I’m very much an empirical, data, statistics, and science type of guy. So, it might be a surprise to learn that I’ve gone ghost hunting a number of times. Now, I’m not a paranormal enthusiast. I’m definitely a skeptic. However, in my view, being skeptical about something does not preclude collecting data about it. I also have friends I trust completely who are sure they’ve experienced paranormal activity. Plus, I don’t need much of an excuse to try something new and unusual!
Three of us skeptical ghost hunters have spent the night by ourselves in a variety of supposedly haunted prisons and insane asylums. We do the standard “lights out” investigations. That’s where you make your way around with nothing but flashlights! The idea is that your senses are most sensitive in these conditions to detect paranormal activity.
Using the statistics mindset of collecting a good amount of data, we bring a lot of equipment: still cameras, audio recorders, video cameras, and electro-magnetic field (EMF) detectors. Our approach is to make serious efforts to debunk anything unusual that we experience, either on the spot or by reviewing the evidence later.
While we don’t collect data that we can analyze in statistical software, these are investigations where we collect data and draw conclusions. I’ve noticed a number of helpful statistics concepts that help keep the conclusions of our paranormal investigations grounded in reality.
Specifically, we test hypotheses, rule out competing explanations, understand effect sizes vs. natural variability, go to great lengths to obtain good quality data, assess false positives, and grapple with unusual observations.
Ghost Hunting and Hypothesis Testing
In both statistics and ghost hunting you evaluate two competing hypotheses by determining which hypothesis is best supported by the data. In ghost hunting these hypotheses are:
- Null: Ghosts do not exist
- Alternative: Ghosts exist
In order to conclude that ghosts exist, you must collect data that provides sufficient evidence to reject the null hypothesis. Further, a lack of evidence doesn’t prove that ghosts don’t exist, just that we didn’t observe them that night. In statistics, you can’t accept the null hypothesis, you can only fail to reject it. Click the link to learn why!
Paranormal enthusiasts take advantage of this inability to prove a negative. On our very first ghost hunt at Rolling Hills Asylum in East Bethany, New York, we did not experience a single, even slightly unusual event. Nada! That result almost ended our ghost hunting adventures on the first hunt!
The owner of the asylum stated that we happened to spend the night during an unusually quiet time. And, that’s the typical response of an enthusiast when they don’t find anything unusual. The ghosts are usually active, just not that particular night.
Ghost Hunting: Small Effects vs. Natural Variation
Even among experienced ghost hunters, reports of blatant ghost activity, such as full-body apparitions, are very rare. Many paranormal researchers hypothesize that ghostly spirits live in an energy impoverished state. It’s hard for them to interact with the physical world.
It’s also possible that ghosts just don’t exist.
Whatever the reason, the ghostly effects that we’re looking for are typically small. In statistics, you can detect small effect sizes but it’s more difficult. Small effects have a hard time standing out from the background variability or noise. Everything has a natural variation, and that can obscure a real effect. You also have to be careful not to confuse normal variation/noise with a non-existent effect. In ghost hunting, we quite literally had to contend with background noise.
Related post: Measures of Variability
Footsteps and Other Quiet Sounds
One of the more common phenomena that ghost hunters report are phantom footsteps. We’ve heard these too, both on wooden floors and metal staircases. However, old large buildings tend to creak and groan. When it’s dark and quiet, your senses are heightened and you notice the background sounds that you normally filter out.
Most of these footsteps can probably be explained by contraction in the cooler night. Also, these unmaintained, older buildings have paint chips that will randomly fall from the ceiling. These create sound and movement in the dim and quiet environment that can trick you.
In statistics, the natural variation is the context in which the effect occurs. You have to understand the normal variation in the data in order to determine whether something is truly unusual. The same is true with ghost hunting. You have to understand the normal noise context for any given building in order to understand what is unusual.
With this in mind, we didn’t classify the creaks as footsteps but rather the normal sounds of an old building.
