THE mind may well be science鈥檚 final frontier. For a century, psychologists and neurologists have recorded people鈥檚 behaviour, invented abstract computer models of mental processes and chopped up dead brains in their quest for clues about the nature of consciousness. But despite all this effort, there have been few returns. Indeed, when it comes to tracing the biological roots of consciousness, neuroscience has a long tradition of being decidedly downbeat about its progress.
But that is changing. Spurred on by advances in brain-imaging techniques, researchers are becoming bolder. They envisage a new era in mind research dominated by machines which use radioactive tracers, magnetic resonance effects or faint electrical signals to map the activity of living brains. There is talk of reaching inside brains to 鈥渇ilm鈥� the trails of neural activity blazed by flitting thoughts and feelings. Brain-scanning systems are even being likened to the accelerators and colliders of particle physics.
It鈥檚 no secret, for instance, that in 1991 CERN came close to creating a brain imaging supercentre as part of its sprawling empire near Geneva. The result would have been one gigantic laboratory studying both the mind and the Universe, but in the end the idea was abandoned. Some blame egos and politics; others point to funding difficulties and a realisation that the field was still rather immature for such an ambitious leap into the unknown.
Advertisement
But such problems haven鈥檛 frustrated the ambitions of imaging researchers in Britain. Far from it. This year, the UK will have its own imaging centre 鈥� a 拢20-million laboratory equipped with the latest brain scanning machines based at the Institute of Neurology in London. Funds are coming from charitable organisations including the Wellcome Trust and the Leopold Muller Estate.
In the past, imaging laboratories have concentrated on medical research, using scanning mainly to study disease and ageing processes. But with an imaging laboratory devoted purely to psychological research, the hunt for the secrets of consciousness can begin in earnest. Similar moves are afoot at other scanning centres, such as the University of California at Los Angeles. Will the mind sciences finally enter an era of big science? Can brain-imaging systems really become the particle colliders of psychology?
Certainly, there are similarities in construction. Brain scanners use the same superconducting magnets and supersensitive radiation detectors as particle accelerators. They both need heavy-duty computing to capture and reconstruct quantum-scale events. Some scanners even need their own dedicated cyclotron, a particle collider, to create the radioactive isotopes used to measure the surges in brain blood flow and glucose consumption that occur when we think.
But there are two key questions to ask about those surges. Just how accurately can scanners measure them? And what, if anything, can they reveal about the biological basis of thought? Critics claim that some early results were hyped when they were published in the late 1980s. Several times, claims made in prestigious journals such as Science later had to be qualified or withdrawn as red-faced imaging pioneers were forced to realise that no new research technique is ever as straightforward as it seems.
Scanners exploit the fact that when the brain goes to work, performing the tasks that create our subjective world of thoughts and awareness, it consumes energy. Indeed, the 50 billion or so nerve cells which make up the human brain are so metabolically active that while the brain accounts for just 2 per cent of our body weight, it demands nearly 15 per cent of our blood supply and a quarter of all the oxygen we breathe. Being conscious may feel effortless, but in fact it is on average the single most energetic thing we do.
What this means is that every mental event leaves a noisy trail. Not only are electrical and magnetic fields generated by the firing of nerve networks, but every wave of firing is accompanied by telltale surges in glucose consumption, neurotransmitter activity and local blood flow.
Imaging systems can track these microscopic fluctuations in several ways. The best-known method is positron emission tomography (PET), which uses radioactivity to label blood, blood sugars, or important neurotransmitters such as dopamine. The tagged substance is injected into volunteers while they lie in a scanner carrying out a mental task. They might be solving a visual puzzle or judging the strength of a tap on the hand. A ring of detector crystals encircling the head then picks up the emitted gamma rays to create a map of the isotope concentrations.
In theory, scanning sounds simple. But the task is actually Herculean. A PET scanner has to sift through 7 or 8 million gamma-ray signals every second to locate concentrations of the tracer. It takes at least a couple of hours to set up and run one subject through a scanner, meaning that a laboratory is lucky to get through four experimental sessions in a day. It then takes a further 36 hours of number crunching on powerful computers to reconstruct the brain images from a single session. A surprising number of support staff are required. From producing the short-lived radioactive isotopes in a cyclotron to the final polishing of the data, the process involves about 15 people.
PET is the best established of the many imaging techniques but others look even more promising. One of these is functional magnetic resonance imaging (functional-MRI), which uses a powerful magnetic field to align the tiny magnetic dipoles of atomic nuclei in the brain. The particles are then probed with carefully tuned radio pulses and at certain frequencies they resonate, revealing the concentrations of elements, such as hydrogen, in different regions. Using this method, it is possible to measure the change in oxygen levels, and hence blood flow, that occurs in capillaries supplying active brain tissue.
