Pictured above: 'Since childhood I was a dangerous type of mechanic' says Indian neuroscientist Champadi Raman Mukundan, who developed the BEOS test.

People are being jailed after lie-detecting brain scans find them guilty. The science is flaky, but this is just the latest instance of neuro imaging being used to 'read' the human mind - and even acclaimed studies are now being challenged as spurious.

To Judge Shalini Phansalkar-Joshi, sentencing her last June to life in prison, Sharma's electro-encephalogram left no doubt: the brain scan revealed "experiential knowledge" which proved that she had to be the killer. Her ex-fiancé Udit Bharati, a 24-year-old fellow student at Pune's Indian Institute of Modern Management, had been found dead after eating sweets laced with arsenic. And Sharma - who had eloped to Delhi with her lover, Pravin Khandelwal - had, according to the prosecution, returned to Pune and lured an angry Bharati to meet her in a McDonald's, where she had given him the poison.

As the judge saw it, the proof was in the science. Sharma had manifested an undeniable "neuro experiential knowledge" of the crime - which the brain could acquire only through direct experience - when she had undergone a brain scan in Mumbai a year earlier. That day, July 25, 2007, 23-yearold Sharma waited while a technician squeezed conductive paste through a syringe into a small button on the red skullcap sitting tightly on her head. This was repeated for all of the 30 wires protruding from the cap, each making a connection between her brain signals and the blue electro-encephalograph (EEG) machine on the table behind her.

An armed police officer waited in the corridor outside, where the air was thick with mosquitoes. Inside a windowless room labelled "Brain Electrical Oscillations Signature", it was a dry 18°C. The air-conditioning was on cool to ensure that the expensive equipment did not overheat. Cloaked in an oversized white apron to keep her warm, Sharma sat alone in a wooden chair, observed through one-way glass.

A tape played a voice reading a series of statements in Hindi, each detailing an aspect of the murder as the investigators understood it. Sharma said nothing as the EEG machine measured her brain activity.

For a while, the statements elicited no detectable EEG response.

Then she heard: "I had an affair with Udit." A section of her brain previously dormant registered a brightly coloured response on the EEG. More statements followed and the voice on the tape each time elicited similar EEG responses: "I got arsenic from the shop." "I called Udit." "I gave him the sweets mixed with arsenic." "The sweets killed Udit."

Throughout the test, she did not say a word. She didn't have to. As each statement was read, the EEG machine measured the frequencies of the electrical signals from the surface of her scalp and fed them through a set of rainbow-coloured wires into the room next door. Here a computer, almost five feet tall, performed a set of calculations and spat out its conclusion in red letters on to its screen: "Experiential knowledge". This meant knowledge of planning the murder, of getting the sweets, of buying the arsenic and of calling Bharati and arranging the fatal meeting. Guilty.

Evidence from the scan took up almost ten pages of the judge's ruling when a year later, on June 12, 2008, he jailed Sharma for life - making her the first person in the world reported to be convicted of murder based on evidence that included a brain scan. "I am innocent and have not committed any crime," she implored Phansalkar- Joshi before he sentenced her. Even he, her lawyer said, had trouble believing that this small, calm, softly spoken student, from a respectable, middle-class family, was capable of killing.

But science had spoken: and in the six months that followed, the same lab would provide evidence that convicted two more people of murder. Neuroimaging as truth teller had come of age.

T he laboratory of the Directorate of Forensic bet365体育赛事 in Mumbai has been running Brain Electrical Oscillations Signature (BEOS) tests on criminal suspects for two years. Business is good: when Wired visits, another room is being added to accommodate a second EEG machine, which sits covered in bubble wrap. "We consider the brain as a computer, where information is stored and can be retrieved," explains Sunny Joseph, the lab's 33-year-old assistant chemical analyser. The psychology department has two other staff members - both in their twenties, both rushed off their feet, with case after case being sent by the courts. "Referral rates have been really high," Joseph adds. "We do possibly 15 cases a month." A growing heap of brown-foldered case reports sit in the corner.

The BEOS test was developed by the Indian neuroscientist Champadi Raman Mukundan. The software, Joseph explains, was designed by collating data from earlier research on memory and translating this into a set of 11 physiological variables. Mukundan's program uses the frequencies and voltages produced by an EEG - which measures slight fluctuations in brain activity caused by neurons firing electrical signals between one another - to determine the results of each of these variables. If all 11 are positive, then the statement being read out to the suspect is assumed, by Mukundan's theories at least, to be true.

