Is the ability to execute a sequence of movements smoothly and accurately and it may involve the senses?

Publisher Summary

This chapter discusses the temporal patterns of rapid movement sequences in speech and typewriting and what these patterns might mean in relation to the advance planning or motor programming of such sequences. The chapter discusses response factors that affect the time to initiate a prespecified rapid movement sequence after a signal when the goal is to complete the sequence as quickly as possible as well as how such factors affect the rate at which movements in the sequence are produced. The response factor of central interest is number of elements in the sequence. The effect of the length of a movement sequence on its latency is based partly on the possibility that it reflects a latency component used for advance planning of the entire sequence: The length effect would then measure the extra time required to prepare extra elements. The idea that changes in reaction time might reflect changes in sequence preparation in this way proposed that simple reaction time increased with the number of elements in a sequence of movements made with one arm. A part of the reaction time includes the time to gain access to stored information concerning the whole sequence: a process akin to loading a program into a motor buffer, with sequences containing more elements requiring larger programs, and larger programs requiring more loading time.

D The Motor Program, Feedback, and Learning

Motor program advocates are concerned with memory but they are not much interested in the learning operations that store the movement sequence in memory in the first place, so they fail to specify how the program is acquired. Is feedback irrelevant throughout the learning of a program? Or does feedback only play a role early in learning, with the program being a developing independence of feedback as learning progresses? An experiment by Adams et al. (1972) was designed to test these possibilities. Linear positioning was the task that they used, and it was learned under either augmented feedback or minimal feedback. With augmented feedback there was full vision, the subject could hear the slide which he manipulated move along its track, and he had spring tension on the slide to give heightened proprioceptive feedback. Minimal feedback was none of these. After low or high amount of practice with knowledge of results, the knowledge of results was withdrawn and the feedback either remained the same or was switched, depending upon the experimental condition. If the motor program is independent of feedback throughout its learning then feedback change should make no difference whether the amount of learning was low or high. But if the learning of a program is gradual liberation from feedback, then feedback change should produce a performance difference at the low level of learning but not the high level. Neither of these things happened. Feedback change made a big difference for both levels of learning. Moreover, the difference was the greatest for the high level of learning, which is contrary to the possibility that a motor program develops with practice. The authors interpret their data in support of my closed-loop theory of motor learning which has a role for feedback throughout all stages of learning.

A Simple Movement Structures

One approach to the problem of representation was investigated by Keele (1975). To capture in abstract form the sequential property of many skills, a task was developed in which sequencing was the primary component. There were eight lights in a horizontal line and beneath the lights was a row of response keys, one key for each light. When a light appeared, the subject pressed the corresponding key, extinguishing the light. Another light requiring a response then came on, and so on. During the initial training the lights appeared in a recurring order. If the lights are designated 1 through 8 from left to right, the order of occurrence for all subjects was 18347562 after which it cycled back to 1 with no break. This task is much like playing a piano or typing with the exceptions that it is simpler, very easy to learn, and cycles repetitively through the sequence much as does walking. Within a half-hour subjects can execute the series at rapid speeds.

Once the skill is well imprinted in memory, how is it represented? Two possibilities were compared by examining response times when the proper order of events was momentarily disrupted and then restarted. One possible representation, event-to-event associations, posits that successive movements are associ-ated with each other. The other hypothesis, event-to-position associations, posits that movements are not associated one with the other but with positions in the sequence.

To obtain a better idea what these hypotheses mean, consider the experimental situation. Suppose a subject is presented sequentially with lights 3475, which are in correct order, and the next light is out of order and is light 8 (recall that the complete correct sequence is 18347562; see Figure 3). Because light 8 is unexpected, people are slow in responding to it. Now there are at least two interesting ways for the sequence to restart following an unexpected event. In the one case the very next light is 3, the one that normally follows light 8. The hypothetical sequence then is 3475834 … and so on till the next out of order event. In the second case, light 2 follows 8, so the hypothetical sequence would be 3475821 … until the next light out of order. The unexpected event 8 replaces the expected event 6 but subsequent events occur in their proper positions. Light 2 is normally the second light after 5 regardless of which light intervened. In the first case, therefore, an event is predictable by the preceding event, even though the preceding event itself may be out of order. In the second case, an event is predictable by the position in sequence regardless of what the preceding event was.

