Some things, some Kantians say about the justification for logic suggest the following superficially attractive idea:
a) Certain sentences have the property that anyone who can think about them must thereby be inclined to accept them (e.g. you can't even think thoughts involving `and' if you aren't willing to accept 'If it's raining and it's snowing then its raining')
b) We are justified in accepting such sentences, because we have no other option.
But actually, its plausible that we can reasonably doubt many sentences with this property. This is because sometimes you can turn out to have been working with an 'incoherent concept' like tonk, or bosh, the naive concept of set or perhaps various philosophical concepts. In such a case, you don't count as thinking with the concept unless you are willing to make certain (bad) inferences.
Now, you might argue that someone who was taken in by this kind of incoherent concept doesn't count as thinking anyway (e.g. there's no proposition which ``it's raining tonk its snowing" expresses). So maybe you only count as *thinking* in the good case, where it turns out that your concepts are coherent. But, given that we know that very smart and conscientious people can wind up with bad concepts, it intuitively seems reasonable to not completely dismiss the possibility that various new concepts you are learning are among the bad ones.
Hence*, it would seem that, we can rationally doubt claims which it would not be possible to deny. (what we're concerned about here is not the possibility that the claim is false - which we can't entertain- but that one of the concepts figuring in it is incoherent, so the claim is nonsense)
*if a) is true