Testing infrasound thrills and chills with a double-blind randomized spook-house

John Huntington and his friends decided to test the reported phenomenon of infrasound (very low-frequency sounds) causing people to feel spooky chills and thrills, a phenomenon blamed for ghost sightings and reports of hauntings. They created a spook-house with a double-blind randomized infrasound generator and used surveys to check for a correlation between infrasound and creepy feelings. John exhaustively documented the experimental setup (and the setbacks encountered in getting things up and running), and the results. Spoiler alert: they didn't find a correlation.

We picked a 19 Hertz (Hz, or cycles per second) sine wave as our infrasound source, since that's the frequency Tandy and Lawrence had found in their initial investigation of a "haunted" space (see Part I of this series for details). For our initial tests last spring, we used an Audio Toolbox as our signal generator, and connected it up to our ancient Apogee P-10 subwoofer processor and a Crown K2 amp, which drove an Apogee AE-10 double 18" subwoofer (these units work just fine, but haven't been made in years.) We got the sound going, but after a minute or so, it would cut out. I figured out that the processor was apparently protecting the subs against the "bad" infrasound, and was cutting out. So, I bypassed the processor and its protection circuits, and (carefully) drove the subwoofers directly from the amp. This worked just fine, but I was a bit concerned about damaging the speakers themselves. Fortuitously, over the summer I managed to get for our department a massive, modern Meyer Sound 650-P powered subwoofer. And after some tests in late summer we determined that the unit could generate quite a bit of 19Hz without clipping, and we figured the level was pretty good because by the time we pushed the 650-P up to its limits, the effects of the infrasound would be obvious to anyone in the room, which wouldn't have been acceptable for our purposes.

Part 1, Part 2, Part 3, Part 4, Part 5

(Thanks, John!)


  1. Years ago I built some reflex speakers with port frequencies of 12-14 Hz, just to be difficult. Driven at those frequencies, listening tests suggested the presence of standing waves in the room (probably the audible harmonics, as this report observed), and demonstrated the ability to make things rattle throughout the house- as well as drive my three cats into hiding.

  2. I have slightly extended hearing range. I believe I’m what’s called a “hummer” in medical terms. I can hear that 19Hz WAV as clear as day at even a moderate volume through my headphones (Philips SHP805s powered by a NuForce μDAC, no EQ) I also always seem to hear those mosquito tones that only kids are supposed to be able to hear. I’m 33.

    Although I completely understand their reasoning, I think 19Hz was a poor choice of frequency, to be honest. Right at the edge of audibility for even average listeners.

    If they’d used a rotary subwoofer (look it up) to get things a couple of octaves down to 4Hz or so, I’d have liked to see the difference in reactions. I’m quite sensitive to noise, and every now and then have definitely experienced sound of a low enough frequency to cause a sort of mild panic that dissipates once I’ve looked around and seen that nothing’s happening.

    I know it’s an anecdote, but that’s my experience. I hear and feel very low frequency sound as a a kind of heaviness and tightness in the ears, chest and throat. A kind of full-body throb. It’s not too intrusive for me, luckily (some have it far worse) but I’m a wreck if there is high frequency noise about.

  3. While this was very interesting, I didn’t really care for the experimental design.

    First of all, judging from the video in Part 2, the exposure time was too brief. I would have wanted to leave the subjects in the room for at least 5 minutes – and preferably 10 or 15 minutes – before having them fill out their survey.

    Second, there were too many distractions that could have muted the emotional impact of the infrasound: Apparently, instead of bringing subjects into the room one at a time, they were in groups; and the subjects had to listen to that spiel about the murdered professor, and then the lights went out and came back on. It would have been better had they brought the subjects into the room one at a time, with no distractions – no narration, no lights going off and coming back on, etc. – just have each subject spend 15 minutes alone in a dimly lit room either with or without the infrasound.

    Third, there was apparently no pre-test survey. The experiment would have been better had there been pre-test/post-test surveys so that any changes in a subject’s emotional state could be measured.

    Fourth, conducting the experiment in a Halloween “haunted house” instead of in a regular psych lab was a bad idea, in my opinion. The haunted house setting, and the narration about the murdered professor and people having weird experiences in this particular room, sets up certain expectations on the part of the subjects. These expectations might influence the emotional responses of the subjects (or at least how they choose to respond to the survey questions).

