Fortunately, the process of motor-skill learning is readily amenable to laboratory study, so Henry became a rich resource. The intriguing evidence from Milner’s 1955 mirror-drawing study and my own 1964 study inspired me to examine whether Henry could learn other motor tasks. In 1966, when he was forty years old, I had the opportunity to examine this question more thoroughly. His parents gave their approval for him to check into the MIT Clinical Research Center (CRC) for two weeks of testing—the first of Henry’s fifty visits to the CRC over the next thirty-five years. On this trip, the purpose of our tests was to pursue the observation that, in the face of his profound amnesia, Henry could still learn new motor skills. With the prospect of testing Henry on fourteen consecutive days, I recorded his day-to-day progress on three measures of skill learning—rotary pursuit, bimanual tracking, and coordinated tapping.7
The apparatus for the first task, rotary pursuit, resembled an old-fashioned record turntable with a metal target roughly the size of a quarter located about two inches from the edge. Henry held a stylus between his right thumb and index finger, and I asked him to rest the tip of the stylus on the target. After a few seconds, the disc began spinning, and for twenty seconds he tried his best to keep the stylus in contact with the target as it turned; I recorded the time that the stylus remained on the target, as well as the number of times it left the target. I tested Henry and control participants twice a day for the first two days and once a day for the next five days. Then, I tested them again a week later to see how well they remembered the task without practicing it (see Fig. 10).8
Over the seven days of testing, Henry’s scores improved, although not as much as the control participants’. A closer look revealed that the number of times he made contact with the target increased with practice; he became more proficient at returning to the target once he lost contact with it. Overall, the control participants stayed on the target longer. Although Henry’s gains were not as dramatic as the others, he retained the new motor skill for one week with no additional training. When I tested him on day fourteen, he performed just as well as he had on day seven.9
The next week, I trained Henry on a bimanual tracking task. The apparatus was an aluminum drum with two narrow, asymmetric tracks painted on it. Henry held a stylus in each hand and placed one on each track. His job was to maintain contact with the tracks as the drum rotated for twenty seconds. This task was especially difficult from a motor-control perspective because Henry’s brain had to coordinate the movements of his left and right hands and his eyes, which moved back and forth from one track to the other. The two sides of his brain thus had to interact continuously. I repeated the test three times at increasing speeds of rotation, recording how many seconds Henry and the controls stayed on each track and how many times they fell off. As before, Henry’s scores were inferior to the control participants’, and he was less consistent, but he again demonstrated a clear improvement from trial to trial in performing this motor skill (see Fig. 11).10
Henry’s suboptimal performance on rotary pursuit and bimanual tracking was not due to his memory problem; these two tasks depended on quick reaction times. When Henry had more time to respond to a stimulus, he performed just fine. But in general, he tended to do everything slowly. His slow tempo was likely due in part to Phenobarbital, a sedative prescribed for insomnia as well as epilepsy. Other patients who had similar lesions—Scoville’s patient D.C., and Penfield and Milner’s patients P.B. and F.C.—also took antiseizure medication and moved sluggishly. But despite his slowness, Henry clearly could learn new motor skills and retain that knowledge over long periods of time. We do not know how he would have performed were he taken off his antiseizure medications, which was not an option because doing so would have put his health and safety at risk.11
Another motor-learning task, coordinated tapping, measured Henry’s ability to tap four targets in turn with a stylus, first with each hand separately, and then with the two hands together. The goal of this study was to see whether, with practice, he would speed up and increase the number of times he tapped in the thirty seconds allowed. The apparatus consisted of a black wooden board with two metal circles, side by side, divided into quadrants. Each quadrant was numbered 1, 2, 3, or 4, but the numbers were arranged differently in the two circles. First, Henry held a stylus in his right hand and tapped the circle on the right in the order 1–2–3–4. He then held a stylus in his left hand and tapped the circle on the left in the order 1–2–3–4. I next asked him to tap the two targets simultaneously, which was especially demanding because Henry was required to tap the two 1s simultaneously, the two 2s simultaneously, and so on. He had to coordinate the movements of his left and right hands, and because the location of the numbers differed between the two circles, each hand had a different trajectory to follow. Henry and the control participants performed the task twice, with a forty-minute break between sessions (see Fig. 12).12
