In August 1977, a single mother named Peggy Hodgson called the police to a council house on Green Street in Enfield, north London, because the furniture was moving. The officer who attended, a WPC named Carolyn Heeps, later signed an affidavit stating that she had watched an armchair wobble and slide roughly four feet across the floor with nobody near it. Over the following eighteen months, more than thirty people, including neighbours, journalists, and investigators from the Society for Psychical Research, reported seeing objects thrown across rooms, heavy furniture shifting, marbles and plastic bricks flying, matches igniting by themselves, and the two Hodgson daughters apparently lifted off their beds. An eleven-year-old girl began speaking in the gravelly voice of a man who had died in the house. The investigators logged over two thousand incidents.
It is the most thoroughly documented poltergeist case in British history, and it is the best available demonstration of something that runs through the entire subject and is almost never named directly. The poltergeist literature contains a structural feature so consistent that it functions as a signature, and it is this: the phenomena are inversely proportional to the quality of observation. Where nobody is watching, furniture flies and people levitate. Where an investigator is present, things happen behind him. Where a camera is running, the events shrink to small objects at the edge of the frame. And where continuous, controlled, automated recording is finally achieved, what tends to get captured is a child bending a spoon. That relationship is not incidental to the phenomenon; it may be the whole of it, which puts poltergeists closer to the beliefs a group can construct together than to the genuinely unmapped territory.
What a Poltergeist Is Supposed to Be
The word is German and means noisy ghost, and the phenomenology is remarkably stable across centuries and cultures. Reports involve percussive sounds with no visible source, knocking and rapping in walls, objects moving or being thrown, furniture displaced, doors opening and closing, small fires starting, water appearing, electrical equipment malfunctioning, and occasionally voices. Cases begin abruptly, escalate over weeks or months, and then stop as suddenly as they started, usually within a year. That arc, sudden onset, escalation, decline, cessation, is one of the few features that everybody in the field agrees on.
The consistency across cultures is worth pausing on, because it is often presented as evidence that something real underlies the reports. Cases with recognisably identical features appear in medieval European chronicles, in nineteenth-century American frontier accounts, in Brazilian and Japanese records, and in modern British council houses, and the resemblance is genuinely close. But a stable phenomenology can arise from a stable cause of any kind, and adolescence, household friction, and the acoustics of buildings are at least as universal as any hypothetical force. A phenomenon that appears wherever there are stressed teenagers and thin walls is not obviously telling you about physics.
What distinguishes a poltergeist from a haunting is where the phenomenon lives. A haunting is attached to a building: the grey lady walks the corridor whoever happens to be living there. A poltergeist attaches to a person, moves with them, and stops when they are removed from the situation. Investigators call that person the agent or the focus, and the pattern is so consistent that it is the primary diagnostic criterion in the professional literature. This makes poltergeist cases genuinely different from most claims filed under the paranormal, including the reports that fill the modern archive of aerial phenomena, because there is a specific, identifiable, available human being at the centre of every case, whose presence can be tested against the timing of events. It also means that whatever explanation you favour has to run through a person’s nervous system, which is a strong constraint of the kind that anchors the science of how brains and bodies actually work. The ghost, in other words, comes with a person attached, and that is the single most important fact any poltergeist case presents to an investigator.
The Person, Not the Place
Look closely at who the agent tends to be and the pattern sharpens considerably. In the great majority of documented cases the focus is an adolescent, more often a girl than a boy, typically between about eleven and nineteen, living in a household under significant strain. Enfield centred on eleven-year-old Janet Hodgson and her thirteen-year-old sister Margaret, in a council house with a recently separated mother struggling in the economic climate of 1977. The Rosenheim case in Bavaria centred on a nineteen-year-old secretary in her first job. The Columbus case in 1984 centred on a fourteen-year-old girl in a difficult family situation. Again and again, the profile is a young person with limited power in a stressed household, at an age of maximum social pressure, in a situation nobody is choosing.
It is also worth noting what is absent from the profile. Poltergeist agents are not reported in proportion to population; they cluster in adolescence and vanish in adulthood, which is a strange constraint for a physical force and a very natural one for a behaviour. Nor do cases cluster around people with unusual physical characteristics, occupations, or medical conditions, which is what you might expect if some rare physiological capacity were involved. What they cluster around is a social position: young, dependent, closely supervised, with little formal power and a great deal of time spent in the house where the events occur.
