• No results found

proceedings 2003/9/19 17:30 page 29 #33

N/A
N/A
Protected

Academic year: 2021

Share "proceedings 2003/9/19 17:30 page 29 #33"

Copied!
18
0
0

Loading.... (view fulltext now)

Full text

(1)

3

Surprising Specifiers and Cyclic Spellout

Gisbert Fanselow

University of Potsdam

Contents

1 Introduction . . . 29

2 German main clauses . . . 30

3 PPT-movement . . . 32

4 An analysis of PPT-movement . . . 35

5 Alternative description I: remnant movement . . . 38

6 Alternative description II: copy-and-deletion approaches . . . 43

7 Concluding remarks . . . 44

1 Introduction

In the Y-model of grammar, the grammatical architecture underlying GB-theory (Chomsky 1981), phonological and syntactic operations are strictly separated from each other, and phonological rules can apply only after the completion of the syn-tactic part of the derivation. This view has dominated generative theorizing for two decades, but more recent grammatical models within the Minimalist Program (Chom-sky 2000, 2001) or Optimality Theory (see, e.g., Samek-Lodovici 2002) allow a more complex interaction of the two submodules of grammar. In particular, Chomsky (2000, 2001) proposes acyclic theory of spellout: whenever the syntactic computa-tion is complete in certain syntactic domains (‘phases’), these domains are interpreted both phonologically and semantically. Syntactic operations in larger domains may thus refer to results of phonological processes that have applied to smaller units.

The purpose of the present paper is to present some evidence from German main clauses that supports such a cyclic concept of spellout. Under certain pragmatic con-ditions, the choice of the element moved to theVorfeld(prefield, the specifier of CP)

0This article summarizes some of the points addressed in the paper presented at GLiP. I would like to

thank the organizers for the invitation, and the participants of GLiP for interesting discussions. Thanks also go to Damir ´Cavar, Caroline Féry, Hubert Haider, Gereon Müller, Stefan Müller, Matthias Schlesewsky and Ewa Trutkowski for helpful comments in various contexts. The research reported in this paper was partially supported by grant FOR 375 of the Deutsche Forschungsgemeinschaft to the Research Group “Conflicting Rules”.

(2)

depends on the outcome of the application of the Nuclear Stress Rule (NSR) to VP. Such an array of facts is expected in a cyclic theory of spellout, while it is incompati-ble with the Y-model of grammar.

The paper is organized as follows. Section 2 briefly summarizes some standard assumptions on the derivation of German main clauses. In section 3, we present data that show that the movement of parts of VP to theVorfeld may serve the function of expressing pragmatic functions borne by the complete VP. Section 4 argues that these data can best be handled by assuming that movement affects the element that bears the tone representing the topichood of VP. This element can, however, only be identified by the NSR. Sections 5 and 6 discuss alternative accounts that assume a global interaction of syntax and phonology. Section 7 briefly sums up our results.

2 German main clauses

Thiersch (1978) and den Besten (1989) have laid the ground for our current under-standing of German main clause structure. Subordinate clauses are verb-final (1a), while in a main clause, the finite verb or auxiliary moves to second position, viz. Comp. In addition to a feature attracting the finite verb, main clause Comps possess an EPP feature, which is responsible for the obligatory further movement of some constituent to the specifier position of CP.

(1) a. dass

that dertheMannman dasthe Buchbookgelesenread hathas b. [CPder Mann [[Comphat] das Buch gelesen]]

‘(that) the man has read the book’

In pragmatically unmarked declarative clauses, the first position can be filled by a subject, a sentence adverb, a temporal adverb, and a few other categories1, as (2)

il-lustrates. These are exactly the elements that can appear in the highest structural position of IP (apart from weak pronouns). Taking up a suggestion of Bhatt (1999) for Kashmiri, Fanselow (2002a) makes the Minimal Link Condition (MLC) of Chom-sky (1995) responsible for the correlation between the set of elements in Spec,CP and those licensed in the highest position of IP. According to the MLC, cannot move to

if there is a closer to

than that can also move to

. Since can always move

to Spec,CP in (3) representing the structure of a German main clause (unless it is an unstressed object pronoun), it blocks the attraction of other categories such as to Spec,CP

.

(2) Was ist geschehen? ‘what happened?’ a. Ein

a Zugtrainistwasentgleistderailed b. wahrscheinlich

probably istis eina Zugtrainentgleistderailed c. heute

todaymorgenmorningistis eina Zugtraineingleistderailed

‘Probably, a train was derailed (this morning)’

(3)

(3) [CP

[COMP [IP [. . . ]]]]

However, Comp may also possess an additional operator feature. In questions, Comp bears a [wh]-feature, in pragmatically marked declarative clauses, it may possess a [focus] or a [topic] feature. These features must be checked by the phrase moved to

, the specifier of CP. In other words, when Comp bears an operator feature f, the class of elements that can move to Spec,CP is restricted to those categories that bear an identical operator feature. Therefore, can move to

in (4) if possesses the feature f while does not (See Bhatt 1999, Fanselow 2002a). For example, an object

may be moved to first position in awh-question (5a). It can also be placed there if it is focussed, as in examples such as (5b)–(5c), or if it is a topic-phrase as in (6).

