14	Control of Movement and Learning of Motor Skill

Rather than viewing perceptual-motor behavior as a series of motor responses made to reach some goal, it is possible, and I believe considerably more profitable, to view such behavior as an information-processing activity guided by some general plan or program.

P. M. Fitts
1964

INTRODUCTION

Any interaction between a person and a machine or the natural environment ultimately requires the person to execute a motor response—to move his or her body. This movement can be as simple as pushing a button or as complex as the coordinated actions required to perform heart surgery or to operate heavy machinery. In all cases, the person must not only perceive information correctly, make suitable decisions, and select appropriate responses, but also successfully carry out the intended actions. Often, the limiting factor in performance will be the speed and precision with which these actions can be executed. Because motor control is a major component of many tasks, the human factors specialist must understand the ways that simple and complex movements of various types are planned and executed.

Every movement requires the cooperation of different muscle groups and the neural mechanisms that control them. The limbs that will execute the action (the effectors) must be selected and prepared, sequences of movements must be timed and coordinated, and the final movements must be executed with the right force and speed to accomplish the goal. The role of the nervous system is to activate the proper muscle groups in a precise order and use feedback from the various senses to coordinate and modify ongoing movements, maintain posture, and plan future actions.

Think for a moment about the skills required to ride a bicycle. Maintaining balance is an important part of the process. Your legs must pedal with force sufficient to attain the speed you want. You need to be able to steer the bicycle with your arms, and brake (with your hands or legs) when necessary. The coordinated performance of all of these actions requires constant monitoring of proprioceptive, visual, and vestibular feedback. When you first tried to ride a bicycle, it probably seemed as if it was impossible to do all of these things at the same time. Yet, most people learn to ride a bicycle very quickly. In this chapter, we will discuss the principles underlying the control of movement and how complex motor skills are learned. Most research in these areas is conducted from one of two perspectives: cognitive science and human information processing, or ecological psychology and dynamical systems (Rosenbaum, Augustyn, Cohen, & Jax, 2006). Our emphasis in this chapter will be on the cognitive science and human information processing approach, but you should keep in mind that the two approaches are complementary rather than antagonistic (Anson, Elliott, & Davids, 2005).

PHYSIOLOGICAL FOUNDATIONS OF MOVEMENT

In Chapters 5 and 7, when we introduced the visual and auditory perceptual systems, we provided a brief description of the sensory structures that underlie perception. We will do the same thing in this chapter for the motor system. We will first discuss how the human body is engineered for movement and those structures in the nervous system that are involved in movement.

THE MUSCULOSKELETAL SYSTEM

There are 200 bones in the adult human skeleton that provide support for the body. The bones are joined by connective tissues called ligaments, and similar tissues, called tendons, connect muscles to the bones. Movement is accomplished by muscular contractions acting on the bones. All movement takes place through the changes that muscles make on the joint angles.

Movements at some joints occur only in two dimensions and are said to involve one degree of freedom (movement in a single plane). The movement of the forearm in relation to the upper arm at the elbow is of this type. Other movements can occur in more dimensions and involve multiple degrees of freedom. For example, movement of the upper arm at the shoulder involves three degrees of freedom (left-right, up-down, and rotation). Because most limb movements involve motion at multiple joints, there are many different combinations of joint motions, and corresponding trajectories, that could move a limb from one location to another. Yet, somehow, the motor system constrains the degrees of freedom to arrive at a single, smoothly executed trajectory. The question of how this is accomplished is called the degrees of freedom problem (Rosenbaum, 2010). Some of the constraints used by the motor system are biomechanical, such as the range of motion of a particular joint or the stiffness of a particular muscle, whereas other constraints are imposed by cognitive processes that coordinate action.

Muscles are arranged in groups with opposite actions. One group (the agonist) engages in flexion, and its opposing group (the antagonist) engages in extension. Movement at the joints is controlled by the coupled actions of the agonists and the antagonists. A muscle contracts when it receives a signal from motor neurons. We will now discuss various mechanisms involved in the control of movement.

CONTROL OF MOVEMENT

The nervous system controls movements through a hierarchy of mechanisms. At the lowest levels, the elastic properties of the muscles themselves and motor neurons control muscle contraction. At higher levels, the central nervous system controls movement through the spinal cord. We will examine these different levels of control, beginning with the physical properties of muscles and bones, and motor neurons, and ending with the higher-level organization and execution of action mediated by the brain.

MASS-SPRING PROPERTY AND MOTOR UNIT

A convenient way to describe the behavior of the muscle is as a mass-spring system (Bernstein, 1967; de Lussanet, Smeets, & Brenner, 2002). The stretchy muscle can be thought of as a spring attached to the bone. Every spring has an equilibrium point, or resting length, the length to which it stretches when a mass is attached to it. When the spring is moved from its equilibrium point and then released, it will return to its resting length. Leg stiffness, a predictor of athletic performance and injury risk, can be modeled effectively with a simple mass-spring system (Brauner, Sterzing, Wulf, & Horstmann, 2014).

The resting amount of flexion and extension of the agonist and antagonist muscle groups at each joint, and hence the resting joint angle, is determined by the muscles’ equilibrium points. When some external force changes the joint angle, the muscles in one group will be stretched, and the muscles in the other will be compressed. When the force is removed, the muscles will return to their equilibrium points, and the joint angle will return to its original position. This is the most basic level of movement, in that it can occur in the absence of any neural signal. However, more complex movements (controlled by higher levels of the nervous system) may also rely on the mass-spring properties of muscles. In some situations, it may be useful to talk about movement, such as elbow flexions, as being accomplished by a systematic change in the stiffness, and hence the equilibrium point, of each muscle. The elastic properties of the muscles cause the position of the limb to change in accord with the changing equilibrium points.

Change in muscle stiffness, caused by contraction of the muscle, is a result of signals from motor neurons. Each muscle is composed of muscle fibers, which are innervated by hundreds of motor neurons. An individual motor neuron innervates many muscle fibers spread throughout a muscle; the neuron and fibers together are called a motor unit. When the motor neuron “fires,” all of the muscle fibers within the unit will be affected. The motor unit is considered to be the smallest unit of motor control. The contraction of a single muscle will be determined by which and how many motor units are activated at the same time (e.g., Gielen, van Bolhuis, & Vrijenhoek, 1998).

Although motor unit activity may seem far removed from ergonomic concerns, a study by Zennaro et al. (2004) illustrates that this is not the case. They measured surface electromyograms (EMGs; an overall measure of muscle activity) and the intramuscular motor unit activity of a shoulder muscle while participants tapped a key for 5 minutes on a desktop. The desktop was either an appropriate height (where the upper arm could hang relaxed, with the lower arm parallel to the desktop) or 5 cm above that height. EMG activity was greater for the too-high desktop than for the one at the appropriate height. This greater EMG was due mainly to individual motor units being active for longer periods of time, rather than to an increased number of active motor units. This study demonstrates that ergonomic issues can be evaluated at the motor unit level and that incorrectly adjusted office equipment can lead to prolonged activity of the motor units, which in turn may result in musculoskeletal disorders (see Chapter 16).

SPINAL CONTROL

The spinal cord controls certain actions by way of spinal reflexes (Abernethy et al., 2013; Bonnet, Decety, Jeannerod, & Requin, 1997). Such reflexes begin with stimulation of the sensory receptors that provide information about limb position (proprioception). Proprioceptive receptors are located within the muscles, tendons, joints, and skin. Their signals are sent to the spinal cord, where a motor signal is quickly evoked and sent to the appropriate muscles. Spinal reflexes allow movements to be made within milliseconds of the initiating stimulus. For example, when a sensory neuron receives a painful signal, the limb withdrawal reflex will cause the appropriate muscle to contract, removing the limb from the source of the pain. This happens very quickly because the signal does not have to travel all the way to the brain and then back again.

Spinal control of movement is not limited to the initiation of reflex responses. Both spinal reflexes and the higher central nervous system contribute to maintenance of posture (Tanaka, 2015). Gait and other movement patterns, though initiated by the brain, are controlled by the spinal cord once they are initiated (Grillner, 1975; Pearson & Gordon, 2000). There is even some evidence that the spinal cord can perform complex controlling operations (Schmidt & Lee, 2011). The sophisticated information-processing capabilities of the spinal cord free the brain so that it can engage in other activities.

CONTROL BY THE BRAIN

After the spinal cord, the next three structures critical to motor control are the brainstem, cerebellum, and basal ganglia (see Figure 14.1). The brainstem controls movements of the head and face, respiration and heart rate, and in part controls movement of the eyes.

FIGURE 14.1The cerebrum (a), basal ganglia (b), brain stem (c), and cerebellum (d).

The cerebellum is involved in several aspects of motor control (Rosenbaum, 2010). These include the maintenance of muscle tone and balance, the coordination and timing of rapid action sequences, and the planning and execution of movement. The cerebellum may help plan and initiate movements, but without input from sensory feedback (Bastian, 2006).

