Martin (1992: 389):
What this argument amounts to saying is that it is more economical to leave SUBSTITUTION and ELLIPSIS in the grammar, generating cohesive ties with respect to grammatical functions defined on that stratum. Since text forming resources are distributed across all three strata in the meaning making model assumed by English Text, there is nothing remarkable in this.
 This betrays Martin's modus operandi of theorising a discourse semantics — taking Halliday & Hasan's ideas from where they fit consistently within the overall theory, and moving them to a new ill-considered theoretical location.
 To be clear, in SFL theory, the text forming resources across all strata are those of the textual metafunction.
 This continues the confusion of stratification and semogenesis.
 There is much that is remarkable in this, not least because the meanings realised this grammatical system need to be modelled by a semantic theory.