(4) [CP

[ [[COMP, +f] [IP [. . . ]]]]

(5) a. was

whathathasdertheMannman gelesenread ‘what has the man read?’ b. NICHTS

nothing hathaserhegelesenread ‘he hasn’t read anything’ c. einen

a miesenbad KRIMIdetective storyhathaserhegelesenread ‘he has read a bad detective story’

(6) (Soll ich was über Hans erzählen? ‘Do you want me to say something about Hans?’)

Diesen

this Verbrechercriminal hathasmanone endlichfinally verhaftenarrest könnencould ‘One has finally been able to arrest this criminal’

To sum up, the element occupying Spec,CP in German main clauses is either the high-est category in the IP below CP, or it is an operator.

There are, however, at least two types of sentences that do not fit this description. First, the category sitting in Spec,CP maydominatean appropriate operator rather than being an operator itself, as the examples in (7) show:

(7) a. [Wessen

whose Lied]song maglikeserheamat-themeistenmost ‘whose song does he like best’ b. [Senaits

Senait’sLied]song maglikeserheamat-themeistenmost ‘he likes SENAIT’s song best’

Such sentences can be dealt with easily, however. They merely illustrate the need for a theory of pied piping. When Comp possesses an operator feature f, the category c which Comp attracts must bear f as well, but c is not necessarily the only element that is displaced phonologically. Rather, conditions on pied piping (such as the ban against extracting possessors from DPs in German) may require that a category dominating c is moved.

(4)

3 PPT-movement

The second class of exceptions is more surprising. As we will see, it is characterised by the fact that somethingsmallerthan the real operator phrase is placed into Spec,CP. Therefore,pars-pro-toto-movement (PPT movement) is an appropriate term for such constructions.

In the most spectacular type of example, Spec,CP is filled by a verb prefix. Recall that a large number of German verbs possesses a separable prefix that is stranded when the verb moves to Comp. This is illustrated in (8) for the verban-fangen(begin, lit. ‘at+catch’).

(8) a. dass

that erhedamitthere-with‘at-catches’an-fängt ‘that he begins with that’ b. er fängt damit an c. *er anfängt damit

In most instances, the separation of the verb and the prefix is the result of stranding: the verb moves to Comp, leaving the prefix behind. Separable prefixes themselves are quite immobile. For example, they cannot undergo scrambling within IP. The claim that verb prefixes cannot be placed into Spec,CP either goes back to Bierwisch (1963), and was repeated over and over again in the syntactic literature on German (see S. Müller 2000 for an overview). Nevertheless, under certain pragmatic conditions, pre-fixescanbe placed into Spec, CP, as (9) and (10) illustrate.

(9) auf-machen (open, lit. ‘open-make’) auf

openhathaserhedietheTürdoorgemachtmade ‘he has opened the door’

(10) a. vor-haben (intend, lit. ‘before-have’) vor

beforehabenhave wirwe dasthatschonwell gehabthad ‘we had intended that’

b. vor-machen (to fool, lit. ‘before-make’) vor

beforekannstcan duyouderherwirklichreally nichtsnothingmachenmake ‘you cannot really fool her’

c. an-haben (wear, lit. ‘at-have’) an

at hattehad siesehnichtnot mehrmorealltoozuvielmuch ‘she did not wear too much any longer’ d. an-fangen (begin. lit. ‘at-catch’)

an

at habenhave wirwe DAMITthere withgefangencaught ‘we have begun with that’

(5)

Under our present perspective, data such as (9) are the least interesting: the prefix-verb combination has a more or less compositional interpretation. The result state described byauf ‘open’ can be considered the focus or the topic of the utterance, de-pending on the context. The syntax-pragmatics correlation of (9) need not differ from standard topic/focus movement at all.

Data such as (10) — in which the interpretation of the prefix-verb combination is not a compositional one — are more problematic for the theory of German syntax sketched in section 2. They are often ignored in the syntactic literature, or claimed to be non-existent, but the careful empirical survey carried out by Stefan Müller (2000) shows that such structures are wellformed, and that they can be found easily in text corpora. Often, such examples come with a certain stylistic flavour (playful language), and it need not be the case that they can be formed with all prefix verbs. See S. Müller (2000) and the references cited there for a discussion of some of the syntactic issues arising in the context of (10) that need not concern us here.

In the structures underlying (9) and (10), the verb prefix is part of the predicate, and doesnotoccupy the highest position in IP. Therefore, if our description of Ger-man main clauses in section 2 is correct, the prefix cannot move to Spec,CP when Comp has no operator feature (other categories are closer to Spec,CP than the prefix). The prefix can reach Spec,CP only if it is attracted by a Comp checking an operator feature. But the verb prefixes in (10) are meaningless elements — they form a mean-ingful unit in combination with the verb only. It is difficult to accept the idea that a meaningless element can be interpreted as a focus or topicoperatorphrase in German syntax. Therefore, it appears as if the data in (10) do not at all fit the idea that material in Spec,CP moves there because it is attracted on the basis of operator features.

The same problem reappears in constructions that do not involve the special syn-tax of particle verbs. Parts of idiomatic verb phrases that are again meaningless in isolation (relative to the idiomatic reading) may be placed into Spec,CP as well, as (11) illustrates.