The basal ganglia are low-level brain structures that are also involved in the planning of movements. Evidence suggests that the basal ganglia form an action-selection circuit that helps choose between actions that use common motor pathways (Humphries, Stewart, & Gurney, 2006). The basal ganglia also control the size or amplitude of movements and integrate perceptual and motor information. They control slow, smooth movements, such as those required for postural adjustments and those requiring the continual application of force. Motor learning requires both the cerebellum and the basal ganglia (Doya, 2000).

After the brainstem, cerebellum, and basal ganglia, the highest levels of motor control are located in the cortex. The motor, premotor, and supplementary motor cortices are located in adjacent regions in the rear part of the frontal lobes (see Figure 14.2). The relationships between these different motor areas are very complicated and not yet well understood (Pockett, 2006), although we are learning more about them every day (Sosnik et al., 2014). All the areas communicate with each other to greater or lesser degrees.

FIGURE 14.2The motor (^), premotor (°), and supplementary motor (·) cortices.

We know that the motor cortex is structured as a topographic map that takes the form of a “homunculus,” or little man (see Figure 14.3). It is involved in the initiation of voluntary movements. The premotor cortex controls movement of the trunk and shoulders. It is also implicated in the integration of visual and motor information, and we believe that it plays a role in turning the body to prepare for forthcoming movements. The supplementary motor cortex is involved in the planning and execution of skilled movement sequences. It is different from the premotor cortex in that its activity does not seem to depend on perceptual information.

FIGURE 14.3The homunculus on the motor cortex.

CONTROL OF ACTION

Motor performance is controlled by cognitive processes. Our understanding of these processes is based primarily on measures of human performance under a variety of conditions. In particular, we want to understand how movements are selected and controlled, and how perception and action are related (Rosenbaum, 2005).

Movements are made to accomplish a task. Different kinds of tasks will require different kinds of movements, which in turn will require different cognitive processes. Some tasks, such as throwing a ball or pressing a key, are discrete, because the action has a distinct beginning and ending. In contrast, tasks such as steering a car are continuous. Some tasks, such as those that might be performed on an assembly line, involve a series of discrete actions and fall in between. To understand the processes involved in motor control, we need to understand the demands placed on the cognitive system, which may not be the same across these different kinds of tasks.

We also need to distinguish between open and closed motor skills (Poulton, 1957). Open skills are performed in dynamic environments, and the speed and timing of the movements tend to be determined by events occurring in the environment. Closed skills are performed in static environments and are self-paced. Football and basketball are sports that require mainly open skills, whereas gymnastics and track and field require primarily closed skills. Because open skills are paced by the environment, they require rapid adaptations to the environment, which closed skills do not. When athletes are asked to perform tasks unrelated to their sport, athletes in open-skills sports are more influenced by environmental factors than are those in closed-skills sports (Liu, 2003).

The type of task, whether discrete or continuous, and the type of skill, whether open or closed, will interact to determine how movements will be controlled. Movement control is referred to as open- or closed-loop, terms that refer to the extent to which feedback is used in the performance of the action (Heath, Rival, Westwood, & Neely, 2005). Do not confuse the kind of control with open and closed skills, because open skills actually use closed- more than open-loop control, and closed skills use open- more than closed-loop control.

CLOSED-LOOP CONTROL

Recall from Chapter 3 that closed-loop systems are characterized by the use of negative feedback to regulate their outputs. To apply this idea to motor control, we assume that a person knows what movement needs to be made and generates a mental representation of the desired movement (see Figure 14.4). During the action, sensory feedback produced as the movement is executed is compared with this representation. A person will perceive as error any difference between the actual and desired positions of her limb. The person will then use this error to select corrective movements to bring her actual limb position closer to the desired one. She continuously makes comparisons between her actual and desired positions until the difference between them is minimized.

FIGURE 14.4A closed-loop system for motor control.

Closed-loop control depends on the sensory feedback produced as a person acts on the environment. What sources of feedback are available? As Smetacek and Mechsner (2004) state, “Our daily doings are coordinated and run by a trinity of independent sensory systems: proprioception, vision, and the vestibular organs of the inner ear” (p. 21). Earlier in this chapter we mentioned the role of proprioception in reflexes, but Smetacek and Mechsner emphasize that “all purposeful movements, both conscious and unconscious, are controlled by proprioception” (p. 21). As a person moves his hand toward an object, visual feedback provides information about the locations of both his hand and the object that can be used to take corrective actions. The vestibular sense provides feedback about his posture and head orientation. Feedback is also provided by touch (Niederberger & Gerber, 1999) and audition (Winstein & Schmidt, 1989). For example, a person walking down carpeted stairs can determine when the uncarpeted floor has been reached by the feedback conveyed by changes in both touch and sound.

OPEN-LOOP CONTROL

In contrast to closed-loop control, open-loop control does not depend on feedback (see Figure 14.5). Movements subject to open-loop control are either too rapid to allow modification from feedback or overlearned. Open-loop control is achieved by developing a sequence of movement instructions generated by a general mental representation called a motor program. The selection of a motor program and generation of a movement sequence occur prior to the initiation of the movement.

FIGURE 14.5An open-loop system for motor control.

The idea of a motor program was first formalized by Keele (1968) and later developed and extended by Schmidt (1975). A motor program is an abstract plan for controlling a specific class of movements. To execute a particular movement within a class, parameters such as the muscles to be used, their order, and the force, duration, and timing of their contractions must be specified. In the rest of this section, we present certain implications of the motor program idea and describe the invariant characteristics of a motor program, its modular, hierarchical structure, and the role of feedback.

IMPLICATIONS

The motor program concept has several implications. First, because the motor program contains a general template, or schema, of the desired movements, coordinated movement should be possible when feedback is not available. Studies have demonstrated that monkeys and humans who cannot process proprioceptive feedback (because of surgical intervention or disease) can still perform some skilled actions, such as grasping, walking, and running (Bizzi & Abend, 1983; Rothwell et al., 1982; Taub & Berman, 1968). These findings are consistent with the motor program concept.

The motor program concept also implies that rapid movements can be made accurately even when the transmission and processing of sensory feedback take more time to accomplish than the movement itself. These kinds of movements are made during keyboarding, in which keystrokes are made very rapidly (Lashley, 1951; see Box 14.1). Based on the speed of the keystrokes, it makes sense to assume that the keystroke movements are programmed in advance. Another implication of the motor program concept is that it should take longer to program more complex movements. It is true that response times to initiate a movement are longer for more complex movements (Henry & Rogers, 1960; Klapp, 1977). Thus, the speed with which a person can react to a stimulus is directly related to the complexity of the movements that must follow.

INVARIANT CHARACTERISTICS

Motor programs have invariant characteristics that specify critical structural aspects of the class of movements that a particular program controls. The invariant characteristics are independent of the particular muscles used to execute the movement. These characteristics may include the order of movement components, the relative amount of time that each component takes, and the way in which force is distributed to each component.

Evidence for invariant characteristics can be seen in writing samples produced by different muscle groups (Merton, 1972; see Figure 14.6). For example, you can see similarities between a writing sample produced by arm and shoulder muscles (as on a blackboard) and a writing sample produced by hand and finger muscles (as on a piece of paper), as long as both samples are produced with the dominant hand. Writing produced with the nondominant hand is also similar to writing produced with the dominant hand, suggesting that many aspects of the motor program for writing are independent of the effectors used for execution (Lindemann & Wright, 1998).

FIGURE 14.6Writing samples produced using the hand (on a piece of paper (top)) and the arm (at a blackboard (bottom)). We thank Bob Hines for contributing these samples.

MODULAR ORGANIZATION

Considerable evidence suggests that motor programs are composed of several modules that reflect the invariant characteristics of a movement. For example, timing of movement seems to involve an independent module or component of control (Keele, Cohen, & Ivry, 1990). The evidence for this comes from studies in which people are asked to produce timed movements with different parts of their bodies. Not everyone can time their movements accurately, and these differences across people show up as differences in the variance of movement times. People who are less accurate have larger movement time variances. Moreover, the extent of variation for a person will be similar regardless of the effector he or she uses to make the movement.

This correlation between variances across parts of the body occurs mostly in tasks for which movement timing is especially salient, such as tapping in beat to a metronome (Ivry, Spencer, Zelaznik, & Diedrichsen, 2002; Zelaznik et al., 2005). Other tasks may also put a strong premium on movement time, but the movement trajectory (or some other feature of the movement) may be more salient. For example, some researchers have asked people to draw circles at a timed rate (e.g., Ivry et al., 2002). The variability of movement time in these tasks is not related to the variability of movement time for tapping tasks. Thus, when movement timing is not as salient as the trajectory of the movement, the timing of the movements seems to arise from the trajectory control process itself.

