saying that they are "absolutely certain"? And in what sense may one describe them as the ultimate ground of all knowledge? 

Let us consider the second question first. If we imagine that I at once recorded every observation -- and it is in principle indifferent whether this is done on paper or in memory -- and then began from that point the construction of science, I should have before me genuine "protocol statements" which stood temporally at the beginning of knowledge. From them would gradually arise the rest of the statements of science, by means of the process called "induction," which consists in nothing else than that I am stimulated or induced by the protocol statements to establish tentative generalizations (hypotheses), from which those first statements, but also an endless number of others, follow logically. If now these others express the same as is expressed by later observation statements that are obtained under quite definite conditions which are exactly specifiable beforehand, then the hypotheses are considered to be confirmed so long as no observation statements appear that stand in contradiction to the statements derived from the hypotheses and thus to the hypotheses themselves. So long as this does not occur we believe ourselves to have hit correctly upon a law of nature. Induction is thus nothing but methodically conducted guessing, a psychological, biological process whose conduct has certainly nothing to do with "logic." 

In this way the actual procedure of science is described schematically. It is evident what role is played in it by the statements concerning what is "immediately perceived." They are not identical with those written down or memorized, with what can correctly be called "protocol statements," but they are the occasions of their formation. The protocol statements observed in a book or memory are, as we acknowledged long ago, so far as their validity goes, doubtless to be compared to hypotheses. For, when we have such a statement before us, it is a mere assumption that it is true, that it agrees with the observation statements that give rise to it. (Indeed it may have been occasioned by no observation statements, but derived from some game or other.) What I call an observation statement cannot be identical with a genuine protocol statement, if only because in a certain sense it cannot be written down at all -? a point which we shall presently discuss. 

Thus in the schema of the building up of knowledge that I have described, the part played by observation statements is first that of standing temporally at the beginning of the whole process, stimulating it and setting it going. How much of their content enters into knowledge