(11) a. ins Bockshorn jagen (intimidate, lit. ‘into-the goat horn chase’) ins

into.theBockshorngoat horn hathaserhesichrefl nichtnot jagenchaselassenlet ‘he did not let himself be intimidated’

b. schöne Augen machen (to make eyes, lit. ‘beautiful eyes make’) schöne

beautifulAugeneyes hathaserheihrherekeinenone gemachtmade ‘he didn’t make eyes at her’

c. den Vogel abschiessen (to surpass everyone, lit. ‘the bird off shoot’) den

the Vogelbird hathasaberhoweverdertheVorsitzendechairman abgeschossenshot off ‘however, the chairman surpassed everyone’

Syntactically, the relevant idiom parts are direct (11b)–(11c) and prepositional (11a) objects, so that they cannot have moved to Spec,CP but by being attracted on the basis of an operator features — the subject is always closer to Spec,CP. Beingpartof an

(6)

idiom, the material placed into Spec,CP in (11) is itself meaningless, however, and therefore it does not constitute a good candidate for operator status.

While the data in (10) and (11) cannot be analyzed as illustrating the movement of an operator to Spec, CP, the pragmatic consequences of fronting a prefix or part of a verbal idiom can be characterized easily: in (10) and (11), thecompletepredicate constitutes the (contrastive) topic of the clause. Thus, (10a) can be uttered in a context in which the intentions of the group of the speaker are the topic (and in which these may be contrasted with what could be realized). Likewise, in (11a), the topic is the process of being intimidated, i.e., the complete lower VP (ins Bockshorn jagen). Similar observations can be made for the other examples in (10) and (11). The data can there-fore be characterized as in (12). They seem to involve something that one may call

pars pro toto-movement (PPT-movement).

(12) Under certain conditions, the attraction of a [+topic] phrase X to Spec,CP im-plies the overt movement of part of X only.

In the light of (12), a number of further mysteries of German topicalization may find a simple analysis. PPT movement cannot only be found in constructions with a (quasi-) idiomatic interpretation, it applies in fully compositional verb phrases as well. Thus, consider the small texts in (13):

(13) a. War er fromm? ‘Was he religious?’ a. In

IndertheBibelBible hathaserhenuronlyseltenrarelygelesen.read ‘He has only rarely read in the Bible.’ b. War er Anarchist? ‘Was he an anarchist?’ b. Häuser

houses hatgaserhejedenfallsin any eventnieneverangezündet.set on fire ‘He has never set houses on fire.’

In (13a,b), a prepositional object (in der Bibel) and a direct object (Häuser) have been

placed into Spec,CP. (13a,b) are felicitous continuations of (13a) and (13b),

respec-tively, in spite of the fact that the phrases in Spec,CP are not the topics of the two sentences in a pragmatic sense. The topic of (13a–a) is a religious activity (reading in

the Bible), rather than a book. The topic of (13b–b) is a typical action of an anarchist

(setting houses in fire), rather than a set of houses. In other words, the topic of the second sentences in (13a)–(13b) is their VP, but only part of that VP has been moved to Spec,CP. The data in (13) are thus further instantiations of PPT-movement.

From (12), it follows that sentences in which the direct object is placed into Spec,CP allow at least three interpretations in terms of information structure. The direct ob-ject might be the focus of the utterance (as in (5)), it may represent the topic of the utterance (as in (6)), and it might be a proper part of the semantic topic of the clause (as in (13)). In (13), the formal topicalization of a DP stands for the semantic topi-calization of a predicate phrase containing that DP. A similar effect can be observed for the scope of focus operators such asnur ‘only’. Consider (14) in this respect. It is ambiguous between the two interpretations given in (14a) and (14b). In (14a), it is

(7)

the semantic content of the DP that is in the scope of the focus operatornur. In (14b), however, the VP dominating the DP moved to Spec,CP (=read the bible) is affected.

(14) Nur

onlydietheBibelbiblehathaserhenienevergelesenread

a. Only for x, x = the bible: he has never read x b. Only P, P = bible reading: he has never done P

Example (14b) thus illustrates that (12) holds for XPs affected by focus operators, too. Finally, (12) also seems to figure prominently in the analysis of discontinuous noun phrases. The formation of discontinuous noun phrases is a perfect means of ex-pressing contrastive DP topics, as (15b) suggests when it is interpreted as an answer to (15a).

(15) a. Wieviele Tiere hat sie? ‘How many pets does she have?’ b. Katzen

cats hathassieshedrei,threeandundHundedogs (hathassie)shezwei.two ‘She has three cats and two dogs.’

In the light of what we have seen so far, it comes as no surprise that the formation of discontinuous noun phrases may serve a second pragmatic function, too: The pred-icate (reading many books, wearing a tie) is the topic of the second sentences in the texts in (16a) and (16b).

(16) a. Ist er gebildet? ‘Is he educated?’ Bücher

books hathaserhejedenfallsin any eventvielemanygelesen.read ‘At least, he has read many books.’

b. Ist er ordentlich angezogen? ‘Is he dressed properly?’ Krawatte

tie trägtwearserhejedenfallsin any eventwiederagain maloncekeine.no ‘Again, he does not wear a tie.’

PPT movement thus appears to be a widespread phenomenon in German syntax.

4 An analysis of PPT-movement

The pragmatic effect that comes with the data presented in the preceding section can also be arrived at by moving the complete verb phrase, as the following pairs of ex-amples show.

