The overall aim of this research project is to identify and analyse recurring prototypical discourse patterns in academic writing in English in order to inform the teaching of writing for academic purposes to non English-speaking background (NESB) students. Chapter 1 presents the motivation for the thesis, and outlines the aims, objectives, and research methods. Two concepts for the classification of discourse are proposed cognitive genre and social genre. In relation to the former, four prototypes (termed Rhetorical Types), characteristic of written academic discourse, are identified. Chapter 2 reviews a number of approaches to the categorisation of discourse, beginning with classical rhetoric, and tracing the progress of discourse-classification constructs over time, focusing primarily on cognitive rather than social genre. Chapter 3 critically reviews recent pedagogic approaches to the classification of extended written discourse in terms of social genre. These focus, at a macro-level, on overall organisational structure and, at a micro-level, on linguistic features. It is concluded that any adequate construct for classifying extended written discourse also needs to involve cognitive frameworks that organise knowledge (see Chapter 4). Chapter 4 reviews cognitive approaches to the classification of knowledge and seeks to establish the centrality of knowledge-organising prototypes and prototype-internal, procedural knowledge and to relate these to the cognitive organisation of written discourse. Chapter 5 proposes a cognitive model for the organisation of procedural knowledge in written academic discourse in English. On the basis of this model, four types of cognitive genre (referred to here as Rhetorical Types) commonly drawn upon in the structuring of academic discourse are discussed. They are: Report, Explanation, Recount, and Discussion. The relationship between the Rhetorical Type construct (as a cognitive genre) and the concept of social genre is also discussed. Chapter 6 examines each aspect of the Rhetorical Type model proposed earlier (see Chapter 5) in terms of its occurrence in a corpus of academic texts and by reference to sample texts from the corpus. Chapter 7 reports on two studies. The first involves examining samples of native-speaker and non-native-speaker writing in terms of the model proposed in Chapters 5 and 6. In the second, the model ratings of a small number of scripts (from the first study) are compared with grades assigned to the same scripts by two expert raters unfamiliar with the model. Chapter 8 provides an overview of the research, and discusses its implications for the teaching and learning of English as a second language.