In recent work, Chomsky proposes a cyclic model of the syntax-phonology interface: phases are spelt out immediately after completion. In this paper, I would like to present some evidence from German that supports this view.

 

The line of argumentation is fairly simple. In verb second clauses, X(P)s other than the subject are attracted to first position because they bear a certain operator feature: wh, focus, topic. Normally, when the word or subconstituent bearing f is attracted to first position, the full maximal projection dominating f must be pied-piped, but this is not necessary in all cases. Sometimes, the only the word realizing a certain feature is moved. The word realizing a a wh-feature is the wh-word itself, but the word realizing a pragmatic feature of XP can be taken to be that element in XP that bears the special intonation connected with the pragmatic function.

 

We can observe now that the pragmatic topicalization of a verb phrase can be realized syntactically in a number of different ways: the complete VP can be placed into first position, but often, the preposing of the stressed particle, or of part of the direct suffices to express the topichood of VP. These elements moved to the left are the words in VP that would receive structurally determined accent by the nuclear stress rule.

 

The selection of elements for movement to first position thus depends on a property that is phonologically determined, viz. the outcome of the Nuclear Stress Rule.