Introduction

We noted in Chapter 1 that experts have often internalized patterns of discourse behaviour characteristic of neurological disorders and their phases which inform their clinical judgements. However, these internalized patterns may not be explicitly recognized and consequently appear as ‘intuitive’ responses. Just as people can often recognize regional or social dialect variation without being able to say precisely what it is they recognize, so the patterns informing clinicians' impressions may not be explicit. Descriptions of discourse can make experts' tacit knowledge about discourse patterns associated with diagnostic groups explicit, in the same way that descriptions of dialect variation can aid dialect recognition. Such descriptions can also be used to model characteristic patterns for healthcare workers and families who may not have experts' breadth of experience.

Beyond explicitness, one role of clinical discourse analysis is to add tools to existing diagnostic resources where diagnosis is still a clinical decision. Another is to provide characterizations for diagnostic categories which are under-investigated. These may help with diagnostic clarification and planning for treatment. A third role for description of discourse is to track change over time both intra-individually and for group applications. There is potential for monitoring developmental and degenerative processes and tracking responses to treatments and interventions. There is also a fourth role which is to improve understanding of relationships between everyday discourse behaviours and neurocognitive function.

In relation to these roles, in this chapter we model study designs which address discourse correlates of diagnoses and monitoring.

Introduction

As we outlined in Chapter 5, the term context may be used to refer to a variety of apparently rather different constructs including the physical and social environments in which communication takes place, common patterns of interaction that might occur in those settings, as well as meanings made relevant by ongoing events and discourse. Our use of the term context includes these senses, but interprets them specifically as information available to speakers. This information is acquired, some of it actively ‘learned’, in interactions which individuals participate in and is represented in episodic, semantic, linguistic and other repertoires. Speakers use this acquired and learned information to interpret what is happening, to shape responses to new, incoming, information and to communicate. We use context of culture, context of situation and phase to describe information relevant for interpretation of particular texts or discourses. Respectively, these reflect broad patterns of culture associated with language and dialect variation, situation types which constitute patterns of culture, and the specific discourses which instantiate culture and situation. We are interested in instances of discourse because new information is acquired ‘live’, moment by moment in situations and because examining instances allows us to observe discourse behaviour which may be clinically relevant and from which generalizations may be abstracted and related to other evidences of neurocognitive integrity and function.

Introduction

This chapter outlines approaches to top-down cognitive modelling and inferencing, and addresses functionally grounded work on affect. We describe each area and illustrate its potential for addressing questions in clinical discourse analysis. We also review recent work from neuroimaging and lesion studies to suggest some of the relevant neural systems. As usual, we draw on various disciplinary perspectives and theoretical models. Our practical motivation here is to use what works, and has potential for coding corpora in the various linguistic contexts and situations encountered doing clinical discourse analysis.

Cognitive models in general characterize information bundles of various kinds. Perhaps the most familiar are those used to represent words or word-like concepts. Models for words may be more or less detailed depending on the tolerance for elaboration within a particular framework, but morphosyntactic class, inflection and distribution features are typically indicated. How a word is pronounced – its phonological form and regular phonetic variants – will be spelt out in phonological and phonetic representations. Semantic features are often specified only at superordinate levels as in THING/EVENT or merely indexed through the use of the ‘CAPS-for-concept’ convention. Thus, the model for the lexeme ‘cat’ will include the information that it is a common count noun, with the inflectional and distributional features of this class – it can occur as head of a noun phrase and it inflects for plural number /s/. It is pronounced /kæt/.

Introduction

In this and the following two chapters we present descriptive resources which allow coding of spoken discourse in terms of conversational, grammatical, pragmatic and contextual features. Features that are coded can be counted. This allows both ‘pattern finding’, a replicable way of proceeding inductively to characterizations of discourse (a specialized kind of data mining), and ‘hypothesis checking’ where analyses in terms of the presence, absence, frequency and/or co-occurrence of discourse features are hypothesized to be associated with particular disorders, neurocognitive states, or changes in neurocognitive states.

Our goal is pragmatic: we do not enter into debates or account for the historical development of particular concepts or constructs, but present criteria for description that allow understanding and reliable coding of discourse features. Often this means that we are presenting ‘hybrid’ models, synthesized from works in functional, ethnographic, interactional and formal linguistic traditions. We refer readers to these works but discuss them only when such discussion seems essential for understanding.

The order of presentation, from conversational to contextual analysis, is intended to proceed from the familiar to, perhaps, less familiar ways of describing discourse. However, as is inevitably the case, some terms appear that presuppose knowledge of others presented later. For the most part, we address this by providing examples and/or glosses of terms and references. We use normative examples rather than examples from clinical contexts in presenting the grammar. Normative examples are practical insofar as they do serve to illustrate without the addition of marked features or other distractions.

What is clinical discourse analysis?

