A language-game is analogous to a chess game, i.e., you're either playing chess or not. To say you're playing chess, when you're making the wrong moves, is not chess. It's not that you're playing chess incorrectly, you're not playing chess at all. My contention is that there are incorrect moves in the game of chess, so if you move the rook diagonally, then you're not playing chess correctly. If you're teaching the game of chess, then it seems obvious that there are correct and incorrect moves based on the rules. — Sam26
Yes! I was going to use this exact image - of playing chess - as an example, but I dropped it for brevity's sake. A few more words on why it's so important to distinguish between "not playing" (no use) and "playing wrong" (incorrect use): the issue turns on how 'use' is understood. If use is 'use in a language-game', then use is always, as it were, something 'positive'. Use for a purpose, as it were - for
doing something. And the point is that if you are not doing something with a word, then in what sense can you be using it - and hence meaning something - at all?
This goes right to the heart of the issue of the 'publicness' of language: the further problem with admitting 'incorrect use' is that this more or less amounts to admitting private language. If there can be no private language for Witty, it is because all meaning is inseparable from doing: and doing is not something that can even in principle, be "private" - which is to say, unintelligible. Doing, for it to count as a 'doing', must be exhibitable. Others must be able, in principle, to 'pick it up', to learn from you how to 'go on', in a similar way.
This is also why it is so detrimental that people so often drop the 'in a language-game' part from 'meaning is use in a language-game'. Language-games are, by definition, public things. In fact, the importance of the idea of the language-game (which I think is so often missed) is that they admit of different
kinds. This is so important in fact, that Witty very early on in the PI writes out a whole list of them: giving orders, reporting, requesting, thanking, etc. This is further why language-games are not just 'contexts' (another term that is almost wholly absent from the PI). Would would it even mean to try and 'give orders' privately? Or request privately?
The inseparability of 'use' and 'language-game': they are mutually, 'analytically' defined by means of one another, means that a use which is not a doing is simply not a use at all. One can of course,
try to do something, but in a wrong way. One can make a wrong move in chess, and one can say: 'that's not a move you can make'. And like
@Banno said, this can introduce novelties. But the introduction of novelties still implies that one must be
trying to play chess - it must be something that others, in the future, can also pick up (the en passant): this use
expands the language-game: it alters chess itself. Chess is something different after the introduction of the en passant. A new use will bring with it a new language-game in tow, after which one cannot say of that use that it is incorrect. Prior to it's introduction, the en passant was simply a move in a language-game not accepted as chess.