That demographic regularity is the strongest fact in the entire subject, and it is the point at which the field divides, because it is exactly compatible with two utterly different explanations. Either adolescence under stress generates some unrecognised physical effect, or adolescence under stress generates behaviour, and the behaviour is producing the phenomena. Both hypotheses predict the same demographic signature, the same onset and cessation, and the same attachment to a person rather than a place, which is why the observation has been claimed with equal confidence by both camps for a century. Adolescents in constrained circumstances finding indirect ways to alter their standing in a group is one of the best-attested patterns in the study of social animals, visible throughout the political manoeuvring of primate troops, and the production of misleading signals to change how others behave is likewise ubiquitous in nature, as catalogued in the natural world’s repertoire of deception. One demographic fact, two incompatible theories, and nothing in the observation that discriminates between them.
The Theory: Recurrent Spontaneous Psychokinesis
The parapsychological account has a formal name and a serious pedigree. William Roll, working at the Psychical Research Foundation in North Carolina in the orbit of Duke University’s parapsychology programme, surveyed 116 written poltergeist reports spanning four centuries and many countries, and proposed the term recurrent spontaneous psychokinesis, or RSPK, to describe what he took to be a real physical effect generated unconsciously by a living agent. The theory holds that psychological tension in the focus person, particularly hostility that cannot be safely expressed, discharges as physical disturbance in the environment, and crucially that the agent does not know they are doing it.
It is worth being fair about the intellectual seriousness of this tradition, because it is easy to caricature. Roll, Bender and their colleagues were not credulous enthusiasts; they built case databases, applied psychological testing to agents, brought in physicists, and attempted to specify falsifiable correlations between phenomena and measurable variables. That is what a research programme looks like, and the reason it failed is not that the people were foolish. It is that the object of study kept declining to appear under the conditions that would have settled the matter, which is a problem no amount of methodological rigour on the investigator’s side can fix.
The theory has real explanatory virtues and one enormous vulnerability. Its virtues are that it accounts for the person-attachment, the adolescent profile, the correlation with household stress, and the sudden cessation when circumstances change. Its vulnerability is contained in Roll’s own characterisation of the agent as someone who is unaware of causing the disturbances while being, in his phrasing, secretly or openly pleased that they occur. Read that description again without the psychokinesis and it is an exact portrait of a child producing effects deliberately and enjoying the result. The theory therefore describes the observable facts perfectly and specifies an unobservable mechanism that does no additional work, which is a difficult position in which to make progress, and it faces the standing problem that inner states are extraordinarily hard to establish from outside, a difficulty that runs through the whole scientific effort to determine what another creature is experiencing, as in research on subjective states in animals. The theory fits the evidence well. So, unfortunately for it, does the entirely boring alternative.
Enfield
Enfield deserves its reputation because both the investigation and the criticism were unusually thorough. Maurice Grosse, a retired inventor and SPR member, arrived in September 1977 and stayed involved for over two years; he had come to psychical research after the death of his own daughter, who was also called Janet, a fact worth stating because it explains his investment without impugning his honesty. He was joined by Guy Lyon Playfair, a writer who had studied poltergeist cases in Brazil and who arrived, by his own account, sceptical. They kept detailed logs, made extensive tape recordings, attempted photographic documentation, and produced the standard account of the case from the believer’s side.
What they reported is genuinely striking, and the Society for Psychical Research’s own account of the affair, in the Psi Encyclopedia entry on the case, records that an SPR committee convened in 1978 to review the investigation concluded there was good evidence for paranormal phenomena described by credible informants. But that same committee, in the phrasing that matters most here, reserved judgement on incidents that could not have been clearly observed, and was wary of the voice. Even the sympathetic institutional review, in other words, sorted the case by observational quality and found that its confidence tracked how well an event had been seen. The voice itself, attributed to a man named Bill Wilkins who had died in the house, was produced using the false vocal folds, a real and learnable technique for generating a deep rasping speech, and a professional ventriloquist who examined the recordings reproduced it without difficulty; the physiology of voice production is thoroughly understood, and lies within the same domain of applied neuroscience as the engineering of neural prosthetics. Playfair also noticed that the voice had a habit of abruptly changing the subject, which was a habit Janet had. Meanwhile the press camped outside, and the machinery of public attention turned two frightened children into performers. The investigation was serious. So were the doubts.
