The notion of "data" is familiar throughout science, and is usually treated by men of science as though it were perfectly clear. Psychologists, on the other hand, find great difficulty in the conception. "Data" are naturally defined in terms of theory of knowledge: they are those propositions of which the truth is known without demonstration, so that they may be used as premisses in proving other propositions. Further, when a proposition which is a datum asserts the existence of something, we say that the something is a datum, as well as the proposition asserting its existence. Thus those objects of whose existence we become certain through perception are said to be data. There is some difficulty in connecting this epistemological definition of "data" with our psychological analysis of knowledge; but until such a connection has been effected, we have no right to use the conception "data." It is clear, in the first place, that there can be no datum apart from a belief. A sensation which merely comes and goes is not a datum; it only becomes a datum when it is remembered. Similarly, in perception, we do not have a datum unless we have a JUDGMENT of perception. In the sense in which objects (as opposed to propositions) are data, it would seem natural to say that those objects of which we are conscious are data. But consciousness, as we have seen, is a complex notion, involving beliefs, as well as mnemic phenomena such as are required for perception and memory. It follows that no datum is theoretically indubitable, since no belief is infallible; it follows also that every datum has a greater or less degree of vagueness, since there is always some vagueness in memory and the meaning of images.