Hypotheses and Hypothesis Testing

Abstract

In this chapter I argue that the least common denominator for all sciences, including social sciences and humanities, is that hypotheses are formulated and tested. This is meaningful only if one is prepared to change one mind after testing, to admit that even one favourite hypothesis was wrong. This state of mind is a crucial part of the scientific attitude.

Testing hypotheses against observations require auxiliary assumptions. They can also be tested, of course, although in a particular case of hypothesis testing they are taken for granted at the beginning.

The result of the test is either that the predictions and observation reports are compatible, or that they conflict. In the former case one may be justified to say that one’s hypothesis is supported. In the latter case one must reconsider something; one must reject the hypothesis, an auxiliary assumption, or the observation report. If one decides that the hypothesis is to be rejected, one has falsified it. It is thus clear that falsification of a hypothesis is no automatic inference from the test result; it is the result of a considered judgement, all aspects considered.

In this chapter I also discuss statistical testing of hypotheses, concluding that the logical structure is the same as in the general case.

We and other animals notice what goes on around us. This helps us by suggesting what we might expect and event how to prevent it, and thus fosters survival. However, the expedient works only imperfectly. There are surprises, and they are unsettling. How can we tell when we are right? We are faced with the problem of error.

W. V. O. Quine