Every device sold to ghost hunters measures something real. Magnetic field strength, air pressure, infrared radiance, antenna capacitance, depth geometry — these are ordinary physical quantities and the instruments measure them competently. The weak link is never the instrument. It is the step immediately after the reading, where an unexplained number becomes a spirit.
This guide takes the standard kit one item at a time and says what it detects, what sets it off, and what the published research has found. It is not a shopping list and it is not a sneer. If you are going to spend a night in a cold building with three hundred pounds of equipment, you may as well know what the equipment is doing.
Some of the links below go to Amazon and carry our affiliate code. We are paid a commission if you buy, it costs you nothing extra, and it has bought nothing in this guide: the devices we take apart hardest are the ones we link.
The meters
EMF meter
Measures the strength of alternating magnetic fields, in milligauss. The K-II, which is the one you have seen on television, covers roughly 50 to 20,000 Hz across five LEDs, from green below 1.5 mG to red above 20.
Two things are worth knowing. It is single-axis, so the reading changes depending on how you hold it — rotate it and the same source gives you a different number. Tri-axis meters sum three coils and give a figure that does not depend on orientation. And the K-II is poorly shielded: its own internal wiring acts as an antenna, so a UHF walkie-talkie at 446 MHz or a phone on a mobile band will light it up despite being far outside its stated range. A rapidly pulsing pattern is characteristic of radio interference rather than a field the meter can actually quantify.
Everything else that sets it off is domestic. Mains wiring, consumer units, ring mains, transformers, fluorescent ballasts, underfloor heating, lift motors, and faulty earthing and bonding — which are common in exactly the kind of old building that gets investigated. Thirty to seventy pounds for a K-II; eighty to two hundred and fifty for a calibrated tri-axis meter.
If you are buying one anyway: K-II EMF meter or tri-axis EMF meter. We earn a small commission, which does not change the price or the verdict above.
REM pod
Not an EMF meter, despite being sold beside them. It drives a telescopic antenna with an oscillator so the antenna radiates its own weak field, and anything conductive entering that field changes the antenna’s capacitance and trips the lights.
That makes it a proximity sensor, and it is a good one. A hand, a sleeve, a cable, a pet, a phone, damp air, condensation on the casing and thermal drift of the oscillator itself will all set it off. It tells you something entered the field. It does not tell you the field was disturbed by anything that was not already in the room. Thirty to a hundred and fifty pounds.
If you are buying one anyway: REM pod paranormal. We earn a small commission, which does not change the price or the verdict above.
Infrared thermometer
Reads the surface temperature of whatever is in its cone, and the cone widens with distance. Point it into the middle of a room and you are averaging the far wall with everything in between. It cannot measure air temperature at all, which is what “cold spot” is usually taken to mean. A logging air thermometer with a known response time is the honest instrument here, and costs about the same.
If you are buying one anyway: data logging air thermometer. We earn a small commission, which does not change the price or the verdict above.
The cameras
Thermal imaging
Measures long-wave infrared emitted by surfaces, and infers temperature from it — accuracy depending on an assumed emissivity for the material. Like the IR thermometer, it cannot image air. A thermal cold spot is a cold surface.
It picks up residual body heat on chairs and handrails, cold bridging, damp patches, draughts across surfaces, and reflections of warm bodies in glass and gloss paint. Those reflections are the ones that produce human-shaped figures. Two hundred to well over a thousand pounds.
If you are buying one anyway: thermal imaging camera. We earn a small commission, which does not change the price or the verdict above.
Full-spectrum and infrared cameras
A full-spectrum camera has had its infrared-cut filter removed, so roughly 350 to 1000 nm reaches the sensor. An IR camera with an illuminator sees near-infrared only.
If the illuminator sits next to the lens — as it does on almost every consumer model — then dust, pollen, insects and moisture droplets close to the lens are lit head-on and image as bright out-of-focus discs. That is what an orb is. Infrared reflectance also makes fabrics and foliage look wrong, and the false colour invites the eye to find faces. A hundred and fifty to four hundred pounds converted.
If you are buying one anyway: full spectrum camera paranormal. We earn a small commission, which does not change the price or the verdict above.
SLS camera
The most interesting one, because the figures it produces are not a malfunction.
An SLS camera is a structured-light depth sensor in the Kinect lineage: it projects an infrared dot pattern, computes depth from how the pattern distorts, and then runs skeletal-tracking software over the depth map. That software was written to fit a human skeleton to a human-shaped blob, and to infer joints it cannot see. It always proposes a best fit.
Give it a coat stand, a chair back, a curtain, a bin bag, foliage or a headstone and it will do exactly what it was built to do and draw a person there, complete with limbs behind the object. Low light, reflective surfaces and daylight degrade the depth map and make misfits more frequent. A hundred and fifty to five hundred pounds for a device whose headline behaviour is a documented property of the algorithm.
If you are buying one anyway: SLS camera paranormal. We earn a small commission, which does not change the price or the verdict above.
Laser grid
A diode laser through a diffractive optic, projecting a fixed matrix of dots. Genuinely useful as a visual reference — if something solid passes through it on camera, you will see it. Be aware that the dots also move with laser speckle, air currents, camera noise and thermal drift in the mount, and that dust crossing a beam glints. Twenty to seventy-five pounds.
If you are buying one anyway: laser grid projector paranormal. We earn a small commission, which does not change the price or the verdict above.
