Quine said that even logic could be revised on empirical grounds. Did he later reverse that position? If so, in which of his writings?
I think it is fair to say that Quine retreated from his earlier position in regard to the revisability of logic, at least to some degree. Probably the most important reference here would be the chapter called Deviant Logics in his book Philosophy of Logic. Quine argues that there cannot truly be deviant logics, because anyone who claims to use different laws of logic is simply using the relevant words differently from ourselves. We might gloss this as saying that anybody who claims to use a different logic is merely changing the meaning of the words involved. Quine would not put it that way, since he does not accept such things as 'meanings'. Rather, he expresses it in terms of translation: if a speaker of an unfamiliar language professes to accept a compound sentence, but not a constituent of it, that gives us a reason not to translate the compound as a conjunction, etc. Quine's idea is that logic, specifically classical logic, is built into any translation manual that we might use to enable us to understand a foreign speaker. Indeed, it is more firmly built in than anything else, including scientific statements. Attempting to attribute a deviant logic to a foreign speaker would be like attempting to interpret them as believing something obviously false, like 1+1 is not equal to 2.
Whether this reasoning is correct is another story. For myself, I am much more receptive to the idea that there can be non-classical logics. It is true to say that if you change a logic then you are implicitly changing the meaning of at least some of the logical constants involved, but we can still recognise a family resemblance between constants. Intuitionistic negation does not mean the same as classical negation, but I can understand and use both, and I am aware that both have a claim to be called negation because of the similar roles they play in their respective logics. The intuitionist is not changing the subject when they use their version of negation, nor would I suppose that they are saying something obviously false if they deny the rule of double negation elimination. Likewise with the difference between relevant implication and material implication, etc. Whether there are specifically empirical grounds for revising logic is a more tricky question.