Can You Trust Your Data?
Trusting your data is key in statistics. If you have bad data, your conclusions are automatically suspect. Unfortunately, the typical ghost hunting environment is conducive to bad data.
When hunting for ghosts, you’re working in a very dark and quiet environment that taxes the limits of your senses and equipment. Your senses are so heightened in the dim, quiet environment that you react to the background noise more than normal. The automatic gain on your equipment is also jacked up to the max, which generates noisy video and audio. Sound in an unfamiliar building can behave very strangely, which can trick you.
Consequently, for both statistical analyses and ghost hunting, you must employ a methodical approach to collect good data and verify the data that you do collect. Below are some examples of the steps we’ve taken.
Good Data is Worth the Price for Ghost Hunting
Just like any research, it costs money to collect data while ghost hunting. It costs more if you want to collect better quality data. In statistics, you want high quality data. So, we generally pay extra to go on private ghost hunts rather than public hunts. The last thing we want is to have our data contaminated by the noises and movements of others. Especially when you’re looking for the subtle signs of ghosts! We also buy extra equipment just to be able to crosscheck our data.
To illustrate the importance of this, we went on a public ghost hunt at the Trans-Allegheny Lunatic Asylum in Weston, West Virginia. A group of us were on one floor when we heard a loud rumbling on the floor above. It sounded like a loaded cart trundling down the hallway. These are very solid floors in a solid, stonemasonry building. It’s practically impossible to hear anything between floors. So, this was significant.
Our group had previously been on the floor where the rumbles originated, and we had been told that people often heard gurney sounds there. Our guide later checked with the group above us and they heard the sound too. None of them saw anything unusual and they couldn’t find anything that could have produced the sound.
If we had been alone in the building, this event could’ve been very meaningful. However, given that there were 11 people on the floor above us, we can’t realistically rule out the possibility that someone made the rumbling sound.
In statistics and ghost hunting, it’s often worth paying more for data that you can trust.
Crosschecking to Detect Bad Data in Statistics
In statistics, you want to be able to identify and remove bad data. When we go ghost hunting, we carry multiple recorders with us and leave other recorders around the site. It’s very informative to crosscheck the different recordings. We do this because the environment can trick both our senses and our equipment. This approach has paid off by helping us debunk a number of mysterious happenings. Below are several cases where crosschecking our data allowed us to debunk what first appeared to be solid evidence.
Electronic Voice Phenomena (EVP)
Ghost hunters try to record ghostly communications with digital audio recorders. I don’t understand the theory of how this supposedly happens. However, the idea is that you can hear the ghosts communicate on the recordings but not with your own ears. These are called Electronic Voice Phenomena (EVP).
Whispery EVPs are definitely one of the more common findings of paranormal investigations. The problem is that the ghostly voices are down in the level of the noise, and the human mind tends to find patterns in random noise. Once again, the question becomes whether the phenomenon is random noise or something paranormal. It’s hard to tell with something so quiet.
By having the audio recorder and video camera running simultaneously we’ve debunked several EVPs. In one case, we were sure that we heard a voice say a specific word. However, when we checked the video camera, the audio on it revealed that the “voice” was a water drip! The audio recorder had been pointed directly away from the drip. It recorded either the echo of the drip or else distorted the drip because it occurred out of the spatial range that the microphone was designed to record. The video camera was pointed at the drip and recorded it accurately.
Another case was debunked because the “voice” turned out to be caused by the rustling of clothes in the video.
These instances illustrate how you can be tricked when you crank up the gain and listen to background sounds without the larger context. Consequently, we generally discount whispery EVPs as being unreliable evidence of the paranormal.
Ghostly Moan?
While we were ghost hunting at the Ohio State Reformatory in Mansfield, Ohio, we all heard a loud moaning sound that reverberated through the prison. It sounded like a stereotypical ghostly moan! It left us chilled. My wife was suddenly less interested in hunting ghosts! In addition to hearing it live, the audio recorder and video camera that we carried both recorded the same moan.