Mukundan, for one, sees no room for doubt: his 20-page patent application for an "Electronic Investigative Device for Identifying Truth", filed on Valentine's Day 2007, explains how it can be used "for investigation of truth from individuals who have committed an act of offence" by "advantageously utilis[ing] the experiential knowledge present in a subject's brain that elicits a bioelectric response".

Aditi Sharma would have been told by police officers about the crime of which she was accused, Joseph explains, but unless she had in fact participated, the test would come up negative. That was because her memory of the crime was hard-wired in her brain as experiential knowledge. "We are sitting and talking here. This is an experience for me. This is an experience for you," Joseph says, pausing for emphasis between sentences. "Now you go and tell your friend. Whatever we discuss here, you can only impart knowledge of this experience. Your friend can never have this experience unless she comes and sits here. This is how it works."

Confessions have been made in at least ten of the 75 or 80 BEOS tests so far conducted, Joseph says. Had other subjects also been accused of murder? "Yes, most of them murder, yes."

He rotates in his swivel chair and looks at the brown wooden chair in the otherwise empty white room. "They are so, so relieved to be here. They're so happy to be here with us, because we're not scary.

We talk to them nicely. Just imagine... You can imagine in India the way the police must be dealing with them."

A colleague of Joseph's later points out that brain-imaging allows an overstretched police force to speed up the conviction process by eliminating innocent suspects from their enquiries and by corroborating evidence. That is why Mumbai is not the only Indian city to have invested in BEOS technology. The government's forensic science directorate in Gandhinagar, in Gujarat, has been using it since 2003 and has now tested 163 subjects in 88 criminal cases.

Support came directly from India's chief forensic scientist, Dr MS Rao. "The technique has great potentiality to become an infallible tool in crime investigation," he wrote in a paper presented to the All-India Forensic bet365体育赛事 Conference in January. "It can become a revolutionary technique like DNA fingerprinting if its evidential strength and judicial acceptability are established." A third such facility opens soon in the northern Indian city of Chandigarh.

And if private enterprise has its way, courtrooms around the world will soon be convicting their own Aditi Sharmas based on brain-imaging evidence. Since last year, Cephos Corp of Massachusetts has been marketing what it calls "commercially available fMRI-based deception-detection services", based on software analysis of subjects' MRI brain scans (the "f" stands for the "functional" interpretation of the scans). In a study of 61 people, Cephos claimed better than 90 per cent accuracy in determining deception. In California, meanwhile, No Lie MRI charges clients $4,000 to $5,000 (£2,750 to £3,500) to conduct lie-detection tests (though neither firm would tell Wired how many it has sold).

Although US and British courts do not admit brain images as evidence, both companies have ambitions in that direction. Steven Laken, CEO and president of Cephos, explained that a third of his "customers" are convicted prisoners seeking lie-detection evidence on which to base an appeal. He added that he was "confident" that lie detection evidence based on brain scans would one day enter America's legal system. No Lie MRI's CEO, Joel Huizenga, said that he hoped to open a series of brain-imaging centres in the UK - after which he would try to get the scans accepted in British courts. Meanwhile, defence lawyers in a current case in San Diego hope that a brain scan by Huizenga's company will prove that their client, an accused sex abuser, is innocent.

In May 1991, Dr Kenneth Kwong, a radiologist at Massachusetts General Hospital, found that he could use an MRI scanner to detect small, local changes in the oxygenated blood flowing through his subjects' brains. When he showed the subjects visual stimuli, such as flashing lights, he detected activity in their visual cortex as they processed the information, activity that reflected changing short-term oxygen levels in these parts of the brain. Measurement is possible because the oxygenated blood cells contain increased amounts of the iron-rich protein haemoglobin. Iron has magnetic properties, which can be detected by the MRI's magnet - the more oxygenated the blood in a specific part of the brain at a specific time, the brighter that area of the brain scan will be.

The following year, Kwong published his influential research on this "blood-oxygen level dependent" (or BOLD) brain imaging in Proceedings of the National Academy of bet365体育赛事s. And so began the modern era of functional MRI, or fMRI, scanning. Along with EEG imaging, fMRI has given neuroscientists a persuasive new tool: by letting them map local brain activity in response to experimental stimuli, they can claim to demonstrate particular thought processes at work. The tool has been vastly popular: in 1992, four peer-reviewed papers were published that mentioned fMRI, but by 2007 the total had reached more than 19,000. In the published literature, brain imaging was being used to prove the "neural correlates" of hate, trust, romantic love, moral sensitivity - even, in a famous Coke versus Pepsi taste test, a behavioural preference for "culturally familiar drinks".

