To Sweeney's collaborator, University of Pittsburgh neurologist Dr. Nancy Minshew, the images Sweeney has produced of autistic minds in action are endlessly evocative. They suggest that essential connections between key areas of the brain either were never made or do not function at an optimal level. "When you look at these images, you can see what's not there," she says, conjuring up an experience eerily akin to looking at side-by-side photographs of Manhattan with and without the Twin Towers.

A MATTER OF MISCONNECTIONS
Does autism start as a glitch in one area of the brain — the brainstem, perhaps — and then radiate out to affect others? Or is it a widespread problem that becomes more pronounced as the brain is called upon to set up and utilize increasingly complex circuitry? Either scenario is plausible, and experts disagree as to which is more probable. But one thing is clear: very early on, children with autism have brains that are anatomically different on both microscopic and macroscopic scales.

For example, Dr. Margaret Bauman, a pediatric neurologist at Harvard Medical School, has examined postmortem tissue from the brains of nearly 30 autistic individuals who died between the ages of 5 and 74. Among other things, she has found striking abnormalities in the limbic system, an area that includes the amygdala (the brain's primitive emotional center) and the hippocampus (a seahorse-shaped structure critical to memory). The cells in the limbic system of autistic individuals, Bauman's work shows, are atypically small and tightly packed together, compared with the cells in the limbic system of their normal counterparts. They look unusually immature, comments University of Chicago psychiatrist Dr. Edwin Cook, "as if waiting for a signal to grow up."

An intriguing abnormality has also been found in the cerebellum of both autistic children and adults. An important class of cells known as Purkinje cells (after the Czech physiologist who discovered them) is far smaller in number. And this, believes neuroscientist Eric Courchesne, of the University of California at San Diego, offers a critical clue to what goes so badly awry in autism. The cerebellum, he notes, is one of the brain's busiest computational centers, and the Purkinje cells are critical elements in its data-integration system. Without these cells, the cerebellum is unable to do its job, which is to receive torrents of information about the outside world, compute their meaning and prepare other areas of the brain to respond appropriately.

Several months ago, Courchesne unveiled results from a brain-imaging study that led him to propose a provocative new hypothesis. At birth, he notes, the brain of an autistic child is normal in size. But by the time these children reach 2 to 3 years of age, their brains are much larger than normal. This abnormal growth is not uniformly distributed. Using mri-imaging technology, Courchesne and his colleagues were able to identify two types of tissue where this mushrooming in size is most pronounced.

These are the neuron-packed gray matter of the cerebral cortex and white matter, which contains the fibrous connections projecting to and from the cerebral cortex and other areas of the brain, including the cerebellum. Perhaps, Courchesne speculates, it is the signal overload caused by this proliferation of connections that injures the Purkinje cells and ultimately kills them. "So now," says Courchesne, "a very interesting question is, What's driving this abnormal brain growth? If we could understand that, then we might be able to slow or stop it."

A proliferation of connections between billions of neurons occurs in all children, of course. A child's brain, unlike a computer, does not come into the world with its circuitry hard-wired. It must set up its circuits in response to a sequence of experiences and then solder them together through repeated neurological activity. So if Courchesne is right, what leads to autism may be an otherwise normal process that switches on too early or too strongly and shuts off too late — and that process would be controlled by genes.

Currently Courchesne and his colleagues are looking very closely at specific genes that might be involved. Of particular interest are the genes encoding four brain-growth regulators that have been found in newborns who go on to develop mental retardation or autism. Among these compounds, as National Institutes of Health researcher Dr. Karin Nelson and her colleagues reported last year, is a potent molecule known as vasoactive intestinal peptide. vip plays a role not only in brain development but in the immune system and gastrointestinal tract as well, a hint that other disorders that so frequently accompany autism may not be coincidental.

The idea that there might be early biomarkers for autism has intrigued many researchers, and the reason is simple. If one could identify infants at high risk, then it might become possible to monitor the neurological changes that presage the onset of behavioral symptoms, and someday perhaps even intervene in the process. "Right now," notes Michael Merzenich, a neuroscientist at the University of California, San Francisco, "we study autism after the catastrophe occurs, and then we see this bewildering array of things that these kids can't do. What we need to know is how it all happened."

The genes that set the stage for autistic disorders could derail developing brains in a number of ways. They could encode harmful mutations like those responsible for single-gene disorders — cystic fibrosis, for instance, or Huntington's disease. They could equally well be garden-variety variants of normal genes that cause problems only when they combine with certain other genes. Or they could be genes that set up vulnerabilities to any number of stresses encountered by a child.

A popular but still unsubstantiated theory blames autism on the MMR (measles, mumps and rubella) vaccine, which is typically given to children at around 15 months. But there are many other conceivable culprits. Researchers at the University of California at Davis have just launched a major epidemiological study that will test the tissues of both autistic and nonautistic children for residues of not only mercury but also pcbs, benzene and other heavy metals. The premise is that some children may be genetically more susceptible than others to damage by these agents, and so the study will also measure a number of other genetic variables, like how well these children metabolize cholesterol and other lipids.

Drugs taken by some pregnant women are also coming under scrutiny. At the University of Rochester, embryologist Patricia Rodier and her colleagues are exploring how certain teratogens (substances that cause birth defects) could lead to autism. They are focusing on the teratogens' impact on a gene called hoxa1, which is supposed to flick on very briefly in the first trimester of pregnancy and remain silent ever after. Embryonic mice in which the rodent equivalent of this gene has been knocked out go on to develop brainstems that are missing an entire layer of cells.

In the end, it is not merely possible but likely that scientists will discover multiple routes — some rare, some common; some purely genetic, some not — that lead to similar end points. And when they do, new ideas for how to prevent or correct autism may quickly materialize. A decade from now, there will almost certainly be more effective forms of therapeutic intervention, perhaps even antiautism drugs. "Genes," as the University of Chicago's Cook observes, "give you targets, and we're pretty good at designing drugs if we know the targets."

Paradoxically, the very thing that is so terrible about autistic disorders — that they affect the very young — also suggests reason for hope. Since the neural connections of a child's brain are established through experience, well-targeted mental exercises have the potential to make a difference. One of the big unanswered questions, in fact, is why 25% of children with seemingly full-blown autism benefit enormously from intensive speech- and social-skills therapy — and why the other 75% do not. Is it because the brains of the latter are irreversibly damaged, wonders Geraldine Dawson, director of the University of Washington's autism center, or is it because the fundamental problem is not being adequately addressed?

The more scientists ponder such questions, the more it seems they are holding pieces of a puzzle that resemble the interlocking segments of Tommy Barrett's Transformer toys. Put the pieces together one way, and you end up with a normal child. Put them together another way, and you end up with a child with autism. And as one watches Tommy's fingers rhythmically turning a train into a robot, a robot into a train, an unbidden thought occurs. Could it be that some dexterous sleight of hand could coax even profoundly autistic brains back on track? Could it be that some kid who's mesmerized by the process of transformation will mature into a scientist who figures out the trick?

— With reporting by Amy Bonesteel/Atlanta

1 | 2 | 3

	FOOD & DRINK Chefs Who Shine Nothing in the food world beats a Michelin star. TIME talks to the five European chefs receiving top honors PROFILE Model Example Waris Dirie wants to alert the world to the plight of millions of young women BUSINESS No Picnic Teddy bear rivals feud over new "make-your-own" models ARTS Urban Legends Young British artists are opting for cityscapes with a hint of darkness