In Autolexical Syntax, Jerrold M. Sadock argues for a radical departure from the derivational model of grammar that has prevailed in linguistics for thirty years. He offers an alternative theory in which the various components of grammar—in particular syntax, semantics, and morphology—are viewed as fully autonomous descriptive devices for various parallel dimensions of linguistic representation. The lexicon in this theory forges the connection between autonomous representations in that a typical lexeme plays a role in all three of the major components of the grammar.

Sadock's principal innovation is the postulation of a uniform set of interface conditions that require the several orthogonal representations of a single natural language expression to match up in certain ways. Through a detailed application of his theory to the twin morphosyntactic problems of cliticization and incorporation, Sadock shows that very straightforward accounts are made possible by the nonderivational model. He demonstrates the empirical success of these accounts by examining more than two dozen morphosyntactic problems in almost as many languages.

Autolexical Syntax will be of interest to those in the fields of theoretical grammar, particularly concerned with the problems of morphology and syntax, as well as philosophers of language, logicians, lexicographers, psychologists of language, and computer scientists.

Actions are described by verbs whose subjects, objects, and other complements refer to various participants in those actions. What linguistic principles determine which participants are referred to by each component of a verb?

Many previous approaches to this problem have employed a set of thematic roles, such as agent and patient, to classify varieties of participants. The alternative developed here fits within the framework of Head-Driven Phrase Structure Grammar while utilizing typed feature structures, certain basic features of verb meaning, a hierarchical classification of verb meanings, and constraints from more general to more specific word classes. Relying on no special mechanisms or components of grammar, this book is unique in its ability to account for the observed range of verb types in human languages with a simple yet widely applicable set of principles.

Mixed category constructions like the English verbal gerund involve words that seem to be central members of more that one part of speech. This poses a problem for the standard view of syntactic categories.

This book presents a novel analysis of this and similar mixed category constructions in languages including Quechua, Tibetan, Arabic, Fijian, Dagaare, and Jacaltec. Under this analysis, Robert P. Malouf shows that verbal gerunds share the selectional properties of verbs and the distributional properties of nouns. He further shows that since different dimensions of grammatical information can vary independently, the behavior of mixed categories creates no paradox. These dimensions are in principle independent. However, certain types of mixed categories are quite common in the world's languages, while others are rare or nonexistent. The book discusses how cross-linguistic variation can be accounted for by a lexical categorial prototype. By stating these prototypes as default constraints in a hierarchy of lexical information, Malouf argues that one can bring insights from cognitive and functional approaches to linguistics into a formal analysis, thus building on the strengths of both approaches.