Reviewed by: Using corpora to explore linguistic variation ed. by Randi Reppen, Susan M. Fitzmaurice, and Douglas Biber Merja Kytö Using corpora to explore linguistic variation. Ed. by Randi Reppen, Susan M. Fitzmaurice, and Douglas Biber. (Studies in corpus linguistics 9.) Amsterdam: John Benjamins, 2002. Pp. xii, 274. ISBN 1588112837. $126 (Hb). Among the increasing number of books dedicated to the study of linguistic variation and aspects of language use, this volume stands out. It offers a well-balanced selection of corpus-based studies that cover a broad range of linguistic features, registers, and dialects (or varieties) of English, but also includes one study in which the target language is Russian. In the editors’ words, ‘adequate descriptions of variation and use must be based on empirical analyses of natural texts’ and ‘on multiple texts collected from many speakers’. Moreover, these descriptions ‘must simultaneously consider the influence of a range of contextual factors on linguistic variability’ (vii). The studies included in the volume are carried out with these methodological aims in mind. [End Page 438] The approach has been gaining momentum over the past few decades, along with the recent advances made in the compilation and exploitation of computerized corpora. The book comprises an illuminating introduction and thirteen chapters divided into three sections. In Part 1 (Chs. 1–9), the main research goal is to explore variation in the use of different linguistic features across various dialects and registers. Two of the nine studies are devoted to hedging, one of which focuses on uses of the modal would. There are two further studies on modals, and five studies devoted to various phraseological and lexical phenomena, ranging from formulaic language to lexical bundles, pseudo-titles, and issues in pattern grammar. The two chapters in Part 2 are devoted to dialect or register variation. Ch. 10 describes syntactic features in written Indian English, and Ch. 11 investigates linguistic variation in academic lectures. The two chapters in Part 3 focus on historical variation. Ch. 12 deals with patterns of negation in eighteenth-century English, and Ch. 13 looks at register variation in nineteenth-century English. Deanna Poos and Rita Simpson investigate the functions of hedging in academic English using a pilot version of the Michigan Corpus of Spoken English (MICASE). Focusing on kind of and sort of, they show convincingly that the hedging frequencies do not depend so much on the gender variable as on academic discipline, type of interaction, and speaker needs within a given academic institution. Hedging is shown to be more characteristic of discourse in the humanities than in the physical sciences; of particular interest in this context are the possible reasons the authors suggest might account for this empirical finding. The other important result presented concerns the multifunctionality of kind of and sort of in spoken interaction. In addition to appearing as indicators of tentativeness, these items also ‘serve a variety of often overlapping sociopragmatic purposes in spoken interaction’ (21). Thus the study brings together the gender and multifunctionality issues, emphasizing the importance of acknowleding the multiplex nature of speaker identity. Fiona Farr and Anne O’keeffe compare the use of would as a hedging device in spoken Irish English (sampled from phone-in conversations on radio and post-observation teacher training interaction) with the uses attested in corpora representative of spoken British and American English. The corpus evidence shows that would is used more in Irish English than in the other varieties, and more specifically, for broader pragmatic functions than merely to mitigate or tone down the force of an utterance. To account for their findings, the authors arrive at a hierarchical three-tiered model where register, setting, and sociocultural norms interact to influence language choice. The model clearly has potential for further research. By contrast, Graeme Kennedy limits his study of modals to one variety, British English. However, as the object of this study is the 100-million-word British National Corpus (BNC), we are presented with a gigantic dataset (1,457,721 tokens), extracted on the basis of the tagging accompanying the text. Kennedy’s study is a fascinating piece of work, one of the few in which...