Motor control, in reference to movements of an organism or motions of a robot, is often conceived of as a computational problem. How is something or someone able to move to achieve various environmental goals? For human movement, in particular, the question of how individuals are able to organize the motor system at multiple levels (e.g., joints, muscles, neurons) defines the study of motor control. In this way, motor control is a solution that is arrived at by the individual, which satisfies numerous and sometimes competing goals (e.g., to remain balanced, to reach a hot cup without getting burned, to avoid obstacles, to move quickly and accurately, to minimize energy expenditure, to avoid injury or uncomfortable positioning). The problem of control is typically conceived at multiple levels and is often distinguished with respect to the voluntary or intentional nature of control versus a more automatic or reflexive level. In the former case, movements are goal-directed and at least part of an unfolding movement designed to reach the goal is under voluntary control of the actor. In the latter case, movements are not under intentional control and take place without perceived control or conscious attention. This issue concerning what is controlled, prepared, or programmed in advance of a movement and what is controlled as a movement unfolds dominates much of the discussion in the motor control literature. For seemingly very simple movements like balance and walking, to more complex movements that are considered to define many sporting accomplishments, the problem of motor control is enigmatic.
Hierarchical Control
Many of the processes underlying human movement take place without explicit awareness on the part of the actor, but many movements are still voluntary. That is, an actor makes a conscious decision to act and this desire ultimately leads to movement. Yet the exact parameters of the movement are usually unknown and not directly controlled by the actor. Therefore, motor control is the problem of which transformations intervene between the thought of attaining a goal and the muscle activations that result in movement. Models and theories of motor control differ, but most if not all of the models are hierarchical in nature, suggesting that (a) the goal of the task is loosely set at a high level (sometimes referred to as the task level); (b) a sequence of movements is specified at a lower level (also known as response programming); and (c) at the lowest level, a specific pattern of muscle activation allows for execution (execution level). Although for clarity we treat these various levels as independent, it will be clear from the following discussion that these processes are neither serial nor mutually exclusive and our distinctions across levels serve primarily to aid thinking about the various problems involved in controlling movement.
Intending and Planning to Act (Response Identification and Selection)
At the task level, the goal of the movement is specified consciously by the actor, although there is evidence that actions are substantially influenced by purely perceptual stimuli without conscious intention on the part of the mover (e.g., different types of handles afford different grasping postures). At the task level, the issue of movement preparation or planning is important, and considerable research has been conducted to explore how movements are planned in advance of execution. We have learned a lot about movement preparation through simple measures of reaction time (RT) (i.e., the time between an environmental event and the actual action). Defining when an action actually begins is not always a simple procedure. For some people, action onset is defined with respect to a change in displacement of a limb in comparison to a baseline resting value, whereas movement onset has also been defined with respect to muscle activation as measured by electromyography (EMG). One of the most robust findings is that RT increases in a predictable, linear fashion as the number of stimulus–response (S–R) alternatives increases, so termed Hick’s law. Hick’s law is a problem of response selection. As choice increases, the time needed to make an appropriate movement response will also increase.
Further inferences about movement planning based on measures of RT have been verified with more sophisticated recording techniques that identify brain areas involved in movement preparation. For instance, areas in the prefrontal cortex(PFC) are particularly involved in the task-level processes of motor control. Human neuroimaging studies show decreased activity in the PFC as a result of motor learning, suggesting that as a skill is practiced there is less reliance on explicit (attention demanding) control of the movement. When people learn a sequence of movements implicitly (i.e., without awareness of learning and sequence regularities), activity in PFC decreases relative to explicit learning conditions (i.e., with verbalizable knowledge of the sequence). Similarly, when implicitly learned regularities in the sequence are explicitly revealed, PFC activity increases to baseline levels. Damage to PFC has also been shown to reduce conscious control of movement, whereby brain-lesioned patients use or respond to objects in a stereotyped manner that is inappropriate for a given context.
Significant attention has also been paid to the role of vision in the planning and execution of actions. One neurophysiologically supported theory is that there are two main processing routes in the brain that are activated, with selection depending on whether actions are required toward an object. When individuals interact with objects or move to targets, processing of visual information is assumed to be primarily governed by a dorsal route in the brain (i.e., a route that involves areas of the brain near the top of the cortex, particularly the posterior parietal cortex). In contrast, when individuals make decisions about a target’s position or identity, ventral route processing in the brain is involved (involving lower level cortical areas, primarily occipital-temporal cortex). Patients with temporal cortex lesions are severely impaired when visually identifying objects but show unimpaired tool use and object manipulation. Conversely, patients with parietal cortex lesions have unimpaired recognition abilities but significantly impaired reaching and pointing accuracy (suggesting a deficit in calculating spatial locations). This double dissociation between recognition in the temporal lobe and perceptual integration in the parietal lobe suggests that the ventral and dorsal streams of vision might be differentially involved in response selection and response programming, respectively.
