Science is inherently atheistic

[polymath257]

But, as always, you can start out with observed behavior and *hypothesize* the laws they operate under, test those laws, etc.

There is simply no reason the scientific method could not be applied to ghosts, goblins, or gods, in theory.

In practice, of course, it fails to study them because of the utter non-existence of them all.

I agree, you could, but you can guarantee "control."  If something is not bound by natural laws, we can't measure it with assumptions of natural  laws. When anything is measured scientifically, it is a measure of relationship.  The starting point is always "null" or "no relationship", and then we try to determine the probability of relationship by controlling things that would skew those results, that way we are only testing the specific relationship in the research between two or more subjects.  If we find nothing, we conclude the null was correct, but we also give a numerical value on the probability that "null" was the right result.  If we see a relationship, we reject the null, and determine the extent of that relationship within those parameters.  That's the beauty of the scientific method.  Once we finish a study, we now the basis for more study by bringing what we know and resetting parameters to advance knowledge of those subjects.  This is also why peer-review is important.  You may (or may not) have done garbage research by not controlling the parameters and someone else can call you on it.  This isn't necessarily a bad thing, because we all make mistakes and learn.  Next time you can go back and do it with the necessary corrections.

Now if you say a vampire exists that can magically turn into a bat, and you believe you saw him in a cave in a mountain, how the heck are you going to set up parameters to test such a thing?  You could try, but I can just imagine there would be at least a few dozen difficulties, and even more so if such a creature actually existed (safety issues, making sure it doesn't walk through a wall, etc...)  Even finding in could potentially be a pain, and cooperation even more so.

Again, I really don't see anything more difficult than dealing with any other sort of dangerous creature that we don't understand. For those creatures in the wild, we seldom are capable of doing struct controls. All we can do, especially at first, is describe, catalog behaviors, and make a few hypotheses. Through more observation, some of the hypotheses may be shown wrong.

We do NOT have to measure under an assumption of natural laws for the creature itself. But, for example, if we can photograph it, that implies some sort of interaction with light. If we can hear it, that implies some sort of interaction with pressure waves in the air. Even if we don't understand the nature of those interactions, that is, at least, an 'in' to understanding what the creature is and how it functions. If we can record a transformation into a bat, we might be able to do a frame by frame analysis to see whether mass is conserved and, if not, what happens to is.

All said, this would lead to some *amazing* insights into physics.

The point is that natural laws are *descriptive*, not *prescriptive*. We find out natural laws by looking at phenomena and finding descriptions of what we observe, looking for patterns all the time. A natural law is a type of general description of patterns that we see. To say there is no 'natural law' operative simply means there are no patterns that we can discern and/or no descriptions that are useful for understanding.