Permanent Present Tense
Page 22
A big gap in scientific knowledge remains, however. How do motor mechanisms in different parts of the brain—the primary motor cortex, the striatum, and the cerebellum—coordinate their individual contributions to achieve the complex enterprise of motor learning necessary in our constantly changing world? Funcional MRI holds great promise for dissecting the individual processes that govern different kinds of skill learning and for documenting when and how the various brain networks work together. These studies indicate that many cortical areas are recruited during motor learning, suggesting that a broad network of motor and non-motor areas supports skill acquisition (see Fig. 15).
Any athletic skill—dribbling a soccer ball, sinking a free throw, serving an ace—requires extensive training. Increasingly improved performance is coupled with changes in the brain. In 1998, a team of neuroscientists at the National Institute of Mental Health set out to examine the neural changes that occur with continued practice, and to discover how much practice is needed to bring about detectable changes during skill acquisition. They chose the primary motor cortex, a strip at the back of the frontal lobe that sends out neural codes for movements, as their area of interest because it controls voluntary movement and also supports motor learning. The researchers asked healthy adults to practice making sequences of finger movements for several weeks. The participants touched their thumb to the other four fingers one at a time in a particular order: pinky, index finger, ring finger, middle finger, and pinky. They practiced this sequence for ten to twenty minutes every day for five weeks, and as the weeks passed, they completed more sequences and made fewer errors in each thirty-second test.33
To capture what was happening in their brains, the researchers conducted a functional MRI study once a week, in which the participants performed the sequence inside an MRI scanner. The resulting MRI images showed activation in the hand area of their primary motor cortex that expanded as the participants’ skills improved, and the change lasted for several months. This finding provided evidence that practicing a motor skill encourages additional motor neurons to be active and incorporates them in a focal brain circuit that represents the trained-motor sequence. This indisputable evidence of neural plasticity in adult brains may represent the kind of modification that is responsible for motor-skill learning. The main function of the primary motor cortex is to tell our muscles what to do, but in addition, neuronal firing in this area during motor learning can change synaptic strength—the ability of one cell to excite its synaptic partner cells—thereby promoting memory consolidation. The neural circuits within the primary motor cortex are adaptable on a moment-to-moment basis during the acquisition, consolidation, and retrieval of motor skills.34
We demonstrated in the lab that Henry could acquire several motor skills, such as mirror tracing, rotary pursuit, and bimanual tracking. His normal primary-motor cortex likely played a role in his capacity to acquire these new motor skills and to use his walker deftly in everyday life. But it is also probable that helpful changes occurred in other areas in Henry’s brain—some dedicated to motor function and others to cognitive processes.
Motor learning usually occurs slowly over many practice sessions, and the complex mechanisms that support skill acquisition change as learning progresses. Training-induced plasticity can be seen in expansions of both gray matter, the cell bodies of neurons, and white matter, the fiber tracts that connect different groups of cells. Initially, the primary-motor cortex and adjacent motor areas are called online, with increased activity in the prefrontal and parietal cortices and the cerebellum as well. Later, when the skilled movements become more automatic, learning still engages the primary-motor cortex and, in addition, the striatum and cerebellum. Movement representations expand in the motor cortex and other cortical areas that are specialized for planning, perceiving movement, controlling eye movements, and calculating spatial relations. These areas work together to achieve motor-memory formation. Multiple circuits throughout the brain are engaged in motor-skill learning, but, as Henry showed us, those in the medial temporal lobes are unnecessary.35
Modern brain-imaging tools have enabled us to observe in healthy individuals what the critical circuits are doing during the course of practice. Researchers wanted to know what areas are active as people advance from novice to expert on a particular task. In 2005, neuroscientists used functional MRI to show that when participants received extensive training on a sequence-learning task (Nissen and Bullemer’s task, described earlier), their brain activity in the novice stage differed from that in the later, automatic stage. Initially, areas in the prefrontal cortex and a deep motor area, the caudate nucleus, were highly active, but this activity decreased when performance became automatic with practice, suggesting less reliance on cognitive control processes. The finding that the striatum (caudate and putamen) plays a key role in the acquisition of motor-sequence knowledge is consistent with the finding of motor-learning deficits in Parkinson disease and Huntington disease, both of which damage the striatum.36
Starting with the idea that learning occurs gradually from one training session to the next, two neuroscientists at Concordia University in Montreal conducted an ambitious study in 2010 to document changes in brain activity across five consecutive days of skill acquisition. Scanning participants at every training session was unnecessary, so they performed the same motor-learning task inside the scanner on days one, two, and five, and outside the scanner on days three and four. The researchers discovered that, as performance improved, several motor areas that were initially active became less active. These decreases may have occurred because the brain was paying less attention to repeated stimuli and no longer needed to correct errors as learning progressed. At the same time, small areas within the primary-motor cortex and the cerebellum showed increases in activity with improvement.37
These pockets of increased activity within a network of overall decline in activity could represent the areas where motor memories are ultimately stored. The researchers speculated that separate populations of neurons within the primary-motor cortex encode and express different facets of motor-sequence learning. One population of neurons, which is activated by performance errors, is dedicated to rapid learning; it talks to a declarative memory network. The other population, which shows resistance to forgetting, is specialized for gradual learning; these neurons talk to a network dedicated to learning procedures—the how to do it. These two populations of neurons work cooperatively.
