The question is: Do we supplement our statement that the word has two meanings by a statement saying that in one case it had this, in the other that meaning? As the criterion for a word's having two meanings, we may use the fact of there being two explanations given for a word. Thus we say the word “bank” has two meanings; for in one case it means this sort of thing, (pointing, say, to a river bank) in the other case that sort of thing, (pointing to the Bank of England). Now what I point to here are paradigms for the use of the words. One could not say: “The word ‘red’ has two meanings because in one case it means this (pointing to a light red), in the other that (pointing to a dark red)”, if, that is to say, there had been only one ostensive definition for the word “red” used in our game. One could, on the other hand, imagine a language-game in which two words, say “red” and “reddish”, were explained by two ostensive definitions, the first shewing a dark red object, the second a light red one. Whether two such definitions || explanations were given or only one might depend on the natural reactions of the people using the language. We might find that a person to whom we give the ostensive definition, “This is called ‘red’” (pointing to one red object) thereupon fetches any red object of whatever shade of red on being ordered: “Bring me something red!” Another person might not do so, but bring objects of a certain range of shades only in the neighborhood of the shade pointed out to him in the explanation. We might say that this person “does not see what is in common between all the different shades of red”. But remember please that our only criterion for that is the behaviour we have described.