Against scrambling as an instance of Move-alpha

1. Introduction "Scrambling" is a cover term that arose from Ross (1967), and that roughly means that two constituents can be permuted if they are clause mates. The languages of the world differ greatly as to the extent to which they allow for such permutation. Syntactic research has recently given an increased attention to this phenomenon which has borne a clearer understanding of the data as well as concrete proposals as to how scrambling can be explained by VG and without recourse to ad-hoc means such as "stylistic transformations" or "PF-operations" (see Williams 1984 and others). At the moment there are two major proposals in GB-type generative grammar. According to the first, scrambling is an instance of Chomsky-adjoining an XP (NP, PP, to a more limited extent also AP or ADV) to VP or to IP (perhaps also AP). In other words, it is movement to an A'-position. This view has recently been advocated for German by Webelhuth (1985, 1989, 1990)-Webelhuth (1989) actually concludes in the end that scrambling may be halfways between A'-and A-movement with respect to binding theory-, Sternefeld (1990) and Felix (1985); Bennis-Hoekstra (1985) take a similar position for the analysis of Dutch. I While there is little doubt that certain instances of scrambling can be adequately captured in this way, there are cases which obviously cannot be the result of" A'-movement. Relevant observations to which we will turn below have led to the proposal that certain subcases of scrambling are instances of A-movement. 2 The goal of this investigation is twofold. We first want to argue that the core cases of scrambling in German can neither be captured adequately by an adjunction operation nor by some movement to an argument position Le., we will deny a movement account altogether. Second we want to sketch a base-generation account that is not committed to the inadequacies of a non-configurational account of German syntax. It will be shown that this proposal makes use of morphosyntactic and lexical properties that are independently attested in the language. We will first turn to evidence against A'-movement, then to evidence against A-movement, and finally propose an alternative account that is not committed to syntactic movement at all.