It wasn’t until we checked an audio recorder that we left elsewhere in the prison that we determined that the “moan” was actually the high pitched whine of a revving motorcycle engine! The sound must have interacted with the cavernous interior and metal bars of the prison to sound like a moan!
Phantom Footsteps and Moving Shadows
While ghost hunting, there have also been a number of times where our senses have been tricked. We’ve heard footsteps approach from ahead of us and have seen shadows move in the background. However, when we carefully double check these anomalies, literally retracing our exact steps and movements, we find the true cause.
In the case of the footsteps, the acoustics of the building made our own footsteps sound like they were coming towards us! Once we figured out the correct arrangement, we could reproduce this at will.
Moving shadows are another ghost hunter staple. With moving shadows, we could always find, and reproduce, a dim but normal object that appeared to move against the background while walking.
Unusual Observations while Ghost Hunting
In regression analysis, the statistical output displays a table of unusual observations. These are observations that don’t fit the model well because they have a large amount of error. Unusual observations don’t necessarily invalidate the model. In fact, you typically expect 5 percent of your observations to be classified as unusual due to random error alone. However, it is always wise to double-check the list of unusual observations to determine whether there was measurement error or whether you need to modify your model.
For skeptical ghost hunters like us, our default model is that ghosts do not exist. However, several unexplained observations don’t fit this model. While they don’t necessarily invalidate our model, they are worthy of a closer look. Here are some of the unusual observations.
Orb
There was a glowing, spherical orb near the ceiling. It was only there for a moment and the person next to me noticed it as well. It was gone before I could video it. I tried to find a possible light source and debunk it but I could not reproduce it. Flashlights pointed up there just didn’t look the same.
Women’s voices
My wife and I both heard two women having a conversation in solitary confinement. Unfortunately, they were too quiet to be recorded. My wife was the only woman in the prison at the time. We could not identify the source of the voices but we wouldn’t be surprised if the sound somehow travelled along, and possibly got altered by, the numerous metal pipes and bars in the prison.
Stabbing/Numbness
My friend suddenly went numb, starting in his neck and spreading to half of his body. Paranormal enthusiasts suggest that this happens when a ghost touches you. This event occurred in the civil war section of an asylum after he provoked the spirits to get a response. Jokingly, I told him that he had just been bayoneted in the throat!
He’s not the nervous type and we doubt it was due to nerves because we were laughing just before incident. However, it’s hard for us to give much credence to a totally subjective experience. But he did feel weak for several hours afterwards.
Loud Sigh
In the same spot where my friend was “stabbed”, we later recorded a loud sigh. The audio recorder and video camera both recorded this sigh. Neither of us heard it at the time even though in the recordings it is as loud as our voices. It is not the typical whispery EVP.
We know it’s not coming from my friend because he was in the video at the time it occurred. Also, the audio channels on the video camera show that the sigh originated to the right of the video camera whereas my voice came from the left of the video camera.
This sigh is what we consider our strongest piece of evidence, yet it’s not enough to make us believe in the paranormal. We don’t have any explanation for it. If this had contained any recognizable words at the same loud volume as the sigh, we would give it more weight. As it is, the sigh really contained no vocal information to indicate intelligence rather than some unexplained noise in the environment.
Notably, most of these unusual incidents are undocumented. Our ability to debunk incidents is positively correlated with the amount of data we have. Consequently, it seems likely that the inability to debunk these unexplained cases is due to insufficient data rather than them being truly paranormal.
Ghost Hunting Closing Thoughts
Our approach to ghost hunting may not be that of a typical paranormal investigator. We are methodical and place a high priority on obtaining good data and using the principles of statistics to adequately test the competing hypotheses.
