Wired.com Dissection column, April 4, 2008Link
It’s easy to be over-dazzled by the brain. Who could be unimpressed by the billions of neurons packed into our skulls, linked together by trillions of connections, capable of encoding memories from decades ago, of playing a saxophone, of sending space probes out of our solar system? We naturally want to know how our brain got to be so good. But there’s an even more interesting question worth asking: How do we manage to survive with a brain that’s so bad?
The job of the brain is to make decisions. It takes in information from its senses, which it then processes in a vast network of neuron circuits, finally producing some kind of output. The output may be as abstract as casting a vote, or as basic as taking a breath. These decisions depend on handling signals with extreme precision. Errors that creep into those signals as they ricochet around the nervous system are called noise. Remarkably, the closer scientists look at the brain, the more noise they discover.
Signals are encoded in the brain with spikes of voltage that travel down the length of neurons. These spikes are something like the digital stream of information that moves through a computer. But instead of silicon and gallium, neurons are made of fat, water and protein. They transmit their voltage spikes by opening channels, letting in charged atoms. The channels create a surge of current, which then causes the neighboring channels to open up. Each channel stays open only for an instant, as the spike of voltage rolls down the neuron, like a wave moving across a stadium.
The trouble with neurons, as scientists from Cambridge University write in the new issue of Nature Reviews Neuroscience, is that the channels don’t always do what they’re supposed to. The channels are continually wobbling and twitching, and sometimes they open up a little earlier than they should, splitting a single wave in two. Sometimes they open late, or not at all. These delinquent channels can make a short, sharp wave blur into a longer, weaker one. Channels sometimes open up when there is no wave, creating an entirely false spike.
The damage done to signals in our neurons is proving to be huge. As a train of voltage spikes travels down the length of a neuron, it can lose more than 25 percent of its information. More noise can creep into the brain's signals at other stages as well. When signals reach the tip of a neuron, they trigger a release of chemicals that flow to a nearby neuron, triggering a new voltage spike that can race onward. But these chemicals don’t work like simple switches; sometimes they fail to cross the gap, and the signal fails as well. As a signal moves from neuron to neuron to neuron, each one can add more noise to the signal, like a mental game of telephone. All this noise can blur our perception of the outer world and throw off the commands our brains send to our muscles.
The noise in our brains is so huge that it puts some hard limits on how well they work. One of the best ways to build a powerful brain is to use tiny neurons. As the size of each neuron shrinks, you can fit more of them in a given space. They can make more connections with one another, and it takes less energy for them to send signals.
It turns out, however, that our neurons could be much smaller than they actually are. If you packed all material necessary for sending signals as tightly as possible, the branches of a neuron (called axons) would measure just .06 microns [about 2.3 millionths of an inch] across. In fact, the thinnest axons are about .1 microns [about 4 millionths of an inch]. Recent studies have shown that it's noise that prevents them from getting thinner. The thinner an axon gets, the noisier it becomes. Below .1 microns, the noise abruptly rises so much that it drowns out any signal. We might be far smarter if noise didn't keep us from growing more neurons.
Scientists are finding that much of the brain's organization is dedicated to fighting noise. One way to fight it is to calculate the average of several signals. When we hear a sound, hair-like structures on neurons in our ears wiggle. Their wiggling creates a pattern of voltage spikes, which the neuron then passes on to 10 to 30 other neurons. All of those neurons then carry the same signal toward the brain, where they can be compared. Each neuron degrades the signal in a uniquely random way, and by averaging all of their signals together, the brain can cancel out some of the noise.
In order to reduce noise even more, our brains do not passively take in impressions of the world, like soft wax stamped with a seal. To perceive, we actually compare. When new signals arrive from our eyes, for example, we compare them with information stored in our brains about what the world usually looks like -- the fact that objects have edges, for example. This comparison allows us to discard the distractions of noise and focus on the authentic signal. Our brains also can't be passive in the way they issue commands to our muscles. A garbled signal could make us take a fatal misstep. Instead, our brains are constantly receiving information about how well our bodies are reaching their goals. To compensate for noise, our brains send out continuously updated commands to correct for previous ones.
Impressive? Absolutely. Our brains unconsciously carry out sophisticated calculations that engineers are trying to mimic to build better computers and communication systems. And yet all of this complex math serves a paradoxical purpose: to make up for the mistakes built into our very biology.