Much work in stylistics involves the in-depth analysis of individual literary texts or extracts, usually in order to relate specific linguistic choices and patterns to potential meanings and effects. To my mind, this kind of work represents one of the main strengths of the stylistics tradition: for all the controversy that sometimes surrounds the linguistic study of literature, explicit, rigorous, and sensitive linguistic analyses provide invaluable insights into the workings of texts and language generally, as well as useful hypotheses and explanations with respect to readers' interpretations.
Inevitably, however, the analysis of specific texts involves implicit or explicit comparisons with other texts. Claiming that particular linguistic choices and patterns are significant because they are deviant, or conventional, or typical of an author or genre inevitably involves claiming that similar choices and patterns will, or will not, normally be found in other (comparable) texts, or in general language use. This is where analysts often have to rely on their own intuitions as language users and literature readers, and on the assumption that these intuitions will be shared by their audience.
The increased availability of corpora provides new resources that can usefully complement analysts' intuitions, and therefore strengthen and refine the conclusions drawn from the intensive linguistic analysis of individual texts. My aim in this article is to demonstrate this by carrying out an in-depth analysis of an extract from Julian Barnes's novel England, England against the background of a relevant corpus. I will focus particularly on the way in which characters' speech and thought is presented, and on how this affects the projection of point of view and the potential for readers' sympathy towards the characters.
The presentation of characters' words and thoughts is a crucially important aspect of narrative, which has received a great deal of attention within stylistics and narratology (see, for example, Cohn; Fludernik; Leech and Short; Page; Rimmon-Kenan; and Toolan). My analysis will benefit particularly from the findings of a corpus-based project on "Speech, Writing and Thought Presentation" (SW&TP) which I was involved in at Lancaster University in the mid-1990s (see, for example, Semino and Short; and Semino et al., "Using"). The project involved the creation of a corpus consisting of 120 extracts of approximately 2,000 words each, for a total of 258,348 words of (late) twentieth-century written British English. The 120 text samples were drawn from three different written genres: prose fiction (87,709 words), newspaper news reports (83,603 words), and biography and autobiography (87,036 words). Each genre section was further divided into a "popular" and a "serious" subsection. In the case of prose fiction, we made a distinction between popular romances and action novels on the one hand, and prestigious, "highbrow" novels on the other (by authors such as Virginia Woolf and Salman Rushdie). The corpus was manually annotated for SW&TP using a specially-designed annotation system (see Semino and Short 26ff.; and Wynne et al.). Thanks to this annotation, it is possible to search the corpus, and all its subsections, for particular forms of SW&TP in order to study their characteristics, frequencies, patterning, and potential effects.
The availability of this corpus will benefit my analysis in two ways. First, I will apply to the extract an updated model of SW&TP that was developed during the corpus project, starting from the account of speech and thought presentation in fiction given by Geoffrey Leech and Mick Short in Style in Fiction (1981). As I will show, this revised model captures a wider range of phenomena than was the case with previous models. Second, I will compare choices and patterns in Barnes's extract with the patterns that occur in the corpus as a whole and in the fiction section in particular. …