What do we do if an observation disagrees with a hypothesis?

If an experiment is carried out and the result disagrees with that predicted by the prevailing hypothesis, there can be several reasons.

1. The hypothetical model is fundamentally wrong, eg. Earth-centred Universe.
2. The model is approximate and can be extended to improve the accuracy in the new situation. The old is an approximation to the new. eg. Special relativity.
3. The experimental conditions are not fully, or incorrectly known. In which case the hypothesis can be correct and will give agreement if all the surrounding conditions are specified. eg. Uranus.

According to Logical Empiricism, as expressed by Popper, a single disagreement between prediction and observation is enough to throw out the hypothesis. This would have caused Newton’s mechanics to be discarded in situation (3) above which would have been clearly inappropriate. Investigating the reason for the discrepancy to see where it arises is what is generally done.

If a succession of experiments give results which disagree with hypothesis then the pressure to seek a new hypothesis grows. It becomes apparently less likely that it is the experiments which are lacking as the hypothesis is either incomplete or wrong. For a number of reasons, some good and some bad, the pressure has to mount to a certain level before the prevailing hypothesis is abandoned. When that happens, there is a rapid shift to a new, and possibly very different, hypothesis which give more accurate agreement and has the virtue of parsimony