Rosenheim
The case usually offered as the strongest in the entire literature happened not in a family home but in a lawyer’s office in Rosenheim in southern Bavaria in late 1967, and it is genuinely the best-instrumented case on record. Sigmund Adam’s law practice experienced light fixtures swinging and bulbs exploding, heavy furniture apparently shifting, fluid leaking from the copier, and above all a telephone problem: the firm was billed for enormous numbers of calls to the speaking clock that nobody admitted making, on the order of six hundred in five weeks, including forty-six in a single quarter of an hour, at a rate that seemed impossible for the mechanical dialling equipment of the period.
It is worth separating the Rosenheim phenomena into two very different classes, because they are usually discussed as one. The electrical and telephone anomalies were recorded by instruments belonging to third parties with no interest in the outcome, which makes them the strongest data in the case. The physical events, furniture moving and pictures rotating, rest on the same witness testimony as every other case in the literature. Those are not equivalent evidentially, and the strength of the first has been used to underwrite the second for nearly sixty years, which is not how evidence is supposed to work.
The response was impressive. The telephone authority installed monitoring equipment. The power company investigated and reported substantial surges. Hans Bender, who directed the institute for the study of borderline areas of psychology at Freiburg, took the case, and two physicists from the Max Planck Institute joined the investigation. The team concluded that events occurred only when a nineteen-year-old secretary was present, and that they ceased when she left the firm. That is an impressive correlation and the case is often presented as decisive because of the instrumentation involved, which is the right instinct, since measurement is what converts an impression into evidence across every serious technical undertaking. The criticisms are equally substantial: Bender’s investigation has been faulted for omitting details and for setting aside naturalistic explanations early, and the telephone anomalies in particular have a mundane candidate, since pulse-dialling systems register calls by counting electrical interruptions, and a system experiencing genuine power surges can generate phantom call records with nobody touching a handset, an interaction between electrical noise and instrumentation familiar from the physics of electromagnetic systems. The best case has the best instruments. It also has an ordinary explanation nobody ruled out.
Nobody Ever Sees It Start
Now to the structural feature, and once it is pointed out it is very difficult to stop seeing. Across the entire case literature, almost nobody ever witnesses an object begin to move. What witnesses report is hearing an impact, turning around and finding something has moved, seeing an object already in flight after it has left its origin, or entering a room to find it disarranged. The initiation, the moment at which a stationary object departs from rest, is the one part of the sequence that is essentially never observed, in thousands of documented incidents across centuries.
The regularity is easy to check against any case narrative. Read Enfield, or Rosenheim, or the Bell Witch, and count how many incidents include a witness who was looking directly at a stationary object at the moment it departed from rest. In a literature of thousands of documented incidents, the number of such accounts is close to negligible, and those that exist tend to come from witnesses who were also the agent, or from moments of confusion in a dark room. Every other element of a poltergeist event is well witnessed. Only the first instant is missing, consistently, across centuries.
This is usually presented as coyness on the phenomenon’s part, and it is treated in some of the literature as an interesting property of poltergeists, that they are shy of direct observation. But there is a much simpler reading available. If the events are produced by a person, then the moment of initiation is precisely the moment that must be unobserved, because it is the only moment at which the mechanism is visible. Everything afterwards, the flight, the impact, the aftermath, can be witnessed by anyone without giving anything away. A thrown object looks identical whether a hand or a poltergeist launched it; only the launch distinguishes them. The perceptual point matters too, since human vision is drawn to motion already underway and is very poor at reconstructing what preceded a sound, which is a general property of how visual attention is allocated, mapped in the science of how sight is assembled and in the study of how animals detect and orient to stimuli. One instant in the sequence is never seen, and it happens to be the only instant that could settle anything.
The Observation Curve
Put that together with the case histories and a quantitative pattern emerges that deserves to be treated as a finding in its own right. Poltergeist phenomena vary in magnitude in inverse proportion to the rigour of observation, and the relationship is startlingly consistent. In the retrospective testimony phase, when a family describes what happened before investigators arrived, the phenomena are spectacular: heavy wardrobes crossing rooms, people carried through the air, objects appearing from nowhere. When an investigator moves in, the phenomena continue but relocate, occurring in the next room, or behind the observer, or in the moments when he steps outside. When cameras are introduced, events become smaller, briefer, and worse-positioned. And under continuous automated recording of a controlled space, the reliable output is either nothing at all or footage of an adolescent doing something with their hands.
