The chief source of error in ordinary inference is an unrecognized complexity of data. Perception which is not experimentally controlled fails to present sufficiently wide data to secure differentia of possible inferences, and it fails to present, even in what is given, lines of cleavage which are important for proper inference. This is only an elaborate way of saying what scientific inquiry has made clear, that, for purposes of inference as to conditions of production of what is present, _ordinary_ sense perception is too narrow, too confused, too vivid as to some _quales_ and too blurred as to some others. Let us confine our attention for the moment to confusion. It has often been pointed out that sense qualities being just what they are, it is illegitimate to introduce such notions as obscurity or confusion into them: a slightly illuminated color is just as irretrievably what it is, as clearly itself, as an object in the broad glare of noonday. But the case stands otherwise when the _quale_ is taken as a datum for inference. It is not so easy to identify a perceived object _for purposes of inference_ in the dusk as in bright light. From the standpoint of an inference to be effected, the confusion is the same as an unjustifiable simplification. This over-simplification has the effect of making the _quale_, as a term of inference, ambiguous. To infer from it is to subject ourselves to the danger of all fallacies of ambiguity which are expounded in the textbooks. The remedy is clearly the resolution, by experimental means, of what seems to be a simple datum into its "elements." This is a case of analysis; it differs from other modes of analysis only in the subject-matter upon which it is directed, viz., something which had been previously accepted as a simple whole. The result of this analysis is the existence as objects of perception of isolated qualities like the colors of the spectrum scientifically determined, the tones of the scale in all their varying intensities, etc., in short, the "sensations" or sense qualities of contemporary psychology textbooks or the "simple ideas" of sensation of Locke or the "objects of sense" of Russell. They are the material of sense perception discriminated for the purpose of better inferences.