The issue, of course, is how well people can respond to the first event back in sequence. If the return to sequence is predicted by events, then people should respond rapidly to the first event back in sequence only if the memory structure consists of event-to-event associations. On the other hand, if the return is predicted by position, people should do well only if the memory structure consists of event-to-position associations.

People in the experiment received training on the cyclical sequence until they could respond rapidly and the skill presumably was well set in memory. On the next day the task remained about the same, but people were told that 20% of the lights would be out of the expected order, and following such an intrusion the sequence would restart. For half the people the event succeeding an out of order one was predicted by the identity of the out of order event. For the other half, the event succeeding the out of order one was predicted by the position in the sequence. An additional variable was the interval between one response and the next light. With a relatively long interval (1500 msec RSI), allowing ample time for anticipation of the following light, it was expected that people could effectively use either type of sequence return. But at short response-stimulus intervals (50 msec RSI) people might experience difficulty in using one type or the other of sequence return.

Reaction times are shown in Figure 4 for unexpected lights, designated the 0 position in the sequence, and for the first, second, third, and fourth lights back in sequence. Although people in the position condition respond slightly faster on the average when a long RSI allows ample time for preparation, they are slower than event condition subjects at the short RSI. In other words, when the subjects are pressed for time, they are unable to effectively use position information, but use of event information suffers little. The improvement in reaction time from the unexpected event to the first back in sequence is a measure of preparation for the succeeding event. At the long RSI both conditions show an improvement on the order of 160 msec. However, at the short RSI, when spare time is not available, people in the event condition again show a large improvement but people in the position condition show very little improvement. The modest improvement that does occur in the latter case is probably in part an artifact: If two successive unexpected events occur, the second one is usually responded to faster than the first as though the unexpected is no longer surprising. When time is short, an event predictable by position is little or no better, therefore, than another unpredictable event.

Our results suggest that in some skills a sequence of events and their corresponding movements are stored in a chain of event-to-event associations. This conclusion is very much in line with the theoretical position of Wickelgren (1969), though he was primarily concerned with the phonetic representation of words. A classic paper by Lashley (1951) suggested in contrast that the memory structure of serial order is independent of the actual events that fit into the structure. A position structure of the sort investigated in this study would be an example of a structure other than event associations but by no means the only possible one. According to the position conception, the skill is represented as a number of slots, one for each movement. As the skill progresses, one slot after the other is examined for content, and the event or movement contained in the slot is prepared for. Preparation for a movement depends not on the preceding event but only on the current position. A position stepping structure is not supported by this experiment, but in other skills with structural regularities, a representation that is neither event associations or purely position associations may be indicated. This point will be returned to later.

The preceding experiment does not support the position hypothesis as outlined, but with other assumptions the position theory might be accommodated to the data. For example, when an unexpected event occurs, a position marker might skip forward in the sequence until the position of the unexpected event is found. According to this view a position hypothesis would make the same prediction as the event association hypothesis for the preceding data. The next experiment dealing with repeated elements was therefore designed as a further test between the two theories.

A conceptual problem faced by the event-to-event hypothesis concerns repeated elements. If the identical movement occurs at more than one place in a movement sequence, and each time it is followed by a different movement, then how are simple one-to-one assocations able to determine what movement should follow the repeated one? In typing, if the letter H follows the letter T in one position and E follows T in another position, how does the typist know whether to type H or E following a T if only event associations are used. Associational hypotheses might be elaborated in several ways to handle this problem. Wickelgren (1969), for example, proposed that seemingly identical elements in different portions of a sequence in fact are not identical and differ slightly depending on their surrounding movements.4 Nevertheless, repeated elements, even though partially differentiated by context, should be more similar to each other than to other elements. While repeated elements in a sequence do not destroy the sequencing, they should constitute weak links in the associational chain. In structural models such as Lashley's, however, there should be no particular problem deciding which movement should follow the repeated one, for there are no event-to-event associations.

