David Rose 22:37 17/2/14 to sys-func:
My view is the elaborate relational theory of meaning that has been developed in SFL, building on Saussure, Firth, Malinowski, Hjelmslev etc. One type of relation is valeur between features in systems, another is between these features and their structural realisations, another between these realisations and instantiations in text, another between unfolding structures in texts. These sets of relations are cycled through relations between functions and classes at each rank, and each stratum of language and social context. Meaning can't be taken out of these relations and defined as in a dictionary, the relations are the meanings.
 A less confusing way to put this is: the valeur of a feature in a system is its relations to other features; the set of oppositions it contrasts with. It is the feature that carries valeur. The advantage of using the term 'valeur' is that it pinpoints a precise definition that would be obscured by the general term 'meaning'. It also avoids the confusion that would arise in speaking of the "meaning" of a phonological feature such as [stop] or [velar].
 The relation between systems and structures is not "meaning", it is realisation: a relation between different levels of abstraction; e.g. the relation between syllable systems and syllable structures is not "meaning", it is the realisational relation between the paradigmatic and syntagmatic axes.
 To the extent to which this is coherent, the relation between 'structural realisations' and 'instantiations in text' is not "meaning", it is realisation as in : i.e. the relation between the paradigmatic — (instance of) the system — and syntagmatic axes.
 As [1,2,3] demonstrate, Rose's view is not that which has developed in SFL, just a misconstrual of it that fosters confusion.