Halliday & Matthiessen (2014: 44):
In general, two approaches to the account of the depth of texts have emerged in various traditions: the depth of layering may be modelled in terms of a semantic rank scale operating with some kind of constituency structure (analogous to the lexicogrammatical and phonological rank scales discussed above), as in Longacre’s work since the 1970s; or it may be modelled in terms of internal nesting of relational organisation — along the lines of Grimes (1975) and Beekman, Callow & Kopesec (1981). Within systemic functional linguistics, we also find these two models of the depth of text — the rank-scale model with rhetorical units proposed by Cloran (1994) and the internal-nesting model derived from Rhetorical Structure Theory (RST, e.g. Matthiessen & Thompson, 1988; Matthiessen, 1992, 2002a). The two are applied to the analysis of the same text by Cloran, Stuart-Smith & Young (2007). They are not, of course, mutually exclusive; they can be interpreted as capturing different aspects of the ‘depth’ of texts. And as grammarians we do not have to choose between the two as long as they provide us with motivated accounts of how to relate semantics to grammar.