(17) a. vor

beforehabenhave wirwe dasthatschonwell gehabthad b. vorgehabthaben wir das schon c. das vorgehabthaben wir schon

‘we had intended that’

(18) a. ins Bockshornhat er sich nichtjagenlassen into-the goat horn has he refl not chase let

(8)

b. ins Bockshorn jagenhat er sich nicht lassen c. sich ins Bockshorn jagen lassenhat er nicht

‘he did not let himself be intimidated’ (19) War er fromm? ‘Was he religious?’

a. In

IndertheBibelbible hathaserhenuronlyseltenrarelygelesen.read b. In der Bibel gelesenhat er nur selten.

‘He has only rarely read in the bible.’ (20) Ist er gebildet? ‘Is he educated?’

a. Bücher

books hathaserhejedenfallsin any eventvielemanygelesen.read b. viele Bücher gelesenhat er jedenfalls

‘At least, he has read many books.’

In this respect, PPT-movement data are reminiscent of patterns involving wh -movement such as the ones in (21)–(23). Wh-movement normally involves the dis-placement of the fullwh-DP in German, but somewh-determiners can also be moved alone.

(21) a. was

whatfürforBücherbooks hasthaveduyougelesenread b. washast dufür Büchergelesen

‘what kinds of books have you read’ (22) a. wen

whovonof denthe Studentenstudents hasthaveduyougesehenseen b. wenhast duvon den Studentengesehen

‘which of the students have you seen’ (23) a. wieviel

how muchGeldmoneyhasthaveduyoudabeitherewith

b. wievielhast duGelddabei [ok in certain dialects only] ‘how much money do you have with you?’

The Comp nodes of the questions in (21)–(23) possess a [wh] feature that must be checked by the element that moves to Spec,CP. Attraction by Comp may trigger the preposing of thewordthat bears the attracted feature (was,wen,wieviel), but the (small-est) phrase dominating that word may be pied-piped (and in many circumstances, pied-piping is obligatory).

Given that this analysis follows from the fundamental assumptions of the Mini-malist Program (Chomsky 1995), one would expect that it characterizes topicalization (and focus movement) as well. In other words, topicalization and focus movement should either prepose thewordthat bears the topic (focus) feature, or some phrase dominating that word (which is small enough for being pied-piped). But what makes a word bear a topic feature? Unlike what holds forwh-words, the [+topic] feature is not linked to individual words by virtue of a lexical property. Furthermore, [+topic]

(9)

does not correspond to a morphological property of a word in German — rather, it is linked to a prosodic property. A word ‘bears’ a topic feature (more precisely: a topic feature manifests itself on a certain word) if that word bears a particular tone, a particular accent.

One therefore expects that topicalization moves the word that realizes the topic ac-cent, or some properly sized phrase dominating that word. How is the word bearing the topic accent identified? The answer to this question has both a pragmatic and a syntactic part. Obviously, the portion of the sentence that is the topic cannot be se-lected on a purely syntactic basis: it is the context and the communicative intention of the speaker that determines what is a topic or a focus. However,withinthe part of the sentence that is pragmatically identified as the topic, accent placement is carried in a purely structural fashion: the Nuclear Stress Rule (NSR) determines which word the accent goes to and it places the accent on thelowestword in the structure (see Cinque 1993, Zubizarreta 1998, among many others). When ZP is a semantic topic, syntactic topicalization should thus affect thestructurally lowest wordin ZP (=PPT movement)-but more material may (have to) be pied-piped (full phrasal movement).

The reader will already have noticed that the PPT data presented above neatly bear out the prediction that VP or the lowest element in VP may move to Spec,CP when VP is a topic. Consider first the VP in (24). By the NSR, a tone that is to be linked to VP must be realized on the lowest complement (the DP), and a tone linked to this DP must be realized by the noun embedded in DP, viz.Bücher. Thus,Bücherwill bear the topic tone associated with VP, and the movement ofBücherto Spec,CP may serve the function of expressing that VP is the topic (20a). Phrases dominatingBücher(up to VP itself) may be pied-piped in this context (20b).

(24) [VP[DPviele [NPBücher]] gelesen]

Determiners such as the definite one that can hardly be stranded in other contexts (‘normal’ discontinuous DPs) have to be pied-piped in PPT-movement as well. Like-wise, a noun cannot leave a prepositional object in the context of PPT-movement. Prepositional phrases simply must not be discontinuous in German (there is no prepo-sition stranding). Consequently, the complete PP-object must be pied piped when a noun expressing a topic feature assigned to VP dominating PP is attracted to Spec, CP.

Whether the word bearing the tone that represents the topichood of VP has a meaning of its own or not does not matter in this analysis at all. Therefore, we expect that the movement of (parts of) direct objects and prepositional objects may serve the function of signalling the topic status of VP independent of whether VP has an idiomatic or a literal interpretation.

Note, finally, that separable prefixes can take up a tone linked to VP. E.g., the structural sentential accent signalling wide focus goes to the verb prefix in (25). (25) a. dass

that sieshedietheTürdoorAUF-machteopen-made ‘that she opened the doors’

(10)

b. dass

that siesheihmhimnichtsnothingVOR-machtebefore-made ‘that she did not fool him’

The prefixes that occupy Spec,CP in (10) (expressing the topicality of the predicate) have therefore moved to that position because they bear the tone that is linked to VP because it is a topic.