HIERARCHICAL ARRANGEMENT

Timing and trajectory control are examples of high-level modules of a motor program. Motor programs are hierarchical, in that high-level modules pass control to lower-level modules. Evidence for hierarchical control comes from studies of tapping. For example, one study asked people to tap their fingers in response to sequences of six numbers (Povel & Collard, 1982). The sequence 123321 was used to indicate the order for three fingers, labeled 1, 2, and 3, respectively. Regardless of the fingers used, the first and fourth response latencies were longer than the others (see Figure 14.7). This finding suggests that a hierarchical program was executed to perform the sequence, with the top level passing control to the first response subgroup (123) and then to the second response subgroup (321). The longer response latencies for the first and fourth taps reflect the time required to shift control to the next subgroup.

FIGURE 14.7Mean latency of intertap intervals for the sequences 123321 (a), 332112 (b), and 233211 (c).

ROLE OF FEEDBACK

Although the motor program idea allows movement to be executed without feedback, feedback is still assumed to contribute to the control of action in several ways. First, information fed back through the senses specifies the position of the body part that is to be moved and where it is to be moved. Without this information, it would not be possible to select the appropriate parameters for the motor program. Second, for slow movements, feedback can be used to correct the parameters of an ongoing program, or a new program can be selected, as appropriate. Third, rapid corrections of movements based on proprioceptive and visual feedback can occur in less than 100 ms in some situations (Saunders & Knill, 2004).

AIMED MOVEMENTS

Aimed movements are those that require an arm or some other part of the body to be moved to a target location. The movement made by an operator to bring a finger to a pushbutton is an example, as is moving the foot from the accelerator pedal to the brake pedal in an automobile. The speed and accuracy with which such movements can be made are influenced by many factors, including the effector used, the distance of the movement, and the presence or absence of visual feedback. To ensure that an operator’s movements will be made within the necessary speed and accuracy limits, a designer must consider the way in which aimed movements are controlled.

Aimed movements were first studied by Woodworth (1899), who was interested in the amount of time needed to use visual feedback. To investigate this issue, Woodworth used a task in which people repeatedly drew lines on a roll of paper moving through a vertical slot in a table top. The lines were to be of a specified length, with the rate of movement set by a metronome that beat from 20 to 200 times each minute. One complete movement cycle (up and down) was to be made for each beat. The role of visual feedback was evaluated by having people perform with their eyes open or closed. Movement accuracy was similar for the two conditions at rates of 140 cycles/min or greater; thus, visual feedback had little or no effect on performance. However, at rates of less than 140 cycles/min or less, the eyes-open condition yielded better performance. This result suggests that the minimum time required to process visual feedback is longer than (60 s/140 cycles) × 1000 ms/s = 428.6 ms/cycle (although modern estimates of this time are much shorter; see below).

To explain these and other results, Woodworth (1899) proposed that rapid aimed movements have two phases, which he called initial adjustment and current control, and which correspond to open- and closed-loop modes, respectively. The initial phase transports the body part toward the target location. The second phase uses sensory feedback to correct errors in accuracy and home in on the target. Elliott, Helsen, and Chua (2001) noted, in a review of the impact of Woodworth’s research on the study of goal-directed arm movements, “his theoretical contribution to the understanding of speed-accuracy relations and the control of rapid goal-directed movements that has stood the test of time” (p. 354).

Not much research was conducted during the first half of the 20th century to follow up Woodworth’s (1899) work. However, since the 1950s, aimed movements have been examined extensively. In addition to the time-matching task used by Woodworth, a time-minimization task has been widely used (Meyer, Smith, Kornblum, Abrams, & Wright, 1990), in which people are to minimize their movement times while approximating a specified target accuracy value. For both tasks, single movements as well as repetitive movements have been examined.

Fitts’s Law

Fitts (1954) established a fundamental relation between aimed movement time and the variables of distance and precision that has come to be known as Fitts’s law. He examined performance in a repetitive tapping task, in which a person was required to move a stylus back and forth between two target locations as quickly as possible. As the distance between the targets increased, movement time increased. Conversely, as the widths of the targets increased, movement time decreased. Because the width of the targets dictates the precision of the movements, this relation can be viewed as a speed–accuracy tradeoff (see Chapter 13).

From these relations, Fitts (1954) defined the index of difficulty (I) for an aimed movement as

$$I={\text{log}}_{\text{2}}\left(\text{2}D/W\right),$$

where:
D	is the center-to-center distance between the targets, and
W	is the width of the targets.

Fitts found this index to be related to movement time (MT) by the linear equation

$$\text{MT}=a+bI,$$

where a and b are constants (see Figure 14.8).

FIGURE 14.8Movement time as a function of the index of difficulty.

According to Fitts’s law, when the distance required for a movement is doubled, the movement time will not change if the target width is also doubled.

Fitts’s law applies across a wide range of tasks. Fitts (1954) obtained similar results with tasks that required washers to be placed onto pegs or pins into holes. Other researchers have found the law to hold in tasks as diverse as angular positioning of the wrist (Crossman & Goodeve, 1963/1983), arm extension (Kerr & Langolf, 1977), positioning a cursor with a joystick or a head-movement controller (Jagacinski & Monk, 1985), working with tweezers under a microscope (Langolf & Hancock, 1975), and making aimed movements underwater (Kerr, 1973). Fitts’s law is also important in the study of human–computer interaction, where many tasks require moving a cursor to a target location (Seow, 2005), and for interacting with a mobile device such as a tablet or a smartphone (Alexander, Schlick, Sievert, & Leyk, 2008).

In 2004, a special issue of the International Journal of Human-Computer Studies was devoted to the 50th anniversary of Fitts’s original study. In that issue, the editors emphasized, “Fitts’ law has … [made] it possible to predict reliably the minimum time for a person in a pointing task to reach a specified target” (Guiard & Beaudouin-Lafon, 2004, p. 747). Newell (1990) characterized the generality of Fitts’s law as follows: “Fitts’ law is extremely robust …. Indeed, it is so neat, it is surprising that it doesn’t show up in every first-year textbook as a paradigm example of a psychological quantitative law” (p. 3).

There have been many explanations of Fitts’s law since 1954. The most widely accepted explanation is the optimized initial impulse model (Meyer, Abrams, Kornblum, Wright, & Smith, 1988). An aimed movement toward a specified target location is presumed to involve a primary submovement and an optional secondary submovement that is made if the initial submovement is “off target.” The submovements are programmed to minimize the average time for the total movement.

The idea that aimed movements are controlled by executing a single corrective submovement has become a central idea in motor control (Hoffmann, 2016). The evidence supporting this idea includes the facts that movements rarely show more than two submovements and that the time to execute these submovements is constrained by Fitts’s law. It is interesting to note that the optimized impulse model is consistent with Woodworth’s (1899) original proposal that movements are controlled in two phases.

Application

Fitts’s law is used to evaluate the efficiency of different movements in a wide variety of real-world situations. Efficiency is measured as the slope (b) of the function relating movement time to the index of difficulty. This measure of efficiency can be used to evaluate different workspace designs. For example, Wiker, Langolf, and Chaffin (1989) noted that many manual assembly tasks require people to use hand tools raised above their shoulders. Wiker and his colleagues examined people’s ability to make repetitive movements of a stylus to a hole with their hands raised to different positions (−15 to +60° relative to shoulder level; see Figure 14.9). Movement times were longer (by 20%) when people performed the task at the highest position (compared with the lowest). Wiker and colleagues attributed the longer movement times to the increased tension in the muscles needed to raise the hand. They recommended that sustained manual activity be restricted to below shoulder level when possible.

FIGURE 14.9Task postures tested in an experiment by Wiker et al. (1989).

Another example involves assistive technology devices such as chin, head, and mouth sticks, which are used by people with limited mobility to press keys on a computer keyboard. Andres and Hartung (1989) asked people to tap between targets of varying width and separation with a chin stick. Fitts’s law still held, but the mean information transmission rate was 7 bits/s. This value is considerably less than what we usually see for hand or foot movements. While part of the reason for the lower transmission rate is due to how neck and shoulder muscles are controlled, part of the problem is in the design of the stick.

Baird, Hoffmann, and Drury (2002) asked people to perform aimed movements with hand-held pointers of different lengths. The longer the pointer, the longer the movement time was for the same index of difficulty. The ends of longer pointers “wiggle” more, because the pointer amplifies small muscle tremors. This wiggling makes it more difficult for people to bring the end of the pointer to a target. This finding suggests that, for assistive technology devices like sticks and for other hand-held tools like screwdrivers and soldering irons, the shorter the length of the tool, the easier it will be to bring the tip of the tool to the object of interest. When the tool cannot be shortened, then the size of the object (the target area) must be increased to compensate.

Visual Feedback

Another issue in the control of aimed movements is the role of visual feedback. Remember that Woodworth (1899) estimated the time to process visual feedback to be 450 ms, because people benefited from having their eyes open only for slow movements. This estimate seems quite long, because we know that people can make accurate choices based on visual information in half that time.