We now have compelling evidence that the evolution of a complex skill from novice to expert is not a single process. Different timescales operate in motor memory, and their contributions change over time. The ability to tease apart neutrally distinct processes helped us to understand Henry’s performance on the reaching task when he had to compensate for the added force in the mechanical arm. Although he could retain the skill, his rate of learning lagged behind that of the controls. The results of the 2010 functional MRI experiment lead me to speculate that Henry’s slow progress could be blamed on his damaged cerebellum, which, in healthy brains, makes an important contribution to the early stages of learning.
For most of us, nondeclarative and declarative memory processes are intertwined. You may not be able to describe what you are doing when you ride a bicycle, but you can think back to the days when you rode with training wheels, or when a parent first let go of the back of the bike and let you ride on your own. Skills, experiences, and knowledge are all linked. What remains fascinating about Henry’s case is that it showed how a skill could bloom in the brain, even as the experience behind it was irretrievably lost.
Nine
Memory without Remembering II
Classical Conditioning, Perceptual Learning, and Priming
From the mid 1980s through the late 1990s, members of my lab and I expanded our thinking and efforts to investigate the nature of learned behavior. In a broad theoretical context, we designed new experiments to unravel the different cognitive and neural mechanisms that account for nondeclarati
ve memory. As we have seen, Henry was able to unconsciously acquire new motor skills. We also found that he could successfully perform other nondeclarative memory tasks. In our studies of classical conditioning, perceptual learning, and repetition priming, Henry demonstrated what he had learned through his performance of the tasks and not through conscious declarative memory. His proficiency indicated that these forms of unconscious learning, like motor learning, occur in brain circuits outside the medial temporal lobes. Henry played a major role in the development of thinking about each of these kinds of nondeclarative knowledge.
During this time, my colleagues and I came to realize Henry’s limitless worth as a research participant. We continued to be amazed at how many different contemporaneous scientific findings could be related to, or strengthened by, further examining him, and our research with Henry was certainly a boon to my lab’s reputation. Although our publications describing his results made up only 22 percent of our total output, these articles were, and continue to be, high profile and widely cited.
Classical conditioning is a learned behavior that capitalizes on a reflex, such as salivation, a kneejerk, or blinking. This form of nondeclarative learning has been for many decades a valuable tool for research in animals and humans. In experiments using classical conditioning, a neutral item, such as the sound of a bell, is paired repeatedly with another item, such as food, which reliably produces a reflex, such as salivation. Eventually, the sound of the bell itself elicits the reflex response. When the subject salivates in response to the bell, we know that during the multiple presentations of the bell with the food, the animal has learned to associate the two.
The Russian physiologist Ivan Pavlov discovered classical conditioning in the early 1900s while studying digestion in dogs. His technique for eliciting this phenomenon capitalized on a simple reflex: when an animal has food in its mouth, it salivates. Pavlov ingeniously observed that a similar reflex could be activated by the smell of the food, by seeing the person who delivered the food, or even by the sound of the person’s footsteps. The dogs learned that these sensory cues meant that food was on the way. In Pavlov’s experiments, his assistant rang a doorbell just before the dogs received their food. After being exposed repeatedly to these paired stimuli—the bell and the appearance of food—the dogs salivated when they heard the bell, indicating they had learned to associate the sound with food.1
Establishing links between items and emotions is a popular strategy for the advertising industry. Picture an ad for a Caribbean resort dominated by beautiful, smiling couples taking sunset strolls on the beach, swimming with tropical fish, and enjoying massages. If we decide to take a tropical vacation, we will likely choose the resort that, thanks to conditioning, we have learned to associate with fun and romance.
My colleagues and I knew from previous experiments that both the cerebellum and hippocampus play a role in the formation of conditioned responses, but we wanted to test how important each area was for this kind of learning. We reasoned that if Henry showed conditioned responses to stimuli without a functioning hippocampus, then his residual cerebellum likely mediated the learning. If Henry did not show conditioned responses, the result would be uninterpretable: we would not know whether the damage to his hippocampus, the cerebellum, or both was responsible for the deficit. All of Henry’s neurological examinations dating back to 1962 uncovered signs of cerebellar dysfunction, and his MRI scans showed marked cerebellar atrophy, indicating cell death. But despite his extensive hippocampal and cerebellar damage, Henry did exhibit conditioned responses in our experiments. Although his learning was much slower than that of a healthy man his age, he demonstrated remarkable retention by showing conditioned responses in studies carried out two years after the initial learning sessions.