In the interest of space, we have to leave open a number of important questions here: what is the exact syntactic status of verb prefixes that allows them to appear in a specifier position? How much material must be pied-piped (if at all) when VP-topicalization is carried out as PPT-movement? What forces the obligatory pied-piping of PPs? What are the locality restrictions for PPT-movement? How do we capture stylistic differences between phrasal and PPT-movement? The answers to these questions are important for a complete description of the PPT-movement. For the restricted purposes of the present paper, the sketch given so far is already suffi-cient, however:

When one wants to determine which element undergoes PPT-movement to Spec,CP in constructions involving an attracting operator feature such as [wh] or [topic], one needs to consult a theory of pied piping, and one needs to know which word realizes the feature in question. For [wh], one can obtain the relevant informa-tion from the lexical entries — but for the topic feature, one must know which word bears the tone coming with topichood. This word, however, is identified by appli-cations of the Nuclear Stress Rule, which belongs to thephonological component of grammar. This is an important result: when one wants to determine whether Z can be moved to

in (26) in constellations in which VP is the topic, one must know the result of the application of the NSR to VP.

(26) [CP

[[Comp. . . + topic] [IP. . . [VP . . . Z . . . ]]]

In other words: phonological rules must have applied to VP before operator move-ment can apply in CP. The interaction of syntax and phonology must (at least) be a of a cyclic nature, as suggested by Chomsky (2000, 2001).

5 Alternative description I: remnant movement

Of course, alternative descriptions can be formulated for the data introduced in sec-tion 3, and some of these alternatives may allow to avoid the conclusion that phonol-ogy and syntax interact cyclically. One of the alternatives that deserves special atten-tion is remnant movement.

PPT- and remnant movement may have the same overt effect: a category C is at-tracted, but overt displacement affects part of the phonological material of C only. Remnant movement was introduced as a descriptive device for German by Thiersch (1985) and den Besten & Webelhuth (1987,1990), see G. Müller (1998) for a detailed discussion (and Fanselow 2002b for a critique). According to their analyses, phrases may be extracted from VP before the latter category moves to Spec,CP, as sketched in (27). In a model that allows remnant movement, the phonetic content of X is not a good indicator for the category of an element X that has undergone overt

(11)

move-ment. Thus, the verbal participle in (27) is the only visible element that has moved to Spec,CP, in spite of the fact that the category that has ‘really’ undergone movement is a full VP.

(27) a. [IPer

henichtnot [VPsiehergeküsstkissed hat]]has b. [IPer nicht sie [VPt geküsst hat]]

c. hat [IPer nicht sie [VPt geküsst ]]

d. [VPt geküsst]

kissed hathaserersie

hernichtnot ‘he has not kissed her’

In a simple tense main clause as exemplified by (28a), the verb moves out of VP to Comp. After this movement, VP no longer contains an overt verbal head. If the objects can also be removed from VP (e.g., because they undergo scrambling), a structure such as (28b) arises in which the prefix is the only element left in VP that has a phonetic matrix. When such a VP is attracted to Spec,CP, structures will be generated in which the verb prefix is the only visible element in Spec,CP (28c).

(28) a. er

hemachtmakes[VPihrhernichtsnothingvorbeforet] ‘he does not fool her’

b. er macht ihr nichts [VPt t [vor t ]]

c. [VPt t t vor] macht er ihr nichts

In other words, data such as (9) and (10) also seem to be compatible with an analysis in which a VP goes to Spec,CP out of which all material but the prefix has been removed before topicalization. Similar evacuation analyses suggest themselves for the move-ment of parts of idioms (11), and the movemove-ment of DP- and PP-objects in sentences in which the VP is the topic from a pragmatic point of view. If the data introduced in section 2 involve remnant VP movement instead of PPT-extraction, there also seems to be no need for placing a tone in VP before VP is fronted by remnant movement. As a consequence, PPT-facts would not constitute evidence in favor of a cyclic interaction of syntax and phonology.

Closer inspection reveals, however, that reference to the location of the tone linked to VP cannot be avoided in a remnant movement theory either. When the idiomatic part of a VP comprises the direct object, a prepositional object, and the verb, it is often the direct object (rather than the PP) that bears mains stress, as exemplified in (29a) and (30a). When this direct object is placed into Spec,CP, the idiomatic interpretation is available, and the sentence links to an information structure in which predicate is the topic. When the PP is placed into the prefield, it can be interpreted in its literal sense only.

(29) a. wir

we habenhave ihmhimdenthe redrotenHAHNcock aufson-theDachroof gesetztput ‘we set his house on fire’

(12)

c. @aufs Dachhaben wir ihmden roten Hahn gesetzt (30) a. ich

I willwantdiryoukeineno stonesSTEINEinindenthe Wegwaylegenput ‘I don’t want to place obstacles in your path’ b. Steinewill ich dirkeine in den Weg legen c. @in den Wegwill ich dirkeine Steine legen

The contrasts in (29) and (30) are immediate consequences of the PPT-approach sketched in section 3, but they require an additional stipulation in the remnant move-ment approach: a VP remains a VP, irrespective of which categories have been ex-tracted from it. There is a need to distinguish [VPtaufs Dacht] from [VPden roten Hahn

t t], but this difference does not follow from the syntax of remnant movement at all. In addition to this difficulty, there are three facts that make a remnant movement account of PPT-data unattractive. First, unlike what seems characterize ‘standard’ remnant movement, the operations evacuating VP cannot always be motivated inde-pendently for the data from section 2. Second, the freezing effects that one would expect to observe in a remnant movement theory fail to show up. The latter property constitutes a general problem for the assumption of remnant movement in German (see Fanselow 2002b) and will not be discussed here in the interest of space.

