It has been claimed that the functional architecture of German clauses involves heads such as [Topic] that are defined by their information structure value, and that movement to the specifier position of such heads is (partially) responsible for free word order in German. This paper argues that this view is misguided. (i) There is no evidence for a syntactically defined topic position in the middle field of German clauses. (ii) The distinction between marked and unmarked serialization must also be made for structures that do not involve movement. (iii) Movement in the interest of information structure is often altruistic
Filling first positions
(2003)
In the recent literature there is a hypothesis that the human parser uses number and case information in different ways to resolve an initially incorrect case assignment. This paper investigates what role morphological case information plays during the parser’s detection of an ungrammaticality or its recognition that a reanalysis is necessary. First, we compare double nominative with double accusative ungrammaticalities in a word by word, speeded grammaticality task and in this way show that only double nominatives lead to a so-called ”illusion of grammaticality” (a low rate of ungrammaticality detection). This illusion was found to disappear when the second argument was realized by a pronoun rather than by a full definite determiner phrase, i.e. when the saliency of the second argument was increased. Thus, the accuracy in recognizing an ungrammaticality induced by the case feature of the second argument is dependent on the type of this argument. Furthermore, we found that the accuracy in detecting such case ungrammaticalities is distance sensitive insofar as a shorter distance leads to a higher accuracy. The results are taken as support for an ”expectationdriven” parse strategy in which the way the parser uses the information of a current input item depends on the expectation resulting from the parse carried out so far. By contrast, ”input-driven” parse strategies, such as the diagnosis model (Fodor & Inoue, 1999) are unable to explain the data presented here.