Many of these studies have confirmed long-held assumptions about human behaviour. For example, that cliché about men being less attentive than women? In 2000, a neuroradiologist at the Indiana University School of Medicine showed that men process language mainly with the left side of the brain, whereas women tend to use both sides. Ever noticed that school bullies seem to get a kick out of kicking their victims? A study by researchers at the University of Chicago demonstrated that the brain's pleasure centres were activated in aggressive males when they saw other people hurt. How does a woman spot a potential partner? An fMRI experiment by a psychologist at Rice University in Texas confirmed assumptions that women can recognise a man's sexual intentions by smelling the pheromones in his sweat. Our thoughts, it seems, are increasingly able to be revealed through brain imaging.

But what if the science behind brain imaging analysis is in fact fundamentally flawed? Enter Edward Vul. Last year, the 26-year-old graduate student took it upon himself to challenge the basic science behind fMRI studies and, in the most devastating terms, describe how the "cherry-picking" of data was creating "inflated estimates of correlations". Just before Christmas, Vul, based in the Department of Brain and Cognitive bet365体育赛事s at MIT, submitted a research paper (with Christine Harris, Piotr Winkielman and Harold Pashler from the University of California) called "Voodoo Correlations in Social Neuroscience" to the journal Perspectives on Psychological bet365体育赛事. The paper was not due to be published until about now but Vul released it online over the Christmas break.

Although others had previously raised concerns about the veracity of some brain-imaging studies, Vul and his colleagues were withering in their rigorous criticism.

He suggested that scientists were producing results that stretched the limits of brain imaging beyond statistical probability.

Frequently, he said, people claimed that scans showed correlations with human behaviour or emotion, with the corresponding brain activity on an fMRI scan close to or above 80 per cent, a figure that Vul thought impossibly high. "These correlations often exceed what is statistically possible assuming the (evidently rather limited) reliability of both fMRI and personality/emotion measures.

The implausibly high correlations are all the more puzzling because method sections rarely contain sufficient detail to ascertain how these correlations were obtained." He and his fellow researchers, who examined 54 studies, were hailed as "methodological whistle-blowers" by other sceptics. One high-profile piece of research, published in 2003 in the journal bet365体育赛事, had claimed that activity in the anterior cingulate cortex - an area in the centre of the brain that has been linked to the negative feelings associated with physical pain - increased when people suffered rejection. The lead author was Naomi Eisenberger, a PhD candidate in social psychology at the University of California. She had subjected 13 undergraduates to fMRI scans while they played an electronic game of catch with two other people. The rejection element came in when the virtual players, which the subjects believed were being controlled by real people, deliberately excluded them and passed the ball to each other. The painful, gut-wrenching feeling you get when you're ignored, the research suggested, is not imaginary. As far as the brain is concerned, it is a genuine, physical pain.

It sounded plausible. But what sparked concerns in Edward Vul's mind was the claim that the strength of the relationship between the feeling of rejection and the activity in the anterior cingulate cortex was as high as 88 per cent. This, to him, seemed mathematically unlikely, for two reasons.

First of all, emotions, in themselves, are hard to measure in numbers. It's not a precise science to give a feeling of sadness, or rejection, or love, or - and this is key - truth a value between 1 and 100. When you make a correlation between two things, Vul points out, its accuracy is a function of the accuracy of the original figures. If one figure is accurate itself to only 60 per cent, the correlation itself can never be more accurate than that.

Previous research by psychologists had shown that, at best and using the most trusted and accurate tests of mental state or personality, those tests would be between only 70 and 80 per cent accurate.

Then come the brain scans themselves. Brain scans, Vul explains from his office at MIT, are divided into between 40,000 and 500,000 "voxels" - similar to the pixels in a computer screen, but in three dimensions. Some voxels indicate more activity than others, depending on which part of the brain they represent, and the aim is to spot which ones display the most activity. In Eisenberger's paper, for example, researchers were looking for a response in the anterior cingulate cortex. The problem is that not all of these tiny voxels give a true picture of the brain. Some are skewed because of noise. Experiments usually correct for these kinds of errors, but Vul charged Eisenberger and others with cherry-picking a small number of noisy voxels that matched their conclusions. "So, based on this procedure, we would be misled to believe that one could predict, for example, 80 per cent of the variability in behaviour by measuring a specific part of the brain. But in reality that number is probably much lower," Vul says.