Since Woodworth’s (1899) original experiment, this question has been revisited several times (Keele & Posner, 1968; Zelaznik, Hawkins, & Kisselburgh, 1983). We now know that the time to process visual feedback depends on the type of task to be performed and whether or not people know in advance that feedback will be available. For example, Zelaznik and his colleagues asked people to make timed, aimed movements toward a target, and on some trials they turned the lights off at the beginning of the movement. When people knew that the lights were going to be turned off, they were able to accurately aim their movements with as little as 100 ms preparation time. In a similar study, when the target of an aimed movement jumped to a different position during the movement, people were able to resolve the mismatch between the movement and the new perceived target position within 100 ms (Dimitriou, Wolpert, & Franklin, 2013). The fact that the time for visual processing in aimed movements can be quite short means that even very rapid movements may be more accurate if visual feedback is available.

Bimanual Control

The experiments investigating manual control that we have discussed up to this point have allowed people to make only single aimed hand movements. Natural body movements usually involve coordination of several limbs. Some tasks, such as light assembly, require people to make two different aimed movements at once, one with each hand. It is not hard to think of situations in which these two different movements would have different indices of difficulty. However, Fitts’s law does not apply to each limb separately.

Kelso, Southard, and Goodman (1979) asked people to perform two movements simultaneously, for which the indices of difficulty differed, one with each hand. The right hand moved to a close, large target (a movement with a low index of difficulty) and the left hand to a distant, small target (a movement with a high index of difficulty). If the two hands moved independently, the right-hand movement should have taken less time than the left-hand movement, because the index of difficulty was lower. Instead, people moved their hands at different speeds, so that both lifted off and reached the targets simultaneously. People also accelerated and decelerated the two movements at the same times. Overall, the time to make both movements was approximately equal to the time to make the more difficult movement when it was performed alone. Thus, the easier movement was coupled to the harder movement.

We can explain why movements are coupled like this by referring back to the motor program. If the program is responsible for executing classes of similar movements, then both movements may require the same program (Schmidt et al., 1979). Movement characteristics like velocity and distance to be traveled can be determined separately for each limb. If similar hand movements are controlled by the same motor program, then we would expect that a hurdle placed in the path of one limb would produce a change in the movement paths for both limbs, and this is what we see (see Figure 14.10; Kelso, Putnam, & Goodman, 1983). Also, with practice, people can learn to move two limbs independently. Schmidt et al. (1998) proposed that, in this case, the motor program is modified so that it allows for different trajectories for the limbs involved in the coordinated movement.

FIGURE 14.10Bimanual movement without (a) and with (b) a hurdle.

These experiments show that the index of difficulty does not have to be equal for different controls (e.g., pushbuttons) when designing a control panel. When these controls must be activated simultaneously, the operator will be able to coordinate his or her movements without too much difficulty. However, although bimanual movements seem to be controlled by a single motor program, they cannot be completely synchronized. Warrick and Turner (1963) had people hold down keys with their left and right index fingers, which were to be released simultaneously at any time after a “ready” signal was presented. The average difference between the release times of the two fingers was zero, so people were able to release their fingers virtually simultaneously. However, there was a considerable amount of variability in the difference between the release times, so that one finger could precede the other by as much as 20 ms. For more complex tasks, Warrick and Turner worried that these differences could be even larger, which could create potential problems for the operation of equipment or machinery requiring near-simultaneous activation of left- and right-hand controls.

GRASPING AND INTERCEPTING OBJECTS

Grasping is a fundamental component of actions as diverse as picking up a cup of coffee, grabbing a hammer, opening a door, or flipping a light switch. Movements that culminate in grasping an object can be broken into two components: a transport phase (reaching) and a grip formation phase (grasping). The transport component is very similar to the aimed movements we have been discussing and involves moving the hand to the object. The grasp component involves positioning the fingers for grabbing the object. The fingers of the hand gradually open to a maximum aperture and then close until the grip conforms to the object size (see Figure 14.11).

FIGURE 14.11The grasping phase of movement.

Researchers have been interested in the factors that influence the transport and grasp components, and how transport affects grasp, and vice versa. Most experiments have looked at movements when the object to be grasped changes size or location (e.g., Castiello, Bennett, & Stelmach, 1993; Paulignan, MacKenzie, Marteniuk, & Jeannerod, 1990). When the object is moved farther away, the transport component takes longer, but the grip component does not change. Regardless of the transport duration, the grip aperture reaches a maximum within the first 60%–70% of the movement (Jeannerod, 1981, 1984). However, the transport and grip components are not independent of each other. When an object unexpectedly changes shape or location after the reach begins, modifications of both the transport and grip components often occur together, suggesting that the components are coupled to each other (Rosenbaum, Meulenbroek, Vaughan, & Jansen, 2001; Schmidt & Lee, 2011).

While most experiments on reach have used stationary objects, people often need to grasp objects that are moving. For a moving object, not only do people need to reach toward the object in the same way as they would if it were stationary, but they also need to time their movements so that they grasp the object at an appropriate point in its trajectory. This may require movements not only of the arm and hand but also of the entire body. For example, an outfielder can catch a fly ball only after he or she estimates the ball’s flight path (e.g., Whiting, 1969). The player must calculate the best location at which to intercept the ball, execute the movements to reach that location, and then execute the appropriately timed grasping movements to make the catch.

A variable important for determining where a moving object can be intercepted is based on how quickly the object’s retinal image is growing. The inverse of the speed of growth, called tau, determines the time to contact with the object (Lee, 1976). If the image is growing very quickly, tau will be small, and “time-to-contact” will be short. If the image is growing very slowly, tau will be large, and time-to-contact will be long. When people are asked to estimate time to contact, tau is important, but so are other factors such as pictorial depth cues (Hecht & Salvesbergh, 2004; Lee, 1976).

OTHER ASPECTS OF MOTOR CONTROL

Human movements can be very simple, but are usually amazingly intricate and complex. We cannot do justice to all of the many aspects of motor control that are relevant to issues in human factors. However, before we leave this topic, there are three important aspects of motor control that we need to discuss briefly: posture, locomotion, and eye and head movements.

Posture

Posture and balance control takes place mostly in the spinal cord and is maintained through closed-loop control. Adjustments to posture are made on the basis of information provided by the proprioceptive, vestibular, and visual senses. Some of the parameters that are controlled by the feedback loop include force, velocity, and distance of corrective movements. It is interesting that these parameters will eventually be modified under low-gravity conditions. After extended space flight, astronauts often have difficulty maintaining posture and balance (Cohen et al., 2012), apparently because their spinal neurons adapt to low gravity and are slow to re-adapt to normal gravity (Lackner, 1990).

Locomotion

Most people spend a great deal of time walking, or locomoting, through their environment. Locomotion occurs in a four-phase step cycle, shown in Figure 14.12. Like posture and balance, the step cycle is controlled by the spinal cord. Like posture and balance, the step cycle can be modified by information from the proprioceptive, vestibular, and visual systems. Visual feedback serves two important purposes in locomotion (Corlett, 1992). People use visual cues to plan routes from their current position to a desired location, and they use these cues during locomotion to initiate each step, which is then executed ballistically (or without modification from visual feedback; Matthis, Barton, & Fajen, 2015).

FIGURE 14.12The step cycle.

Eye and Head Movements

We have emphasized the importance of visual feedback in maintaining posture and balance and in locomotion. Delivering this information to motor control centers requires frequent and extensive eye and head movements to ensure a complete picture of the environment. For example, consider the eye movements a person must make to track a moving target. His smooth pursuit eye movements must match the velocity of the eye with that of the target, and if the target moves a significant distance, his head must move as well.

Eye movements are coordinated with head movements through the vestibulo-ocular reflex, which is triggered by rotation of the head or body while looking at an object. The eye will move in the direction opposite to the head, and so compensate for any change in the visual image caused by the head movement. This compensation is not very accurate when the head is turning very quickly (Pulaski, Zee, & Robinson, 1981). Furthermore, when a person fixates on a target image that rotates with the head, like the images in a helmet-mounted display, the vestibulo-ocular reflex is suppressed. Because the reflex is suppressed, the person will no longer be able to track objects in the environment well.

MOTOR LEARNING

Part of understanding how actions are controlled involves understanding how people learn to make complex movements. Some of the questions addressed by researchers in this area include how movements are represented and retained in memory, what role feedback plays in the acquisition of motor skill, what kinds of practice and feedback result in optimal learning, and how motor skill relates to other skills. The answers to these questions have implications for the structuring of training programs and design of equipment for use in the workplace (Druckman & Swets, 1988; Druckman & Bjork, 1991; Schmidt & Bjork, 1992).

A lot of contemporary research in motor learning has been inspired by Schmidt’s (1975) schema theory. At the heart of schema theory is the concept of the motor program. Recall that a motor program is an abstract plan for controlling a specific class of movements. Accurate performance requires not only that the appropriate motor program be selected, but that parameters such as force and timing be specified correctly. Two kinds of motor schemas act to determine these parameter values: recall and recognition schemas.

