ABSTRACT
The Sydney architecture of systemic functional linguistics (SFL) is known for its wide applications in educational linguistics and social research, with a focus on power relations, the social distribution of semiotic resources and the building of communities. The model is strongly influenced by J.R. Martin’s reading of Hjelmslev (1943, 1961) and Hartford stratificational linguistics (Gleason 1968; Gutwinski 1976), alongside Halliday’s systemic functional grammar. It is commonly characterised by its text-orientation, discourse semantics (for example Martin 1992; Martin and White 2005; Martin and Rose 2003, 2007) and the modelling of genre (Martin and Rose 2008) as a contextual stratum. Building on Halliday and Hasan’s (1980) initial work on cohesion and texture (see Clarke, this volume), the model takes an innovative approach to text as a unit of meaning with its own systems and structures (discourse semantics) that transcend the boundaries of the clause, and mediate between lexicogrammar and its context.1 As such, descriptions of discourse semantics relate systematically to both of its adjacent strata, and it is therefore necessary to consider the way in which context is modelled in this approach to understand the motivation underlying the particular treatment of discourse semantics in this model and to locate it within its overall architecture. Most distinctly, the stratified model of context in the Sydney architecture theorises two levels of context – register and genre – as semiotic systems in themselves that are expressed by linguistic resources, including discourse semantics.2 Beyond a mere redefinition of some SFL terminology, this difference in the formulation of context has significant implications for the understanding of the role and boundaries of linguistics. This chapter will outline the key concepts of context and discourse semantics by drawing attention to the differences between the model and other systemic models, before reviewing their practical implications for current directions in research (for example Bednarek and Martin 2010).
