Martin (1992: 147):
The reference structure of [3:88] is presented in Fig. 3.14. … The two participants in possessive nominal groups are aligned with their respective chains with a dotted line showing that the identity of the participant realised in the Thing is recoverable via the participant by which it is possessed (e.g. his bed in [3:88i]). Where the Thing participant has already been identified (as with his pet frog in 3:88xxvii]), it is shown as well as presuming preceding items in this chain. Note that the analyses of reference structure consist only of phoric items and any non-phoric items they presume; non-phoric items which are not included since they do not enter into semantic dependencies related to participant identification. In [3:88] only four minor participant chains, the glass, the rock, the reindeer and the baby frogs are introduced non-phorically.
 This is inconsistent with both Martin's own model and SFL theory, according to which, a participant is realised by a nominal group, not the Thing (bed) of a nominal group.
 This is inconsistent with both Martin's own model and SFL theory, according to which, a participant is realised by a nominal group, not the Deictic (his) of a nominal group. The function of the inconsistency is to hide the fact that Martin's model, in mistaking nominal groups (e.g. his bed) for reference items (e.g. his), necessarily leads to the nominal groups Tommy, he and his bed all realising the same identity.
 To be clear, Martin has, to this point, used the term 'phoric nominal group'. In switching to the term 'phoric item', Martin signals that he is aware of the problem that arises from mistaking a nominal group for a reference item, and hopes to hide that fact from the reader.
 Trivially, unknown to Martin, there is a fifth minor participant chain that fits his criterion:
a hollow tree