Chapter 4

Learning by Doing

"When starting a kiss, the rule of thumb is to start slow. This just makes sense, and it lets everyone get used to the dynamics of that particular kiss. A slow start is a good introduction... and sometimes the kiss should just stay slow. Jumping into rapid tongue maneuvers can scare your partner, and is rude to boot. Athletes always warm up before moving onto serious play... why should kissing be any different?"

(Hays, Allen, & Hanish, 2005)

DO YOU REMEMBER YOUR FIRST KISS? For some, it is a magical memory. For others, that kiss was a somewhat awkward experience in which a single thought kept recurring: "Am I doing this right?" Kissing is simple enough in concept. Take your lips and press them against someone else's lips. What could be easier? After just a few experiences with bad kissers, however, it becomes clear that this apparently simple ability is not one that humans are born with. By the same token, a single encounter with an especially good kisser is enough to make you appreciate that kissing requires some skill.

The success of a first kiss may depend in part on the setting and the partner, but most young people are savvy enough to know that they need to practice if they want their first real kiss to be a good one. Practice might consist of kissing one's own hand or arm, a pillow, or a stuffed animal. The hope is that these practice sessions will give you an edge when a real opportunity comes along. Practicing by kissing your hand or arm is a good strategy, because that way you get feedback about what your lips feel like. Attending a Kissing 101 class might also help, but you will not become an adept kisser by memorizing lists of rules about how to kiss. To become an expert, you need to kiss (a lot), you need to get feedback about your kissing, and, most important, your brain has to store memories of your kissing successes and failures.

This chapter describes how repeated experiences incrementally enhance the performance of a skill by gradually modifying memories of how the skill can best be executed. As you will discover, repeated experiences not only can change how a person performs a skill, such as kissing; they also can change the structure of the brain circuits that are used to perform that skill. Skill memories are formed and processed by several brain regions, including the basal ganglia, the cerebral cortex, and the cerebellum. People with damage in one or more of these brain regions have trouble learning new skills, as well as performing skills already learned.

4.1 Behavioral Processes

The previous chapter dealt with memories for events and facts in other words, information a person remembers and knows. Skill memory, in contrast, consists of what a person knows how to do. By reading this sentence you are exercising a skill that you learned a long time ago. Reading may seem so effortless now that you can hardly recall the challenge of learning to read. When you turn a page, highlight a sentence, type or write notes, or think about what you'll need to do to remember the contents of this chapter, you are accessing memories of several different skills.

Qualities of Skill Memories

A skill is an ability that you can improve over time through practice. Skill memories are similar in many respects to memories for events (also called episodic memories) and facts (semantic memories), but they also possess some unique qualities (Table 4.1). Like memories for facts, skill memories are long lasting and improved by repeated experiences. Unlike memories for events and facts, however, skill memories can't always be verbalized; moreover, skill memories may be acquired and retrieved without conscious awareness. As you'll recall from Chapter 3, psychologists sometimes classify skill memories as nondeclarative memories, because these memories are not easily put into words.

All memories for events and facts depend on skill memories, because the abilities to speak, write, and gesture to convey information are learned abilities that improve over time with practice. In contrast, skill memories do not necessarily depend on verbalizable memories, although memories for events and facts can play an important role in acquiring skills. Given the dependence of memories for events and facts on skill memories, perhaps it would be fairer to describe declarative memories as "non skill" memories, rather than calling skill memories "nondeclarative."

Table 4.1

Comparison of Memories for Skills, Events, and Facts

Memories can be classified in many ways, but often don't fit neatly into the conventional classification schemes. Contemporary researchers generally classify skill memories into two basic types: perceptual-motor skills and cognitive skills (Gabrieli, 1998; K. M. Newell, 1991; Rosenbaum, Carlson, & Gilmore, 2001; van Lehn, 1996; Voss & Wiley, 1995).

Perceptual-Motor Skills

The kinds of skills you are probably most aware of are those that athletes demonstrate when they compete. More mundane skills include opening and closing doors, driving a car, dancing, drinking out of a glass, and snapping your fingers. These are all examples of perceptual-motor skills: learned movement patterns guided by sensory inputs.

Consider dancing. An important part of dancing is being able to move your body in certain established patterns. This requires significant voluntary control of your movements. If you can't control where your arms go, you'll end up being more of a spectacle than a dance sensation. Dancing is more than just repeatedly moving your feet and arms in a pattern, however; you also have to move to the beat (that is, respond to auditory inputs). In addition, some well-established dances, such as the Hokey Pokey or the Macarena, require specific movements to be performed at specific points in a song. The goal in learning these kinds of dances is to perform a consistent sequence of movements in a prescribed way. Professional ballet dancers, too, learn precisely choreographed dance sequences. Psychologists classify skills such as ballet dancing, which consist of performing predefined movements, as closed skills. Other kinds of dancing, such as salsa or swing dancing, also involve particular movement patterns, but dancers may vary the way they combine these movements when they dance, at least in social dance settings. Such dances depend to some extent on the dancers' predicting (or directing) their partner's next move. Researchers classify skills that require participants to respond based on predictions about the changing demands of the environment as open skills.

These classifications apply to a wide range of perceptual-motor skills. For example, athletes who are gymnasts or divers are perfecting closed skills, whereas athletes who participate in coordinated team sports such as soccer or hockey depend heavily on open skills. Dogs can learn to catch a Frisbee (an open skill), and they can also learn to play dead (a closed skill). Catching a Frisbee is an open skill because many environmental variables—such as quality and distance of the throw, wind speed, and terrain characteristics—determine which movements the dog must make to perform the skill successfully. Most perceptual motor skills contain aspects of both closed skills and open skills, and so it is better to think of any particular skill as lying somewhere along a continuum from open to closed (Magill, 1993).

Most research on perceptual-motor skills focuses on much less complex skills than those needed to dance or play soccer. Skills studied in the laboratory might consist of pressing buttons quickly or tracking the position of a moving object (Doyon, Penhune, & Ungerleider, 2003). It's not that knowing how a person learns to dance is uninteresting to psychologists. Rather, research psychologists want to keep things as simple as possible so they can control the relevant variables more precisely. This gives them a better chance of understanding how experience affects an individual's ability to perform a particular skill. For example, it is much easier to assess quantitatively whether someone's tracking abilities are improving than to measure improvements in their dancing abilities.

Cognitive Skills

What are some other activities that improve with practice? How about playing cards, budgeting your money, taking standardized tests, and managing your time? These are all examples of cognitive skills, which require you to use your brain to solve problems or apply strategies, rather than to simply move your body based on what you perceive (J. R. Anderson, Fincham, & Douglass, 1997; Singley & Anderson, 1989). Researchers often conduct experiments on cognitive skills that participants can learn relatively quickly, such as those used to solve simple puzzles like the Tower of Hanoi (Figure 4.1). In this puzzle, the objective is to move different sized disks from one peg to another, one disk at a time (we discuss this task in greater detail in Chapter 5). The puzzle would be trivially easy, except that one of the rules is that you cannot put a larger disk on top of a smaller one. The numbered sequence in Figure 4.1 shows one solution to the puzzle. Normally, people get better at this puzzle with practice. This is not because they are getting better at moving the disks from one peg to another (a perceptual motor skill), but because they are learning new strategies for moving the disks so that they end up in the desired position (J. R. Anderson, 1982).

Psychologists usually associate cognitive skills with the ability to reason and solve problems. Descartes proposed that the ability to reason is what distinguishes humans from other animals. Descartes would probably have been willing to accept that dogs can store memories for perceptual motor skills such as how to catch a Frisbee, but he would have considered it impossible for a dog or any other nonhuman animal to learn a cognitive skill. Following Descartes' lead, many psychologists assume that only humans can reason. Certainly, this is one reason that most of what we currently know about cognitive skills comes from studies of humans.

Nevertheless, humans are not the only animals with cognitive skills. To give an example, it was once thought that only humans used tools and that this particular problem solving ability played a key role in the evolution of the human mind. In the past two decades, however, psychologists and animal behavior researchers have described tool use in many animals (Beck, 1980; Hart, 2001; Hunt, Corballis, & Gray, 2001; Krutzen et al, 2005; Whiten et al, 1999). Researchers have observed chimpanzees in the wild that learn how to use stones to crack nuts (Whiten & Boesch, 2001). In the lab, experimenters have taught primates and other animals to use various tools. There is also recent evidence that, in the wild, animals can teach themselves to use tools—for example, dolphins have learned to use a sponge while foraging (Krutzen et al, 2005), as shown in Figure 4.2.

Tool use is an ability that typically involves both perceptual-motor and cognitive skills. Movement patterns required to use a tool improve with practice, and the recognition that a particular tool (or strategy) can be useful in solving various problems also improves with practice. Some animals can use tools more flexibly and imaginatively than others. By comparing different animals' abilities to learn perceptual-motor and cognitive skills, and by exploring which neural systems they use when forming and retrieving memories of different skills, scientists are beginning to gain a clearer understanding of the brain systems underlying skill memories.

Historically, philosophers and psychologists have distinguished perceptual motor skills from cognitive skills. However, recent evidence suggests there are many more similarities in how humans learn and remember both types of skills than was previously thought (Rosenbaum, Carlson, & Gilmore, 2001). As you read this chapter, consider what, if anything, makes memories of perceptual motor skills different from memories of cognitive skills. Is it how they are learned, how they are remembered, how they are forgotten, or something else? Perhaps the differences lie not in how a person forms and recalls these memories but where in the brain the memories are formed and recalled. We will return to these questions about how and where memories for skills are processed later, in the Brain Substrates section.

Now that you know how researchers classify different kinds of skills, let's consider the question of what allows some people to excel at a particular skill. You won't be surprised to hear that practice is an important factor. We'll examine how different kinds of practice affect performance and retention of skill memories, and why people who are great at one skill are not necessarily as good at other, similar skills. We'll also describe a classic psychological model of skill learning.

Expertise and Talent

You might be able to dance as well as an all-star basketball player, a virtuoso pianist, or a Nobel Prize-winning scientist, but they clearly have mastered other skills at a level that would be difficult for you to match. Different individuals start with different skill levels, and the extent to which practice can improve their performance levels also varies from one person to the next. People who seem to master a skill with little effort (the way Mozart mastered anything related to music) are often described as having a talent or "gift" for that skill, and people who perform a skill better than most are considered to be experts. The people who start off performing a skill well are often those who end up becoming experts, but someone who initially has little ability to perform a skill may, with practice, become better at that skill than someone who seemed destined to become a star. So, if your significant other is currently lacking in the kissing department, don't lose hope! Additional practice may yet unleash his or her full potential.

