The Man Who Wasn't There: Investigations into the Strange New Science of the Self (24 page)

David Cohen knows only too well of this inability of some autistic children to represent the human body. When I met Cohen, the head of child and adolescent psychiatry at the Pitié-Salpêtrière Hospital in Paris, in the autumn of 2011, he was still smarting from the ill effects of an article that appeared in the British journal
Lancet
in 2007,
investigating the French practice of
le packing,
or packing therapy. The practice involves wrapping a child from feet to neck in cold, wet sheets, leaving only the head free to move. The child is then covered in a rescue cover and a dry blanket, thus allowing the body to warm up. Each such session lasts about an hour, and the therapy may involve multiple sessions over days or weeks.
Le packing
has been used in France as an additional therapy to help calm down severely autistic children with self-harming behaviors.
Lancet
illustrated the article with a photograph of the Chapelle de la Salpêtrière, a famous landmark in Paris at the entrance to the hospital, linking packing therapy with Cohen, though the story was hardly about his hospital unit (in fact, Cohen had yet to publish a single academic paper on packing therapy). Those who found the practice cruel and barbaric directed their ire at Cohen. Further negative publicity came in the form of a stern letter to the editor of the
Journal of the American Academy of Child and Adolescent Psychiatry
, in which a number of prominent autism researchers decried the “
alleged therapy” as unethical.

But Cohen stands by the therapy. He pointed out to me, as he has
done in articles he has published since the outcry in the
Lancet
, that the child is treated under the supervision of a
psychomotricien
(a specialist trained in psychomotor disturbances) and at least two members of the team caring for the child.

One of Cohen’s patients, John, was an adolescent diagnosed with pervasive developmental disorder. He was catatonic when he was admitted to the hospital. Given the seriousness of his condition, electroconvulsive therapy (ECT) was an option, but John’s parents refused it. Instead, they chose packing therapy, along with drugs (benzodiazepine and Prozac). John’s condition improved while on this combination of therapies: he even agreed to draw after each session of packing therapy. The drawings showed something very interesting. After the second session, John wrote letters and words. It was only after session twelve that the first hints of a body appeared in the drawing: John drew a hand. He drew a stick figure after session sixteen, and a more realistic body after session twenty-three. It was as if the packing sessions were bringing
John closer to his body.

Cohen views John’s catatonia and indeed some of the sensory-motor disturbances observed in autism as a consequence of the inability of the brain to properly integrate all the various senses. The basic idea here is that the brain combines all the various sensations, both internal and external, such as touch, vision, vestibular and proprioception, to create a
body percept
(the sense of the body as an entity in itself—the bodily self), which then becomes the foundation for learning and behavior. Any disturbance of this multisensory integration process doesn’t just disrupt perception of stimuli and one’s own body, it also has behavioral and cognitive consequences.

Packing therapy, then, is helping reintegrate the senses, serving to “
combine the body and the image of the body” and “
to reinforce
children’s consciousness of their own body limits.” In the parlance of the self, packing therapy is helping the self-as-subject form clear perceptions of the bodily self, the basic constituent of the self-as-object.

I asked Susan if this made sense to her, given her experience with Alex. She thought it did. “I believe that Alex’s difficulties in understanding his body in relation to his physical environment caused delays in his language development, his abilities to self-organize, and caused his creative/imaginative and social impairments. These challenges must have had a big impact on the development of his sense of self,” said Susan. “The reason his drawing of the human body was so primitive is because he may not have experienced fully the sensation of his limbs.”

The consequences of sensory integration problems, in this way of thinking, could also cause a deficient theory of mind in children with autism. “Yes, they have that problem. Wouldn’t you have it if you couldn’t feel your body?” cognitive psychologist and computational neuroscientist Elizabeth Torres of Rutgers University asked me rhetorically.

Torres displays impatience with the status quo when talking about autism. To her, the practice of diagnosing autism by observing and cataloging behavioral aberrations “
based on clinical observations with shifting criteria” is not helping. In fact, it’s missing the point. According to Torres, the behavioral disturbances observed in those with autism are the result of one not being grounded in a stable perception of one’s own body. Simple as that.

“People speak of behaviors in a very disembodied manner,” Torres told me. “They talk about it as if it was something esoteric, but
behavior is a combination of movements that flow continuously like a stream. Movements [in turn] are a combination of things that we do with a purpose and things we don’t even know we are doing.”

So, can one see anything in the way children with autism move that hints at an underlying problem? Torres’s answer is an adamant yes.

Her assertion is based on measurements of movements of children diagnosed with autism spectrum disorder. Torres was looking for what she calls micromovements: barely perceptible fluctuations in how we move. For instance, if you were to reach out to touch a target on a computer touchscreen, at some moment
t
after you began moving your hand, it would reach a peak velocity
v
, and then it would slow down and eventually stop at the screen. Both
v
and
t
are examples of parameters that describe the motion of your hand. Interestingly, these parameters have tiny variations, or what Torres calls micromovements. If you reach out for the computer screen a hundred times, the values for
v
and
t
will be ever so slightly different each time. These variations in the parameters that describe our movements and the rates at which their statistical signatures change from moment to moment are unique to each person. Torres has argued that this variability in micromovements is a type of sensory input from the periphery of the body to the central nervous system.

