Review of "Ergativity: Argument Structure and Grammatical Relations. Christopher D. Manning. Stanford, California: CSLI Publications, 1996. Pp. xiii + 222."

This book is a revised version of Christopher Manning's 1994 Stanford University dissertation. As a syntactician who is not an expert on ergativity I enjoyed the book, finding it clearly written and carefully argued. While a large amount of data is included to back up the empirical claims made, it is carefully presented not to overwhelm the reader. Also, while one of the central thrusts is a theoretical one, the claims are made in as theory-neutral a way as possible, making this book useful to linguists of various stripes. Part 1, Cutting the Ergative Pie, outlines the core claims of the book, backed up by various theoretical and empirical cross-linguistic considerations; Part 2, Inuit (West Greenlandic), is an in-depth look at Inuit, a well-studied ergative language; there Manning compares his account of ergativity to others from the literature. The basic claim is that a syntactic representation is organized into two levels of information: grammatical relations structure (gr-structure) and argument structure (a-structure) and that one locus of variation among languages is in the linking between the two levels of representation. Gr-structure corresponds roughly to a surface level of grammatical relations, like the final grammatical relations of Relational Grammar (RG), the level of f-structure in Lexical-Functional Grammar (LFG), or the level of S-structure in Government-Binding (GB) theory. His notion of a-structure is a syntactic level of representation, as in some work in LFG (Bresnan & Zaenan (1990)), and not strictly semantic, as in other LFG work (Alsina (1993:85)). He suggests that it corresponds roughly to the VP-internal relationships in recent versions of GB and Minimalist work. He notes (p. 35) that, as he sees it, gr-structure and a-structure are grammaticized representations of two different sorts of information. Gr-structure is a grammaticization of discourse roles, while a-structure is a grammaticization of notions of semantic/thematic prominence. The book focuses on (1) arguing for these two syntactic levels of representation, and (2) exploring the mapping or "linking" between these two levels. Many languages have a case pattern that groups the single argument of an intransitive verb (called 'S') with the less patient-like argument of a transitive verb (called 'A'), marking them with one case, nominative; the more patient-like argument (called 'O') is marked differently, with accusative case. This is called an "accusative" pattern. However, a number of languages group together the single argument of an intransitive verb (S) with the more patient-like argument of a …