When Thiersch, den Besten, and Webelhuth proposed remnant movement as an analysis for (27), their model won widespread support because it allowed to ex-plain an apparently problematic construction (heads appearing in a specifier posi-tion) in terms of two independently motivated operations (scrambling of objects, VP-movement to Spec,CP). When one tries to replace PPT-VP-movement by remnant move-ment, one must postulate movement operations that arenotmotivated on indepen-dent grounds, and that may even be quite problematic from the overall perspective on German syntax.

A first problem stems from the fact that the operation removing heads from VP that is crucial in the remnant movement analysis is motivated independently for simple examples such as (28c) only. In constructions such as (31), however, a finite auxiliary appears in second position, so that the prefix cannot have been separated from the lexical verb by verb movement to Comp. In order to account for (31a) in terms of remnant movement, one would have to assume that non-finite verbs (machen) adjoin to the verbal head (können) selecting them in overt syntax, and strand their prefixes in this context. For (31b), one would have to make the additional assumption that the head the lower verb has adjoined to (kannst) may excorporate from the complex

machen kannst. (31) a. vor

beforehasthaveduyouderhernichtsnothingmachenmake könnencould ‘you have not been able to fool her’

b. vor

beforekannstcan duyouderherwirklichreally nichtsnothingmachenmake ‘you really cannot fool her’

(13)

as-sumptions, and there are reasons for doubting that they are correct. Non-finite verbs can always be placed into Spec,CP (see (32)), which shows that the incorporation of lower verbs into higher ones is not overt in German.2

(32) a. küssen

kiss [hat

has dentheACC

Fritz

Fritzniemandnobody [t

können]

could t

]

‘nobody has been able to kiss Fritz’ b. zu

to küssen

kiss [hat

has dentheACC

Fritz

Fritzniemandnobody [t

gewagt]

dared t

]

‘nobody has dared to kiss Fritz’

Two further types of contrasts suggest that there is no string-vacuous head movement of verbs in German and Dutch. Haider (1993) and Koopman (1995) draw this conclu-sion from data such as (33) (and their Dutch counterparts). Certain verbs come with two prefixes rather than one, and they typically must not appear in second position, as the contrast between (33a) and (33b)–(33c) suggests. Haider and Koopman derive this and similar contrasts from the assumption that verbs likevoranmeldencannot undergo overt movement at all. If correct, the wellformedness of (33d) suggests that non-finite verbs do not adjoin to the heads selecting them in overt syntax.

(33) a. dass

that erheseinhis Kindchildvoran-meldetpre-at-reports ‘that he pre-registers his child’ b. *er voranmeldet sein Kind c. *er meldet sein Kind voran d. er

hewirdwill seinhis Kindchildvoranmeldenpre-at-report könnencan ‘he will be able to pre-register his child’

A detailed discussion of the contrast between (34a) and (34b) can be found in Haider (1997) and Meinunger (2001). Apparently, verbs must not be moved out of the scope of certain adverbial (?) operators such asmehr als‘more than’. (34a) suggests that finite verbs stayin situin embedded clauses (when there is no overt movement to second position), and by the same logic, (34c) implies that non-finite verbs do not undergo overt head movement either.

(34) a. dass

that erhedenthe Gewinnprofit mehrmorealsthatverdreifachtetripled ‘that he more than tripled his profit’

b. *er verdreifachte seinen Gewinn mehr als c. er

hehathasseinenhis Gewinnprofit mehrmorealsthanverdreifachentriple könnencould ‘he has been able to more than triple his profit’

Data such as (32)–(34) thus corroborate the view that movement to Comp is the

2See, e.g., Haider (1993) for further arguments against non-finite verb movement in clause union

(14)

only type of overt verb movement found in German. This is in line with the typo-logical generalizations formulated in Julien (2002). If one accepts this view, many PPT-data cannot be reanalyzed as involving remnant movement: there is no process which removes the verb from VP.

The postulation of an operation which removes complements and adjuncts from VP in the context of remnant movement leads to the same difficulty. Despite first ap-pearance, the evacuation of VP cannot be the result of scrambling. The scrambling of XPs out of VP always serves an information structure function. An XP may scram-ble when it is topical, or when it needs to avoid the information structure function it would receive if it stayedin situ. In addition, there is focus scrambling in German3.

This intrinsic connection to information structure implies that meaningless elements cannot undergo scrambling. In particular, parts of idioms do not scramble. Conse-quently, (35b) does not really allow the idiomatic interpretation that (35a) has. Quite in contrast to what we observe in (35b), the remnant movement analysis of (35c) pre-supposes thataufs Dachcan be extracted from VP — after all, the VP [VP den roten

Hahn aufs Dach gesetzt] needs to be transformed into [VP den roten Hahnt t] in a

rem-nant movement analysis of (35c). (35) a. wir

we habenhave ihmhimdenthe redrotenHAHNcock aufson-theDachroof gesetztput ‘we set his house on fire’

b. @wir haben ihm aufs Dach den roten Hahn gesetzt c. den roten Hahnhaben wir ihmaufs Dach gesetzt