The curve operates within cases as well as between them, which is the part that makes it hard to dismiss as a selection artefact. Enfield began with heavy furniture crossing rooms in the weeks before investigators arrived, moved to objects thrown while adults were present but turned away, narrowed to events at the edges of camera coverage once recording began, and produced, under an automated camera in a room the photographer had left, footage of ordinary hands. That is a single case declining in magnitude, step by step, in exact proportion to how well it was being watched, over eighteen months.
That is a dose-response curve, and dose-response curves are how you identify a causal variable. If the phenomena were independent of the observer, improving observation should improve the record, which is what happens with every real phenomenon ever studied: better instruments produce better data on lightning, on animal behaviour, on subatomic particles, on anything. Poltergeist research is close to unique in producing the opposite relationship, where the more carefully you watch, the less there is to see, and that inversion is itself a measurement, of exactly the kind that separates genuine detection from artefact in every field that relies on it, from the training of biological detectors to the laboratory work that separates a real material transition from a spurious signal in the search for exotic states of matter. The curve itself is the most robust data the field has ever produced, and almost nobody has thought to publish it as a result in its own right.
What the Cameras Caught
The specifics are instructive, and Enfield again provides them. Investigators eventually installed a video camera in an adjoining room, and it caught Janet bending spoons and attempting to bend an iron bar. Grosse himself observed her banging a broom handle against the ceiling and hiding his tape recorder. Separately, a remote-controlled still camera set to fire every fifteen seconds, with the photographer deliberately absent from the room, produced a sequence that the investigator Melvin Harris presented as showing the girls pranking. The famous levitation photographs, examined critically, are consistent with an athletic girl bouncing off a bed, and Janet was a school sports champion. Both sisters later told journalists they had faked some incidents, putting the figure at around two percent, and said they had done it partly to see whether the investigators would notice.
The pattern repeats elsewhere. In the Columbus case of 1984, a newspaper photographer left his camera running and captured the fourteen-year-old agent pulling a lamp toward herself at a moment she evidently believed she was unobserved. Amityville was acknowledged by one of the participants to have been constructed over drinks. Borley Rectory, for decades billed as the most haunted house in England, dissolved under scrutiny of the investigator’s own conduct and a resident’s admitted faking. The sceptical case has been argued most persistently by investigators with stage-magic backgrounds, and the Skeptical Inquirer’s examination of the Enfield evidence sets out the specific points at which the documentary record and the paranormal reading diverge, including the observation that what was attributed to supernatural interference with the tape recorders was a known mechanical fault of reel-to-reel machines of that era. It is important to say clearly that none of this makes the children villains; an eleven-year-old in a chaotic household who discovers that a knock on a wall summons adults, journalists, and sympathy is responding rationally to an extraordinary incentive structure. But it does mean that when observation is automated and the human observer removed, the record fills with ordinary hands doing ordinary things, which is the sort of unglamorous documentary finding that decides cases in forensic investigation and in the exposure of long-running deceptions such as the cipher company that was never what it claimed. The cameras did eventually catch something clear and repeatable, and in every instance it turned out not to be a ghost.
Why Investigators Become Participants
The methodological problem here is genuinely hard and deserves sympathy rather than mockery, because poltergeist investigation is structurally almost impossible to do well. The investigator must live in a family home for weeks or months, forming relationships with distressed people, in a situation where the phenomena occur unpredictably and cannot be induced. Controls that would be trivial in a laboratory, such as isolating the subject or restricting movement, are intrusive and often cruel when applied to a frightened child in her own bedroom. So the investigator ends up as a houseguest rather than an experimenter.
There is a further difficulty that has no clean solution, which is that the obvious control is ethically intolerable. The decisive experiment would be to place the suspected agent under continuous observation with their hands visible, in a stripped room, for an extended period, and see whether the phenomena continue. Applied to an adult volunteer this would be unremarkable. Applied to a frightened eleven-year-old in the middle of a family crisis it is something closer to an interrogation, and any investigator who proposed it would be right to feel uneasy. So the one procedure that would settle the question is the one that decency forbids.
