ABSTRACT
In the previous two chapters we have seen how a corroboration-based account of scientifi c method can stand without appeal to induction, and that corroboration values should be based on intersubjective probabilities formed by suitable groups and procedures. The discussion has emphasised the signifi cance of testing both ideas and people, and has suggested that these activities are at the heart of the scientifi c enterprise. Just as a theory will not make it into a textbook (except perhaps as an example of a mistake) without having stood up under fi re, so a scientist will not make it into the laboratory without having jumped through several professional hoops. Of course, accidents do happen. We have some bad theories, and some bad scientists. But we do our best as a collective-and with a margin of success, it would seem-to minimize these. (Although it should be noted that it can sometimes serve collective ends best to admit people who violate the canons of individual rationality; I discuss this idea in Chapter 6.)
There are still several outstanding problems for a corroboration-based account of scientifi c method, however. One especially noteworthy example, which will be discussed in depth in the next chapter, is Duhem’s thesis that a hypothesis cannot be tested in isolation. In the present chapter, we will discuss three further problems. First, when exactly should an observation statement be understood to be worthy of counting as an e in the corroboration function? We have already seen that Popper thought that not just any old observation statement will do, but instead that we should only use the results of tests (at least to positively corroborate theories), and then only the severest tests at that.1 As we will see, however, this view is not so easy to sustain as it may fi rst appear. One key problem we will encounter is what I have elsewhere called ‘The Problem of the Big Test’: that the severest test of any hypothesis is to perform all possible tests of said hypothesis (when ‘possible’ is suitably interpreted).