Wickelgren (1966) investigated the issue of repetition in short-term memory for lists of nine letters. In some lists, a letter was repeated twice and followed by a different letter in each case. During recall when an error was made on an item following a repeated item that error was often an intrusion of the item that followed the repeated one at the other place in the list. Such associative intrusions were more common than in control lists without repeated elements. These results are consistent with an event-to-event associative model.

Short-term memory may be rather different than well learned skills, however, so we explored the repeated elements problem in our skills paradigm.5 Six lights and six response keys were used instead of eight as in the earlier experiment. In this case, however, one of the lights occurred three places in a sequence, each time followed by a different light, so the total event sequence was eight in length. As before, the sequence recycled after the last item. The order of lights was different for each of 12 subjects, but all sequences obeyed the same structure. An example of one sequence is 14364542…. Each subject practiced on the fixed sequence over 400 cycles in one session. During the second session, when all subjects were quite proficient on the task, 30% of the lights were out of order and the remaining lights occurred as expected in the learned sequence.

Of particular interest are the reaction times and errors to the out of sequence lights, and they were classified into three types as illustrated in Figure 5. On some occasions the repeated light occurred in its proper order but was followed by an out-of-order light. If the out-of-order light normally followed the repeated event at another place in the sequence it was called an associated light. Thus, for the illustrative sequence, if 14364 is followed by 3 or by 2, then 3 and 2 are associated events. If a repeated light is followed by one that normally does not follow the repetition in any position (lights 1 and 6 in the illustration sequence), it was called an unassociated event. Finally, any out-of-order light following a nonrepeated light is a control. If Wickelgren's (1969) associative hypothesis is correct, out-of-order but associated lights should result in faster reaction times and fewer errors than either unassociated or control lights, since occurrence of the repeated light should elicit all of its associates.

The reaction times and proportion of errors to the three types of out-of-order events are shown in Table I. The results conform to expectations, but statistically not all of them are significant. The associated events were faster than control events for all 12 subjects (p < .01) and more accurate for 9 subjects, with 2 ties and 1 reversal (p < .05). Although 8 of 12 subjects show faster reaction times and greater accuracy for associated than for unassociated events, these two comparisons are not significant. Examination of the table shows, however, that unassociated events are faster and less error prone than control events. Why would that be since both are controls in the sense that neither is associated with a repeated event? One possibility is that events following the repeated one are relatively ambiguous and hence are not strongly prepared for. In the absence of a strong expectation, any out-of-sequence event following a repeated light tends to be responded to faster than out-of-sequence events following nonrepeated lights, although associated events are responded to fastest of all. This tentative explanation is consistent with the Wickelgren hypothesis that repeated elements produce weak links in the associative chain.

Table I. Mean Reaction Times (msec) and Proportion of Errors for Out-of-Order Events

To check the explanation, the data were analyzed from the first session in which the subjects responded as rapidly as they could to lights that always occurred in their proper order. If the linkage between a repeated event and ensuing events is rather weak, because of associative ambiguity, those events should be responded to rather slowly or with high errors. The relevant data are shown in Table II. Although reaction times do not differ among event types, errors do differ. As predicted, 9 of 12 subjects exhibited higher error rates for events following repeated events than for controls (p < .05). When these data are put together with the earlier data there is sufficient consistency to conclude that a repeated event elicits associations of all the events that follow it in different positions.

