Not another book on Verb Raising
This thesis provides a novel analysis of the word order variation in three-‐verb clusters reported in the Syntactic Atlas of the Dutch Dialects (SAND). On the basis of distributional correlations between order variation in verb clusters and interruption of the verb cluster by non-‐verbal constituents, it is argued that only 1-‐2-‐3 and its mirror image 3-‐2-‐1 are truly verbal clusters. All other orders attested in SAND are argued to involve non-‐verbal elements: adjectival participles and nominal infinitives. This analysis dispenses with movement in the derivation of verb clusters, an improvement over many previous accounts, as movement in this domain is unmotivated and, in certain cases, makes wrong predictions.
There are more crucial differences between the approach taken here and the existing literature. It has a different starting point, the geographic distribution of word orders. It uses a different methodology: a comparative ranking task, which includes word orders that do not occur in the informants’ own varieties. It presents several new findings. Speakers turn out to possess knowledge of word orders that do not occur in their own language varieties. This is shown to follow from properties of human grammar. Neither familiarity nor properties of language processing can account for these results. Verb clusters are base-‐generated in a low structural position in the clause. An analysis in which auxiliaries are generated directly as functional heads is untenable for Dutch varieties. There is a cut-‐off point for cluster interruption, which is parameterized. In West-‐Flemish, it lies at vP, only elements that are merged below vP can interrupt the verb cluster.