"No linguistic behaviour can logically determine its own sequel, since no past time can logically determine the future. A man may 'follow a rule' as we do, and yet, at some future time, diverge from us, insisting all the while that what he is doing is the same as what he has always done. We cannot establish, once and for all, and with no possibility of doubt, that another really does understand a word as we do -- whether that word be 'he' or "I". The only point is that, if he begins to make mistakes in his use of "I", this shows either that he has ceased to understand the word (and there are psychotics of whom this is true) or else that he always understood it wrongly (a most disturbing possibility). The problem of distinguishing between those alternatives is acute: but it is a general problem in the theory of meaning, and has nothing special to do with 'I'."