A person in an old house hears a step on the stairs, or sees a figure at the end of a hall, or feels a sudden chill and a strong sense that someone is there. Reports of hauntings include seeing a figure, odd smells, sudden changes in temperature and a strong sense of presence. Each of these is something a person truly felt, and each is a report. Belief is common: a 2001 Gallup poll found about 38 percent of Americans believed ghosts exist, and in a June 2005 Gallup phone poll of 1,002 adults, 37 percent said they believe houses can be haunted. A poll counts what people believe. Evidence asks a second question, which is what else could produce the same experience, and whether anyone can check.

Long ago, on a stage, people made a see-through ghost with a sheet of glass and a hidden actor in bright light. Henry Dircks first made the glass effect by accident, around 1858. John Henry Pepper, who ran the Royal Polytechnic in London, turned Dircks's idea into a stage effect in 1862. The trick was a bright light on a hidden actor, and a big angled pane of glass, so the audience sees the actor's reflection on the stage. That is how a see-through figure appeared, with glass and light and a person hidden from view, and it shows that a convincing figure can be made from ordinary materials.

The Society for Psychical Research was founded in London in 1882, to study such reports with care, and it grew out of a conference held in London in January 1882, and its first president was Henry Sidgwick. Eleanor Sidgwick, a mathematician and later principal of Newnham College, Cambridge, was one of the Society's leading investigators. The Society's Census of Hallucinations asked thousands of people about unexplained experiences, and the SPR says it is still the largest survey of its kind. Its advice to an enquirer is plain: seek a clean account with as much information as possible, and consider every normal explanation first.

Houses make creaks and pops. Wood, metal and glass get a tiny bit bigger when they warm up and a tiny bit smaller when they cool down. Building materials in a house, such as wood, steel, masonry and glass, expand in the heat of the day and contract again at night when it is cooler. Cold pipes can make noise when warm or hot water runs through them. Winter air dries wood out, so the boards shrink, and that is why some floors make more noise in cold weather.

Seeing faces in things that have no faces has a name: pareidolia; people see faces in clouds, rocks and tree bark. The brain area that notices faces gets busy when we only think we see a face, like one in a cloud. The visual system groups small parts, like a mouth and eyes, into a bigger picture, which is how it can build a face out of clouds or bark. One research team says the brain evolved to help us deal with other people, and that shapes how we see the world, so we see faces in objects. Carl Sagan and others suggested that spotting faces fast helped early humans stay alert to others nearby. This is an idea, not a settled finding.

A hill on Mars looked like a face in one old photo because of shadows; the hill is 1.5 kilometers (one mile) across, and the Viking photo was taken with the sun about 20 degrees above the horizon. The speckled look of the first Mars picture came from bit errors in the data, made more obvious when the photo was enlarged. NASA says the hill looks like a face only in low-resolution pictures with very specific lighting. The hill has dozens of hills like it nearby, and sharper photos showed it is just a hill.

On a mountain, your own shadow can fall on fog or cloud below you; a mountain climber's shadow on cloud can look huge and have a colored ring around it, yet it is only the climber's own shadow. The Brocken spectre is an observer's shadow cast on fog or cloud, often with a ring of colored light around it called a glory. Researchers note that surprisingly realistic imagery can be seen on the threshold of waking. Sleep paralysis is when you cannot move or speak as you are waking up or falling asleep. It can be scary, but it is harmless. During sleep paralysis some people feel as if someone is in the room, and the feelings can last up to several minutes.

People generally hear sounds between 20 Hz and 20,000 Hz; infrasound is below that range: it is sound with a pitch below 20 hertz, lower than the sounds people usually hear. Infrasound is wrongly thought of as completely inaudible, because people can perceive it when it is strong enough. Infrasound comes from nature too: volcanic eruptions, earthquakes, falling meteorites, tornadoes and lightning make it. A scientist named Vic Tandy traced eerie feelings in a lab to a faulty extractor fan that made very low sound. In 1998 Tandy and Tony Lawrence proposed that a standing air wave of 19 Hz may create the sensations people take for a ghost. Tandy later reported that infrasound of about 19 Hz had been found in a Coventry cellar where people had reported odd feelings, and he noted that only a minority of people are sensitive to the effects of infrasound.