Table II. Mean Reaction time (msec) and Average Number of Errors per Subject to Different Event Types

This second experiment using a somewhat different method supports the conclusion of the first experiment that some skills are represented in memory as associations between successive movements or events. The particular skills studied in the two experiments are of a particular type, however. They have no inherent structure other than linear ordering of events. Thus, we have mainly shown that for unstructured sequences of eight or nine events a position representation is not adopted and instead event-to-event associations are formed. This outcome might have been expected from general memory theory and observations of absolute judgments (Keele, 1973; Miller, 1956). When the number of events exceed about half a dozen there are too many for strict position recall, since that involves only one level or one dimension of organization. Instead they are recalled in a highly structured manner. If the events are quite unrelated to one another, associations between successive events are likely to be used. But when some other basis exists for categorizing events, recall is structured around categories or in a hierarchical organization (e.g., Mandler, 1967; Nelson and Smith, 1972). The issue arises, therefore, whether movement sequences with a possible structural basis other than linear positions lead to storage modes other than event-to-event associations.

E Conclusion

These, then, are the four legacies of Edward Thorndike. Two of them we accept without thinking: the one that causes us to shy from punishment operations, and the one that has us reward the success of the act, ignoring the details of the movement sequence. I have suggested that we cast these restrictive legacies aside. The other two legacies on why KR affects behavior as it does, and the need for objective KR, have been challenged. We have new ideas on these matters, we are trying to refine them, we are arguing about them, and we are doing experiments on them. We are moving slowly on these new ideas, however. Maybe the new ideas are wrong, but maybe the power of the Thorndike tradition makes it difficult for us to adopt new ways of thinking. I submit that a scientist who can hold our minds so firmly for so long deserves respect.

Feedforward

As just indicated, whenever performance is consistently accurate though feedback is removed the performer must have relied on feedforward control. The accuracy of performance provides an indication of the accuracy of the performer’s feedforward control.

Under open-loop conditions, a number of movement sequences can be performed surprisingly well. For example, monkeys deprived of sensory feedback from their limbs can walk and climb, though they do so less gracefully than monkeys with sensory feedback (Taub & Berman, 1968). Sensory feedback is eliminated in these animals by cutting the nerve fibers that transmit sensory signals from the limbs into the spinal cord. When these same nerve fibers are damaged in humans as a result of accident or disease, some movement control may still be possible (Lashley, 1917). For example, a man who could not feel his body because of a disease affecting his sensory nerves could draw complex figures on command, could sequentially touch his thumb with each finger of the same hand, and could touch his nose. He could do all these things without the aid of visual, auditory, or other forms of feedback (Marsden, Rothwell, & Dell, 1984).

Abilities such as these indicate that the gross features of some movements can be performed under feedforward control. When these same movements are performed with feedback, however, they are usually performed more precisely and gracefully. This outcome suggests that everyday movements reflect the combined use of feedforward and feedback.

Can feedforward be inferred only by removing feedback? The answer is no, as can be demonstrated through the simple act of looking into a mirror and trying to watch your eyes move. You cannot see your own eyes move, as was noted over a century ago (Dodge, 1900). If you have a friend look at your eyes while you move your eyes, he or she will have no trouble seeing your eyes dart about. This shows that eye movements are not simply too quick to be seen.

Why can’t you see your own eyes move? An intriguing hypothesis (Volkmann, Schick, & Riggs, 1969; see Chapter 6) is that your brain suppresses visual inputs arising from your eye movements when your eyes jump from place to place or, said another way, when your eyes make saccades (the French word for jumps). There could be a distinct functional advantage of such saccadic suppression. Since the retinal image is smeared during saccades, the smear might serve no useful purpose for perception. Saccadic suppression could help reduce the damage to visual perception caused by such retinal smearing. Chapter 6 provides a more extensive discussion and critical evaluation of this hypothesis.

Suppression effects are not limited to eye movements. Chewing sounds are loud, yet we barely hear them. The reason is that during chewing, there is internal suppression of auditory feedback (Rosenzweig & Lehman, 1982). Similarly, during active hand movements, sensitivity to tactile stimuli is reduced (Demaire, Honoré, & Coquery, 1984). Such reduced sensitivity to self-produced actions may explain why we can’t tickle ourselves (Blakemore, Wolpert, & Frith, 1998).

