But we might've expected the contrary result to be possible. Consider the following passage taken out of context from Laurence Bonjour's In Defense of Pure Reason (pp.57-58):
I am presently looking at two books on my desk. Both are darkish blue, but not quite the same shade of darkish blue, though my rather meagre color vocabulary contains no names for these specific shades nor any other way of indicating them linguistically. On this basis, I come to believe and, so far as I can see, to know a priori a certain proposition that I can only indicate indirectly but cannot adequately express in language, the proposition that nothing could be both of these colors all over at the same time...
(Bonjour uses this to show that a priori justification cannot be accounted for merely by "linguistic convention". My current purposes are completely different.)
Now, one way that I would be tempted to state the described a priori thought is as follows:
(D) Nothing could be that colour [mental nod towards first book] and that colour [mental nod towards second book] all over at the same time.
Is D a priori? It seems to depend on whether the demonstrated colours are themselves part of the tokened thought, or whether the demonstrations are mere placeholders. If the colours are "built in", then the resulting thought is certainly a priori. (Those two colours are themselves incompatible in all scenarios.) However, if the thought instead contains "placeholders", then it is not a priori, since this allows different targets to "fill the gap" in different scenarios. In particular, it allows a scenario in which we pick out the same colour twice. Such a scenario falsifies D, so understood, since things can of course be one and the same colour all over at the same time. Failing to be true at all scenarios, D thereby fails to be a priori.
As I understand it, Chalmers' framework requires that we adopt the second interpretation: demonstratives are variable placeholders, and so "non-trivial" demonstrative thoughts cannot be a priori. Is that a problem? Should we want to allow the first interpretation? We will at least want to allow some thought or other to have this "built in" content. But perhaps we should deny that those are "demonstrative thoughts" any longer. We might instead suggest that Bonjour was thinking something along the lines of:
(D*) "Nothing could be colour1 and colour2 all over at the same time." (Those "colour" labels might be replaced by appropriate items of direct phenomenology, say mental images of the colours in question. We want epistemically rigid designation, anyway, however that is done. The two colours should be held fixed across all scenarios.)
Note that there's nothing essentially "demonstrative" about D*. Lacking the vocabulary to express the thought directly, perhaps the best we could do verbally is to say something like D instead, and point to the two colours we have in mind. But really it is the two colours, and not the pointing, that we do in fact have in mind. So I think that provides some independent support for the claim that, contrary to my initial speculation, Bonjour's a priori thought is not a demonstrative one.
Is that the best way to make sense of this problem? (Is the answer to that question -- or this one -- a priori? Heh.)