Martin (1992: 145-6):
Finally, a comment on generic chains. The second paragraph of [3:16] above has been extended to illustrate the way in which these chains are re-initiated each time a non-phoric nominal group is used to realise the generic group in question. The group cool deserts does this twice in the extended version of [3:16]. Recalling that the is not phoric in generic groups, this is a common textual pattern.
…Cool deserts are found further polewards in the deep interiors of large continents like Eurasia or where mountains form rain-shadows, which keep out rain bearing winds that might otherwise bring wet conditions… [For much of the year they may not look like deserts because they are lightly covered with snow. But the snow is not very deep and because it does not melt and run off gives a false impression of how wet cool deserts are. In winter they simply save whatever precipitation they get. Cool deserts can in fact be every bit as dry as the Sahara and some of them are even drier.]
This means that as far as reference structure is concerned, [3:16] includes three independent chains, all realising the same generic participant. Since cool deserts is not a phoric group, it cannot be shown to depend on previous realisations of this participant; but the chains can be aligned directly under one another by way of showing that the same participant is being realised, though not continuously presumed. This strategy brings out the experiential similarity between generic and specific chains alongside their textual differences.
 As previously explained, the distinction between generic and specific reference, which Martin (1992: 103) defines as:
Generic reference is selected when the whole of some experiential class of participants is at stake rather than a specific manifestation of that class …
though labelled as a relation of delicacy (generic vs specific), confuses part-whole relations ('whole') with token-type relations ('manifestation of a class'), the former, the extending logical relation of composition, the latter, an elaborating logical relation of instantiation; see Halliday & Matthiessen (1999: 145-6).
 The claim here is that a nominal group breaks a reference chain of the participant it realises simply because it doesn't include a reference item itself. Even in terms of Martin's model, where a reference chain is a chain of nominal groups denoting the same identity, this is is nonsensical, since the presence or absence of the has no bearing on the identity of the participant being realised.
In terms of SFL theory, each of the six highlighted personal reference items in [3:16] — they they they they them— makes anaphoric reference to cool deserts.
 This again misunderstands ideational denotation ("realising participants") as textual reference.
 This is manifestly untrue. Whenever the appears in a "generic" nominal group, such as the cool deserts, it is either homophoric, anaphoric or cataphoric (including Martin's "esphoric", as in the cool deserts we talked about earlier).
 This is a bare assertion, made without evidence from corpora, based on a falsehood; see .
 To be clear, no argument has been made for why a reference chain can be interpreted as a structure. In the source of Martin's ideas, Hasan (1985: 84), such cohesive chains are not presented as structures. Moreover, a structure is the internal syntagmatic organisation of a unit, such as the Theme^Rheme structure of a clause, whereas Martin's structure is a succession of related units (participants).
 This again mistakes nominal groups ("phoric") for reference items.