Science: I can see for miles

Autistic savants have many talents that have long defied explanation. Until now. By Rita Carter

Gordon Cherry has a skill few of us can comprehend, let alone explain. He can, if you ask him, tell you which day of the week it will be on any future date you care to mention - say, 18 June, 2099. He is one of that select group of people who can calculate, within seconds, the weekday of more or less any date, past or future.

Unfortunately many of them, including Mr Cherry, have severe difficulty doing almost anything else. They're autistic, and their astonishing talent for calendrical calculation appears as an isolated chip of genius in an otherwise severely curtailed intellect. Other "autistic savants" - as they're known - can reproduce works of art to perfection, play any piece of music faultlessly after a single hearing; recall almost limitless lists of objects at a single glance and recite the entire British Rail timetable, along with connections and buffet cars.

Savant skills are among the most puzzling phenomena. They are not limited to autistics - about one in 100 normal people (known in autistic circles as neurologically typicals or NTs) have some such talent. But savant ability is about 10 times more common among those with autistic disorders. Now a new theory has been put forward to explain this. The startling thing is that it suggests we might all have such abilities - if only we could access them.

Two Australian psychologists, Allan Snyder and D John Mitchell, writing in the Proceedings of the Royal Society, propose that savants are able to access brain functions that each and every one of us use, but which in most of us never surface. Savant skills are, they say, part of the normal cognition, but a part that is usually done in the dark back rooms of the brain, never brought to consciousness. They base their argument on brain-imaging studies revealing the complex unconscious processing that goes on before we're aware of perceptions, feelings and thoughts. A visual stimulus, say, a face, takes at least 150 milliseconds (ms) - about one-seventh of a second - to pop up in a person's mind as a conscious image. In that time an enormous amount of brain-work goes on. First the visual signal is sent from the eyes to the primary visual processing areas at the back of the brain, in about 50ms. These areas extract some basic information: whether it is a single object or a group; its distance, and roughly, its size, in about 20ms. Next this semi-constructed image is shunted forward to secondary visual processing areas that sort out - over the next 30 ms - its colour, any movement and, vaguely, its form. The signal then goes to sophisticated processing areas that detect patterns and start to clothe the signal with meaning. It is here that a roundish object, with a blob on each side, a vertical protuberance beneath and a slit at the bottom, is recognised as a face.

Consciousness kicks in when processing stops being just perception-building, and starts to involve a meaningful context. By the time we become conscious of seeing a face, we're already attaching to it associations and feelings: is it a familiar face? Is it attractive? Is it friendly? So, while unconscious processing involves assembling the building blocks of thought, conscious processing involves using them to make concepts. The advantage is that our sophisticated, conscious brains are not cluttered with routine perception- building. But it means that the way our ideas are formed is largely hidden from us, and the concepts that fill our conscious minds often blind us to what lies beneath.

One demonstration of this is the difficulty most people have in drawing accurately. When we see something we immediately clothe it in concepts. So we see the sea as blue and clouds as white - whether or not they really are. Anyone who has ever attended an art class knows one of the first things they have to learn is to ignore these prejudices and learn to look further back, at the image itself rather than the idea of it. Snyder and Mitchell think savants who can draw perfectly do so automatically, and for autistic savants it may be because they do not form the concepts that, in most of us, obscure the view of our earlier brain processes. A similar effect is seen with music; some autistic savants can reproduce a piece note for note, after a single hearing. The music seems to be "frozen" in their brains at the stage in which it is still just a series of notes, unelaborated by emotion, and a series of notes can be reproduced by simple rote memory. There is plenty of evidence to show we all store vast amounts of information that's never brought to consciousness. So the ability to regurgitate long sequences of notes or train times does not need any extra brain faculty, just access to what in most of us is buried or hastily junked.

So could any of us learn to reach back in our brains and pull out early, concept-free perceptions? Snyder and Mitchell think not. NTs who develop savant-like talents via practice usually employ quite different brain strategies. People who do rapid calculation or memory feats for, say, entertainment usually use clever formulae to do their sums, or use mnemonics. They have to work at what they do, whereas savants seem to do it without effort.

One celebrated case history of Oliver Sacks, the psychologist, involved savant twins who exchanged prime numbers up to 20 digits long - which would challenge a supercomputer. But their truly extraordinary mathematical talent was revealed most strikingly when Sacks dropped a box of matches. As they cascaded to the floor the twins both muttered "111". Sacks counted the matches and found exactly 111. "How could you count the matches so quickly?" he asked. "We saw them," was the reply. Just as we'd consciously "see", say, two or three matches without having to count them, the twins could "see" almost any number at a glance.

Even if we all "see" such things at an early stage of brain processing, bringing them to consciousness would presumably be as difficult as accessing the perception of a visual stimulus at the stage when it is still in fragments of colour, shape and motion. Snyder and Mitchell believe that accessing low-level, savant-type processing can be done only by "unusual people in whom mental integration is incomplete owing to retardation, brain damage or some other kind of mental disturbance".

But Niels Birbaumer, the German neurologist whose team recently fitted two paralysed stroke victims with brain-activated communication devices, believes we may all be able to train ourselves to access our unconscious state. Commenting on Snyder's and Mitchell's work, he points out that "after many training sessions" his patients and healthy controls were able to control their own unconscious brain activity to communicate from the "mysterious early state of non-consciousness".

He says we may discover techniques to let us all tell within seconds what day of the week it'll be 100 years hence. Then again, it might be easier to call up Mr Cherry.