Left Peripheral Focus: Mismatches between Syntax and ... - CiteSeerX

Left Peripheral Focus: Mismatches between Syntax and Information Structure* Gisbert Fanselow (Potsdam) & Denisa Lenertová (Leipzig) Abstract In Czech, German, and many other languages the leftmost accented part of the focus can be moved to the left periphery of the clause. We propose an analysis of this construction by modifying the theory of cyclic linearization developed by Fox & Pesetsky (2003, 2005). Crucially, our analysis makes no reference at all to concepts of information structure or to prosodic categories. We show that the partial fronting of topics fits into our model as well. Furthermore, the complete fronting of the focus or the topic turns out to be a mere special case of partial fronting. This implies generally that movement to the left periphery in Czech and German is not formally related to information structure, which is in line with the claim of Chomsky (2005) that UG specifies no direct link between syntax and information structure.

1.

Introduction

In Czech (Cz), German (Ge), and other languages, virtually any category can appear in the left periphery of declarative clauses, as (1) illustrates for direct objects.1 In the standard analysis of such sentences, the mechanics of the movement of the object is described along the lines designed for wh-fronting, i.e., the fronting of the object establishes an agreement relation between an operator-like feature of the object and Comp. In particular, the features relevant for movement encode notions of information structure such as ‘focus’ or ‘topic’ in such an analysis. (1)

a. Den Josefi mag jeder ti . the.acc Josef likes everybody (každý) uznává b. Josefai Josef.acc everybody.nom appreciates ‘Everybody appreciates Josef.’

(Ge) (každý) ti . everybody.nom

(Cz)

The present paper argues that this view is misguided. Accentuation rather than informational status determines which categories can be fronted, but this special role of accentuation does not imply that syntax refers to prosodic features. Accentuation comes into play indirectly only, its relevance for fronting stems from the fact that accentuation is a side-effect of cyclic linearization in the sense of Fox & Pesetsky (2003, 2005), Müller (2007). In addition to offering a new analysis of Czech and German constituent order, our paper therefore also * The research reported here was partially supported by DFG grants FOR 375 (A3), SFB 632 (A1) [Potsdam] and FOR 349 (D1) [Leipzig]. Numerous discussions with our colleagues have helped us to improve the present paper considerably. In particular, we are indebted to Joanna Błaszczak, Caroline Féry, Werner Frey, HansMartin Gärtner, Beáta Gyuris, Andreas Haida, Daniel Hole, Roland Hinterhölzl, Shin Ishihara, István Kenesei, Katalin É. Kiss, Ivona Kučerová, Maria Lesinski, Gereon Müller, Stefan Müller, Andreas Pankau, David Pesetsky, Pawel Rutkowski, Vieri Samek-Lodovici, Matthias Schlesewsky, Tue Trinh, and Malte Zimmermann. Thanks also go to the four anonymous reviewers, and to the participants of various workshops and lectures in Berlin, Gniezno, Leipzig, Stuttgart, and Warsaw, for their very helpful comments and suggestions. 1 Czech is not a verb-second language, thus word order variation behind the fronted element in (1b) is possible.

1

addresses a broader theoretical issue. Our findings support the view (see, e.g., Chomsky 2005) that notions of information structure do not figure in the syntactic derivation, at least not in the sense of being formally responsible for movement. Second, the analysis proposes a treatment of accent-related locality constraints on movement, which is relevant for the issue of the interaction of syntax and phonology. The paper is organized as follows. Section 2 sketches the standard analysis of movement to Spec,CP in Czech and German, and argues that five observations call for a revision of this standard analysis. A construction we call ‘subpart of focus fronting’ (SFF) turns out to be particularly problematic. A detailed discussion of its properties follows in section 3, leading to the main generalization that SFF constructions allow only the leftmost accented part of the focus constituent to be placed into Spec,CP, and to the conclusion that it is a proper syntactic process. We account for its crucial properties in section 4 by modifying the cyclic linearization theory of Fox and Pesetsky. The analysis is then extended to other types of leftperiphery-movement. In section 5, we dismiss alternative analyses of SFF as an attraction of accents, remnant movement or scattered deletion. Section 6 argues that there is no need for focus or topic movement rules in addition to the movement triggered by an unselective edge feature of C. German and Czech thus lend further support to the claim that there is no direct link between information structure and syntactic movement (Baker 2001, Chomsky 2005), a point also corroborated by Italian and Hungarian. We conclude the paper with a brief look at SFF in further languages. 2.

The standard analysis of the left periphery

The grammatical systems of the SVO language Czech and the SOV language German differ in many respects, yet they share the basic grammar of the clausal left periphery. Both languages have second position effects manifesting themselves in the (finite) verb second constraint of German main clauses, and the obligatory placement of clitics (pronominals and auxiliaries) into the second position in Czech. The position preceding the finite verb/the clitics is occupied by a wh-phrase in constituent questions, and by some maximal projection that can bear virtually any grammatical function (subject, adjunct, direct and indirect object, etc.) in declarative clauses. For German wh-questions, there is no reason to work with anything but the standard generative account.2 The wh-phrase moves to Spec,CP, and the second position is filled by movement of V to C. (2)

Peter ti eingeladen tV ]]? [CP [ wen]i [C hat] [TP der who.acc has the.nom Peter invited? ‘Who has Peter invited?’

(Ge)

The idea suggests itself that the fronting of XPs to the left periphery of declaratives finds the same analysis, as was first proposed by Thiersch (1978). Most generative approaches to German subscribe to that view (but cf. Müller 2004). One remaining issue would be whether wh-phrases and non-wh-phrases target the same slot, a question answered negatively in socalled cartographic approaches (Rizzi 1997), which split up Comp into at least four different heads. Abstracting away from such questions, den Josef appears in Spec,CP in examples such as (1a), and the finite verb is placed in C. (3)

[CP [den Josef]i [C mag] [TP jeder ti tV ]]

2

However, not all instances of wh-phrase movement to Spec,CP in German seem amenable to a treatment in terms of wh-agreement, cf., e.g., the discussion in Müller & Sternefeld (1996).

2

The analysis for Czech is similar. Toman (1999) argues that auxiliary clitics in Czech locally relate to the complementizer and are in the highest position below C0 (FinP/MoodP in his account), a similar proposal is made in Lenertová (2004).3 Wh-phrases have been argued to target Spec,CP in Czech in Meyer (2004), cf. (4a). This can be applied to fronted elements generally, (cf. 4b). (4)

Petr pozval ti ]]? a. [CP [ Koho]i [FinP [Fin by] who.acc aux.cond Peter invite ‘Whom would Peter invite?’ (každý) uznával (každý) ti ]]. b. [CP [ Josefa]i [FinP [Fin by] Josef.acc aux.cond everybody appreciate everybody ‘Everybody would appreciate Josef.’

(Cz)

Wh-movement to Spec,CP is linked to the wh-criterion requiring that a [+wh] C agrees with a [+wh]-phrase in its specifier, and vice versa (Rizzi 1991, 2004). The apparent parallelism between wh-movement and the movement to the left periphery in declaratives suggests that the latter is also licensed by establishing an agree relation between features of C(-heads) and the fronted XP (Rizzi 1997, 2004), and (non-subject) fronting in declaratives indeed seems to have an information structure function (e.g., Grewendorf 1980). The sentences in (6) can be used as answers to context questions such as (5), showing that the fronted elements (6) are interpretable as topics, which can be defined as ‘aboutness topics’ (cf. Reinhart 1981) or contrastive topics/‘I-topics’ (cf. Jacobs 1996, Büring 1997, Reich 2003, see also Speyer 2007). The former case is illustrated by the question-answer pair (5a)-(6), the latter case by the pairs (5b-c)-(6). (5)

a. I’ve heard the mayor has been arrested. Who reported him to the police? b. I know who reported the secretary to the police, but who reported the mayor? c. Who reported the mayor, his secretary, and their lovers to the police?

(6)

a. Den Bürgermeister hat wohl der Villenbesitzer angezeigt. the.acc mayor has supposedly the.nom owner-of-the-villa reported b. Starostu prý udal majitel vily. mayor.acc supposedly reported owner.nom villa.gen ‘The owner of the villa supposedly reported the mayor.’

(Ge) (Cz)

The fronting in (6) might establish agreement for a [+topic] feature. However, focused phrases can also move to the left periphery, as (7a,c) show (the versions b,d illustrate the ‘neutral’ order – with focus in situ – in the same context). In order to avoid confusion, we will thus use the term ‘left periphery movement’ (LP-movement) rather than the usual ‘topicalization’ for movement to Spec,CP in declaratives. (7)

What did you see there? a. LAvinui jsme viděli ti !4 avalanche.acc aux.1pl seen.pl

(Cz)

3

See also Kučerová (2006). Some accounts treat second position (2P) clitic auxiliaries in Slavic analogously to Germanic verb second and locate auxiliary clitics in C0, see, e.g., Wilder & Ćavar (1994). However, there is no complementary distribution of second position clitics and complementizers in Slavic, clitic auxiliaries follow the complementizer in embedded clauses. See Bošković (2001) and Franks & King (2000) for a discussion. 4 In the Czech and German examples, syllables carrying the main stress will be capitalized. Where relevant (cf. section 5.2), small caps will mark syllables with prenuclear stress. In the interest of space, word order variations of previous versions will lack glosses and context questions will be given in English only rather than also in the target language.

3

b. Viděli jsme LAvinu! c. Eine LaWIne haben wir ti gesehen! an avalanche have we seen d. Wir haben eine LaWIne gesehen! ‘We saw an avalanche!’

(Ge)

The data in (6)-(7) are compatible with several models concerning the syntactic encoding of information structure. Comp may optionally possess a focus or a topic feature with which the fronted phrases agrees (Rizzi 2004), and Comp may be split into independent Topic and Focus heads, which attract XPs with the corresponding informational value (Rizzi 1997). For several reasons, however, information structure driven models do not successfully capture LP-movement. First, they fail to account for important differences between whmovement and LP-movement (see also Horvath 2007). LP-movement of arguments is clausebound in most dialects of German, while wh-movement is not (Kvam 1983:81). Whmovement is obligatory in both Czech and German, while topic and focus fronting to Spec,CP are optional, cf. (7c-d). More importantly, wh-movement rarely applies optionally in the world’s languages, while the optional nature of focus or topic fronting is the rule rather than the exception. Where focus fronting seems obligatory – as in Hungarian – it goes hand in hand with additional semantic properties such as exhaustivity. Arguably, this additional semantic property rather than focality is responsible for the obligatoriness of movement (Horvath 2007). Second, reference to focus/topic features violates the inclusiveness condition (Chomsky 1995): only those features can figure in syntactic computations that represent properties of lexical items. On obvious grounds, being a focus or a topic is not a lexical property – words and phrases can be classified as such only when used in a specific context. ‘Topic’ and ‘focus’ are not likely candidates for features figuring in syntactic computations (see Brunetti 2004 and Neeleman & Szendrői 2004 for related arguments). Third, the island and binding data discussed by Wold (1995), Rooth (1996), and Alboiu (2004) imply that there is no covert focus movement equivalent to covert wh-movement.5 If the overt movement of focus phrases was driven by a focus feature in the same way as whfeatures drive wh-movement, we would be left without an explanation for the fact that one kind of movement has a covert counterpart, while the other one does not. The optionality of focus and topic fronting implies the fourth difficulty. The slot preceding the second position must always be overtly realized in German, and in Czech at least in the presence of second position clitics. When neither the focus nor the topic move to first position (because these movements are optional), other elements go there. Based on a suggestion of Hubert Haider, Fanselow (2002) and Frey (2005) argue that it is the leftmost element of TP that undergoes LP-movement in such cases, see also Müller (2004) for a proposal with a similar consequence. This accounts for the option of fronting full lexical subjects6 – (8) and high adverbs – (9)-(10), but not direct objects7 – (11), in out of the blue contexts, i.e. when (8)-(11) answer questions like What’s new/What happened. Only (8)-(10) involve fronting of the highest constituents in TP.8 We will follow Frey (2005) in calling this construction formal fronting (FF). In the absence of a pragmatic marking of the leftmost elements in (8)-(10), FF-constructions cannot involve the checking of or agreement with 5

See Drubig (1994) for argumentation in favour of island-sensitivity of focus, and Reich (2003) and Horvath (2007) for counterarguments. 6 Craenenbroek & Haegeman (2007) show that subject-initial verb second clauses cannot be analysed as TPs. 7 But see below for possible exceptions. 8 Czech additionally allows fronting of verbal participles across the second position clitics (in (i) it is the reflexive si and the auxiliary) in sentences without overt subjects or high adverbs: (i) Chytila jsem si zajíce. caught.Sg.fem aux.1Sg refl. rabbit.acc ‘I caught myself a rabbit.’

4

topic/focus features. (8)-(10) show that there is an operation attracting XPs to the left periphery independent of their informational status, and if we want to maintain the idea that (6) and (7) involve topic/focus movement in a strict sense, FF-constructions force us to postulate two different processes filling Spec,CP, clearly an inelegant solution.9 (8)

a. Ein Kind hat einen Hasen gefangen. a.nom child has a.acc rabbit caught b. Nějaké/Jedno dítě si chytilo zajíce. a/some child refl. caught rabbit.acc ‘A child has caught a rabbit.’

(Ge)

a. Wahrscheinlich hat ein Kind einen Hasen gefangen. probably has a child a.acc rabbit caught b. Asi si nějaké/jedno dítě chytilo zajíce. probably refl. a/some child caught rabbit.acc ‘A child has probably caught a rabbit.’

(Ge)

(10) a. Heute früh hat ein Kind einen Hasen gefangen. today early has a child a.acc rabbit caught b. Dneska ráno si nějaké/jedno dítě chytilo zajíce. today morning refl. a/some child caught rabbit.acc ‘A child has caught a rabbit today in the morning.’

(Ge)

(11) a. *Einen Haseni hat (wahrscheinlich) ( heute) ein Kind ti gefangen. a.acc rabbit has probably today a child caught b. *Zajícei si (asi) ( dneska) nějaké/jedno dítě chytilo ti. rabbit.acc refl. probably today) a/some child caught

(Ge)

(9)

(Cz)

(Cz)

(Cz)

(Cz)

While FF-constructions have figured in the discussion of LP-movement since Travis (1984), there is a further, less known set of data that also defies an analysis in terms of focus/topic fronting. We will discuss the phenomenon in detail in the following sections 3-4. Here, it suffices to note that accented objects and other categories can undergo LP-movement in wide focus contexts as induced by questions What’s new/What happened, if they do not cross another accented category. The fronted elements in (12) are neither a topic nor the semantic focus, what is fronted is part of the semantic focus. (12) What’s new?/What happened? a. [ Einen Hasen]i habe ich ti gefangen. a.acc rabbit have I caught b. [ Einen Hasen]i hat wer ti gefangen. a.acc rabbit has someone caught c. Zajícei jsem chytil ti. rabbit.acc aux.1Sg caught d. Zajícei prý kdosi chytil ti. rabbit.acc prt. somebody caught ‘I caught a rabbit’/‘Somebody has caught a rabbit.’

9

(Ge)

(Cz)

In a split CP model, the FINiteness head accepting pragmatically unmarked specifiers might capture such data.

