Another approach is [*semantic externalism*](https://en.wikipedia.org/wiki/Semantic_externalism): "meanings aren't in the head". Hilary Putnam [argued for it](https://en.wikipedia.org/wiki/Twin_Earth_thought_experiment):

Imagine a planet mostly the same as Earth, but with no H2O, only a alternative substance (say XYZ) that's a biologically adequate substitute. The humans living there even call it water. But on the planet, water doesn't mean what it means here; it means XYZ, because of the world shaping their language.

Putnam argued that a brain in a vat doesn't have the right experiences to have accurate definitions of terms, so statements it thinks in its language will be false. So if it thinks "I'm a brain in a vat", it's wrong, just as its thoughts are wrong in general. But if even a brain in a vat can't accurately think it is one, then such brains don't exist.

This argument is related to the *principle of charity*. If you try to translate a recently discovered tribe's language, you have to assume their beliefs about the world are largely accurate, like your own, otherwise you can't fathom a guess at what they're likely saying when they say statements, or asking when they ask questions. Even though a brain in a vat remains logically possible, Putnam has rendered it too unworkable a scenario to reason intelligibly about. We can't really get anywhere if we think our experiences are too misleading for us to learn from them. By contrast, realism, even if only representational realism, is much more operable than any attempt at near-global scepticism about the empirical world.

I like externalism because of the light it casts on Quine's attempted dissection of the analytic-synthetic distinction. His "married bachelors" thought experiment involves a world different enough for terms to mean something different there, so that the fact that "all bachelors are unmarried" has a different truth value therein may be ascribed to the definitions being different rather than to the truth value depending on more than terms' definitions, which is what would happen if the statement were synthetic.