Feedforward helps distinguish perceptual changes due to motion arising in the external environment from perceptual changes arising from one’s own motion. The disambiguation occurs by subtracting perceptual changes caused by motor commands (or internal perceptual representations leading to motor commands) from perceptual changes caused by changes in the external environment.

Support for this hypothesis came from a remarkable experiment with flies (von Holst & Mittelstaedt, 1950). The experiment was prompted by the observation that when a fly stands still and a drum with vertical stripes turns around it, the fly turns with the drum, presumably to keep itself stationary with respect to the external world. This behavior is known as the optomotor reflex. When the stripes are stationary, the same fly moves freely in front of them. The visual stimulus is approximately the same when the fly moves and the stripes are stationary or when the fly is stationary and the stripes move. Why does the fly turn with the stripes when the stripes turn, but seems to disregard the stripes when they are stationary?

To answer this question, von Holst and Mittelstaedt twisted the fly’s head 180 degrees and glued the head in this new position (see Figure 2.12). Under this condition, the fly’s behavior was, to say the least, strange. When the fly stood still and the vertical stripes were turned to the right, the fly turned to the left, and when the fly stood still and the vertical stripes turned to the left, the fly turned to the right. However, when the fly attempted to move on its own, it took a step one way or the other and kept doing this indefinitely.

How did von Holst and Mittelstaedt (1950) explain these results? According to the authors, the fly “expects a quite specific retinal image displacement, which is neutralized when it occurs” (p. 179). In other words, when the fly turns to the right, it has a reference signal for a retinal displacement to the left. Similarly, when the fly turns to the left, it has a reference signal for a retinal displacement to the right. Obtaining the expected retinal displacement indicates to the moving fly that the world has remained stationary. By contrast, if the fly is stationary and the retinal image moves, the shift of the retinal image indicates to the fly that it has lost its bearings with respect to the external world. Consequently, the fly turns to realign itself with its surroundings. If the eyes are spatially interchanged, as in the experiment of von Holst and Mittelstaedt, the expected and obtained retinal image displacements are reversed. The result, as von Holst and Mittelstaedt (1950, p. 179) put it colorfully, is a “central catastrophe.”

Considering the behavior of the flipped-head fly shows how sophisticated the perceptual-motor system can be, and at how early a stage of evolution this sophistication took hold. Not surprisingly, internal subtraction processes of the kind proposed for the fly have also been attributed to higher animals, including people (von Holst & Mittelstaedt, 1950; Sperry, 1950; Wolpert & Flanagan, 2001). For human and nonhuman animals, the perceptual consequence of one’s own actions is sometimes called reafference. The first person to recognize the importance of reafference and the role of feedforward in distinguishing reafference from exafference (perceptual input arising from outside influences) was the great German physiologist Hermann Helmholtz (1866/1962).

Motor Programs

I have allowed a word to enter this discussion which has become the subject of intense debate among students of human motor control. It is motor program; the related term is motor programming. In the late 1960s, the motor program was defined as “a set of muscle commands that are structured before a movement sequence begins, and that allows the sequence to be carried out uninfluenced by peripheral feedback” (Keele, 1968, p. 387). Keele wrote this definition when the major issue in motor-control research was the extent to which skilled movement depends on sensory feedback. He reasoned that if a movement sequence can be performed skillfully even when sensory feedback is unavailable, then one can conclude that the sequence can be controlled centrally. The word program denoted the set of commands within the central nervous system that allowed for such performance.

A number of complaints arose about the programming concept (Kelso, 1981; Meijer & Roth, 1988). One is that sensory feedback has an effect on movement. Much of the grace and subtlety of movement that is present when feedback is available deteriorates when feedback is withdrawn. This suggested that the concept of motor program as defined by Keele (1968) has limited utility.

The problem with this challenge is that a careful reading of Keele's (1968) definition shows that though a motor program may allow a movement sequence to be carried out uninfluenced by peripheral feedback, it does not require movement sequences to be uninfluenced by peripheral feedback. Consider a conventional computer program. Such a program is designed to carry out procedures differently depending on what input values it receives or depending on the outputs it produces at earlier stages. Thus a conventional computer program is not immune to feedback and neither is a motor program.

