This book presents a theory of grammatical relations among sentential constituents which is a development of Chomsky's Government-Binding Theory. The cross-linguistic predictive power of the theory is unusually strong and it is supported in the examination of a wide range of languages. Within the syntax of a language, grammatical relations determine such things as word order, case marking, verb agreement, and the possibilities of anaphora (co- and disjoint reference) among nominals. Other approaches to grammatical relations have considered them to name classes of constituents that share clusters of properties, including most prominently structural positions or case marking. Still others have claimed that grammatical relations are primitives in syntactic theory, but are related essentially to semantic roles. Rejecting these approaches, this monograph develops a theory which includes at its core a "projection principle": The syntax of a language is assumed to be a (direct) "Projection" of the compositional sematics, and the mechanisms of projection are explicitly spelled out. Chapters cover the two asymmetries and two lexical features on which the theory is built; semantic and syntactic data from a wide variety of languages that support the universal applicability and explanatory power of these asymmetries and features; features of passive, antipassive, dative-shift, anticausative, causative, and applied verb constructions in the worlds' languages explained by the theory; confirmations of the theory's predictions in languages for which alternative approaches to grammatical relations fail to provide successful analyses; and, comparison of the book's conception of grammatical relations to those in the GB framework, Montague Grammar, Relational Grammar, and Lexical-Functional Grammar.