5

We follow the common practice of identifying the focus of an utterance with that part of the utterance that corresponds to the wh-phrase in a congruent constituent question.10 Thus the semantic focus in an answer to the question What did you buy is the DP corresponding to what. Wh-questions like What does/did X do and What’s happened/What’s new invoke VPand IP-focus, respectively.11 (12) constitutes a further mystery for the standard theory of LP-movement. How can the fronting of just some part of the focus fulfil an agreement requirement for a focus feature? Remnant movement and scattered deletion cannot solve the problem, as we shall see in section 5. How can the role accentuation plays be captured within the focus movement theory? Furthermore, given that all focus constituents are accented, could focus movement be considered a special case of the operation illustrated in (12)? 3.

Subpart of focus movement

Before we develop our account of LP-movement, a close look at the data illustrated in (12) is necessary. (13)-(14) show the fronting of a direct object and a prepositional object in sentences with VP- and IP-focus, respectively. The fact that focus fronting does not always affect the whole focus constituent has not gone unnoticed in the literature on German (cf. Büring 1997:46, 72, Gärtner 1996, Jacobs 1991,12 Krifka 1994: 145-6, Fanselow 2004), exactly the same was observed for Czech by Lenertová and Junghanns (2004, 2007). The fronted element is part of the focus, so we will label the construction ‘subpart of focus fronting’ (SFF). (13) What did he do? [ Ein BILD]i hat er ti zerrissen. a picture has he torn ‘He tore a picture.’

(Ge)

(14) What happened? [ Im GRAben]i ist er ti gelandet! in ditch has he landed ‘He drove into the ditch!’

(Ge)

However, SFF is not limited to VP- or IP-focus, we can observe subpart-of-focus movement out of smaller focus domains like a DP-focus. Thus, the context question (15a) imposes narrow object (DP-) focus on possible answers, yet it is compatible with answers such as (16 a,c)13, in which only a part of the DP was fronted, with the determiner being stranded. Importantly, all the answers in (16) are congruent with both context questions in (15). (15) a. What have you bought? b. What have you done?/What’s new? 10

Depending on the context, the semantic focus may include given (unaccented) elements. Crucially, the constituent corresponding to the semantic focus must contain the main prominence of the clause. See Büring (2006b, 2007), Schwarzschild (1999), and Reich (2003:ch2), among others, for a discussion. 11 DP-focus is also referred to as ‘narrow’ and VP/IP-focus as ‘wide’/‘broad’ focus throughout the paper. Note that this terminology is only descriptive and is not meant to correlate with distinctions like ‘informational’ vs. ‘contrastive’ focus (cf. Kiss 1998). The latter distinction is orthogonal to the discussion of focus in this paper. 12 As pointed out by an anonymous reviewer, Jacobs (1991), quoting Susanne Uhmann (p.c.), speculates that such constructions presuppose that the relation between the object and the verb is ‘typical’ (like in ein Bild malen ‘to paint a picture’). We think that felicitous examples such as (13a) speak against this view. 13 In (16a,c), the traces only indicate the base-position of the moved element and we leave it aside how the derivation proceeds, e.g., whether the fronted phrase undergoes remnant movement, in which case the bracketing would be different.

6

(16) a. BÜcheri hab’ ich mir [ ‘npaar ti] gekauft.14 books have I refl. a few bought b. [‘N paar BÜcher]i hab’ ich mir ti gekauft. jsem si [pár ti] koupila. c. KNÍžeki book.gen.pl aux.1sg refl. a few bought.sg.fm d. [Pár KNÍžek]i jsem si ti koupila. ‘I have bought a couple of books.’

(Ge) (Cz)

Thus, a complex picture of the interaction between semantic focus and movement emerges. With both context questions in (15), the displaced category is smaller than the semantic focus in (16a,c). Even when the whole DP is fronted, as in (16b,d), the displaced category is smaller than the semantic focus in the context (15b). The fronted category is identical with the semantic focus (this is the situation analyzed as attraction of a focus feature in many alternative accounts) only in the case of (16b,d) in the context of (15a). In the following subsections, we will first argue that SFF cannot be successfully explained away by postulating another pragmatic feature being borne by the fronted element (3.1). Then, we show that SFF is subject to locality restrictions concerning accentuation (3.2) and finally argue that, in spite of this generalization, SFF is a syntactic movement (3.3). 3.1

Pragmatic properties of SFF

In an SFF construction, the element in Spec,C is not the semantic focus and therefore does not agree with C for a focus feature. In order to maintain the idea that Spec,C and C agree for some pragmatic feature that is responsible for fronting, one might attempt to identify an additional information-structural property that might be made responsible for the SFF. E.g., one could analyse the preposed XP as a ‘topic in a focus’, or assume that particular ‘salience’ is attributed to the left-peripheral part of the focus. In general, such suggestions are not really helpful, however. The unspecific indefinite ein Buch ‘a book’ in (17) is neither a likely candidate for an aboutness topic, nor for a contrastive topic. Furthermore, no particular salience is attributed to ‘a book’ (while ‘book-reading’ may be particularly focused or contrasted). (17) What do you want to do in your holiday? [ Ein Buch] würde ich gerne ti lesen. a book would I eagerly read ‘I would like to read a book.’

(Ge)

SFF movement also cannot be explained away if we try to identify the additional pragmatic property not with respect to preceding text (as notions such as topic and focus do) but with respect to the effect the fronting might have on the following text (for the importance of such a perspective, see the Centering Theory of Walker, Joshi, and Prince 1998), such as preparing a topic shift. In a dialogue such as (18), the fronted non-specific DP is neither a backward looking topic or the focus of the utterance, nor does it fulfil any function for the following text. Nevertheless, the sentence with the fronted DP is perfectly well-formed in this context.

14

See also Puig-Waldmüller (2006) for similar data. The above examples are a serious challenge for the distributed deletion account of discontinuous noun phrases (Fanselow & Ćavar 2001, 2002) because the two parts of the DP have the same pragmatic function. Given that the base-generation of a focal phrase in Spec,CP is quite unlikely (at least, it has never been argued for in the literature on discontinuous DPs), our examples show that the discontinuity of DPs sometimes arises as a consequence of SFF, but this in not true in general, see. e.g., section 5.2.2.

7

(18) What had you wanted to do in your holiday? [ Ein Buch]i hätte ich gerne ti gelesen, aber ich musste immer Windeln wechseln. (Ge) a book had I eagerly read but I had-to always diapers change Und dann wurde das Kind noch krank... and then became the child even ill ‘I had wanted to read some book, but I always had to change diapers. And then the child even fell ill.’ A second argument against the idea that the left peripheral element in an SFF construction bears a pragmatic function of its own can be derived from the fact that, by their very nature, concepts such as topic, focus, salience, etc. are applicable only to elements with semantic content. Nevertheless, idioms can be split by SFF, in spite of the fact that the fronted parts of the idiom are by themselves meaningless. It is only the idiom as a whole, not its individual parts, that bears an interpretation. Since the moved element in such cases is meaningless, it cannot be a topic, a focus, or be particularly salient. Examples such as (19) refute the idea that SFF movement is triggered by a pragmatic property of the moved item in a direct way.15 (19) Why did you quarrel with him? a. BOUdui na mě ušil ti! hut.acc for me.acc stitched.sg.ms ‘He has cheated me!’ b. MÁSloi má ti na hlavě! butter has on head.loc ‘He is corrupt!’ c. [ Schöne AUgen]i hat er ihr ti gemacht. beautiful eyes has he her made ‘He made eyes at her.’ d. [ Den GARaus]i hat er ihr ti gemacht. the garaus has he her.dat made ‘He killed her.’

(Cz) (Cz) (Ge) (Ge)

Moreover, prosodic changes are the primary expression of information structure in Czech and German. If the basic distinctions of information structure were applicable to parts of idioms, we would expect that parts of idioms can be optionally deaccented, which is not the case. Prosodic facts thus corroborate the view that parts of idioms are meaningless and as such incompatible with distinctions of information structure. Still they can be fronted in SFF constructions.16 We conclude that the fronting attracts a part of the semantic focus and that it is not plausible to postulate an additional pragmatic feature associated with the fronted element. 3.2

Formal aspects of SFF

15

Frey (2005) points out that some idioms have a quasi-compositional structure (Nunberg et al. 1994) relative to which parts of idioms could have an informational value of their own. However, Horne (2003) has shown that thematically non-compositional idioms are transparent for movement in German as well. 16 Controlled experiments with written (reported in a separate paper) and auditory (Féry 2007) material reveal that the difference between a narrow focus and an SFF context has no influence on the acceptability of object initial sentences. Native speakers sometimes characterize SFF constructions as being more ‘emphatic’ than their narrow focus counterparts, but this emphasis affects the predicate as a whole and never the fronted part of the predicate alone. SFF constructions may come along with specific pragmatic functions, but these do not pertain to the fronted element, but the category the fronted element was extracted from.

8

The most remarkable property of SFF is a locality restriction concerning accentuation: it is always the leftmost accented XP17 that undergoes SFF movement, this element cannot cross another accented element. Provided this condition is fulfilled, not only direct and prepositional objects but all clausal constituents may figure as the displaced subpart of focus. The restriction comes out clearly in idiomatic sentences. In idioms that involve (at least) two phrasal parts, like (20), it is always the leftmost XP that is fronted, as observed by Müller (2003). Compare (20b) and (20c). (20) a. Er hat den NAgel auf den KOPF getroffen. he has the.acc nail on the.acc head hit ‘He clearly expressed the truth.’ b. [Den NAgel]i hat er ti auf den KOPF getroffen. c. #[Auf den KOPF]i hat er den Nagel ti getroffen.

(Ge)

As already mentioned, there is no deaccentuation licensed by context for parts of idiomatic expressions (only the whole idiom can be affected). Importantly, fronting of the leftmost accented element in cases like (20a) or (21a) does not cross accented material (see below for cases with non-pronominal subjects) and SFF is felicitous (20b)-(21b). In contrast, the fronting of the lower accented element in (20c)-(21c) leads to the loss of the idiomatic reading and to inacceptability,18 because the crossed element would have to be deaccented for crossing to be grammatical. (21) a. Er ist vom REgen in die TRAUfe gekommen. he has from rain in the eaves come ‘He jumped out of the frying pan into the fire.’ b. [Vom REgen]i ist er ti in die TRAUfe gekommen. c. #[In die TRAUfe]i ist er vom Regen ti gekommen.

(Ge)

Note that the ‘primary’ accent of the clause goes to the second rather than the first accented DP/PP in such constellations in German, (cf., e.g., Truckenbrodt (2005) and the references there). This means that the right PP is the ‘focus exponent’ in (20)-(21) according to many theories of focus. The examples in (20)-(21) thus show that the element fronted in SFF construction is not necessarily the focus exponent (in contrast to claims of Fanselow 2004 and Lenertová & Junghanns 2004, 2007, a point also made by Frey 2005). The crucial property is rather that no other accented category be crossed. PP objects can thus be moved across direct objects when the latter are contextually given and deaccented as in (22). (22) What have you done with the plates? [ In die KÜche]i hab ich sie ti gebracht. in the kitchen have I them brought ‘I brought them in the kitchen.’

17

(Ge)

The leftmost accented XP is normally also the structurally highest XP with an accent. The model developed in section 4. implies that accented XPs must c-command and precede YP for exerting intervention effects on the leftward movement of YP. Given the strong correlation between hierarchy and linear order, it is very hard (if not impossible) to find constructions in which a phrase following ZP c-commands ZP that would allow us to check whether intervention conditions are adequately characterized by the conjunction of precedence and c-command. 18 Some speakers rate the b-versions of (20)-(21) as less acceptable than the neutral a-versions. However, they accept the idiomatic reading of the b-versions, in a contrast to the c-versions, which no speaker can interpret idiomatically.

9

Selkirk (1995) and Kratzer & Selkirk (2006),19 show that accents are obligatorily distributed on all arguments that are not discourse given (see also Féry & Kügler 2006 for deaccentuation of arguments as a means of givenness-marking). Therefore, accented higher arguments block the SFF movement of lower arguments in wide focus domains. Transitive verbs in wide focus, on the other hand, are unaccented20 (see also Jacobs 1999, Wagner 2005, a.o.), thus they do not block SFF even when they c-command the direct object, as is the case in Czech (but not in German). The SFF examples with fronted objects used so far involved subjects in the form of unstressed or silent pronouns, as in (23a). In contrast, (23b) with fronting across a full lexical subject is not felicitous in a wide focus context. (23) What’s new? a. GUláši jsem uvařila ti. goulash aux.1sg cooked.sg.fm ‘I cooked goulash.’ b.#GUláši matka uvařila ti. goulash mother cooked.sg.fm c. Matka uvařila GUláš.21 mother cooked.sg.fm goulash

(Cz)

Now consider the Q-A pair in (24), with a context imposing narrow focus on the fronted element. Here, object initial answers are acceptable: (24)

What did mother cook? matka uvařila ti. GUláši goulash mother cooked.sg.fm

(Cz)

The minimal difference between (23) and (24) stems from the contextual status of the intervening subject: the wh-question in (24) renders the subject given,22 thus it need not be accented and does not count as a prosodic intervener for fronting of a lower accented element. Indefinite subjects like ‘somebody’ (25) and epithets (26) (which are generally not accented, see Ladd 1996:174-204 and Truckenbrodt 2005:16, among others), support our generalization: they do not block SFF. (25) a. What’s new? hledal ti. KARlai někdo Karel.acc somebody.nom looked-for.sg.ms ‘Somebody was looking for Karel.’ b. How was the party yesterday? [ Ein BUCH]i hat jemand ti vorgelesen. a book has somebody read out ‘Someone read out a book.’

(Cz)

(Ge)

(26) What did Fritz do on Sunday? 19

For supporting experimental evidence on German, see Truckenbrodt (2004). Rhythmical accents possible on verbs do not count here, see also section 4. 21 Note that the paradigm of past auxiliaries in Czech has a null form for 3.Person Sg. and Pl. The question arises whether sentences like (23c) are to be analysed as CPs or smaller structures. As the discussion would go beyond the scope of this paper, and the issue is orthogonal to our subject, we leave it open. 22 The question What’s new with mother?, imposing wide focus on the answer but rendering the subject given, would have the same effect. 20

10

[Ein BUCH]i hat der Idiot ti gelesen anstatt schwimmen zu gehen. a book has the idiot read instead swim to go ‘The fool read a book, instead of going swimming!

(Ge)

Deaccentuation of an overt subject within broad focus is felicitous if a rich enough context enables its accommodation. (27) is a possible case of SFF across an overt subject: a party with very loud music evokes an expectation of annoyed neighbours.23 On the other hand, the same sentence as an out-of-the-blue utterance (e.g., at the beginning of a phone-call) would sound odd, cf. (28). (27) A: You know Peter was giving a party yesterday and he prefers a very loud music. B: So what happened? [ Den STROM]i hat uns der Nachbar ti ausgeschaltet … (Ge) the.acc electricity has us the neighbor switched-off ‘Our neighbor has switched off the electricity.’ (28) What’s the matter? Why are calling me so late at night? a.#Den STROM hat uns der Nachbar ausgeschaltet … the.acc electricity has us.dat the neighbor switched-off b. Der NACHbar hat uns den STROM ausgeschaltet.