    Here’s how I would have designed the experiment: Bring subjects into the psych lab one at a time. Tell them that you’re doing an experiment on emotional responses to specific stimuli; but don’t elaborate. Have them fill out a pre-test survey of their emotional state prior to the experiment in order to get a baseline. (For added authenticity – and additional data – you could also have a nurse check their pulse and blood pressure at this time.) Then tell them that, in order to insure that they are in a relaxed emotional state for the test, they must sit alone in a dimly lit waiting room for 15 minutes. This room is actually the lab, which should be monitored by video camera at all times to see what the subject is doing. While in the waiting room, the test subject will either be exposed to infrasound (test) or to no infrasound (control). At the end of the 15 minutes, the subject should be given a post-test survey (and the pulse and blood pressure checked again). But this time, there should be a more in-depth interview (by an interviewer who does not know whether the subject was exposed to the infrasound or not), who will ask probing questions about how the subject felt while sitting in the room. At this point, the true nature of the test should be explained to the subject, who will now be free to leave.

    After the data have been gathered, the analysis can begin. Obviously you want to compare the pre-test survey results to the post-test results and see if there was any change. (Likewise, you want to compare pre-test and post-test pulse and blood pressure readings, if you took them.) You also want to examine the details of the post-test interview to see if those subjects exposed to the infrasound reported any unusual feelings or experiences that the subjects in the control group did not. Finally, you want to examine the video tapes of the subjects as they were sitting in the room. Find someone who is an expert on body language and have that person watch the tapes; but make sure that this expert does not know what the experiment was about, or which subjects were part of the experimental group and which were part of the control group. Have this body language expert rate each subject for signs of either relaxation/boredom or anxiety/agitation. Then compare the results for subjects in the experimental and control groups.

    Anyway, that’s how I would do it. But it wasn’t my experiment; and I’m sure the researchers had their reasons for designing the experiment the way they did. While it might not be an ideal research design, in my opinion, it did yield some interesting results.

    1. Well, its a field study, not an experiment. You give up some control to put things in a more natural context. I suspect that if they try to get this research published, they’d have to run a lab study as well, not to mention have to come up with a study that explains why they didn’t find anything in their field study.

      With regard to your first point, why would you expect longer exposure to generate a more pronounced response to infrasounds?

      By random assignment there is no reason to expect any differences between any of the cells before exposure to the stimuli, and the stimuli used were the same except for the presence or absence of the infrasound. Any of the variation that came from the presence of the people/sounds/story/whatever should have averaged out across cells, especially considering the whopping 200+ N for a two-cell design. This also means they don’t need to use a pre-test/post-test design, since they can just compare across the cells. They aren’t interested in changes within the cells over time, they want to show that infrasound is affecting people’s emotional experiences. Using a within design like that would give the study unnecessary power, which given the enormous cell sizes it already has in spades.

      You’re right, however, that this isn’t a particularly clean test of their research hypothesis, that exposure to infrasounds causes emotional discomfort. Your suggested study would be more effectively able to get at this issue, but if there is any awareness of the infrasounds, its very likely to generate demand effects in the setting and analysis you describe. Given what brought the authors to their research question, I think their interest is more in testing whether infrasounds can actually drive the sorts of effects that are attributed to them, which is another reason why they used the design that they did. However, I suspect the reason they didn’t find anything is because there probably isn’t anything there.

    2. I can see a few reasons for doing the study this way. Here is one.

      The initial hypothesis may be that EITHER the sound by itself induces fear, OR it sensitizes you to fear-inducing stimuli. Thus, the haunted house and the story may be an important element of the design.

      I appreciate your argument for the lab design as a means of reducing the “noise” in the system. However, sometimes when you’re too careful to remove noise, you also lose the signal.

    3. To follow up on what I just posted, in effect this kind of study allows you to test a hypothesis that is less specific than the lab setting you describe… to test that idea that infrasound plays SOME kind of role in fear reactions, whether that is causing or augmenting or sensitizing. If they had found a positive result here, the next experiment would have been to dissect it further in the lab, and get at which possible mechanism is involved. It’s often not a bad idea to start with a design like this when there is a range of possible mechanisms, and you need to just confirm that there is something there to study.

