Ghost-Hunting Gear Explained: EMF Meters, Spirit Boxes and EVP

Amateur ghost investigation has a standard toolkit: a meter for electromagnetic fields, a digital voice recorder, a modified radio known as a spirit box, a thermometer or thermal camera, and cameras that see in infrared. The reasoning behind it is simple enough. If a ghost is real and can affect the physical world, perhaps it disturbs something an instrument can register, even when people see nothing.

The difficulty is that every one of these devices was built to measure something ordinary, and none has ever been shown to respond to a ghost, because no one has established what a ghost's signal would look like. Without that, a reading cannot confirm anything. A spike on a meter is a spike; whether it means a presence or a refrigerator depends entirely on what else could have caused it. That is why the most useful thing to know about each tool is what it actually measures, and what else in an old building produces the same reading.

An EMF meter measures electromagnetic fields, and the inexpensive handheld models popular with ghost hunters mostly respond to the low-frequency fields produced by alternating current: 60 hertz from mains power in North America, 50 hertz in Europe. Those fields surround wiring in walls, breaker panels, transformers, extension cords, appliances and electric motors, and they weaken quickly with distance. Some meters also react to radio-frequency signals from phones, walkie-talkies and wireless networks, which investigators often carry themselves.

So a reading that jumps as someone crosses a room usually means the meter passed near a cable inside the wall, a lamp cord under a rug or a phone in a pocket. Many cheap meters measure along a single axis, so the reading can also change when the meter is simply turned in the hand. The belief that ghosts cause field spikes has no experimental basis. A separate and more interesting idea is that strong or fluctuating magnetic fields might affect how people feel, and so produce haunted experiences. Tests of that idea have been mixed: a 2005 double-blind study at Uppsala University found that the mystical experiences people reported while wearing a field-emitting helmet were predicted by their suggestibility, not by whether the field was switched on.

A spirit box is a radio receiver modified, or built, to sweep continuously through AM or FM frequencies, pausing on each station for only a fraction of a second. The result is a stream of static broken by short fragments of whatever is being broadcast: a syllable of a news bulletin, half a word of a song, a burst of a talk show. Users ask questions aloud and listen for replies in the fragments.

Everything coming out of the speaker is real radio, so the device reliably produces speech sounds; the only question is how they are interpreted. Fragments that happen to follow a question will sometimes sound like answers, especially to a group that is listening hard for names and is primed by a dark, atmospheric setting. Investigators in the same room often disagree about what a given burst said, and clips played to people who do not know the question rarely produce the same reading. The device is, in effect, a machine for generating ambiguous speech, which the human brain is very good at completing.

Electronic voice phenomena, or EVP, are sounds on an audio recording that listeners interpret as voices nobody heard at the time. The idea took hold after 1959, when the Swedish painter and film producer Friedrich Jürgenson said he had heard human voices on tapes he made while recording birdsong. The Latvian psychologist Konstantīns Raudive went on to collect thousands of such recordings; his book, published in German in 1968, appeared in English as Breakthrough in 1971. Critics pointed out that stray radio broadcasts could account for many of his voices, which often mixed several languages.

Psychologists explain most EVP as auditory pareidolia, a form of apophenia: the tendency to perceive meaningful patterns, in this case words, in random noise. Recordings made in old buildings are full of hums, creaks, breathing, the rustle of clothing and the investigators' own whispers and stomach noises, often amplified and filtered until they are barely recognizable. Experiments bear this out. In a 2015 study published in Applied Cognitive Psychology, simply telling listeners that a study was about purported ghost voices made them more likely to hear voices in ambiguous recordings, and they could not agree on what the voices said. Once someone is told what a clip says, they tend to hear exactly that.

Thermal cameras and infrared thermometers measure the temperature of surfaces, not of the air, by reading the infrared radiation those surfaces give off. A cold patch on a wall usually means a draft, a window, damp plaster, a pipe or an exterior wall; a warm, figure-shaped patch can be heat left on a chair or a wall where someone leaned. Shiny surfaces such as glass, tile and polished wood also reflect heat from elsewhere, including from the person holding the camera. A cold spot felt on the skin is usually moving air, and old buildings, with their chimneys, cellars and gaps under doors, have plenty of it.

Night-vision and so-called full-spectrum cameras have their own artifacts. An infrared lamp mounted beside the lens lights up dust, pollen, insects and water droplets close to the camera and far out of focus, turning them into the glowing discs known as orbs. Long exposures in low light smear anything that moves into a translucent streak. Those effects, and the way the eye finds faces in shadows, are covered in the article on pareidolia, orbs and ghost photos.

None of this means a building with a haunted reputation has nothing interesting going on, only that instruments need controls before their readings mean anything. A careful investigation starts with a baseline: walk the building in daylight with the same equipment, map where wiring, appliances, ducts and pipes run, log temperatures and fields room by room, and note drafts, traffic, plumbing and when the heating cycles on. A spike where a cable runs is explained; a spike where nothing does is worth a second look.

Recording should be kept honest. Log the time, place and person for every reading; run two recorders in different positions so a sound on one can be checked against the other; keep phones and radios off or well away; and say aloud whenever anyone whispers, coughs or moves, so those sounds are not mistaken later. Decide in advance what would count as a meaningful result.

Analysis should be blind. Clips should be played to listeners who do not know what question was asked or what anyone else thought they heard, and their transcriptions compared. Readings should be compared with the same equipment in a similar room that has no haunted reputation. Investigations run this way rarely produce dramatic results, which is informative in itself: the effects tend to shrink as the controls improve. Other things worth measuring, such as low-frequency sound and carbon monoxide, are discussed in the article on infrasound and haunted feelings.