To date, we haven’t found any “smoking gun” evidence, nothing that made us believers. We’ve experienced a number of unusual things, but we’ve debunked most of them. Despite not finding anything paranormal, it’s a lot of fun exploring creepy old buildings at night with nothing but flashlights. When we listen back to our recordings, we mostly hear us laughing and having a good time. As an added bonus, ghost hunting gives us a small, but enticing, chance of making a discovery that shatters our worldview.
So far, we’ve found that ghost hunting is primarily an interesting psychological exercise where we’ve seen how easy it is to be deceived. Keeping yourself well-grounded with a methodical statistical/research mindset can help prevent you from being tricked!
Have you seen any ghosts?
Matheus says
Amazing post, Jim!
I´d like to see you in Gypsy tents testing:
Null hyphotesis: previsions are random
Alternative hypothesis: previsions are correct
And giving then the “Jim´s P-value < 0.05" quality seal…
Fran Baker says
Great use of a metaphor to promote interest and build understanding of statistical concepts challenging to many students!
Jeremy says
I’ve experienced ghosts on two occasions. One was footsteps of a heavy-set man, in the hallway of our older apartment building with wooden floors, over a few weeks. I later learned the neighbors had been hearing it too, and a neighbor said one day there was a knock on her door and when she answered it, no one was there. Spooky! Another was in the upstairs storage floor of an old shop, where employees had heard footsteps for years. I went up there to assist my girlfriend at the time, and the room she was in had been locked from the outside! Totally spooky!
Thanks for the stories of your ghost-hunting. There’s some beautiful architecture in those old buildings. One thing though, about your statistical approach (or reasoning approach) is that you say the null hypothesis can never be proved, only failed to be rejected. True enough, and appropriate for most of the statistical testing we use such as in clinical trials. But in some experiments you need to assume the null hypothesis is false unless it’s “proven” true, for example in very dire or hazardous situations where death is most likely to occur if you don’t do something, like risky medical treatment for highly fatal diseases or space flight.
Jim Frost says
Hi Jeremy,
That does sound spooky! Thanks for sharing! I was definitely impressed with some of the architecture. Old stone buildings!
We need to distinguish between hypotheses in the scientific versus the statistical senses. In statistics, the alternative hypothesis is always what the researcher is hoping to be true. It’s the finding that they want and that they’ll need to gather good quality evidence in sufficient numbers to show that it is true. The null is the one they don’t want. Typically, the null is no effect because the researchers are hoping that the medicine, procedure, or what ever will have an effect. So, they hope to reject the null, and need strong evidence to do so.
However, in some cases, the researchers want to prove that two things are equal. Maybe that a cheaper new drug is as effective as the more expensive established drug. In that case, the null and alternative switch! The null states that the two drugs having unequal effects and the alternative states that they’re equal. And then we use a specialized hypothesis test to assess their equivalence (i.e., equivalence testing). But, the researchers still hope to reject the null, and must collect strong evidence to do so, even in this type of test.
You definitely need to use subject area knowledge to correctly state the hypotheses for your statistical tests. And there is a translation between the scientific hypotheses and the statistical hypotheses for a hypothesis test.
Krishnabharathi says
Me and My friend have seen orb in a very dark room where light can’t enter but we were very scared before that also..
Jim Frost says
That sound scary. Thanks for sharing!
Phillip says
Great mixture of stats and the unknown!
For my part, I saw a human-sized column of mist cross my path at night. I pointed this out to my companion but it had gone by the time he looked around. We continued walking and it happened again, this time my friend saw it too! I am a skeptic so think it was marsh mist. My friend was more open to a supernatural explanation. Either way, we were both terrified at the time!
Jim Frost says
That’s creepy! Being followed by human sized columns of mist!
maskan says
I encountered ghosts in a shadowy form during a broad daylight, est. 11 am. I could not find concrete, logical objects that could possibly form those shadows. Ghosts exits, my belief…
Thanks Jim for the interesting post.