Gereon Mueller (Universitaet Leipzig)
A standard assumption of earlier versions of generative grammar was that transformations can remove syntactic structure — either complete subtrees of a phrase marker (as in transformations like Passive, Equi NP Deletion, or Indefinite Object Deletion), or just certain nodes or shells (see, e.g., Tree Pruning as envisaged by Haj Ross, or Chomsky’s concept of S-bar Deletion). Subsequently, based on a more principled approach to syntactic transformations with the Projection Principle at its core, the hypothesis that trees could genuinely shrink in this way was generally abandoned. Current minimalist approaches maintain this assumption by postulating that syntactic operations can make trees grow (via Merge) or stay the same size (via Agree) but can never make them shrink. Against this background, and with the Projection Principle long gone, I will argue that there is every reason to postulate a general operation Remove that is as a complete mirror image of Merge and obeys exactly the same constraints (e.g., featural triggers, strict cyclicity). Evidence for Remove is partly conceptual (upon closer inspection, such an operation turns out to be implicitly adopted in standard approaches to minimalist syntax anyway), but mainly empirical (also see Chomsky (2014), Pesestky (2016)). I will illustrate the empirical relevance of Remove on the basis of six case studies from the syntax of German: (i) passive, (ii) applicatives,
(iii) ellipsis, (iv) restructuring, (v) complex prefields, (vi) tough-movement (= modal passive).