The audio
Digital voice recorder, for EVP
Records pressure waves. What turns up on playback that nobody heard at the time comes from microphone self-noise, automatic gain control pumping, lossy compression artefacts, handling and clothing noise, plumbing, building services, distant radio or television bleed, and the investigators’ own whispering.
The important research here is Nees and Phillips (2015), who primed listeners to expect spirit voices and then played them pure noise. About thirty-six per cent reported hearing voices, and belief in the paranormal predicted who did. You do not need a signal to hear speech. You need an expectation.
If you are buying one anyway: digital voice recorder. We earn a small commission, which does not change the price or the verdict above.
Spirit box
A radio scanning AM or FM in steps of a few hundred milliseconds. It is sampling live broadcast, so fragments of speech and music are not an anomaly, they are the designed output. Slow the sweep down and the stations resolve into stations. Add phonemic restoration — the brain’s habit of completing partial words — and the expectation the device was bought to satisfy, and you get sentences. Seventy to a hundred and thirty pounds.
If you are buying one anyway: spirit box P-SB7. We earn a small commission, which does not change the price or the verdict above.
What the research actually found
Three claims circulate as settled science and are not.
That electromagnetic fields make people feel a presence. This comes from Michael Persinger’s “God helmet” work, applying weak complex fields over the temporal lobes. Granqvist and colleagues ran a properly double-blind replication in 2005 and found no field effect at all: who reported a sensed presence was predicted by suggestibility and personality, not by whether the field was switched on. Gendle and McGrath found the same in 2012. Then in 2009 French and colleagues built a purpose-made chamber and manipulated Persinger-style fields and infrasound together in a controlled design with seventy-nine participants. Neither condition predicted anomalous sensations. The only thing that did was a suggestibility measure.
That 19 Hz makes you see ghosts. Vic Tandy’s 1998 paper is a genuinely good piece of noticing — he felt watched in a Coventry laboratory, saw a figure at the edge of his vision, and found a fencing foil vibrating in a vice, which led him to an extractor fan and a calculated 18.98 Hz standing wave. But Steven Parsons has shown that Tandy computed only the fundamental from the room’s length, ignoring its width and height; a full solution gives around twenty-seven standing-wave modes below 100 Hz, with 19 Hz carrying about half the acoustic power. The foil’s own resonance is nearer 190 Hz. Tandy called his own measurements quick and dirty, at plus or minus ten per cent. When 18.9 Hz was played to 439 people at GhostFest in 2007, one reported a visual anomaly.
Infrasound does seem to affect mood — Wiseman, Angliss and the National Physical Laboratory injected it into live music in 2003 and listeners who did not know reported more coldness, anxiety and shivering. That is a real finding. It is not the same as seeing a figure.
That mundane causes have been ruled out. They generally have not been looked for. Carbon monoxide is the established one: Wilmer’s 1921 paper documents a family in an 1870s house suffering headaches, depression, footsteps in empty rooms, voices, the feeling of being watched, a woman in black, and the sensation of being attacked in bed — all from a faulty furnace venting into the house, all resolved when they left. Toxic mould has been proposed as another and remains an unpublished hypothesis; treat it as a lead, not a finding.
What has replicated is duller and more interesting. Wiseman’s teams studied Hampton Court’s Haunted Gallery and the South Bridge Vaults in Edinburgh with hundreds of participants, and found reported experiences clustering in the reputedly haunted areas beyond chance — correlating with local physical features: draughts, air movement, lighting, temperature, ceiling height, floor area. Magnetic field variance correlated at Hampton Court and then failed to replicate in the vaults, which is worth stating because reporting only the positive half is the exact error this guide exists to avoid.
What a careful investigation does
Document the building before you record anything: age, construction, ventilation, heating, the consumer unit, where the circuits run, known earthing faults, and what is on the other side of every wall.
Take baselines over hours rather than seconds — logged air temperature, humidity, sound, field strength on a calibrated tri-axis meter — so that a later spike has something to be a spike against. Without a baseline, a reading is a number.
Use control areas with no haunting reputation and treat them identically. Keep people blind to which is which where you can. Interview witnesses separately, without leading them, and before they hear each other. Log every instrument’s mode, orientation and settings, and log radio and phone use, because it explains a great deal of what an unshielded meter does. Decide in advance what would count as an anomaly, so it is not decided afterwards.
Fit a carbon monoxide alarm and check for damp, as a duty of care rather than as a debunk. On the ethics of all this, Baker and O’Keeffe’s 2007 guidelines are the reference point: consent, confidentiality, never charge the person who reported the problem, work in pairs, refer on where someone needs medical or mental-health support, and leave the witness no worse off than you found them.
The device that would actually matter
There is no instrument on sale that detects a ghost, and the reason is not that the instruments are cheap.
A detector needs a specified physical quantity to detect — a mass, a charge, a wavelength, a pressure. No proposal about what a ghost is specifies one. So every device in the bag measures something else that is well understood, and the ghost enters at the inference: this reading has no obvious cause, therefore the cause is a spirit. That is unfalsifiable in both directions. Nothing detected proves nothing; something detected always has a mundane explanation available too.
A device worth having would be built the other way round. State the physical property first. Predict what the instrument should read if it is present. Then show the reading appears when it should and is absent when it should not, blind, repeatably, in someone else’s hands.
Nothing marketed to investigators is built that way, which is why the most valuable thing in the bag is the notebook.