Martin (1992: 98-9):
Comparing Tables 3.1 and 3.2, it can be noted that there are correlations between phoric and non-phoric groups and first and subsequent mention: all non-phoric groups for example are associated with first mention. But several participants in [3:1] are realised phorically at first mention: e.g. the writer (I), the zoo (the zoo), the tiger (the tiger) and the mum (Mum). A number of these are perfectly appropriate and will be taken up in 3.3.3 below (for a thorough discussion of the correlations between phoric/non-phoric items and first/subsequent mention in a related narrative context, see Du Bois 1980).
 This continues Martin's confusion of reference items with the nominal groups in which they appear, as a result of confusing cohesive reference with nominal group deixis, as demonstrated in previous posts.
 Here Martin invalidates his own model, in terms of theoretical consistency, by unwittingly confusing two different senses of linguistic reference, from two different traditions. The sense of 'reference' used by Du Bois is the Bedeutung of Gottlob Frege (1892), as demonstrated by this quote from Du Bois (1980: 208-9):
Having used the term referential on several occasions, it will be well to specify the particular meaning I attach to it.
(2) A noun phrase is referential when it is used to speak about an object as an object, with continuous identity over time.
The object here may be a physical object or an objectified concept; it may be specifically known or it may be unknown; it may exist in the real world or in some hypothetical world; there may be one or more than one object. As long as a noun phrase is used to speak about such objects and the objects are conceived of as having continuity of identity, the noun phrase is referential.
That is, Du Bois is concerned with the use of noun phrases to refer to extralinguistic "objects", physical or conceptual. This is the view of meaning as 'transcendent' of semiotic systems. Halliday & Matthiessen (1999: 415-6):
We can identify two main traditions in Western thinking about meaning (see Halliday, 1977):
(i) one oriented towards logic and philosophy, with language seen as a system of rules;
(ii) one oriented towards rhetoric and ethnography, with language seen as resource. …
These external differences are associated with internal differences as well.
(i) First, the orientations differ with respect to where they locate meaning in relation to the stratal interpretation of language:
(a) intra-stratal: meaning is seen as immanent — something that is constructed in, and so is part of, language itself. The immanent interpretation of meaning is characteristic of the rhetorical-ethnographic orientation, including our own approach.
Many traditional notions of meaning are of the second kind — meaning as reference, meaning as idea or concept, meaning as image. These notions have in common that they are 'external' conceptions of meaning; instead of accounting for meaning in terms of a stratum within language, they interpret it in terms of some system outside of language, either the 'real world' or another semiotic system such as that of imagery.(b) extra-stratal: meaning is seen as transcendent — something that lies outside the limits of language. The transcendent interpretation of meaning is characteristic of the logico-philosophical orientation.