(Ge)

The following examples illustrate wide focus with a fronted accented subject. (29) a. Why did you do that? MATkai mi to ti poručila. mother.nom me.dat it ordered.sg.fm ‘Mother ordered it to me.’ b. That is a really nice sweater. ANtjei hat mir den ti geschenkt. Antje has me.dat it.acc given ‘Antje gave it to me as a present.’ c. What happened? [ Eine KRANKENschwester]i hat ti einen PaTIENten getötet. a nurse has a.acc patient killed ‘A nurse killed a patient.’

(Cz)

(Ge)

(Ge)

In (29a-b), the subject is the only available accented category. SFF movement of the subject is acceptable in the presence of further accented material as in (29c), too, because the subject is the leftmost accented element. Its SFF movement crosses no accented phrase. In the case of accented subjects as in (29c), the outcome of SFF movement is identical with the outcome of formal fronting, for obvious reasons. The accentuation condition on SFF leaves only one option for verbs to undergo SFF. As mentioned above, even focused verbs are unaccented and we observe that their fronting is not felicitous in wide focus contexts – (30b) is a pragmatically odd answer to the question in (30). However, when both the subject and the direct object are weak pronominals or unaccented indefinites, the verb receives the main accent and can undergo SFF, as it crosses no accented material on its way up to Spec,CP – (31). (30) What did you do on Sunday? 23

Speakers vary in their judgments of structures comparable to (27). We attribute this variation to differences in the ability and willingness to ‘confabulate’ contextual properties that license the deaccentuation of the subject.

11

a. [Den WAgen]i hab’ ich ti gewaschen. the.acc car have I washed ‘I washed the car.’ b.#Gewascheni hab’ ich den WAgen ti.

(Ge)

(31) a. What has happened last Sunday? VerLETZTi hab ich mich ti. hurt have I myself ‘I hurt myself.’ b. Why did they fire him? BeLEIdigti hat er wen ti. insulted has he someone.acc ‘He insulted someone.’ 3.3

(Ge)

SFF is a syntactic movement

Although the crucial locality restrictions on SFF is based on accentuation, we will show that the fronting cannot be excluded from the syntax proper and analysed as a post-syntactic, PFoperation. Concerning models in which PF movement and syntactic movement differ only in terms of whether LF has access to their outcome (Holmberg 2000), we would not gain much from analysing SFF movement as a PF process. The special role accentuation plays for SFF movement would by definition not be explained any better than by a syntactic movement operation.24 In other models, PF movement differs substantially from syntactic movement. E.g., head movement was taken to be PF movement by Chomsky (1998, 2000) because it does not respect the extension requirement. However, with the exception of the accent-sensitive intervention effect, SFF behaves like a normal A-bar-movement operation. E.g., in contrast to Stylistic Fronting (Holmberg 2000), SFF is not clause-bound but enters long distance dependencies. (32) a. What has she been doing there so long? [Das AUto]i denk ich hat sie versucht the car think I has she tried ‘I think she has tried to repair the car!’ b. And what happened then? [Ein TAxi]i hat sie gesagt dass sie sich ti a taxi has she said that she refl. ‘She said she would take a taxi.’

zu ti reparieren! to repair

(Ge)

nehmen wird. take will

SFF is also island sensitive. In Czech, topic and (narrow-)focus fronting (in contrast to whmovement) is compatible even with weak islands (cf. Meyer 2004: 188-195). SFF movement is not blocked by weak islands either, as illustrated in (33) with a wh-island and in (34) with a factive-island. On the other hand, SFF movement obeys strong islands like adjunct clauses (35). (33) Why are you so angry? 24

Furthermore, SFF would have to be distinguished from other movements to the first position in German that cannot be PF processes. Empty topics (i) (see Cardinaletti 1990) can occupy the 1st position of German, while there is a consensus that the possession of a phonetic matrix is a crucial visibility condition for PF processes. (i) ∅ Hab ich schon gesehen! ‘I already saw that.’ have I already see

12

Ále, [ jeden blbej FORmulář]i nevím, jak mám vyplnit ti. prt one stupid form not-know.1sg how shall.1sg fill-in.inf ‘I don’t know how to fill in one stupid form.’

(Cz)

(34) If we were moving, it would have to jolt. A [ VZDUCH]i bysme cejtili, že rozrážíme ti ! and air would.1pl feel.pl that cut.1pl ‘And we would feel that we cut through the air!’ (35) Are you not worried that Peter still does not know what he really wants? a. *ŠKOlui se rozhodne, až dokončí ti! school.acc refl. decide.3sg.fut when finish.3fut ‘He will decide himself when he has finished the school!’ b. Rozhodne se, až dokončí ŠKOlu! decide.3sg.fut refl. when finish.3fut school.acc

(Cz)

Finally, parasitic gaps lend further support to a syntactic analysis of SFF movement. Parasitic gaps cannot be constructed for Czech, but German infinitivals as in (36) are often analysed as parasitic gap constructions. (cf. Bayer & Kornfilt 1994, Felix 1985, Webelhuth 1989).25 (36) Sie hatte diesi ohne ei gelesen zu haben ti zurückgewiesen. she had this without read to have rejected ‘She had rejected this without reading it.’

(Ge)

(37) shows that parasitic gaps can be licensed by SFF movement. There is no reason to doubt, then, that SFF movement is a proper syntactic process. (37) What did he do? AKteni hat er [anstatt ei zu bearbeiten] ti in den Papierkorb geworfen. (Ge) documents has he instead-of to process in the basket tossed ‘He threw documents into the basket without having processed them.’ We are thus confronted with a syntactic movement that cannot be analysed as an attraction of a syntactic focus (or other pragmatic) feature (see below for a discussion and rejection of possible feature-checking accounts of SFF) and shows locality restrictions sensitive to accentuation. In the following section, we will propose an analysis of left-peripherymovement that explains the properties of SFF. 4.

A Linearization-based Analysis of SFF

4.1 Architecture of the grammar and SFF The primary way of focus realization in languages like Czech and German is prosodic prominence. In syntactic structures, focus is standardly represented via F-markers26 which are freely assigned to syntactic nodes (cf. Jackendoff 1972). F-markers are not formal features introduced on lexical items, and thus not in active in the syntactic derivation (see also Horvath 2007 for a discussion). They were designed as tools for grammatical model 25

Fanselow (2001), Haider & Rosengren (2003), and Kathol (2001) propose to not analyse (36) as a parasitic gap construction. 26 See also Féry & Samek-Lodovici (2006) and Selkirk (2006) for the proposal of G(ivenness)-marking in addition to F-marking.

13

formulation at a time when no direct communication was possible between LF and PF (see Chomsky 1981), so that the information whether some category is accented or not had to be transported through syntax to the interpretive component of grammar. In a minimalist approach, the syntactic, prosodic, and semantic derivations operate in a quasi-parallel fashion, being synchronised with the completion of phases. Lexical items, and, therefore, also syntactic options are triplets specifying phonological, syntactic, and semantic information. The information whether a certain category K dominates a node that is accented can thus, in principle, be read off the syntactic tree when phases are interpreted, so that the status of K as being (potentially) in focus can be computed on the basis of the presence and location of accents in K, taking into account interface constraints (see Büring 2007 and Schwarzschild 1999). In principle, F-marking can thus be dispensed with (see also Szendrői 2001 for a discussion), but for the ease of exposition, one can assume that accented elements are F-marked, and that there is a component of ‘focus projection’ (Selkirk 1995) that allows the percolation of F-marking in the syntactic tree. We assume an interpretive approach to focussing in which accentuation applies freely in the course of the syntactic derivation. The freely assigned prosodic shape of a sentence then determines –as expressed via F-marking- which pragmatic contexts the sentence is compatible with. The differentiation between default structural accent (phrase stress) and pragmatically determined accent in sentences containing contrastive focus and given elements is crucial for the proposed account of movement restrictions in SFF contexts. The most notable property of SFF movement - that it can affect the leftmost accented phrase only - cannot be accounted for by a direct reference to accentuation restricting the syntactic movement. The intervention restrictions will be analysed as a side-effect of linearization, the syntactic property underlying default/structural accentuation. In other words, the locality constraint on SFF will be attributed to restrictions on ordering statements in linearization. The other syntactic ingredients of our analysis of SFF movement have been introduced in the preceding discussion. The first position of a Czech and German clause has to be filled, but the choice of the element that goes there is relatively free in declaratives: the leftmost accented element can go to Spec,C (SFF and focus movement), elements situated at the edge of TP can appear there (FF-movement), and so can topic phrases. We conclude from our discussion that at least SFF and FF movement is altruistic, in the sense that the displacement does not satisfy any requirement for the moved phrase. Specifically, we assume that SFF movement is triggered by an edge feature of C (Chomsky 2005) requiring a filled Spec,CP in all matrix and some embedded clauses. 4.2

Intervention effects in SFF and Cyclic Linearization

Our account of the intervention effects with SFF is based on insights of Fox & Pesetsky (2003, 2005), who propose that linearization is coded in the form of ordering statements X > Y established in a phase-related fashion. Crucially, once an ordering statement has been created, it cannot be altered or abandoned later. If the derivation cannot proceed but by entering an ordering statement incompatible with previous ones, the derivation crashes. Fox & Pesetsky show that such a linearization model explains intervention effects for a number of constructions such as Scandinavian Object Shift. The linearization model of Fox & Pesetsky (2003,2005) can be modified such that it captures SFF movement in Czech and German. Fox & Pesetsky (2003, 2005) share the standard minimalist assumption that Spellout is a cyclic process. They argue that there are Spellout Domains such as CP, roughly corresponding to phases. Each time a Spellout Domain has been constructed, a complete linearization of the phrases included in that domain must be computed. The leading concept of cyclic spellout is compatible with the view that ordering statements may (in principle) be entered at any point in the derivation. In principle, UG leaves it open 14

when two phrases are linearized, but the derivation cannot proceed beyond a Spellout Domain (a phase) unless the material in that Spellout Domain has been completely linearized. This is the version of cyclic Spellout that we will (crucially) assume in this paper. Once α and β have been linearized, their relative order can no longer be changed. In order to prevent an accented category α from crossing a further accented category β, it must only be made sure that accented phrases are linearized before they move. The assignment of a structural accent indeed presupposes linearization. After all, the relative prosodic prominence of a category depends on whether it sits on a left or a right branch, in a language- and category specific fashion. (38) is our proposal for capturing this connection. XPs can be linearized at any point, and when they are linearized immediately (directly after merge, but see below), structural (= noncontrastive, etc.) stress assignment is an automatic consequence. Likewise, structural accents have to be assigned in the context of immediate linearization. The failure of immediate linearization (and consequently the absence of stress assignment) may be considered the structural correlate of givenness (but see below).27 (38) Structural accents are assigned in the context of immediate linearization. The connection that (38) establishes between stress assignment and immediate linearization implies the intervention effects for SFF movement in an obvious way. Crossing movement is not ruled out as such, since the MLC is no longer considered a condition on syntactic movement (Chomsky 2005). Suppose, that Comp possesses an edge feature in (39), and that α and β are structurally accented. (39) [Comp … [W … [α ... [Z … [… β …] … ] …] …] …] Because of their accentuation, (38) implies that α and β have been linearized immediately when they were merged. In particular, this means that an ordering statement α > Z has been created. Z dominates β, and because β has already been linearized in Z (because it is structurally accented), the ordering α > β is implied, which must not be changed later. Consequently, α can move to Spec,C (because this leaves the ordering statement α > β intact), but β cannot move across α. Suppose, on the other hand, that α in (39) does not bear a structural accent. (38) then implies that α has not been linearized immediately. In principle (if α is not included in a Spellout Domain different from CP) the linearization of α need not happen before β is attracted (only then will the Spellout Domain CP be completed), in which case there will be no linearization statement α > β that could prevent β from crossing α. The movement of accented β across unaccented α is therefore not ruled out, as required. The basic distinction we observe in SFF movement is thus derived: an element can be fronted only if it is the leftmost accented category. Let us flesh out the system we propose in more detail. We assume that linearization is immediate if it applies directly after (or as a side-effect of) an external merge operation. This seems to conflict with the idea that phrases are to be spelt out in the context of feature valuation (or at some point after valuation), see Epstein & Seely (2002) for a discussion. Consider subjects of unaccusative verbs in this respect. They are merged as complements of 27

(38) requires that all non-given XPs will be accented when they are merged. In that respect, the goal of (38) differs from a number of proposals which try to explain accentuation in a phase-based model (Adger 2006, Ishihara 2006, Kahnemuypour 2004, Kratzer & Selkirk 2006). These approaches attempt to identify the location of the ‘sentence accent’, the most prominent structural accent of a clause. Since sentence accents play no particular role in SFF movement (it is the leftmost accent that moves, not the prosodically most prominent one), we need not discuss here which of these models is best compatible with our approach.

15

V, but their Case depends on agreement with T, so they cannot be spelt out completely before T is merged. However, the subject rather than the verb is stressed in wide focus unaccusative constructions, see (41), in contrast to unergative constructions such as (42). The explanation of this difference implies that unaccusative subjects are accented in the position of β rather than α in (40) (cf. e.g., Adger 2006). (40) [TP α T [ … [V β]] (41) What’s the matter? a. The BABY’s disappeared. b. *The baby’s DISAPPEARED. (42) a. The BABY cried. b. The baby CRIED. The demands of immediate linearization and spellout in the context of/after complete valuation are compatible, however. After all, the determination of linear order and accents is just one aspect of spellout, that can precede the computation of the final phonological realization. In particular, the presence of linearization statements does not exclude further featural valuation in an obvious sense. As in the original proposal of Fox & Pesetsky (2003, 2005), the size of the Spellout Domains determines which nodes must be targeted as cyclic movement steps. An accented and thus linearized phrase β can cross α in (39) as long as β moves before α has been linearized relative to β. The structural accentuation of α linearizes α relative to β, as required. However, α must always have been linearized before the derivation proceeds beyond W, if W is a Spellout Domain. In that case, β must move to the edge of W before α is linearized, i.e., it must undergo cyclic movement. Note that this cyclic movement step helps to overcome intervention effects of α only if α is not structurally accented. So far, we have confined our attention to arguments (and adjuncts). Verbs play a special role because they are usually unaccented even when they are [-given]/F-marked, as remarked above. There, we also saw that verbs do not block SFF-movement of an argument. In an SOV language like German, such a blocking effect is not to be expected in any event, because the verb follows its arguments and is thus unable to trigger intervention effects. Things are different in the SVO language Czech: the verb precedes its object (43b), so that the mobility of XP indeed depends on the status of V28. (43) a. [XP V] b. [V [XP]]

(Ge) (Cz)

For our purposes, it suffices to assume that verbs cannot participate in immediate linearization, The failure of immediate linearization also explains why verbs move easily across accented arguments29 and adjuncts, e.g., in a verb second construction which requires that the German vP (44a) be eventually mapped onto (44b). If den Hund and sah were 28

The failure of the verb to trigger intervention effects for SFF movement in Czech shows that SFF movement cannot be recast in terms of a feature [-given]: the verb is [-given] in wide focus contexts, but it can be crossed by accented objects. The category that can go to the left edge of CP in SFF constructions is correctly identified by accentuation only. 29 Once V is ordered in Comp or Tense, no element in TP and vP can be moved across V to Spec,CP and Spec,TP, respectively. UG allows the serialization of the finite verb after the specifiers of TP and CP have been filled (or after the insertion of an expletive), and for most cases, this is the only derivation that does not crash. See below for the role Spellout Domains play for verb movement.