Worse, the investigator’s arrival changes the incentives of everyone in the house. Attention arrives, sympathy arrives, and, if the case becomes public, the press arrives. If a child is producing phenomena for attention, the investigation itself supplies a vastly increased reward, which means the standard research method actively amplifies the behaviour it is meant to measure. Then there is the emotional dimension. Grosse and Playfair spent two years with a family in real distress, came to care about them, and were criticised by their own colleagues for compelling the girls to retract a confession of pranking, a response that is much easier to understand as protectiveness than as dishonesty. Investigators embedded long enough with a subject to gain access reliably lose the detachment that made the access worth having, which is a well-known failure mode in institutional oversight and audit, visible in the anatomy of a bank whose examiners saw too little too late and in every case where the record only became clear once outsiders read the documents, as in the great document-driven exposures. Nobody involved in these investigations was a fool, and that is rather the point: the method itself is the trap.
Poltergeists in 2026
Which brings us to a prediction the observation curve makes, and which has been tested more thoroughly in the last decade than in the previous century, without anyone designing the experiment. Contemporary homes are saturated with continuous recording. Doorbell cameras watch entrances. Indoor cameras watch pets and children. Phones sit on every surface, and smart speakers and sensors log activity around the clock, with autonomous devices proliferating alongside drones and robotic systems. These are precisely the conditions under which a genuine, physically real poltergeist should finally be documented beyond argument: multiple angles, continuous coverage, timestamped, with no observer to be shy of.
It is worth being careful about how much weight this can bear, because absence from the record has more than one possible cause. Reporting channels have changed, the Society for Psychical Research is a smaller presence than it was, newspapers no longer fund long investigations, and a family experiencing something strange in 2026 is likelier to post about it than to write to a learned society. Some of the decline is certainly a change in how such experiences surface rather than in how often they occur. But that cannot account for the whole of it, because the modern environment would preserve the evidence even if the reporting route changed, and what continuous domestic recording has produced is not a better class of poltergeist case but effectively none.
The result is that the classic poltergeist case has all but disappeared from the record. There is no modern Enfield, no modern Rosenheim, nothing occupying the position those cases held in their decades, despite a vastly larger population, far better equipment, and enormous public appetite for exactly such a story. What has replaced them is instructive: the modern domestic haunting is overwhelmingly a smart-home phenomenon, in which lights turn themselves on, speakers emit voices, thermostats change, and cameras trigger for no visible reason, all of which have prosaic explanations in networked devices, shared accounts, automation routines, and radio interference. The noisy ghost, in short, did not survive the arrival of continuous recording, and what replaced it was the firmware update.
The Curve Is the Finding
Strip the subject down and the useful result is not a verdict on whether poltergeists exist but a measurement that the field has been generating for a hundred and fifty years without treating it as data. Across the whole literature, the magnitude of the reported phenomena is a decreasing function of the quality of the observation, and the moment of initiation, the single instant that would distinguish a hand from an unknown force, is essentially never witnessed. Those two facts together do not prove fraud in any particular case, and it is worth being precise that they cannot, because absence of good observation is not the same as proof of what happened in its absence.
The practical test follows directly and can be applied to any case, historical or current. Ask whether anyone saw an object leave rest, as distinct from hearing it land or seeing it in flight. Ask what the observation conditions were at the moment of each reported event, and whether the impressive incidents cluster in the poorly observed periods. Ask who was present at every event and who was present at none. And ask what happened to the phenomena when recording improved. Four questions will sort almost any poltergeist case into its component parts.
What they do is locate the burden. Any hypothesis about poltergeists must explain why the effect degrades as the watching improves, and the psychokinetic account has no principled reason for that relationship, while the mundane account predicts it exactly, since a person producing phenomena needs unobserved moments and produces less as those moments become scarcer. That asymmetry is why this belongs among the largely resolved entries in the catalogue of Fortean phenomena, while acknowledging honestly that a residue of well-witnessed incidents in the best cases has never been satisfactorily accounted for. Something real was happening in that house on Green Street: a family was frightened, a mother was overwhelmed, two children were at the centre of a national story, and adults kept arriving to watch. The most remarkable thing about the Enfield poltergeist may simply be that everyone spent two years looking at the wrong thing in the room.