which is a rendition of Büring and Hartmann’s (1997) account, and makes use of the proposal that both TP and AspP undergo licensing movement into the C-domain. A case of extraposition is illustrated in (23). At the end of the derivation neither Tense nor Aspect c-command the Tense within the “extraposed” CP, as demanded in (22). From its “scope” position in (23) the CP-complement, being outside of the verb cluster, can be topicalized on its own as well as be stranded or pied-piped by topicalization of the verb cluster, that is, by movement of the AspP into [Spec,CP].

  1. (22) A Tense-head may be neither in the checking domain nor in the scope (defined by C-command) of an Aspect- or Tense-head.



[CP C [MoodP M [StatP S [TP [AspP V CP]]]]]


[CP C[MoodP M [StatP S [CP [TP T [AspP V t]]]]]]



[CP C[MoodP M [StatP [AspP V t] S [CP [TP T]]]]]

licensing of the AspP


[CP C[MoodP [TP T] M [StatP [AspP V t] S [CP]]]]

licensing of the TP

Another issue concerns the topicalization of verb-projections in coherent infinitives. As we have seen in chapter 5, topicalization of the verb cluster or the dependent infinitive alone is one criterion for detecting a coherent construction. However, the test of topicalization also shows that large parts of the embedded clause can also be topicalized, even with verbs that restructure obligatorily. This is a hard problem, which to my knowledge has not been given a satisfactory explanation so far. In section 7.2, I outline an account that takes advantage of the availability of the two kinds (p.228) of infinitives in West Germanic. I argue that the two forms are in partial competition with each other, with the gerund functioning as a means of last resort. Though some questions of a mostly technical nature remain, this account paves the way to a general solution to this problem.

8.5.2 VP-topicalization

VP-topicalization data raise two kinds of problems for the account that I have developed in this book. First, VP-topicalization leads to a bleeding of the IPP-effect in Dutch and West Flemish, while in German the IPP-effect is only voided with perception verbs. Second, the dependent infinitive can be topicalized together with one of its arguments to the exclusion of the VR-verb. This is unexpected since verb and argument do not form a constituent anymore in my account after restructuring has applied.

The solution that I provide makes use of the different fine structure of verb clusters in German and Dutch/West Flemish, as it is outlined in chapter 6, the PIC, and the availability of the nominalized infinitive (the gerund) as a means of last resort.

In German, allowing for VPR, dependent infinitive plus argument embedded within an IPP-infinitive can be extracted out of the verb cluster due to last-resort movement to the left edge of the V-domain. Thus, VP-topicalization generally does not lead to a bleeding of the IPP-effect in German. Only in the case of perception verbs, movement of last resort of the infinitive that invokes the IPP-effect is blocked by a more economic derivation that involves the gerund, which generally fails to induce an IPP-effect, since it is licensed directly in the left edge of the V-domain.

In Dutch and West Flemish, however, movement of the dependent infinitive to the left edge of the V-domain from [Spec,F3P] is blocked by its own copy in [Spec,F2P]. Therefore, the gerund, which as a phrasal affix can attach to any extended projection of an infinitive, is inserted in the course of the derivation as a means of last resort. This gerund, on the one hand, will not induce an IPP-effect in cases of VP-topicalization, since it is licensed directly in the left edge of the V-domain. On the other hand, it is blocked by the more economic derivation that involves the (verbal) infinitive in restructuring constructions without VP-topicalization in Dutch (and West Flemish), since it is not selected by the restructuring verb, explaining why Dutch verb clusters may only contain verbs (and verb particles) when untopicalized but may contain arguments and adjuncts when topicalized.

Summing up, the differences in VP-topicalization between German, on the one hand and Dutch and West Flemish, on the other hand, follow from the fine structure of the verb clusters in these languages and the Phase Impenetrability Condition.

8.5.3 A unified analysis of the gerund

In chapters 6 and 7, I discuss different occurrences of nominalized infinitives. I assume that some occurrences of gerunds are selected (cf. the so-called Doelfoarms) and that sometimes they can be used as a means of last resort. Furthermore, I propose that coherent to-infinitives involve a nominalized infinitive as well. The common assumption that I made about these different occurrences is that the gerund is a (p.229) phrasal affix of nominal nature that morphologically selects for an infinitive and that this requirement is satisfied via adjacency. Section 7.4 proposes a unified account of the different occurrences of the gerund. The gerund is treated as a phrasal affix that always nominalizes full clauses. I propose that parallel to complementizers, this nominal affix is inserted in the head position of the Status Phrase above TP. In this analysis, gerunds can be taken to be nominalized clauses that license adverbs and Case. This is especially important for the analysis of coherent to-infinitives in German. Remember that the latter can license adverbs but fail to license structural Case. Now, the licensing of adverbs follows since the nominal affix attaches above the infinitival TP while the failure to license structural Case is relegated to the categorial nature of the infinitival marker in coherent to-infinitives.