Analogizing motor programs to computer programs is a second source of dissatisfaction with the motor program concept. The nervous system is quite different from a computer, the argument goes, so the term program is misleading. There clearly are differences between the nervous system and most modern computers. Computer systems today primarily rely on serial processing, whereas an important feature of the nervous system is that it relies extensively on parallel processing. Moreover, the individual processing elements of a computer are very fast, whereas neurons are comparatively slow. Nevertheless, computers are likely to change dramatically in the next few years. Indeed, there is a concerted effort to make them more “brainlike.” When this happens, the complaint that human motor programs aren't like computer programs will no longer apply. The implication is that the complaint about computers is too narrowly related to current technology.

A third grievance about motor programs is related to the use of the term muscle commands. Use of this term led to the reproof that information guiding movement is more abstract than commands for muscle contractions (Tuller, Turvey, & Fitch, 1982). As was seen in the last chapter, only some efferent signals directly activate motor neurons; efferent signals also influence gains in feedback loops, for example. Moreover, as will be seen in later chapters, there is reason to doubt that information governing movement and stability is defined with respect to the activity of particular muscles or groups of muscles (Klapp, 1977b). Accepting these observations, the larger point is that difficulties with a particular term (muscle commands), used by one author (Keele, 1968) more than 20 years ago, need not rule out the underlying concept of a motor program. A motor program can now be viewed as a functional state that allows particular movements, or classes of movements, to occur.

The word command—whether viewed as a command for muscles or commands for perceptual states–has also raised criticism. Who decides what command to issue? And what defines the functional boundary between the sources of commands and their targets? Introducing the notion of motor program seems to require the notion of a motor programmer, which begs the question of how movements are generated. I agree with this concern, but I think that if the program is viewed in the broad terms given at the end of the last paragraph, the question of who or what does the programming becomes irrelevant.

Viewing the motor program as a functional state need be no more controversial than the concept of memory. To say that we have memories makes no particular claims about what form the memories take or how they are embodied physically. To say that the term memory is useless because it is not specific has the potential of endangering commitment to understanding the nature of information storage, and this, in my view, would be a grave mistake. Likewise, it would be most unfortuante if one were to deny the importance of detailed, testable hypotheses about the functional states underlying motor control, including the functional states allowing for the preparation of forthcoming movements and movement sequences. The term motor program is a convenient label for these states, and the term motor programming is a useful term for the processes by which the states evolve. Though neither term is specific (Kugler & Turvey, 1987), both terms invite inquiry into the detailed nature of motor preparation and control. Questions to be answered about motor programs are how information about the biomechanical properties of the skeletomuscular system are represented, how physical interactions with the external environment are taken into account or exploited, and so forth. As these questions show, research on motor programs need not devalue or ignore physical factors, as some critics have charged. Presumably we know a great deal about mechanics, even if only implicitly. Motor programs incorporate this knowledge so that movements can be prepared and carried out efficiently.

Motor Programs

A word used in the last sentence—”programmed”—was the subject of intense debate among students of human motor control in the 1980’s and 1990’s. Earlier than this, in the late 1960’s, the motor program was associated with the control of ballistic movements. Keele (1968) defined the motor program as “a set of muscle commands that are structured before a movement sequence begins, and that allows the sequence to be carried out uninfluenced by peripheral feedback” (p. 387). Keele wrote this definition when the major issue in the field was the extent to which skilled movement depends on sensory feedback. Keele reasoned that if a movement sequence can be performed skillfully even when sensory feedback is unavailable, one can conclude that the sequence was centrally controlled. For him, the word “program” designated the set of commands within the central nervous system that allowed for such control.

A number of complaints arose about the programming concept (Kelso, 1981 Meijer & Roth, 1988). One was that sensory feedback has an effect on movement. Much of the grace and subtlety of movement that is present when feedback is available deteriorates when feedback is withdrawn. Kelso and Meijer and Roth argued from this outcome that the motor program, as defined by Keele (1968), had limited utility.

