It is testimony to the stature of Corpus Linguistics (henceforth ‘CL’) as a thriving subdiscipline of linguistics that Cambridge University Press should publish a ‘critical overview’ volume of such vision and authority. The formidable duo of Biber and Reppen have assembled an impressive cast of forty-one authors, the fruits of whose labours is a 639-page book that is a must-read for corpus linguists and an essential resource for those teaching courses in CL. Each contributor was asked by the editors to both provide an up-to-date survey of previous literature and describe an empirical study, and even though the editors are forced to concede that this ‘has not always been possible’ (p. 6), the vast majority have succeeded in doing so. This is an attractive feature of the book, with contributors approaching it in a variety of ways: some making use of their own previous work, some offering entirely new studies and some focusing on applications.
In their Introduction, Biber and Reppen take the opportunity to foreshadow some of the issues and debates that are raised by a number of their contributors, such as the extent to which CL is, or is not, a predominantly atheoretical, methodologically driven, field. They also discuss – and correct – the fallacious assumption that CL originated in the 1980s and 1990s as a reaction against intuition-based linguistics. Long before the advent of electronic corpora, they correctly observe, there existed ‘corpora’ (albeit specialized collections rather than representative and readily searchable corpora) such as the OED citation database and C. C. Fries’ 1940s letters collection.
The book consists of twenty-eight chapters organized into four parts, the second and third of which account for approximately two-thirds of the book, excluding the references and index: (Part I) ‘Methodological considerations’ (three chapters); (Part II) ‘Corpus analysis of linguistic characteristics’ (eleven chapters); (Part III) ‘Corpus analysis of varieties’ (nine chapters); and (Part IV) ‘Other applications of corpus analysis’ (five chapters).
For Part I the editors could hardly have selected anyone more eminently qualified to comment on methodology than the three authors, Mark Davies, Paul Rayson and Stefan Th. Gries. In his chapter (‘Corpora: an introduction’) Mark Davies – who has earned an enviable reputation for his capacity to assemble very large corpora such as COHA (Corpus of Historical American English) and COCA (Corpus of Contemporary American English), and most recently the massive 1.9 billion-word GloWbE (Global Web-based English) corpus – comments on the debate over the relative merits and usefulness of, on the one hand, smaller multigeneric ‘representative’ corpora such as the million-word corpora belonging to the ICE (International Corpus of English) and the Brown families, and on the other hand the typically more generically limited large corpora of the type compiled by Davies himself.
Paul Rayson (‘Computational tools and methods for corpus compilation and analysis’) provides a balanced account of the growth of CL methodologies from the 1970s/1980s when – he observes ruefully – making a concordance was an extremely time-consuming task, to the present-day when – he wrily notes – the end-user ‘would be mildly annoyed if a concordance from a 1-billion-word corpus tool took more than five seconds to be displayed’ (p. 32). A survey by Rayson of papers published in 2012 in CL journals finds a dominant concern with retrieval methods, less so with compilation and even less with annotation. He criticizes CL researchers for their tendency to rely on pre-built software tools rather than tools created by themselves, and current CL for being too tool-driven (‘counting only what is easy to count’). Nevertheless, Rayson envisages a spread of CL techniques in the future to other disciplines such as psychology and social sciences.
Stefan Th. Gries (‘Quantitative designs and statistical techniques’), a major contributor to the topic of his chapter and to whom we are indebted for leading the way, raises the statistical-knowledge-requirement-bar for corpus linguists higher than many will probably be able to clear. Gries argues forcefully that CL could benefit greatly if more attention were paid to under-utilized statistical methods and if researchers were to renounce monofactorial analyses in favour of more explanatorily powerful multifactorial analyses.
Winnie Cheng's chapter (‘Discourse intonation: a corpus-driven study of prominence on pronouns’), the first in Part II, suggests that in Hong Kong English the lower the frequency of a pronoun the greater the unlikelihood that it will carry nuclear prominence. The applicability of the suggestion to the pronoun category is undermined somewhat by the limitation of her study solely to personal and possessive pronouns. Cheng does not attempt to comment upon the much-explored interrelationship between prosodic prominence and information structure.
Jonathan Culpeper and Jane Demmen (‘Keywords’) confirm the value of small corpora, noting that in research on keywords a one-million-word corpus such as FLOB is as good as a one-hundred-million-word corpus like the BNC. The point is well made that determining keyword frequencies is not an end in itself, but rather a precursor to the qualitative analysis of social, cultural and other implications.
Richard Xiao (‘Collocation’) argues for the importance of studying collocation, in research on lexicology, fluency and other topics. Espousing the advantages of significance-driven analyses of collocation (rather than concordance-driven, as in the Firth/Halliday tradition), he devotes much space to discussing statistical methods such as log-likelihood.
Bethany Gray and Douglas Biber (‘Phraseology’) show how, with the advent of CL, researchers have been able to empirically confirm the extent to which everyday language is composed of prefabricated expressions (casting serious doubt on the Chomskyan notion of creativity), using both corpus-based studies of pre-selected expressions and corpus-driven studies that seek to identify the full set of multi-word sequences in a corpus. Gray and Biber address the role of register and discourse function, along with the core issues that preoccupy phraseology researchers, and provide a case study of discontinuous lexical frames.
Geoffrey Leech's chapter (‘Descriptive grammar’) is the first of four chapters dealing with grammar (his unifying comments on the other three seemingly usurping the role of the volume editors in their introductory chapter). No contributor to this volume can speak with greater authority than Leech, and it is indeed fitting that the volume should be dedicated to his memory. The chapter is a delight to read, a fascinating first-hand historical account from one who has lived through and pioneered developments in CL from the beginning of the electronic era to the recent past. Leech stresses the important role of the ‘total accountability’ principle in descriptive – as opposed to theoretical – grammar, in the shaping of comprehensive corpus-informed grammars such as Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985) and in opening up areas of grammar previously ignored by theoretical linguists (e.g. discourse markers). The principle is traceable back to the citation corpora used by neo-traditional grammarians such as Jespersen (whose database, Leech observes, might more appropriately be described as a ‘shoebox corpus’ in recognition of Jespersen's famous method of filing his citation cards). Leech concedes that despite the boost to corpus-based research on grammar provided by the compilation of readily available corpora and the advent of POS-tagging, progress has been hampered by continuing difficulties in the development of syntactic parsing. Insightful observations in this chapter – too numerous to discuss here – include the importance of statistical modelling in CL in adding value to descriptive grammar by providing an empirical basis for frequency statements, and the role of CL in shedding light on such fundamental issues in linguistics as ‘gradience’ and ‘prototypicality’.
Daniela Kolbe-Hanna and Benedikt Szmrecsanyi (‘Grammatical variation’) conceive of this area very broadly, encompassing five sub-areas (variationist sociolinguistics, diachronic linguistics, register/genre/text-type, dialectology and knowledge processing), and for each one they describe in some detail a study they consider to be representative. They trace the progression from traditional sociolinguistic research based on ‘private’ corpora to replicable studies based on publicly available corpora. The most significant finding reported is that grammatical variation is, like phonological variation, sociolinguistically conditioned. Most of the chapter is devoted to a case study modelling the constraints on complementizer-that deletion that incorporates both language-internal and language-external perspectives. Curiously, the authors do not venture explanations for their intriguing geographical finding that the prevalence of that-omission in central England contrasts with that-retention in ‘peripheral’ areas (Wales, Cornwall, Scotland and Outer Hebrides).
Martin Hilpert and Christian Mair (‘Grammatical change’) enter a timely warning about the influence of corpus selection on findings obtained, alluding to the important debate in CL – often referred to as ‘the Millar/Leech debate’ – prompted by Neil Millar's (Reference Millar2009) challenge to findings by Geoffrey Leech and his colleagues (Leech et al. Reference Leech, Hundt, Mair and Smith2009; Leech & Smith Reference Leech, Smith, Renouf and Kehoe2009) of a decline in modal verb frequencies in late twentieth-century English based on data from Brown family corpora. Millar, reporting a study of the modals based on the 100 million-word TIME Magazine Corpus (1923-2006), found that some modals – most notably may – were in fact rising in frequency. Responding, Leech (Reference Leech2011) acknowledged that the Brown family corpora suffer from limitations of size and widely spaced sampling points, but argued that a large monogeneric (in fact single-publication) corpus of the type used by Millar does not provide representative results. His claims enjoy support from an analysis of COHA, a 400 million-word multi-genre corpus whose design is quite similar to that of the Brown family, which finds an overall fall – of 20.9 per cent for the twentieth century as a whole – for the modals as a category.
Susan Hunston (‘Lexical grammar’), following Biber (Reference Biber2012), distinguishes between a ‘grammar-to-lexis’ approach (which ‘views lexis through the lens of grammar’ (p. 210) as epitomized in Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999), and in which ‘construction’, and more recently ‘collostructure’ – see Stefanowitsch & Gries (Reference Stefanowitsch and Gries2003) – are central concepts), and on the other hand ‘lexis-to-grammar’ (as exemplified by Pattern Grammar, which uses the COBUILD corpus to identify grammatically relevant word usages and phraseologies).
Alan Partington and Anna Marchi (‘Using corpora in discourse analysis’) discuss the use of CL methodologies (word frequencies, recurrent patterns, KWIC concordances, etc.) to advance our understanding of, inter alia, the ideological biases of sociopolitical discourse and the essential properties of different discourse-types.
Brian Clancy and Anne O'Keeffe (‘Pragmatics’) demonstrate clearly that research on topics such as pragmatic markers, speech acts and conversation analysis has both benefited from and been greatly facilitated by the compilation of spoken corpora such as the Bergen Corpus of London Teenager Language (COLT) and the Cambridge and Nottingham Corpus of Discourse in English (CANCODE).
Irma Taavitsainen (‘Historical pragmatics’) claims that CL is now firmly established as the main methodology in (the relatively new field of) historical pragmatics. Developments have been influenced by the split between the predominantly European approach to pragmatics as communication, and the predominantly Anglo-American view that pragmatics is a ‘component’, or level, of language. With rapid developments in CL methodologies and the preparation of the second generation of post-Helsinki Corpus historical corpora (a useful list of which is supplied on pp. 267–8), increasingly sophisticated studies have appeared yielding novel insights into such topics as grammaticalization, speech acts, and language and ideology.
Part III begins with Shelley Staples’ chapter (‘Spoken discourse’), a comprehensive survey of advances in CL on the study of speech (including research on the characteristics that differentiate speech and writing, the characteristics of particular spoken registers and specific individual features associated with speech). Staples also explores new directions and challenges for spoken speech research, and provides a helpful list (and critique) of available spoken corpora, along with a case study of stance features in nurse–patient interaction.
Ken Hyland (‘Corpora and written academic English’) shows how CL studies offer researchers a means of avoiding the impressionism, partiality and prescriptiveness of traditional methods of text analysis. Hyland identifies key studies in the field associated with such topics as cross-disciplinary genres and disciplinary discourses, and presents a case study of the role of CL research in the study of identity via a comparison of the published works of two well-known linguists, Cameron and Swales.
Susan Conrad (‘Register variation’) attempts to bring some order to the morass of terms and definitions associated with the study of situation-based varieties. She proposes a taxonomy that is based on a distinction between two categories: linguistic features and text category. The former includes individual features (a striking example of which is Biber et al.’s Reference Biber, Johansson, Leech, Conrad and Finegan1999 Longman Grammar), and co-occurring features (dominated by MDA research). The latter includes single registers, and comparisons of registers and subregisters. Conrad offers a frank assessment of the accomplishments and challenges in register studies, along with a case study of registers in civil engineering.
Merja Kytö and Eric Smitterberg (‘Diachronic registers’) investigate central matters that arise in applying register to diachronic material, such as achieving comparability and representativeness. One perennial problem has been the dearth of spoken registers (the closest corpus compilers can get for earlier periods being speech-related genres such as drama and court records). Another problem is that some registers are more or less ephemeral; for example, the compilers of B-LOB (1931) found it difficult to find texts corresponding to the science fiction and sociology texts of LOB (1961). Previous single-feature and multi-feature studies are surveyed and a case study of thou versus (singular) you presented.
Michaela Mahlberg (‘Literary style and literary texts’) examines the current state of the art in corpus stylistics, which she situates within the context of CL and computational stylistics, illustrating her points with numerous pertinent examples, the most extended being a case study of characterization in Dickens. Mahlberg predicts, plausibly, that the increasing availability of digitized literary texts will promote interest in corpus approaches to research in this field.
Jack Grieve (‘Dialect variation’) observes the challenges produced by more recent corpus-based research to assumptions made in earlier questionnaire- and interview-based research. Grieve suggests, controversially, that there has been relatively little corpus-based research in dialect variation. While this may be true of studies of regional dialect variation based on the Helsinki Dialect Corpus, the Freiburg English Dialect Corpus and others, it seemingly overlooks the multitude of national dialect studies based on the Brown family and ICE corpora.
Marianne Hundt (‘World Englishes’) scrutinizes the widely used ICE corpora, their design and suitability for research. Worryingly, she identifies a number of failures to achieve comparability, including diachronic biases over a period of more than two decades and cross-cultural differences in the interpretation of text types. The chapter includes a case study of the present perfect which aims to broaden the scope of previous research – the results of which are often in conflict – by including varieties not hitherto subjected to comparative investigation, such as Ghanaian and Fiji English. The study is nevertheless limited by its restriction to a single genre, news, and its lack of a diachronic dimension. The latter is a general limitation of ICE-bound studies, the only relevant strategy adopted by some researchers being the assumption of speech versus writing differences as a proxy for ongoing change.
Anna Mauranen, Ray Carey and Elina Ranta (‘New answers to familiar questions: English as a lingua franca’) reveal that only recently have researchers moved beyond the dominantly qualitative study of English as a lingua franca to corpus-based examinations thereof, hampered in large part by a dearth of corpus resources. This research has identified a number of ‘known’ features of English as a lingua franca (including non-standard uses of articles and prepositions), while a case study of syntactic universals in spoken English as a lingua franca invokes such features as extended uses of the progressive, of would in hypothetical if-clauses, and of singular agreement in existential there-constructions.
Gaëtanelle Gilquin and Sylviane Granger (‘Learner language’) claim that ‘Learner Corpus Research’ emerged in the 1980s, encroaching on SLA territory, but differing from it in its preoccupation with performance rather than competence. While there are some similarities with English as a lingua franca, this research is interested more in the use of L2 for communication than in the learning process. Representative studies are critiqued, new directions predicted and challenges acknowledged. The chapter ends with a case study of two-word discourse markers in native and non-native speech.
The first chapter in Part IV, by Ron Martinez and Norbert Schmitt (‘Vocabulary’), supports the longstanding belief that the first words to be taught should be the most commonly used, the determination of which has benefited greatly from CL-derived frequency information. The authors trace the development of key corpus-informed vocabulary lists, from older ones such as the General Service List of 1953, to newer ones such as the New General Service List of 2013, and pinpoint how the lists have improved (in large part aided by the switch from word families to lemmas).
Magali Paquot (‘Lexicography and phraseology’) focuses on electronic dictionaries, with a detailed description of Sketch Engine, the corpus query system most frequently used by lexicographers today. She shows how CL has aided development of new access routes to phraseological units in learners’ dictionaries, and overviews studies of word combinations in English lexicography.
Thomas Cobb and Alex Boulton (‘Classroom applications of corpus analysis’) identify and discuss three classroom applications of CL methodologies: improved descriptions of language varieties and features, the new availability of corpora and corpus tools for teachers, and the same for learners. The chapter focuses on the third application – noting the recency of research on the potential benefits of corpus consultation in language learning – and provides an interesting ‘meta-analysis’ which seeks to reduce subjective bias through its quantitative presentation of corpus survey results.
Fanny Meunier and Randi Reppen (‘Corpus-informed versus non-corpus-informed pedagogical materials: grammar as the focus’) present a comparison of four corpus-informed versus four non-corpus-informed grammars which lacks depth until the comparison is limited to a single topic (the passive). The chapter contains several inaccuracies (on p. 512, the passive of ‘Major bank announces 200 job losses’ is incorrectly given as ‘200 job losses have been announced by a major bank’), and in the Conclusion on p. 513 the premise upon which it is argued that pedagogical grammars need to be corpus-informed is the unsupported claim that ‘Corpus information on registers, frequency, and lexical preferences is key to a good understanding and use of grammar’).
Silvia Bernardini (‘Translation’) argues that the emergence of corpus-based translation studies (from the early 1990s) was prompted by the perceived subjectivity of traditional approaches and the perceived desirability of moving away from a focus on individual texts and their translations to coherent sets of texts. Bernadini discusses the types of corpora that are both available and needed, and addresses questions specific to translation research, rather than neighbouring areas such as a contrastive linguistics. Five particularly influential studies are overviewed.
This fascinating volume is a must-read for corpus linguists. One might quibble that some emergent areas in CL are over-represented (such as pragmatics/discourse analysis, with three chapters), or that there are occasional instances of overlap across the chapters. But these do not detract from the richness and consistent quality (the only erratum we spotted being the misspelling ‘Inhalainen’ on p. 365!) of this book.