It’s a concept that goes back to work done in 1950 by Erich von Holst and Horst Mittelstaedt. We saw in the context of schizophrenia that the brain makes copies of motor commands, predicts the sensory consequences of these commands, and compares them to the actual sensations to generate a sense of agency. So, the brain would have to rely on some sort of error feedback from the body. Where exactly is the feedback coming from? “
There could be error feedback from kinesthetic receptors in joints, tendons, and sensory muscle spindles.
Inputs from these receptors could indicate whether the moment-to-moment position of the arm corresponds to the intended position indicated by motor commands (efference copy). . . . Its use requires a learned association between motor commands and kinesthetic inputs.”

The brain uses such kinesthetic signals, the argument goes, to build and maintain a stable percept of the body or an internal model of the body, so that it can then effectively issue motor commands and also accurately predict the consequences of carrying out the commands. This means that the micromovements should have valuable information, or a high signal-to-noise ratio.

When Torres studied variations in these micromovements in children of varying abilities and ages, something very intriguing emerged.

The signal-to-noise ratio increased with age. In three-to-four-year-old typically developing children, the kinesthetic error feedback signals were very noisy. But in four-to-five-year-olds, the signals were significantly less noisy. In adults, these kinesthetic inputs were reliable and predictable. But autism affects this progression. The feedback from the micromovements is extremely noisy in both children and adults diagnosed with autism. If the brain is indeed working with internal models of the body, then these inputs are not helping to keep these models updated. They provide very little information for the brain about prior behavior on which it can base future behavior. It’s as if the child with autism is struggling to make sense of every experience. “That’s the way they must be experiencing the world. It’s constantly new. They can’t make sense of it. They can’t come to a stable percept that enables one to predict ahead,” Torres said.

The self-as-subject is unable to get a fix on the self-as-object.

The body anchors the self-as-object; it’s our point of reference.
Everything we perceive is in relation to it. According to Torres, a glitch during development can disturb this point of reference, and depending on when that glitch happens, it can lead to a bewildering array of consequences as the child grows up—which helps explain the huge range of behavioral symptoms that get clumped under autism spectrum disorder, ranging from problems processing sensations to theory-of-mind disturbances and difficulties in relating socially to others. “If you don’t have that point of reference, and everything is always new to you, there is no anchor,” said Torres. “This must be happening to people with autism, because their information about the body is noisy and it’s random. We have measured very precisely; it doesn’t depend on my opinion, and it is what it is. And it really is present in every autistic person we have seen and it gets worse with age.”

Fortunately, said Torres, this also means that detecting the noisy system early, with objective measurements, not a clinician’s subjective observations—as well as developing therapies to train the body and reduce the noise—is likely to be extremely beneficial.

Torres’s work sits well with the idea of the Bayesian brain, the idea that the brain could be making probabilistic inferences about the likely causes of sensory inputs. We saw how this framework could be applied to emotions to explain depersonalization disorder. Applying it to autism is proving valuable too.

If you accept that the body’s imperative is to survive, then the brain’s job (in close collusion with the body) is to maintain the body in a state suitable for survival. For any given biological organism, survival means existing in one of a finite set of physiological states. For instance, if you take internal parameters like blood pressure and heart
rate, and external parameters like temperature, and define body states based on these parameters, then there are only a limited set of states in which all these parameters are within acceptable limits. Looked at another way, “
there is a high probability that a [biological] system will be in any of a small number of states, and a low probability that it will be in the remaining states,” according to Karl Friston of University College London. The process of staying within the bounds of physiologically viable states, again, is known as homeostasis.

Friston posits that the brain achieves homeostasis by minimizing what he calls free energy, which allows biological systems (or indeed any system that can learn and adapt) to “
resist a natural tendency to[ward] disorder.” In the context of a population out in the wild, systems that minimize free energy survive; those that don’t, die.

Friston has shown that minimizing free energy for a biological system is equivalent to minimizing the amount of surprise it encounters as it navigates its environment. “
Biological agents must avoid surprises to ensure that their states remain within physiological bounds,” he says.

To review, the way a Bayesian brain avoids surprises is by maintaining an internal model of the body, the environment, and of itself. This is a probabilistic model that can generate a number of predictions about the causes of sensory inputs based on a set of prior beliefs as to the causes. Then, given actual sensory data, the brain assigns new probabilities to its predictions, and the prediction with the highest probability is what we perceive as the cause of those sensations. Of course, now the brain has a new set of prior beliefs, updated to reflect its new understanding of the body, brain, and environment. Where does surprise enter this picture? Well, if you found yourself in a state that was extremely surprising and detrimental to your existence (an
example from Friston is that of a fish out of water, an extremely surprising state for the fish), your brain would initiate action to suppress that element of surprise—the action could either involve modifying its internal models or making the body move (a fish out of water had better flip-flop its way back into water; modifying internal models will not help). So, the greater the discrepancy between the brain’s predictions about the causes of sensory inputs and the actual sensory inputs, the greater is the element of surprise. Minimizing surprise is akin to minimizing prediction error—which implies a brain whose models are in tune with external and internal reality.