But there is yet a third thing which may be meant, and which I am willing to admit may be true. It may be said: "I believe many generalisations of the following kind. I believe that when I have a perception A, some other person has generally had a feeling X; I believe that the existence of the feeling X is generally followed, in the same person, by that of the feeling Y; and I believe also that when another person has the feeling Y, I generally have the perception B. I believe all this." And it must, I think, be admitted that we do believe generalisations of this kind, and generalisations in which there are not merely two steps between A and B, but a great number of steps. "But then," it may be said, "my belief in this generalisation causes me, when I observe my perception A, to expect that I shall have the perception B; and such expectations, I observe, are constantly realised." And this also, I think, must be admitted to be true. "But, finally," it may be said, "beliefs which produce expectations which are constantly realised are generally true. And hence the fact that these beliefs of mine about the connection of feelings in other persons with perceptions of my own do lead to expectations which are realised, gives me reason to suppose that these generalisations are true and hence that other persons do have particular kinds of feelings." And I am willing to admit that this also is true. I am willing to admit that true predictions can, as a rule, only be produced by true beliefs. The generalisation that this is so, is, indeed, one which can only be justified by the observations of beliefs, which are, in some way, independently proved to be true; and hence, if it is to be justified, without assuming the existence of anything other than my own perceptions, thoughts, and feelings, it can only be justified by my observation that beliefs with regard to the manner in which _these_ succeed one another generally lead to true predictions. Whether the observation of such beliefs _alone_ could give sufficient reason for it, is, I think, doubtful; but I am willing to admit that it may be so. One thing, however, is, I think, quite plain: namely, that this generalisation "Beliefs which lead to true predictions are generally true" cannot be true, _unless_ some other of the "contents" which I observe, beside my own perceptions, thoughts, and feelings, do exist. That is to say, in giving a reason for supposing the existence of other people, this generalisation also gives a reason for the very theory which I am advocating, namely, that some of those data which I have called "sense-contents" do exist. It does this, because it is quite certain that beliefs in generalisations about the existence of sense-contents _can_ (and do) constantly lead to true predictions. The belief that when I have observed a fire of a certain size in my grate, something similar to what I have observed will continue to exist for