The problem with this challenge is that a careful reading of Keele’s (1968) definition shows that though a motor program may allow a movement sequence to be carried out uninfluenced by peripheral feedback, it does not require peripheral feedback to have negligible effects.

Consider a conventional computer program. Such a program is designed to carry out procedures differently depending on what input it receives. Thus, a conventional computer program is not immune to feedback; quite the opposite is true. The analogy to computer programs is a second source of dissatisfaction with the motor program concept. The nervous system is quite different from a computer, the argument goes, so the term “program” is misleading.

Clearly, there are differences between the nervous system and most modern computers. Most computer systems today rely on serial processing, whereas one of the most important features of the nervous system is that it relies on parallel processing. Moreover, the individual processing elements of a computer are very fast, whereas the individual processing ­elements of the nervous system—neurons—are comparatively slow. Nevertheless, computers are likely to change dramatically in the next few years. Indeed, there is a concerted effort to make them more brain-like. When this happens, the complaint that human motor programs aren’t like computer programs may no longer apply. The implication to be drawn is that the word “program” may be offensive in large part because of the current state of technology.

A third complaint about motor programs is related to the use of the term “muscle commands.” Use of this term led to the reproof that information guiding movement is more abstract than commands for muscle contractions (Tuller, Turvey, & Fitch, 1982). Indeed, as was seen in the last chapter, only some efferent signals directly activate motor neurons; efferent signals also influence gains in feedback loops, for example. In addition, as will be seen later in this book, there is reason to doubt that information governing movement and stability is defined with respect to the activity of particular muscles or muscle groups (Klapp, 1977b). These observations having been made, the larger point is that difficulties with a particular term (“muscle commands”) used by just one author (Keele, 1968) need not rule out the underlying concept of a motor program, which can be defined here as a functional state that allows particular movements, or classes of movements, to occur. For the author of this book and other investigators who subscribe to the cognitive approach to the study of motor control (i.e., an approach that does not shun memory representations or memory codes), understanding these functional states—that is, understanding motor programs—is arguably the single most important aim in the study of motor control (human or otherwise).

Task-Oriented Training

High number of repetitions has been proven successful for training of the lower extremity [16]. However, simple repetitive motions alone do not induce plasticity to the same extent in the upper extremity. Neurophysiological and neuroanatomical changes in the motor cortex are rather induced by skill acquisition [16,17]. Skill acquisition can be defined as practice-dependent modification of temporal and spatial organizations of physiological muscle synergies. The results are movement sequences that are smooth, accurate, automatized, and persistent over time [18,19]. There is a continuum of increasing difficulty from more basic to complex tasks. Tasks performed in activities of daily living (ADL) are often complex and involve several DOF. They may be too complex to be mastered within a single training session [20]. However, training of complex functional movements seems necessary to increase the transfer of a learned skill to daily life. Difficulty is not objective, and particularly after brain injury, the complexity of a task may be perceived considerably different between subjects depending on how challenging it is relative to the skill level of the subject (“functional task difficulty” [13]). Therefore, tasks should be chosen and continuously adapted in accordance to the abilities of an individual subject. Furthermore, there is a transfer of skill from one task to another, or in other words, learning a skill can be facilitated by prior practice of a similar skill [21]. Thus, task-oriented training in a therapy setting should incorporate movement components and an environment that resembles the targeted task in the relevant functional context (i.e., the home setting) [22].

4 HAND-ARM CONTROL

PSIGHN is mounted on the Intelligent Arm, which was developed for the Intelligent Hand. The Intelligent Arm has 8 DOF and consists of base frame, shoulder frame, upper arm frame, and forearm frame. The rotation mechanism is on the base frame for horizontal rotation movement. For the shoulder and the elbow joint, universal joints are applied for independent longitudinal and transverse rotation. A coaxial differential gear mechanism is designed for a rotation-flexion-rotation mechanism at the wrist. The length ratio of each frame is designed similarly to a human arm. Seven AC servomotors and one DC motor are mounted to drive the joints.

