protected from any temptation to turn logical specification into metaphysical atomism. The need for the protection is sufficiently great to justify spending some energy in pointing out that the brute objective facts of scientific discovery are discovered facts, discovered by physical manipulations which detach them from their ordinary setting. We have stated that, strictly speaking, data (as the immediate considerations from which controlled inference proceeds) are not objects but means, instrumentalities, of knowledge: things by which we know rather than things known. It is by the color stain that we know a cellular structure; it is by marks on a page that we know what some man believes; it is by the height of the barometer that we know the probability of rain; it is by the scratches on the rock that we know that ice was once there; it is by qualities detected in chemical and microscopic examination that we know that a thing is human blood and not paint. Just what the realist asserts about so-called mental states of sensations, images, and ideas, namely, that they are not the subject-matter of knowledge but its agencies, holds of the chairs and tables to which he appeals in support of his doctrine of an immediate cognitive presentation, apart from any problem and any reflection. And there is very solid ground for instituting the comparison: the sensations, images, etc., of the idealist are nothing but the chairs, tables, etc., of the realist in their ultimate irreducible qualities.[6] The problem in which the realist appeals to the immediate apprehension of the table is the epistemological problem, and he appeals to the table not as an object of knowledge (as he thinks he does), but as evidence, as a means of knowing his conclusion--his real _object_ of knowledge. He has only to examine his own evidence to see that it is evidence, and hence a term in a reflective inquiry, while the nature of knowledge is the _object_ of his knowledge.