The prosody of Mandarin Chinese By Xiao-Nan Susan Shen (review)

662 LANGUAGE, VOLUME 67, NUMBER 3 (1991) struction, and a relationship between cases of polysemy and dead metaphors is discussed. The book's 'Summary and conclusion' (Ch. 8, 175-81) provides a detailed summary of each chapter, which combines with a 'Bibliography' and detailed indexes of subjects and personal names (182-205) to make this book remarkably usable. In search of a theoretically sound descriptive model, P pays little attention to the diagnostic aspect of lexical description, the most difficult problem of semantics. While the corpus-based statistical approach lends some credence to judgments on feature analysis, often P must fall back on intuitive judgments which are unassisted by the functional model. P's functional model is, however, interesting in its combination of theoretical and practical approaches. Overall, the book can be useful as an introduction to lexical semantics because of its coverage oflinguistic semantic theory, and to researchers in lexical semantics for its insights and refreshing eclectic approach. [Donalee H. Attardo, Purdue University.] Eine selbstlernende generativ-phonologische Grammatik. By Jürgen Rolshoven. (Linguistische Arbeiten , 218.) Tübingen: Max Niemeyer , 1989. Pp. viii, 295. Paper $66.00. Rolshoven attempts to construct an adaptive ('self-learning') and flexible algorithm as a model for language (or at least its phonology) and language change. In addition to a Regelwerk (= rule system + lexicon) and langue, which parallel the langue/parole and competence/performance dichotomies of structuralism and generative-transformational grammar , R's system comprises a third component, a so-called Steuerwerk ('steering mechanism '). This component checks the parole— which is the product of the Regelwerk and which, in this recursive system, also constitutes the incoming data for the Steuerwerk—and measures its degree of communicative success. It then creates/modifies hypotheses for the Regelwerk , which is equivalent to modifying the rules as necessary in order to describe the grammar more accurately and thus generate the language . Because the Steuerwerk is so crucial to R's model, nearly half the study (Ch. 3) is devoted to the meticulous description and construction of this component. This is achieved through a progressive, step-by-step refinement of problem -solving component algorithms, from the more complex to the simplest. The model 'learns' first through trial and error, then creates optimal archimorphemes and hence recognizes allomorphs; the resultant decline in redundancy leads in turn to determination of suppletive forms and revision of any relevant phonological rules. The fourth chapter then tests the entire Steuerwerk-algorithm on two sets of linguistic data, the first garnered from one dialect of Rhaeto-Romansch and the second from three minor Rumanian dialects. R then records all lexical items and phonological rules that the algorithm obtains from the necessarily limited corpus. A very brief fifth chapter evaluates the efficacy of the model. Although the model can only do so much with a small amount of actual data, Rolshoven stresses its strengths, namely its dynamic (adapt)ability to accommodate new data and hence revise the grammar. Thus it serves as both a synchronically and a diachronically descriptive mechanism of a given language. As an algorithm, it is easily transferable to computer language format and would be of interest primarily to those dealing with artificial intelligence . [Robert G. Hoeing, State University of New York at Buffalo.] The prosody of Mandarin Chinese. By Xiao-Nan Susan Shen. (University of California Publications in Linguistics, 118.) Berkeley: University of California Press, 1990. Pp. vii, 95. Paper $15.00. This monograph, a revision of the chapter 'Tone, stress, and intonation' in Shen's University of California, Berkeley, dissertation (Contrastive study of Mandarin Chinese and French interrogative intonologies: A contribution to the prosody of Chinese), must be regarded as a valuable addition to the understanding of the interrelationships among tone, stress, and intonation in Mandarin Chinese. Up to this point, there has been very little tightly controlled experimentation and instrumental analysis of the prosody of Mandarin Chinese, especially with regard to the question ofhow intonation interacts with tone. Thus, this work represents a significant attempt to fill the void. BOOK NOTICES 663 S begins in Ch. 1, 'Introduction', by stating that previous studies have suffered from 'imperfect investigation methods', e.g. 'impressionistic observations' ; 'confusion of multiple uses of pitch'—in other...