Studying the relation of the movements of the human upper limb has set a target on designing a control for human-like movement sequence of the multiple D.O.F. in the hand and arm. We arranged a X-Y-Z coordinate system at the shoulder that divides the area around the body, and classified the movements of the arm into 14. This method is also applied to the wrist and digits to classify the movements and setting a reference model. The classified movements are then described by linking the required joint actions and to the modes of hand. The five-digit hand performs a broader range of modes for prehension, but the modes are strongly restricted by the arm and wrist orientation. Therefore, the classification and networking of the hand-arm relation becomes the key in controlling the anthropomorphic hand.

Joint angles are previously computed at grind points set around the orthogonal coordinates. The arm movement will be driven by the differentiation of the present and target wrist position. The palm and elbow are reoriented at target grid point based on the hand mode and object approaching direction, and finely positioned toward the object. This method is implemented to reduce the time delay in response for positioning the arm.

Tactile sensory information is feedback to adjust the digits for performing the hand modes. The multiple sensors laid out in a continual diamond pattern works as a slippage detection sensor by observing the time-series behavior of the sensory response. The classification of the hand mode based on the contact area provides a focused monitoring and digit control for gripping the target object. The hand and arm is controlled to change its posture to overcome the slippage. The composite sensor for tracking the orientation of the arm is in development.

Theories of Basal Ganglia Function

In view of the unfortunate consequences of basal ganglia disease, what can we say are the functions served by this set of neural structures? From the outset, it is important to observe that while the basal ganglia are involved in motor control, they may serve non-motor functions as well. Patients with basal ganglia disease don’t just have difficulty performing movement sequences; they also have difficulty recalling sequences of symbols (Côté & Crutcher, 1985; Gunilla, Oberg, & Divac, 1981). Also, as mentioned earlier, dementia is one of the symptoms of Huntington’s disease. That this is true should not be surprising in view of the fact that the basal ganglia appear to be important in learning and maintaining skills, as shown in neuroimaging of the two main structures of the basal ganglia, the caudate and putamen (Grafton, Hazeltine & Ivry, 1995; Poldrak, Prabakharan, Seger, & Gabrieli, 1999).

One hypothesis about the role of the basal ganglia in motor control is that the structures in this neural constellation contribute to the activation or retrieval of movement plans (Marsden, 1982). Consistent with this hypothesis, neurons in some basal ganglia structures (the globus pallidus and the zona reticulata of the substantia nigra) have been found, in monkeys, to discharge before voluntary movements of the arm or leg and before chewing or licking movements (Iansek & Porter, 1980). Similarly, Parkinson’s patients have difficulty beginning voluntary movements. For example, they may have difficulty starting to walk when asked to do so, although, paradoxically, they may start walking with no difficulty in a different intentional context—when asked simply to leave the room. Visual stimuli can also be effective in helping Parkinson patients start to walk. Figure 3.18 shows that the placement of markers on the floor at regularly spaced intervals can help Parkinson’s patients locomote in a nearly normal fashion. Providing the markers can be nearly as effective as L-DOPA medication (Forssberg, Johnels, & Steg, 1984).

Besides serving to retrieve or initiate movement plans, the basal ganglia may serve to scale the amplitudes of movements. Basal ganglia disease may disrupt the overall size and timing of movements in such tasks as handwriting, reaching, grasping, and manual positioning (Brooks, 1986). The pallidum appears to be the structure within the basal ganglia that is mainly responsible for this scaling function.

A final function served by the basal ganglia pertains to perceptual-motor integration. Cells in the caudate nucleus of the basal ganglia have been found, in sedentary cats, to respond to light brushing of the face, but to respond at different levels if the same stimulus is applied during mouth or head movements (Manetto & Lidsky, 1989). Manetto and Lidsky (1989) suggested that the basal ganglia may be a site where perceptual inputs are gated by motor activity.