At 11:44 on 8/5/14 David Rose wrote to sysfling:
One reason I think why children can readily recognise clauses from an experiential perspective, is the tacit metalanguage in their own grammar… who or what it's about, what they're doing, where, when, how. But these are not directly related to transitivity categories, which are specified by process type or ergativity. Rather they are discourse semantic elements, that Ruqaiya Hasan has referred to as 'message parts', and Jim Martin as nuclear relations.
From this existing intuitive knowledge of children it is a simple step to consciously recognise a clause as a process involving people and things in places and times.
 The elements Rose talks about — 'who or what it's about, what they're doing, where, when, how' — are not "the tacit metalanguage in their own grammar". Metalanguage is not "in their own grammar". Metalanguage is language about language. In this instance, the metalanguage is Rose's language about the language of children, not the language of children.
 On the SFL model, these are experiential meanings, and as such, are indeed directly related to the "transitivity categories" of the lexicogrammar; as meanings, they are of a higher stratum of symbolic abstraction, semantics, and thus related to lexicogrammar by realisation.