Martin (1992: 506):
(iii) Making genre rather than register variables responsible for generating schematic structure makes it easier to handle changes in experiential, interpersonal and textual meaning from one stage to another in a text. There are many text types where this occurs: a teacher may shift fields to explain a point by analogy; a salesperson will manipulate tenor in order to close a sale; sports commentators shift rhythmically from play to play description to critique and evaluation. Underlying register, genre can be used to predict these changes, stage by stage, while at the same time accounting for a text's overall coherence. Halliday and Hasan's (1976: 23) observation about coherent texts being "consistent in register" cannot in other words be interpreted literally as "the same register throughout"; rather the text must be motivated by its register, changes in which can in turn be motivated by genre.
 The use of 'generating' here confuses semogenesis with the realisation relation between the paradigmatic and syntagmatic axes. The relation of system to structure on a given stratum is one of realisation; system is a higher level of symbolic abstraction than structure. Generation, on the other hand, is the unfolding of text, logogenesis, at the instance pole of the cline of instantiation. This is a departure from previous confusions of semogenesis and realisation, which involve the realisation relation between strata, rather than axes.
 To be clear, by 'register variables' Martin means the metafunctional systems of context: field, tenor and mode.
 The claim here is that contextual changes of field, tenor and mode during the logogenesis of texts are easier to "handle" if the text structure of text types (genres) is modelled as a higher level of symbolic abstraction than the culture as semiotic system. The problems with this model include:
- one aspect of semantics, text structure, is removed from semantics and relocated to a level even more abstract than context (the culture as a semiotic);
- the stratification model is reversed: semantics (text structure) is realised by context (field, tenor and mode);
- cultural context (field, tenor, and mode — misconstrued as register) is modelled as the interface between two levels of language, in Martin's model: between genre and discourse semantics.
 Martin's claim that his higher stratum (genre) underlies his lower stratum (register) betrays his misunderstanding of stratification.
 This again reverses the relation between strata. Contextual changes of field, tenor and mode during the logogenesis are realised in changes in text structure at the level of semantics.
 The coherence of a text is not accounted for by an incoherent model.
 This is misleading in a way that suits Martin's argument. The interpretation is not Halliday and Hasan's, though Martin's use of quotation marks invites this attribution.
 In addition to all the above-mentioned inconsistencies, this presents the relation between strata as one of motivation rather than realisation. That is, it misconstrues an intensive (elaborating) identifying relation as a circumstantial (enhancing: cause) identifying relation.