Localism versus Globalism in Morphology and Phonology

2. A Localist Theory

2 A Localist Theory This part of the book develops a theory of allomorphic locality that is centered on the interaction of cyclic and linear locality domains. This theory is developed as an account of a number of empirical generalizations that are presented in the course of the discussion. If something like this theory is on the right track, then morphology and phonology show the kinds of properties that are expected in a localist view of grammatical architecture. In particular, if the key generalizations about allomorphy in natural language can be explained in a theory with sharp locality conditions and do not require a theory that makes reference to (e.g.) competing forms or to the phonological properties of outputs—things that can be referred to in globalist architectures—then we have support for a localist view. This part of the book presents the details of a localist view; explicit comparisions with globalist alternatives are made in part II. The theory of allomorph selection that is developed in these chapters is part of a localist, serialist theory of grammar. An important aspect of this theory, a version of Distributed Morphology, is that the syntax generates hierarchical structures that are subjected to further computations in the interface components PF and LF. I assume that the syntax operates in terms of locality conditions that arise from cyclic derivation. A further assumption, one that is automatic in a syntactic approach to morphology like the one advanced here, is that conditions on locality in syntax also define behavior in the interface components . By reducing at least a certain amount of morphological interaction to cyclic derivation, this theory follows a long line of earlier theories, originating with the theory of the transformational cycle in Chomsky and Halle’s (1968) The Sound Pattern of English and other pioneering works in generative phonology. The basic empirical question that is addressed in this chapter and the next concerns the conditions under which a node may have its phonology determined by items in its context. That is: (1) Locality of allomorphy question For the contextual allomorphy of some node, what factors in the environment of that node are visible? Given the architectural premises of the theory that I assume, the key theoretical questions center on how morphological e¤ects are determined in a system that has (i) cyclic derivation, (ii) structural (i.e., hierarchical) relations determined by the syntax, and (iii) linear relations derived from the hierarchical structure (in the PF component of the grammar, by hypothesis ). It is important to distinguish (i)–(iii) in this way because conditions stated in terms of cyclic, hierarchical, and linear representations enforce conditions on locality that are in many cases distinct from one another. The relations that are important for di¤erent types of e¤ects in morphology broadly speaking could thus be defined in di¤erent ways, and, ultimately, empirical evidence must determine which of (i)–(iii) (or a combination) is active for any particular phenomenon. The theory presented below explores the idea that a kind of strict linear adjacency is required for contextual allomorphy, in a way that interacts with a cyclic theory of what is ‘‘active’’ at a particular stage of a derivation . The central idea is that a node can be sensitive to another node for the purposes of allomorphy only when the two nodes are linearly adjacent to one another. There are, however, cases in which surface linear adjacency is not enough, and this is where cyclic structure plays a role: it is only when two nodes are present in the same PF cycle that they may potentially interact. The cyclic and linear notions of locality appealed to in this theory are logically independent of each other. It is an empirical hypothesis of this work that these two distinct types of locality interact to account for attested patterns of allomorphy in natural language. 2.1 Syntax and Morphology The theory presented here is a piece-based, syntactic theory of morphology : Distributed Morphology, along the lines of Embick and Marantz 2008 in particular. Complex expressions are built out of discrete pieces (morphemes), and it is in the syntax (or in terms of relations derived from syntactic structures) that the composition of morphemes takes place. Another fundamental component of the theory is the idea that morphology is realizational. This means that at least some morphemes possess no phonology as part of their basic representation; rather, phonological material is added to such morphemes in the PF component of...