1The Longman Grammar of Spoken and Written English (LGSWE) is a major study of contemporary English, incorporating many new and revealing findings, and should prove to be an essential reference work for many years to come. The authors define their approach unambiguously as descriptive, rather than theoretical, and present the work as a companion to A Comprehensive Grammar of the English Language (Quirk et al. 1985), adopting similar terminology and concepts to those of the earlier work. In several important respects, however, LGSWE is innovative, and clearly reflects the increasing attention that has been paid to register variation of language in use, ‘the lexical end of grammar’ and phraseology, in linguistic studies of English over the past two decades.

2The grammatical description is corpus-based, using the 40 million-word Longman Spoken and Written English corpus (LSWE). This corpus comprises 37,244 texts, all (apart from a few texts in fiction) collected after 1980, and is broken down into four core registers (~5 million words each) chosen because they are the most important, productive, and different from each other – conversation (BrE only), fiction (AmE and BrE), news (BrE), and academic prose (AmE and BrE) – plus sub-corpora of AmE conversation and news (~7.7 million words) and two supplementary registers of non-conversational speech (BrE) and general non-fiction prose (AmE and BrE), which together account for over 12 million words. The analyses of register variation are based principally on the 4 core registers, while the AmE sub-corpus is used for dialect comparisons, and the two supplementary registers mainly for overall findings from the full corpus. All of the findings are normalised to frequency of occurrence per 1 million words. The automated computer analysis of the tagged corpus was complemented by manual checking of the KWIC concordance files, an interactive grammar-checking tool, and a qualitative, functional interpretation of the quantitative results.

3The book is divided into 5 sections totalling 14 chapters. Section A, the introduction (Chap. 1), presents the aims and principles on which LGSWE is based: the implications and methods of a computer-based approach to English grammar, the interplay between structure and use in grammatical description, the importance of register variation and a functional interpretation of language in use. The authors discuss their choices frankly and take pains to make their approach, key concepts, and decisions concerning the display and normalisation of the results clear.

4The following 4 sections, B to E, entitled respectively ‘Basic grammar: description and distribution’ (Ch. 2 & 3), ‘Key word classes and their phrases’ (Ch. 4–7), ‘More complex structures’ (Ch. 8–10), and ‘Grammar in a wider perspective’ (Ch. 11–14) present the grammatical analysis in detail. As the titles of these sections indicate, the development is bottom-up, working progressively upward from small-scale units to larger, more complex ones. Each grammatical feature is presented in a similar tripartite format: an initial description of form and structure of the feature is followed first by the corpus findings and frequency counts, then by a discussion of the findings. The Corpus Findings sub-section presents the distributional patterns of the feature and frequently includes barcharts and tables which provide very legible graphic displays of the quantitative results. For each feature, the endnotes specify the method of analysis used and whether the results are based on the full corpus, on the 4 core registers, or on other samples and sub–sets only. These empirical, corpus-based results amply illustrate one of the main conclusions of LGSWE, namely that language in use exhibits recurrent patterns which vary systematically with register, and, to a lesser extent, with dialect. In the Discussion of Findings sub-section, these quantitative results are interpreted in functional terms: the work or task that a feature performs in discourse (textual, ideational, personal...), the processing constraints that it reflects (constraints that are markedly different between conversation and academic prose, to take the two extreme cases), and lastly, the situational and social distinctions that the feature conventionally conveys.

5The book also contains very useful reference sections: a list of abbreviations and symbols used, a contents summary which gives an overview of the book, followed by a much more detailed table of contents, and explanations of the symbols and notational conventions, all at the beginning of the book; and at the end, an appendix of contractions, endnotes, a bibliography of corpus-based and corpus-informed studies and grammars of English, plus two indexes (lexical and conceptual). These well-designed sections mean that finding one’s way around LGSWE is remarkably straightforward, despite the imposing size of the work.

6The most exciting section of LGSWE, for the present reviewer at least, is Section E, which investigates the interface between grammar and the wider issues of discourse, lexis and pragmatics, by looking at information structure, the grammatical marking of stance, ‘lexical bundles’ (the phraseological component), and in particular, the final chapter on the grammar of conversation. The authors’ approach to the grammar of conversation is less radical than that of, for example, David Brazil (1995), who argues for a fundamental difference between the grammar of speech and that of writing, proposing a linear model which develops incrementally over time, rather than the traditional constituency model in terms of hierarchies of units. A rather similar approach is developed by Carter & McCarthy (1995), who propose a dynamic structure for conversation, with utterances composed of a core clause frequently preceded by a pre-clause dislocated topic and followed by a post-clause ‘tail’. LGSWE is somewhat more conservative, opting instead for a ‘same grammar’ point of view between speech and writing, but the work stresses the highly different use made of grammatical features in the spoken medium, and incorporates many of the insights of the above-mentioned studies into the looser, add-on composition of speech compared to the architectural structure of the written sentence, and proposes the ‘C-unit’ (composed of clausal and non-clausal material) as the governing principle of conversation. This refreshing new look, combined with the strong corpus basis of LGSWE which provides reliable evidence for their findings, should profoundly influence future studies of English.

7LGSWE will no doubt prove to be more immediately useful to teachers and materials writers, both in general English and in ESP/EAP (thanks to the inclusion of the academic prose register), and to corpus linguistics and natural language processing studies (thanks to its probabilistic approach), than it will be to foreign learners of the language, as the descriptive method leaves many aspects of structural analysis un- or underspecified and implicit. However, the strong phraseological orientation of the work and the constant attempt to bridge the gap between lexis and syntax, by highlighting the systematic patterns of association or co-occurrence between syntactic structures and lexical items, as well as the functional interpretations of register variation, will enrich and broaden learners’ perception both of grammar and of the English language itself.