What role does talent play in achieving expertise in cognitive or perceptualmotor skills? Even child prodigies are not born able to perform the skills that make them famous. Like everyone else, they learn to perform these skills. Mozart's father, a professional musician, trained Mozart extensively from a young age. So it's difficult to determine whether Mozart's musical abilities were a result of heredity or of his father's teaching abilities.

Psychologists have attempted to gauge the role of genetics in talent by conducting studies with twins—some identical (sharing 100% of their genes) and some fraternal (sharing, like other siblings, 50% of their genes)—who were raised in different homes. Other twin studies look at the differences between twins reared together.

In one large study of twins reared apart, researchers at the University of Minnesota trained participants to perform a skill in which they had to keep the end of a pointed stick, called a stylus, above a target drawn on the edge of a rotating disk, as shown in Figure 4.3a (Fox, Hershberger, & Bouchard, 1996). Researchers frequently use this task, known as the rotary pursuit task, to study perceptual-motor skill learning. The task requires precise hand-eye coordination, much like the coordination used by potters to shape a clay pot on a pottery wheel. When individuals first attempt the rotary pursuit task, they generally show some ability to keep the stylus over the target, but often have to adjust the speed and trajectory of their arm movements to do so. With additional practice, most individuals rapidly improve their accuracy, increasing the amount of time they can keep the stylus tip over the target (Figure 4.3b).

The researchers found that when they trained twins to perform the rotary pursuit task, identical twins' abilities to keep the stylus on the target became more similar as training progressed, whereas fraternal twins' abilities became more dissimilar. That is, during training, the performance of one twin became more correlated with the performance of the second twin only when the two twins shared 100% of their genes (Figure 4.3c). Put another way, if you were to view videos of the participants' hands, after training, as they attempted to keep the stylus above the rotating target, you would judge the movements of identical twins' hands to be the most similar. If you saw a pair of identical twins performing this task side by side after training, their movements might remind you of synchronized swimming. In the case of fraternal twins, however, you would probably judge their movements after training to be very dissimilar. For example, one twin might keep the stylus over the target continuously, while the other twin increased her speed every few seconds to catch up with the target.

One interpretation of these data is that, during the experiment, practice decreases the effects of participants' prior experiences on the accuracy of their tracking movements and increases the effects of genetic influences. Identical twins have identical genes, so when practice increases the role of their genes in behavior, their behavior becomes closer to identical. Because fraternal twins have different genes, increasing the role of their genes in behavior makes their behavior more different. Researchers have tested for such effects only in tasks, such as the rotary pursuit task, that require individuals to learn simple perceptual motor skills. It is possible, however, that practice has similar effects on more complex perceptual-motor and cognitive skills. It could be that you have hidden talents that you're unaware of because you have never practiced the skills that require those talents, or have not practiced them enough. Perhaps future genetic analyses will discover biological correlates of specific talents, permitting identification of individuals who have an inherited propensity to perform certain skills exceptionally well. Currently, however, the most common way of evaluating an individual's potential to excel at a particular skill is the nonscientific one of asking someone with expertise in the skill to make a subjective assessment of that person's ability.

Some psychologists argue that innate talent plays no role in expertise and that practice alone determines who will become an expert (Ericsson, Krampe, & Tesch-Romer, 1993; Ericsson & Lehman, 1996). Until more is known about how practice affects skill memories, it will be difficult to reliably predict either an individual's maximum level of skill performance or the amount of practice someone needs to reach peak performance. In any case, scientists investigating skill memory in experts suggest that practice is critical in determining how well a person can perform a particular skill. Researchers often conduct studies of skill memories in athletes or chess masters, or other professional game players, for several reasons: (1) people who learn to play games outside a research lab provide good examples of "real world" memories; (2) it is not difficult to find people with widely varying levels of expertise in these games, which can often be quantitatively measured through performance in competitions; and (3) games require a variety of perceptual-motor and cognitive skills.

A person must practice thousands of hours to become a master chess player, learning more than 50,000 "rules" for playing chess in the process (Simon & Gilmartin, 1973). Researchers studying expert chess players found that experts and less experienced players scan the game board (a visual-motor skill) differently (Charness, Reingold, Pomplun, & Stampe, 2001). When chess masters look at chess pieces, their eyes move rapidly to focus on a small number of locations on the board, whereas amateur chess players typically scan larger numbers of locations and do so more slowly. When experts stop moving their eyes, they are more likely than non-experts to focus on empty squares or on strategically relevant chess pieces. Similarly, inexperienced soccer players tend to watch the ball and the player who is passing it, whereas expert players focus more on the movements of players who do not have the ball (Williams, Davids, Burwitz, & Williams, 1992).

Humans may need to practice many hours to become experts at chess, but practice is not a universally necessary prerequisite for expert chess performance. Computer programmers have designed software that can compete with the best chess players. For example, Deep Blue, a chess playing computer designed by IBM, defeated world champion Garry Kasparov in 1997. Computers access large databases of stored information to replicate some of the abilities of human experts. If a skill is an ability that improves with practice, chess playing computers can be considered experts without skills, unless they are programmed to improve their performance based on past experiences. Although humans also make use of large amounts of information in performing certain skills, the way their brains store and access information differs greatly from the way computers do this. For example, if one computer can be programmed to perform a particular task, the same ability is usually easy to replicate in another computer. If only humans could acquire abilities so easily! For better or for worse, information can't yet be copied from one brain to another. If you want to become an expert at a particular skill, you'll probably have to do it the old fashioned way: practice, practice, practice.

Practice

In The Karate Kid, a classic movie from the 1980s, a teenage boy asks a karate master to give him a crash course in martial arts. The master reluctantly agrees and begins by making the student wax his collection of cars, sand his woodfloored yard, and paint his large fence. When setting each task, the master demonstrates the exact movements he wants the student to use. The student does as he is told, and later discovers that the movements he has been laboriously repeating are the karate movements he needs to know to defend himself. Because he has repeated these movements hundreds of times while doing his chores, he is able to reproduce them rapidly and effortlessly. He has learned the skills of karate without even knowing it!

Hollywood's portrayal of the relationship between practice and skill memories in this movie is similar to several early psychological theories. The basic idea is that the more times you perform a skill, the faster or better you'll be able to perform it in the future. Is this how practice works? Or is there more to practice than just repetition? To address this issue, Edward Thorndike conducted experiments in which he repeatedly asked blindfolded individuals to draw a line exactly 3 inches long (Thorndike, 1927). Half of the participants were told when their line was within one-eighth of an inch of the target length, and the other half were not given any feedback about their lines. Both groups drew the same number of lines during the experiment, but only the participants who received feedback improved in accuracy as the experiment progressed. This simple study suggests that waxing cars and sanding floors may not be the most effective way to learn karate moves. Feedback about performance, what researchers in the field usually call knowledge of results, is critical to the effectiveness of practice (Butki & Hoffman, 2003; Ferrari, 1999; Liu & Wrisberg, 1997; A. P. Turner & Martinek, 1999; Weeks & Kordus, 1998).

Acquiring Skills

The earliest detailed studies of how practice affects performance were conducted by military researchers who were interested in the high speed, high precision performance of perceptual motor skills such as tracking and reacting to targets (these studies are reviewed by Holding, 1981). One of the basic findings from this early research was that, with extended practice, the amount of time required to perform a skill decreases at a diminishing rate. For example, Figure 4.4a shows that as participants practiced a reading task, the amount of time spent reading each page decreased (A. Newell & Rosenbaum, 1981). Initially, there was a large decrease in the time required to read a page, but after this initial improvement, the decreases in reading time gradually got smaller. Figure 4.3 b shows a similar pattern in individuals learning the rotary pursuit task—the initial gain in performance is the largest. This "law of diminishing returns," also known as the power law of learning, holds for a wide range of cognitive and perceptual motor skills, both in humans and in other species.

When you first learned to use a computer keyboard, you had to search for keys, and the number of words you could type per minute was probably low. After your first year of using a keyboard, you probably had doubled or tripled the number of words you could type per minute. If your typing speed doubled after every year of practice, you would be typing incredibly fast by now! The power law of learning, however, predicts that this does not happen. According to the power law, each additional year of practice after the first produces smaller increases in typing speed; learning occurs quickly at first, but then gets slower.

It may seem obvious that as you become more proficient at a skill, there is less room for improvement. What is surprising about the power law of learning is that the rate at which practice loses its ability to improve performance is usually predetermined, regardless of the skill being practiced or the type of animal learning the skill. In many cases, psychologists can use a simple mathematical function (called a power function) to describe how rapidly individuals will acquire a skill; the number of additional practice trials necessary to improve a skill increases dramatically as the number of completed practice trials increases.

The power law of learning provides a useful description of how practice generally affects performance. It is possible to overcome this law, however, and enhance the effects of practice. For example, in one experiment, researchers asked a participant to kick a target as rapidly as possible. With feedback about his kicking speed, the rate at which he was able to decrease the time required to kick the target was predicted by the power law of learning (Hatze, 1976). When the man stopped improving, the researchers showed him a film comparing his movements with movements known to minimize kicking time. After seeing the film, the man improved his kicking time considerably (Figure 4.4b). This is an example of observational learning, a topic we discuss in detail in Chapter 11. The participant observing the film forms memories of the observed performance techniques that he later uses to improve his own performance. These memories act as a powerful form of feedback about how successfully he is performing the learned skill relative to what is physically possible.

All feedback is not equally helpful, and the kinds of feedback provided can strongly determine how practice affects performance. The secret to improvement is to discover what kinds of feedback will maximize the benefits of practicing a particular skill. Experiments show that frequent feedback in simple perceptual motor tasks leads to good performance in the short term but mediocre performance in the long term, whereas infrequent feedback leads to mediocre performance in the short term but better performance in the long term (Schmidt & Wulf, 1997; Schmidt, Young, Swinnen, & Shapiro, 1989). For the most part, however, instructors, coaches, and their students discover through trial and error what types of feedback work best in each situation. For example, dance instructors have discovered that the visual feedback provided by mirrors enhances the effects of practicing dance movements, and most dance studios now have mirrors on the walls. Can you think of any similar advances that college professors have made in the last century in providing feedback to improve students' cognitive skills? An example might be online tutorials that provide immediate feedback; some research suggests that these can produce faster learning and greater achievement levels than classroom instruction (J. R. Anderson, Corbett, Koedinger, & Pelletier, 1995).