When a movement is to be made, the initial conditions (where a person is now) and outcome goal (where a person wants to be) are used by a recall schema to select the response parameters for a motor program. A recognition schema specifies the expected sensory consequences of the movement. The recall and recognition schemas are used in different ways for fast and slow movements. For fast movements, the recall schema both initiates and controls the movement. Then, after the movement is completed, perceived consequences can be compared against those expected from the recognition schema. Any mismatch between the two is used as the basis for correcting the recall schema. The recall schema initiates slow movements as well. However, comparison between sensory feedback and the sensation predicted by the recognition schema can occur during the movement, and correction to the movement parameters can be made as soon as an error is detected.

The schema theory assigns a prominent role to sensory feedback, incorporates motor programs, has two memory components (one involved in movement initiation and the other in evaluating feedback), considers feedback to be important during learning, and provides a way for error to be detected during or after execution of the movement. Though we know now that schema theory is not completely accurate (Shea & Wulf, 2005), these general features are central to contemporary views of motor-skill acquisition.

CONDITIONS AND SCHEDULES OF PRACTICE AND TRAINING

The term practice refers to repeated execution of a task with the goal of attaining mastery of that task. How a person practices a motor skill determines how quickly he or she will attain mastery, how long he or she will remember the skill, and the extent to which the skill will result in improved performance for other tasks. There are many ways that specific motor skills can be taught. Different methods of practice can be viewed as different kinds of training programs. Not all training programs are equally effective. Consequently, a lot of effort has been devoted to investigating different training programs for different kinds of motor skills. In particular, most training programs are designed with the goal of optimizing performance with a minimal amount of training time.

Training programs often are evaluated by the amount of practice required to attain a criterion level of performance and not how long that level can be retained. The level of performance that a person reaches during training can be influenced by many factors and does not always indicate the amount that the person has learned. Therefore, it is important to distinguish between variables that affect learning and cause a relatively permanent change in behavior and those that only temporarily affect performance. For example, a person’s performance may deteriorate near the end of a long, difficult practice session due to fatigue, but that person’s performance may be much improved once he or she is no longer fatigued.

The extent of learning can be demonstrated by measuring performance after a delay following training, a procedure referred to as a retention test. More effective training programs will result in better performance after such a delay. Another way of measuring learning is to look at a person’s ability to perform new tasks that are related to, but different from, the tasks learned in the training program, a procedure referred to as a transfer task. In the rest of this section, we will focus on how different practice conditions contribute to retention and transfer.

AMOUNT OF PRACTICE

Usually, retention will increase with more practice. Even after a performer has attained an acceptable level of skill, if she continues to practice (or “overlearns” the skill), she may retain the skill better. One experiment examining this issue required soldiers to disassemble and assemble an M60 machine gun (Schendel & Hagman, 1982). Three groups of soldiers had to disassemble and assemble the gun until they had made no errors. One group, the control, received no further training and was retested 8 weeks later. The two other groups received overtraining; soldiers in these groups performed additional assemblies equal to the number of assemblies they had performed before the first errorless execution. For one group, the overtraining was performed on the same day as the initial training, whereas for the other, it was performed 4 weeks later, halfway between the initial training and the retention test. Both experimental groups had greater retention than the control group. Although overtraining may seem excessive, it is an effective way to ensure that skills will be remembered.

Another experiment asked college basketball players to perform free throws (a distance of 15 ft) and shots at six other distances (three closer and three farther; Keetch, Schmidt, Lee, & Young, 2005). The players made 81% of their shots from the free throw line, a percentage that was much higher than their performance at any other distance. One explanation for this is that, because basketball players have more practice at free throws than at any other kind of shot, free throws are overlearned. Note that it is difficult to explain this result with schema theory, because the movements are all from the same class, differing only in their distance (Keetch et al., 2005).

FATIGUE AND PRACTICE

Extended physical activity results in fatigue. Consequently, we must ask whether or not practice is as effective when the learner is fatigued as when he is rested. Fatigue has a detrimental effect on the speed with which a person can acquire a new motor skill, at least for some laboratory tasks (Pack, Cotton, & Biasiotto, 1974; Williams & Singer, 1975). However, skills acquired under conditions of fatigue are retained almost as well as those acquired under conditions of non-fatigue (Cotten, Thomas, Spieth, & Biasiotto, 1972; Heitman, Stockton, & Lambert, 1987).

As one example, one study asked people to perform a task that required them to rotate a handle in a clockwise direction and then rotate it in a counterclockwise direction, and finally to knock down a wooden barrier (Godwin & Schmidt, 1971). On the first day, people in a fatigued group had to turn a stiff crank for 2 minutes before each performance of the task. People in a non-fatigued group only tapped their fingers. The fatigued group performed the task more slowly than the non-fatigued group (see Figure 14.13). However, after a three-day rest, there was only a small difference in performance between the two groups. This difference was eliminated by the fourth retention trial, indicating that fatigue had little effect on learning.

FIGURE 14.13Effect of fatigue on initial performance and learning.

The effects of fatigue on learning are greatest at highest levels of fatigue. One study asked people to perform a ladder-climbing task that required them to maintain their balance (Pack et al., 1974). Three levels of fatigue were induced in three groups of people by having them perform strenuous exercise that maintained their heart rates at 120, 150, or 180 beats/s between trials of the task. People who maintained their heart rates at the two highest levels did not perform as well on retention trials as people who maintained their heart rates at the lowest level, who showed no difference from people who did not perform any strenuous exercise.

DISTRIBUTION OF PRACTICE

Distribution of practice refers to the influence that scheduling of practice sessions and work periods has on the acquisition of motor skill. There are two kinds of practice sessions: massed or distributed. With massed practice, a person practices the same task repeatedly for an extended period of time, whereas with distributed practice, the person rests occasionally between trials. When we compare performance for massed versus distributed practice, we will look at conditions in which the number of practice trials is the same whether massed or distributed. Under distributed practice, the trials are performed over a longer period of time than under massed practice, where the trials are performed all at once.

Massed practice can result in a much slower rate of skill acquisition than distributed practice. Lorge (1930) asked his subjects to trace a star by watching their movements in a mirror. People who used distributed practice performed the tracing task better than people who used massed practice. Then, Lorge shifted some people from distributed to massed practice, and their performance dropped to the level of the people who had used massed practice all along. This suggests that the difference in performance under massed and distributed practice is only temporary.

The extent to which the type of practice schedule influences retention is unclear. Several reviews have suggested that massed versus distributed practice will influence mainly tasks that require continuous movements. These reviews showed that massed practice hurts retention for these kinds of tasks, and that distributed practice improves retention (Donovan & Radosevich, 1999; Lee & Genovese, 1988). Distributed practice of continuous tasks may be more beneficial when the sessions are separated by a day rather than shorter periods within the same day (Shea, Lai, Black, & Park, 2000).

A few studies have examined distribution of practice effect using tasks with discrete movements. One asked people to pick a dowel up out of a hole, flip the ends of the dowel, and put it back in the hole (Carron, 1969). In contrast to what happens in continuous tasks, massed practice produced slightly better retention than distributed practice. This result was replicated when a stylus was to be moved from one metal plate to another in 500 ms (Lee & Genovese, 1989). However, Dail and Christina (2004) found that people acquired and retained golf-putting ability better under distributed practice. This suggests that there is no simple relation between task type (discrete or continuous) and whether massed or distributed practice produces better learning. Some evidence suggests that training schedules incorporating both massed and distributed practice may be best for some perceptual-motor tasks (Paik & Ritter, 2016).

VARIABILITY OF PRACTICE

Variability of practice refers to the extent that the movements required for each practice trial differ. When a person executes the same movement on each trial, there is little variability, but when he or she performs different movements on each trial, there is greater variability. According to schema theory, variable practice should lead to better performance than practice with only a single movement. This is because variable practice will produce a more detailed recall schema that can be used when a new variation of the movement is encountered.

Variability of practice has its greatest influence on transfer tasks, when people are asked to apply their skills to a task for which they haven’t practiced. People who receive variable practice perform better on transfer trials than those who do not. For example, one study asked people to perform a two-part timed movement in which they knocked down two barriers (Lee, Magill, & Weeks, 1985). They were told how long each part of the movement should take. One group of people practiced the movements under four different time requirements (the random practice group), and another group practiced the movements under a single time requirement (the constant practice group). Each group performed exactly the same number of trials and then was shifted to a new, unfamiliar time requirement. People in the random practice group were better able to meet the new time requirement than people in the constant practice group (see Figure 14.14). A third group of people received variable practice, but time requirements were blocked so that people performed all of the trials with one set of time requirements before moving on to the next block of trials (the blocked practice group). The blocked practice group showed no better performance on the transfer trials than the constant practice group (see Figure 14.14).