On this test, Henry scored as well as the control participants, and when I retested him after a break, he was faster than he had been initially. He consolidated the motor memory of the tapping skill, allowing him to demonstrate his learning of this motor behavior forty minutes later. Why was Henry’s learning comparable to that of the control group on the tapping task, but not on rotary pursuit and bimanual tracking? A major difference was that the tapping task was self-paced; Henry went at his own speed. On the other two tasks, however, the movement of the apparatus dictated his moves. The rotary pursuit apparatus turned at three different speeds, and the drum on the bimanual tracking apparatus advanced automatically in short steps. On these two tasks, he also had to predict quickly where the target was going, and this need to anticipate the future may have required input from declarative memory.13
These early studies with Henry illuminated the distinction between declarative and nondeclarative learning. Declarative knowledge requires medial temporal-lobe structures for its expression, whereas nondeclarative, procedural knowledge is independent of that network. Learning new skills, new procedures, occurs without conscious awareness. When we ride bicycles, play tennis, or ski, we demonstrate our expertise—or lack thereof—through performance. If we try to analyze what we are doing millisecond by millisecond, we may crash, miss a shot, or catch an edge. Similarly, musicians find that their performance falls apart if they try to think about a difficult piece of music note by note; instead, they execute a complex motor sequence without thinking about it. When concert pianist Peter Serkin performs a Mozart concerto with the Boston Symphony Orchestra, his interpretation is driven by his brain’s extensive procedural knowledge acquired over years of rigorously practicing that piece; he has integrated the individual key presses into a fluent whole, and performs without conscious reference to individual finger movements.
Before neuroscientists investigated the distinctions between different types of learning, other thinkers in philosophy, computer science, and psychology theorized more abstractly along these lines. The British philosopher Gilbert Ryle wrote about a particular division in his 1949 book The Concept of Mind, scolding theorists of the mind for putting too much emphasis on knowledge as the foundation of intelligence and for failing to consider what it means for an individual to understand how to carry out tasks. Ryle termed this difference knowing that versus knowing how. When we learn a skill, such as a new dance move, we may be unable to articulate the sequence of commands the brain sends to our muscles and the resulting feedback—knowing that—but we can show off the new move to our admiring friends—knowing how.14
Henry’s ability to learn new motor skills demonstrated convincingly that the areas that had been excised in his operation—the hippocampus and surrounding structures—were not necessary for learning new motor skills. So of course the next question we wanted to answer was, what critical brain circuits do support motor learning? In order to research this question, we focused on non-amnesic patients whose brains were damaged in other ways.
Since the dawn of the twentieth century, scientists
have known that two structures, the striatum and the cerebellum, play important roles in motor control. The striatum includes the caudate nucleus and the putamen, two collections of neurons under the cortex. They receive signals from above and below—neurons in the cortex and neurons lower in the brain. The striatum receives messages from specific cortical areas and sends signals back to the same areas by way of the thalamus, an area in the center of the brain that integrates sensory and motor activities. As a result, the striatum is well informed about what is going on in the body and in the world, and as such is well qualified to learn difficult motor skills.
The cerebellum, Latin for little brain, is a large, complex structure at the back of the brain under the visual cortex. Henry’s cerebellum was greatly reduced in size, but we could not tell from his MRI scans exactly where the damage was. This structure is directly connected to the striatum and to several areas in the cortex by means of closed circuits. Because the cerebellum receives information from many parts of the brain and spinal cord, it stands at the frontline of motor control.
