In case anyone it interests anyone, here is the book (particularly section 3.2) from which I learnt to consider atleast two kinds of "representations". More specifically, this paragraph, but the entire section (or the book*) might be an interesting read.
Whether we call any retinal or other neural layout a representation is partly a question of terminology, and anyone is free to call it a representation if they wish, on the grounds that NLs usually carry information about something in the world to which it is causally connected. What does matter is not the terminology, but the distinctions we need to make with respect to the role NLs play in explanations. If we use the term “representation” to refer to any information-bearing state, then we will still need to distinguish another, stronger, sense of representation. The main distinction we still need is between states whose representational content plays a role in explanations and those in which the content (if any) does not. If we gain no explanatory advantage by specifying what a NL represents, then it is not relevant to the question of explaining how mental representations connect with the represented world . The fact that the NL may look like a map – even if places on the NL correspond to places in the world – is still not enough for it to be a representation in the strong sense.