The Ganzfeld effect happens when the eyes see a uniform, featureless field and the brain starts filling in the blanks on its own; in a Ganzfeld session, people may see zigzag lines, dots or a blob of color within a few minutes, and after 10 to 20 minutes may lose sight intermittently. The explanation offered is that with nothing to see, the higher visual cortex amplifies whatever information is available. With no incoming signal the brain keeps searching for input, which is why people with healthy eyes can see patterns that are not in the room. Researchers report that prior expectations shape perceptual decision-making: what we expect to see changes what we decide we saw. In an fMRI study, confidence in a perceptual decision rose as expectations increasingly supported it, a signal tracked in the right inferior frontal gyrus.

The 2003 Wiseman study had two parts: Hampton Court Palace and the South Bridge Vaults in Edinburgh; the Hampton Court part used 462 usable questionnaires from members of the public in 36 groups, and they reported 431 unusual experiences. The visitors who took part were members of the public who chose to join, tested in groups walking the rooms. The investigators did not know which areas were labelled haunted until all data had been collected, which prevents their expectations leaking into the results. At Hampton Court Palace, people who already believed in ghosts reported significantly more unusual experiences than people who did not. In the Hampton Court study, believers in ghosts also were more likely to attribute their unusual experiences to a ghost.

At Hampton Court, of 215 people who reported something, 8 said it was definitely a ghost, 22 probably, 80 were uncertain, and about half said probably or definitely not. About two thirds of the unusual experiences people reported at Hampton Court were a change in temperature. The average strength of the magnetic field did not differ between haunted and control areas, but its variation was larger in the haunted areas, with a variance of 12.71 against 2.16. The authors say the Hampton Court study had low statistical power because visitors had to be analysed as groups, so they tested again in Edinburgh with people alone.

The Edinburgh part took place in April 2001 during the Edinburgh International Science Festival, with each visitor spending about 10 minutes alone in one vault; in Edinburgh, 95 of the participants reported at least one unusual experience in the vaults, 172 experiences in all. The vaults were ranked from most to least haunted. That ranking matched how many experiences visitors reported, even when visitors who knew the reputation were removed, with a correlation of rho = .87. The study found significantly more reports of unusual experiences in areas with a haunted reputation, and this was not explained by what visitors knew in advance. The authors said the findings strongly suggest that alleged hauntings may not necessarily be evidence of ghostly activity, and could be, at least in part, people responding to normal features of their surroundings.

In 2003 a London concert played infrasound under some pieces, and neither performers nor audience were told which; the Purcell Room concerts were counterbalanced: infrasound was added to different pieces in each of two concerts, and nobody in the hall was told when. The infrasound at the concert was set at 17.5 hertz, and was only played during certain pieces. About 22 percent of the audience reported more unusual experiences during pieces with infrasound. Wiseman said the results suggest that low frequency sound can cause people to have unusual experiences even though they cannot consciously detect infrasound. The concert's co-organizer called the result a tantalising but inconclusive result and said there was likely considerable priming and suggestion at play.

Walking from a well-lit corridor into a much darker vault may give mild sensory deprivation, which can cause unusual feelings like those in the Ganzfeld; the authors suggest vault features may match a person's idea of a haunted place and so bring on mild suggestion effects. The Wiseman team said future work should separate competing interpretations: suggestion by the setting, or direct effects of light, dark, size and field changes. They also noted that the pattern might come from another factor that varies with the rooms, such as the production of unusual shadows. When the 2003 paper appeared, well-controlled systematic research on high-profile haunting claims was very scarce. A person who reports a spooky feeling did feel something, and nobody has to be mocked for that. In 2003 the team listed what a later study would need to separate, and the possible causes they named were suggestion, light, dark, size, field changes and shadows.

Words to know

• priming: getting ready
• Ganzfeld: blank-field effect
• pareidolia: seeing faces
• infrasound: very low sound
• suggestion: idea from setting
• census: big count