system of science only so long as they are supported by, or at least not contradicted by, other hypotheses. We therefore always reserve the right to make protocol statements subject to correction, and such corrections, quite often indeed, do occur when we eliminate certain protocol statements and declare that they must have been the result of some error. 

Even in the case of statements which we ourselves have put forward we do not in principle exclude the possibility of error. We grant that our mind at the moment the judgment was made may have been wholly confused, and that an experience which we now say we had two minutes ago may upon later examination be found to have been an hallucination, or even one that never took place at all.

Thus it is clear that on this view of protocol statements they do not provide one who is in search of a firm basis of knowledge with anything of the sort. On the contrary, the actual result is that one ends by abandoning the original distinction between protocol and other statements as meaningless. Thus we come to understand how people come to think that any statements of science can be selected at will and called "protocol statements," and that it is simply a question of convenience which are chosen. 

But can we admit this? Are there really only reasons of convenience? It is not rather a matter of where the particular statements come from, what is their origin, their history? In general, what is meant here by convenience? What is the end that one pursues in making and selecting statements? 

The end can be no other than that of science itself, namely, that of affording a true description of the facts. For us it is self-evident that the problem of the basis of knowledge is nothing other than the question of the criterion of truth. Surely the reason for bringing in the term "protocol statement" in the first place was that it should serve to mark out certain statements by the truth of which the truth of all other statements comes to be measured, as by a measuring rod. But according to the viewpoint just described this measuring rod would have shown itself to be as relative as, say, all the measuring rods of physics. And it is this view with its consequences that has been commended as the banishing of the last remnant of "absolutism" from philosophy;' 

But what then remains at all as a criterion of truth? Since the 
proposal is not that all scientific assertions must accord with certain 
definite protocol statements, but rather that all statements shall accord with one another, with the result that every single one is considered