Halliday & Matthiessen (1999: 6):
But to show that a grammar is a theory of experience we use a functional, semantically motivated grammatics, since this allows us to seek explanations of the form of the grammar in terms of the functions to which language is adapted.