Organization of Movements

Human beings always on the move. We were once clumsy toddlers who eventually learn and readapt our movements to become more skillful. This adaptation occurs not only to make us more skillful, but also to adjust our motor system to the change in our growing body structure. There are 3 specifics domains of knowledge of 'motor behavior': motor control & planning, motor learning, and motor development across age. Motor control and motor learning have become popular, forming an intersection between neurophysiology and kinesiology.

To begin, there are three basic types of movements: 

1. Reflexive. Reflexes are born as a result of external stimuli and they are involuntary. The most popular example is the knee-jerk or patellar reflex and it employs the simplest form of lower-level connection called the monosynaptic circuit. Another example of reflex, the stretch reflex, occurs as a response to overextension to prevent muscle injury, e.g. it limits your movement when you flex your hand. A more complex movement will be the withdrawal reflex when our hand accidentally touches a sharp or hot object. 
2. Rhythmic. The second type, rhythmic movement, includes chewing, swallowing, breathing, scratching, and walking. The spinal cord and the brainstem participate in the production of these repetitive rhythmic motions and they are mostly stereotyped.  
3. Voluntary. The last type of movement includes any type of movement done to achieve certain objectives or a response to external stimuli. Voluntary movements are goal-directed and under the control of the cerebral cortex. We perform goal-directed movements with the blessing of the ability to learn, improve, or make corrections.

Simplified Sequence of Events

It is long postulated that movement production involves a series of events. At rest, motor unit activity is at a subthreshold level. The muscle state at rest is called muscle tone. What happens when one decides to move? Being the highest in the movement hierarchy, the cerebral cortex defines the objectives or goals and the intention to move. The prefrontal cortex is thought to be the place that this happens. This area receives projections from the parietal lobe that provides the body spatial information and body image. Not only do you need tactile and kinesthetic information of the body (the somatosenses), you also have to depend on the visual and auditory inputs from the environment. This information converges in the parietal region where multisensory integration occurs.

The next step is the preparation of a movement. The secondary motor areas (premotor cortices) are involved in this, and the basal ganglia play a role in providing supporting signals, i.e. the go or no-go behavior. The supplementary motor area (SMA), in particular, defines sequential information of muscle activation. Lastly, the primary motor cortex comes into play to activate specific muscles required for producing the movements. It determines how much force each muscle group must exert, and then sends these commands as many sets of action potentials to alpha motor neurons. 

At the level of the spinal cord, the upper motor neurons (the pyramidal cells) make contact with the lower motor neurons that innervate the muscles, either directly or indirectly via the spinal cord interneurons. Alpha motor neurons are neurons that generate actual movements.  At the neuromuscular junction, acetylcholine is released from the presynaptic terminal. Subsequently, cross-bridges are formed and the muscles will contract or stretch out, producing the intended movement. 

Lastly, it should be noted that movement production is not a one-time off thing, but spans for a certain duration. While performing the movement, there are two other parts of the CNS that are quite essential. Postural maintenance and balance while making certain movements are ensured by the brainstem. The temporal and spatial accuracy of the movement is monitored continuously by the cerebellum, the area responsible for error-detecting mechanisms. 

Movements as Sensorimotor Transformations

Movements are created by motor outputs, basically, neural commands that act on the muscles, causing them to contract and then produce movements. How do we get these outputs? They are derived from sensory inputs. The sensory information has gone through sensorimotor transformations. Sensory inputs provide both extrinsic states of the world as well as intrinsic (bodily) information. Two important components of intrinsic information: kinematics (position, velocity, acceleration, joint angles, muscle stretch) and dynamic or kinetic (forces) generated and experienced by the body.

Unlike reflexes, complex motor actions such as reaching to a cup of coffee in front of you involve a series of processes:

1. Spatial localization of items in the workspace in front of you. Our visual system is typically crucial for this. We also depend on the somatosensory system to tell where our arm is. You know you want to reach the cup with your hand (the end-effector) connected with your upper limb segments.
2. Movement planning based on the direction and distance to reach the cup based on the visual and proprioceptive information.
3. Inverse kinematics and inverse dynamics to achieve this movement goal. The first operation is to get the end-point trajectory of your hand. The second one deals with joint torque and muscle activities (forces) necessary to move your arm along the trajectory.

The end-point location can be computed from a set of joint angles of the arm through a forward kinematic transformation. The opposite, to know the joint angles to achieve the end-point, is called the inverse kinematic transformation. The same holds true for generating motor commands that are translated into joint torques and muscle forces to move your arm to the target, i.e. forward and inverse dynamic transformation. It is thought that our nervous system is able to perform both inverse and forward operations.

Fig-1: Two types of internal models that have been prescribed by theorists lately, taken from [1], 5th ed.

           
Recently, scientists have adopted engineering theories to model motor control and learning. It is thought that sensorimotor transformations involve internal models that serve as a controller to do forward and inverse transformations.

Feedback and Feedforward Mechanism

There are essentially two mechanisms involved in performing goal-directed movements, i.e. feedback and feed-forward. These mechanisms are performed by the central nervous system using the internal models. Updates of internal models take place during motor learning.

