Why do old theories take priority over new ones?

Question

Given two scientific theories that both agree with experiment, it seems that unless the new one is much simpler, the older one is accepted. Why is this so?

Examples:

• Special Relativity wasn't accepted until experiments contradicting Newtonian Mechanics where avaible, despite electromagnetism being more in support of special relativity.
  • Likewise for general relativity, which would have been selected by occam's razor due to not having special reference frames.
• Our current theory is Standard Model+General Relativity. That "+" is really complicated, as it sort arbitrarily transitions from the Standard Model to General Relativity. On the other hand, string theories apply at both large and small scales. Of course there are many string theories, but one could be selected that is consistent with experiment.

One could say "if you don't make predictions, you aren't falsifiable", but the theories above due make predictions; that just so happen to coincide with the predictions of the older theory for the most part. So both the new theory and old theory are equally falsifiable.

Why are old theories given priority over new theories?

Answer 1

The point of science is not just the generation and selection of theories, it is giving explanations of observations. Kuhn calls this work of applications 'normal science' and contends that it really is most of science, and the most important part of the whole body of science. It is, after all, upon 'normal science' that we base engineering disciplines like bridge building, automotive design and pharmacology.

From a Kuhnian point of view, existing accepted theories are not independent, they are woven together into a paradigm. So when you propose replacing one with a new theory, you are challenging not just the theory being replaced, but its contributions to all of the other applications in which it has been combined with other theories to produce explanations.

All of those explanations seem to work, or they would not have been accepted. Unlike the Popperian model you allude to that chooses to focus primarily upon falsification, a more realistic model (such as Kuhn's) considers the overall faith in an existing theory and its track record of success, as well. So in order to choose a different basic explanation, you need some excuse for why those existing explanations worked, or you need for the new theory to be clearly enough better to require re-deriving all of the explanations that involved the old one, despite the widespread faith in and dependence upon the existing work.

This means that there is a good reason for science in general to be conservative, and to maintain older theories not just until there is adequate evidence against them, but until that evidence overwhelms the success they have had, and they are truly no longer tenable.

Kuhn's historical work itself focusses on what happens at the point where old theories do finally break down, and considers the results disruptive enough that he characterizes these as 'revolutions', with all the metaphorical blood shedding that implies.

Answer 2

There's a lot of pragmatic reasons. One overarching reality is that the older theories are better known and better understood than the new ones. If we switched to new-but-equivalent theories all the time, we'd end up with dozens of shifts in notation and whatnot slowing the scientific community down.

The other good reason is that the new theories tend to be far more complicated than the old ones. You mention electromagnetics makes special relativity seems natural, and thats true in hindsight. Find me one physicist from the 1800s that would give up simultaneity in return for SR. SR makes some really counter-intuitive demands which weren't really worth the cost until experimental evidence showed that the existing theory was insufficient.

As for new theories of everything, like string theory, there's currently no way to pick between them. You say one can simply be selected, but on what merit? If I may use circular logic, old theories prevail, so this theory may be around for quite a long time. Did we actually gain any clarity by replacing the existing models?