Brain Buys

by Dean Buonomano

  While these studies were initially met with great skepticism, cortical plasticity is now accepted as one of the key principles of brain function. Studies in humans have further established the importance of the cortex’s ability to reorganize for many types of learning. For example, using noninvasive techniques, studies have compared the amount of somatosensory cortex devoted to the fingers of musicians and nonmusicians. People who started playing string instruments at an early age were found to have more cortical area devoted to their fingertips. Similarly, an expansion of the fingertip areas was observed in people who learned to read Braille when they were young.9

  In the early days, computer programmers had to preallocate the amount of memory to be devoted to a given program. That is, they had to estimate how much memory would be used, and some early pieces of software had a limit on the amount of information they could handle. Over the decades, more sophisticated programming languages have been developed that allow the dynamic allocation of memory: as I type more and more words into a word processor, the amount of memory dedicated to the file is dynamically increased. In terms of the allocation of computational power, the brain has used this strategy for tens of millions of years—though the dynamic allocation of cortical areas is a gradual process takes place over weeks and months.

  The brain, of course, has no supervisor to oversee the distribution of cortical real estate. So how does it figure out exactly how important different body surfaces are? It seems to use a rule of thumb. Since the degree of activity in a given zone of the somatosensory cortex roughly reflects how much the corresponding body part is used, the brain can assign importance based on activity.10 Let’s consider what happens in the somatosensory cortex of a person who has the phantom sensation of an index finger that was lost in an accident. Normally, the neurons in the zone of the primary somatosensory cortex that represent the index finger would be driven by inputs originating in that finger. But, now deprived of their source of input, these cortical neurons should fire much less than they once did. For argument’s sake, let’s assume that the index finger neurons in the somatosensory cortex went totally silent after the accident. Neurons abhor silence. A neuron that never fires is mute; it barely deserves to be called a neuron since neurons are all about communication. So it is not surprising that neurons are programmed to do their best to avoid being mute for long periods of time. In the same manner that there are compensatory or homeostatic mechanisms in place to allow the body to regulate its temperature depending on whether it is warm or cold outside, neurons are able to homeostatically regulate their level of activity.

  A neuron in the somatosensory cortex that is devoted to the index finger receives thousands of synapses. Many of these synapses convey information from the index finger, but some synapses originate from neighboring neurons in the cortex that represent other parts of the body. In this case, because of the topographic organization of the cortex, neurons surrounding the index finger neurons tend to be those representing the thumb and middle finger. These neighboring neurons should exhibit their normal levels of activity (or perhaps more, because someone who lost an index finger will start using the middle finger in its place). The silenced neurons of the index finger will amplify the inputs from the neighboring areas that are still active. This is what allows them to “change teams”—the ex-index finger neurons can become thumb or middle finger neurons by strengthening the synapses from neurons that already responded to the thumb or middle finger.

  The exact mechanisms responsible for the amplification of previously weak inputs continues to be debated, but once again they seem to rely on the same synaptic and cellular mechanisms underlying learning and memory, including strengthening of existing synapses and the formation of new ones.11 Recall Hebb’s rule from Chapter 1: neurons that fire together, wire together. But suppose a neuron stops firing completely, as in our hypothetical example of the index finger neurons that went silent. Homeostatic forms of synaptic plasticity allow previously weak synapses to become strong even in the absence of much postsynaptic activity—essentially overriding Hebb’s rule.12 If a strong signal is lost, the response to weak inputs can be amplified.

  Now let’s return to our question of why neurons in the somatosensory cortex that have lost their original input continue to fire and mislead the brain into believing that the amputated limb, or finger, is still present. One hypothesis is that the neurons that used to be activated by the index finger are being driven by activity in the thumb and middle fingers, but in an exaggerated fashion. So even in the absence of their normal source of input the neurons in the primary somatosensory cortex that were previously responsible for conveying information about the index finger may still be active! Downstream or “higher-order” areas of the cortex continue to interpret this activity as evidence that the index finger is still in place. It is presumed that these higher-order areas somehow create the conscious experience of feeling one’s body, but no one knows how or where this comes about. Nevertheless, in people with a phantom limb it is clear that these areas never get the message that the body has changed—the master map is never updated. Just as a king who is never told that part of his empire has been captured might continue to “rule” over a territory he no longer controls, some part of the brain persists in generating an unaltered illusion of the body, blissfully unaware that some part of the body no longer exists.

  THE FANTASTIC PLASTIC CORTEX

  The discovery that the somatosensory cortex is continuously remodeled throughout life was seminal because it revealed a general feature of the cortex and the mechanisms of learning and memory: cortical plasticity is not restricted to the somatosensory cortex; it’s a general feature of the entire cortex. Many studies have demonstrated that the circuits in other cortical areas also undergo reorganization in response to experience.

