Halliday’s description of register as ‘a variety of language, corresponding to a variety of situation’, with situation interpreted ‘by means of a conceptual framework using the terms “field”, “tenor” and “mode”’ (Halliday, 1985/89: 29, 38) is revisited to reflect on the theoretical work the term ‘register’ does within the SFL paradigm. In doing so, we recognise that the concepts of a linguistic theory are ‘ineffable’ (Halliday 2002[1988]); i.e. that ‘providing definitions of a theoretical term... requires that it be positioned vis-à-vis other concepts in the theory’ (Hasan, 2004: 16). It follows that changing the position of ‘register’ in the theory changes the nature of the concept. So while alternative uses of the term ‘register’ – such as in Martin’s genre model (e.g. 1992) and Halliday’s model – may advance a shared program for language description and explanation as a route to social change, they must be seen as more than terminological variants. One consequence of the productivity of Martin’s approach has been that the Hallidayan line of register theory has not had sufficient critical explication. This paper therefore begins with a brief review of the register concept. It then exemplifies the term, as postulated by Halliday, with a registerial analysis of surgical interaction, drawing on Hasan’s context modelling (e.g. Hasan 1995, 2004, 2009a), and adopting what Matthiessen (1993) calls a ‘metafunctional slice’ with ‘multistratal coverage’. By accounting for choice at different strata, we seek to ‘relate wording to context via meaning which acts as the interface between the two’ (Hasan 2009a: 182).