Let’s change tack here. Science doesn’t attempt to prove hypotheses, it attempts to disprove them.
So how do we test the “ghost” hypothesis? What specific predictions does it make that we can measure and check?
I mean, sort of.
Here's a quick rundown of the scientific method:
Step 1: Observe
Step 2: Question
Step 3: Make a hypothesis
Step 4: Experiment
Step 5: Analyze
Step 6: Draw a conclusion
Let's assume we have already observed something we can't explain (step 1), then we asked, "woah, what was that?" (step 2), and a friend said, "I think it was a ghost" (step 3). So we can skip ahead to step 4, building the experiment. And here's where it all goes to hell.
Experiments must define and measure variables. So if you are trying to explain something you saw, your experiment would focus on visible objects and phenomena. To make sure you are measuring
only ghosts, you would need to eliminate all non-ghost variables. Let me reaffirm that part: I said
eliminate, as in "remove from the data set completely." This is not the same thing as "explain it." It has to be removed from the experiment completely, or it will influence the results of your data.
First, since this was an observation, you need to eliminate anything that could affect cognition. Was it dark? Was it foggy, rainy, or windy? Do you wear eye correction, and if so, how old is your prescription? Are you taking any medication? How long had it been since you last ate something? Were you intoxicated? Were you well-rested? Were you disoriented or anxious? Any of these will affect your ability to see something and/or interpret what you saw--which is why so many ghost sightings happen late at night (when lighting is poor, it's been several hours since the observer last ate something, it's past the observer's bedtime) and they're stressed about running late or being lost. If you can't eliminate these variables from your experiment, it won't be measuring ghosts...it'll end up measuring your own vision and cognitive ability in that moment.
Then you do this again for environmental factors: was it windy, was the building drafty, were there curtains near an open window, etc. And then you do it again for social factors: do you own a pet, are you certain you were alone, was the area open to the public, was any wildlife in the area? And so on, and so on, until you have eliminated all non-ghost variables from your experiment....otherwise you aren't measuring ghosts. (It's no coincidence that so many ghost sightings happen in public areas like parks, graveyards, or roadsides where the observer can't guarantee they were alone, or in abandoned places where vagrants, scavengers, and teenagers are often found.)
I think it's the rigorous elimination of variables that folks are referring to when they say "disproving theories." You're not trying to discredit the observer, you are trying to refine the experiment.
Then there's confirmation bias, and I think that's the worst variable of all. All of the "paranormal investigators" you read about have already decided that the paranormal exists (there's a clue in their name, you see), so they are only interested in proving what they have already accepted as fact. This will shape the experiment in several ways, but it usually takes the form of cherry-picking variables and data to support the hypothesis. Because
it just has to be a ghost, right?
What else could it
possibly be? (except for that, or that, or that, or that, or that, or that, or...) And bias is almost impossible to detect and eliminate, and usually requires parallel tests and peer review.
Sorry for the long essay about How To Do A Science. I would love to see an actual scientific experiment involving ghosts, but so far, nobody has been able to design one.