Sjef Barbiers (Meertens Instituut and Utrecht University): Restructuring Finite Thought and Will

The finite complements of bridge verbs such as THINK and WANT look completely identical superficially in many languages and share the property that they allow long dependencies . This is illustrated for Dutch in (1a-c).
(1) a. Ik denk [dat Robben speelt].
I think that Robben plays
b. Ik wil [dat Robben speelt].
I want that Robben plays
?I think/want that Robben will play.?
c. Wie denk/wil je dat er ____ speelt?
who think/want you that there ____ plays
In this talk I present a number of observations, partly new, that show that the clausal complements of the two verbs must have radically different syntactic structures in Dutch: (i) only THINK allows for long Wh-doubling and Wh-scope marking constructions; (ii) only THINK allows for Neg-raising; (iii) Both types of verbs allow for adverbs that surface in the embedded clause but have obligatory matrix scope, but this involves MoodSpeechAct adverbs in the case of THINK and AspRepetitive adverbs in the case of WANT; (iv) Both verbs allow for overt raising of embedded adverbs into the matrix field, but the types of adverbs that can do this are partially distinct for the two verbs; (v) The complement of THINK but not WANT can contain the higher Mood, Modality and Temporal adverbs (in the sense of the Cinque hierarchy). To capture these differences I propose a new analysis for bridge verbs with two main characteristics: (i) I extend Cinque?s (2001) analysis of infinitival restructuring constructions to finite complements. Thus, THINK and WANT are generated in different designated functional positions in the Cinque hierarchy INSIDE the embedded clause and raise to the position of the matrix verb (by internal merge). (ii) The finite complement of WANT in Dutch is defective in that it lacks the ForceP level (in the sense of Rizzi 1997) and the Mood, higher T and alethic modality levels. This analysis explains the bridge property of bridge verbs, i.e. why structures like (1) show both monoclausal and biclausal properties.