    1. I said 19Hz. I can hear both 19Hz and 19kHz very easily. 19kHz is not ultrasonic.

      There’s very little in the way of affordable equipment that generates a single-digit-Hz tone, which I cannot “hear” as a “tone” but can certainly sense physically as a kind of oppressive atmosphere and something similar to an altitude pressure change discomfort in my ears. It’s very common in large parking garages, I find.

  4. Used to produce lottsa rock’n’roll shows from ’85 to ’95; venue sizes from 500 to 5000. Our “infra theory” back then was that sustained exposure to high-volume output of 12Hz-18Hz made girls a bit itchy in the pants. Our big rig for this was 4 Community sub cabinets running eight EV 18s (can’t remember the cabinet designation; the big 4X6 boxes with fibreglas “horns” for loading dual 18s), powered by 4 Crown PSA-2 amps (those puppies could push 2 ohms all night long), plus whatever other 18s we coud spare. Source was a 16Hz “drone” from a sig-gen plus whatever infra was coming from the stage-source-mixer.

    Three common results: when doing test-builds of a system — usually at 2 in the morning, ripped — our pole barn shop would be surrounded by mammals, just sitting; transfixed. Without the infrasound we’d get no audience. (Squirrels, racoons, feral cats, young deer, an occasional possum, all just chillin’). Second, it was lottsa fun to tune the crossover to find resonant points at the venue — hit the right spot and you could pop all the doors open or knock the cleanup crew off their feet. Third, it did make the girls itchy…they wanted something to happen. (Unless you got it wrong and over-excited the whole crowd, at which point a large number of M/F just seemed to become surly.)

    Seems like the experimenters here didn’t give enough exposure to notice an effect, and it’s a shame they couldn’t run something better than that +/-4dB 28Hz-100Hz 650-P sub.

  5. Infrasound is one of those scientific corners like subliminal advertising or backmasking on recordings: everybody knows about it, most people believe that it is being used at this moment to mold out minds, but when you try any actual experiment nothing happens.

  6. Infra-sound has been standard tech for religious organisations since the year dot. They mostly use reverberations and harmonics from audible frequencies such as echoes from the surf and so on. Often bi-aural beats are mixed in. As others have noted, it’s horrible trying to get audio equipment to do this – especially since most of it now only wants to produce the sound range encoded on CDs. Schumann resonance, for example, tends to shake apart computer speaker enclosures.

    The experimenter has an odd choice of frequencies. For example, in Fortean circles 11 Hz is widely known as the fear frequency: often traced back to malfunctioning gas appliances in “haunted” houses.

  7. @jemather & Gregory Goldmacher:

    Of course, researchers always have to be pragmatic about their research designs, since there are practical limits – in terms of time, expense, facilities, equipment, scheduling, etc. – to what they can actually do. So, I’m perfectly willing to accept that many of the imperfections in this research design were due to pragmatic compromises that the researchers had to make. But it’s still worth pointing out that ideally the research design could have been better; even though we must acknowledge that practically they may have done the best they could do given the constraints they were working under.

    But, even given the practical constraints on the researchers, there are three core principles that must always be followed in any research:

    (1) The research design has to be driven by the research question you’re trying to answer. So, when evaluating any research, we have to ask: What question are the researchers trying to answer? and Is their research design adequate for answering this question? In this case, if the research question is whether, and under what circumstances, infrasound can induce feelings of anxiety, dread, or agitation in humans, I don’t think the design is quite adequate. If, however, the question is whether infrasound can enhance the experience of a Halloween “haunted house”, then perhaps it was a decent research design (though, again, I think the exposure time was too limited to draw any firm conclusions; and, as S2 pointed out, the intensity of the infrasound might not have been sufficient; or, as Nadreck pointed out, they might not have used the correct frequency).

