Literacy in school and society: Multi-disciplinary perspectives Edited by Elisabetta Zuanelli Sonino (review)

BOOK NOTICES 421 In this provocative work, S sets out to construct the framework for a new approach to language description, one that would serve as an alternative to a traditional structuralist account of linguistic behavior—or, more precisely, to a rule-based account. For those among us reared in the structuralist tradition, analogy in language often becomes a nebulous cover term used to account for diverse types of linguistic behavior outside the realm of a rule-oriented description. In AMoL S argues that the principle of analogy is operative in all types of linguistic behavior, and he attempts to provide a solid mathematical basis that can accurately predict actual linguistic behavior, thus eliminating the need for the competence /performance distinction. The book is divided into six chapters and three appendices. In Ch. 1 S identifies three basic types of linguistic behavior—a distinction which appears throughout the book, but which would ultimately be eliminated in the analogical approach: categorial, exceptional/regular, and idiosyncratic. He compares and contrasts ways in which the structuralist and analogical approaches handle these three types. Ch. 2 (22-50) presents an overview of the analogical approach and defines its major theoretical constructs. This is clearly and directly accomplished with artificial data of the exceptional /regular behavior type. The analogical set for a given context is constructed by first examining the occurrences in the data set, determining the network of pointers, and calculating the uncertainty of the data set; this is based upon the number of disagreements between classes of data occurrences. The next step is to determine which data classes are homogeneous, i.e. do not increase the number of disagreements. The sum of these homogeneous data classes, supracontexts, comprise the analogical set, which is subsequently used to calculate the analogical EFFECT-the probability that the predicted behavior will occur in the given context. This theoretical introduction to the analogical approach is readily accessible, if somewhat lacking in depth. The theoretical foundations and ramifications are not fully addressed in this general overview. The remainder of Ch. 2 is devoted to demonstrating how, again with artificial data, the analogical approach would eliminate the tripartite distinction in linguistic behavior mentioned above. Chs. 3-5 deal primarily with the application of the analogical approach to natural-language phenomena of diverse types. Ch. 3 contains examples ofthe analogical approach applied to the three basic types of linguistic behavior in English : the indefinite article, the spelling of IhI. and voice onset time. There are, however, serious gaps in the leap from artificial to actual language data. One example of such a gap is the discussion of the variables that will determine the occurrences in adata set (51-4). S discusses the computational limit on the number of variables and introduces the general rule of proximity in the selection of variables— namely, those variables which are closest to the variable whose outcome is to be predicted. I would like to have seen a more rigorous treatment of this particular aspect of the analogical approach, since it is precisely these variables that are assumed to precipitate the outcome of a given context. In general, more attention is paid to computational exigencies than to the possible cognitive processes involved in analogical modeling. With the increasing recognition of the importance of cognitive processes in a viable theory of language, this issue should be addressed. In sharp contrast to the disappointing and incomplete examples in Ch. 3, Ch. 5 presents an excellent and comprehensive case study. 'Analogical predictions of the past tense in Finnish' (101-36). This case study convincingly demonstrates the potential for the analogical approach to account for language behavior— historical drift in this instance. AMoL suffers under the weight of the task S has undertaken. It does not adequately answer all the theoretical issues that it raises, for these issues are numerous . In his attempt to address all of them, S treats some cursorily: however the number of issues not comprehensively treated should not be viewed as a sign of neglect, but rather as a function of the radical, innovative nature of this work. The analogical approach definitely merits serious consideration, and I hope that further scholarship to clarify and develop it...