This is a comprehensive description of a language spoken some 450 km offshore from the mainland of Papua New Guinea. The language is remarkable for its phonological, morphological and syntactic complexity. As the sole surviving member of its language family, and with little historical contact with surrounding languages, the language provides evidence of the kind of languages spoken in this part of the world before the Austronesian expansion. The grammar provides detailed information on the phoneme inventory, morphology, syntax and select semantic fields. Remarkable features include a 90 phoneme inventory including unique sounds, a morphology with thousands of non-compositional portmanteau elements, complex rules for negation, and extensive ergative syntax. Unusual patterns are also found in the organization of semantic fields, for example in partonymies of the body, taxonomies of the natural world, verbal semantics and kinship terms. The combination of linguistic 'rara' suggest that linguistic evolution under low contact can yield baroque and unusual patterns. The volume should be of special interest to linguists, typologists, sociolinguists, anthropologists and researchers in Oceania and Melanesia. Endorsement: This long-awaited grammar is a major contribution to Papuan and general linguistics, providing as it does by far the most comprehensive and accurate grammatical description of a language that has already assumed a position as one of the world's most complicated. Hitherto, the most extensive grammatical description of the language has been the survey-like Henderson (1995), and while Levinson explicitly acknowledges his debt to this earlier grammar and to unpublished work by Henderson, his own detailed grammar clearly takes the level of description and analysis of the language to a completely new level. In particular, Levinson's grammar makes clear precisely to what extent and in what ways the language's morphology is complex beyond even what most studies on morphologically complex languages envisage. In addition, it provides a much more detailed account of the language's syntax, based on a judicious combination of corpus attestation and careful elicitation (incl. using the kits developed by Levinson's group at the MPI for Psycholinguistics). The grammar thus not only fills a major lacuna in our knowledge of the non-Austronesian languages of the New Guinea area, but also provides grist for future studies on the implications of the language's complexities.Bernard Comrie, University of California, Santa Barbara