Feedback is critical to the acquisition of skill memories because it affects how individuals perform the skills during practice. Certain forms of information that precede practice, such as instructional videos, can have similar effects. Skill memories do not depend only on the way skills are practiced, however. They also depend on how effort is apportioned during practice. Concentrated, continuous practice, or massed practice, generally produces better performance in the short term, but spaced practice, spread out over several sessions, leads to better retention in the long run.

Consider the following classic experiment. Four groups of post office workers were trained to use a keyboard to control a letter sorting machine. One group trained for 1 hour a day, once a day, for 3 months. The other three groups trained either 2 or 4 hours a day for 1 month (Baddeley & Fongman, 1978). Contrary to what you might guess, the group that trained for only 1 hour a day (spaced practice) required fewer total hours of training than any other group to become proficient at using the keyboard (Figure 4.5). The downside was that this group had to be trained over a longer period—3 months instead of 1. Although researchers have conducted many studies to determine what kind of practice schedule leads to better learning and performance, there is still no consensus about an optimal schedule for any given individual attempting to learn any given skill.

Researchers observe similar kinds of effects when participants practice with a very limited set of materials and skills, called constant practice, versus a more varied set, called variable practice. Constant practice consists of repeatedly practicing the same skill—for example, repeatedly attempting to throw a dart at the bull's-eye of a dartboard under fixed lighting conditions, or attempting to master a single trick shot in pool. Variable practice consists of practicing a skill in a wider variety of conditions, such as attempting to hit each number sequentially on a dartboard under various levels of lighting, or trying to improve one's performance at interviews by applying for a diverse range of jobs. Several studies have shown that variable practice leads to better performance in later tests. In one such study, individuals tracked targets that were moving along various paths. People who used variable practice to learn this task performed better, both in training sessions and in later tests, than individuals who trained with constant practice (Wulf & Schmidt, 1997). Variable practice is not always more effective than constant practice, however (van Rossum, 1990); researchers have not discovered how to reliably predict when variable practice will lead to better learning and performance. Researchers and coaches alike continue to vigorously debate which schedules and which types of practice are most effective.

Implicit Learning

Typically, when you acquire a skill, it is because you have made an effort to learn the skill over time. In some cases, however, individuals can learn to perform certain skills without ever being aware that learning has occurred. Learning of this sort, called implicit learning, probably happens to you more often than you think. Given that, by definition, implicit learning is learning that you are not aware of, you'd be hard pressed to estimate how many skills you've acquired in this way. For all you know, you're implicitly learning right now!

Implicit skill learning comes in at least two forms (Knowlton et ah, 1996; Pohl, McDowd, Filion, Richards, & Stiers, 2001; Willingham, 1999; Wulf & Schmidt, 1997). In one type, individuals perform some task, such as washing windows, and incidentally learn an underlying skill that facilitates their performance: maybe they learn that circular rubbing movements shine the window brighter and faster than random rubbing. The learners may or may not realize that they have discovered a faster, better manner of execution.

A task that psychologists commonly use to study implicit skill learning of this kind is the serial reaction time task, in which participants learn to press one of four keys as soon as a computer indicates which key to press. The computer presents the instructional cues in long sequences that are unpredictably ordered (the so-called random condition) or occur in a fixed sequence of about 12 cues (the implicit learning condition). For example, if we designate the four keys from right to left as A through D, then the fixed sequence might be ABADBCDACBDC. Participants eventually begin to get a feel for the repeating patterns and anticipate which key to press next, as reflected by faster reaction times for implicitly learned sequences relative to random sequences (Figure 4.6). When researchers interviewed participants after training, however, the participants typically showed no awareness that any of the sequences were repeating patterns (Exner, Koschack, & Irle, 2002).

The second form of implicit learning is seen in individuals with amnesia. We described in Chapter 3 the problems that individuals with anterograde amnesia have with learning and remembering events and facts. However, such individuals can acquire skills relatively normally from one session to the next, even if they show no awareness that they have practiced the skill in the past or have ever seen the task before (Cohen, Poldrack, & Eichenbaum, 1997; Seger, 1994; Sun, Slusarz, & Terry, 2005). The individuals make an effort to learn the skill during each session, but always think they are trying it for the first time. The fact that their performance improves with each session demonstrates that they are forming skill memories even though they can't verbally describe their prior practice sessions. H.M., the patient with amnesia whom we introduced in Chapter 3, was able to learn new perceptual-motor skills, but he did not know that he had learned them (Corkin, 2002; Gabrieli, Corkin, Mickel, & Growdon, 1993; Tranel, Damasio, Damasio, & Brandt, 1994).

The ability of people with amnesia to learn complex skills without being aware that they have learned them suggests that the neural systems underlying memories for skills are different from the systems involved in storing and recalling memories for events and facts. This form of implicit learning in amnesia differs from implicit learning in people with no memory impairment in that the individual with amnesia may be explicitly taught how to perform the skill that is being "implicitly" learned. Nevertheless, because the individual shows no evidence of recalling these training sessions, the learning is typically considered implicit. For example, if you learn to sing a song by practicing it, psychologists would not say you are implicitly learning the song (because you are explicitly encoding and recalling the song and are aware that your ability to sing the song is improving over time). However, neuropsychologists would describe this type of learning in patients with amnesia as implicit learning, because the patients are not aware that they are improving at singing a particular song, nor do they remember learning the song. The fact that individuals with amnesia can learn skills implicitly but cannot recall recent events has often been cited as evidence that skill memories are fundamentally different from memories for facts and events.

As mentioned earlier, people often have difficulty verbalizing what they have learned after acquiring a perceptual-motor skill, which seems to suggest that perceptual-motor skills are more likely than cognitive skills to be learned implicitly. But people can also acquire many features of cognitive skills by implicit learning. No one becomes a chess master simply by reading the rules of chess and listening to other players explaining why they made particular moves. Mathematical whizzes do not become experts by simply hearing about mathematical axioms (Lewis, 1981). Development of both of these skills requires practice during which certain improvements are learned implicitly. In the case of acquiring cognitive skills, it is difficult to assess which abilities are improving independent of awareness, because the changes in thinking produced by practice are not easy to observe. (See "Unsolved Mysteries" on p. 138 on the ability or inability to verbalize learned skills.) Moreover, the learner is often unaware of these changes and therefore cannot report them. Consequently, there is currently no way to assess whether implicit learning is more likely to occur during the learning of perceptual-motor skills than the learning of cognitive skills.

Retention and Forgetting

Like memories for facts and events, the memorability of a skill—how well the skill is performed on a later occasion—depends on the complexity of the skill, how well the skill memory was encoded in the first place, how often it has subsequently been recalled, and the conditions in which recall is attempted (Arthur, Bennett, Stanush, & McNelly, 1998). The common wisdom that once you learn to ride a bicycle, you never forget how to do so, is not accurate. Although skill memories can last a lifetime, they do deteriorate with non-use. Generally, retention of perceptual-motor skills is better than retention of cognitive skills, but unless you actively maintain your bike-riding skills, the skill memories you created when you first learned to ride will gradually deteriorate.

Researchers have studied the forgetting of events and facts much more than they have studied the forgetting of skills. Perhaps this is because if someone loses the ability to do something, it is hard to judge whether he has forgotten how to do it, or forgotten that he knows how to do it, or lost the physical control or mechanisms necessary to perform what he recalls. Loss of motor control does not imply that a skill memory is forgotten. To the outside observer, however, it may be impossible to distinguish whether someone knows how to perform a skill but has impaired movement abilities or has never learned to perform the skill. In fact, the only way to distinguish between these two possibilities is by observing differences in neural activity during the performance or nonperformance of a skill.

Psychologists call loss of a skill through non-use skill decay. Most of the data collected so far indicate that skill decay follows patterns similar to those seen in the forgetting of memories for events and facts. Motor deficits and injuries can clearly affect skill decay, because they are likely to lead to non-use of learned skills.

In some ways, forgetting a skill is like learning it in reverse. Not performing the skill is almost the opposite of practice: if you don't use it, you lose it. Most forgetting occurs soon after the last performance of the skill; as time goes by, less and less forgetting occurs. Thus, forgetting curves are similar to learning curves. Forgetting occurs quickly at first, then gets slower.

Does the passage of time simply cause a skill to be "unlearned"? It often may seem this way, but forgetting can also result when new memories interfere with the recollection of old memories. As time passes, you perform more new skills, creating more memories that potentially interfere with the recollection of earlier skill memories. (Recall from Chapter 3 that interference and decay are also involved in the forgetting of memories for events and facts.) Much of this interference can occur without any awareness on the part of the person attempting to recall a skill. For example, you might have difficulty recalling some of the dances you learned when you were younger, but easily recall dance steps you learned recently. Rather than thinking this recent learning is hampering your ability to perform the old dances, you'd probably assume that you can't remember an older dance simply because it has been so long since you last did it. However, there is no subjective way for you to distinguish whether your forgetting results from the passage of time or from interference.

Recently, researchers observed that interference of skill memories can occur even within a single day. Students trained to perform a finger-tapping task, similar to the serial reaction time task discussed above, demonstrated more rapid and accurate pressing times after a period of sleep (Walker, Brakefield, Hobson, & Stickgold, 2003; Walker, Brakefield, Morgan, Hobson, & Stickgold, 2002; Walker, Brakefield, Seidman, et al., 2003). However, if students learned to press keys in two different sequences on the same day, sleep-dependent enhancement of their performance was seen only for the second sequence learned. If participants learned the second sequence one day after the first sequence, sleep enhanced the performance of both sequences. Interestingly, if on the second day the students reviewed the first day's sequence immediately before learning the new sequence, then on the third day sleep enhanced their accuracy on only the second sequence. Thus, not only can practicing two skills on the same day interfere with retention of memories for the first skill, but reviewing a recently learned skill before beginning to practice a new one can interfere with subsequent recall of the skill that was reviewed! These findings highlight the intimate relationship between skill acquisition and skill recall, and the fragile nature of newly acquired skill memories. Note, however, that athletes and musicians commonly practice multiple skills in parallel with no evidence of interference, and variable practice generally leads to better long-term performance than constant practice. Thus, skills more complex than learning a sequence may be less susceptible to interference effects.

Research has also shown that a major determinant of whether a person will recall a particular skill is the similarity between the retrieval conditions and the conditions she experienced while learning the skill. In many situations, of course, the conditions under which a skill must be recalled are not the same as the training conditions. In this case, trained performance must "transfer" to the novel conditions.