The great advantage of functional-MRI over PET is that while only a few dozen hospitals in the world can afford their own isotope-producing cyclotrons, thousands already have MRI body scanners that could be converted into brain machines for just 拢500 000 or so. 鈥淚t鈥檚 imaging for the masses. Soon every Tom, Dick and Harry with a big magnet is going to be doing psychological activation studies,鈥� says one prominent PET specialist.
A third imaging approach is magnetoencephalography (MEG), which uses delicate liquid-helium cooled superconducting sensors to pick up the faint magnetic fields generated by active nerve networks. These fields are a billion times weaker than the natural field produced by the Earth. MEG machines are so sensitive that even the moving metal of traffic in the street or a nearby lift can upset their readings. So, despite the use of rooms shielded by wire cages, experiments often have to be conducted in the dead of night to minimise interference. Even so, many believe that MEG 鈥� especially when used in combination with the more traditional electroencephalography (EEG), which monitors the brain鈥檚 far stronger electrical fields 鈥� will eventually prove to be the most powerful technique.
All three methods of measuring brain activity are improving rapidly. In the late 1980s, PET produced blurred images with a resolution measured in centimetres. Worse still, more than half-an-hour鈥檚 worth of recording was required to produce an image; bursts of activity at different times during this period were inseparably merged. PET images could give a broad-brush picture of which regions of the brain 鈥渓it up鈥� while speaking or moving a hand, but that was about all. Now, however, PET resolution is down to millimetres and tens of seconds. Functional-MRI should do better than this, and eventually MEG-EEG machines will offer a millisecond by millisecond picture of brain activation.
This promise of pinpoint clarity is what makes it legitimate to talk about reaching inside the head and filming human thought. Harvard psychologist and enthusiastic scanner convert, Stephen Kosslyn, says that it should be possible to ask someone to form a mental image and then go in and capture that as a picture on a computer screen.
Visual images are thought to begin as picture-like representations in the primary visual cortex of the brain. Etched out in a pattern of firing nerve connections is a mapping of the scene captured by our eyes. While completing the act of seeing 鈥� or constructing a mental image 鈥� involves at least another dozen levels of processing in surrounding areas of the brain, there is no reason why a scanner should not be able to grab a snapshot of this representation. In fact, Kosslyn has already made a start by measuring the changes in brain surface area activated in subjects who were asked first to imagine something very small, and then imagine the same object filling their whole field of view. Even with the relatively low resolution of present-day PET machines, Kosslyn has been able to demonstrate detectable differences.
Some critics have dismissed this kind of experiment as a party trick, arguing that other areas of the brain, such as the areas associated with hearing, smell or more abstract mental processes, are hardly likely to yield such an obvious correlation between patterns of nerve connections and subjective experience. The imaging of a visual impression is not going to advance our understanding of consciousness very far. However, Kosslyn believes that the power of brain imaging is that its evidence is hard to argue with. 鈥淚t鈥檚 not abstract. You can do the experiment and discover where things are happening in the brain.鈥�
Public health warning
Yet even optimists such as Kosslyn agree that brain imaging ought to come stamped with a public health warning. The colourful brain maps that adorn so many current research papers are used as 鈥渉ard鈥� evidence to back up whatever theory the researcher happens to hold. But the creation of every image involves many steps and there are a hundred ways that error, or even plain wishful thinking, can creep into the process.
Some of the blunders of the past few years include the mistranslation of a statistical program used to separate PET signals from background noise. A decimal place was omitted, leading to a serious overestimation of activity. Yet because the experiment appeared to produce the expected result, the mistake was not spotted for some time.
Then there are the more routine problems of interpretation. A hotspot of activity in the brain, for example, could just as well be a flush of inhibitory nerve messages as a pattern of excitation 鈥� to a scanner, an 鈥渙ff鈥� signal looks much the same as an 鈥渙n鈥� signal. Worse still, there is evidence that as the brain becomes practised in a skill or process, the network of nerve sells needed to perform it shrinks. Unnecessary connections are pruned, making the pathway more efficient. An apparently busy area of brain could just be a zone that is struggling with something new or unfamiliar; the real work could be taking place in areas that look silent.
The two other imaging techniques, functional-MRI and MEG, bring extra problems. This year critics have been arguing that functional-MRI may be inherently unreliable, as even the smallest body movement 鈥� the natural tremor of neck muscles or pulse in the forehead 鈥� can create a flurry of false readings.
MEG has an equally fundamental problem in that there is no single correct interpretation of an image. Several brain states could give much the same result and the experimenter has to chose the one that seems the most likely. Sometimes the images show activation in unexpected locations 鈥� even outside the skull. Researchers using MEG feel that it is reasonable simply to ignore such obvious recording glitches, but the wider world of science is often aghast at such a cavalier attitude to data. 鈥淢EG images can be a shocking mess. It鈥檚 not good enough to see this cloud of activation outside someone鈥檚 head and tell people simply not to pay any attention to it,鈥� says Peter Mansfield, professor of physics at the University of Nottingham and a founding father of scanner technology.
