Volker Struckmeier (U of Cologne)

Deriving German scrambling without information structure It has been argued by many authors that information structure (IS) controls scrambling movements in German. IS factors were therefore introduced into syntactic derivations as functional heads (Top0, etc.). We show that these IS approaches' predictions are fundamentally wrong. Prosodic and semantic factors, we demonstrate, influence scrambling and derive the effects of word order changes accurately. Focussed arguments allegedly cannot scramble in German (cf. already Lenerz 1977). We show, however, that bearing main stress (marked by CAPitals) prevents arguments from scrambling, not their IS status – wherever the two properties can be dissociated. E.g., unstressed arguments that are F-marked (in the sense of Schwarzschild 1999) scramble freely: 1.Q: What did you give to who? A': ?Ich habe dem MannF das GELDF wohl gegeben. A'': Ich habe dem MannF wohl das GELDF gegeben. I have to-the man part the money given 'I have probably given the money to the man'.

(focusses both objects in answer) (scrambling stressed objectF odd) (scrambling unstressed objectF ok)

IS analyses also claim that topics scramble obligatorily (judgement by Frey 2004): 2. Context: {I will tell you something about Fritz.} Cont'd: #Morgen wird wohl eine reiche Dame den Fritz heiraten. tomorrow will part a rich lady the Fritz marry 'Tomorrow, a rich lady will supposedly marry Fritz.'

(Fritz designated topic) (Fritz odd in-situ)

The problem with the discourse in (2), we argue, is much too subtle to be syntactic in nature (many speakers accept the cont'd sentence). Subtle problems are, however, caused by the stress placement. In (2), neither default stress on the (Given) object is possible, nor any other (e.g. narrow/contrastive) stress placement. As soon as contrastive foci (!) become possible in a context, the placement of the topic (!) becomes unproblematic, as in (3): 3.Context: {Fritz has never found a woman. But let me tell you something NEW about Fritz:} Cont'd: Morgen wird wohl eine reiche Dame den Fritz HEIraten. (Fritz now ok in-situ) Topicality thus is irrelevant in (2/3), and prosody is the relevant factor in both cases. To represent scrambling movements, we propose a double-sided approach: Elements can move syntactically to achieve semantic effects on outcome (Chomsky 1999). Since German is scope-transparent (Bobaljik & Wurmbrand 2012), movement of the QP to the edge of vP is both acceptable at the semantic interface and reflected in the PF spellout in (4): 4.

PF:

C...

DP...

[QP

[Neg [vP DP QP V]

T]]]]]]

E.g.:

wenn [ein Arzt]DP [allen Patienten]QP nichtNeg helfenV kannT when a doctor all patients not heal can 'when for [all patients]x, a doctor cannot help themx' (salient reading: ∀¬)

The second scrambling operation we assume is a-semantic in nature: Following Biberauer & Richards (2006), we assume that vP moves to SpecTP. vP contains copies of all arguments, too – but since these arguments are too deeply embedded to scope (or bind) out of vP, spellout decisions for these arguments are made not on a semantic basis, but for prosodic reasons: 6. wenn [TP [vP ein Arzt /ALLen Patienten...] [NICHTF [vP ...allen Patienten helfen]] kann] 'when a doctor cannot help all the patients' (...but potentially some, salient reading ¬∀!) Spellout achieves a prosodic bridge contour (/SLASH... CAPS), when allen Patienten is spelled out in the high vP copy. Scope transparency cannot determine word order for vPinternal elements, so the observed mismatch between linearization and interpretation ensues. This 'subtractive architecture', we show, correctly predicts the interplay of semantic and prosodic scrambling factors in languages like German, without IS features or IS positions: • Syntax derives hierarchical configurations of constituents, and restricts these structures only on formal syntactic grounds. Syntax is not involved in representing IS properties. • The semantic interface rejects derivations that fail to adhere to semantic principles (Chomsky's duality of semantics) and therefore further restricts the set of possible structures. • For the subset of derivations converging at LF, semantic transparency is implemented in languages like German (in contradistinction to, say, English): Some linearization requirements are entered at LF (and implemented by PF in scope-transparent languages). • However, these syntacto-semantic structures are only partially deterministic with regard to word order: Some configurations (e.g., vP moving to TP, as pointed out above) do not lead to semantic effects and therefore do not trigger semantic linearization statements. • For word order properties undecided by core syntax and LF, PF will determine the remaining linearization options on prosodic grounds – explaining both the semantic 'mismatches' and the appearance of 'prosody-driven' scrambling (without look-ahead to PF or IS). In sum, possible word orders in German show syntactic, semantic and phonological requirements at work. Analyses that implement scrambling by IS projections made word order predictions that in general turned out to be much too harsh – and often quite simply false. These older analyses also never even attempted to describe the complex interplay between syntax, semantics and phonology in German scrambling that our analysis sets out to explain. References Bobaljik, Jonathan & Susi Wurmbrand 2012: "Word Order and Scope: Transparent Interfaces and the 3/4 signature". In Linguistic Inquiry 43.3, pp. 371-421. Chomsky, Noam 1999: "Derivation by Phase". In: MIT Occasional Papers in Linguistics 18, pp. 1-42. Frey, Werner 2004: "A Medial Topic Position for German". In: Linguistische Berichte 198, pp. 153-190. Lenerz, Jürgen 1977: Zur Abfolge nominaler Satzglieder im Deutschen. Tübingen: Gunter Narr. Schwarzschild, Roger 1999: "GIVENness, AvoidF and Other Constraints on the Placement of Accent". In: Natural Language Semantics 7, pp.141-177.