16

ordered with respect to each other when merged, the verb could never move across the direct object.30 (44) a. [vP er [VP [ den HUND] sah]]] he the dog saw b. Er [sah [VP den HUND tV]. ‘He saw the dog.’ All XPs can be accented in a (wide focus) construction, but they do not have to be so. The absence of an accent on a phrase is visible to, and can be made sense of, by the interpretive component, which can take unaccented XPs to be given. Furthermore, the interpretive ‘focus projection’ can also apply in the usual fashion – with a phrase being a potential focus if it contains an accented word in certain structural constellations allowing it to be maximally prominent (see Büring 2006b for a discussion of focus projection). Our model thus adheres to current explanations of the relation between accentuation and focus (projection).31 4.3

SFF vs. other types of movement

Phrases with a structural accent cannot cross each other in our model. This captures SFF movement in a direct way. For the interaction of accented phrases, our proposal has much in common with the analysis proposed in Fox & Pesetsky (2003, 2005): there is no crossing movement when the phrases have been serialized relative to each other before movement. However, in our model, accented phrases are serialized immediately, so they cannot even escape each other’s intervention effects by moving to the edge of a Spellout Domain. How is (structurally) unaccented material treated in our model? Such syntactic objects are not serialized immediately. The absence of immediate linearization gives them some freedom of movement, and prevents them from acting as intervenors for the movement of material already linearized. This greater flexibility distinguishes our model from the proposal of Fox & Pesetsky. Unaccented material must eventually be serialized as well, and the fact that spellout is bound to certain domains excludes the idea that unaccented XPs are linearized at the end of the complete derivation of a sentence. We will call linearization of material that has not undergone immediate serialization ‘secondary linearization’. Let us discuss three relevant subcases in turn: (a) wh-words and scrambling, (b) FF-movement, and, (c) the movement of given and of contrastive elements. 4.3.1 Wh-movement and Scrambling Wh-words like who or what (and their German and Czech counterparts) are typically unaccented when they go to Spec,CP (Kahnemuyipour 2004). Immediate linearization is 30

This conclusion seems to lead us into a paradoxical situation. It must be possible to immediately linearize the complement of V (e.g. in a wide focus context), but the verb must not be immediately serialized in VP. It is unclear, then, what the object is ordered with when it enters ordering relations in [V V DP]. There are several solutions for this problem. We propose that the linearization of α in a tree T means that α is entered into the set S of elements ordered in T. If S has a single member only, ordering statements are trivial. Once further elements are entered into S, they must be ordered relative to the other members of S. 31 By distinguishing XPs that are serialized immediately from those that are not, our account resembles the theory of relativized linearization proposed by Müller (2007). In his model, ‘merge status’ matters: Only the elements with the same ‘merge status’ are ordered relative to each other when they are merged. A syntactic tree thus can contain elements that are not yet ordered relative to each other. Our account shares this property, but does not employ the notion of a merge status. Elements that are not immediately linearized are unordered with respect to each other (up to a certain point). Therefore, our notion of linearization is also rather ‘partial’ than ‘relative’. When α is ordered in T, it is ordered relative to all elements in T that have been ordered so far.

17

optional, which means that wh-phrases do not have to be linearized directly after being merged. If they fail to be so, the interpretive component will so far mark them as [+given] because they are not accented, which is fatal given their meaning. If wh-phrases in Spec,CP are, however, exempt from being interpreted as given (because they are in a local agree relation with a [+wh]-feature overriding the accentuation-based interpretation), they need not be structurally accented, as required.32 Because the wh-phrase moving to Spec,CP is unaccented and thus not serialized immediately, its movement is not blocked by intervention effects from accented XPs, as long as the wh-phrase has not itself been serialized. Such a flexibility of movement is required for (45) and similar sentences, but it is not guaranteed in the system presented so far. (45) Wen mag jeder twh? who.acc likes everyone ‘Who does everyone like?’

(Ge)

The difficulty stems from the assumption of Fox & Pesetsky (2003, 2005) that Spellout Domains are linearized completely once they have been completed. Suppose that Z in (39) – repeated here for convenience- is a Spellout Domain, and that α is structurally accented. Then β cannot move across α even if β is unaccented and moved to the edge of Z, because the relative linear position of β in Z will be fixed when Z is completed, and because the ordering statement α > Z (hence α > β) will be established when α is merged an accented. (39) [Comp … [W … [α ... [Z … [… β …] … ] …] …] We will take up the proposal of Müller (2007) and assume that phases rather than Spellout Domains are relevant for linearization. According to Müller, the completion of a phase (CP, vP) implies that the complement of the phase’s head must be spelt out.33 The intermediate movement steps of a wh-phrase moving cyclically place the wh-phrase into the edge of a phase, so that it does not have to be linearized when the phase is completed. Not being linearized, unaccented wh-phrases are not subject to intervention effects triggered by ccommanding XPs with a structural accent, as required. Eventually, the unaccented wh-phrase is serialized when it checks the edge feature of Comp. (A-)Scrambling is restricted to given phrases (see Haider & Rosengren 2003 for German and Kučerová 2007 for Czech). The category that moves is unaccented, and has thus not undergone immediate serialization. Therefore, intervention effects triggered by structurally accented XPs are not to be expected. This prediction is borne out, as (46) shows.34

32

See Zubizarreta (1998) for a related proposal. Holmberg’s generalization on Scandinavian Object Shift (the object moves only if the verb moves, too), which figured as one of the major arguments for cyclic linearization in Fox & Pesetsky (2003, 2005) can still be derived if one makes the following two assumptions: (a) object shift moves the object out of vP, without intermediate movement steps, and (b) V does not reach the head position of vP when it does not move out of vP. A detailed discussion is beyond the scope of the present paper. 34 If scrambling is triggered by an edge feature, ultimate serialization can take place when the edge feature is checked. Scrambling may also arise as a side-effect of linearization. When a phase is completed, its complement domain Г must be spelt out, so that the linearization of the material in Г that has not been immediately serialized must be determined. In languages without scrambling, the new ordering statements added for Г must not only be compatible with previous ordering statements, rather they must also be compatible with the hierarchical relations in Г (a > β is impossible if β c-commands α). In scrambling languages, on the other hand, the new statements need only respect previous serialization. A statement a > β entered for a pair of elements in which β c-commands α then yields the scrambling of α to a position c-commanding β. 33

18

(46) Who tore the book up? Ich meine, dass das Buchi der Ingo ti zerrissen I think that the book the.nom Ingo t torne ‘I think Ingo tore up the book.’

hat. has

(Ge)

4.3.2 ‘Formal Fronting’ The examples (8)-(10) have shown that elements located at the left periphery of TP can be placed into Spec,CP in the absence of any particular pragmatic specification (FF-movement). One can find deaccented pronouns in Spec,CP (47). In addition, any other type of subject, temporal adverb, and sentence adverb (48) can appear in Spec,CP in German without any pragmatic force of its own, the same is true for Czech (49b), with the addition of verbal participles (49a). Scrambled objects behave in the same way (see (52) below). Note that FF and SFF overlap when the leftmost item of TP is structurally accented, thus focused. (47) Why does Anna cry? Sie hat viel Geld verloren. she has much money lost ‘She has lost a lot of money.

(Ge)

(48) Why is everyone so excited? Wahrscheinlich ist ein UFO gelandet. Probably is a ufo landed “Probably, an UFO has landed” (49) Why does Anna not want it? a. Přišla by prý o spoustu peněz. lost.Sg.fem aux.cond.3P particle for plenty.acc money.gen ‘She would supposedly lose a lot of money.’ b. Prý by přišla o spoustu peněz.

(Ge)

(Cz)

Travis (1984) proposed to place of the pronoun in (47) into the structural subject position, but not all of the elements taking part in FF-constructions are likely candidates for specifiers of TP (see Fanselow 2002, Müller 2004, Frey 2005). Fanselow (2002) assumes a structural grid such as (50), and invokes the MLC of Chomsky (1995) to guarantee that only the leftmost element of the left periphery is attracted in FF. Müller’s (2004) quite different model has exactly the same consequence. (50) [C [TP (sent. adverb) [(temp. adverb) [(scrambled) [TP subject [vP(scrambled) …]]]]]]] In contrast, the present approach predicts no privileged access of the leftmost element in TP to Spec,CP: if α has not received a structural accent, it is not linearized immediately, so it will neither be subject to intervention effects nor trigger them. Any α not included in the complement of the vP-phase can move to Spec,CP, unless there are intervention effects triggered by structurally accented material. Deciding between left-periphery theories of FF (Fanselow 2002, Müller 2004) and the present approach turns out to be a non-trivial task. Note that the observations brought forward by Fanselow and Müller only show that elements in Spec,CP must also be able to figure as the leftmost element in TP. This does not imply that FF-movement to Spec,CP originates in the left periphery of TP. Even if FF-movement could start only there, the empirical effect of this restriction would be small: unaccented XPs can be scrambled and adjoined to any

19

segment of TP, so they can easily reach the leftmost point of TP in an intermediate scrambling step. However, there is an ordering constraint that may decide the issue. Frey & Pittner (1998). Ernst (2002, 2007), Engels (2004) point out that the subject-oriented interpretation of adverbs such as klugerweise ‘cleverly’ arises only when the adverb is c-commanded by the subject (51a vs. 51b). Engels continues to observe that this restriction need not be respected by movement to Spec,CP (51c-d), and notes that this is incompatible with the MLC-based model of FF, in which (51b) is the only source of (51d). In our present approach, unaccented Anna triggers no intervention effects, so that (51d) can also be derived from (51a). The reading difference between (51b) and (51d) thus supports the present approach. (51) a. dass Anna klugerweise Maria einlud. that Anna cleverly Mary invited ‘that Anna cleverly invited Mary.’ b. dass klugerweise Anna Maria einlud c. Anna lud t klugerweise Maria ein. d. Klugerweise lud Anna t Maria ein.

(subject oriented reading - ok) (subject oriented reading - *) (subject oriented reading - ok) (subject oriented reading - ok)

Consider also the data in (52). Given objects can cross a given subject in adverb focus contexts. Because given objects can undergo scrambling, the derivation of (52b) might involve an intermediate movement step in which the object scrambles to the left of the subject, and is then mapped to Spec,CP by FF-movement respecting the MLC. However, (52a) cannot possibly be derived in this way. Subject pronouns cannot be preceded by scrambled material in TP (52c). Movement of den Fritz to Spec,CP therefore has to cross the subject sie. This is incompatible with MLC-based approaches (in particular, sie might have moved to Spec,CP, too) but it can be easily captured in our proposal: neither sie nor den Fritz are structurally accented, so their movement to Spec,CP is unrestricted. (52) Tell me something about Anna. When did she call Fritz? a. Den Fritz hat sie am FREITAG angerufen. (Ge) the.acc Fritz has she on-the Friday called ‘She called Fritz on Friday.’ b. Den Fritz hat (t) die Anna am Freitag t angerufen. the.acc Fritz has the Anna on-the Friday called ‘Anna called Fritz on Friday.’ c. Wahrscheinlich hat ( den Fritz)* sie ( den Fritz) am FREItag angerufen. probably has the.acc Fritz she the.acc Fritz on-the Friday called ‘Probably, she has called Fritz on Friday.’ FF and SFF constructions thus turn out to be due to the same operation, triggered by an unspecific edge feature of Comp. Wh-movement is also effected by the very same edge feature of C, but the wh-criterion forces that only a wh-phrase can move when C possesses a +wh feature (and excludes the attraction of wh-phrases in case of a [-wh] C, but see Müller & Sternefeld 1996). 4.3.3 Multiple foci/contrastive topics Up to now, we have worked with the assumption that material participating in secondary linearization cannot be accented, and this is correct if one confines one’s attention to structural accents. Structural accentuation is, however, not the only way by which a phrase may receive stress – it can also be stressed as a contrastive or corrective focus, or if it represents a (contrastive) topic. The accents corresponding to these functions do not depend 20

on structural position, rather, they are assigned freely, and map onto specific pragmatic functions. Thus, an XP that has not been serialized immediately may remain unaccented, or receive an accent reflecting contrastive, corrective, topical, etc., interpretation. In case a phrase ends up being unaccented, it is interpreted as given, and if it receives a non-structural accent, that accent corresponds to the appropriate information-structural interpretation.35 We have already seen that given/unaccented material undergoing late linearization can cross structurally accented material, so the extension of our model to non-structural accents leads to the expectation that this crossing option should also be open to contrastive and topical material. This prediction is borne out. For a comparison of structural with non-structural accents, it is helpful to employ a distinction of two types of multiple questions/focus structures discussed in detail by van Hoof (2003:528-533). The first type involves a semantic representation as sketched in (53), in which a single wh-operator unselectively binds two variables (so-called ‘conjoined questions’, Comorovski 1996). It is exemplified in (54). No contrastivity is involved, and both accents of the answer may be assumed to be assigned in immediate linearization. Consequently, no inversion of subject and object is possible. (53) whx,y, x a person, y a dish [x cooked y] (54) So you came in and smelled food. Who cooked what? a. # GUlaschi hat KARL ti gekocht. goulash has Karl cooked ‘Karl cooked goulash.’ b. KARL hat GUlasch gekocht.

(Ge)

In the second type, exemplified by a multiple question such as (55), two distinct operators are present semantically, one taking scope over the other. Importantly, each operates on a different set of alternatives, which leads to contrastive implicature. Thus, in (56), Hubert is contrasted with other members of the set of the party-participants and nobody is contrasted with other quantifiers in the set that fulfil the selection criterion requirement of the discourse. As one would expect for a contrastive sentence, it is pronounced with a so-called bridge (risefall) contour (cf. Büring 1997, among others), and the accented object niemanden is able to cross the accented subject der Hubert in (56a). (55) a. [ Wieviel Leute]i hat wer nicht ti erkannt? how-many people has who not recognized b. whx, x number of people, why, y a person, [y did not recognize x persons] (56) At the class reunion, many did not recognize some of their school friends. Did somebody not recognize anyone? a. NIEmandeni hat nur der HUbert ti wiedererkannt. nobody.acc has only the Hubert.nom recognized ‘Only Hubert did not recognize anyone.’ b. Es hat nur der HUbert niemanden wiedererkant. it has only the Hubert nobody.acc recognized

(Ge)

(Ge)

The crossing constellation in (56a) is possible because the subject der Hubert is the only category that receives a structural accent, i.e., which is linearized immediately. Therefore, there is no further linearization statement that could block the movement of niemanden across 35

Models assuming F-marking additionally postulate T-marking for the interpretation of (contrastive) topics, see. Büring (2007), Reich (2001), a.o.