    (2) The research design has to try its best to isolate the effect of the independent variable, controlling for any potentially confounding variables. It isn’t enough simply to make sure that nothing varies between the test subjects except for the independent variable. You’ve got to try your best to eliminate any factor that might mask the effect of the variable you’re trying to study. Suppose you’re conducting a double-blind taste test between Coke and Pepsi. But suppose that you hold this taste test at a buffalo wing eating contest, and that all of the participants in the taste test have just eaten as many extra-hot, habanero-flavored wings as they could force down their gullet. Under those circumstances could you trust the results of your taste test? Probably not – the intense flavor of the wings would probably overwhelm the relatively subtle difference in flavor between Coke and Pepsi. You really need to try your best to eliminate any factor that might mask the effect of your independent variable. In my view, this research design fails to do that. The “haunted house” setting, with the spooky story and the lights going out, is itself designed to elicit the same emotional response that the researchers are trying to measure. It is quite possible that the emotional response to the setting itself might mask any effect that the infrasound might have. (Unfortunately, without a pre-test, we don’t how much the setting itself might have affected the emotional state of the test subjects.) This test does appear to demonstrate that a relatively short exposure to a relatively low intensity of infrasound at a very specific frequency does not really enhance the feelings of fearfulness produced by a “haunted house” scenario. But it doesn’t really demonstrate anything beyond that.

    (3) The research design has to give the experiment a fair chance of working. Imagine this: You take two very large pots and fill both to the brim with tap water. Set them both on the stove, turning one burner on the lowest setting and leaving the other off. Wait 1 minute, then place thermometers in both pots to see if there is any difference in temperature between them. You see no significant difference; so you conclude that stoves do not, in fact, heat water, as is commonly believed. Myth busted. Right? Of course not! Everyone knows that it takes a lot longer than a minute for a low flame to significantly increase the temperature of a large pot of water. The research design was not adequate, because it didn’t allow enough time to produce a measurable effect. Affective states are a lot like large pots of water: They start out at a given emotional “temperature”, and they tend to remain more-or-less at that temperature unless something “heats them up” or “cools them down”. Certain extremely intense stimuli have the ability to heat up or cool down someone’s emotional temperature very rapidly; but most stimuli are not so intense, and take a while to alter a person’s affective state. If infrasound can affect someone’s emotional state, I wouldn’t expect it to have a measurable effect after only a brief exposure. That’s probably the main criticism I have of this experimental design – the exposure to infrasound was too brief to give it a fair chance of working.

    1. Of course, any design – particularly one in psychology – could be made “ideally” better, but that’s not realistic. Saying that it could be made better ideally doesn’t really say anything about their design choices.

      (1) “The purpose of our proposed research is to examine the role of infrasound (sound waves below 20 cycles per second) on emotional states in audience members in our existing Gravesend Inn ‘Haunted Hotel’ attraction. ” Near as I can tell, that’s what the experiment did. Again, what basis do you have for thinking that additional exposure will increase the effect?

      In the Tandy and Lawrence paper they cite, they specifically give an 18.9hz frequency as the source of the effect, so I don’t see how criticisms of the frequency not being “correct” holds any water, and the Tandy paper list a 19hz of 38db over ambient, which is consistent with the 40db over ambient they had for their experiment.

      (2) Actually, it is. While reducing noise is helpful in demonstrating significant effects, when measuring something like psychological DVs, noisy measurements are the norm. The “noise’ in this study just makes the task of demonstrating a significant effect harder. However, their SDs are low (1 point on a five point scale or less), which suggests that noise is not a factor here. They have more than enough power that if anything was there, it would have shown up. Given the lack of theoretical guidance in this area, I can’t see how you can claim that two minutes is insufficient to provoke an emotional response.

      A pretest won’t tell you anything, unless you have some reason to believe that there was a difference in the two groups before the experiment and then the infrasound made them go to exactly the same emotional state after the experiment, in which case people probably weren’t randomly assigned. It absolutely does not matter how much the stimuli affected the subjects, because that’s not the research question. The researcher wants to know if an infrasound wave will affect emotional states. They aren’t interested in whether it will amplify emotional responses to the existing emotion-inducing stimuli of the haunted house itself.

      (3) The real problem is that there is no theoretical link offered in the existing literature for affective response to infrasound, which you have alluded to. We have a pretty big body of literature on how water heats up, so we can easily say that 1 minute is insufficient to see a significant difference in temperature. The link between infrasound and emotional responses is pretty tenuous in a brief perusal of the papers they are citing – most of the research has studied physiological responses.

