<<   >>   TOC

6. Concluding Remarks

The evidence against movement analyses of scrambling seems to be overwhelming:

        the characteristics of scrambling structures do not fit into the standard analytical framework that otherwise seems to be descriptively adequate for movement operations

        the data speak against the assumption of chain formation, since scrambling structures do not show relevant trace effects

        it is hard to see how scrambling as movement could be mechanically motivated, since analyses involving feature checking would grant either too little or too much optionality to scrambling.

Thus, I am led to believe that scrambling structures are base generated. Since individual verbs show strong preferences for argument orderings which speak against a structural symmetry between syntactic representations and propositional or thematic relations, it appears reasonable to either assume direct encoding of deviating word orders in the lexicon or presyntactic mechanisms that would ensure such word orders. Such mechanisms may be identifiable as e.g. definiteness or specificity effects, which, however, require further investigation.

To be able to account for certain V2-phenomena, I committed to a VP/IP structure for German that on the one hand allows to categorize both the verb alone as well as any verb/object conjunction as a maximal (moveable) projections, and on the other hand provides a structural distinction for arguments that can or cannot move in conjunction with the verb. I proposed that VP and IP form interdependent tiers in a multidimensional sentence structure, where each individual tier may introduce independent constraints into the syntax. Further research will have to show independent evidence for syntactic tiers, but it seems reasonable to hope that a tier analysis provides a more powerful apparatus than the traditional 2-dimensional framework. It is hard to see how e.g. intonation and focus could be integrated into a plain tree structure, in spite of the fact that they undoubtedly have semantic effects. 

<<   >>   TOC

Philipp Strazny 1997