21

the subject. For the purposes of movement, unaccented (given) categories do not differ from contrastively accented XPs. The examples (57a) from Jacobs (1996) and its Czech equivalent in (57b) show a by now familiar property for the movement of contrastively stressed material: parts of an idiomatic expression can be fronted in the bridge contexts under discussion. What is contrasted in (57) is the complete idiomatic expression, as the possible continuation of the text in the parentheses of the translations in (57) suggests. The fronting in (57) therefore cannot be driven by a pragmatic feature, it is reminiscent of SFF movement. (57) How will Grass react to the bad reviews? a. Nun, [ die HAAre]i wird er sich NICHT gerade ti raufen … (Ge) well the hair will he himself not really pull-out ‘Well, he won’t be completely upset (but he will be a bit angry).’ (Jacobs 1996:8) zrovnaj vznášetk NEbude tj tk ti… (Cz) b. [ V sedmém NEbi]i se asi in seventh heaven refl probably really fly.inf not-will.3Sg ‘Well, he won’t be completely happy (but he will cope with it.)’ The examples in (58) illustrate the same point. What is moved is neither the topic nor a contrastive element nor anything else directly definable in terms of information structure. Rather, it is part of a predicate (reading the bible, etc.); only the latter has the relevant pragmatic function of being a topic or contrasted.36 (58) a. Are they anarchists? HÄUseri hat von denen KEIner ti angezündet. houses has of these nobody set-on-fire ‘No one of them has set houses on fire.’ b. Is he religious? [ Die BIbel]i hat er noch NIE ti gelesen. the bible has he yet never read ‘He has not ever read the bible.’ c. BIblii teda rozhodně NEčte ti. bible particle definitely not-read.3Sg ‘He definitely does not read the bible.’

(Ge)

(Ge) (Cz)

The same is true for further examples discussed in Fanselow (2004), Müller (2003) involving particle verbs. Although the particles make no identifiable semantic contribution, they can be fronted separately in rise-fall constructions, the pied-piping of the verb being optional: (59) a. VORi haben wir es SCHON ti gehabt. before have we it well had ‘We have intended that.’ b. VORgehabti haben wir es SCHON ti.

(Ge)

We conclude that SFF, Formal Fronting and the movement of contrastively stressed material are due to the same syntactic process: attraction by an (unselective) edge feature of Comp. In case Comp happens to attract a category with a structural accent, the intervention effects 36

Examples (58b-c) invite an alternative interpretation in terms of the leftward movement of a topic standing in a POSET-relation (Prince 1999) to elements introduced in the preceding discourse. Talking about religion makes priests, bibles, and the like, prominent. This alternative is less convincing for (58a), and certainly excluded in the case of idiomatic expressions as in (57). Again, the partial fronting of idiom material forces an interpretation as a purely formal movement operation.

22

derived above arise. If Comp attracts a category without a structural accent, no such blocking effects exist. 5

Alternative accounts of SFF based on feature-checking movement

5.1 SFF movement is not due to the attraction of accents Phrases act as interveners in SFF construction depending on their accentuation. One might be tempted to account for this in a more direct way than in the model we are proposing, viz. by assuming that ‘accentuation’ figures as a syntactic feature and licenses movement. Several reasons militate against the idea that SFF can be licensed by agreement for an ‘accentuation’ feature. First, being accented is not a natural lexical feature,37 so the use of accentuation as a trigger of syntactic movement violates the inclusiveness condition.38 A theory of movement that refers directly to phonological properties also has to explain why accentuation is the only aspect of phonology that triggers/licenses movement. The obvious answer is that accentuation itself is determined by syntactic properties. But then, it is a priori more plausible to explain movement in terms of the syntactic factors controlling accentuation rather than in terms of accentuation itself. SFF-like movement can also occur in constellations resembling second occurrence focus (SOF). As pointed out by an anonymous reviewer, in contexts like (60), the direct object einen Bären (part of the idiom einen Bären aufbinden) can be fronted even if it is not accented, cf. (60b).39 (60) a. Peter sagte einen BÄren habe man ihm aufgebunden. Peter said a.acc bear.acc has one him bound-on ‘Peter said they lied to him.’ b. Auch HANS sagte einen Bären habe man ihm aufgebunden. ‘Also Hans said they lied to him.’

(Ge)

The accentuation of elements occurring as second occurrence foci can be fully suppressed in postnuclear positions, and it is reduced in prenuclear positions. However, the elements are prosodically marked by, e.g., duration (see Beaver et al. 2004, Féry & Ishihara 2005). Regardless of the quality of the prosodic marking on SOF, crossing of SOF by first occurrence focus is possible. Consider examples like (61). (61a) belongs to cases discussed in 4.2.3 above: the verb phrase beleidigt is contrastively accented and thus need not be linearized immediately. In (61b), the reappearing bound focus Peter is crossed by another bound focus occurring for the first time, thus receiving the main prominence. As second occurrence foci are always given, along the lines of our linearization-based account we 37

This difficulty could be overcome by assuming that all morphemes are accented in the lexicon, with phonological rules deleting (rather than adding) the accent features relative to syntactic structure and information structure (Gereon Müller, personal communication). 38 Prosodically driven movement as proposed by Zubizarreta (1998) and Szendröi (2001) is different: it does not displace a phrase on the basis of a prosodic property already present in that phrase. Zubizarreta’s account concerns removing elements from the domain of stress assignment to enable a constituent to receive stress in situ, Szendröi considers cases in which, arguably, a constituent moves because it can be stressed in the landing position only. In Czech and German SFF, phrases do not move in order to escape or receive stress. 39 A typical SOF construction is illustrated in (ib): the focus Sid bound by the focus particle only reappears in (ib) as a bind focus in the semantic domain of a higher bound focus (even the state prosecutor). (i) Both Sid and his accomplices should have been named in this morning’s court session. a. But the defendant only named SID in court today. (Beaver et al., 2004) b. Even the state PROsecutor only named Sid in court today. Note that the fronted element in (60) is not bound by a focus particle. See Büring (2006a) for a discussion of unbound secondary foci.

23

conclude that they need not be linearized immediately (and their prosodic marking is not structural). Such distinctions cannot be captured in an account postulating a feature checking attraction of accents. (61) The teacher is angry at Peter’s friends. a. Aber beLEIdigt hat sie nur PEter. but offended has her.acc only Peter ‘But only Peter has offended her.’ b. Sogar beLOgen hat sie nur Peter. even cheated has her.acc only Peter ‘But only Peter has even cheated her.’

(Ge)

A third argument against reference to accentuation in the formulation of SFF lies in the fact that it also occurs in languages that do not have (obligatory) focus marking by stress (for focus in situ). Hausa (cf. Hartmann & Zimmermann (2004)), Trinidadian English, and Haitian Creole (Cozier 2006) allow SFF constructions, as shown by (63) for Haitian in which the fronted material is part of the focus of the construction. The fronted element in Trinidadian English need not be accented (though it can), while Haitian cannot use stress for focusing. Gurune (Andreas Haida, p.c.) and Somali (Svolacchia et al. 1995) behave in the same way.40 (62) What happened? Dabboobi-n jeejìi nee mutàanee su-kà kaamàa. (Hausa) animals-of bush prt men 3pl-rel.perf catch ‘The men caught wild animals.’ (Hartmann & Zimmermann, 2007) (63) What are you doing there? a. Se kuit m ap kuit it is cook I pres.prog cook ‘I am cooking chicken.’ b. Se poul m ap it is chicken I pres.prog ‘I am cooking chicken.’

poul. chicken

(Haitian)

kuit. cook (Cozier, 2006)

That SFF movement also occurs in languages without a prosodic marking of focus would be left unexplained in a theory that directly relates the construction to a prosodic property, even if the latter is understood in a rather abstract way. The German and Czech data discussed in 4.3.2. involving accent crossing constitute a final type of evidence against a prosodic view of SFF movement. For these reasons it is not plausible to account for SFF in term of a syntactic movement triggered by an accent-feature.

40

Apparently, the only important difference with respect to SFF between Czech and German on the one hand, and Trinidadian English and Haitian on the other is that verbs undergo SFF movement in the latter (as evidenced by (35a)), while deaccentuation due to prosodic integration usually blocks SFF movement in the former, see (i). (i) What are you doing? a. N HÜhnchen tue ich mir bruzzeln. ‘I’m roasting a chicken.’ (Ge) a chicken do I me.dat roast b. #Bruzzeln tu ich mir n’HÜhnchen.

24

5.2

Focus-Feature checking approach to SFF

In this section, we argue that SFF (and FF) cannot be captured in terms of feature checking of focus or topic phrases. First, no other type of operator movement allows it that just a subpart of the relevant element checks the operator-related feature. One might see certain parallels between SFF and ‘partial’ wh-movement available in languages like French – (64a) or Czech. However, the wh-operator is the word bearing the wh-feature, and not the DP dominating this word. Therefore, the English equivalent in (64b) involves pied-piping of a DP (for empirical and conceptual reasons against a wh-feature percolation analysis of piedpiping see Heck 2004: 98-113), while (64a) exemplifies wh-movement without pied-piping – and not partial fronting. This is not parallel to SFF phenomenon, where a subpart of a focusmarked phrase undergoes fronting. (64) a. Combien as-tu lu de livres? how-many have-you read books b. How many books have you read?

(Fr)

In order to cope with this problem in a model that analyses SFF constructions as focusfeature checking movement, one would have to assume that it is always the focus XP that is attracted, and that ‘partial’ fronting is the result of constituents of XP being removed from that category before it moves, i.e., that SFF involves remnant movement in the sense of Müller (1998), or some sort of scattered deletion. We will argue that neither of these possibilities can capture SFF appropriately. 5.2.1 Evidence against SFF as remnant movement41 Remnant movement constructions bear a certain resemblance to SFF constructions. In an SFF construction, the overt material showing up in Spec,CP is just part of the focus. Remnant movement theory as discussed in Müller (1998) proposes that the overt fronting of part of an XP must often be analysed as the movement of the complete XP, out of which the parts apparently remaining in situ have been scrambled before XP was moved. In this sense, Spec,CP is filled by VP in (65), but the object was moved (scrambled) out of that VP before the remnant was fronted. (65)

[ti gesehen]VP hat er nur eine EUlei tVP. seen has he only an owl ‘He only saw an owl.’

In an analysis of SFF as remnant movement of focus phrases, some material leaves the attracted focus VP/IP before the focus moves to the left periphery (cf. Cozier’s 2006 account of focus movement in Trinidadian English). The constellation of a remnant movement analysis for SFF is given in (66)-(67) for VP- and IP-focus contexts, respectively. (66)

What did she do? a. [Den GARaus ti tj ]VP/FOC hat sie ihmi gemachtj tVP. the garaus has she him.dat made ‘She killed him.’ b. [Den GARaus ti gemacht]VP/FOC hat sie ihmi tVP.

41

Thanks are due to the anonymous reviewers for prodding us into a more detailed discussion of the relation between SFF and remnant movement or scattered deletion.

25

(67)

What happened? a. [ti den GARaus tj tk ]IP/FOC hat siei ihmj gemachtk tIP. b. [ti den GARaus tj gemacht]IP/FOC hat siei ihmj tIP.

Note, first, that focus preposing indeed often involves remnant movement in German and Czech. A clear and undisputable case of remnant movement involves constructions such as (68a) (taken from Müller 2003) in which, descriptively speaking, more than a single constituent surfaces to the left of the finite verb. However, as shown by Fanselow (1993) and Müller (2003), the material preceding the finite verb IS a constituent, viz. a VP that lost its verbal head, as sketched in (68b-d). (68) a. [Trocken][durch die Stadt] kommt man mit der BVG. dry through the city comes one with the BVG ‘One gets through the city dry with the Berlin City-Transport Company.’ b. [VP [PP mit der BVG] [[trocken][PP durch die Stadt] kommt]]]] Æ PP extraction c. [PP mit der BVG]i ... [VP ti [[trocken][PP durch die Stadt] kommt]]]] ÆV to C movement d. kommtj ... [PP mit der BVG]i ... [VP ti [[trocken][PP durch die Stadt] tj ]]]] Consider now (69). The overt material in Spec,CP consists of two parts of an idiomatic VP. A remnant VP-preposing analysis with the verb gegangen leaving VP before VP moves is inevitable if one wants to maintain that single constituents only can reach Spec,CP. Pragmatically, (69) thus exemplifies the partial fronting of a focus due to remnant movement. (69) What did you do on Helgoland in the evenings? [VP [PP Mit den Hühnern] [PP ins Bett] tV] bin ich gegangenV. with the chicken into bed am I gone ‘I went to bed early.’ However, not all SFF constructions are amenable to such a treatment. For remnant movement of XP to Spec,CP, the operation evacuating XP before movement must be scrambling.42 German has no long-distance scrambling (cf. Müller & Sternefeld 1993). This restriction explains the ungrammaticality of remnant VP-fronting examples such as (70), because the category left behind by remnant movement would have to be extracted by scrambling out of an embedded finite clause (70a) or a CP-infinitive (70b,c), as sketched in (70d). (70) a. b. c. d.

*[ Gesagt, dass sie sich ti nehmen wird] hat sie ein TAxii. said that she refl. take will has she a taxi ‘She said she would take a taxi.’ *[ Gedrängt t aufzugeben] hat sie ihn seinen JOBi. urged give-up.inf has she him.acc his job *[ti Aufzugeben gedrängt] hat sie ihn seinen JOBi. ‘She urged him to give up his job.’ hat sie [VP gesagt dass sie sich ein Taxi nehmen wird] hat sie ein Taxi [VP gesagt dass sie sich t nehmen wird] [VP gesagt dass sie sich t nehmen wird] hat sie ein Taxi

Corresponding SFF examples are grammatical, however, as (71) shows. A remnant movement analysis for (71) would have to involve illicit long scrambling, cf. the derivation in (72). (71) thus cannot be interpreted as remnant VP movement. 42

Further options are head movement and extraposition (Müller 1998), but these are not relevant for (70).

26

(71) a. And what happened then? [Ein TAxi] hat sie gesagt dass sie sich nehmen wird. a taxi has she said that she refl. take will ‘She said she wants to take a taxi.’ b. How did she apologize? [Ein UNfall] sagte sie dass es gewesen war. an accident said she that it been was ‘She said it had been an accident.’ c. How did she upset him? [Seinen JOB] hat sie ihn gedrängt aufzugeben. his.acc job has she him.acc urged give-up.inf ‘She urged him to give up his job.’ (72) hat sie [VP gesagt dass sie sich ein Taxi nehmen wird] hat sie [VP ein Taxi [VP gesagt dass sie sich t nehmen wird]] hat sie [VP gesagt dass sie sich t nehmen wird] [VP ein Taxi tVP] [VP ein Taxi tVP] hat sie [VP gesagt dass sie sich t nehmen wird] The parasitic gap data introduced above also militate against a remnant VP-movement analysis of SFF. The parasitic gap bound by Akten in the SFF example (37) repeated here has the category DP, it is not a VP. (37) What did he do? AKteni hat documents has

er [anstatt ei zu bearbeiten] ti in den Papierkorb geworfen. (Ge) he instead-of to process in the basket tossed

Another problem for the remnant movement idea arises from restrictions on VPtopicalization. As observed by Haider (1990) and Diesing (1990), nominative subjects can be part of a fronted constituent if they are not definite, cf. the difference between (73a) and (73b). No such restriction applies to fronted objects, cf. (73c). The question is whether the definiteness restriction on subjects is constant with all predicates, as shown by the contrast in (74). Importantly, SFF structures show no such definiteness effects with subjects, regardless of the predicate, cf. (75).43 (73) a.

