I recently dreamt that a friend of mine was -- despite all appearances to the contrary -- a cat. I don't recall the superficial events of the dream as providing any supporting evidence for this (I was surprised to learn that she had children, but they looked like baby owls, not kittens, so go figure). Nevertheless, I must insist that it was true in the dream that she was a cat. This raises the philosophically interesting question: how? What determines the "facts" internal to a dream?
It won't do to appeal to the simplest or most charitable interpretation of the sense data. In any remotely realistic possible world where events unfold largely as experienced in my dream, I am simply mistaken to believe my friend is a cat. She has every appearance of being human (well, until you factor in the avian offspring! But a revisionist might as well claim that I was mistaken to think the owls were truly her offspring, too.), and there is nothing in the dream -- apart from my firm belief -- to suggest that she is actually a cat instead.
Still, it seems to me, authorial intuition trumps all. Dreams don't have to be simple, sensible, or even consistent. If a dreamer intuits (with psychological certainty) that P, then P is true in the dream. As subconscious authors of our own dreams, we have a special authority and privileged access to their content. Appearances can be misleading, even in dreams, and dreamers may have "authorial intuitions" to inform them when this is the case.
This raises difficulties, because we must also recognize the possibility of being deceived within a dream. Here I think we must appeal to a sort of 'duality' to the dreamer's sense of self. This arises from the multiple roles we play in our dreams, as both authors and actors, third-personal narrators and first-personal characters.
(This strikes me as supported by phenomenology. It seems like I sometimes switch between third- and first-person perspectives while dreaming, going from spectator to participant and back again. I may know things as a spectator that cannot influence the actions of my 'character' -- perhaps due to a kind of compartmentalization, or perhaps due to my refusal as an "actor" to break from the "script".)
With this distinction in place, I want to say that our authorial, spectating, third-personal self is the ultimate arbiter of what's true in the dream. While in the first-personal mode, we might sometimes be deceived or mistaken about the internal facts of the dream. But even the participating dreamer can sometimes attain infallible insight into the dream-facts, by way of what I call "authorial intuitions". These are accompanied by a distinctive phenomenology, a feel that this is how things are, which allows the dreamer to recognize them as certain facts (within the dream). We might consider this a sort of communication between the two parts of the dreaming self, an exception to the "compartmentalization" posited above.
Does any of that sound even remotely plausible?