A common definition that teachers use is in relation to similes, that compare one thing with another with ‘like’, while metaphors say one thing is the same as another. My strategy is to show them that metaphors have ‘two meanings’, the ‘literal' meaning of the wording and an ‘inferred' meaning that is different. And comparison is only one type of metaphor. They can be recognised because the literal meaning doesn't make sense in the co-text. The relation to SFL theory that teachers recognise is three levels of comprehension, literal, inferential and interpretive, where literal is defined simply as recognising meaning with a sentence, inferential as recognising connections from sentence to sentence and page to page, and interpretive as recognising connections to one’s knowledge and values. In technical SFL terms these are explained stratally as lexicogrammar, discourse semantics and register. This and much much more can be found in… Rose, D. (2017). Reading to Learn: Accelerating learning and closing the gap. Teacher training books and DVDs. Sydney: Reading to Learn
 This misconstrues (literary) metaphors as similes — the comparison ('same') of phenomena — despite having already identified similes as concerned with comparison ('like'). The general frame of literary metaphor X is a Y, not X is the same as a Y.
A similar misunderstanding of metaphor can be found in Working With Discourse (Martin & Rose 2007: 45), as demonstrated here.
 Here Rose misrepresents levels of symbolic abstraction in the modelling of language as levels of comprehension of readers.
 In SFL theory, the "meaning in a sentence" is not lexicogrammar; it is the semantics (sequence) realised by the lexicogrammar (clause complex). 'Sentence', on the other hand, is a graphological unit.
Moreover, the (lexicogrammatical) wording may be congruent with the meaning ('literal') or incongruent (grammatical metaphor).
 In SFL theory, the "connections from sentence to sentence" are those of lexicogrammatical cohesion, the non-structural resource of the textual metafunction. Martin (1992) misinterprets these non-structural textual systems of the lexicogrammar as structural experiential, logical and textual systems of his 'discourse' semantics, as demonstrated here.
 In SFL theory, 'register' is a functional variety of language, and as such, is modelled as a sub-potential of language, midway on the cline of instantiation between system and instance. Martin (1992) confuses registers of language with the cultural contexts they realise and misconstrues register as context potential — i.e. neither language nor sub-potential — as demonstrated here.
 The use of the term 'accelerating' is interesting here, given that Reading To Learn is largely Rose's rebranding of Brian Gray's Accelerated Literacy. Rose's doctoral research was concerned with providing an SFL description of an indigenous language; he worked with Brian Gray after its completion.