[ Ein Orden verliehen] wurde a.nom medal awarded was b. *[ Der Orden verliehen] wurde the.nom medal awarded] was c. [ Den Orden verliehen] hat man the.acc medal awarded has one

(74) a. *[Der Papst angekommen] ist hier the pope arrived is here ‘The pope has never arrived here.’ 43

ihr erst gestern. (Ge) her.dat only yesterday ihr erst gestern. her.dat only yesterday ihr erst gestern. her.dat only yesterday noch NIE. (Ge) never-before

Wurmbrand (2001) attributes the restriction on topicalization in (73)-(74) to the ungrammaticality of TPfronting in German. If definite subjects have to overtly move to TP, cf. Diesing (1990), (73b) would have to involve TP-fronting. In contrast, indefinite subjects and definite objects stay in the vP, which can be fronted. Thus (73a) and (73c) are expected to be grammatical. Given such a ban on TP-fronting in German, SFF cases like (75) should be excluded, as they involve (remnant) TP-fronting, but such examples are fully grammatical.

27

b. [Der Papst begegnet] ist mir noch NIE. the pope encounter is me.dat never-before ‘I have never met the pope.’ (75) What happened? a. Der PAPST ist mir begegnet! (Ge) the pope is me.dat encountered b. Der PAPST ist angekommen! the pope is arrived Further differences between SFF movement and VP-fronting can be observed in Czech but not in German. A question inducing VP-focus as in (76) can be answered by SFF structures, cf. (76a-b). However, Czech allows quantifier stranding only if its genitive complement is fronted alone (76a), not if the (infinitival) VP is fronted, cf. (76c) vs. the German (77c). (76a) thus has no obvious remnant movement source. (76)

What else could I do? a. [RŮží ] bys mohl pár koupit. (Cz) roses.gen aux.2sg could a-couple buy.inf ‘You could buy a few roses.’ b. [pár RŮží] bys mohl koupit. a couple roses.gen aux.2sg coud buy.inf c. *[Koupit RŮží ] bys pár mohl. buy roses.gen aux a-couple could.2sg d. [Koupit pár RŮží] bys mohl. buy.inf a-couple roses.gen aux.2sg could

(77) a. [ROsen] könntest du ein paar kaufen. (Ge) roses could.2Sg you a-couple buy.inf b. [Ein paar ROsen] könntest du kaufen. c. [ROsen kaufen] könntest du ein paar. d. [Ein paar ROsen kaufen] könntest du. 5.2.2 SFF is not scattered deletion As pointed out by a reviewer, a related possibility of explaining SFF in terms of focus-feature checking consists of the creation of discontinuity by scattered deletion (Bošković 2002, Fanselow & Ćavar 2001, 2002). If the deletion following copying can affect both copies, SFF constructions could involve focus phrases in which all material but the leftmost accented XP has been deleted in the left copy, with the remainder being realized in the right copy, as in (78). (78) A: What did he do? B: [Ein BUCH zerrissen] hat er [ein BUCH zerrissen]. a book torn up has he a book torn up “he tore up a book” Scattered deletion is, however, no improvement over remnant movement with respect to the parasitic gap argument and the subject-definiteness problem. The locality differences between partial VP-fronting and SFF movement also find no satisfactory explanation. 28

Furthermore, scattered deletion has been introduced for discontinuous DPs and PPs, and does not always yield SFF-like results in this context. Czech allows discontinuous DPs in answers to discontinuous wh-phrases like ‘whose-X’ and ‘what kind of-X’ – (79a), although piedpiping of the whole DP is a valid option – (79b) (as well as leaving the DP in situ – (79c)): (79) Q: Jakou jsi koupil růži? (Cz) what.acc aux.2sg bought.sg.ms rose.acc ‘What kind of a rose did you buy?’ a. ČERvenou jsem koupil růži. red.acc aux.1sg bought.sg.ms rose.acc b. ČERvenou růži jsem koupil. red.acc rose.acc aux.1sg bought.sg.ms c. Koupil jsem ČERvenou růži. bought.sg.ms aux.1sg red.acc rose.acc In (79), the wh-word imposes a narrow focus reading on the fronted modifier in the answer, whereas the part of the noun phrase stranded in (79a) belongs to the background. Note, however, that structures like (79a) are infelicitous in wide focus contexts, cf. (80), with Q1 imposing DP-focus and Q2 VP/IP-focus (although the prosodic make up of the answer is parallel, compare (79c) and (80a)). (80) Q1: Karel invited Petra to the cinema and Jan brought her a red rose. What did Emil buy? Q2: Peter invited Hanna to the cinema and John brought her a red rose. What did Emil do/ What did Emil buy? a. Koupil jí BÍlou růži. (Cz) bought.sg.ms her.dat white.acc rose.acc ‘He bought her a white rose.’ b. #BÍlou jí koupil růži. white.acc her.dat bought.sg.ms rose.acc c. BÍlou růži jí koupil. The contrast between (79a) and (80b) could be explained if the upstairs deletion may apply only to non-focused parts. Then even the deaccented noun rose in the answers of (80) would resist deletion, being part of the focus phrase. But note that such a rule would exclude other cases of SFF that are possible. For example, the above discussed example (16c) repeated here, should be impossible, due to the deletion of a part of the focus upstairs.44 (16)

6.

What have you bought? KNÍžek jsem si [pár e] koupila. (Cz) book.gen.plaux.1sg refl. a few bought.sg.fm Eliminating Focus And Topic Movement?

In the preceding section, we have shown that SFF, FF and wh-movement find a satisfactory unified analysis in terms of the following model: movement is triggered by an unspecific edge feature of Comp, and is not confined by the MLC (Chomsky 2005). The locality effects of FF and SFF movement are a consequence of the idea that the complement domain of a 44

(80b) is also relevant for a potential remnant movement analysis of SFF. Left-branch extraction like (79a) has been analyzed as movement of a DP-remnant, cf. Corver (1990). SFF is obviously not compatible with this operation. Thus we would have to postulate different kinds of remnant movement for the two phenomena.

29

phase must be spelt out when the phase is completed, combined with the idea that linearization statements must not be changed in the course of a derivation (Fox & Pesetsky 2003, 2005). The main ingredient added by the present paper lies in the assumption that the linearization operation that is required for structural accentuation must happen immediately, i.e., as soon as a syntactic object is merged. The fronting of an XP is triggered by an edge feature of C, so information structure has no influence on the choice of the displaced category. However, movement is constrained by linearization theory, with a direct link to accentuation, and, consequently, an indirect link to pragmatics. An XP with a structural accent moved to Spec,CP may be part of the focus of an utterance (=SFF construction), but it may also happen to be the narrow focus of the utterance. In the latter case, the surface effect of movement is identical with the one a special focus movement operation would have. Likewise, the edge feature of C can map elements with a non-structural accent to Spec,CP. The phrase that moves in this context may be part of a contrastive focus, or part of a topic, but it can also happen to be the contrastive focus or topic itself. In other words, the set of movement options licensed by the edge feature of C cannot be prevented from having the overt effect that a topic/focus fronting process would have, so that the latter operations should be abandoned on grounds of parsimony. The question of whether this model for left-periphery-movement is correct has two different aspects. First, we need to establish that the choice of the category fronted to Spec,CP is pragmatically adequate, and we need to explain the differences between the various movement types in binding or locality. 6.1

Topic vs. Focus: Size of the fronted category

A focus or a topic movement rule displaces exactly the focus or the topic of an utterance. The movement triggered by an edge feature of Comp is completely unselective in this respect. The fronted XP may happen to be the focus or the topic, but it can also be a category smaller or larger than the actual focus/topic. The data in (81)-(82) illustrate once more a point made repeatedly in the present paper: the focus XP (=VP, in our case), can remain in situ (81d)(82c), or it can be fronted partially (81/82a,b), or completely (81c). Any of the options – focus F remains in situ, or F is fronted partially or completely – is pragmatically acceptable in the context of F-focus. (81) What have you done in the city? a. Bücher hab ich mir ein paar gekauft. (Ge) books have I me a couple bought ‘I bought a few books.’ b. Ein paar Bücher hab ich mir gekauft. c. Ein paar Bücher gekauft hab ich mir. d. Ich hab mir ein paar Bücher gekauft. (82) What have you done in the city? a. KNÍžek jsem si [ pár t] koupila. book.gen.pl aux.1sg you.dat a few bought.sg.fm ‘I have bought a couple of books.’ b. [Pár KNÍžek] jsem si koupila. c. Koupila jsem si pár KNÍžek.

(Cz)

Given the unspecific nature of the edge feature of C, a category larger than the focus can be fronted, too. As expected, a question imposing a narrow focus on the direct object is also

30

congruent with (83b), a sentence in which the VP dominating the focused direct object has been moved.45 (83) What have you bought? a. Ein BUCH hab’ ich mir gekauft. (Ge) a book have I myself bought b. [VP Ein BUCH gekauft] hab ich mir.46 ‘I bought a book.’ Examples such as (83b) lend further support to our proposal. In theories in which the focus feature is decisive for movement to Spec,CP, (83b) would have to be analysed as involving the pied-piping of a VP, but such an analysis is not in line with a general restriction on piedpiping (see Heck 2004): if α is attracted to β, a Γ c-commanded by β and dominating α can be pied piped only if α (or a category D dominating α and dominated by Γ) cannot be extracted from Γ. In other words, pied piping is restricted to categories that are syntactic islands for the attracted category. Consequently, DP can be pied piped when its wh-specifier is attracted, because German does not allow violations of the left branch condition (84a). VP, however, is not a barrier for movement, and cannot be pied-piped. (84) a. *Welches hast Du Dir Buch gekauft? which have you yourself book bought b. Welches Buch hast Du Dir gekauft? c. *Welches Buch gekauft hast Du Dir? ‘Which book have you bought?’ The contrast between (83b) and (84c) is difficult to explain if the two derivations differ only in that +wh-agreement triggering movement is replaced by +foc-agreement: why should the movement of a [+foc] category not be subject to the standard conditions on pied piping? In contrast, the difficuly with (83b) does not arise in our model, because the unselective edge feature of C can attract any accessible phrase. Thus, it can also attract the VP directly: (83b) is not an instance of pied-piping (cf. Horvath (2007) for a related point concerning Hungarian).47 Let us now consider topic movement. The topic can be moved to Spec,CP, and we also saw above that topic fronting can be partial, and that the topic does not have to move to Spec,CP. In these respects, topic and focus behave alike. However, Frey (2004) argues for German that a VP dominating a topic object cannot be moved to Spec,CP. This suggests that 45

As discussed above, such examples are not possible to construct in Czech. The fronted constituent in (83b) cannot serve as term answer in the same dialogue, cf. (i). Term answers may be analysed as sentential answers, in which the string identical with the congruent wh-question has been deleted (see Reich 2003, among others). The string is identical only in (ib). (i) Was hast du gekauft? What have you bought? a. *Ein BUCH gekauft. b. Ein Buch habe ich gekauft. 47 The degraded status of (ia) shows that the prosodic integration of the verb plays a crucial role in determining the status of (83b) and (ib). Syntactically, there is no major difference between these sentences, but prosodically, there is: the verb integrates into the phonological phrase of the direct object, but the indirect object does not, and if it would, the resulting phonological object would not have the strongest accent at its left edge. (i) What did you give to Mary? a. ?Der Maria ein Buch geschenkt habe ich. the.dat Mary a book presented have I ‘I presented Mary with a book.’ b. Ein Buch geschenkt hab ich der Maria. a book presented have I the Mary 46

31

the moved category must not be larger than a topic (except in cases of obligatory piedpiping), in contrast to what holds for focus movement. (85) What about Tom? a. Den hat niemand eingeladen. him has nobody invited b.#[VP den eingeladen] hat niemand. ‘Nobody invited him.’ c.#[VP den Idioten eingeladen] hat the-acc idiot invited has ‘Nobody invited that idiot.’

(Ge)

niemand. nobody

However, German topics seem to always leave VP (Wurmbrand 2001, Frey 2004), perhaps because topics must be prosodically set apart from the rest of the clause. Consequently, when a VP is fronted, it cannot carry along the topic, because the topic has moved out of VP. TPs, on the other hand, arguably cannot be moved to Spec,CP. The data in (85) thus find an independent explanation. If this is correct, one predicts that a fronted subordinate CP can contain a topic, and this prediction is borne out. The co-occurrence of (86b’) and (86b’’) shows that the topic can go along with the rest of CP in contexts where it could itself be extracted, i.e., (86b’) is not amenable to a pied piping analysis. (86) a. As for your book [CP das lesen zu müssen] war eine Zumutung. that read to must was a cheek ‘It was a bit too much to have to read it.’ b. As for generative syntax b.’ Die zu unterrichten ist mir ein Gräuel. it to teach is me.dat a horror b.’’Die ist mir zu unterrichten ein Gräuel. ‘I hate teaching generative syntax.’

(Ge)

LP-movement thus turns out to not be adequately describable in terms of pragmatic features: what goes to Spec,CP may be smaller, larger, or identical with the focus and the topic, and neither the topic nor the focus have to undergo LP-movement at all. 6.2

Topic vs. Focus: Binding and Reconstruction

A further potential difference between the descriptive types of LP-movement lies in an observation we owe to one of the anonymous reviewers: parts of a focus moved to Spec,CP undergo obligatory reconstruction, while (partial) topic movement can, but need not be reconstructed (Czech behaves the same with respect to this point, we will only discuss the German example provided to us by the anonymous reviewer). (87) a. What did the trainer say after they got kicked out of the tournament? Seine sieben SAcheni sollte jeder ti packen. his seven things should everyone pack ‘Everybody should leave.’ b. He said that every inhabitant has his duties. What does he mean? *Jedes ZImmeri sollte seini Bewohner ti aufräumen. every room should its inhabitant clean up. ‘Every room should be cleaned up by its inhabitant.’

32

c.

What do the regulations say about the cleaning of each room? Jedes Zimmeri sollte seini BEWOHNER aufräumen.

The contexts in (87a-b) yield wide focus interpretations of the answers with a parallel prosodic shape: the fronted object DPs carry the main prosodic prominence, while the remaining arguments are deaccented or not prominent. In (87a), with partial fronting affecting an idiomatic expression, the lower quantifier jeder can bind the fronted variable seine. In contrast, binding is impossible in (87b), where the quantifier is contained in the fronted phrase and its base position is lower than the subject. This means that SFF movement undergoes obligatory reconstruction at LF. With the bridge contour linked to a topicality, (87c) seems fine, however. Does that constitute a good reason for distinguishing different types of LP-movement syntactically? Note that the possibility of creating new binding options in contexts such as (87c) rather than the need for reconstruction in (87b) (which one expects for A-bar-movement in any event) is the remarkable property. After all, Spec,CP is an A-bar position and therefore irrelevant for the A-binding of pronouns. Haider (1981) proposed that the ‘new’ binding option is established before the phrase moves to Spec,CP. In current terms, this means that new binding options are due to an intermediate derivational step (e.g., scrambling) that has properties of A-movement (Fanselow 2001, Haider & Rosengren 2003, Frey 2004), as sketched in (88). (88) Merger of subject [[pronoun N]nom .. [DPacc … ]] Æ Object scrambling, New binding DPacc-i [[pronouni N]nom … [DPacc … ]] Æ Object moves to Spec,CP [CP DPacc-i … [DPacc-i [[pronouni N]nom … [DPacc … ]]] Æ ‘LF’- Reconstruction to scrambled position retains new binding. This explains the differences in (87). Phrases cannot scramble when they have a structural accent, because their linearization is then fixed when they are merged. If a phrase with a structural accent moves to Spec,CP, it therefore cannot have undertaken an intermediate scrambling step. The movement step necessary for yielding a new binding option cannot been carried out. LP-movement displacing part of a focus thus always reconstructs semantically, which explains (87a-b). If a phrase has no structural accent it can be scrambled and mapped to Spec,CP subsequently. From this scrambled position, new binding options can be created (e.g. an object binding a pronoun in a subject), which further movement to Spec,CP does not destroy (87c). The difference between the various instances of movement to Spec,CP with respect to binding are thus predicted in our model. 6.3

Focus vs. Topic fronting: locality effects

Rizzi (1997) argues on the basis of Italian data that focus and topic movement differ syntactically in general. Topic but not focus fronting involves clitic doubling (89), Focus movement (capitalized) shows weak crossover effects (90b), while topic movement does not (90a). (89) a. Il tuo libro, l’ ho comprato. the your book it have.I bought ‘(As for) your book, I’ve bought it.’ b. *Il tuo libro, ho comprato. c. *IL TUO LIBRO, l’ ho comprato. d. IL TUO LIBRO ho comprato. 33

(It)

‘It is your book that I have bought.’ (90) a. Giannii, suai madre loi ha sempre apprezzato. Gianni, his mother him has always appreciated ‘As for Gianni, his mother has always appreciated him.’ b. ??GIANNIi, suai madre ha sempre apprezzato. ‘It is Gianni that his mother has always appreciated.’

