The term "sense-data" is ambiguous; and therefore I think I had better begin by trying to explain what the class of entities is whose status I propose to discuss.
There are several different classes of mental events, all of which, owing to their intrinsic resemblance to one another in certain respects, may, in a wide sense, be called "sensory experiences," although only some among them would usually be called "sensations." There are (1) those events, happening in our minds while we are awake, which consist in the experiencing of one of those entities, which are usually called "images," in the narrowest sense of the term. Everybody distinguishes these events from sensations proper; and yet everybody admits that "images" intrinsically resemble the entities which are experienced in sensations proper in some very important respect. There are (2) the sensory experiences we have in dreams, some of which would certainly be said to be experiences of images, while others might be said to be sensations. There are (3) hallucinations, and certain classes of illusory sensory experiences. There are (4) those experiences, which used to be called the having of "after-images," but which psychologists now say ought rather to be called "after-sensations." And there are, finally, (5) that class of sensory experiences, which are immensely commoner than any of the above, and which may be called _sensations proper_, if we agree to use this term in such a way as to exclude experiences of my first four sorts.
Every event, of any one of these five classes, consists in the fact that an entity, of some kind or other, _is experienced._ The entity which is experienced may be of many different kinds; it may, for instance, be a patch of colour, or a sound, or a smell, or a taste, etc; or it may be an image of a patch of colour, an image of a sound, an image of a smell, an image of a taste, etc. But, whatever be its nature, the entity which _is_ experienced must in all cases be distinguished from the fact or event which consists in its being experienced; since by saying that it is experienced we mean that it has a relation of a certain kind to something else. We can, therefore, speak not only of _experiences_ of these five kinds, but also of the entities which _are experienced in_ experiences of these kinds; and the entity which is experienced _in_ such an experience is never identical with the experience which consists in its being experienced. But we can speak not only of the entities which _are_ experienced in experiences of this kind, but also of _the sort_ of entities which are experienced in experiences of this kind; and these two classes may again be different. For a patch of colour, even if it were not actually experienced, would be an entity _of the same sort_ as some which are experienced in experiences of this kind: and there is no contradiction in supposing that there are patches of colour, which yet are not experienced; since by calling a thing a patch of colour we merely make a statement about its intrinsic quality, and in no way assert that it has to anything else any of the relations which may be meant by saying that it is experienced. In speaking, therefore, of _the sort of_ entities which are experienced in experiences of the five kinds I have mentioned, we do not necessarily confine ourselves to those which actually _are_ experienced in some such experience: we leave it an open question whether the two classes are identical or not. And the class of entities, whose status I wish to discuss, consists precisely of all those, whether experienced or not, which are _of the same sort_ as those which are experienced in experiences of these five kinds.