Word order by Jae Jung Song (review)

This book presents a very useful overview of a considerable body of literature. To survey a topic as big as word order in a book of manageable size, it was necessary to make choices about what to omit. Jae Jung Song’s choices will not make all readers happy, but they are not unreasonable ones. First, the literature covered is largely confined to work of the last thirty years. Second, the only theories of grammar considered are the minimalist program (MP) and optimality theory (OT).¹ As a consequence, some extremely interesting ideas about how to handle word-order variation never get mentioned. In particular, there is no mention of the idea of decoupling linear precedence and immediate dominance in a phrase structure grammar, an idea that led to some productive research in generalized phrase structure grammar (Gazdar & Pullum 1981), lexical-functional grammar (Falk 1983), and head-driven phrase structure grammar (Reape 1993).

The book has seven chapters: brief introductory and concluding chapters sandwich one surveying what is known about word-order typology (which S abbreviates LT, for linguistic typology), two on MP, one on OT, and one on what S calls ‘the performance-based approach’—that is, corpus and experimental research on word order.

Although S discusses these as though they were alternative theories of the same thing, they are in many ways not really comparable. LT seeks inductive generalizations over directly observable patterns, with relatively little attention devoted to explaining those generalizations. The performance-based approach tries to explain word-order patterns (both within and across languages) on the basis of processing efficiency²—that is, what makes utterances easy or hard to produce and comprehend. MP, by contrast, has little concern with what is directly observable or with the efficiency of linguistic processing. Rather, it seeks to deduce properties of language from three ‘dimensions to the minimalist position: (1) virtual conceptual necessity; (2) economy; and (3) symmetry’ (80). Facts about languages play a role in this enterprise only to the extent that they can be shown to follow from or contradict the analyses so deduced. The OT research S presents shares a largely top-down approach with MP, but, as S writes, ‘OT needs to take into account what word order actually looks like on the surface’ (183). Later on the same page, S characterizes LT as ‘data-driven’, and MP (and its transformational predecessors) as ‘theory-driven’, and says ‘OT seems to strike a balance’ between the two.

As a practitioner of the ‘performance-based approach’ who has not kept up with the other literature S surveys, I learned a great deal from reading Chs. 2–5. Ch. 6 covers material I was already familiar with, allowing me to assess the accuracy and completeness of S’s coverage. In what follows, I comment on these chapters individually.

Ch. 2 examines the linguistic-typological approach. The method of LT, though inductive, presupposes prior theoretical choices. Going back to Greenberg’s (1963) work in this area, typo-logical generalizations about word order have been expressed in terms of subject, object, verb, preposition, postposition, and so on. These categories and their application to particular utterances involve implicit theorizing. Moreover, typologists’ claims about the word order of a particular language are claims about basic or predominant orders—claims that require examination and analysis of a great deal of primary data. Given these complexities, it is striking how much progress has been made in LT over the half century since the publication of Greenberg’s seminal paper. The number of languages surveyed has vastly increased; the set of elements whose relative orderings have been tested and correlated has expanded; general formulations unifying various ordering correlations have been proposed and tested; family and areal tendencies have been explored; and some proposals have emerged for explaining why certain typological generalizations hold.

This progress in LT has not been without internal disagreements—for example, over whether high-level generalizations about constituent ordering should be stated in terms of heads...