FIGURE 14.14Accuracy of movement time for constant, blocked, and random practice.

Schema theory can explain why variable practice is helpful mainly for random practice if we consider the concept of contextual interference introduced by Battig (1979) to explain verbal learning performance (Sherwood & Lee, 2003). This concept refers to disruption of short-term memory, and thus performance, as a consequence of practicing multiple-task variations within the same practice session. Contextual interference will be least when successive movements are identical and greatest when they are different. When contextual interference is minimized, a movement can be performed with relative ease, even retained well, but the resulting schema may not be very detailed or accurate. The schema may contain only the requirements for a single movement because the movement parameters needed for earlier blocks of trials have been “written over.” Conversely, when contextual interference is maximized by using random practice, movements are more difficult to learn, but the resulting schema contains the parameters for all of the movements required.

The more elaborate and flexible schema arising from random practice has its greatest benefits in performance of transfer trials. These benefits imply a range of practical considerations, such as dealing with problems that arise with aging. Older adults show the same improvements in performance on transfer trials with random practice schedules as do younger adults, suggesting that appropriate design of training schedules is an important consideration for offsetting declining motor skills in older adult populations (Pauwels, Vancleef, Swinnen, & Beets, 2015).

Most accounts of the benefits of random variable practice on motor skill acquisition emphasize how cognitive factors contribute to learning. However, motivational factors are also important (Holladay & Quinones, 2003; Wulf & Lewthwaite, 2016). Consider the concept of self-efficacy, which refers to judgments about one’s capabilities for performing various tasks. Random variable practice conditions, which lead to better learning, also lead to higher generality of self-efficacy across a range of transfer conditions. The implication is that a person may be more highly motivated to try to do a transfer task well if their assessment is that they are capable of performing the task than if it is that they are not.

Traditionally, repetitive drill-type training has been used to teach motor skills. For example, the Web Institute for Teachers’ (2002) instructions for teaching keyboarding skills to elementary-school students state, “Teachers must provide repetitive drill for developing skill.” Although repetitive training works, the studies we have reviewed in this section suggest that learning and retention of the skills can be improved significantly by varying the routine on a trial-to-trial basis.

MENTAL PRACTICE

Mental practice is the term used to describe mentally imaging the execution of a desired action for performing a task. Athletes routinely engage in mental practice before performing a difficult routine or motor sequence. Despite how common this practice is, there is some question about the extent to which it actually improves performance or facilitates the acquisition of a skill.

To answer this question, experimenters compare the performance of groups of people who have acquired a skill with and without the assistance of mental practice. For example, one group of people would physically practice a task, a second group would mentally practice the task for the same amount of time, a third group would perform both physical and mental practice, and a control group would receive no practice at all (Druckman & Swets, 1988). Mental practice usually results in better performance than no practice at all (e.g., Driskell, Copper, & Moran, 1994; Feltz & Landers, 1983; Wulf, Horstmann, & Choi, 1995). However, for transfer tasks, where we would expect that learning one skill would make it hard for people to perform a new skill, mental practice does not seem to hurt performance as much as physical practice (Shanks & Cameron, 2000). This means that while mental practice can be helpful, it is not the same as performing physical practice.

One benefit of mental practice is to allow rehearsal of the cognitive components of the practiced task (Sackett, 1934). For example, a tennis player must not only execute her backhand flawlessly, but she must also be able to anticipate her opponent’s return and position herself appropriately. This means that mental practice should be more effective for motor tasks that have a large cognitive component (e.g., card sorting) than for those that do not (e.g., repetitive tapping).

A study of sequence learning confirmed this hypothesis (Wohldmann, Healy, & Bourne, 2007). People in this study either practiced rapidly typing four-digit strings or only imagined typing the same strings instead of actually typing them. On a later test, those who had “practiced” using mental imagery showed the same improvement in performance as those who physically practiced the task. The lack of difference between the mental practice and physical practice conditions, as well as other results in the study, suggests that the benefits of practice were on cognitive representations and not the physical effectors.

Another prediction from the hypothesis that mental practice benefits a cognitive component is that the extent of mental practice should be independent of the movements required by the task: If two tasks share the same cognitive component, but require very different movements, the benefit of mental practice should be the same for the two tasks. One motor task with a large cognitive component is reading sentences in a foreign language. MacKay (1981) asked bilingual people to read sentences in German and English as rapidly as they could. Silent reading (mental practice) of the sentences in one language not only decreased reading time; it also resulted in faster reading of those same sentences translated into the other language. Mental practice resulted in a greater decrease in reading time for the translated sentences than did physical practice: The benefits of mental practice were not dependent on the different patterns of muscular activity required to read the sentences in German or English.

Acquisition of motor skill is usually best when performers combine mental practice with physical practice (Allami, Paulignan, Brovelli, & Boussaoud, 2008; Druckman & Swets, 1988). It may be that the combination of mental and physical practice results in more detailed and accurate motor programs. An optimal training routine will use some combination of both mental and physical practice. Apart from improved acquisition, there are other benefits to mental practice, including no need for equipment, no physical fatigue, and no danger.

TRAINING WITH SIMULATORS

We have talked about transfer and the extent to which practice of one set of movements improves the performance of a novel set of movements. How well practice transfers is of particular concern in the design and use of military and industrial simulator-training devices (Baudhuin, 1987; Rogers, Boquet, Howell, & DeJohn, 2010), such as those used to train pilots. Simulators are used for situations in which it is not feasible to have operators train in the real system. For example, student pilots should not train in real, fully loaded Boeing 767s, but they can operate a simulator.

The goal of using a simulator for training is to ensure the greatest possible amount of transfer to the operational system that is being simulated at the lowest possible cost. If training on a simulator transfers to the operational system, then money that would have been spent can be saved for the operation of the system itself. Moreover, the risk of physical harm and damage to the real system can be minimized. A “crash” in a flight simulator causes no real harm. For these reasons (and because of their increasingly wide availability), simulators are used extensively for research in air-traffic management (Vu, Kiken, Chiappe, Strybel, & Battiste, 2013), driving (Rendon-Velez et al., 2016), construction equipment operation (So, Proctor, Dunston, & Wang, 2013), and laparoscopic surgery (Luursema, Verwey, & Burie, 2012).

A major issue in simulator design involves the fidelity of the simulation to the real system. Designers often assume that physical similarity is important, and that the physical characteristics of the simulator should closely resemble those of the real system. This is clearly the case for flight simulators used for commercial aviation, which attempt to duplicate the natural cockpit environment closely (Lee, 2005). The cockpit of a full-motion flight simulator is an exact replica of that of the simulated aircraft. High-fidelity, wrap-around visual displays are used, auditory cues are provided, and changes in forces on the controls that would occur in real flight are simulated. The cockpit is mounted on a platform that moves in three dimensions, simulating the forces on the vestibular system that arise in flight. The end result is an experience that closely approximates that of an actual flight.

However, high fidelity is not necessary for effective simulator training. Practice on low-fidelity simulators such as desktop flight simulators and flat-screen construction equipment simulators results in positive transfer to real environments (Rogers et al., 2010; So et al., 2016). The extent to which simulated practice transfers to real systems depends on the extent to which the procedures to be executed are the same in the simulated and operational environments, even if the specific stimulus and response elements of the tasks are not identical.

Given the cost of high-fidelity simulation, along with the technological limitations that prevent perfect resemblance between the simulated and operational environments in many situations, training programs must emphasize functional equivalence over realism: the equivalence between the tasks that the operator will be required to perform in the simulation and in the real systems (Baudhuin, 1987). Functional equivalence, not realism, will determine how well practice will transfer.

With the widespread availability of relatively low-cost computers with powerful image-generation systems, it has become easy to develop inexpensive simulators based on personal computers. For example, X-Plane is a powerful flight simulator that will run on a personal computer. It can be used as a desktop simulator on any computer powerful enough to support it, and even to provide displays for an unmanned aerial vehicle (UAV) simulator (Garcia & Barnes, 2010). Lower-fidelity simulators like these provide only a restricted view on the computer monitor and omit many of the sensory cues available in a high-fidelity, full-motion simulator. However, they are sufficient for teaching basic perceptual-motor control skills, spatial orienting skills, and how to read flight instruments (Bradley & Abelson, 1995).

Virtual environment systems, based on virtual reality generators, construct a simulated environment in which a person is completely immersed. Within the virtual environment, a person interacts with a system in much the same way as with a “hard” simulator. Virtual environments can be used in place of more expensive training (Lathan et al., 2002). Because these environments do not have many of the constraints of physical simulators, they have the potential to ensure good skill transfer to the operational environment very cheaply.

