f we keep an eye on the actual enterprise of knowing, it is clear that only sense-data can supply evidential subject- matter} ideas of what is not presented in sense interpret evi- dence, but they cannot constitute it. The whole history of science shows, however, that material directly and originally observed does not furnish good evidential material} as we saw, it was the essential mistake of ancient science to suppose that we can base inference upon observed objects without an artificial prior analytic resolution. Hence there is need of a distinctive type of experimental operations which detach some qualities of the object} these form sense-data in the technical meaning of the word.