A Derivational View on Movement Constraints

Abstract

This paper explores conditions on movement in a strongly derivational organization of grammar. The setting is Ulf Brozsiewski's model of syntactic derivations [4]. It offers a local encoding of movement dependencies that combines properties of transformational and feature-based approaches. It is based on simple assumptions about syntactic expressions and the operations defined on them, and is thus suitable to investigate how much is needed to derive, or at least express, fundamental characteristics of movement. Moreover it reduces the amount of representations needed for syntactic operations to a minimum and thus constitutes an exploration of how derivational syntax can be.The outline of the paper is the following. In section 1, Brosziewski's view on derivations will be introduced. After that, I address a problem that it faces with extraction from phrases that are extracted themselves, and show how to solve it. Then I will turn to generally assumed constraints on movement. In section 3, I review how the Condition on Extraction Domain is expressed in Brosziewski's model. Then, in section 4, I will consider how weak islands can be captured. Furthermore, section 5 deals with accounting for minimality effects. Finally, in section 6, I will also show how across-the-board extraction can be incorporated.1 On derivations Syntactic expressionsIn approaches like HPSG and the Minimalist Program, features play a central role in constructing and manipulating syntactic structures. Brosziewski's model of syntactic derivations shares this assumption. Simple expressions are taken to be feature bundles containing phonological, syntactic, and semantic features. In the following, phonological and semantic features will be ignored and focus will be on syntactic features only. With respect to these, Brosziewski sticks to a very simple feature system. For encoding syntactic dependencies he assumes that features come in two varieties: as plain features f , and as features ∧ f with a prefix that indicates that it wants to be matched with a corresponding plain feature. Following Adger [2] a. o., I add a distinction between two types 0 For thorough remarks and helpful suggestions, I am very grateful to Gereon Müller, Eric Reuland, and Andreas Pankau.