All science depends upon the record of the past, and a record other than that in the memory is plainly something which cannot be verified by direct observation.
…Verification is the watchword of Positivism. But it is easy to see that a proposition is no more verifiable by direct observation for being such as we can suppose (by a recognized falsification) to be observed unless it is also such as really can be observed. This maxim, therefore, must refer to really possible observations, not such as are supposably possible, for the proof they give leads to that or nothing.
…It is not a question capable of being decided by direct observation, what is and what is not direct observation. The logical rule, therefore, which is the whole basis of Positivism appears to me to be entirely false.
~C.S. Peirce, Notes on Positivism, in Charles S. Peirce: Selected Writings (Values in a Universe of Chance)
Peirce’s refutation is notable in several ways. First, his claim is only that Positivism “appears” false, a modesty born from his doctrine of fallibilism. Second, his refutation is a precursor of the famous Duhem-Quine thesis, which states that a scientific theory is not refuted on the basis of a single negative experiment because it’s always possible that some mistake was made in the measurement process and, in addition, the theory can be “rescued” by adding auxiliary hypotheses post-hoc. Similarly, Peirce’s point seems to be that Positivism fails to live up to its own standards because if we suppose the gold standard for knowledge is “direct observation”, how can we be sure that our observation was really and not seemingly direct? To verify that our observation was direct, we need a direct observation that our observation was direct. Thus, Positivism will either lead to an infinite regress or bottom out at a direct observation that we haven’t directly observed is a direct observation.