Abstract
Narrative Progression in the Short Story by Michael Toolan investigates, using corpus tools and methods, how narratives guide readers’ expectations and responses. It explores a range of possibilities for computer-assisted detection of the parts of a text that are important for narrative progression, which are developed into an eight parameter model. Through the course of the book, a number of short stories are analysed, including: James Joyce’s Two Gallants; Katherine Mansfield’s Bliss; Raymond Carver’s Cathedral, Boxes, and A Small Good Thing; John Updike’s A&P; and Alice Munro’s The Love of a Good Woman. The book fits into the growing literature concerning computer assisted analysis of literary texts, offering a number of suggestions and directions for the analysis of narrative.
Chapter 1 sets the theoretical and analytical scene with a discussion of the key issues and concepts concerning narrative construction and coherence, and is directed towards how readers actively engage with a text through ‘narrative prospecting’, that is, using a combination of textual cues and background knowledge to make judgements and form expectations about the text.
Chapter 2 focuses on textual coherence and introduces some of the important concepts to do with lexical patterns and the associated corpus linguistic approaches developed to discover them. This prepares the reader for the following chapters, which attempt to link lexical patterns to progression in narratives. Included in the discussion are collocation, lexical priming and semantic prosody, with reference to the work of Sinclair, Hoey and Louw. Toolan also introduces his corpus of prose fiction, which is used as a comparator in the investigations carried out in later chapters. This is a 500,000-word corpus of novels and short stories by authors including Joyce, Hemingway, Faulkner, Woolf, Mansfield, Updike, Bowen and Atwood.
Chapter 3 moves on to explore lexical patterns in short stories, starting with an investigation of Joyce’s Two Gallants. The analysis begins with simple word frequencies and considers high and low frequency lexis, textual richness (type token ratio), lexical innovation, clusters (or n-grams) and unusual collocations. The investigation then makes an initial foray into keywords analysis. This turns out to be rather unrevealing with regard to narrative progression but acts as a prelude to a more productive keywords analysis in the following chapter.
In Chapter 4, evidence from respondent testing is used to evaluate the coherence of top keyword abridgements of Joyce’s Two Gallants, Updike’s A & P, and Carver’s Boxes and Cathedral. These abridgements are constructed using just keyword sentences, in other words, sentences that contain the most significant keyword for a particular story, derived via comparison with the author’s fiction corpus. The testing provides evidence towards the abridged stories being coherent, suggesting that keyword sentences are more relevant to narrative progression than much of the rest of the text, and such abridgements can help us to better understand the way narrative is constructed. Toolan notes that top keywords often tend to be names or main character referents, and that important plot-moving action or language happens around major figures in stories. This suggests that major figures coincide with narrative progression.
Chapter 5 explores lexical patterning in a story and how the lexical structure can guide a reader through a narrative. A number of respondents were asked to identify sections of Alice Munro’s The Love of a Good Woman that they felt would be important or relevant in the narrative that followed. Certain words and phrases are reported to be frequently singled out and Toolan discusses these at some length before eventually considering whether the priming elements of a story can be detected using corpus methods. Keywords are problematized in this respect since they highlight repetition of a single lexical item and ignore the use of (near) synonyms or variant terms, which might be important in narrative priming and prospecting. A suggested solution is to identify the most common collocates of the story-keywords in a large reference corpus (COBUILD is used by the author), then search for these collocates in the short story. In this way, it is not only keywords that are identified, but also what is commonly associated with them (according to COBUILD). This is seen by Toolan as a kind of textual cohesion, and a way in which a story might raise and meet a reader’s expectations.