We first studied Henry’s capacity for classical conditioning in 1990, using an eyeblink conditioning experiment, which tested whether he showed a conditioned reaction by blinking in response to a tone that preceded a puff of air in his eye. For testing, Henry sat in a comfortable chair in a quiet room at the MIT Clinical Research Center. He wore a headband that held an air-puff jet and a monitor to record his eyeblinks. The researcher gave Henry these instructions: “Please make yourself comfortable and relax. From time to time, you will hear some tones and feel a mild puff of air in your eye. If you feel like blinking, please do so. Just let your natural reactions take over” (see Fig. 16).2
Over an eight-week period, we administered two kinds of conditioning tasks: delay conditioning and trace conditioning. During delay conditioning, a tone came first, immediately followed by an air puff, and the two stopped together. Each training session lasted about forty-five minutes and included ninety trials. On eighty of those trials, Henry experienced both the tone and the air puff, giving him the opportunity to unconsciously associate the two. If an eyeblink occurred in the very short interval—less than a second—between the tone and the air puff, we counted it as a conditioned response. This blink indicated that Henry had learned to associate the sound with the forthcoming air puff to his eye, and unconsciously blinked in anticipation of the air puff. On the ten remaining trials, Henry heard the tone alone, and if he blinked immediately, we scored that blink also as a conditioned response. Tallying up the results was simple: we counted how many times he showed a conditioned response in the tone-plus-air-puff condition and in the tone-alone condition. During trace conditioning, a silent interval occurred between the tone and the air puff, meaning that Henry’s brain had to hold the tone online for half a second to associate it with the air puff that followed. As before, Henry got credit for a conditioned response if he blinked right after the tone.3
Throughout the conditioning sessions, we showed Henry films to keep his attention focused on something pleasurable. One of his favorites was The Gold Rush, a Charlie Chaplin comedy, and he liked a documentary about the 1939 New York World’s Fair, which he had attended with his mother. Although we turned off the sound so Henry could hear the tone, he did not complain and enjoyed the testing experience. All the while, Henry was unaware that he was in a memory experiment, confirming that this task indeed tapped his nondeclarative memory processes. We compared his conditioning scores with those of a healthy sixty-six-year-old man to see whether Henry was impaired and to what extent.4
Henry produced conditioned responses in both the delay and trace procedures, an achievement tied to modifications in his brain during the nondeclarative training experience. But overall, his performance was inferior to that of the control participant. He required more trials than the control to reach the criterion of learning—blinking in eight out of nine consecutive trials when the tone was presented alone. For delay conditioning, the control attained the goal—eight out of nine correct trials—in 315 trials, whereas Henry required 473. Five weeks after the delay-conditioning experiment, we introduced the trace procedure. For trace conditioning, Henry’s control participant reached the criterion of learning on the first trial, whereas Henry required ninety-one trials. It seemed he was impaired on both delay and trace conditioning.5
We gained some understanding of Henry’s performance during delay and trace conditioning by looking at trials that included the tone but no air puff. On some of these trials, although he produced an eyeblink after the tone, it came too late—after our cutoff for conditioned responses, 400 milliseconds; consequently, these blinks did not count as conditioned responses. His slowness to respond explains, at least in part, why he required over a hundred extra trials to reach the learning criterion for delay conditioning. But a second measure of learning, how much Henry remembered after a five-week break, showed that some of the learning in the delay procedure carried over. This time, Henry required 276 trials to show conditioning—197 fewer than before—while the control participant took ninety-one trials, twenty-four fewer than previously. Although Henry’s 42 percent gain was less impressive than the control’s 79 percent, he clearly made substantial progress in acquiring the conditioned response. This experiment tells u
s that the hippocampus is not essential for classical conditioning to occur in either the delay or trace procedure. Henry’s ability to learn, although reduced, forces us to speculate about what parts of his remaining cerebellum could have supported this learning.6
Two years after the initial conditioning experiment, we examined the durability of this learning. In the new experiments, Henry gave us a striking demonstration of nondeclarative learning: in only nine trials, he reached the criterion of learning for trace conditioning, showing that over the two-year period, the learned conditioned responses were consolidated and securely stored in his brain. This unambiguous result indicated that the hippocampus is not essential for storing a trace of the tone for half a second to associate with the air puff. For conditioning to occur, Henry must have engaged his remaining cerebellum and cortical areas to retain the learned responses for two years. He showed unconscious, nondeclarative learning despite having no declarative memory of the experience: he did not recognize any of the researchers, apparatus, instructions, or procedures, and was unaware of what he had learned.7
To better understand the differences between delay and trace conditioning, we turned to the work of three memory researchers at the University of California, San Diego. In 2002, they marshaled evidence from experiments in animals and humans—including patients with amnesia—to highlight that awareness is necessary for trace conditioning but not for delay conditioning. Awareness in this learning task is the declarative knowledge of the relation between the tone and the air puff—the tone signals the imminent arrival of the air puff. Our control participant must have had this declarative knowledge—this awareness—because he acquired trace conditioning in one trial. During the course of the experiment, healthy participants came to understand, on a conscious level, that the tone forecast the occurrence of the air puff, and would come to expect it.8