The authors of the paper disputed his criticisms, but accepted that the strength of the links shown by studies such as theirs may be inflated. The very best correlation that researchers could hope to achieve between a human behaviour or an emotion and a brain scan, Vul and his colleagues said, was 74 percent - and even this assumed that the relationship between the two was perfect. "Over half of the investigators in this area," they wrote, "used methods that are guaranteed to offer greatly inflated estimates of correlations."

Long before the hard copy of "Voodoo Correlations", its impact resounded online. Many science writers and bloggers wrote about it positively, including the widely read Sharon Begley of Newsweek.

Vaughan Bell of the Mind Hacks blog predicted that the paper had "the potential to really shake up the world of social cognitive neuroscience". And it did: after all, Vul had the audacity to name check specific studies, many of which had been published in leading journals such as Nature and bet365体育赛事, and to call some of them "entirely spurious".

Shocked by the speed with which this paper was being disseminated and discussed, two groups of neuroimaging scientists wrote rebuttals and posted them on several blogs. Vul followed up, linking his rebuttal to theirs. The discourse, initiated at breakneck pace, continues.

For Adita Sharma, such dialogue now offers at least some hope for a belated challenge to the legal "proof" of her guilt. "Suddenly there's been a burst of these cases where the police have used lie detectors," her lawyer, Revati Dere, says despairingly. "Somewhere, someone down the line should understand it's the human mind that you're talking about. It can't be tested with that much accuracy. It's an easy solution, a very easy solution. A short cut. "Two different people could react to it differently. You can't say that you're testing the human mind on the basis of 'that does this'. People will undergo stress and parameters will go haywire."

Her eyebrows furrow. "She [Sharma] must have been under tremendous stress at that moment, when she was made to undergo the test."

In September 2008, a report by a committee at India's National Institute of Mental Health and Neuro bet365体育赛事s (NIMHNS) declared that brain scans of criminal suspects were unscientific. The committee warned that they should not be used as evidence in a court of law. Following Sharma's conviction, Dere says, she immediately lodged an appeal with the high court, complaining that the brain-imaging tests conducted by the scientists in Mumbai were "bad science". Six months later, Sharma was released on bail pending that appeal.

Dere hands me a bunch of papers about the case, including Sunny Joseph's expert testimony. "So, do you know how this brain mapping thing works?" she asks.

In all honesty, despite months of research, I could not say that I understood the inner workings of the software being used at the Mumbai lab. So I took the question back to Sunny Joseph. But Sunny Joseph won't tell. The technology is under patent, he explains. He cannot leak trade secrets.

Perhaps the inventor of BEOS himself, the Bangalorean scientist Champadi Raman Mukundan, can? "He knows so much about the brain... and he is so different from most others," Joseph eulogises, his eyes lighting up when talking about him. "He made his own EEG machine. There was no equipment, no funding. This was about 30 to 40 years back. Because he was very good at physics and electronics and things like that."

We track down Mukundan in Gandhinagar, in the neighbouring state of Gujarat. The visionary who dared decipher the human mind - indeed, who built a machine to automate the process, and then convinced the Indian government to use his system in its courts - is, in person, a short, shy 67-year old with salt-and-pepper hair. "They think they can sit and fool us," Mukundan laughs, squeezed into his chair at the offces of the Gandhinagar Forensic Laboratory, as he looks through the one-way glass into the room into which suspects are brought for testing. He's talking about the criminal suspects he has seen in the lab. One arrived for a brain scan having drunk a bottle of cough syrup, in the hope that this would fool the EEG (he had to retake it later). "In my family, they say that I was this boy born with a screwdriver in my mouth," he explains, sipping a cup of tea fetched by a student. "Since my childhood I was a mechanic, a dangerous type of mechanic. I used to get punished every day. If I found I needed a plank of wood or something, I'd go and remove the plank from the back of the wooden wardrobe. Suddenly my parents open the wardrobe and find there is no..." He trails off into a belly laugh. "So nobody understood why I took to psychology. That seemed to be against my spirit."

Forty-three years ago, Mukundan switched to studying psychology after reading physics and mathematics. He was, by his own admission, one of the weakest students in his year. Unlike the others in his postgrad clinical psychology class, Mukundan hated meeting patients. He preferred the laboratory, where he could tinker with electrical equipment. For his PhD project, he hooked up schizophrenics to a galvanometer to measure the resistance across their skin - a possible indicator of emotional stress - when a small current was applied.

