<<   >>   TOC

1. German Sentence Structure


It is commonly assumed that embedded (verb final) clauses exhibit the basic sentence structure of German:


..., daß Dornröschen      schlief.

     that sleeping beauty slept.


The categorial status of the embedded clause has been cause of much debate, especially since the standard CP/IP-analysis as shown in (2) runs counter the widely held suspicion that tree structures are right-recursive in UG (cp. Kayne 1994).


Since German does not show a subject-object asymmetry like English, e.g. with respect to subject-island effects (German subjects can undergo V2), Haider (1993: 150) argues against the assumption of a structural position reserved for subjects (i.e. [Spec, IP]). Furthermore, it seems doubtful that German needs an independent sentence-final I-position (Haider: 60). He proposes that the finite verb is generated in situ, fully inflected and does not need to move for feature checking. Rather, it is licensed in its base position by a functional head which takes a VP-complement. This functional head ("F" for "finite") can be filled by either a typical complementizer (and thus functions as the counterpart to the English C-position) or by a verbal element in V2-constructions. Verbal elements in F-position trigger movement/generation of another element into [Spec, FP] in declarative sentences. Matrix counterparts of (1) would thus allow the following syntactic representation:



For sentences with auxiliaries or multiple verbs, Haider (1993) assumes that the verbal elements merge to form a complex verb as in (4a). In matrix sentences, the lowest verbal element is moved into F-position,[1] triggering [Spec, F] to be filled by either an expletive (4b), the subject (4c), or the verbal complex (4d).



            a.         ..., daß Dornröschen [VP [V geschlafen [V haben [V wird ]]]]

                             that sleeping beauty slept              have         will

                        ...that sleeping beauty will have slept.

            b.         Da wirdi Dornröschen geschlafen haben [V ei ].

            c.         Dornröschenj wirdi ej geschlafen haben [V ei ].

            d.         [VP Geschlafen haben [V ei ] ]j wirdi Dornröschen ej.[2]


Complex or simple verbs can be fronted alone and in conjunction with an object. Thus not only V in German is to be considered Vmax (Haider 1993: 280ff., cp. also Chomsky 1995a for a similar proposal), but also combinations of objects and V are maximal projections:


            a.         ..., daß der Prinz Dornröschen einen Kuß gegeben hat.

                             that the prince sleeping beauty a kiss   given      has

                             That the prince has given sleeping beauty a kiss...

            b.         [VP Gegeben [V ei ] ]j hati der Prinz Dornröschen einen Kuß ej.  

            c.         [VP Einen Kuß gegeben [V ei ] ]j hati der Prinz Dornröschen ej.

            d.         [VP Dornröschen einen Kuß gegeben [V ei ] ]j hati der Prinz ej.


Frey and Tappe (1992) and  Haider (1993) thus posit (6), allowing VP to be an 'open' projection (reminding of Fukui and Speas' 1986 proposal of multiple spec-positions):  



In my further discussion, I will assume that (6) accurately describes the basic sentence structure for German at least insofar as the internal organization of FP as well as the lack of an independent I-position to the right of VP is concerned. However, scrambling data (i.e. word order variations within VP) require modifications of the internal structure of VP. Therefore, I will later commit to a structure slightly deviating from (6) in its details and implications.

<<   >>   TOC


© Philipp Strazny 1997

[1]               It is necessary to assume complex verb structures if the Shortest Move condition is in fact a universal, and if there is truly no evidence for a higher I position. Only in complex verbs as shown in (4a), all verbal elements would be considered equidistant from [Spec, FP], such that the lowest element can move. 

[2]               Haider (1993) never explicitly states that fronting of the verbal complex involves also fronting of the trace left by wird. This, however, follows from his assumptions that (a) tree structures are right-recursive and (b) only maximal projections are subject to topicalization (i.e. V2-fronting). Tagging along of traces may not pose a problem in itself, since it is commonly assumed that applications of Move-a allow full reconstruction at LF.