Halliday & Matthiessen (2014: 437-8):
There are syntagms that may look as if they need to be interpreted in terms of a multivariate sentence structure, involving textual or modal Adjuncts or Vocative elements as Theme:
||| However << after the results of many studies were published, >> there was a shift towards the theory being quite unacceptable. |||
||| Interestingly, <<< as I left my small town || and explored the world via the military >>> I realised || I really like to learn, || and I was good at it. |||
||| Larry, Larry, Larry, << when you’re in the public eye >> you don’t do anything. |||
However, these can all be analysed as hypotactic clause combinations, where the dependent clause is included within the main clause after the textual and/or interpersonal Theme and before the topical Theme: main clause << dependent clause >> – more specifically, main clause [textual + interpersonal Theme] << dependent clause >> main clause [topical Theme ^ Rheme]. The motivation behind such sequences with included dependent clauses is thus textual: the main clause is powerfully contextualised first by its own textual and/or interpersonal Theme, and then, within the domain of the clause complex, by the dependent clause that qualifies it, and finally by its own topical Theme.
Nevertheless, it will be seen later that there is one type of clause complex that is interpreted as a multivariate (Theme^Rheme) structure — though not acknowledged as such. Halliday & Matthiessen (2014: 552):