Abstract for Owen Rambow's talk

Zu-infinitives in German: Lexicon or syntax?
Owen Rambow (Columbia)
September 20, 2005 (Tuesday)
6:30 PM - 8:00 PM; Room 7102, The CUNY Graduate Center

This talk addresses the issue of embedded zu-infinitival clauses in German from the perspective of a formal grammatical framework, Tree Adjoining Grammar (TAG) and related tree rewriting systems. Two radically different analyses have been proposed for this construction. According to the "syntactic'' analysis, embedded infinitivals in German are analyzed essentially as in English, namely as clausal complementation. The "lexical'' analysis suggests that there is a process by which two verbs are combined into a single lexical unit (in some sense), which is head of a single syntactic projection. The incorporation analysis has become widely (in fact, nearly universally) accepted in one form or another in both the transformational and non-transformational literature, and has an undeniable intuitive appeal to the native-speaker linguist. However, this paper argues that there is empirical evidence against the incorporation analysis. Furthermore, methodological parsimony requires that the introduction of machinery to handle the merging of argument lists of two verbs (as required under the incorporation analysis) be motivated by the data, and that no alternate account (which does not rely on the additional mechanism) be available. Unfortunately, the status of much of the crucial data is quite murky. As a consequence, theoretically significant choices in the machinery of syntactic theories need to be made on the basis of difficult grammaticality judgments. This paper does not argue for a syntactic solution as such. Instead, it suggests that the stark contrast between the two analyses is in fact an artifact of the grammatical frameworks in which the construction has been analyzed. The paper proposes an analysis in a grammatical formalism in which all phrase structure is built incrementally in a formal derivation, and in which node labels are represented largely as features. There is no separate "lexical" phase of the derivation. In such a system, it is argued, the difference between the syntactic and the incorporation analyses can be interpreted as a difference in the ordering of steps in the derivation.