Let us suppose that I have certain visual and tactual sensations which produce in my mind an association of ideas. For example, when sitting at the table in my study, I have those visual sensations which I call seeing the table and the tactual sensations which I call touching or feeling the table. And an association is set up such that when I have a visual sensation of this kind, a tactual sensation is present as a possibility. Conversely, when I have only a tactual sensation, as when the room is completely dark, a visual sensation is there as a possibility. Further, when I leave the room and later re-enter it, I have similar sensations.
Hence an association is formed in my mind of such a kind that when I am out of the room, I am firmly persuaded that, if I were at any moment to re-enter it, I should or could have similar sensations. Further, as these possible sensations form a group, and as moreover the group is found to enter into various causal relations, I inevitably think of the permanent possibilities of sensations as an abiding physical object. Actual sensations are transient and fugitive. But the possibilities of sensation, associated as a group, remain. The definition of matter as a permanent possibility of sensation is, however, ambiguous. For it easily suggests the idea of a permanent ground of possible sensations, a ground which is itself unknowable.
As for science, this would become a study of the relations between my sensations. But is it credible that if an anatomist looks at a human brain, the object of his examination is simply a set of his own subjective states, actual and possible? In short, the logical result of defining physical objects in terms of sensations is solipsism. And nobody really believes that solipsism is true.