Chapter 6 looks more closely at priming, cohesion and repetition in the structure of The Love of a Good Woman, firstly, by using Wordsmith’s ‘Plotlink’ utility to find keywords that collocate with other keywords within a given span of text, and secondly by investigating ‘para-repetitions’, which occur when similar entities are described using different (but related) language. Such repetitions form what are called intratextual links, and noticing such links involves making semantic relationships between words, which need ‘to be worked out, by a truly active reader’ (p. 104). Therefore, they elude the usual corpus methods. Toolan discusses this point with reference to a number of examples from the story. He also explores the priming function of story openings, working from the premise that the beginning of a story is important with regard to thematically related words and phrases and narrative purpose. Toolan suggests that judgements about what is important are based on where one’s narrative interest is drawn, which, in the case of The Love of a Good Woman, is, according to Toolan, human entities. These are investigated in the rest of the story, but the results are mixed and inconclusive. Toolan acknowledges this, noting that the analyses carried out so far suggest that the focus on lexical frequency and keywords is too narrow. This points towards the need for a model that takes into account a variety of factors.
Chapters 7 and 8 describe such a multi-factor model. Building on some of the analyses in previous chapters, Toolan proposes a checklist of textual elements that are likely to assist narrative prospection:
Sentences containing references to the main character when the reference used (e.g. name, pronoun, noun-phrase) is the top-keyword.
Sentences that contain a narrative-tense active verb where a main character is a participant in the clause; the character must be a frequently occurring keyword.
The initial sentence of a narrative paragraph.
Sentences containing frequent lexical keywords. Frequent is defined as keywords that represent 0.1% of the text, or that occur a minimum of 5 times.
Sentences that contain the presentation of character thought, especially free indirect thought and direct thought.
Presentation of direct speech that forms questions, requests/directives, undertakings and future-orientated statements.
Sentences where negation occurs.
Sentences that contain narrative verbs of modality and/or mental processing.
These are the main textual resources for guiding a reader’s expectations about narrative continuation. Of particular interest to me was the discussion of item 5 on the checklist, because this incorporated an exploration of the possibility of automatically detecting free indirect thought. This involves a three-step procedure which Toolan explains in some detail. He reports that initial findings from testing the procedure on a number of short stories, ‘give some cause for optimism’ (p. 144). However, further, more thorough testing would need to be carried out on a larger amount and a broader range of examples. Using the Lancaster Speech, Writing and Thought presentation (SW&TP) corpus, which is marked-up for SW&TP categories as a result of rigorous, manual analysis, could be a good way to continue with this line of research.
Chapter 9 represents a testing out of the eight-parameter checklist by investigating the creation of suspense and surprise using abridged versions of Joyce’s Two Gallants, Mansfield’s Bliss and Carver’s A Small Good Thing. The chapter first describes in some detail the conditions for suspense and surprise, drawing on a number of examples. Suspense is created when (i) a fork is reached in the plot, which has (usually) two easily foreseeable outcomes; (ii) an agent (character, reader) cares about the outcome; and (iii) the outcome is delayed. Surprise occurs when a story, which seems to be progressing in a ‘script-like’, ‘predictable’ and ‘foreseeable’ way, does not continue as expected.
The rest of the chapter is devoted to exploring the textual features likely to coincide with suspense and surprise in abridged versions of the short stories produced using the eight-parameter checklist. The results are again promising and indicate that the model can capture some of these textual features.
Finally, Chapter 10 sets out the next steps for studying narrativity using corpus tools and methods. Here Toolan evaluates the model and suggests a number of amendments and further refinements. Some of the suggested areas for future consideration and/or inclusion include modality, evaluation, and point of view. Also high on the list of directions for future research is the testing of the prospection model against readers’ responses to stories.
As Toolan states in the final chapter, this book represents a work in progress, and presents a number of investigations that turn out to be rather inconclusive in their findings. Nevertheless, it offers some potentially interesting avenues and ideas for corpus-based investigations of fictional texts. While the investigations in the book are directed specifically towards narrative structure and how fictional prose guides the reader, the book as a whole is likely to be of interest to anyone engaged in the analysis of prose fiction using corpora or corpus tools.