Abnormalities in the striatum are responsible for more than twenty disorders, including two progressive brain diseases, Parkinson disease and Huntington disease. Within the striatum, the putamen is most affected in Parkinson disease and the caudate nucleus in Huntington disease.
Parkinson disease is a common affliction with an unknown cause that typically strikes people in their fifties, men more than women. Someone afflicted with this disease often has an expressionless face, slow movement, shaking of the hands, stooped posture, and shuffling steps. In the brain, Parkinson begins with a loss of neurons in the substantia nigra, a bundle of gray matter under the cerebral cortex that normally sends out fibers, which carry the neurotransmitter dopamine up to the striatum. But when cells in the substantia nigra die, as they do in Parkinson, the supply of dopamine transmitted to the putamen is diminished, causing motor abnormalities.15
Huntington disease is a rare hereditary disorder caused by neuronal loss in the caudate nucleus, accompanied by cell death in the cortex. The cause is a defect in the HTT gene on chromosome 4. A particular segment of DNA in this gene is repeated up to one hundred twenty times in people with the disease, but only ten to thirty-five times in unaffected people. While the hallmark of Parkinson disease is too little movement, the salient feature of Huntington disease is too much movement. The most striking symptom of Huntington is involuntary, jerky movements of the face, arms, and hips, which makes it seem as if the afflicted person is doing a dance.16
Studying Parkinson and Huntington diseases side by side is instructive because the initial damage occurs in different parts of the striatum—the putamen in Parkinson, and the caudate in Huntington—providing complementary evidence about the localization of different skills.
In the early 1990s, to explore the role of the putamen in motor-skill learning, my lab studied mirror tracing in early-stage Parkinson patients. We asked the patients to perform a test similar to the one Milner conducted with Henry. The task was to trace around a six-pointed star as quickly as possible without veering off the track. Because the Parkinson patients had a motor disorder, they took longer to get around the star, traced more slowly, and paused more frequently than control participants. These deficits, which we had anticipated, were measures of motor performance, not motor learning. To see whether Parkinson affected their motor-skill learning, we documented their improvement over three consecutive days of training and then compared their rate of change to that of control participants. A digitizing tablet placed under the star told us precisely where the stylus was from the start to the finish of each trial, millisecond by millisecond. These data allowed us to calculate several different indices of motor-skill learning. These measures were uncontaminated by deficits in motor performance because they focused strictly on each individual’s rate of progress, regardless of the performance level at which he or she started.
Although the Parkinson patients improved on all of these measures over the three days of training, their progress was slower than the controls’. On several measures of learning—how long it took them to trace around the star, how long it took them to get back on the path when they veered off, and how much time they spent going backward—the Parkinson patients showed less improvement over the three days than control participants. The difficulty that the patients experienced on this tracing task provided direct evidence that the striatum participates in complex motor-skill learning, giving credence to the idea that Henry enlisted his striatum to learn the motor skill.17
The finding that our Parkinson patients were impaired on mirror tracing did not necessarily mean that they would perform poorly on all motor-learning tasks. Many areas throughout the brain support motor behaviors, and it would not make sense for all these areas to be dedicated to a single, all-purpose motor-learning function. The brain is an efficient machine that does not give its component parts redundant tasks. We, therefore, hypothesized that different motor skills engage separate cognitive and neural processes. It was possible that the particular brain circuit within the striatum recruited for mirror tracing would not be necessary to perform a different skill-learning task, such as learning a specific sequence of responses.
