Halliday & Matthiessen (1999: 615):
A prerequisite for the semiotic construal of experience is generalisation: the move from "proper" to "common" as the basic principle of referring. The protolanguage, as already remarked, is non-referring; children move into reference by gradually deconstruing the proto-linguistic sign in a sequence of steps such as 'I want Mummy to ...', 'I want Mummy!', 'Where is Mummy?', 'Mummy' (see Halliday, 1992). The sign has now become a word, functioning as a proper name.
Typically one or two other signs have been deconstrued at the same time in similar fashion, e.g. 'I want my (toy) bird!', 'Where is my bird?', 'My bird'; and by a further step these then become common names 'bird(s)'. The child has now learnt to name a class of things; this then opens the way
(i) to constructing hierarchies of classes — a 'pigeon' is a kind of 'bird', and so on, and(ii) to naming other kinds of element, processes and qualities, which can be construed only as "common" terms.
Since these other elements have distinct and complementary functions it becomes possible to combine them into organic structures, as complex elements or as figures, such as 'blue bird', 'birds flying', 'tiny bird flew away'. The resources are now in place for construing experience in lexicogrammatical terms.
This principle of generalisation — that is, naming general classes rather than specific individuals — is what makes it possible to construct an ideation base. When they have reached this stage, children can make the transition from protolanguage to mother tongue, building up figures and sequences of figures, and simultaneously structuring these as moves in dialogic exchanges (question, statement, etc. — the interaction base), and as messages or quanta of information (the text base). In other words, they learn "how to mean" according to the metafunctional principle of adult human semiosis.