Transfer of Training

Skills are often highly constrained in terms of how they can be applied (Goodwin, Eckerson, & Voll, 2001; Goodwin & Meeuwsen, 1995; Ma, Trombly, & Robinson-Podolski, 1999). You may have mastered the culinary skills needed to make great Italian food, but this will not make you a great sushi chef. In some cases, skill memories are so specific that the introduction of additional informative cues can disrupt performance. For example, after individuals were trained to touch a target with a stylus without visual feedback about their arm movements, their performance was worse when researchers allowed them to see their arm moving as they carried out the task (Proteau, Marteniuk, & Levesque, 1992). Most people normally use visual feedback when learning to aim at a target, so it is surprising that providing such information can interfere with the recall of skill memories.

In other cases, skills seem to be easily transferable to novel situations. For example, you learned to write with your right or left hand, and you may even have practiced with each hand, but have you ever written with your mouth or feet? If you try, you'll discover that you can write semi-legible text using these and other body parts. You are able to transfer what you have learned about writing with one hand to other body parts, despite large differences in the specific movements you must perform to do so. In sports, teams spend much of their time practicing in scrimmages, with the hope that these experiences will transfer positively to similar situations in real games. If skills learned in scrimmage did not transfer to real games, it is unlikely that so many coaches in so many different sports would train their teams in this way.

The restricted applicability of some learned skills to specific situations is known as transfer specificity. This phenomenon led Thorndike to propose that the transfer of learned abilities to novel situations depends on the number of elements in the new situation that are identical to those in the situation in which the skills were encoded (Thorndike & Woodworth, 1901). Thorndike's proposal, called the identical elements theory, provides one possible account of why transfer specificity occurs. It predicts that a tennis player who trained on hard courts might suffer a bit if a game were moved to clay courts, and would do progressively worse as the game was changed from tennis to badminton or table tennis. Conceptually, transfer specificity is closely related to transfer-appropriate processing, described in Chapter 3. The main differences between the two stem from whether the memories being recalled are memories of skills or memories of facts.

When you apply existing skill memories to the performance of novel skills, you are generalizing based on past experience. Generalization of learning is a topic that psychologists have studied extensively (you will learn more about this in Chapter 9). Nevertheless, we do not yet know how one skill generalizes to another, or what factors limit how well a learned ability can be generalized. Even if Thorndike's identical elements theory is on the right track, it doesn't tell us what the "elements" of skill memories are or how to assess the similarities and differences between those elements.

When you perform a skill that you have learned in the past, you are generalizing from a past experience to the present. From this perspective, every performance of a skill involves transfer of training. For example, each time you open a door you are making use of memories you acquired by opening doors in the past. Practice improves performance and recall, and thus increases the stability and reliability of skill memories over time. How might elements of skill memories be made stable? Current theoretical models of skill acquisition suggest that an individual stabilizes skill memories by converting them from memories for events and facts into memories for predefined sequences of actions called motor programs (J. R. Anderson, 1982), as discussed below.

Models of Skill Memory

In the previous chapter, we described how psychologists have modeled memories for facts using semantic networks, with facts represented as nodes within the network, and connections between different facts represented as links between nodes. This type of model is useful for describing how facts are organized in memory. Scientists studying skill memories have developed similar models, but most models of skill memory focus on how individuals learn skills over time rather than how they organize what they have learned.

Motor Programs and Rules

When you practice a skill, you probably do so because you want to become better at performing that skill. To most people, "becoming better" means that their performance becomes more controlled and effortless. Say the skill you are practicing is juggling. The goal is to keep the objects moving in the air, and in and out of your hands. Ideally, you'd probably like to be able to juggle while casually talking to a friend. In this case, your friend would know you are an expert juggler because you don't need to pay attention to what you are doing. The skill has become automatic. Some might even say that your juggling actions have become reflexive. Reflexes, however, are inborn, involuntary responses to stimuli, distinct from highly learned responses. Sequences of movements that an organism can perform automatically (with minimal attention) are called motor programs. Unlike reflexes, motor programs can be either inborn or learned. Releasing an arrow from a bow is not an inborn reflex, but for the expert archer it has become as automatic and precise as a reflex. More complex action sequences such as juggling can also become motor programs. One way to determine whether a skill has become a motor program is to remove the stimulus during the action sequence and observe the results. For example, if someone grabs one of the balls in midair as you are juggling, does your arm still "catch and throw" the nonexistent ball? If so, it suggests that your juggling skill has become a motor program.

Classifying highly learned perceptual motor skills as motor programs is straightforward, but what about highly learned cognitive skills? Might they also, with extended practice, become motor programs? The surprising answer is yes. Think back to when you learned the multiplication tables. This probably required some practice, but now if someone asks you, "What is two times three?" you will respond promptly: "Six." You no longer need to think about quantities at all. You perceive the spoken words, and your brain automatically generates the motor sequence to produce the appropriate spoken word in response. Similarly, in the laboratory, once a person has solved the Tower of Hanoi problem many times, she has learned that particular movement sequences always lead to the solution. Eventually, practicing enables her to perform these motor sequences rapidly, without thinking about which disk goes where. In both cases, a cognitive skill has become a motor program.

The learning of new skills often begins with a set of instructions. You give your great aunt a new microwave oven, and later discover that she refuses to set the time on the display. Why? Because she doesn't know how to do it. The manufacturer of the oven predicted that your great aunt might not possess this skill, and so it provided written rules—a list of steps to take to display the correct time. In a perfect world, your great aunt would read the manual and acquire the skills necessary to set the time on the microwave oven. More likely, though, the rules are ambiguous and open to interpretation, making her first attempts at setting the time awkward and possibly unsuccessful; but, with your encouragement, she finally does manage it. However, at a later date, when she wants to reset the time, she may recollect that the manual was little help in providing rules she could understand, and she may prefer trying to recall the steps from memory. Because she will depend on her memories for events and facts to perform the skill, you could say that her skill memories are her memories of the rules. In other words, skill memories can be memories for events and facts!

Following a recipe in a cookbook provides another example of how memories for facts can serve as skill memories. A recipe teaches you the facts you need to know to prepare a certain dish: what ingredients you need, in what proportions, and how to combine them. However, after some practice—with baking cookies, for example—you no longer need to depend as heavily on the written "rules."

How can skill memories lead to reflex-like automatic movements, but also consist of remembered events and facts? A classic model proposed by Paul Fitts in 1964 suggests that this is possible because practice transforms rules into motor programs.

Stages of Acquisition

Fitts proposed that skill learning includes an initial period when an individual must exert some effort to encode a skill, acquiring information through observation, instruction, trial and error, or some combination of these methods (Fitts, 1964). This period is followed by stages in which performance of the skill becomes more "automatic" or habitual. Fitts called the first stage the cognitive stage, to emphasize the active thinking required to encode the skill. When your great aunt is setting the time on the microwave based on instructions or memories of the steps that were previously successful, she is in the cognitive stage of skill acquisition. During this stage, she bases her performance on what she knows, as well as on her ability to control her movements and thoughts so as to accomplish specific goals. Humans are likely to depend on memories of verbalizable facts or rules at this stage, but this is not what happens in other animals. A monkey can learn to change the time on a microwave oven, if motivated to do so, but the goals and strategies the monkey employs to learn this skill are very different from the goals and strategies available to your great aunt. Researchers do not yet know the degree to which memories for facts or events are important for skill acquisition in nonhumans or preverbal children as compared with adult humans.

Fitts called the second stage in his model of skill acquisition the associative stage. During this stage, learners begin using stereotyped actions when performing the skill and rely less on actively recalled memories of rules. The first few times you play a video game, for example, you may need to keep reminding yourself about the combinations of joystick movements and button presses necessary to produce certain outcomes. Eventually, you no longer need to think about these combinations. When you decide that you want a particular action to occur on the screen, your hands do what is necessary to make it happen. Flow do they do this? Your brain has encoded specific combinations and is recalling them as directed. What began as a process of understanding and following verbalizable rules has become a process of remembering previously performed actions.

Of course, mastering the skills needed to play a video game requires far more than simply memorizing hand movements. You must be able to produce very rapid sequences of precisely timed combinations to achieve specific outcomes. To reach high levels of performance, your movement patterns must become rapid and effortless. In Fitts's model, this level of skill is represented by the third stage, the autonomous stage—the stage at which the skill or subcomponents of the skill have become motor programs. At this stage it may be impossible to verbalize in any detail the specific movements being performed, and performance may have become much less dependent on verbalizable memories for events and facts. If you can juggle while having a casual conversation, you have reached the autonomous stage. You can perform the skill without paying much attention to what you're doing, and if someone unexpectedly snatches a ball, your arms will continue to move as if the missing ball were still there.

In the autonomous stage, the actions of a monkey trained to set the time on a microwave oven might be almost identical to the actions performed by your great aunt when she is setting the time. The monkey and your great aunt may have learned this skill through different strategies, but their end performance is very similar. Is your great aunt's motor program substantially different from the monkey's? The observable skills look the same, but the memories underlying the skills may be very different. Comparing neural activity in your great aunt and in the monkey would be one way to determine whether they are accessing information similarly while performing the same learned skill.

Fitts's model of skill acquisition (summarized in Table 4.2) provides a useful framework for relating skill performance to practice. Although psychologists have developed this model extensively over the past 40 years, many recent versions retain the same basic progression of stages. The "three stages" are, of course, abstractions. There is generally no single performance that can be identified as the last performance belonging to, say, stage one. Additionally, like semantic network models, the three-stage model of skill learning is primarily descriptive. It won't help you predict how much practice you need to convert your skill memories to motor programs or give you pointers about how and when you should practice. The model does suggest, however, that learned abilities may rely on different kinds of memory as practice progresses. Different kinds of memory may in turn require different kinds of neural processing, or activation of different brain regions. By examining the neural activity associated with skill acquisition, scientists have explored the idea that skill memories take different forms as learning progresses.

Table 4.2

Fitts's Three-Stage Model of Skill Learning

Interim Summary

When you learn a skill, you form memories that allow you to take advantage of your past experiences. Two major kinds of skills are perceptual-motor skills and cognitive skills. You are born with certain talents and can use and enhance them by developing appropriate skills. With extensive practice you may even become an expert. It is unlikely that you will become an expert at every skill you practice, but you can retain many skills in memory for extended periods of time. How long after learning skills you can retrieve the skill memories depends on how well you learned the skills, how often you've recalled them, and how complex the skills are.

4.2 Brain Substrates

What neural systems do humans and other animals need in order to acquire memories of perceptual-motor and cognitive skills? Is there something special about the human brain that allows us to acquire skill memories more effectively than other animals? Or do humans use the same brain systems as other animals to learn skills, but use them in slightly different ways? How might one judge whether the skill memories that underlie a dolphin's ability to use a sponge differ from those of a window washer?