Response Programming
The history of response programming in motor control dates back to the late 19th and early20th century. Programming is thought to be the process preceding voluntary actions whereby action plans are organized and potentially stored in cortical or subcortical structures in the brain, ready to be released when a response is required. Evidence for programming is that people are unable to inhibit actions when they are close to their release point, studies of RT show that as complexity of a movement increases so does RT, stereotyped movement features are observed when an action is repeatedly executed, suggesting that at least part of the response is programmed in advance and is not influenced by feedback (termed open-loop control), and features of a movement remain when the movement is unexpectedly blocked or is initiated early due to a startling acoustic stimulus (i.e., the start-react effect).
The term motor program has had varied usage in theories of motor control, although there is little consensus as to what this motor program might be and even whether such terminology helps in our understanding of motor control. Some of the discussion has been with respect to the specific or general nature of motor programs (i.e., are programs for throwing specific to certain distances, ball sizes, limbs), the neuroanatomical location of such a motor program (cortical, subcortical, spinal), as well as the necessity of motor programs, if one considers actions to be assembled on the fly in response to real-time constraints in the environment (i.e., the principle of self-organization in dynamical systems theory). More recently, a language of control has emerged that is based on the idea of internal models, rather than programs. This language and associated ideas were primarily a result of computational motor control theorists who emphasized predictive processes occurring as an action unfolds (i.e., forward, anticipatory models) as well as the idea of control processes that arguably function like motor programs (i.e., responsible for programming motor commands). Despite the debate about the nature or even existence of motor programs, it is evident that actions are at least in part assembled prior to action execution. The nature of the programming that is required is discussed next.
One of the first processes that must occur at the programming level is the translation of objects in allocentric space (the relation of objects in space to each other) into egocentric space (the relation of objects in space to the actor). This translation is tremendously important for sport. Consider asoccer player who is trying to connect their body to volley an incoming ball. Making contact with the ball depends on an accurate calculation of where the ball is in space, where the player’s foot is in space, and calculation of the relationship between the moving ball and the moving leg. Neurological evidence suggests that multiple egocentric spatial frames exist centered around the head, the arm, or the leg depending on the effector that is being used in the action. Other physiologically driven research has led to the suggestion that rather than the whole trajectory of a moving limb being computed in advance, that only the endpoint of the movement is programmed. Rather than individual joints or muscles being represented in the cortex, many features of a movement arise from the peripheral nervous system (rather than the brain) and intrinsic properties of muscles and joints (e.g., equilibrium point theory). However, there is evidence that the sequence of programming is hierarchical in nature, based on such parameters as the extent and direction of a movement as well as the choice of limb required. It is not possible to program movement amplitude if one does not know the limb required to make the movement.
Programming time is affected by the complexity of the response required. Movements that involve more components or that are of a longer duration than others take longer to assemble or program. This programming time is however influenced by the skill level of the actor. After extended practice, action–response sequences become grouped or unitized (requiring less time to program), whereas for novices these responses exist as small(er) movement “chunks” requiring more time to prepare or assemble. A skilled tennis player is therefore expected to take less time to prepare a return lob in tennis than a more novice player, if for the novice player the lob is conceived as a sequence of components potentially involving a backswing and forward swing phase.
Also at the programming level, the correct sequence of movements needs to be specified in order to generate the correct endpoint and movement effect. As motor sequences are learned, the movement becomes more accurate, faster, and more automatic. Neurophysiologic data show that the PFC and presupplementary motor area (preSMA) are important for learning new sequences. As these sequences are learned, however, there is a decrease in these areas and a corresponding increase in activity in the supplementary motor area (SMA) and primary motor cortex. The connection of these cortical structures to lower levels of the brain, (i.e., the basal ganglia and the thalamus) creates feedback loops that are important for the programming of movement sequences in a goal-directed fashion (i.e., learning to calculate the appropriate “next-step” given a current task goal). Interestingly, activity in the SMA is found when individuals physically perform an action and when they mentally imagine performing an action. This suggests that mental imagery may be an effective way to practice certain aspects of a motor skill because mental imagery, to a limited extent, requires programming of the motor sequence.
