Martin (1992: 98):
Thus in text [3:1] above, aside from Rhinocerous, which is not a grammatical English nominal group, every participant is marked by the grammar as "given" or "new". Basically, indefinite nominal groups code the identity of the participant being realised as not recoverable, whereas pronouns, demonstratives, the definite article and proper names signal that the participant's identity is in some sense known.
The 7-year-old's text under discussion is:
[3:1] at the zoo
One day I went to the zoo and I saw Rhinocerous I moved to a Hippopotamus I touched him and he is hand and he is big and so I went on and I saw the tiger and this man was feeding him it was eating it up Mum tod me mv on and next came then a gorilla. I had a baby gorilla. Mum tod me to move on. I saw a watch. It was 5 ock. …
 Here Martin confuses newsworthiness (the structural system of given and new information) with identifiability (the non-structural system of cohesive reference). See Halliday & Matthiessen (2014: 623).
 Here Martin again confuses non-specific nominal group deixis ('indefinite') with cohesive reference ('not recoverable').
 Here Martin distinguishes between demonstratives and the definite article. As Halliday & Matthiessen (2014: 630) point out, the definite article is a demonstrative:
The word the is still really a demonstrative, although a demonstrative of a rather particular kind.
 This foreshadows a major misunderstanding that pervades Martin's model and invalidates it theoretically: the confusion of referents (proper names) with reference items.
 This confuses what the author presents as identifiable with what is known by the reader.