I imagine that if error never occurred in inferences of this practical sort the human race would have gone on quite contented with them. However that may be, errors _do_ occur and the endeavor to control inference as to consequences (so as to reduce their likelihood of error) leads to propositions where the knowledge-object of the perceived thing is not something to be done, but the cause which produced it. The mother finds her baby peacefully sleeping and says the baby didn't _make_ the noise. She investigates and decides a swinging door _made_ it. Instead of inferring a consequence, she infers a cause. If she had identified the noise in the first place, she would have concluded that the hinges needed oiling. Now where does the argument stand? The proper control of inference in specific cases is found (_a_) to lie in the proper identification of the datum. If the perception is of a certain kind, the inference takes place as a matter of course; or else inference can be suspended until more adequate data are found, and thus error is avoided even if truth be not found. Furthermore (_b_) it is discovered that the most effective way of identifying datum (and securing adequate data) is by inference to its cause. The mother stops short with the baby and the door as causes. But the same motives which made her transfer her inference from consequences to conditions are the motives which lead others to inferring from sounds to vibrations of air. Hence our scientific propositions about sensory data. They are not, as such, about things to do, but about things which have been done, have happened--"facts." But they have reference, nevertheless, to inferences regarding consequences to be effected. They are the means of securing data which will prevent errors which would otherwise occur, and which facilitate an entirely new crop of inferences as to possibilities--means and ends--of action. That scientific men should be conscious of this reference or even interested in it is not at all necessary, for I am talking about the logic of propositions, not about biography nor psychology. If I reverted to psychology, it would be to point out that there is no reason in the world why the practical activity of some men should not be predominantly directed into the pursuits connected with discovery. The extent in which they actually are so directed depends upon social conditions. III We are brought to a consideration of the notion of "primitive" sense data. It was long customary to treat the attempt to define true knowledge in terms derived from sense data as a confusion of psychology--or the history of the growth of knowledge--with logic, the theory of the character of knowledge as knowledge. As matter of fact, there _is_ confusion, but in the opposite direction. The attempt involved a confusion of logic with psychology--that is, it treated a phase of the technique of inference as if it were a natural history of the growth of ideas and beliefs.