Localism versus globalism in morphology and phonology by David Embick (review)

David Embick’s book presents a detailed contrastive analysis of two approaches that are currently explored in linguistic theory: localism (Distributed Morphology and the Minimalist Program) and globalism (Optimality Theory). The focus of this book lies in the assessment of these approaches, which provide two opposing views of the architecture of grammar. The author aims to identify the best way to account for allomorphy in morphosyntactic data of the world’s languages. Comparing the mechanisms and theoretical implications of Distributed Morphology (DM; Halle and Marantz 1993) and Optimality Theory (OT; Prince and Smolensky 1993), Embick concludes that the localist theory is better equipped to capture morphosyntactic generalizations with regards to allomorphy. Without a doubt, this book makes a contribution to the advancement of linguistic theory as it engages the reader in the debate over which of these two influential theories is on the right track.

The book consists of seven chapters divided into two Parts (Part I, “A Localist Theory” and Part II, “Phonologically Conditioned Allomorphy”). In the Introduction, Embick discusses the main differences between localism (also called serialism) and globalism (also called parallelism): the localist theory assumes that the phonology interprets the output of the morphosyntactic derivation, whereas in the globalist view phonological constraints are put on the same line as morphological and syntactic constraints. Localism views morphology and phonology as two separate modules, while globalism claims that there is an interaction between them. In this chapter, the author provides a description of the two types of allomorphy (grammatically versus phonologically conditioned allomorphy), to which he refers in the subsequent chapters. Here, the nature of competition is also discussed: in DM competition is considered to be local (between Vocabulary Items competing for insertion), while in OT competition is global (between output candidates going through the evaluation of phonological, morphological, and syntactic constraints for a given output).

While in the Introduction the author introduces the reader into the debate over which approach is better, in Part I (Chapters 2 and 3) he focuses specifically on a localist theory. The ultimate goal of this book is to convince the reader that a localist theory is best equipped to account for allomorphy. Embick discusses the most recent developments (e.g., cyclicity, local competition, linear adjacency) in DM, which he implements to account for allomorphy patterns, and proposes C₁-LIN (“Cyclic-Linear”) theory. The central tenet of C₁-LIN Theory is that abstract morphosyntactic material is spelled out cyclically. Allomorphy is possible only for material within the same cycle; each cycle represents a local domain. Once spelled out (after undergoing the process of Vocabulary Insertion), the material from a cycle is not accessible to the nodes of a higher cycle and allomorphy is predicted not to occur.

According to this proposal, heads are either cyclic or non-cyclic. Only cyclic heads can be spelled out and can trigger the spell out of the structure below them. The cyclic head y triggers the spell-out of the cyclic head x in the complement of y. The spelled-out material includes the complement of x, the head x and any noncyclic nodes attached to x prior to its merge with the cyclic head y. It is emphasized that the cyclic head y triggering spell-out is not itself present in the same cycle as the complement of x and hence cannot show sensitivity to it. As an example, in the structure [[√Root x] y] where x and y are cyclic heads, y cannot show Root-sensitive allomorphy. Although the notion of cyclicity is not novel, Embick brings many invaluable details to his account and gives a very detailed description of possible scenarios of allomorphy allowed under the localist view.

In addition to cyclicity, Embick discusses the linear notion of locality. He claims that contextual allomorphy is restricted to cases where a node can see another node only when it is concatenated with it (linearly adjacent). Embick admits...