Again, a movement operation necessary for the remnant movement theory cannot be motivated. Consider also (36) in this respect, the construction in which a part of a DP-object signals the topicality of VP. The verb phrase must haven been merged as [VP[DP

keine Krawatte]getragen]. The noun must then be separated from the determiner in the course of the derivation of (36). Given that German disallows left branch extractions, the separation can only arise by extractingKrawattefrom the DP, yielding [VPKrawatte

[VP [DP keinet] getragen]. The target remnant VP [VP Krawatte[VP ]] can then be

generated by extracting the remnant DP and the verb from VP. This derivation of (36) leaves it open why neither [VP [DP keinet]getragen] nor [VP [DPkeinet]] can be found

in VP-contexts in German, as (37) illustrates. (36) Krawatte

tie hathaserhekeineno getragenwear ‘he hasn’t worne a tie’

(37) *keine getragen hat er Krawatte keine hat er Krawatte getragen

We conclude that remnant movement does not really constitute an alternative to PPT extraction. See Fanselow (2002b) for further arguments against remnant movement as a process figuring in German syntax.

(15)

6 Alternative description II: copy-and-deletion approaches

Let us briefly discuss a further alternative. Fanselow & ´Cavar (2001, 2002) have pro-posed a model for the analysis of discontinuous noun phrases such as (38a) (and re-lated constructions such aswh-copying) that exploits the insight that movement is a two step process: a copy operation followed by deletion. They argue that most of the descriptive problems that arise in the context of discontinuous noun phrases find a straightforward solution if the derivation of (38a) proceeds as indicated in (38b)– (38e). Thecompleteobject-DP is copied to Spec,CP (38b)–(38c). After that operation,

partof the phonetic matrix of the upper copy is deleted (38c)–(38d), and this step is then followed by a partial deletion in the lower copy (38d)–(38e).

(38) a. Bücher

books hathaserhevielemanygeschriebenwritten ‘he has written many books’

b. hat er viele Bücher geschrieben c. viele Bücher hat er viele Bücher geschrieben d. viele Bücher hat er viele Bücher geschrieben e. viele Bücher hat er viele Bücher geschrieben

Partial deletion avoids some of the theoretical problems that arise with remnant movement, so that it is tempting (at least for the present author) to replace PPT-movement by an analysis involving partial deletion of VP. Our standard examples would be analyzed as indicated in (39).

(39) a. [das vor

beforegehabt] habenhave wirwe [dasthatvor- gehabt]had ‘we had intended that’

b. [ihr schöne

beautifulAugeneyes gemacht] hathaserhe[ihrherschöne Augen gemacht]made ‘he has made eyes at her’

c. [keine Krawatte

tie getragen] hathaserhe[keineno Krawatte getragen]worn ‘he has not worn a tie’

Partial deletion is a very powerful tool, so that it is acceptable in grammatical analyses only if properly restricted. Fanselow & ´Cavar (2001, 2002) have identified a number of factors that restrict partial deletion. In their proposal, complete deletion of the lower (=overt movement) or the upper (=covert movement) copy is always the preferred option, that is ignored only under restricted circumstances. There are principles that require that phonetic material be realized in certain positions, and there are princi-ples that constrain phonological complexity in certain positions. Furthermore, their model presupposes that the identification of the optimal form (which material is kept in which position?) applies in a cyclic fashion: when a copy operation has applied, it must be followed by the corresponding (partial) deletion operation before the cycle (phase) it has applied in is completed.

(16)

Recall that the identification of the phonological material that appears in Spec,CP in (39) is done on the basis of the Nuclear Stress Rule: at least those words must be placed into Spec, CP that realize the tone linked to topical VPs. The partial deletion operation must be able refer to this phonological information, and it applieswithin

the syntactic cycles. If one wants to analyze the phonetic chains of (39) in terms of partial deletion, the interaction of phonology and syntax must thus be cyclic, as it is in a model assuming PPT-movement.

7 Concluding remarks

In the preceding sections, we have discussed a type of German main clause in which a category bearing a certain pragmatic function is only partially moved to the position that corresponds to that pragmatic function. We have proposed a slight revision of the minimalist theory of movement: when a head (such as Comp) overtly attracts a certain feature (say: [topic]), the minimal element that has to move is the (phonological) word that ‘realizes’ this feature. Thewh-feature is a lexical property of certain words, but in languages like German, the topic feature is linked to a certain word if that word bears the tone that expresses topicality. The placement of this tone is partially determined by the Nuclear Stress Rule. Consequently, movement operations can often be applied properly only if they can make reference to the result of phonological computations. Syntax and phonology must interact in a cyclic fashion.

There are numerous other constructions that corroborate this conclusion. For ex-ample, recall that certain prefixesmustbe stranded when V is attracted to Comp in German — but others (the so-called inseparable prefixes) must always be pied-piped. Whether a particle is strandable or not depends on its phonological properties, viz. the question of whether it is stressed. Likewise, Stylistic Fronting seems to be trig-gered phonologically (Holmberg 2000). The application of scrambling may have a phonological component, too (Samek-Lodovici 2002, Zubizarreta 1998). The view that phonology and syntax interact in a complex and interesting way is thus strongly sup-ported.