To further explore the scope of the skill-learning deficit in Parkinson disease, my lab adapted a sequence-learning procedure in the early 1990s that Mary Jo Nissen and Peter Bullemer first introduced in 1987. Our patients with Parkinson disease sat in front of a computer terminal and saw four small white dots arranged horizontally across the bottom of the screen. A custom-designed keyboard just below held four response buttons corresponding to the four dots. Participants rested their left middle and left index fingers on the two left buttons, and their right middle and right index fingers on the two right buttons. On each trial, a small white square appeared below one of the four dots, and the task was to press, as quickly as possible, the key that corresponded to the location of the square. Unbeknownst to the participants, the squares appeared in a ten-item sequence that repeated ten times on each trial, for a total of one hundred key presses. We knew that if the participants were learning the sequence, their response times would become progressively faster on trials that contained the repeated sequence but would not speed up on other trials in which the sequences were random (see Fig. 13).18
Parkinson patients and controls performed the sequence-learning task on two consecutive days. The performance times of the two groups did not differ; the Parkinson patients performed normally. Response times for the repeated sequences decreased during the first day, and participants retained this learning overnight, performing as well at the start of the second day as they had at the end of the first. The Parkinson patients’ reduction in response times for repeated sequences indicated that they had acquired procedural knowledge in a normal fashion.
A comparison of the performance by Parkinson patients on the mirror-tracing and sequence-learning tasks, impaired in the former and preserved in the latter, indicates that skill learning is not a homogeneous concept and that different kinds of skill learning have different neural underpinnings. The memory network in the striatum that normally supports the acquisition of the mirror-tracing skill was dysfunctional in our Parkinson group, but a neural circuit that was spared in the same patients mediated normal sequence-specific learning. We then asked, what is this circuit, and is it compromised in other diseases?
Research on Huntington disease provided a clue about the substrate for sequence-specific learning, informing us about the effect of damage to the caudate nucleus on this task. When Nissen administered her sequence-learning task to a group of Huntington patients, they showed impaired learning. Although their motor function was sufficient to perform the test satisfactorily, they were slower and less accurate than the twenty-one control participants. Their deficit was unrelated to cognitive dysfunction. This result tells us that the caudate nucleus plays a critical role in sequence-specific learning.19
The experiments in Parkinson and Huntingto
n diseases, examined side by side, illustrate how different pathologies within the striatum can exert distinct effects on sequence learning: Parkinson patients are normal, while Huntington patients are deficient. This dissociation suggests that the caudate nucleus, affected early in Huntington, is a critical substrate for sequence learning, and that the putamen, affected early in Parkinson, is not.
Brain areas beyond the striatum are also engaged in motor learning. Since the late 1960s, neuroscientists have gained another perspective on the brain localization of skill acquisition from studying animals and patients with abnormalities of the cerebellum. Their symptoms include poor coordination, slow movement, tremor, and slurred speech. People who are heavily intoxicated demonstrate these symptoms, as did Henry, minus the tremor. One group, patients with cerebellar degeneration, are impaired at sequence learning, but several studies show that their fundamental deficit may differ from and be more severe than that in Parkinson patients. Patients with cerebellar degeneration were also slower and less accurate than control participants when tracking a simple geometric pattern that they saw in mirror-reversed vision, similar to the conditions under which Henry successfully traced around the star. In 1962, we learned from Henry that medial temporal-lobe structures are not necessary for learning the mirror-tracing skill, and thirty years later we learned that the cerebellum is necessary for that kind of learning.20
The deficit in mirror tracing in cerebellar disease reflected the patients’ inability to use the feedback they received during the test to guide their movements. Although they could see the visual display and feel the changing position of their arms and hands, they could not convert this information into new commands to activate their muscles. They could not overcome their ingrained responses. This difficulty is not specific to a single task but essentially a general failure to integrate input from the senses with commands to the muscles. Consider typing on a keyboard as an example. When we perform this skill, we receive information from several sources—the feel of the keys on our fingertips, the position and movement of our fingers and hands, and the visual image of our hands and the document on the computer screen. When we type, our brains automatically combine all these inputs, telling our fingers how to move to hit the right keys in the right order and with sufficient force to make the desired letters appear on the screen. Healthy people have no difficulty acquiring this complex motor skill if they practice.
Permanent Present Tense Page 19