  Most of our knowledge of cortical plasticity comes from studies of the sensory areas, specifically the somatosensory, auditory, and visual cortices. Among these, vision is the notorious cortical hog. In primates, for example, the total amount of cortex devoted to visual processing far exceeds that of any other sensory modality. By some estimates close to half of the entire cortex may be devoted primarily to sight.13 Thus, if these areas went permanently offline as a result of blindness, there would be billions of very bored neurons. Due to cortical plasticity, however, these visual areas can be put to work toward nonvisual tasks. For a person who lost her eyesight at an early age, the tactile task of determining whether the object in hand is a pen or pencil may activate “visual” areas (the part of the brain that would normally process sight). As proof of this, temporarily interfering with normal electrical activity in the “visual” cortex of blind people has been shown to degrade their ability to read Braille. The visual cortex is also more robustly activated by sounds in blind people.14 In other words, a blind person may have more cortical hardware to devote to somatosensory and auditory processing, which likely contributes to superior performance on some somatosensory and auditory tasks.15 The extent to which people can improve processing in sensory modalities, and use these to compensate for a lost modality such as vision, is illustrated in the extreme by the ability of some people to “see” using echolocation. Some animals, including bats and dolphins, can navigate their environment and discriminate objects in the absence of light. Dolphins can even perform a sonogram and “see” through some materials, which is why the U.S. Navy has trained dolphins to find mines hidden underneath layers of mud on the ocean floor.

  Echolocation uses the same principles as sonar. Bats and dolphins emit sounds, wait for the echo of these sounds to return after they have rebounded off objects, and use their auditory system to interpret the scene rendered by these echoes. The delay between the sound emission and the returning echo is used to determine the distance of the object. Remarkably, some blind humans have learned how to echolocate. They emit clicklike sounds from their mouth (or use a walking cane to “tap”) and wait for the echo. One boy who lost both his eyes to cancer at the
age of two was able to walk around and distinguish objects such as a car from a garbage can without touching them.16 Although this ability has not been carefully studied, it likely relies on the brain’s ability to allocate cortical area according to an individual’s experience. It should be pointed out, however, that extraordinary sensory abilities are not simply a result of having more cortical space available to perform certain computations. They are also a product of intense practice and the immersive experience that comes with living in what amounts to an entirely different world.

  The ability of the cortex to adapt and reorganize is among its most powerful features. Cortical plasticity is why practice makes perfect, why radiologists see pneumonia in x-rays that look like out-of-focus blotches to the rest of us, and why Braille readers have more sensitive fingertips. Cortical plasticity is also the reason why a child born in China ends up with a brain well suited to decipher the tonal sounds of Mandarin, which sound indistinguishable to the average English speaker. However, cortical plasticity is also responsible for some of the brain bugs that emerge in response to mild or serious injury. The pain of a phantom limb is the brain’s own fault, produced by a glitch in the brain’s attempt to adapt to the missing limb. The brain’s extraordinary ability to reorganize can be maladaptive.

  Glitches in brain plasticity may also underlie a much more common medical condition: tinnitus. Roughly 1 to 3 percent of the general population experiences the annoying and incessant buzzing or “ringing” that characterizes tinnitus. It is the number-one disability reported among Iraq war veterans.17 The consequences of tinnitus can be serious, and include the inability to concentrate, loss of sleep, and depression.

  Sound is detected by hair cells located in the sensory organ of the ear, the cochlea. Tiny “hairs” or cilia on top of each of these cells respond to minute changes in air pressure; their movement results in the generation of action potentials in the auditory nerve that conveys information to the brain. Different groups of hair cells tend to be activated by specific frequencies. Like the keyboard of a piano, the cochlea represents low frequencies at one end and high frequencies at the other. In the same manner that the somatosensory cortex contains a topographic map of the body, the primary auditory cortex contains a tonotopic map of the cochlea. If a neurosurgeon stimulated your auditory cortex, instead of feeling that someone touched you, you would hear a sound, and depending on the exact location of the stimulation it would be low or high in pitch.

  One might be inclined to speculate that the grating ringing that sufferers of tinnitus experience is produced by overactive sound detectors in the ear; that for some reason some of the hair cells in the cochlea are continuously active, generating the illusion of a never-ending sound. Despite its plausibility, this hypothesis is not consistent with much of the evidence. Tinnitus is generally accompanied by a decrease in activity in the cochlea and auditory nerve, and associated with the death of hair cells.18 The loss of these cells can be produced by certain drugs, chronic or acute exposure to very loud sounds, and normal aging. The ear is particularly sensitive to environmental hazards and the aging process because we are born with precious few hair cells—each cochlea only contains around 3500 of the most important type of hair cell: inner hair cells (in contrast to the 100 million photoreceptors in each retina, for example). Damage to hair cells that respond to high frequencies, of course, results in impaired hearing of high-pitched sounds. The ringing people experience usually corresponds to the same pitch in which people suffer their hearing loss.19 That is, loss of the hair cells at the base of the cochlea, which respond to high-frequency sounds, may result in a continuous high-frequency ringing. At this point, the parallel with phantom limbs should be clear: both tinnitus and phantom limbs are associated with the damage to or absence of normal sensory inputs. Tinnitus is the auditory equivalent of a phantom limb—a phantom sound.