References

den Besten, Hans. 1989.Studies in West Germanic Syntax. Amsterdam: Atlanta. den Besten, Hans & Gert Webelhuth. 1987. Remnant topicalization and the constituent

structure of VP in the Germanic SOV languages. Paper, presented at the X. GLOW Colloquium, Venice.

den Besten, Hans & Gert Webelhuth. 1990. Stranding. InScrambling and Barriers, eds. Günter Grewendorf & Wolfgang Sternefeld. 77-92. Amsterdam: Benjamins.

Bierwisch, Manfred. 1963. Grammatik des deutschen Verbs (=Studia Grammatica II). Berlin: Akademie Verlag.

Bhatt, Rakesh Mohan. 1999. Verb Movement and the Syntax of Kashmiri. Dordrecht: Kluwer.

Chomsky, Noam. 1981.Lectures on Government and Binding. Dordrecht: Foris.

Chomsky, Noam. 1995. Categories and Transformations. InThe Minimalist Program, 219-394. Cambridge, Mass.: MIT-Press.

(17)

Chomsky, Noam. 2000. Minimalist Inquiries: The Framework. In Step by Step, ed. Roger Martin, David Michaels & Juan Uriagereka, 89-155. Cambridge, Mass.: MIT Press.

Chomsky, Noam. 2001. Derivation by Phase. in Ken Hale. A Life in Language, ed. Michael Kenstowicz, 1-52. Cambridge, Mass.: MIT Press.

Cinque, Guglielmo. 1993, A null theory of phrase structure and compound stress.

Linguistic Inquiry24: 239-297.

Fanselow, Gisbert. 2002a. Quirky Subjects and Other Specifiers. InMore Than Words

eds. Barbara Kaufmann & Barbara Stiebels, 227-250. Berlin: Akademie Verlag. Fanselow, Gisbert. 2002b. Against Remnant VP Movement. InDimensions of Movement,

eds. Artemis Alexiadou, Elena Anagnostopoulou, Sjef Barbiers, & Hans-Martin Gärtner, 91-127. Amsterdam: Benjamins.

Fanselow, Gisbert & Damir ´Cavar. 2001. Remarks on the Economy of Pronunciation. InCompetition in Syntax, eds. Gereon Müller & Wolfgang Sternefeld. Amsterdam: Benjamins.

Fanselow, Gisbert & Damir ´Cavar. 2002. Distributed Deletion. InTheoretical Approaches to Universals, ed. Artemis Alexiadou, 65-107. Amsterdam, Benjamins.

Haider, Hubert. 1993.Deutsche Syntax — generativ. Narr, Tübingen.

Haider, Hubert. 1997. Typological Implications of a directionality constraint on projec-tions. InStudies on universal grammar and typological variation, ed. Artemis Alexiadou & Tracy Hall, 17-33. Amsterdam: Benjamins.

Haider, Hubert & Inger Rosengren. 1998. Scrambling.Sprache und Pragmatik49. Lund: Germanistisches Institut.

Holmberg, Anders. 2000. Scandinavian Stylistic Fronting: How any category can be-come an expletive.Linguistic Inquiry31: 445-483.

Julien, Marit. 2002.Syntactic Heads and Word Formation. Oxford: OUP.

Koopman, Hilda. 1995. On verbs that fail to undergo Verb-Second.Linguistic Inquiry

26:137-163.

Meinunger, André. 2001. Restrictions on verb raising. Ms., ZAS Berlin. Müller, Gereon. 1998.Incomplete category fronting. Dordrecht: Kluwer.

Müller, Stefan. 2000.Complex Predicates: Verbal Complexes, Resultative Constructions, and Particle Verbs in German. Habilitation thesis, Universität des Saarlandes, Saar-brücken. To be published by CSLI Publications, Stanford.

Samek-Lodovici, Vieri. 2002. Prosody-Syntax Interaction in the Expression of Focus. To appear inNLLT.

Thiersch, Craig. 1978.Topics in German Syntax. Dissertation, MIT, Cambridge, Mass. Thiersch, Craig. 1985. VP and Scrambling in the German Mittelfeld. Ms., University

of Tilburg.

Zubizarreta, Maria-Luisa. 1998.Prosody, focus and word order. Cambridge, Mass: MIT Press.

(18)

Gisbert Fanselow

Linguistics, University of Potsdam P.O. Box 60 15 53

14415 Potsdam Germany

References

Related documents

The survey asked what attributes of a pharmacoeconomics or outcomes research PhD program are most important for its success and quality and also what attributes should be con-

If word order in the Sandawe irrealis is affected by focus, we would expect to see variation in word order in the set of the 13 examples of target sentence A as these items

This is in line with the evidence discussed in Section 3 that shows that the minimum wage cuts deeper into the informal sector wage distribution (see Graph 3) and as a consequence

Inhibition of succinate accumulation and/or oxidation by dimethyl malonate (DMM), a cell permeable prodrug of the SDH inhibitor malonate, can decrease I/R injury.. However, DMM

However, recent studies have highlighted a lack of education on caring for older adults and people with dementia in undergraduate paramedic programmes (Abbey et al. 2006; Annear et

David (1999) was amazed at the change in roles she experienced when she switched from her role as a reading specialist to that of a literacy coach. For the past 15 years, I have

The purpose of language learning is communication (therefore students need to learn how to ask questions as well as

As long as the table space is in backup mode Oracle will write the entire block is dumped to redo when the ALTER TABLESPACE TBSNAME BEGIN BACKUP MODE is entered but later