      1. A few years ago, MythBusters tested the myth that Nicola Tesla built a small device that, if attached to a structure, could cause that structure to vibrate so violently that it would produce effects similar to an earthquake in a matter of minutes. In order to test the myth, they built a device that met Tesla’s specifications, and attached it to various structures to see what effect it would have. It caused some vibration, but didn’t produce effects even remotely like an earthquake. Myth busted. But here’s the important part: Even though the myth said that the earthquake effect should occur within a matter of minutes, Adam and Jamie let the device run for nearly an hour before they finally declared the myth busted. Why? Because, when you’re trying something new and don’t really know what to expect, you give it every possible opportunity to work before declaring that it doesn’t work. If your hypothesis says that it ought to work within minutes, and you give it an hour and it still doesn’t work, then you can be pretty sure that your hypothesis has been falsified. While science (quite rightly) is willing to accept a considerable risk of Type II error in order to avoid committing Type I error, it shouldn’t invite unnecessary Type II error when it can be avoided. Having the subjects remain in the room for 10 minutes instead of 2 minutes would not add one iota to the risk of Type I error; but it might dramatically reduce the risk of Type II error.

        You point out that we know that it takes more than 1 minute for a large pot of water to heat up. True. But how do we know this? We know it because the experiment has been done many times before. If everyone who had ever attempted to heat up a pot of water had given up after 1 minute, we still wouldn’t know how long it takes to heat up a pot of water. Unless you have some compelling reason to believe that infrasound is supposed to produce an emotional effect very quickly, then it only makes sense to give it more time.

        Keep in mind that human emotion is far more complex than the heating of water. Emotion is affected by many factors which interact in complex ways; and we don’t fully understand how this works. So, when studying emotion, you’ve got to be very sensitive to possible interaction effects. A stimulus that produces a very distinct emotional response in one setting might produce a completely different emotional response in another setting, and might not produce any measurable effect at all in a third setting. Watching a beautiful sunset while holding hands with the person you’ve just fallen in love with is likely to produce a very different emotional reaction than watching a beautiful sunset while holding a photo of a loved one who has recently died, or watching a beautiful sunset while having a heated argument with your spouse. In the first case, the sunset is likely to elicit feelings of bliss, in the second feelings of sadness, and in the third no measurable feelings at all.

        I find it extremely unlikely that emotional response is either linear or additive. You can’t simply assume that a stimulus that is believed to elicit a fear response is going to have the same effect on someone who is standing with a group of people in a “haunted house” as it would have on someone who was sitting alone in a quiet, dimly lit room. If you see blood trickling down the wall of a haunted house on Halloween, you’ll probably shrug it off as something to be expected, and may even turn to the person next to you and laugh about it; but if you see blood trickling down the wall as you’re sitting alone in a quiet, dimly lit room, you’re gonna freak out. For emotional stimuli, the setting matters.

        1. hey that was great what you said. I found mythbusters give up sometimes and don’t try again, but they don’y want people doing stuff at home to endanger others. Tesla’s technologt worked, but it was a bit of tunneing. David Sereda has done alot of work on harmonic ressonance. In stead of 19Hz, they can be 2 or more frequencies working together like a guitar or piano chord. This gets into hyperdimentional physics. David claims to have made a Tesla Coil with about 800 watt input with 1300 watt output. The possible source is torsion physics. I plan to make a Tesla coil from a Neon power suply 9000v @ 30mA AC. Would like to try to get certain ratios with the coil. Yes I have seen ghost, orbs and energies before, no infra sound can do that. I started with some Zen in my younger years. Our brain frequencys change and resonate during experiences. Keep asking questions !?

  8. Infrasound. Emotional effects…emotional response…emotional reaction…emotional stimuli.

    Infrared. Emotional rescue.

  9. We talked about this in the graduate course about the auditory system. (cognitive neuroscience)

    You have to employ a standing wave to achieve effects. Passing through the standing waves is really uncanny.

    Why does nobody ever think of this when talking about this phenomenon?

Comments are closed.