In a feedback mechanism, we have our desired state that acts as a reference signal. A comparator compares this against any incoming information fed back from the sensory afferents. The difference, called the error signal, will go through a controller with a specific gain. The gain represents the proportion of future output w.r.t current error. The output signal produced by the controller subsequently controls the actuator, i.e. our skeletal muscles. A movement is produced and our sensory system re-evaluates the performance. The online feedback mechanism then continues.

For example, a ball is landing on the hand. The desired state is to grab the ball with the correct amount of force and to position our arm at a certain angle in space. Our muscle and cutaneous receptors provide feedback information on the weight and current position of the ball in space. The comparator in the brain first compares the current or actual situation with the desired state. The motor system in the brain then sends appropriate commands to the skeletal muscles accordingly. In control engineering, high gain causes the controller to be easily unstable if there are large delays across the loop. It is known that our nervous system has certain delays, so the gain has to be kept small. Fortunately, IMHO we are amazingly flexible. It is unthinkable how we become unstable (‘to miss’, will be a better word maybe).

Fig-2: Two types of hypothetical mechanisms on how humans perform a movement, taken from [1].

Unlike the feedback system which is adaptive in nature, the feed-forward mechanism is an anticipatory control that relies on the information before the feedback loop is put to work. This feature is important for rapid and sudden movements and usually depends on prior knowledge. In other words, feed-forward control allows us to predict the sensory consequences. E.g. we want to catch a ball. Relying heavily on visions, we position our arms in a certain way. Only when the ball hits our hand can the feedback system begin working.

Characteristics of Goal-directed Movements 
The motor system can be seen as the opposite of the sensory system. Our central nervous system resembles a black box where sensory afferents as the inputs and motor efferents as the outputs. What is sensory perception? It is the representation of external stimuli coming in contact with our five senses. We utilize this representation to produce a goal-directed action. In other words, there is a transformation involved from sensory maps into motor commands, a process that makes use of the internal model in the brain.

It is thought that the brain has extensive representations to plan voluntary movement called motor program. Goal-directed movements improve with practice and this implies that we always readjust our sets of motor programs. Not only does the motor program code movement specification, but it also contains the simplest forms of action spatially and temporally called motor primitives. These primitives can be in the form of kinematic (strokes, submovements with varying speed), dynamic (force, muscles-joint torque synergies, and control policies), or both. The program defines the intricate relationships among the spatial extent and speed, acceleration, skeletal joint angles, and force.

The time delay between the stimulus onset and the required movement is called the reaction time (RT). It varies with the amount of information processed. At the lower level, it reflects the number of synapses, and thus, neural circuits, involved. Voluntary movements due to proprioceptive stimuli are typically ~120 msec, and 150 – 200 msec due to visual stimuli. The shortest latency of a monosynaptic reflex is about 40 msec. Reaction time can be shortened by providing a person with the knowledge of the event and reducing the number of choices to make (minimizing choice effect).

What are the salient physical features of a voluntary movement? Typically,
1. The spatial trajectory is smooth. If it's not smooth, it can be an indication of movement disorder.
2. The velocity profile resembles a bell-shaped curve with acceleration and deceleration steps.
3. The endpoint variability increases as movement distance increases.
4. It displays a speed-accuracy trade-off (Woodworth, 1890).
We make less accurate movements when we are asked to move faster. Using the appropriate term, there is an increase in motor variability. This may be due to more recruitment of neurons and muscle, and the noise of the nervous system is well-known.

Lastly, in the context of voluntary movements of the upper limb, reaching and grasping are widely studied for many decades. Grasping usually follows reaching for an object. Are the two terms independent? Recent anatomical studies and neuroimaging have shed new light on this debate. We now know that reaching and grasping have two different neural substrates,  These two channels are:
(1) Reaching primarily involves the visual cortex → superior parietal lobule → dorsal premotor (PMd).
(2) Grasping primarily involves the visual cortex → supramarginal gyrus → ventral premotor (PMv).

Prelude: Motor Learning

Our motor system must adapt to our own biological development such as growth in limb size, muscle power, etc., and the environment through experience. More recent concepts say that motor learning involves adapting internal models for novel kinematics and dynamic conditions. As mentioned, sensorimotor transformations have kinematic and dynamic components. In learning dynamic tasks, tasks that require manipulation of force and joint torque, we depend heavily on our proprioception. Such tasks can be learned well with and without vision. Deafferented patients have difficulty controlling the dynamic properties of the limb without vision. Kinematic tasks involve vision more to define movement trajectories and speed. 

References
[1]  Chapter 33 - The Organization of Movement. In Kandel E.R. et. al. (2000). Principles of Neural Science 4e, McGraw-Hill.
[2]  Tamar, F. and Hochner, B. (2005). "Motor Primitives in Vertebrates and Invertebrates". Current Opinion in Neurobiology, 15:1-7.
[3]  The Brain: From Top to Bottom.