Abstract

this paper, we will argue for the second view by presenting a processing model for German. German is an interesting language to study from the point of view of both competence syntax and performance because it not only allows scrambling, but also topicalization of arguments. Topicalization refers to the movement of a single element into the sentence-initial position in the root clause. Because German is a verb-second language, in every sentence, some element must topicalize. Like scrambling (and like topicalization in English), topicalization in German can create unbounded dependencies. However, the two types of movement differ in terms of their linguistic properties; for example, scrambling can create new anaphoric bindings, while topicalization cannot (Webelhuth, 1989). In addition to the linguistic differences, there is also a processing difference: long-distance topicalization into sentence-initial position appears to be easier to process than long-distance scrambling of the same element over a similar distance. A simple processing account that somehow measures the number of intervening lexical items must fail. Thus, not only does the German data call for a processing model, but the model must be sensitive to subtle differences in the constructions involved. Our processing model for free-word order languages has two important properties: ffl The processing model provides a metric that makes predictions about processing difficulty on an openended scale. This property allows us to verify our model with respect to the results from psycholinguistic experiments as well as from native-speaker intuition.

Citations