Making the Movement
At the execution level, there is little or no access to awareness and conscious control as this level consists of detailed specification of motor unit recruitment in both time and space. The type of representation used at this level is not clear, but there is research to suggest that rather than programming individual muscle contractions or joint torques, the execution level relies on stereotyped relationships and patterns of muscle activity referred to as synergies or “motor primitives.” For simple movements, much of this execution-level programming appears to occur at the level of the spinal cord. Electrical stimulation of the spinal interneurons in animals leads to relatively smooth movements of the whole effector toward a consistent endpoint. In contrast, stimulation of the alpha-motor neurons, the next level below interneurons, leads to motor unit contractions within a single muscle at a constant force dependent on the magnitude of stimulation.
Reflex loops and hardwired neural pathways exist at the execution level that differ depending on the type of control being examined. That is, there are specific neural pathways that generate relatively consistent patterns of muscle activity across individuals that maintain balance and posture, stabilize patterns of locomotion, or coordinate movements of the eyes relative to the head in order to maintain gaze. These fast, low-level motor control processes operate outside of awareness and can be very difficult to override if they conflict with voluntary motor actions. However, there is evidence that these pathways can be modified through experience and can operate in a goal directed manner (i.e., the result of a reflex loop can change depending on the current goal of the motor system).
Sensory feedback, as well as the prediction of what sensory feedback will look, feel, and sound like, is critical at the execution level so that actions are controlled appropriately to achieve specific functions and goals. The motor system has its own internal error correction mechanisms that facilitate success in movement and learn from experience to progressively fine-tune goal-directed actions. The exact process by which these modifications are made to the motor system is not clear, but several variables must be calculated in order to detect errors and correct the transformations from task level to execution level. These transformations might be conceived of as internal models or dynamic biological processes involving both predictive and control processes. In order to tune these processes, the motor system must represent the actual state of the system, the desired state of the system, and the predicted “next state” of the system. By comparing these different states, error signals can be generated that tune both the predictors and the controllers. Predictors are tuned through the comparison of the predicted state and the actual state (i.e., Did that do what I thought it would?). Controllers (or motor commands) are tuned through comparison of the desired state with the actual state.
It appears that these processes associated with internal models take place at a mostly subconscious level. Neuroimaging data show that when a movement sequence is explicitly attended to, there is an increase in the activity of the PFC and a decrease in the SMA (i.e., a shift to the neural structures that were involved early in the learning process) and a corresponding reduction in the speed at which a movement or sequence of movements is produced. Behavioral evidence also suggests that attending to the details of movement, compared to attending to the desired goal, leads to worse performance and less efficient motor control.
One final consideration for movement execution is critical task variables that appear to affect how the movement is executed. For example, it is well established that movement time (MT) is (predictably) dependent on the size and extent of a movement required (so called index of difficulty). Large movements to small objects are slower to execute than small movements to large objects, arguably because more time is needed to process visual feedback to accurately hone in on the object. Measurement of movement kinematics (or movement process measures), such as displacement, velocity, and acceleration of a limb, has allowed researchers to infer how movements are controlled, based on sensory feedback, as an action is executed.
“Offline” Success Evaluation
This general, hierarchical framework provides a basis for understanding how a command makes the transformation from thought to action. However, in order to achieve the desired goal accurately and reliably, the motor system needs to be able to modify (“tune”) these transformation processes based on sensory feedback errors about the success of an action. This evaluation can happen after the movement in what is termed an “offline” fashion. Correcting subsequent movements can be an onerous task for the motor system because there is considerable ambiguity in what sensory errors mean from a motor control standpoint. Errors convey information about the success of a movement (e.g., hits vs. misses) and some feedback about what needs to be corrected in the movement (e.g. a miss to the left versus the right), but the error does not specify which aspect of the movement needs to be modified (the credit assignment problem) or when in the movement a modification needs to occur (temporal credit assignment). This is where a coach can play an important role, using his or her personal experience and expertise to identify where and when an error occurs and then provide additional (augmented) feedback to an athlete to fix the error.
References:
- Frith, C. D., Blakemore, S.-J., & Wolpert, D. M. (2000). Abnormalities in the awareness and control of action. Philosophical Transaction of the Royal Society, 355, 1771–1788.
- Latash, M. A. (2012). Fundamentals of motor control. Boston: Academic Press.
- Milner, A. D., & Goodale, M. A. (2006). The visual brain in action (2nd ed.). Oxford, UK: Oxford Psychology Series.
- Rosenbaum, D. (2009). Human motor control (2nd ed.). San Diego, CA: Academic Press.
- Schmidt R. A., & Lee T. D. (2011). Motor control and learning: A behavioural emphasis (5th ed.). Champaign, IL: Human Kinetics.
- Willingham, D. B. (1998). A neuropsychological theory of motor skill learning. Psychological Review, 105,558–584.
- Wolpert, D. M., & Kawato, M. (1998). Multiple paired forward and inverse models for motor control. Neural Networks, 11, 1317–1329.
See also: