At a recent conference, a delegate commented quietly in my ear, ‘The problem with those who work in corpus stylistics is that they don't really analyse the data . . . they let the computer do that for them.’ This was intended as a criticism and yet it overlooks the fact that while computer software can analyse the data in a fraction of the time it would take to manually identify items in a text, it is still down to the individual to make the interpretations and offer original insights. A successful corpus analysis therefore, whether for literary or linguistic purposes, requires the expert insight of the researcher into the various lexical patterns, collocations or keyword lists beyond what the computer is able to produce. In other words, a quantitative approach to data analysis is not enough on its own if the general properties of the text are to be identified. As McIntyre (Reference McIntyre2013: 134) asserts, ‘for the approach to work to its best advantage, it needs to be used in conjunction with qualitative analysis’.

In her latest work, Corpus stylistics and Dickens's fiction, Michaela Mahlberg applies corpus stylistic methods to the works of Charles Dickens in an advanced and highly detailed discussion of textual analysis. She focuses mainly on his novels, to examine and reveal the techniques of characterisation, body language and speech presentation used in his fiction to create such memorable characters. But before reviewing her book, it might be useful to say a little about corpus linguistics as an approach to language study to understand how Mahlberg's research in corpus stylistics contributes to this field of study.

Corpus linguistics is generally associated with the use of computer technology. However, this was not always the case and literary works, specifically, have been subject to close scrutiny through a concordance, best defined as ‘a collection of the occurrences of a word-form, each in its own textual environment’ (Sinclair Reference Sinclair1991: 32) as well as key lists of particular parts of speech or particular semantic domains, for example. As early as Reference Becket1787, Andrew Becket produced a Concordance to Shakespeare by laboriously counting words manually to map their location in book, verse and line. Since then, the development of computer software has enabled huge amounts of corpora to be analysed and has led to the growth of corpus linguistics beyond the work undertaken by literary scholars or lexicographers. Corpus linguistics has broadened as an approach to language study and analysis across numerous disciplines including English language teaching and testing, teacher education, stylistics (both literary and non-literary language), forensic linguistics, political discourse and health communication. Recent work using corpus stylistic approaches that illustrate how quantitative methods to texts can highlight their distinctiveness include Bednarek's (Reference Bednarek2012) The language of fictional television and Montoro's (Reference Montoro2012) Chick lit, the stylistics of cappuccino fiction. In her research, Mahlberg continues to develop this area of study by merging her knowledge of stylistics with corpus linguistic approaches to reveal facts about the language of Dickens in a thorough investigation of his fiction. These facts, which would otherwise remain hidden because of the sheer volume of the corpus, foreground the literary techniques Dickens used in the creation of his fictional worlds and characters.

Corpus stylistics and Dickens's fiction comprises nine chapters, each of which are sub-divided into smaller sections to discuss in detail aspects of the chapter's main theme. Chapter 1, ‘Corpus stylistics’, presents an overview of corpus linguistics and explains how literary stylistics and corpus linguistics differ because one is interested in linguistic form, which can be separated from the context (corpus linguistics), whereas the other (literary stylistics) is concerned with form, function, meaning and effects and so context plays an integral part in this understanding. Mahlberg points out that there is a shared ground between both approaches because ‘the specificity of a single text is always seen in relation to general patterns of a language and the range of possibilities that a language offers’ (pp. 7–8). In other words, the patterns of language constrain the possibilities of the choices made in the texts, a point which is thought-provoking in itself. The chapter goes on to discuss Leech's (Reference Leech and van Dijk1985) levels of ‘deviation’ that create the foregrounding, and the psychological effects these may bring about, thus opening the discussion on the notion of the reader's response to a text. The chapter goes on to look at units of levels of description, helpfully defining key terms with examples from other corpus work in literary stylistics.

Chapter 2, ‘Textual building blocks of fictional worlds’, turns to Dickens's characters and the techniques used by the author to make them memorable. Here, Mahlberg discusses the various strategies of characterisation including idiolects, habitual phrases and actions, which are described as external features. Interestingly, these elements of characterisation would be similar to how humans are perceived by others. An alternative way of understanding characterisation is through cognitive processing that draws on approaches in cognitive stylistics and a combination of top-down and bottom-up processing. This is where a reader's knowledge of the world is merged with linguistic items in the text to form impressions of characters in their mind. This knowledge would also require a ‘theory of mind’ to interpret fictional characters, who exist in a fictional world, a discussion that leads nicely onto Text World Theory (Werth Reference Werth1999; Gavins Reference Gavins2007) and an understanding that fictional characters exist in their own physical and social context. Here, Mahlberg argues convincingly that the ‘textual building blocks for fictional worlds’ (p. 26) are achieved through clusters, a repeated sequence of words that are usually strings of five words that have particular functions in creating characterisation through a text-driven approach such as repeated phrasing.

Having presented an overview of how corpus stylistics can and has contributed to the literary and linguistic analysis of fiction in the first two chapters, Mahlberg goes on to describe her methods in chapter 3, ‘Starting with the texts: corpora, clusters, and lexical bundles’, as she highlights the difference between corpus linguistics and corpus stylistics. She lists the twenty-three Dickens texts under scrutiny – all of Dickens's novels and seven stories and sketches – referred to as the DCorp (listed on p. 43). Project Gutenberg provides the eBooks for her corpora, the advantage of using this source being that they are freely available and come as plain text files, which allows for their selection and manipulation. WordSmith Tools is used for the retrieval of five-word clusters in a corpus consisting of 350,000 words. From section 3.3, ‘Approaches to clusters and functional interpretations’, to the end of the chapter, the discussion becomes rather more technical. Mahlberg presents a detailed description of approaches to analysing clusters referred to as n-grams, which are repeated word sequences of a defined length (where n stands for any number). This discussion may be of interest to readers who want to find out more about previous research in this area. One area, in particular, is research on lexical bundles and their ‘pervasive nature’, as an alternative/additional approach to looking at linguistic patterns. Malhberg refers to Biber's (Reference Biber2006) earlier study to provide some interesting facts: that different registers appear to have differing amounts of lexical bundles, so, academic prose has the lowest number of different lexical bundles; conversation has twice as many as academic prose; classroom teaching has twice as many as conversation; and textbooks fall between academic prose and conversation (p. 52). Mahlberg discovers that fiction has fewer different lexical bundles than the four registers in Biber's study and a discussion ensues to provide an understanding of why this might be the case.

Chapter 4, ‘Groups of clusters for the identification of local textual functions’, turns to the Dickens corpus and the formal criteria for their functional analysis in the chapters that follow. There is an explanation of what is counted as clusters and ‘how key clusters can serve to select an initial sample’ (p. 61), which can then be applied to individual texts across the corpus. Key clusters, like ‘key words’, are calculated by comparing their frequency across two corpora; a cluster is ‘key’ if it is ‘significantly more frequent in one corpus than in another’ (p. 63). For the other corpus used in this comparative study, Mahlberg examines a selection of twenty-nine novels from the nineteenth century that includes Eliot's Daniel Deronda and Thackeray's Vanity Fair, referred to as ‘19C’. The point of this exercise is to establish the distinctiveness of the types and distribution of clusters in Dickens's work in comparison to other texts within the same historical period. Mahlberg draws on her earlier research on Dickens to describe and present the same five functional groups of clusters for the task of analysing the clusters in her latest research: Labels, Speech clusters, Body part clusters, ‘As if’ clusters, Time and place clusters and Other clusters.

The following four chapters provide the analysis and interpretation that readers have been waiting for before concluding with a final chapter. Mahlberg presents her findings from the different functional groups, which in turn provide original insights into Dickens's fictional worlds and characters. Chapter 5, ‘Character speech’, examines character speech on the basis of ‘surface features’ of speech clusters, such as first or second pronouns. Four functions become apparent, three of which are categories that imitate real, spoken language and would support the claim that Dickens was a writer who could skilfully represent the characteristics of natural speech. The fourth category, ‘Politeness formulae’ (p. 76), perhaps provides the most original level of insight as work on pragmatics in corpus stylistics is relatively new and open to further study. Chapter 6, ‘Body language’, turns to the non-verbal to examine body language in body clusters that contain body part nouns and looks to Korte's (Reference Korte1997) framework for the analysis of body language in literature. By highlighting the textual patterns in this category, the presentation of body language such as her apron over her head and (with) his hand to his (forehead) can be seen to be a result of narratorial choices in creating an authentic fictional world. Chapter 7, ‘As if and the narrator comment’, extends the discussion on body language but focuses on collocations of as if rather than clusters, taking a lexico-grammatical perspective. As if clusters join two clauses, and Mahlberg pays special attention to the left collocates, that is, the clause appearing to the left, e.g. ‘She made a slight movement, as if she were trying to speak . . .’. Here, the focus is on the behaviour of the characters as well as the narrator's comment on their body language, all of which come together to form an aspect of characterisation. Chapter 8, ‘Labels: contextualising and highlighting functions’, discusses labels or ‘a cluster that contains a name or is a part of an expression that can be used to talk about a fictional character’ (p. 152). Dickens is particularly well known for giving his characters names that are loaded with character-defining connotations – whether intentional or not. This would also include labels describing a character's physical attributes. Other character tags include body language labels that may also be informative about a character's behaviour. The final chapter, ‘Conclusions and outlook’, does just that. Mahlberg succinctly brings together the salient arguments and findings in her study of Dickens's characters in a text-driven fictional world. In her précis, Malhberg also raises a number of pertinent questions about the acquisition, storage and retrieval of linguistic units that researchers in psycholinguistics and text processing may be able answer to help further our understanding of the interplay between text analysis and the processes for their understanding. The final appendices that complete the book provide the lists of data such as the various clusters retrieved and examined in the DCorp corpus.

Mahlberg presents a thorough and advanced study of Dickens's fiction using corpus stylistic approaches to textual analysis and provides an important contribution to both corpus linguistics research and stylistics. By focusing on clusters as a way of understanding textual functions, and foregrounding patterns that reveal Dickens's literary techniques and linguistic strategies for creating characterisation, her book offers original and groundbreaking insights into the work of this popular author. Corpus sylistics and Dickens's fiction will be of particular interest to those who work and research in the areas of corpus stylistics, literary linguistics and literature as well as anyone interested in Dickens generally, who wants to understand how he artfully creates the physical and social contexts of his fictional worlds.