FEEDBACK AND SKILL ACQUISITION

Many sources of sensory feedback are available both during a person’s movement (concurrent feedback) and upon its completion (terminal feedback). These sources of information are intrinsic: They come from the performer. Intrinsic feedback is not only important in control of movements, as described earlier in the chapter, but it also provides a basis for the learning of motor skill (e.g., Anderson, Magill, Sekiya, & Ryan, 2005). However, motor learning solely on the basis of intrinsic feedback can be very slow, and so most training programs use some form of augmented feedback that comes from a trainer or other source. This feedback usually takes the form of knowledge of results (KR), knowledge of performance (KP), or observational learning.

KNOWLEDGE OF RESULTS

KR refers to feedback about a performer’s degree of success in achieving a desired goal. Such feedback can be provided by an instructor or by an automated device. For example, a flight instructor may tell a student whether or not the goal of a particular maneuver was accomplished, or a flight simulator may indicate whether a landing was accomplished safely. KR reliably improves both the initial performance of a motor-learning task and its subsequent retention (Newell, 1976; Salmoni, Schmidt, & Walter, 1984). However, there are many ways that KR can be presented, and some forms of presentation are better than others. Research has focused on the effects of the precision, frequency, delay, and control of KR.

Precision of KR

There are two kinds of KR. Qualitative KR provides general information about the quality of performance (e.g., correct/incorrect), and quantitative KR specifies the direction and magnitude of error. Quantitative KR is therefore more precise than qualitative KR. Typically, quantitative KR produces better performance during acquisition than qualitative KR (Salmoni et al., 1984). Quantitative KR also leads to better retention, even under conditions where people receiving quantitative and qualitative KR perform equally well during acquisition (Magill & Wood, 1986; Reeve, Dornier, & Weeks, 1990).

Frequency of KR

Schema theory suggests that KR should be most beneficial when it is given after every trial, and that the benefits of KR will decrease as the percentage of trials on which KR is given decreases. This is generally true if performance is measured by the rate at which a skill is acquired (Salmoni et al., 1984), but more KR results in poorer retention. This suggests that less KR may produce better learning.

One experiment asked people to learn a pattern of lever movements (Winstein & Schmidt, 1990). This pattern consisted of four movements that were to be produced in 800 ms. KR was given on a computer that showed the actual movement together with the goal movement. One group of people received KR after each trial, while the other group received KR on only half of the trials. While both groups learned the task equally well, the group that received KR less often retained the task best. Another study showed similar results when people learned a pronunciation task (Steinhauer & Grayhack, 2000). However, some very complex tasks, such as a slalom-type task performed on a ski simulator, are not retained as well with low-frequency KR. This suggests that if a motor task is very complex, then more frequent KR will be beneficial (Wulf, Shea, & Matschiner, 1998).

The fact that retention is often better when KR is not provided on every trial suggests that it might be most effective to provide summary KR only after sets of trials are completed. Several experiments demonstrated that, in fact, this is true (Lavery, 1962; Vieira et al., 2014). One of these experiments (Schmidt et al., 1989) had people learn a timed lever-movement task like the one we just described (Winstein & Schmidt, 1990). They provided summary KR after sets of 1, 5, 10, or 15 trials to 4 groups of people. Everyone’s performance improved during the acquisition phase of training, but people who received KR less frequently did not perform as well as those who received it more frequently (see Figure 14.15). However, people’s performance on a delayed retention test showed an inverse relation between the length of the set and accuracy. Performance was best when summary KR was given every 15 trials and worst when KR was provided every trial.

FIGURE 14.15Effects of summary knowledge of results.

Delay of KR

After a movement has been executed or a trial completed, there is some delay before KR is presented. This interval is called KR delay. KR delay is only important when it is very short (Salmoni et al., 1984). When KR is provided immediately, it interferes with learning the task (Swinnen, Schmidt, Nicholson, & Shapiro, 1990). In one experiment, people performed a timed movement. One group of people received KR immediately, and the other half got KR after a brief delay. After a delay, people who had received immediate KR showed poorer retention than those who had received delayed KR. The researchers hypothesized that the time after a trial was important for evaluating intrinsic performance feedback, and that this evaluation helps people to detect their own mistakes. Immediate presentation of KR may interfere with this process.

In another experiment, Swinnen (1990) asked people to perform an attention-demanding secondary task during the KR delay. These people showed poor retention for the primary motor task, demonstrating that the secondary task interfered with learning. However, when the secondary task was performed after KR and before the next trial, retention was much better. In another study, people were asked to perform an extra movement during KR delay, and this extra movement affected retention only when it had to be remembered along with the primary task (Marteniuk, 1986). When people were asked to solve a number problem during KR delay, their retention performance was equally poor. All of these results demonstrate that part of learning a movement involves processing information about that movement after it has been attempted, and that any higher-level cognitive activity performed during the KR delay that is unrelated to the movement will interfere with retention.

KR and Self-Control

Allowing people to choose when they receive KR can also improve learning. One study paired people together to learn a task and allowed one person of each pair to choose when they would receive KR. The other person did not get to choose, but received KR at the same time as her partner. The person who chose when to receive KR learned better than the person who didn’t, even though they both received KR on exactly the same schedule (Sanli, Patterson, Bray, & Lee, 2013). One explanation for the benefit of self-control attributes better learning to improved motivation and perceived self-autonomy.

However, another study found that when people were required to perform a mentally demanding task during the interval after motor execution but prior to KR, the benefit for self-control of the KR schedule was eliminated (Carter & Ste-Marie, 2017). This finding suggests instead that the information-processing activity in which a person typically engages immediately after motor execution (e.g., evaluating how accurately the task was performed) is what determines the benefit, and not just a person’s self-control over whether feedback is provided.

Role of KR

Clearly, KR is important for the acquisition of movement skill. There are three major roles that KR may play (Salmoni et al., 1984). KR may improve motivation, one of the explanations of the effect of self-control, which in turn may result in greater exertion or effort when KR is present than when it is not. KR may help the formation of associations in memory. This is especially important for schema theory, in which KR helps form associations between stimulus and response features to create recall and recognition schemas. Finally, KR may provide guidance and help direct performance during acquisition (Anderson et al., 2005). When KR is provided for every trial, this can allow accurate performance without requiring the deeper processing necessary for learning to occur.

In summary, you should remember the following three points. First, a person learning a motor skill must actively process the information that provided by KR if it is to be of any benefit. Second, KR will be most effective if it is precise, controlled by the learner, and presented when the learner is not required to process other information at the same time. Third, when KR is presented too frequently, the learner may fail to process intrinsic information about his or her performance and instead rely only on the guidance provided by KR.

KNOWLEDGE OF PERFORMANCE

KR provides information about the outcome of a movement, but KP provides information about the performance of the movement, such as how the movement was controlled and coordinated (Nunes et al., 2014). We have good reason to believe that KP should be more effective than KR (Newell, Sparrow, & Quinn, 1985; Newell & Walter, 1981), but we must distinguish between kinematic KP, which describes some aspect of the motions involved in an action, and kinetic KP, which describes the forces that produce those motions. Studies of the effectiveness of KP show people kinematic and kinetic feedback about their movements, often with the kinematic and/or kinetic information for a successful movement presented for comparison.

Kinematic KP includes information about the spatial position, velocity, and acceleration of the limbs. An old, classic study of kinematic KP watched workers who operated a machine used to cut discs from tungsten rods (Lindahl, 1945). The machine was operated through a coordinated pattern of hand and foot movements. The foot movements were particularly important in determining the workers’ cutting efficiency and the quality of the discs. Lindahl recorded the feet of the most skilled workers (see Figure 14.16) and used these records to train new workers. This kinematic KP not only resulted in new workers learning the task faster and performing better (see Figure 14.16), but it also improved the performance of more experienced workers.

FIGURE 14.16Foot action patterns for (a) an expert operator and (b) a new operator after various amounts of practice with kinematic KP.

You may already have noticed that the kinematic KP in Lindahl’s experiment was actually just a kind of quantitative KR, because the goal of the movement was the same as the movement itself. Also, Lindahl did not examine the operator’s retention of the skill, but only acquisition performance. If our goal is to determine whether kinematic KP is better than KR, or in what circumstances KP or KR should be used, then we need to separate the movement goal from the movement and look at what happens with retention and/or transfer (Schmidt & Young, 1991). One experiment that did this is shown in Figure 14.17. People were shown a sequence of illuminated light-emitting diodes that looked like a moving ball. They were asked to manipulate a horizontally mounted lever in a way similar to a tennis racquet. When the light began to move, the performer made a backswing with the lever, bringing it anywhere to the left of the lights. Then the performer attempted to “bat the ball” by swinging the lever forward to intercept the moving light. In a task like this, KP is easily separated from KR. KR tells the performer whether or not the ball was successfully hit, whereas KP gives information about the swing. Note that KP is important in this task, because the apparatus did not constrain the performer’s movement much: the performer was required to select the position of the backswing, and the force and timing of the forward swing.

FIGURE 14.17Schematic diagram of the coincident-timing apparatus used by Schmidt and Young.

