Martin (1992: 156):
Reference structures make use of semantic dependency structures of this kind. Phoric items depend on the co-text they presume, but are not themselves predicted by it. Given a nominal group realising a participant in a text, there is no way of knowing whether or not it will be presumed, aside from the occasional selection of presenting reference marking certain participants as central to a discourse (signalled through this, these, a certain and certain). The dependency relation between presuming and presumed is thus very like that between a tracking or challenging move and the move on which it depends.
Old Blogger Comments:
 The use of dependency (logical metafunction) for reference (textual metafunction) creates a theoretical inconsistency. It misconstrues textual meaning as ideational meaning.
 The dependency relations do not form structures; they merely obtain between units.
 Martin's insight here is that a reference item depends on there being a referent to refer to. To be clear, this is distinct from the SFL notion of dependency, and from the functional relation between the reference item and its referent.
 Martin's insight here is that there is no way of predicting which elements in a text will subsequently be referred to.
 As explained in previous posts, "presenting reference" does not involve reference. It is the first appearance of a potential referent. See Misconstruing The Absence Of Reference As "Presenting" Reference.
 For the confusions on which this false claim is based, see Confusing Experiential Content With Textual Reference.
 To be clear, Martin's argument is: A is very much like B, where
- A = a reference item depends on there being a referent to refer to, and
- B = an interruption (tracking or challenging move) depends on there being a move to interrupt.