Parsons on Reppert

What is a supernatural hypothesis? I will limit attention to hypotheses that postulate the existence of supernatural persons or powers. Instances of supernatural persons would include gods, ghosts, demons, angels, spirits (like Ariel), and souls. Instances of supernatural powers would include mana, qi, astrological influences, telekinesis, ESP, and the creative power attributed to God in Genesis where God says “Let there be…” and there is. But what is it for a person or power to be supernatural? By “supernatural” I mean “capable of existing or operating independently of, unrestrained by, or even in violation of, the laws of nature.”
So, are supernatural hypotheses as characterized above testable? What do we mean by a “testable?” I mean “testable” in the rather strict sense of “confirmable or disconfirmable by rigorous experiment, experiment of the sort typically employed to evaluate hypotheses in the physical and biological sciences.”
http://secularoutpost.infidels.org/2011/09/testing-supernatural-hypotheses.html

I find this deeply confused. Although there are times when a theory or hypothesis coincides with an existential proposition, in many situations that's not the case. For instance, there's a elementary difference between confirming the existence of ball lightning, and testing a hypothesis regarding the nature of ball lightning, i.e. how it's generated. 

Likewise, while it may (at some point) be feasible to reproduce ball lightning under rigorous, laboratory conditions, you don't need experimental evidence to confirm the existence of ball lightning. Anecdotal or testimonial evidence should suffice. 

After all, if ball lightening exists, it normally exists in nature, not in the lab. Therefore, observing ball lightening in nature would be a perfectly legitimate method of confirming its existence. That's where we'd expect to find it–assuming it exists. 

Of course, we'd still need to apply the usual criteria for testimonial evidence. 

Now maybe Parsons would say supernatural entities are disanalogous. But that's a different argument.