  As is the case with phantom limbs, maladaptive cortical plasticity seems to be one of the causes of tinnitus.20 The hypothesis is that if a specific part of the cochlea is lesioned, the corresponding location in the auditory cortex is deprived of its normal source of input. This area might then be “captured” by the neighboring regions of the auditory cortex. The phantom sound may be generated by the neurons in the auditory cortex (or other stations in the auditory pathway) that lost their original input source, and came to be permanently driven by inputs from neighboring neurons. The causes of both phantom limbs and tinnitus are, however, not fully understood, and each is likely to have more than one underlying cause. Nevertheless, brain plasticity gone awry contributes to both syndromes.

  GRACEFUL DEGRADATION VERSUS CATASTROPHIC BREAKDOWNS

  The great majority of the brain’s 90 billion neurons,21 close to 70 billion, are quite frankly, simpleton neurons called granule cells, which reside in the cerebellum (a structure that among other things plays an important role in motor coordination). If you had to part ways with a few billion neurons, these are the ones to choose. Whereas your average cortical neuron receives thousands of synapses, a granule cell receives fewer than 10.22 But granule cells make up for their rather narrow view of the world with sheer numbers. Of the remaining 20 billion or so neurons, most reside in the cortex. This number is not quite as impressive as it sounds. Today, a single computer chip routinely possesses billions of transistors, so some parallel computers have more transistors than the brain has cortical neurons. I’m not implying that one should think of a transistor as being in any way the computational peer of a neuron (even a granule cell), but in terms of component computational units, your average desktop currently exceeds the number of neurons in the brains of many animals, including mice.

  Until the 1990s the reigning dogma was that all mammals were born with their maximal complement of neurons; no new neurons were made after birth. We now know that this is not the case. Some neurons continue to be generated throughout life, but mostly in restricted areas of the brain (the olfactory bulb and part of the hippocampus).23 But, truth be told, the contribution of these neurons to the total number count is probably not significant. If it were these structures would have to grow throughout our lifespan, which they don’t. So in terms of absolute numbers it’s a downhill voyage from cradle to grave. It has been estimated that we lose 85,000 cortical neurons a day, and that total gray matter volume progressively decreases by about 20 percent over our adult life.24 The amazing thing about these facts is how little impact they have on our day-to-day lives. Despite the constant circuit remodeling, cell death, brain shrinkage, and the inevitable whacks to the head, each of us remains as we have always been. For the most part we retain our important memories, core personality traits, and cognitive abilities. Scientists and computer scientists refer to systems that can absorb significant amounts of change and damage without dramatic effects on performance as exhibiting graceful degradation, but brains and computers differ considerably in their ability to degrade gracefully.

  Computers depend on the extraordinary reliability of transistors, each of which can perform the same operation trillions of times without making a single mistake or breaking. However, if a few of the transistors etched into a modern CPU chip did break, depending on their location on the chip, the consequences could be highly ungraceful. In sharp contrast, losing a few dozen neurons in your cortex, independent of location, would have no perceptible consequence. This is in part because neurons and synapses are surprisingly noisy and unreliable computational devices. In sharp contrast to a transistor, even in the well-controlled environ of a laboratory, a neuron in a dish can respond differently over multiple presentations of the same input. If two cortical neurons are connected by a synapse, and an action potential is elicited in the presynaptic neuron, the presynaptic component of the synapse will release neurotransmitter that will excite the postsynaptic neuron. The truth is, however, that there is a significant probability that the synapse between them will fail, and that the message will not make it across to the postsynaptic neuron. This so-called failure rate is dependent
on many factors, and is generally around 15 percent but can be as high as 50 percent.25 The unreliability of cortical neurons should probably not be presented in an overly negative light, because this variability is in place by evolutionary design—synaptic transmission at some synapses outside the cortex can be vastly more reliable. Some neuroscientists believe that like someone trying to find the next piece of a puzzle by trial and error, the fallibility of cortical synapses helps networks of neurons explore different solutions to a computational problem and pick the best one. Furthermore, the unreliability of individual neurons and synapses may be one reason the brain exhibits graceful degradation, since it ensures that no single neuron is critical.

  The brain’s graceful degradation is often looked at with some envy by computer scientists. But the envy is somewhat misplaced; in some cases the brain’s degradation is not at all graceful. True, only massive damage to the cortex (or of critical areas of the brainstem) can produce a system crash (coma or death), but small lesions can lead to stunning breakdowns in specific abilities.

  One example of a fantastical syndrome that can arise when certain areas of the brain are injured is called alien hand syndrome. It is a very rare disorder that can have a number of different causes, including head injuries and strokes. Patients with alien hand syndrome experience a dissociation between themselves and one or more of their limbs. The limb isn’t paralyzed or incapable of fine motor movements, but rather, it is as if the limb has acquired a new master—one with some warped hidden agenda. Patients with alien hand syndrome have been known to be buttoning a shirt with their unaffected hand while the alien hand proceeds to unbutton it, or to simultaneously try to open a drawer with one hand and close it with the alien hand. The syndrome often results in perplexed and frustrated reports from patients: “I don’t know what my hands are doing. I’m no longer in control”; “The left hand wants to take over when I do something, it wants to get into the swim”; and as one patient with a wayward hand reported to the nurse, “if I could just figure out who’s pulling at my hair because that hurts.”26