The best batting accuracy was achieved when the performer moved the lever back to approximately 165°. People were given only KR about the accuracy of their swing or the KR plus KP about position of their backswing. The people who received KR and KP performed better on acquisition trials and also on a retention test. Therefore, we can conclude that kinematic KP provides a benefit to motor learning over and above that provided by KR alone.

The benefits of kinetic KP have been examined for skills that require control of force and movement durations. For example, some isometric tasks (which require changes in muscle force without limb movement) should benefit greatly from kinetic KP, because accurate performance of such tasks is completely determined by kinetic variables. An early experiment demonstrating this was conducted on new U.S. Army recruits learning to shoot a rifle (English, 1942). Good marksmanship requires that the soldier squeeze the stock of the rifle at the same time as the trigger. Because this technique is difficult to learn, English implemented a new training program that provided kinetic KP. The stock of a rifle was hollowed out, and a fluid-filled bulb was placed within it. The bulb was attached to a fluid-filled tube that displayed the amount of force applied to the stock. The soldier could compare the level of the liquid in the tube when he shot the rifle with the level produced by an expert marksman. This method was remarkably effective, with even soldiers “given up as hopeless” achieving minimum standards quickly.

Some more recent experiments looked at the role of kinetic KP for tasks that required not only the application of a specific amount of force but also a gradual change in the application of force over time (Newell & Carlton, 1987; Newell, Sparrow, & Quinn, 1985). People performing one task produced a maximum force of 30 N against an immovable handle, whereas people in the other produced a specific force–time curve. Both groups received kinetic KP in the form of their force–time curves. KP improved both initial performance and retention for the task with the force–time criterion but not for the task with the peak force criterion. Further investigation into the way that this kind of KP should be presented evaluated the efficiency of performance when the actual and desired force–time curves were superimposed. Superimposing the curves was only beneficial when the desired force–time curve was asymmetric, or of a shape that was unfamiliar to the performers.

One popular way to present KP is by showing the performer a video replay of his movements (Newell & Walter, 1981; Rothstein & Arnold, 1976). However, in many situations this kind of KP can provide too much information, and can therefore confuse the performer rather than clarify what he needs to do to improve his performance. Video replay is most effective when the performer is told to pay attention only to the specific aspects of his performance that are important for learning.

The general principle emerging from this work is that the success of any type of augmented feedback will depend on the extent to which it provides relevant information in a form that is useful for improving performance. This means that before deciding what types of KR or KP to provide, a trainer must first analyze in detail the requirements of the task.

OBSERVATIONAL LEARNING

Sometimes people learn how to perform a motor task by observing someone else (a model) performing it. This is called observational learning. Explanations for how people learn through observation can be based on Bandura’s (1986) social cognitive theory. According to this theory, the observer forms a cognitive representation of the task to be learned by attending to the salient features of the model’s performance. This representation can then guide the production of the action when the observer is asked to perform it. The representation also provides a referent against which feedback from the observer’s own performance can be compared. Not too surprisingly, many of the same variables that affect motor learning when a person performs a task, such as frequency of KR, have similar effects on motor learning when a person watches a model perform the task (Badets & Blandin, 2004).

A movement sequence can be learned partly through observation, but not entirely (Adams, 1984). This is because important task factors (such as static force and muscle tension, and any unseen components of the movement) can only be learned by performing the task, and the inaccuracies or ambiguities in the cognitive representation of the movement cannot be resolved until the task is performed. Consequently, it is often helpful to combine observational learning with physical practice (Shea, Wright, Wulf, & Whitacre, 2000).

To the extent that a trainer can provide information about relevant task factors and resolve ambiguities during observational learning, a learner’s performance might benefit. One experiment explored this idea and demonstrated how observational learning can be improved by showing the observer the parts of a movement that are otherwise unobservable to a learner (Carroll & Bandura, 1982). This experiment asked learners to manipulate a paddle device in a complicated way (see Figure 14.14–14.18). A demonstration video recording was made of a model performing the components of the paddle-manipulation task, with the recorded image being of the back right of the model’s body (as in the figure) such that the orientation of the model’s arm and hand corresponded exactly to the observer’s arm and hand. Each learner saw the demonstration video of the modeled pattern six times, and after each demonstration, he or she had to execute the movement pattern. The experimenters presented simultaneous video of the learner executing the movement during his or her performance (visual feedback) on none of the trials, the first three trials only, the second three trials only, or all six trials. After the six acquisition trials, the learner had to execute the movement pattern three additional times without the demonstration video or visual feedback. After each set of three trials, the experimenters measured the accuracy of the learner’s cognitive representation of the task by asking him or her to put in order nine photographs representing the components of the action sequence.

FIGURE 14.18Response components of the action pattern investigated by Carroll and Bandura (1982), with the components numbered in the order in which they were enacted.

Visual feedback during only the first three acquisition trials did not help performance, but visual feedback during only the second three trials was as helpful as visual feedback on all six trials. The accuracy of ordering the photographs also was higher after the second and third sets of trials than after the first set. The authors interpreted their findings as indicating that, as implied by social cognitive theory, an accurate cognitive representation of the observed behaviors must be established before visual feedback of one’s own behavior can be beneficial.

Carroll and Bandura (1985, 1987) showed that the video augmentation was an effective training tool, but only when it was provided simultaneously with the learner’s movements. When it was delayed by about a minute, performance was no better than when the video was not provided. The improvement in learning obtained by having the learner match the actions of a model was equivalent to that obtained by the video augmentation, but this did not depend on whether the model’s action was presented simultaneously or later. Further investigation also showed that the more frequently the learner is shown the movements of the model, the better his performance will be (Carroll & Bandura, 1990).

Finally, it should be noted that observational learning may occur in either a more “bottom-up” or “top-down” fashion. People in one study practiced a motor sequence with a computer mouse (Roberts, Bennett, Elliott, & Hayes, 2015). The sequence trajectory either mimicked natural, biological motion or was an artificial trajectory that moved at a constant velocity. When the learners were led to think that the biological motion was human-generated, the learning was bottom-up: that is, it was primarily in the sensorimotor system and relatively automatic. But when they were led to think that the movement was computer-generated, both the biological and artificial motion trajectories were learned in a more top-down manner: that is, the learning required effortful cognitive processing. These results imply that a person’s understanding about agency in reproducing a motor pattern influences how they learn from observation.

To summarize, observational learning can be an effective training tool, but only to the extent that it promotes the learner’s development of an accurate cognitive representation of the task. The extent to which observational learning is useful for the acquisition of complex movement skills, relative to other KP methods, is still an open question. It may be that while observation is effective for learning the coarse aspects of tasks, such as the order and extent of different movements in a sequence, it will not be useful for learning exact details (Newell, Morris, & Scully, 1985).

SUMMARY

Understanding how people execute movements and control their actions is a fundamental part of understanding human factors. We have presented several important ideas in this chapter. First, control of action is hierarchical. The motor cortex receives proprioceptive feedback and delivers signals for control and correction of movement. These signals travel through the spinal cord, which alone can control movement to some degree. Our current understanding of higher-level motor control is that the brain develops plans for the execution of complex actions, whereas the spinal cord is involved in control of the fine adjustments.

The brain’s action plans, called motor programs, are hierarchical and modular, just like the organization of the nervous system. Complex actions involving more than one muscle group can be controlled by a single motor program. Motor programs rely on sensory feedback to determine the appropriate parameters, such as force and distance, for a particular movement, and the way that sensory feedback is used depends on whether an action requires open- or closed-loop control. Sensory feedback can be used to modify slower closed-loop actions as they are being executed, but it plays a smaller role in actions that are executed very quickly.

Perhaps the most interesting questions about how people control their movements are directed toward understanding how highly skilled behavior is learned. The ease with which motor skills are acquired varies greatly with the kind of training program used. High levels of practice variability will lead to better performance, retention, and transfer of similar types of movements. Learning and performance also will benefit from augmented feedback. Providing knowledge of both results and performance can improve learning when the information provided is chosen appropriately. The human factors specialist has an opportunity to provide input on optimal training programs that can speed an operator’s progress through the phases of skill acquisition.

RECOMMENDED READINGS

Enoka, R. M. (2015). Neuromechanics of Human Movement (5th ed.). Champaign, IL: Human Kinetics.

Jeannerod, M. (Ed.) (1990). Attention and Performance XIII: Motor Representation and Control. Hillsdale, NJ: Erlbaum.

Jeannerod, M. (1990). The Neural and Behavioural Organization of Goal-Directed Movements. New York: Oxford University Press.

Magill, R. A., and Anderson, D. I. (2014). Motor Learning: Concepts and Applications (10th ed.). New York: McGraw-Hill.

Rosenbaum, D. A. (2010). Human Motor Control (2nd ed.). San Diego, CA: Academic.

Schmidt, R. A., and Lee, T. D. (2011). Motor Control and Learning (5th ed.). Champaign, IL: Human Kinetics.