This analysis of coherent to-infinitives highlights the importance of the role of the Status Phrase for sentential complementation. The Status Phrase is not only responsible for checking the finiteness of the verb but also the place where the complementizer and nominalizing affixes are inserted. These elements are essential for qualifying an embedded proposition as an argument of the selecting verb by nominalizing the embedded clause.

8.6 The connection between scrambling, remnant movement, and restructuring

In this section, I would like to discuss how the analysis of VP-topicalization and the connection between scrambling, remnant movement, and restructuring that I sketched in chapter 1 are accounted for in the approach that is developed in this book.

8.6.1 The analysis of remnant topicalization

A typical case of VP-topicalization (also simply called remnant topicalization) is given in (24a). Its analysis in the standard account following Den Besten and Webelhuth (1987) is illustrated in (24b). In this account, arguments and adjuncts are scrambled into the matrix clause, while the infinitival clause—however big it is supposed to be—that contains the remnant infinitive is moved into [Spec,CP] of the matrix clause. In this account, (long-distance) scrambling permitted by the specific properties of a restructuring infinitive feeds remnant topicalization.

In my account, only the infinitival AspP, a remnant category created by the clausal split triggered by restructuring, is topicalized. The topicalized AspP is a remnant category that only contains traces of licensing movement. Arguments and adjuncts of the infinitival are not scrambled into the matrix clause but arrive there via remnant movement of the containing infinitival TP, as is illustrated in (24c). Additional scrambling of a constituent can then obtain according to its referential and quantificational properties, either within the infinitival TP, as is illustrated in (24c), or into the matrix TP.

In this account, scrambling does not feed remnant topicalization. Remnant categories are solely created by standard licensing movements and licensing movements induced by restructuring.




besuchen will Hans die Maria morgen

visit wants Hans-NOM the Maria-ACC tomorrow

‘Hans wants to visit Maria tomorrow’


[tScr tScr besuchen] will Hans die MariaScr morgenScr


[AspP besuchen [VP tCase]] will [TP1 Hans [TP2 die MariaScr morgen tScr]]

As I have noted in chapter 6, bare infinitives can be topicalized together with their direct object, while to-infinitives fail to do so. This difference follows from the different licensing status of direct objects in bare infinitives and to-infinitives.

8.6.2 Differences between bare infinitives and to-infinitives

For the sake of illustration, let us compare the derivations of the minimal pairlike sentences in (25).



die Maria besuchen wollte Hans morgen

the Maria-ACC visit wanted Hans-NOM tomorrow

‘as for visiting Mary John wanted to do it tomorrow’


??die Maria zu besuchen wünschte Hans morgen the

Maria-ACC to visit wished Hans-NOM tomorrow

With the bare infinitive in (25a), the direct object is Case-licensed by the infinitive and both VR and VPR can apply to the infinitival complement. If VPR applies, the direct object is pied-piped by movement of the Aspect Phrase into the matrix clause and can thus be topicalized with the dependent infinitive by extracting out of the verb cluster created by restructuring. The essential movements of the derivation of (25a) are given in (26).

  1. (26)

    Summary and Conclusions

In (25b), however, the direct object, due to the presence of the infinitival marker and its categorial nature in restructuring contexts, cannot be licensed in the embedded clause. The direct object is pied-piped by movement of the infinitival AspP into the matrix clause but has to undergo Case-licensing movement into the matrix TP (p.231) after verb cluster formation. Thus, “VP-topicalization,” that is, movement of the infinitival Aspect Phrase into the matrix [Spec,CP], cannot affect the direct object, deriving that only verbs can be topicalized with coherent to-infinitives. The essential movements of the derivation of (25b) are given in (27).

  1. (27)

    Summary and Conclusions

8.6.3 Conclusions

The complex interplay between restructuring, remnant movement, and scrambling that is evidenced in cases of remnant topicalization is the result of licensing movements of various types that occur in coherent as well as in non-coherent clauses. What is special about restructuring infinitives is that the main constituents of the infinitival that are otherwise licensed in the embedded C-domain are licensed in the matrix clause. The movements that ensue from these licensing requirements lead to the formation of verb clusters and are responsible for the general transparency of coherent infinitives. Verb cluster formation in turn is the basis of remnant topicalization, as we have seen earlier. Scrambling, however, while not playing any role in restructuring and verb cluster formation itself, is the operation that is responsible for the so-called “matrix clause-interpretation” of arguments that belong to the infinitive and applies after the licensing movements of the infinitival TP and AspP have “restructured” the clause. Moreover, I have shown that the facts of remnant topicalization provide convincing evidence for the unified biclausal analysis that I have proposed for coherent bare infinitives and to-infinitives.

Finally, I have argued in this book that all movement operations involved in the complex derivations that lead up to remnant topicalization can be given a coherent account within a phase-based minimalist framework that only employs leftward movement into unique Specifiers of dedicated functional positions that are motivated by feature-checking requirements of the main constituents of the clause. (p.232)