Halliday & Matthiessen (2004: 497n):
The traditional analysis was to treat the primary group as Predicator in its own right and the secondary group together with elements following it in the clause as an embedded non-finite clause serving as Complement, and this type of analysis was taken over in modern formal approaches, at least initially. … we can note that while the traditional analysis is forced on us if our only model of structural organisation is that of constituency, the analysis we present here becomes possible once we recognise tactic interdependency structures. It allows us to show the analogy, and agnation, between sequences of verbs, and sequences of clauses, with areas of indeterminacy between the two. It enables us to shed light on so-called ‘serial verb constructions’ in a range of languages, interpreting the findings that have emerged, particularly in the last 15 years or so. Further it enables us to show how categories of the simple verbal group have evolved from verbal group complexes. And it also makes it possible to avoid one of the major drawbacks of the traditional analysis: the secondary verbal group plus the elements following it do not in fact behave as Complements. For example, if they were Complements, the wh-interrogative should be what is she trying? [eg]; but it is not: it is what is she trying to do?.