Do the premises of a theory have to be empirical?

Cinitiator

Since there's no board dedicated to the methodology of science, I thought that this board would be the best fit. Here's my question:
Do the premises and elements of a theory have to be empirical? Or do they not have to be, as long as the theory is consistent with the observed data? By consistent I mean that the conclusion seems to be true, but the premises aren't empirical.

I'm asking because from what I've seen, cognitive psychology relies on mental "modules" to explain empirical data. These modules can't actually be observed empirically, but the results put forward by the theories which rely on these non-empirical concepts seem to be more or less consistent with the evidence up to date.

mfb

How can premises be empirical? They can be inspired by experiments, of course.

One example:
"The speed of light is constant for all observers". How can you ever observe this? You cannot measure the speed of light in any possible reference frame. You can measure it in many, and assume that it is true for all (and check it with more measurements).

phinds

No, they don't HAVE to be empirical, but it wouldn't make a lot of sense for them NOT to be since the FIRST thing you have to do with a theory is test it against reality.

I mean, the first step in a theory is a guess and you could just make some wild guess, but it's not likely to reflect reality if it isn't in some way based on reality and as soon as you take the next step and COMPARE it to reality, most times, you won't have a theory any more, you're just back to having a guess that's wrong.

EDIT: look up the on-line video of Feynman talking about the scientific method.

Cinitiator

Thanks for the video suggestion, I will watch it.

A theory can explain the data and yet not be based in the real world, though. It can be logically equivalent for a given moment with our data. Cognitive and evolutionary psychology is a good example of this.

A.T.

No. You can postulate whatever you want. If it makes predictions which are testable and agree with experiment it is a good physical theory, no matter how you came up with that stuff.

However:
1) Only the postulates that affect the quantitative predictions are physics.
2) A simpler theory, with less postulates, making the same predictions, makes a complicated one obsolete.

phinds

And likewise a more complicated one that explains more of the real world can make a less complicted one obsolete. E.G. GR vs Newton's gravity (may not the BEST example, since Newton isn't exactly obsolete, just known to be very limited).