Karl Friston, a statistical analysis expert at the Medical Research Council鈥檚 (MRC) brain imaging unit at Hammersmith Hospital in London, estimates that only a third of all the scanner studies reported so far are going to stand the test of time. A third more, he says, are downright flaky. The rapid spread of functional-MRI over the next couple of years could compound the problem. The ease with which standard hospital MRI scanners can be converted into brain activation machines may mean a hundredfold increase in imaging research and some people fear there could be a flood of poorly controlled work from novice investigators.
However, Friston believes the teething troubles of imaging should not be exaggerated. Many of the problems of correct location and interpretation are fast disappearing as the resolution of machines improves and the processing of results becomes more standardised. Besides, says Friston, most of the troubles of brain imaging are born largely of its own success. The field is only a few years old, yet because of the sheer pace at which the technology is developing, imaging researchers are already mounting frontal assaults on areas about which the old school would hardly dare even speculate.
Eric Reiman, associate professor of psychiatry at the University of Arizona, says many traditionalists are horrified by the way imagers are leaping in at the deep end, asking exactly what happens inside a person鈥檚 head when they feel a pain, make a decision, or try to recall a memory. Reiman鈥檚 own PET studies on emotion are typical of the more controversial work. To discover the pathways involved in the neural mapping of different emotions, Reiman simply scanned a group of subjects as they watched short film clips 鈥� or even just recalled memories 鈥� designed to evoke feelings of disgust, sadness and happiness.
Pictures of the mind
鈥淚t seems too easy just to ask subjects to feel something and then take pictures of their brain,鈥� says Reiman. 鈥淚t鈥檚 almost a psychological barrier. Other scientists say 鈥榟ey you shouldn鈥檛 be able to do that 鈥� it shouldn鈥檛 work鈥�. We鈥檝e got used to not being able to ask that kind of direct question about the mind.鈥�
Reiman cautions that imaging studies cannot provide the whole answer. He says that every picture from the scanner should be backed up by evidence from other techniques such as the animal lesion studies of neurology and the behavioural studies of psychology. Reiman鈥檚 experiments are matched against traditional measures of emotion such as recordings of involuntary body reactions and facial expressions, and also against lesion studies in which known emotion pathways are stimulated or destroyed in laboratory animals. However, what has changed is that the mind sciences now have a direct approach to complement the many indirect research methods that have been employed over the years.
MRC researchers are particularly keen to explore what happens in the brain as something makes the transition from being subconscious or subliminal to being fully conscious. Everyone is familiar with the phenomenon of only noticing the background hum of a fan or computer once it has been switched off, leaving a sudden hush. To be able to realise something has fallen silent, our brains must have been registering its presence at some level at least. Scanning a person鈥檚 brain during such a transition could prove extremely revealing about exactly what counts in bringing something into full consciousness.
One possible answer is that processing has to reach certain locations in the brain before it becomes part of focal awareness. Enough is known about the brain to be sure that it has no single 鈥渃onsciousness centre鈥�. But at least three key cortex structures seem regularly to light up whenever something enters awareness.
PET studies by Michael Posner of the Washington University School of Medicine in St Louis, Missouri, have shown that a part of the frontal lobes known as the anterior cingulate is particularly active when a task demands careful attention. Another part of the brain, the post parietal cortex, seems to be crucial in bringing about shifts in our focus of awareness. Other work has suggested roles for the hippocampus and additional areas in the frontal cortex. It seems that consciousness may depend on the teamwork of a cluster of convergence zones 鈥� high-level areas in the brain where the various processing streams involved in sensation, motor output planning and memory each come to a head.
However, others believe that while convergence zones will be important to the story, the critical step that marks consciousness probably takes place at a lower and more distributed level. Following the speculation of Nobel prizewinner Francis Crick and others, some imaging researchers are hoping to test the theory that consciousness depends upon synchronised firing by neurons throughout the brain. The problem is that while scanners already have the resolution to look at convergence zones, it may be a few years before they can be used to test theories about synchronized oscillations.
Friston鈥檚 colleague Richard Frackowiak says that with hindsight, given the openness of the questions, it may be a good job that the CERN venture did not come off. It is true that a European imaging centre would have allowed brain researchers to build a magnificent scanner. 鈥淲e had all sorts of ideas like using a helmet of sensors, rather than a ring, so we could cover the brain from all angles. We talked about using supercomputers to process the images in realtime, allowing us to change experiments in response to what was happening. Or using virtual reality headsets to give the subjects interesting psychological tasks.鈥�
But Frackowiak says it has since become clear that imagers are not ready for such a technical quantum leap. 鈥淚n fundamental physics, you can justify such an approach because you have the theory and you know exactly what you want to test. But in brain science, we are still building the theory. Physics made its theoretical breakthroughs in the 1920s but did not go from university-level laboratories to international collaborations until the 1950s. I think imaging will follow the same pattern and CERN-type collaborations may develop in a decade or so.鈥� But for the moment, says Frackowiak, imaging is moving quite fast enough.