Data are not those things of which our consciousness is earliest in time. At every period of life, after we have become capable of thought, some of our beliefs are obtained by inference, while others are not. A belief may pass from either of these classes into the other, and may therefore become, or cease to be, a belief giving a datum. When, in what follows, I speak of data, I do not mean the things of which we feel sure before scientific study begins, but the things which, when a science is well advanced, appear as affording grounds for other parts of the science, without themselves being believed on any ground except observation. I assume, that is to say, a trained observer, with an analytic attention, knowing the sort of thing to look for, and the sort of thing that will be important. What he observes is, at the stage of science which he has reached, a datum for his science. It is just as sophisticated and elaborate as the theories which he bases upon it, since only trained habits and much practice enable a man to make the kind of observation that will be scientifically illuminating. Nevertheless, when once it has been observed, belief in it is not based on inference and reasoning, but merely upon its having been seen. In this way its logical status differs from that of the theories which are proved by its means. In any science other than psychology the datum is primarily a perception, in which only the sensational core is ultimately and theoretically a datum, though some such accretions as turn the sensation into a perception are practically unavoidable. But if we postulate an ideal observer, he will be able to isolate the sensation, and treat this alone as datum. There is, therefore, an important sense in which we may say that, if we analyse as much as we ought, our data, outside psychology, consist of sensations, which include within themselves certain spatial and temporal relations.