The aim of this article is to explore how exactly the idea of distinguishing different coding levels in language has been theorized in different stages of Hallidayan systemic functional grammar (SFG), focusing on its view of the syntax-semantics interface. This is done by juxtaposing the levels of the Hallidayan model and the various components of Hjelmslev's model of stratification, on the basis of Halliday's re-interpretation of Hjelmslev's theory at various stages in the development of SFG. In this exploration, specific attention is paid to two important theoretical aspects of the design of Hjelmslev's and Halliday's models: (1) the different dimensions along which semiotic distinctions are made in the two models, i.e. dimensions along which language, as a semiotic system, is 'partitioned' into different components in order to explain and describe it; and (2) the semiotic relationships between these strata and components as defined by Hjelmslev and re-interpreted by Halliday, viz. the relations of 'manifestation', 'exponence', 'realization' and 'instantiation'. It is shown that Halliday's multi-stratal model blurs fine-grained distinctions which play a crucial role in Hjelmslev's theory, and that Halliday's concept of 'semantics remains underspecified in comparison to Hjelmslev's model. By taking Halliday's model of language as a basis, but 're-connecting' it to the detailed semiotic framework laid out by Hjelmslev, by which it was originally inspired. I argue that in the three different stages of SFG, three different types of 'semantics' have been highlighted. (C) 2010 Elsevier BM. All rights reserved.