(It)

There is nothing unexpected in the focus data: weak crossover is typical for A-bar-movement in languages without scrambling, and foci cannot be linked to clitics in situ either. Furthermore, (91) illustrates that Italian allows SFF, too. In (91b) (which we owe to Vieri Samek-Lodovici) the accented part of an idiom has moved into the ‘focus’ position. This parallel between German/Czech and Italian suggests that ‘focus’ fronting is effected by a simple edge feature of Comp in Italian, too. (91) a. Why are you so late? Un blocco stradale, ha a block road, has ‘The taxi ’ b. What have you done to the pig? La festa, gli abbiamo the feast, to-him have.we ‘We killed him.’

trovato il tassí. found the taxi

(It)

fatto. done

Further evidence for this claim comes from the ban against overt non-pronominal subjects in Spec,TP that characterizes various constructions in Italian, viz. resumptive preposing of aboutness topics (92) (= 40(a-c),(54a) from Cardinaletti 2007, cf. also Cinque 1990), whmovement, focus movement (93) (Cardinaletti 2002, Belletti 2004) or left peripheral focus in Sicilian (Cruschina 2006). (92) a.*La stessa proposta il partito di maggioranza fece poi. the same proposal the party of majority made then ‘The majority party then made the same proposal.’ b. La stessa proposta pro fece senza consultarsi con nessuno. the same proposal made without consulting with anybody ‘(S)he made the same proposal without consulting anybody.’ c. che la stessa proposta tu possa fare in tutta autonomia. make in full autonomy that the same proposal youweak can ‘that you can make the same proposal in full autonomy.’ d. La stessa proposta fece poi il partito di maggioranza.

(It)

(93) a. *Chi who b. *A to

(It)

Gianni ha invitato? Gianna has invited MARIA Gianni parlò Mary Gianni talked

no not

na to

Paolo. Paul

These operations can only apply across empty or weak pronominal subjects, or across traces of heavy NP-shift. In situ deaccentuation of lexical NPs as we find it in Germanic or Slavic is absent in Romance.48 This means, in particular, that non-pronominal subjects are immediately linearized when they are placed into Spec,TP. The blocking effect of intervening lexical 48

There is, however, a postfocal operation of ‘marginalization’, cf Antinucci & Cinque (1977).

34

subjects for the constructions in (92) and (93) is therefore, in principle, amenable to a treatment in terms of linearization theory that excludes the movement of structurally accented categories past an XP that has been serialized. One way of understanding the Italian topic data in (89)-(90) could thus be related to the idea that XPs are immediately serialized and equipped with structural accents in TP. Topics bearing a non-structural accent thus need to be merged outside TP, i.e. a construction such as CLLD has to be chosen. By and large, Italian fits well into the model proposed here, but the absence of in situ deaccentuation forces a non-movement origin for topics in most contexts. 6.4

In situ vs. ex-situ focus

For languages that have a choice between an in situ and an ex situ focus strategy, it has been claimed that ex situ focus comes with additional pragmatic properties (see Drubig 2003, among others): it must be ‘exhaustive’ or ‘contrastive’. This is still subject to discussion (see, e.g., Wedgwood 2003 and Brunetti 2004 for arguments against a strict grammatical encoding of such additional interpretive properties in Hungarian and Italian), but for the purposes of the present paper, it suffices to observe that partial and ‘complete’ focus movement do not differ in this dimension. Hungarian is among the languages tolerating partial focus movement, as is clear from examples such as (94b-c) that we borrow from Kenesei (1998).49 The fronting of the object triggers an exhaustive interpretation, irrespective of whether the sentence comes with a wide or a narrow focus. (94) What did Peter do yesterday afternoon? a. Péter [fel-olvasta a Hamletet a kertben] (nem pedig úszott). Peter PV-read the Hamlet-ACC the garden-in not rather swam 'What Peter did was read out Hamlet in the garden (rather than swim). b. Péter [a Hamletet olvasta fel a kertben] (nem pedig úszott). c. Péter [a kertben olvasta fel a Hamletet] (nem pedig úszott). Hungarian supports our model in a further respect. According to Szendrői (2001) and others, an XP moves to the preverbal ‘focus’ position in order to pick up stress there. The movement to the focus position thus does not start in a slot in which a structural accent is assigned, rather, it terminates there. Consequently, there can be no intervention effects. Indeed, in the structure (94b-c), any of the constituents can be moved into the preverbal slot while the wide focus interpretation is maintained. That each constituent is able to move to the preverbal slot in a wide focus context is best illustrated with idiomatic expressions, as (95) shows: According to I. Kenesei (p.c.), both linearizations are possible with a wide focus reading as long as both NPs are stressed. (95) a. Anna a “gombhoz varrta a “kabátot. Anna the button.all sewed the coat.acc b. Anna a “kabátot varrta a “gombhoz. ‘Anna sewed the coat to (fit) the button.’ = ‘She was attentive to the details rather than the whole picture.’

(Hu)

The situation in Hungarian is more complex than just described. In particular, the subject must be moved to preverbal position in wide focus contexts when the subject is neither pronominal nor a topic (Kenesei, p.c.) and there are further facts to be accounted for (see, e.g., Horvath 2007). A detailed treatment of Hungarian is beyond the scope of the present 49

The existence of such sentences was first noticed by Szabolcsi (1981), as was pointed out to us by K. Kiss.

35

paper. For our purposes, the parallels of fronting in wide and narrow focus contexts are the crucial observations. 7.

Concluding Remarks

In the preceding sections, we analysed the movement to the left periphery of clauses in Czech, German, Hungarian, and Italian, with different degrees of detail. We found that what matters in Czech and German is not information structure but accentuation, and this relevance of accentuation was a side effect of immediate serialization. Certain data in Hungarian and Italian suggest that these languages allow essentially the same analysis. 7.1

SFF in other languages

In fact, many languages pattern with German and Czech in this respect. The existence of partial fronting of a focus, in particular partial idiom fronting, is a good indication that movement to the left is not driven by information structure. Above, we have already presented or cited pertinent data for Hausa, Gurune, Somali, Trinidadian Dialectal English, and Haitian. One can easily add more languages. Russian dialogues such as (96) are a case in point.50 In contrast to Czech, elements bearing the nuclear accent in Russian (Ru) are frequently moved to a preverbal, sentence-medial rather than initial position (cf. Mehlig 1993, Janko 2001), so in wide-focus sentences like (96b), partial focus movement (capitalized) need not clash with accented subjects: (96) a. What have the children done? Cvety oni sobrali. flowers.acc they have-plucked ‘They have plucked flowers.’ b. What’s the noise outside? Deti v MJAČ igrajut. children in ball play ‘Children are playing ball.’

(Ru)

(Janko 2001:195)

Likewise, Polish and Slovenian possess partial focus movement. Our three Serbo-Croatian informants had some additional conditions on SFF in their language. While the fronting of a direct object in case of a VP- or IP-focus is grammatically possible, it is linked to the expression of annoyance. Such an additional effect is absent with the narrow focus interpretation of fronted direct objects. It is difficult if not impossible to pack such usage differences in the syntactic derivation itself: we do not want an annoyance-feature in a syntactic derivations. Rather, it suffices to observe that partial focus movement constructions can be identified by their surface representations, so that it is possible –but not necessary– that additional conditions of use are linked to them. Ideally, the restricted nature of SFF movement in languages such as Dutch or Swedish in which SFF constructions are rather marked (as compared to narrow focus object fronting) can be linked to conditions of use as well.

50

Thanks to Konstantin Kazenin for having this tested in a linguistics class at Moscow University. Thanks also go to Tijana Ašić, Damir Ćavar, Luka Szucsich (Serbocroatian), Joanna Błaszczak, Pawel Rutkowski (Polish), Andrej Stopar (Slovenian), Hans Broekhuis, Norbert Corver, Ruben van de Vijver (Dutch), Elisabeth Engdahl, Christer Platzack (Swedish), Stavros Skopeteas (Greek), Kaja Kohler (Estonian) and Henna Kairanneva (Finnish).

36

(97a) is also a perfect dialogue for speakers of Greek, though not for all informants. The accented part of an idiom can also be preposed without a loss of the idiomatic interpretation, cf. (97b). (97) a. What did you do on the weekend? Tin efimeriDa Djavasa. the.acc newspaper I-have-read ‘The newspaper I have read.’ b. What about the neighbours? Mijes varane I jítones. fly.acc.pl.fm hit.3pl def.nom.pl.ms neighbour.nom.pl.ms ‘They are bored.’

(Gr)

According to Williams (2003:34) instances of English Heavy NP-Shift can figure as a reply in (98a). We concur with Williams’ claim that the moved bracketed constituent is part of the focus, not the focus, and in that respect, the construction resembles SFF. (98b) taken from Birner & Ward (1998:36) may involve an instance of partial leftward focus movement. It exemplifies the preposing of the accented phrase six dollars in a situation in which ‘costs six dollars’ is the focus.51 (98) a. What did John do? John gave to Mary [all of the money in the satchel]. b. Where can I get the reading packet? In Steinberg [Gives directions]. Six dollars it costs. While Estonian appears to allow partial fronting easily, Finnish speakers tend to reject simple examples of SFF constructions. However, dialogues like (99) are acceptable with the contrastive particle hän. Perhaps, this is due to the strong contrastive nature of the left peripheral position in Finnish (see Kaiser 2000) that triggers the need for the additional presence of a contrastive particle. If correct, this implies that Finnish belongs to the set of languages allowing partial focus movement, too. (99) A: Have you heard that Peter left his wife yesterday? B: Ei, talonsahan hän my-i! no house.3s.poss-part 3s sell.past.3s ‘No, he sold his house!’ 7.2

Outlook

In descriptive terms, the choice of a marked word order goes hand in hand with the expression of a marked information structure in many languages. That such correlations of form and function are reflected in the mechanics of syntax may seem very plausible, but we have shown that theories of this sort do not capture the wealth of empirical facts in languages with free constituent order. For the languages we have considered, it seems safe to conclude that notions of information structure play no role in determining what is fronted to the left periphery of a clause, and they neither figure in determining the locality of fronting nor in any other syntactically relevant aspect of the construction. Movement to the left periphery is triggered by an unspecific edge feature of Comp, and the choice of the category to be fronted is arbitrary from a syntactic point of view. This finding is in line with the observation that constructions do not have a uniform function in natural language. Baker (2002) points out 51

Under our analysis, but not the interpretation offered by Birner and Ward.

37

that Mohawk sentences have the syntactic structure of Italian clitic left dislocation constructions without sharing their pragmatic force. The conclusion we draw is that notions of information structure play no role in the functioning of syntax (cf. also Chomsky 2005). Focus features/focus marking was added to syntactic representation in the past when the Government & Binding architecture of syntax had no direct means of transporting information from the PF-representation of a clause to its LF, and vice versa (see Horvath 2007 for a lucid reconstruction of the history of focus marking in the syntax). More recent syntactic models allow for a parallel computation of prosodic, semantic and syntactic properties of a sentence, so they need no ‘mediating’ features. There is no reason left for coding information structure in the syntactic representation, and our results show that there is also no empirical motivation for it. Prosody is the primary means of expressing information structure. There can be no doubt that prosody has an influence on the well-formedness of syntactic representations. We have proposed that this influence is effected in an indirect way, too: prosodic structure is computed on the basis of linearization information, and linearization is a crucial factor in determining syntactic locality (Fox & Pesetsky 2003,2005; Müller 2007).

References Adger, David 2006. Stress and Phasal Syntax. Manuscript. Queen Mary College. University of London. Alboiu, Gabriella. 2004. Optionality at the Interface: Triggering Focus in Romanian. In: Henk van Riemsdijk and Anne Breitbath (eds.), Triggers, Mouton de Gruyter, Berlin. 4975, Antinucci, F., and G. Cinque, 1977. Sull’ordine delle parole in Italiano: l’emarginazione. Studi di Grammatica Italiana VI, 121–146. Baker, Mark. 2001. The Natures of Non-Configurationality. In: Mark Baltin and Chris Collins (eds.), Handbook of Syntax. Blackwell, Oxford, 407-438. Bayer, Josef & Jaklin Kornfilt. 1994. Against scrambling as an instance of move alpha. In: Henk van Riemsdijk and Norbert Corver (eds.), Studies on Scrambling. De Gruyter, Berlin, 17–60. Beaver, David, et al. 2004. When semantics meets phonetics: Acoustical studies of second occurrence focus. Ms., Stanford University. Belletti, Adriana. 2004. Aspects of the low IP area. In: Luigi Rizzi (ed.), The Structure of CP and IP. OUP, Oxford, 16-51. Birner, Betty and Ward, Gregory. 1998. Information Status and Noncanonical Word Order in English, Benjamins, Amsterdam. Bošković, Željko, 2001. On the Nature of the Syntax-Phonology Interface. Elsevier, Amsterdam. Bošković, Željko. 2002. Clitics as non-branching elements and the Linear Correspondence Axion. Linguistic Inquiry 33:329-340. Brunetti, Lisa. 2004. A unification of Focus. Unipress, Padova. Büring, Daniel. 1997. The Meaning of Topic and Focus. The 59th Street Bridge Accent, Routlege, London. Büring, Daniel. 2006a. Been There, Marked That – A Tentative Theory of Second Occurrence Focus. Ms., UCLA. Büring, Daniel. 2006b. Focus Projection and Default Prominence. In: V. Molnár and S. Winkler (eds.), The Architecture of Focus, Mouton de Gruyter, Berlin, 321-346. Büring, Daniel. 2007. Semantics, Intonation and Information Structure. In: G. Ramchand and C. Reiss (eds.), The Oxford Handbook of Linguistic Interfaces. OUP, Oxford