His career in brain research began in 1974 at NIMHNS. Then in 1979 he founded India's first electrophysiology laboratory to study the electrical variations associated with the body's physiological changes. He patched together old bits of kit for five years until he understood it. He recalls proudly that he became known as the crazy guy on campus, the maverick who fascinated the students and attracted curiosesity from his colleagues. In later years, they would criticise his invention. Mukundan will not disclose the inner workings of his brain-imaging software. His decision not to publish his research or subject his ideas to peer review has prevented others from verifying his results. But he does not care, he says, because he would rather his peers condemned him than he lost control of his invention before it is patented.

Mukundan pored over decades of literature about how the brain processes memories before starting work on the BEOS test. "See, I am basically an electronics man, not a psychology person. So it was easier for me to conceptualise the neural organisation." He claims that he saw things that other psychologists and neuroscientists could not - the different functions of the brain as the elements of a circuit board. Unlike his colleagues, he has always been a "hardcore reductionist", he says - paring down the human brain to its nuts and bolts. "One day we will have an explanation for everything. And we may even be able to control many things." He pauses. "I don't believe in a soul," he says. He is that rarity in India - an atheist. "There's this famous biblical saying that God created man after his image, in the image of God. And man later said that he created the computer after his own image - the brain."

When it comes to understanding the brain, the cult of reductionism has a long history. The world's first phrenologist, Franz Gall, born in 1758, believed that feeling around the surface of the skull would give him a better understanding of the mind inside (a bulge, for example, indicated that part of the brain was particularly well-developed). In 1791 he published a study which hypothesised that the brain comprised around 30 separate organs, each relating to different faculties, including digestion and memory.

His pseudoscientific theories were discredited by the end of the 19th century. Trying to read someone's character through the swellings and indents in their head was no more reliable than palmistry.

The underlying concept that physiological traits can be localised in different glands, nerves and groups in the brain, however, has survived. Researchers studying brain-injury patients have proven some links between mind and matter. In his 1998 book Phantoms in the Brain, brain researcher VS Ramachandran described a patient who had suffered a stroke in the right half of her brain. Afterwards, her family discovered she would take the same time and care over perfectly arranged hair, painted nails and make-up as ever, but only on the right side of her body - her left half was ignored. And if a plate of food were placed in front of her, the left half would be untouched. And when asked to draw a flower, she sketched only the right half.

The reason for the problem, Ramachandran learned, was that the stroke had damaged her right parietal lobe, the part of the brain near the top of the head that helps us recognise the layout of our environment. This had caused "hemineglect", which prevented her from noticing the left half of her world.

The right parietal lobe is only the start. Neuroscientists know that the hippocampus is linked to how memories are formed and retrieved; that the lateral hypothalamus is the corner of the brain that makes us feel hungry - it contains nerve cells that are sensitive to glucose levels; and we also know that the posterior nucleus of the hypothalamus controls our body's response to cold (if it is damaged, you get hypothermia).

In Phantoms, Ramachandran pinned down people's claims of religious experiences to epileptic seizures in the temporal lobe (journalists dubbed it the "God spot").

As for Mukundan's claims about what he can tell about the brain, they leave some senior neuroscientists utterly unconvinced "There is nothing in the history of brain imaging to say that we could ever get the degree of precision needed to detect lies," says Geraint Rees, professor of cognitive neurology at University College London. He laughs when I tell him that brain scans are being used as lie detectors in Indian murder cases.

"The technology is only at the research level, helping us to better understand the brain," he says. "I find the commercial use of brain scans more worrying because it taps into the public perception that the technology is more advanced than it really is.

It's like saying, 'Let's find a machine that could find out the physical state of my computer'," he says. "Let's say some device could read all the files on my computer just by measuring the activity on its case. Well, we're not even remotely close to having that - and the brain is of course several orders of magnitude more complicated."

Aditi Sharma was on remand for a year before she was convicted and she then spent another six months in jail before she was bailed.

India's slow judicial process means that it could be a long time before she is back in court. "It can take anything between five and ten years," says Revati Dere. When Sharma is finally back in court, her only hope will be if one of India's higher courts rejects the brain scan evidence. She declined to be interviewed for this article but her father said, via her lawyer, that she wants only to get on with her life.

So what does Mukundan feel about the woman whose life hangs in the balance because of his invention?

Sitting in the empty forensics lab in Gandhinagar, his BEOS machine on the floor beside him, he is philosophical. "Man is not destined to be controlled by nature. Man is destined to control nature," he says. "This is the big departure between man and the animals. Human beings are destined to create a nature and then live in that nature."

But does he not see any conflict between his scientific views and the complexities of human psychology? Could he perhaps have got the human mind wrong? "They are not in conflict," Mukundan replies sharply. He interrupts an attempt to challenge him: "I found that there is no conflict."

This article was originally published by WIRED UK