I recently had the opportunity to be interviewed by some students taking a course at Champlain College in Burlington, VT called Scientific Revolutions: X-files edition. I had met one of the students at the Ghostacular Paracon there last Fall. The course uses the paranormal as a focal point for discussions about the role of science in a broader context. Inspired by this discussion, I decided to take this opportunity to reflect on the role of science in paranormal investigation.
Since I have, for several years, managed a paranormal investigation team called Scientific Paranormal, I am often asked what I think makes us “scientific.” Let me first make a disclaimer that I did not name the group. I joined a group that was already working and eventually came to lead it. If I were starting my own group, I probably would not use the word “scientific” because it tends to raise a lot of hackles. Nevertheless, we do try to approach our investigations as scientifically as the case allows.
Also, we do not do experimental science. Although the work done by experimental parapsychologists in lab settings is great, Paul and I are out slogging through swamps and crawling around attics and basements. It is an uncontrolled environment. This means that we cannot formally test using known statistical distributions or the like. As we have discussed here before, our primary goal is to help the witnesses to paranormal phenomena to integrate their experiences. The home of a frightened family is not a lab for doing scientific experiments. Still, part of assisting the clients is coming to the best understanding that we can of what actually occurred. This involves analyzing the data, deciding what it can tell us and what it can’t tell us. This is where taking a scientific approach to the investigation can be our best guide.
The null hypothesis:
Contrary to popular use, there is no one thing called “THE” scientific method. Different sciences have to use different approaches, though there are common elements between them. One commonality is that a scientific approach works by falsifying hypotheses.
My own field is statistics, which offers an approach that has both benefits and drawbacks for the study of the paranormal. I approach a paranormal investigation in the manner we use for statistical hypothesis testing. A key concept in statistical work is that data can never prove anything. It can only disprove things (and often not even that, as we will discuss later).
A formal statistical hypothesis test begins with what we call the null hypothesis, or just the null for short. The null is what we will assume to be true unless we can disprove it using data as evidence. This means that the null hypothesis must be one that is falsifiable using data as evidence against it. So, we have to choose the proper null. We can’t go into an investigation using the null hypothesis that the location is haunted. This would not be a scientific approach, since this hypothesis cannot be falsified with data. It would mean that if we go in and observe nothing strange and capture no evidence at all, we maintain our assumption that it is haunted. That would not be proof of anything.
So, if we want to find evidence for the paranormal, we have to begin by assuming that there is no paranormal activity occurring. Then, we attempt to disprove this assumption.
The role of data
Having established the proper null, we then set out to collect data against it, and we ask ourselves how likely we are to observe this data if in fact our null hypothesis is true. This is the point of the investigation and much of the equipment.
If we observe data that is highly unlikely to occur under the null hypothesis, then we reject the null. If we do not observe such data, then we maintain the null hypothesis. In statistics, we say we fail to reject the null. Does this mean we have proved the null to be true? Not necessarily. It just means that we don’t have enough evidence to say otherwise. We reject the null or we fail to reject the null. That is all. That is the only decision we ever make. In a formal sense, the data does not tell us anything about any other hypothesis.
One key concept in statistical work (that is what I like about statistics) is that, whatever decision we make, we must always recognize that we could be wrong. If we reject the null based on evidence we observe, we could be wrong. And if we fail to reject the null based on the evidence, we could also be wrong. This is a scientific attitude, and it is also crucial to keep in mind when dealing with witnesses. Ultimately, whatever data we collect, we don’t know what they experienced.
Bringing it together
The only scientific way to investigate the paranormal as far as I can tell is to start with the null hypothesis that there is no paranormal activity occurring, and anything the witnesses are experiencing has normal (i.e. not paranormal) explanations. Then gather data to try to “disprove” that hypothesis, or data that seems highly unlikely to be witnessed under this assumption. If we find such evidence, then we reject the null. If not, we fail to reject it. This, of course, is not proof that there is no paranormal activity, just that we did not capture any evidence of it.
We also often have particular null hypotheses for specific claims, such as that the knocking on the walls is caused by the hot water pipes, or the screams from the forest at night are a barred owl. If we hear a roar from the forest that sounds like King Kong, we reject the barred owl hypothesis (or reject what we believe about barred owls).
Extraordinary claims require extraordinary evidence
A common rallying cry for Skeptics is the phrase made popular by Carl Sagan that extraordinary claims require extraordinary evidence. Let me say that I am in complete agreement with this. If you are going to tell me that your house is inhabited by interdimensional beings or that there is an 8-foot primate unknown to science living in your backyard… you need to back that up a bit. But we need to make a decision as to what constitutes “extraordinary evidence.”
In formal statistical tests, the strength of the evidence is measured by the p-value, which is the probability of observing the data we observe if in fact the null is true (for stats nerds, it is the probability of a type-I error). The lower the p-value, the stronger the evidence against the null. A common convention in statistics is to reject the null if the probability of observing the data under the null hypothesis is less than 0.05 (5%), or 1 in 20. But if we want “extraordinary evidence,” maybe we can require a p-value of .01 or even 0.001 (1 in 1000). Of course, in formal statistical (or experimental) work we are able to put numerical probabilities on these things. In a field investigation, we can’t. But the same idea can be applied.
For instance, in one of our cases described in our book, we captured one of our camera tripods sliding on the floor on its own when no one was in the building. Under the null hypothesis of no paranormal activity, how likely is that to happen? Seems unlikely to me. Is it possible? I guess anything is possible. A localized earthquake maybe? Or a pipe burst under the floor directly below where our tripod was set up? We could always come up with some story. But how unlikely is it? If I set up my tripod and filmed it for an hour somewhere where I knew there is no paranormal activity (impossible to prove in itself), how many times do we think it would move on its own? Less than 1 time in 20? Less than 1 in 1000 times?
The point is that if the probability is low enough, we must reject the null hypothesis. That is the scientific approach. In my opinion, the error that Skeptics often make is that they will never reject the null under any circumstances. This is unscientific. Just being able to find a possible natural explanation does not equate to “solving” a case. It gives us a null hypothesis to work with, but that is only the beginning of the investigative process. Very few cases can really be “solved.”
This raises the other issue with field evidence, which we discuss at length in our book: field evidence always exists in a context. What is convincing to someone who was there and knows the conditions under which the evidence was gathered will not be at all convincing to anyone looking at it second-hand. It is for this reason that we say that no second-hand evidence can ever produce what we call the Rabbit Hole Experience.
The ultimate question
In the vast majority of cases we investigate, we do not find any evidence to reject our null hypothesis. In these cases, we will suggest possible natural explanations for the activity to the client. We will also admit we have no idea if that is the case. I have no problem with that. And we will stay in contact with the witnesses to continue to collect data. Because whatever we conclude, at the end of the day, we could always be wrong.
But suppose we do go out and investigate using the approach I propose. And suppose we do find strong evidence against the null that there is no paranormal activity and we reject it. This does not “solve” the case at all. This leaves us with an even larger question: what is the alternative? Strictly speaking, it is just that the null is false. So if our null is that there is no paranormal activity, and we reject this, that implies the presence of some paranormal phenomenon. But in a strict sense, this simply means something outside of normal experience has occurred. As Jason Hawes put it in one of my favorite early episodes of Ghost Hunters, “something happened.”
There are all kinds of possible paranormal hypotheses, from spirits to psi to residual imprints. It could be one of these, or a combination, or something we have not even thought of yet. All we can do is take our best educated guess, make this our new null hypothesis and then go out and gather more data.
And the search continues…