38

Cardinaletti, Anna. 1990. Subject/object asymmetries in German null-topic constructions and the status of specCP. In: Juan Mascaró and Marina Nespor (eds.), Grammar in progress: GLOW essays for Henk van Riemsdijk. Foris, Dordrecht, 75-84. Cardinaletti, Anna. 2002. Against optional and zero clitics. Right dislocation vs. marginalization. Studia Linguistica 56: 29-57. Cardinaletti, Anna. 2007. Topic or Focus? On some preposing phenomana in Italian, compared to Germanic. Paper presented at the Comparative Germanic Syntax Workshop 2007, Stuttgart. Chomsky, Noam. 1981. Lectures on Government and Binding. Foris, Dordrecht. Chomsky, Noam. 1995. Categories and Transformations. In: The Minimalist Program, MIT Press, Cambridge, Mass., 219-394. Chomsky, Noam. 2005. On Phases. Ms., MIT. Cinque, Guglielmo. 1990, Types of A’-Dependencies. MIT-Press, Cambridge, Mass. Comorovski, Ileana. 1996. Interrogative Phrases and the Syntax-Semantics Interface, Kluwer, Dordrecht. Corver, Norbert. 1990. The Syntax of Left Branch Extractions, Katholieke Univ. Brabant. Cozier, Franz. 2006. The Relationship between Focusing and Clefting in Trinidadian Dialectal English. Ms., MIT. Craenenbroek, Jeroen van, and Liliane Haegeman. 2007. The Derivation of Subject Initial V2. Linguistic Inquiry 38, 167-178. Cruschina, S. 2006. Informational Focus in Sicilian and the Left Periphery. In: Mara Frascarelli (ed.). Phases of Interpretation. Mouton de Gruyter, Berlin, 363-385. Diesing, Molly. 1990. The syntactic roots of semantic partition, PhD thesis, University of Massachusetts, Amherst. Drubig, Hans Bernhard. 2003. Toward a Typology of Focus and Focus Constructions. Linguistics 41, 1-50. Engels, Eva. 2004. Adverb Placement. An Optimality Theoretic Approach. Doctoral Dissertation, Universität Potsdam. Epstein, Samuel D. & T. Daniel Seely. 2002, Rule Applications as Cycles in a Level-free Syntax. In: Samuel D. Epstein, & T. Daniel Seely (eds.). Derivation and Explanation in the Minimalist Program. Blackwell, Oxford, 65-89. Ernst, Thomas. 2002. The Syntax of Adjuncts. Cambridge University Press, Cambridge. Ernst, Thomas. 2007. On the role of semantics in a theory of adverb syntax. Lingua 117: 1008–1033. Fanselow, Gisbert. 1993. The Return of Base Generators. Groninger Arbeiten zur Germanistischen Linguistik 36, 1-74. Fanselow, Gisbert. 2001. ‘Features, θ-Roles, and Free Constituent Order,’ Linguistic Inquiry 32 (3), 405-437. Fanselow, Gisbert. 2002. ‘Quirky Subjects and Other Specifiers’, in B. Kaufmann and B. Stiebels (eds.), More Than Words, Akademie Verlag, Berlin, 227-250. Fanselow, Gisbert. 2004. Cyclic Phonology-Syntax Interaction: Movement to First Position in German, In: Shin Ishihara and Michaela Schmitz (eds.), Interdisciplinary Studies on Information Structure [Working Papers of the SFB 632, 1]), Potsdam, 1-42. Fanselow, Gisbert and Ćavar, Damir. 2001. Remarks on the economy of pronunciation. In: G. Müller & W. Sternefeld (eds.), Competition in Syntax, Mouton de Gruyter, Berlin, 107150. Fanselow, Gisbert and Ćavar, Damir. 2002. Distributed Deletion. In: Artemis Alexiadou (ed.), Theoretical Approaches to Universals. Benjamins, Amsterdam, 65–107. Felix, Sascha W. 1985. Parasitic Gaps in German. In: Werner Abraham (ed.): Erklärende Syntax des Deutschen. Narr, Tübingen, 173-200. Féry, Caroline. 2007. Wide Focus Object Fronting. Ms., University of Potsdam. Féry, Caroline and Ishihara, S. 2005. Interpreting Second Occurrence Focus, Ms., Potsdam. 39

Féry, Caroline & Frank Kügler. 2006. German as a Tone Language, Ms., Potsdam. Fox, Danny & David Pesetsky. 2003. Cyclic Linearization and the Typology of Movement. Ms., MIT. Fox, Danny & David Pesetsky. 2005. Cyclic Linearization and its Interaction with other Aspects of Grammar. Theoretical Linguistics 31, 235-262. Frey, Werner. 2004. A medial topic position for German. Linguistische Berichte 198, 153190. Frey, Werner. 2005. Zur Syntax der linken Peripherie im Deutschen. To appear in: F. d’Avis (ed.), Deutsche Syntax: Empirie und Theorie, Göteborger Germanistische Forschungen. Frey, Werner & Karin Pittner 1998: Zur Positionierung der Adverbiale im deutschen Mittelfeld. Linguistische Berichte 176, 489-534. Gärtner, Hans Martin. 1996. Review of M. Reis (ed.), Wortstellung und Informationsstruktur, Niemeyer, Tübingen, 1993, Linguistische Berichte 161, 90-94. Grewendorf, Günter. 1980. Funktionale Satzperspektive und deutsche Wortstellung. Linguistische Berichte 66, 28-40. Haider, Hubert. 1981. Empty Categories. On some differences between English and German. Wiener Linguistische Gazette 25:13-36. Haider, Hubert. 1986. Affect alpha. Linguistic Inquiry 17, 113-126. Haider, Hubert. 1990. Topicalization and other puzzles of German syntax. In: G. Grewendorf and W. Sternefeld (eds.), Scambling and Barriers, 93-112. Haider, Hubert and Inger Rosengren, Inger. 2003. Scrambling: Non-triggered Chain Formation in OV languages. Journal of Germanic Linguistics 15:203-267. Hartmann, Katharina and Zimmermann, Malte. 2007. In Place - Out of Place: Focus in Hausa. In: Kerstin Schwabe and Susanne Winkler (eds.), On Information Structure, Meaning and Form: Generalizations across languages. Benjamins, Amsterdam. 365-403. Heck, Fabian. 2004. A Theory of Pied-Piping. Doctoral dissertation, Universität Tübingen. Holmberg, Anders. 2000. Scandinavian Stylistic Fronting: How Any Category Can Become Expletive, Linguistic Inquiry 31 (3), 445-483. Horn, George. 2003. Idioms, metaphors and syntactic mobility. Journal of Linguistics 39, 245–273. Horvath, Julia. 2007. Separating Focus Movement from Focus. In: Karimi, Simin, Vida Samiian and Wendy K. Wilkins (eds.), Phrasal and Clausal Architecture: Syntactic derivation and interpretation. Benjamins, Amsterdam. 108–145. Ishihara, Shinichiro. 2006. Major phrase, focus intonation and multiple spellout. Ms. Universität Potsdam. Jacobs, Joachim. 1991. Focus Ambiguities. Journal of Semantics 8, 1-36. Jacobs, Joachim. 1996. Bemerkungen zur I-Topikalisierung. Sprache und Pragmatik 41,1-48. Jacobs, Joachim. 1999. Informational Autonomy. In: P. Bosch and R. van der Sandt (eds.), Focus: Linguistic, Cognitive, and Computational Perspectives, CUP, Cambridge, 56-81. Janko, Tat’jana. 2001. Kommunikativnye strategii russkoj reči, Jazyki slav. kul’tury, Moskva. Kahnemuyipour, Arsalan. 2004. The syntax of sentential stress. PhD dissertation, University of Toronto. Kaiser, Elsi. 2000. The Discourse Functions and Syntax of OSV Word Order in Finnish. In: A. Okrent and J. Boyle (eds.), Proceedings from the Main Session of Chicago Linguistic Society’s 36th Meeting, vol. 36 (1), Chicago Linguistic Society, Chicago, 179-194. Kathol, Andreas. 2001. On the Nonexistence of true Parasitic Gaps in Standard German. In: Peter W. Culicover & Paul M. Postal (eds.). Parasitic Gaps. MIT Press (= Current Studies in Lingustics 35), Cambridge, 315–338. Kenesei, István. 1998. Adjuncts and Arguments in VP-Focus in Hungarian. Acta Linguistica Hungarica 45, 61-88. Kratzer, Angelika and Elisabeth Selkirk, 2006. Default phrase stress, prosodic phrasing and the spellout edge: The prosodization of verbs. Ms., University of Massachusetts, Amherst. 40

Krifka, Manfred. 1994. Focus and Operator Scope in German. In: P. Bosch and R. van der Sandt (eds.), Proceedings of the Conference on Focus and Natural Language Processing, Volume 1: Intonation and Syntax, ILL, Heidelberg, 133-152. Kučerová, Ivona. 2006. The T Extension Condition. Ms., MIT. Kvam, Sigmund. 1983. Linksverschachtelung im Deutschen und Norwegischen. Eine kontrastive Untersuchung zur Satzverschränkung und Infinitivverschränkung in der deutschen und norwegischen Gegenwartssprache. Niemeyer, Tübingen. Ladd, D. Robert. 1996. Intonational Phonology, CUP, Cambridge. Lenertová, Denisa. 2004. Czech pronominal clitics. Journal of Slavic Linguistics 12 (1-2), 139-176. Lenertová, Denisa and Junghanns, Uwe. 2004. Left-peripheral Focus Exponents. Ms. Universität Leipzig. Lenertová, Denisa and Junghanns, Uwe. 2007. Fronted Focus Exponents with Maximal Focus Interpretation in Czech. In: K. Schwabe and S. Winkler (eds.), 347-363. Mehlig, Robert. 1994. Explikative Äußerungen. Überlegungen zur Informationsstruktur. In: R. Mehlig (ed.), Slavistische Linguistik 1993, Sagner, München, 181-224. Meyer, Roland. 2004. Syntax der Ergänzungsfrage. Empirische Untersuchungen am Russischen, Polnischen und Tschechischen. Sagner, München. Müller, Gereon. 1998. Incomplete category fronting. Kluwer, Dordrecht. Müller, Gereon. 2004. Verb Second as vP First. Journal of Comparative Linguistics 7(3), 179-234. Müller, Gereon. 2007. Towards a Relativized Concept of Cyclic Linearization. Ms., Leipzig. Müller, Gereon & Wolfgang Sternefeld. 1993. Improper Movement and Unambiguous Binding. Linguistic Inquiry 24, 461-507. Müller, Gereon & Wolfgang Sternefeld. 1996. A-bar chain formation and economy. Linguistic Inquiry 27: 480 – 511. Müller, Stefan. 2003. Mehrfache Vorfeldbesetzung. Deutsche Sprache 31, 29-62. Neeleman, Ad and Szendrői, Kriszta. 2004. Superman sentences. Linguistic Inquiry 35, 149159. Nunberg, Geoffrey, Sag, Ivan, and Wasow, Thomas. 1994. Idioms. Language 70, 491–538. Prince, Ellen F. 1999. How Not to Mark Topics: ‘Topicalization’ in English and Yiddish. Texas Linguistics Forum, University of Texas Press, Austin, chapter 8. Puig-Waldmüller, Estela. 2006. Wörter stehen da einige. On nominal split constructions in non standard Viennese German. Diploma thesis, Universitat Pompeu Fabra, Barcelona. Reich, Ingo. 2003. Frage, Antwort und Fokus, Akademie Verlag, Berlin. Reinhart, Tanya. 1981. Pragmatics and Linguistics: An Analysis of Sentence Topics. Philosophica 27, 63-94. Rizzi, Luigi. 1991. Residual Verb Second and the WH Criterion. Technical Rep. 3, Geneva. Rizzi, Luigi. 1997. The Fine Structure of the Left-Periphery. In: Liliane Haegeman (ed.), Elements of Grammar: Handbook of Generative Syntax, Kluwer, Dordrecht, 281-337. Rizzi, Luigi. 2004. On the Form of Chains: Criterial Positions and ECP Effects. Ms., Siena. Rooth, Mats. 1996. Focus. In: Shalom Lappin (ed.), Handbook of Contemporary Semantic Theory, OUP, Oxford, 271-296. Selkirk, Elisabeth. 1995. Sentence Prosody: Intonation, Stress, and Phrasing. In: J. Goldsmith (ed.), The Handbook of Phonological Theory, Blackwell, Cambridge, MA, 550-569. Speyer, Augustin. 2007. Die Bedeutung der Centering-Theorie für Fragen der Vorfeldbesetzung im Deutschen. Zeitschrift für Sprachwissenschaft 26, 83-115. Szabolcsi, Anna. 1981. The semantics of Topic-Focus-Articulation. In: Jan Groenendijk et al. (eds). Formal Methods in the Study of Language. Amsterdam: Matematisch Centrum. Szendrői, Kriszta. 2001. Focus and the Syntax-Phonology Interface. PhD thesis, University College London.

41

Svolacchia, Marco, Mereu, Lunella and Puglielli, Annarita. 1995. Aspects of Discourse Configurationality in Somali. In: Katalin É. Kiss (ed.), Discourse Configurational Languages. OUP, Oxford, 65-98. Thiersch, Craig. 1978. Topics in German Syntax. Ph.D. dissertation, MIT. Toman, Jindřich. 1999. On clitic displacement. In: Mila Dimitrova-Vulchanova and Lars Hellan, eds. Topics in south Slavic syntax and semantics. Amsterdam: Benjamins, 205-28. Travis, Lisa. 1984. Parameters and Effects of Word Order Variation. PhD dissertation, MIT. Truckenbrodt, Hubert. 2004. Final Lowering in Non-Final Position. Journal of Phonetics 32, 313-348. Truckenbrodt, Hubert. 2005. The Syntax-Phonology Interface. To appear in P. de Lacy (ed.), The Cambridge Handbook of Phonology, CUP, Cambridge. Vallduví, Enric. 1992. The Informational Component, Garland, New York. van Hoof, Hanneke. 2003. The Rise in the Rise-Fall Contour: Does it Evoke a Contrastive Topic or a Contrastive Focus? Linguistics 41 (3), 515-563. Wagner, Michael. 2005. Asymmetries in Prosodic Domain Formation. MIT Working Papers in Linguistics 49, 329-367. Walker, Marilyn, Joshi, Aravind K. and Prince, Ellen. 1998. Centering Theory in the Discourse. OUP, Oxford. Webelhuth, Gert. 1989. Syntactic Saturation Phenomena and the Modern Germanic Languages. Ph.D. dissertation, University of Massachusetts, Amherst. Wedgwood, Dan. 2003. Predication and Information Structure. A Dynamic Account of Hungarian Pre-verbal Syntax. PhD thesis, University of Edinburgh. Wilder, Chris and Damir Ćavar. 1994. Long Head Movement? Verb Movement and Cliticization in Croatian. Lingua, 1-58. Williams, Edwin. 2003. Representation Theory. MIT Press, Cambridge, MA. Wold, Dag. 1995. Binding Foci in situ. Ms., MIT. Wurmbrand, Susi. 2001. AGREE - The other VP-internal subject Hypothesis. In: K. Megerdoomian and L.A. Bar-el (eds.), WCCFL 20 Proceedings, 635-648. Zubizarreta, Maria Luisa. 1998. Prosody, Focus and Word Order, MIT Press, Cambridge, MA.

42