Clinical discourse analysis is the term we use to describe the analysis of language behaviour observed in clinical contexts. Language behaviour includes well-defined areas of clinical research addressing syntax, vocabulary, phonology, conversation skills and cohesion. It also includes areas less commonly described in clinical research such as argument roles, situational features and functional variation. The focus of clinical discourse analysis is natural language behaviour which requires examination of all these aspects of language use. Even the smallest of texts require analysis that can explore multivariate features.

Consider the following sentence:

(1) I can remember my Mom.

It is extracted from a brief sample of spoken discourse elaborated below. The sentence has an interactional function: it makes a statement which is modalized for capacity (can). It has a predicate that references a cognitive process (remember) and two argument roles, an experiencer (I) and a percept (my mom). It has an unmarked clause order: subject occurs first in English statements unless there is a reason to emphasize some other element of structure. One might also assume that the speaker interprets the situation as an informal one in that she chooses to refer to her mother as Mom rather than mother.

Contexts of culture and situation

All discourse is produced in context and interpretation depends on contexts of production and interpretation being, in some measure, shared. Early ethnographic work addressed context dependency by positing contexts of culture and context of situation (Malinowski 1923; 1935; Firth 1957). Context of culture accounted for sets of culturally specific beliefs, expectations and practices in terms of which people interpret events around them. Context of situation referred to patterns of behaviour and talk which appear so regularly in association with a particular activity that they are understood as (abstract characterizations of) the function of the situation type. Behaviours which do not reflect some expected pattern can be interpreted as irrelevant, and behaviours which appear totally unrelated to the contexts in which they occur may be judged uninterpretable.

Later work by Halliday, Hasan, Gregory, Martin and others refined and developed ideas of context. Our view is once again a hybrid, informed by Halliday's ethnographic perspective (e.g. 1977; 1978; 1984; 1994), by our awareness that contexts are significantly matters of what speakers know (e.g. van Dijk 1977; 2006; van Dijk and Kintsch 1983; Gregory 1988), and by work in AI, psychology and discourse analysis on top-down cognitive models as to what ‘contextual knowledge’ might be like. The latter approaches (and ours) differ from traditional functionalist and ethnographic approaches in explicitly situating context in neurocognitive domains of semantic and episodic memory (see also van Dijk 2006).

Introduction

In this chapter, we outline in sequence a grammar for specifying interactional, ideational and organizational features of discourse in English. We use a functional organization for the grammar because we find this perspicuous for discourse analysis. The grammar owes much to Michael Gregory (2009a; 2009b). However, in many cases our approach is once again hybrid; we present a model that draws on social and cognitive perspectives on language, based on our experience of what works. Section 4.1 presents a grammar of interaction. In Section 4.2 we outline the grammar of ideation and Section 4.3 presents the grammar of organization. We presuppose that readers have a basic syntax for English. As much as possible, technical and model-specific terms are limited. Appendix A presents basic grammatical terminology and Appendix B lists the coding options suggested in this chapter.

A grammar of interaction

The interaction relationship has to do, initially, with two types of activity speakers can engage in. The first involves the negotiation of role relationships relative to an addressee and the second involves speakers' expressions of attitude and evaluations. Role relationships are most easily understood in institutional settings with highly generic situation types such as the classroom, court room or clinic, where boundaries may seem relatively fixed: teacher/ student, judge/defendant, doctor/patient, nurse/patient are examples of such role relationships (Goffman 1959; 1961). Less obviously hierarchic and institutionally dependent are peer relationships among colleagues, friends, siblings and partners.

GRAMMATICAL TERMINOLOGY

CLAUSES AND UTTERANCES

In Chapter 3 we defined the terms Clause and Utterance (pp. 29–32). Here we include brief definitions for convenient reference. The reader is referred to Chapter 3 (pp. 29–32) for broader definitions and examples.

English Clauses can be defined syntactically as consisting of a verb, its arguments and adjuncts as in Cosmo bit Piper yesterday. Independent clauses can also be used alone to ask a question, make a statement or exclamation, or give a command. A simple Clause is also typically spoken as a single tone group.

Meaningful speech phenomena which do not meet the criteria for clause such as incomplete utterances, minimal responses, idiosyncratic vocalizations, and isolated hesitation fillers are labelled as utterances. Utterance refers to any unit which can be assigned a speech function, and/or has a distinct tone group, and/or is a linguistic signal of ideational, interactional, or organizational information about a speaker's message.

MORPHEMES

Morphemes are smallest contrastive units in the grammar. Rabbit, dog, -s, in dogs, -ed in called, -ity in fatality, un- in unhappy, brush in toothbrush are all morphemes. None of these items can be further analysed. For instance, rabbit refers to a small furry long-eared animal that hops, but no part of rabbit is associated with one of these meanings. Morphemes may or may not be words: Whereas rabbit is a word, -s in dogs signifies plural but does not occur as a word.

Words may be simple (e.g. rabbit/dog), compound (toothbrush) or complex (fatality/unhappy).