Neuroscientists have used neuroimaging and neurophysiological recording techniques to identify brain systems involved in the formation and recall of skill memories. These techniques allow researchers to monitor brain activity in humans and other animals during the performance of skills. Researchers have also compared brain activity in experts and amateurs, as well as in individuals before and after they have learned a particular skill. Neuropsychological studies of skill learning by patients with brain damage are also an important source of information. Through these kinds of research, neuroscientists hope to associate stages of skill acquisition with changes in brain activity.

All movements and postures require coordinated muscle activity. As you saw in Chapter 2, a major function of the nervous system is to initiate and control muscle activity. The spinal cord and brainstem play a critical role in skill performance by controlling and coordinating movements. Brain regions dedicated to sensation and perception, including the sensory cortices, are also involved, processing information that contributes to skill learning. Remember the experiment described earlier in this chapter in which researchers instructed the participant to kick a target as quickly as possible? He improved at the task by processing visual feedback about how effectively he was coordinating the muscles in his leg.

The importance of the peripheral nervous system and spinal cord to humans' performance of perceptual-motor skills is illustrated by the total paralysis that results when the spinal cord becomes disconnected from the brain. A well-known case is that of the actor Christopher Reeve, who suffered a spinal cord injury after falling from a horse. This injury caused him to lose the ability to feel and move his limbs, as well as the ability to breathe on his own.

Christopher Reeve also serves as an example of the possible effects of practice on the nervous system. Several years after his accident, Reeve regained some sensation in parts of his body, as well as the ability to move his wrist and one of his fingers. Some researchers believe that new rehabilitation techniques, in which a person's muscles are electrically stimulated to generate movements simulating bicycle pedaling, caused this recovery of function (J. W. McDonald et al., 2002). In other words, practicing movements may help the brain and spinal cord replace or repair lost or damaged connections.

In this section we describe how practicing skills can change neural circuits in less extreme circumstances. Although you can form skill memories in ways other than practice (such as studying videos of expert athletes or expert kissers), neuroscientists have focused much of their effort on understanding the incremental effects of practice on brain activity during skill learning.

Sensory processing and motor control by circuits in the spinal cord are clearly necessary for learning and performing skills. However, the core elements of skill learning seem to depend in particular on three other areas of the brain: the basal ganglia, the cerebral cortex, and the cerebellum (Figure 4.7).

The Basal Ganglia and Skill Learning

"Basal ganglia" is one of the few terms for a brain structure that literally describe the region (or in this case regions) to which they refer. The basal ganglia are ganglia (clusters of neurons) located at the base of the forebrain (the most prominent part of the human brain). As you'll recall from Chapter 2, the basal ganglia are positioned close to the hippocampus. Like the hippocampus, the basal ganglia receive large numbers of inputs from cortical neurons. In fact, most cortical areas send inputs to the basal ganglia. These inputs provide the basal ganglia with information about what is happening in the world—in particular, about the sensory stimuli the person is experiencing. Unlike the hippocampus, the basal ganglia send output signals mainly to the thalamus (affecting interactions between neurons in the thalamus and motor cortex) and to the brainstem (influencing signals sent to the spinal cord). By modulating these motor control circuits, the basal ganglia play a role in initiating and maintaining movement.

The basal ganglia are particularly important for controlling the velocity, direction, and amplitude of movements, as well as for preparing to move (Desmurget, Grafton, Vindras, Grea, & Turner, 2003; Graybiel, 1995; R. S. Turner, Grafton, Votaw, Delong, & Hoffman, 1998). For example, suppose you are performing the rotary pursuit task. You need to move your arm in a circle at a velocity that matches that of the rotating target. In this task, your basal ganglia will use information from your visual system about the movements of the target, the stylus, and your arm, as well as information from your somatosensory system about the position of your arm, to control the direction and velocity of your arm movements. Similarly, if you dive into a pool to retrieve a coin, your basal ganglia will help you avoid colliding with the bottom of the pool.

Given all the interconnections between the basal ganglia and motor systems, it's not surprising that disruption of activity in the basal ganglia impairs skill learning. Such disruption does not, however, seem to affect the formation and recall of memories for events and facts. Consider the case of Muhammad Ali. Ali was one of the most agile and skilled boxers of his era, but his career was ended by a gradual loss of motor control and coordination. Doctors identified these deficits as resulting from Parkinson's disease, a disorder that disables basal ganglia circuits (we discuss this disease in more detail later in the chapter). Over time, the loss of basal ganglia function resulting from Parkinson's disease affects even the most basic of skills, such as walking. Whereas H.M.'s hippocampal damage (described in Chapter 3) prevents him from reporting on his past experiences, Muhammad Ali's basal ganglia dysfunction prevents him from making use of skill memories and learning new skills; it has not affected his memory for facts or events.

Many researchers suspect that processing in the basal ganglia is a key step in forming skill memories, although the specific processes whereby sensory inputs lead to motor outputs are currently unknown (Barnes, Kubota, Hu, Jin, & Graybiel, 2005; Graybiel, 2005). Most researchers agree, however, that practicing a skill can change how basal ganglia circuits participate in the performance of that skill, and that synaptic plasticity is a basic neural mechanism enabling such changes (Conn, Battaglia, Marino, & Nicoletti, 2005; Graybiel, 2004). We describe here the experimental results that show the importance of the basal ganglia not only for performing skills but also for forming and accessing skill memories.

Learning Deficits after Lesions

Much of what is known about the role of basal ganglia in skill learning comes from studies of rats learning to navigate mazes, such as the radial maze shown in Figure 4.8a. In the standard radial maze task, rats learn to search the arms in the maze for food, without repeating visits to the arms they have already searched. This task simulates some features of natural foraging, because food does not magically reappear at locations where a rat has just eaten. However, the entrances to the arms of the maze are all very similar, so unless the rat remembers specifically which arms it has visited, it is likely to go to the same arm more than once. In early sessions, this is just what rats do. They often go to the same arm multiple times, and consequently waste a lot of time running back and forth along arms that contain no food. With practice, the rats learn that they can get more food for their effort by keeping track of where they have been, and they make fewer repeat visits to the same arm. Food acts as a kind of feedback in the radial maze task, in that correct performance leads to food. (This is a particularly important class of feedback that is of great interest to learning researchers, as you'll learn in Chapter 8.)

To learn to navigate the radial maze efficiently, rats must remember certain aspects of past events. Not surprisingly, rats with hippocampal damage have major problems with this task (Figure 4.8b). Even after many sessions, they continue to visit arms they have visited before. In contrast, rats with basal ganglia damage learn this task as easily as rats with no brain damage. This shows that basal ganglia damage does not disrupt rats' memories for events, nor does it prevent them from performing the skills necessary to find food in a radial maze.

Researchers can modify the radial maze task slightly, to make it less dependent on memories of past events. If instead of putting food in all the arms, the experimenter places food only in arms that are illuminated, rats quickly learn to avoid the non-illuminated arms (Figure 4.8c). Rats with hippocampal damage can also learn this version of the task, because they only need to associate light with food, which does not require keeping track of arms they've visited. Surprisingly, rats with basal ganglia damage have difficulty learning this "simpler" version of the task. They continue to search non-illuminated arms even though they never find food in those arms (Packard, Hirsh, & White, 1989). Basal ganglia damage seems to prevent rats from learning the simple perceptual-motor skill of avoiding dark arms and entering illuminated arms.

Rats may show similar learning deficits in another task: the Morris water maze. In the standard version of this maze, experimenters fill a circular tank with murky water. They then place rats in the tank, and the rats must swim around until they discover a platform hidden just beneath the water surface. Once a rat finds the platform, it no longer has to swim, and the trial is over. Researchers measure the time it takes a rat to find the platform, and use this as a measure of learning. Intact rats gradually learn the location of the hidden platform after repeated trials in the tank. Rats with hippocampal damage have severe difficulties learning this standard task, but have no problem learning the task if the platform is visible above the surface of the water. Rats with basal ganglia damage can learn the location of the platform whether it is visible or not. This seems to suggest that basal ganglia damage does not affect a rat's ability to learn this task.

Tests of transfer of training, however, tell a different story. If experimenters move a visible platform in the Morris water maze to a new location during testing, rats with hippocampal damage (or no damage) swim directly to the platform to escape the water. Rats with basal ganglia damage, however, swim to where the platform used to be, and only afterward do they find the platform in its new location (R. J. McDonald & White, 1994). One interpretation of this finding is that rats with basal ganglia damage have difficulty learning to swim toward a platform to escape the water (even when the platform is clearly visible), and instead learn to swim to a particular location in the tank to escape. This study illustrates how two animals may seem to be performing a skill in the same way, but their skill memories and their ability to use them in novel situations are not necessarily equivalent. Your great aunt and a trained monkey may be using very different motor programs to set the time on a microwave oven, even though their actions might look the same.

The findings from these experiments with rats illustrate the effects of damage to the basal ganglia on the formation of skill memories. Such studies have led researchers to conclude that the basal ganglia are particularly important in perceptualmotor learning that involves generating motor responses based on environmental cues. The basic assumption behind such research is that there is nothing unique about the way in which the basal ganglia function in rats learning to navigate mazes, and consequently basal ganglia damage should disrupt skill learning in similar ways in humans.

Neural Activity during Perceptual-Motor Skill Learning

Measures of neural activity in the basal ganglia during learning provide further clues about the role of the basal ganglia in the formation of skill memories. Experimenters can train rats to turn right or left in a T-shaped maze, by using a sound cue that the rats hear just before reaching the intersection where they must turn (Figure 4.9). For example, an experimenter releases a rat in the maze, and then a computer plays a specific sound, instructing the rat to make a right turn. If the rat turns to the right, the experimenter gives the rat food (as noted earlier, this is a particularly effective form of feedback). With practice, rats learn to perform this simple perceptual-motor skill accurately. In a recent experiment, researchers implanted electrodes in the basal ganglia of rats before training them in the T-maze. They then recorded how neurons in the basal ganglia fired as rats learned the task (Jog, Kubota, Connolly, Hillegaart, & Graybiel, 1999).

These recordings revealed four basic patterns of neural activity when rats were in the T maze: (1) some neurons fired most at the start of a trial, when the rat was first released into the maze; (2) some fired most when the instructional sound was broadcast; (3) some responded strongly when the rat turned right or left; and (4) some fired at the end of a trial, when the rat received food. During the early stages of learning, about half of the recorded basal ganglia neurons showed one of these four patterns of activity. Most of these neurons fired only when a rat turned right or left in the maze (Figure 4.9a). The remaining neurons fired in ways that were not clearly related to the rats' movements or experiences in the maze. As the rats' performance improved with practice, the percentage of neurons that showed task related activity patterns increased to about 90%, with most neurons firing strongly at the beginning and at the end of the task rather than during turning (Figure 4.9b). These measurements show that neural activity in the basal ganglia changes during the learning of a perceptual motor skill, suggesting that encoding or control of skills by the basal ganglia changes as learning progresses.

The increased neural activity seen in the beginning and end states during the maze task suggests that the basal ganglia develop a motor plan that the rat's brain initiates at the beginning of each trial. The motor plan then directs the rat's movements until the trial ends (Graybiel, 1997, 1998). This hypothetical process is consistent with Fitts's model of skill learning, in which automatically engaged motor programs gradually replace active control of movements (Fitts, 1964). Someone learning to juggle might show similar changes in basal ganglia activity - that is, if we could record signals from her neurons, which currendy is not possible. In a novice juggler, basal ganglia neurons might fire most strongly when the balls are in the air (when an action must be chosen based on visual information). In an expert juggler, basal ganglia neurons might fire most strongly when she is catching and tossing the balls.

Earlier in the chapter we raised the question of whether cognitive skills might involve some of the same brain regions and neural mechanisms as perceptualmotor skills. The data presented above show that the basal ganglia do indeed contribute to learning of perceptual-motor skills. Do the basal ganglia also contribute to cognitive skill learning?

Brain Activity during Cognitive Skill Learning

Neuroimaging studies of the human brain reveal that the basal ganglia are active when participants learn cognitive skills (Poldrack, Prabhakaran, Seger, & Gabrieli, 1999). In these experiments, participants learned to perform a classification task in which a computer presented them with sets of cards and then instructed them to guess what the weather would be, based on the patterns displayed on the cards (Figure 4.10a). Each card showed a unique pattern of colored shapes. Some patterns appeared when rain was likely, and others appeared when the weather was likely to be sunny. As each card was presented onscreen, participants predicted either good or bad (sunny or rainy) weather by pressing one of two keys. The computer determined the actual weather outcome based on the patterns on the cards. Participants had to learn through trial and error which patterns predicted which kind of weather. The task mimics real-world weather prediction, in that no combination of "patterns" (that is, of cloud cover, temperature, wind, and so on) is 100% predictive of the weather that will follow; meteorologists must develop a wide range of cognitive skills to accurately forecast the weather. For participants in the study, the task may have seemed more like reading Tarot cards than learning a cognitive skill, but they usually improved with practice.

Although each card was associated with the likelihood that a particular kind of weather would occur, there was no simple rule that participants could use to make accurate predictions. Instead, to improve at the task, participants gradually had to learn which cards tended to predict certain types of weather. Brain imaging data showed increased activity in the basal ganglia as individuals learned to make these judgments (Figure 4.10b). This and similar imaging studies suggest that the basal ganglia contribute to both cognitive and perceptual-motor skill learning. But how?

Despite considerable evidence that the basal ganglia enable skill learning, their specific function in this learning is still under debate. For example, since the basal ganglia are involved in the control and planning of movements, perhaps damage to the basal ganglia leads to changes in performance that impair learning processes in other brain regions: if you can't control how your arms are moving, you will have difficulty learning how to juggle. In short, changes in skill learning caused by lesions to the basal ganglia, as seen in rats learning the radial maze task, do not definitively prove that this region is critical for encoding or retrieving skill memories. Similarly, learning-dependent changes in the activity of basal ganglia neurons, as seen in rats learning to follow instructions in a Tmaze, could reflect changes in the information coming from the sensory cortex rather than changes generated in the basal ganglia.

Are basal ganglia neurons doing most of whatever is required to form memories of skills, or are other brain regions such as the cortex and cerebellum doing the bulk of the encoding and retrieval? Could it be that the basal ganglia contribute as much to skill-memory formation as do other brain regions, but the basal ganglia are specialized for specific aspects of the learning process? We need to take a closer look at different cortical regions during and after practice sessions to shed some light on these issues. (See "Learning and Memory in Everyday Life" on p. 152 for some insight into how video-game playing develops perceptual-motor skills and cognitive skills alike.)

Cortical Representations of Skills

How important is the cerebral cortex for the learning and performance of skills? Given that most animals don't have a cerebral cortex, and that animals born with a cortex can make many movements after surgical removal of all their cortical neurons, you might conclude that the cortex isn't very important for skill learning. In fact, mammals are the only animals that make extensive use of cortical circuits for any purpose, so whatever the role of the cerebral cortex in skill memory, it probably plays this role most extensively in mammals. Coincidentally (or not), mammals are highly trainable compared with most other species.

Neural circuits in the cerebral cortex that are active when you run, jump, or sing change over time in ways that enhance the activities you perform most often, as well as the activities you find most rewarding. From this perspective, skill memories are the neural outcomes of repeated performances. A simple analogy is the way your body shape changes in response to a bodybuilding regimen. Just as increasing the strength and flexibility of your muscles can affect how well you jump, changes in networks of cortical neurons can also influence your jumping ability.

Cortical Expansion

If cortical networks are like brain "muscles," you'd expect the practice of different skills to affect different regions of cerebral cortex, just as different physical exercises affect different muscle groups. This seems to be true. Regions of the cerebral cortex involved in performing a particular skill expand in area with practice, while regions that are less relevant to the skill show fewer changes.

Neuroimaging techniques such as fMRI reveal this expansion by showing increased blood flow to particular regions. As one example, brain imaging studies of professional violinists showed that representations in the somatosensory cortex of the hand used to control note sequences (by pressing individual strings with different fingers) are larger than in non-violinists (Elbert, Pantev, Wienbruch, Rockstroh, & Taub, 1995). Interestingly, the cortical maps of violinists' bow hands (the fingers of which always move together) showed no such elaborations: the changes are specific to the hand that moves the fingers separately.

Measures of blood flow reveal larger areas of cortical activation after extensive practice, which implies that experience is affecting cortical circuits. These measures do not reveal what physical changes occur, however. Recent studies using structural MRI techniques indicate that practice can change the amount of cortical gray matter (where the cell bodies of neurons are found). For example, after about 3 months of training, people who learned to juggle three balls continuously for at least 1 minute showed a 3% increase in gray matter in areas of the visual cortex that respond to motion (Draganski et al., 2004). No comparable structural changes were observed in the motor cortex, basal ganglia, or cerebellum. It is not known whether expansion of gray matter reflects changes in the number or size of synapses, changes in the number of glia (the cells providing functional and structural support to neurons), or changes in the number of cortical neurons.

Electrophysiological studies also show that practice can expand cortical representations. In one such study, researchers trained monkeys to perform a tactile discrimination task (Recanzone, Merzenich, Jenkins, Grajski, & Dinse, 1992). The task required the monkey to release a handgrip whenever it felt a stimulus on its fingertip that differed from a standard stimulus. During each trial, the monkey initially felt a surface vibrating at a fixed speed on one of its fingers, for about half a second. This initial tactile stimulus, always the same, provided a standard for comparison. The initial stimulus was followed by a half-second interval of no stimulation, and then a series of one to four additional vibrating stimuli, each vibrating either at the same rate as the standard or faster. The monkey was given fruit juice if it released the handgrip when vibrations were faster than the standard. This task is similar to the T-maze task described earlier, in which researchers recorded the activity of basal ganglia neurons in a rat as it learned to turn right or left in response to acoustic cues. Both the T-maze and the tactile discrimination task require the animal to perform one of two responses (in one task, turn right or turn left; in the other, grip or release) based on specific cues provided to a single sensory modality (sound in one task, touch in the other).

When a monkey learned to respond to a vibrating stimulus that predicted the delivery of juice, the area of the somatosensory cortex that processed the cue increased. As a result, monkeys that learned the tactile discrimination task had enlarged cortical representations for the finger they used to inspect tactile stimuli.

Studies such as these show that perceptual-motor skill learning is often associated with the expansion of the regions of the sensory cortex involved in performing the skill. Similarly, practicing a perceptual-motor skill can also cause regions of the motor cortex to expand. For example, electrical stimulation of the motor cortex in monkeys trained to retrieve a small object showed that the area of the cortex that controlled movements of the fingers expanded (Nudo, Milliken, Jenkins, & Merzenich, 1996). In monkeys that learned to turn a key with their forearm, cortical representation of the forearm expanded. Researchers don't know how many different cortical regions are modified during learning of a particular skill, but the current assumption is that any cortical networks that contribute to performance of the skill are likely to be modified as training improves (or degrades) performance. Researchers also have yet to determine exactly how cortical expansion occurs and what it consists of, but most neuroscientists believe that the expansion reflects the strengthening and weakening of connections within the cortex resulting from synaptic plasticity.

Are Skill Memories Stored in the Cortex?

Many experiments have shown that cortical networks are affected by practice, but this tells us only that the two phenomena are correlated, not that changes in the cerebral cortex improve performance. Such studies also do not establish that skill memories are stored in cortical networks. As you saw earlier, changes in neural activity in the basal ganglia also take place during skill learning. The cerebral cortex clearly influences skill learning and performance, but knowing this is not the same as knowing what cortical circuits do during skill learning.

One way to get closer to understanding cortical function is to measure cortical activity during training. Much of what is known about skill learning relates to how different practice regimens lead to differences in the rate of skill improvement and in the rate of forgetting. If it were possible to show that changes in the cortex parallel behavioral changes, or that improvements in performance can be predicted from cortical changes, we could be more certain that skill levels and cortical activity are closely related. Initial investigations in this direction suggest that the behavioral stages of skill acquisition are indeed paralleled by changes in cortical activity.

Data from brain imaging studies show that when people begin learning a motor skill that requires sequential finger movements, the portion of the motor cortex activated during performance of the task increases rapidly during the first training session and more gradually in later sessions. Avi Karni and colleagues required participants to touch each of their fingers to their thumb in a fixed sequence as rapidly and accurately as possible (Karni et al., 1998). In parallel with the changes seen in the motor cortex, participants' performance of the task improved rapidly in early sessions and more gradually in later sessions (Figure 4.11a), consistent with the power law of learning. Imaging data collected over 6 weeks of training suggested that additional practice resulted in additional, more gradual increases in the representation of learned movements in the motor cortex.

Overall, the region of motor cortex activated during performance of the practiced sequence expanded relative to the area activated by different, untrained sequences of identical finger movements (Figure 4.11b). Kami and colleagues hypothesized that the period of "fast learning" involves processes that select and establish the optimal plans for performing a particular task, whereas the subsequent slower stages of learning reflect long-term structural changes of basic motor control circuits in the cortex. Recent data from studies of perceptualmotor skill learning in rats are consistent with this interpretation. Rats trained in a reaching task showed significant differences in their motor map only after practicing the task for at least 10 days (Kleim et al., 2004). This finding suggests that structural changes in the cortex reflect the enhancement of skill memories during later stages of training.

Circuits in the cerebral cortex are activated by many sensory and motor events, so it is not surprising that these brain regions contribute to skill learning. However, until researchers look at interactions between the cerebral cortex and the basal ganglia while individuals are learning a wide variety of perceptualmotor and cognitive skills, assessing the respective roles of the cortex and basal ganglia in forming and recalling skill memories will remain a difficult task.

The Cerebellum and Timing

What about skill learning in animals such as birds and fish that don't have much cortex? Researchers can train pigeons to perform a wide range of perceptualmotor skills, and fish can rapidly learn to navigate mazes. Animals without much cortex must rely on evolutionarily older parts of the brain to learn skills. One region that seems to be particularly important in this process is the cerebellum.

The cerebellum is probably one of the most basic neural systems involved in encoding and retrieving skill memories. Even animals as lowly as fish and frogs, which may seem to have little potential for skill learning, have a cerebellum. Although you aren't likely to see a fish or a frog performing in a circus, this doesn't mean these animals cannot learn perceptual-motor skills; for example, with practice, fish can learn to press little levers for food. You are more likely to have seen parrots riding tricycles or heard them producing intelligible sentences. Birds, too, have a cerebellum, which may facilitate their ability to learn such tricks. In fact, most animals that have a spine also have a cerebellum. Yet there are relatively few studies of cerebellar function in nonmammals. Consequently, much less is known about how the cerebellum contributes to skill-memory formation in animals with little cortex than is known about cerebellar function in mammals that make extensive use of cortex.

Most of the inputs to the cerebellum are from the spinal cord, sensory systems, or cerebral cortex, and most of the output signals from the cerebellum go to the spinal cord or to motor systems in the cerebral cortex. Experiments conducted in the early 1800s showed that cerebellar lesions impair the performance of motor sequences. People with cerebellar damage, for example, have difficulty writing or playing a musical instrument. (Chapter 7 provides further details on how cerebellar damage affects human performance.) Collectively, these anatomical and neuropsychological data indicate that the cerebellum contributes to the performance of perceptual-motor skills in mammals. Because the structure of the cerebellum is organized similarly across different species, it is presumed to serve similar functions in both mammals and nonmammals (Lalonde & Botez, 1990).

Other evidence suggests that, in addition to facilitating the performance of skills, the cerebellum is involved in forming memories for skills. For example, early brain imaging studies of systems involved in motor learning showed that there is a sudden increase in cerebellar activity when humans begin learning to perform sequences of finger movements (Friston, Frith, Passingham, Liddle, & Frackowiak, 1992). Similarly, rats that learn complex motor skills to navigate an obstacle course (for example, balancing on tightropes and seesaws) develop predictable physiological changes in cerebellar neural circuitry, such as increased numbers of synapses (Kleim et al., 1997).

Cerebellar changes in acrobatic rats seem to depend on skill learning rather than on activity levels, because rats that run in an exercise wheel for the same amount of time do not show such changes. More generally, animals such as birds and dolphins that routinely perform three dimensional acrobatic skills—flying between branches; rapidly jumping or diving while spinning—typically have a larger cerebellum than animals that do not learn such skills. The cerebellum is especially important for learning movement sequences that require precise timing, such as acrobatics, dancing, or competitive team sports. A person with cerebellar damage might be able to learn new dance moves but would probably have trouble learning to synchronize those moves to musical rhythms.

The cerebellum is also important for tasks that involve aiming at or tracking a target. A task that psychologists commonly use to assess such abilities is mirror tracing. In this task, individuals learn to trace drawings by looking at their hand, and the figure to be traced, in a mirror (Figure 4.12 a); meanwhile, the hand and the figure are hidden from their view. It's hard to draw well under these conditions, but if the cerebellum is working properly, the participant will gradually improve at this task. In contrast, a person with cerebellar damage would find learning this task difficult. For example, Robert Laforce and Julien Doyon found that patients with cerebellar damage were much slower at performing a mirror tracing task than individuals in a control group, even after several sessions of training, as shown in figure 4.12b (Laforce & Doyon, 2001).

It is interesting to note in Figure 4.12b that the rate of learning for patients with cerebellar damage was comparable to that of the control group. This seems to suggest that the learning process in the patients with cerebellar damage was similar to that of the control group, and that the patients simply performed more poorly. However, subsequent transfer tests in which both groups traced more complex figures revealed that the individuals in the control group benefited more from their training experiences than did the individuals with cerebellar damage. Thus, although both groups were learning at a similar rate, they were not learning the mirror tracing skill in the same way.

A simple way to show that disrupted cerebellar activity diminishes the ability to learn and perform perceptual motor skills such as those used in the mirror tracing task is to temporarily disable a person's cerebellum with an alcoholic drink, then require the person to learn such a task. The cerebellum is one of the first brain regions affected by alcohol, which is why police officers often use tasks that involve tracking (walking along a stripe in the road, or touching finger to nose) as tests for drunkenness.

So far, we have discussed how the cerebellum contributes to perceptual motor skill learning. Recent brain imaging studies show that activity in the cerebellum also changes when individuals learn certain cognitive skills, such as mirror reading. In the mirror reading task, individuals learn to read mirror reversed text. Researchers found that cerebellar changes that occur during learning of the mirror reading task are lateralized that is, are different in each hemisphere (Figure 4.13), with the left cerebellum showing decreased activity and the right cerebellum showing increased activity with training (Poldrack & Gabrieli, 2001). Are you assuming that both sides of your brain are doing the same thing while you're reading this chapter? .nifigfi dnidT How such hemisphere specific differences in cerebellar processing contribute to skill learning or performance is not yet known.

Keep in mind that almost all cognitive skills require the performance of some perceptually guided movements, such as eye movements. Remember learning earlier in this chapter how chess masters move their eyes to scan a chessboard more efficiently than less experienced players? Similar perceptual motor skills may also be important for tasks such as mirror reading. So, it is possible that changes in cerebellar activity during the learning of cognitive skills might partially reflect the learning of motor sequences required for performing the cognitive activity.

In summary, then, the cerebellum, cerebral cortex, and basal ganglia are each critical, in different ways, to skill learning. If you're having trouble learning a skill, which part of your brain should you blame? Currently, there is no cut-and-dried division of labor between these three brain regions. How critical each region is for encoding or performing any given skill probably depends on the particular skill and your level of expertise. Nevertheless, the cerebellum seems most critical for timing; the cerebral cortex, most critical for controlling complex action sequences; and the basal ganglia, most critical for linking sensory events to responses. Knowing this, which brain region do you think would be most critical for learning to run downstairs? The answer is probably all three, at some point in the learning process. Early on, the cerebellum, visual cortex, and motor cortex may work together to coordinate the timing and sequencing of leg movements. After extensive practice, the basal ganglia may begin to initiate and control more automatic sequences of leg movements. How these three brain regions work together during the acquisition and retention of skill memories is a question that researchers are still attempting to answer.

One feature that all three systems have in common is that skill learning is associated with gradual changes in the firing of neurons in these areas during performance of the skill. This finding means that practice can change the structure of neural circuits to make the control and coordination of movements (or thoughts, in the case of cognitive skills) more accurate and efficient. The most likely mechanism for such changes is synaptic plasticity. Understanding how and when the brain is able to adjust specific synapses within and between the cerebellum, basal ganglia, and cortex will clarify how humans and other animals learn skills.

Interim Summary

Three brain regions involved in the formation and recall of skill memories are the basal ganglia, the cerebral cortex, and the cerebellum. The basal ganglia direct interactions between sensory and motor systems during the learning process, and different cortical networks are specialized for particular functions in controlling and coordinating movements. The cerebellum is critical for learning skills that depend on precise timing of motor sequences.

4.3 Clinical Perspectives

In Chapter 3 you learned how damage to the hippocampus and surrounding brain regions can disrupt memories for events and facts. Damage to the cerebral cortex and basal ganglia resulting from injury or disease can similarly interfere with the formation and use of skill memories. In this section we explore the types of deficits caused by damage and dysfunction in these two brain regions. (We defer discussion of cerebellar disorders to Chapter 7, on classical conditioning.)

The disorders reviewed here have a major impact on society, affecting millions of individuals. Experiments conducted with groups of patients with these disorders provide unique opportunities for understanding how the neural systems involved in skill learning can be disrupted. Unlike the various types of amnesia, in which memory loss can be measured with standard tests, disorders that affect skill learning are difficult to distinguish from disorders that impair skill performance. Nevertheless, clinical studies of patients with skill deficits provide clues about the neural systems responsible for skill learning—information that would be difficult or impossible to obtain through experiments with unimpaired individuals.

Apraxia

Damage to the cerebral hemispheres, especially the parietal lobe of the left hemisphere (Figure 4.14), can lead to problems in the coordination of purposeful, skilled movements. This kind of deficit is called apraxia. The most common causes of apraxia are sharp blows to the head (a typical outcome of motorcycle accidents) and interruption of blood supply to neurons (as occurs during a stroke). Tests for apraxia generally require asking patients to perform or mimic specific gestures. A patient with apraxia can usually voluntarily perform the individual steps that make up the movement or gesture requested by the experimenter, but most such patients cannot combine these steps in appropriately sequenced and coordinated patterns when instructed to do so.

The position and extent of cerebral cortical damage determines what abilities are affected. For example, in patients with left parietal lesions, the greatest impairment is in the ability to imitate actions, whereas in patients with lesions in more frontal areas, the greatest loss is in the ability to pantomime actions that involve the use of both hands (Flalsband et ah, 2001). Sometimes patients are unable to perform a skill with one hand and yet can perform it quite easily with the other. These patients understand what the neuropsychologist is instructing them to do, but they are unable to comply.

Early case studies describing patients with apraxia, such as this description by Pick in 1904, give some sense of the severe problems associated with this disorder:

The patient is requested to light a candle in its holder. He takes the match, holds it with both hands without doing anything further with it. When asked again, he takes the match upside down in his hand and tries to bore it into the candle…..A box of matches is put in his hand. He takes a match out and brushes his beard with it, and does the same thing when given a burning match. Even though he burns himself doing this, he continues, (quoted in Brown, 1988)

Cortical damage clearly causes deficits in skill performance. For example, researchers have found that apraxia can affect individuals' abilities to perform both perceptual motor skills and cognitive skills (Leiguarda & Marsden, 2000; Zadikoff & Lang, 2005). What is less clear is how the cortical damage might be affecting the memories of skills that are lost or the ability to form new memories. One hypothesis for why individuals with apraxia have difficulty performing skills is that they cannot flexibly access memories of how to perform those actions (Rothi, Ochipa, & Heilman, 1991). For example, patients with apraxia who were unable to pantomime gestures such as flipping a coin also had difficulty identifying when an actor in a film performed a specific gesture, such as opening a door (Heilman, Rothi, & Valenstein, 1982). This inability to recognize actions suggests that these patients have not simply lost the ability to generate certain actions, but instead can no longer access memories of those actions.

Studies of skill learning in patients with apraxia suggest that cortical damage interferes with the control and execution of skills more than with the learning and recalling of skills. For example, with practice, such patients can improve at performing skills, and their rate of improvement is comparable to that of unimpaired individuals. The highest level of performance they can reach, however, may be lower than the levels at which unimpaired individuals can perform with no training (Jacobs et al., 1999). How well someone with apraxia can learn a particular task seems to depend on both the nature of the person's deficits and the nature of the task. It remains unclear whether learning in individuals with apraxia occurs through the same neural processes as in unimpaired individuals. Patients with apraxia might make do with a subset of these processes, or they might use alternative mechanisms, such as adopting different strategies during practice sessions.

One way to investigate the conditions leading to apraxia is to create temporary states of apraxia in healthy individuals by inactivating cortical circuits, and then examine the effect on skill learning and recall. This strategy has recently been made possible by transcranial magnetic stimulation, (Figure 4.15), a procedure in which a brief magnetic pulse (or series of pulses) applied to the scalp produces small electrical currents in the brain that interfere with normal patterns of activity over an area of about 1 square centimeter. The disruption lasts for just a few tens of milliseconds, but if timed properly, it can impair skill learning and performance. Researchers can disrupt activity in different cortical regions simply by changing the position of the stimulating device. Transcranial magnetic stimulation is a powerful way of studying how cortical deficits affect the formation and recall of skill memories; however, its use is currently limited, because the stimulation has caused seizures in some participants and the physiological effects of repeatedly disrupting cortical function are unknown.

Currently, the main technique for helping patients with apraxia overcome their deficits is behavioral training that involves extensive repetitive practice. Knowing how different variables such as feedback, pacing, and variety of practice can influence learning (as described above in the Behavioral Processes section) is important for developing appropriate behavioral therapies. Future advances in understanding the cortical networks underlying skill memory will probably suggest important ways of enhancing existing treatments for people with apraxia and developing new therapies.

Huntington's Disease

Huntington's disease is an inherited disorder that causes gradual damage to neurons throughout the brain, especially in the basal ganglia and cerebral cortex. The disease leads to a range of psychological problems (including mood disorders, hypersexuality, depression, and psychosis) and a gradual loss of motor abilities over a period of about 15 years. Facial twitching usually signals the onset of the disease. As Huntington's progresses, other parts of the body begin to shake, until eventually this shaking interferes with normal movement.

Patients with Huntington's disease show a number of memory deficits, some affecting skill memory. Such patients can learn new perceptual-motor and cognitive skills (with performance depending on how far the disease has progressed), but they generally learn more slowly than healthy individuals (Willingham & Koroshetz, 1993). People with Huntington's have particular difficulty learning tasks that require planning and sequencing actions, and they cannot perform the mirror reading or weather prediction tasks (described above) as well as healthy individuals (Knowlton et al, 1996). For example, Barbara Knowlton and colleagues found that an experimental group of 13 patients with Huntington's disease who performed the weather prediction task showed no signs of learning over 150 trials, whereas a control group of 12 healthy persons rapidly improved at the task (Figure 4.16). Recall that experimental studies with animals show that lesions of the basal ganglia can greatly impair skill learning. The basal ganglia damage in patients with Huntington's may explain why they find the weather prediction task so difficult to learn. However, they can learn some other cognitive skills that require similar abilities, such as the Tower of Hanoi task (Butters, Wolfe, Martone, Granholm, & Cermak, 1985).

Individuals with Huntington's typically show large deficits in perceptualmotor skill learning that seem to be related to problems with retrieval and decreased storage capacity. They have difficulty learning the serial reaction time task, the rotary pursuit task, and most other skills that require aiming at or tracking a target. However, it is difficult to determine to what extent deficits in learning of perceptual-motor skills are a direct result of cortical or basal ganglia damage, as opposed to being a side effect of patients' inability to move normally. Imagine trying to learn to throw darts with someone randomly pushing your arm. You would probably have problems improving your throw under those conditions—not because you've lost the ability to store and recall memories of past attempts, but because you don't have control of your movements. It's also not easy to know how the combination of abnormal psychological states and damaged neural systems might affect learning in persons with Huntington's.

Scientists have made great progress in using genetic markers to diagnose Huntington's disease, but prevention and treatment of symptoms are still rudimentary. Using knowledge about the genetic abnormalities found in people with Huntington's, researchers have produced mice and fruit flies with similar genetic abnormalities. Experiments with these genetically engineered animals may provide critical new information about how Huntington's disease affects skillmemory systems and how the deficits caused by this disorder might be overcome. For example, recent experiments with Huntington's disease mice have revealed severe deficits in perceptual-motor skill learning and in the changes in cortical circuits that should occur during learning (Mazarakis et al., 2005). Synaptic plasticity mechanisms such as long-term potentiation and long-term depression (see Chapter 2) are also abnormal in these mice (Murphy et al., 2000). Thus, learning and memory deficits in patients with Huntington's may reflect not only basal ganglia damage but also more fundamental deficits in the ability to modify synapses based on experience.

Parkinson's Disease

Parkinson's disease is another nervous system disease involving disruptions in the normal functions of the basal ganglia and progressive deterioration of motor control. Unlike Huntington's disease, however, Parkinson's does not seem, in most cases, to be the result of heritable genetic abnormalities, or to involve large-scale neuronal death in either the cerebral cortex or the basal ganglia. The main brain damage associated with Parkinson's disease is a reduction in the number of neurons in the brainstem that modulate activity in the basal ganglia and cerebral cortex. These brainstem neurons normally determine the levels of dopamine in the basal ganglia, and when these neurons are gone, dopamine levels are greatly reduced. (We'll give more details on the contribution of dopamine neurons to learning in Chapter 8.)

Patients with Parkinson's disease show increasing muscular rigidity and muscle tremors, and are generally impaired at initiating movements. Symptoms of the disease usually do not appear until after the age of 50, but can arise much earlier (for example, the actor Michael J. Fox was diagnosed with Parkinson's when he was in his thirties).

Not surprisingly, people with Parkinson's have many of the same skill-learning impairments as people with Huntington's. Both diseases make it harder to learn certain perceptual-motor tasks, such as the serial reaction time task and tracking tasks (including the rotary pursuit task). On the other hand, individuals with Parkinson's can learn some skills, such as mirror reading, that cause problems for those with Huntington's (Koenig, Thomas-Anterion, & Laurent, 1999). This suggests that although both diseases affect processing in the basal ganglia and cerebral cortex, the damage each causes leads to different but overlapping deficits in skill-memory systems.

Currently, the main treatments for Parkinson's disease are drug therapies for counteracting the reduced levels of dopamine and surgical procedures aimed at counteracting the disruption caused by lack of dopamine in the basal ganglia. One recently developed surgical technique, deep brain stimulation, seems to hint at a way of curing Parkinson's disease, but scientists do not yet know exactly why it works. It involves delivering an electrical current through one or more electrodes implanted deep in the patient's brain. Neurosurgeons place the end of the electrodes near neurons that are part of the basal ganglia-cortical loop (for example, in the thalamus or basal ganglia), as shown in Figure 4.17. When electrical current from an implanted stimulator passes through these electrodes, many of the motor symptoms associated with Parkinson's disease, such as tremors, disappear within seconds, although they eventually return. One theory of how this technique works is that without proper levels of dopamine, interactions between neurons in the cerebral cortex and the basal ganglia become locked into fixed patterns (Dowsey-Limousin & Poliak, 2001). This creates a situation similar to the endless back and forth of young children arguing (Child 1: "No, you be quiet!" Child 2: "No, you be quiet!" Child 1: "No, you..." ad infinitum) and disrupts the control of movements. Stimulation from the electrode is thought to quiet both brain regions, allowing normal brain activity to resume. Deep brain stimulation is still in the early stages of development, but it illustrates how increased knowledge of the brain systems underlying skill memories can help doctors treat these systems when the systems go awry.

Conclusion

Kissing requires both perceptual-motor and cognitive skills, acquired and improved through observation and practice. Differentiating the cognitive aspects from the perceptual-motor ones can be difficult, as this chapter shows. Cognitive skills often depend on perceptual-motor skills (and vice versa), and may even become transformed into perceptual-motor skills over time.

Certainly, one cognitive aspect of kissing is the use of social skills to motivate someone to want to kiss you or be kissed by you. Once you solve this problem—which in some cases may be as strategically challenging as a chess game—you face the perceptual-motor challenge of coordinating your own kissing movements with those of your partner, based on what you perceive of your partner's maneuvers. Your skills at this point will depend on how much and how often you have practiced, as well as on the types of feedback you have received from past partners. Perhaps you are in the cognitive stage of learning to kiss, still thinking carefully about each move you make; or perhaps in the associative stage, feeling comfortable with your performance but knowing there is room for improvement. Possibly you are at the autonomous stage of skill acquisition, having become an expert—your kissing depends on various motor programs that you perform without thinking. If you are an experienced kisser, the skill memories you rely on are dependent on the coordination of several brain regions, including the basal ganglia, the cerebral cortex, and the cerebellum.

In short, there is more to kissing than simply recalling and executing a fixed series of movements. Kissing is an open skill in which the recent actions and reactions of your partner provide important feedback that you can use to guide your own actions. Keeping track of what has happened in the recent past is thus a key component of skillful kissing. The ability to maintain and flexibly use memories of the recent past depends on different brain regions from those we have focused on thus far in our discussion of skill learning and performance. You will learn more about these kinds of memories and their neural substrates in the next chapter, on working memory.