Here's something I've been puzzled by for ages (well, since taking Selim Berker's awesome metaethics class).
Suppose that Jim likes strawberries* best on even months and chocolate best on odd months, and each time the month changes, he says that he has discovered that chocolate has/lacks the property of deliciousness (I mean: he uses exactly the same verbal forms of expression that he would to say that he discovered that my keys were in my backpack all along, draws logical consequences the same way etc). How can we understand someone with this kind of practice?
What's so crazily perplexing here, is that Jim's use of a single sentence "chocolate is delicious" seems to be associated with two norms:
1) take present and past self to be disagreeing about same proposition.
2) judge for yourself in a way that agrees with your dispositions to act/enjoy (i.e. it's not ok to say this is more delicious but I don't like it)
If you have a practice with both these norms, psychological instability gives rise to factual uncertainty. But it's extremely hard to disentangle things.
You can capture aspect 1) of Jim's use, if you interpret him as talking about the property of being 'the kind of thing people enjoy eating', but you can't make sense of 2). On the other hand, you can make sense of 2) if you interpret him as saying something like 'I like chocolate' or 'yay, chocolate' but you can't make sense of why he speaks as though he is discovering that some proposition he previously thought true was false (In the first case, you can't make sense of why he says he's disagreeing with his past self, in the second case it looks surprising that 'chocolate is delicious' logically embeds in the ordinary way, whereas 'yay chocolate' doesn't - this is called the frege geach problem).
Like someone who uses the word 'Tonk' (Pryor's connective that has the introduction rules for or, and the elimination rues for and) you can sort of describe how Jim behaves, and understand all scientific facts about it, but you can't translate his propositions as expressing any proposition.
[This reminds me of how McDowell says that the ethical antirealist can't understand the ethical realist. Maybe we are like the ethical antirealist, in this situation. But McDowell says that the antirealist can't understand the realist as even having a practice/'going on in the same way'. Whereas I think I have a very clear notion of what Jim's practice is, and what it would take for him to go in in the same way, there's just no practice of mine that's sufficiently similar to his for me to translate him by equating his sentence 'chocolate is delicious' with any sentence of mine. ]
However, unlike Tonk (which lets you infer any proposition in your language from any other), accepting Jim's practice doesnt seem prone to lead him to false beliefs that don't involve the word 'delicious'. (Assuming you don't count false beliefs like: 'there's something I learned about chocolate which I didn't know last month') So, maybe a better analogy would be the world 'gleb' which has the introduction rule: if it's sunny infer that it's gleb, and the elimination rule: if it's gleb, then infer that you should mail Sharon 5 dollars. Or the word 'chasitity' as used by someone who thinks that's a virtue.
But what exactly is going wrong here? And is it the same thing in all three cases? It's very tempting to say there's something wrong with all practices "of this kind", but what's the relevant kind?
*I picked the example of what you might call 'realism about deliciousness', since it's close to a real life example most people I've met seem to be instinctively anti-realists about deliciousness. But, if you happen to share Jim's way of using the word delicious, just pick some other arbitrary combination - e.g. imagine someone who accepts the norms 1) believe that whatever your actual height is, is the X-est height for a human 2) take yourself to be disagreeing with with past selves who accepted different claims about what the X-est height is.