Halliday & Matthiessen (2014: 44):
Globally, a text is structured according to the situation it operates in; the contextual structure is projected onto the text, and the contextual elements are realised by patterns of meaning in the text. As a semantic unit, the text consists of semantic domains of different sizes. It is likely to consist of rhetorical paragraphs (or parasemes (see Halliday, 2002d), which may or may not correspond to orthographic paragraphs in writing). In turn, these consist of sequences – sequences of figures, i.e. configurations of processes, participants involved in these and attendant circumstances. These more local domains, sequences and figures, are typically realised grammatically: sequences are realised by clause complexes, and figures by clauses.