Programming Rapid Movements in Advance
In chapter 3, we introduced the idea that in a response-programming stage the person selects, organizes, and initiates a program of action that will produce a series of muscular activities resulting in an action. According to this model, the program must be structured completely (or almost completely) in advance, before the movement can be initiated, and very little modification will occur in the movement for the next few hundred milliseconds or so. We saw evidence for this also in situations in which the limb was blocked from moving, with the pattern of EMGs being unaffected (as compared to unblocked moves) for the next 100 ms or so (Wadman et al., 1979). We also saw evidence that a startling stimulus can “release” a complete, preplanned action far earlier than would be the case if it was released “normally” via the stages of processing (Valls-Solé et al., 1999). Another line of support for this hypothesis is the evidence that certain variables, related to the “complexity” of the movement to be made (i.e., the number of limbs or movement segments, the movement’s duration, or both), tend to affect the time between stimulus and the beginning of the movement (i.e., the RT; Henry & Rogers, 1960). More complex, and longer-duration (but still rapid) movements produce longer RTs (Christina, 1992; Klapp, 1996). It has not been possible to explain how these effects could occur except by the hypothesis that the movement is programmed in advance, with these variables affecting the duration of the stage necessary for completing this preprogramming (Henry & Rogers, 1960; Schmidt, 1972b; Schmidt & Russell, 1972). Many of these ideas are far from new; the original notion dates back to thinking by James (1890) and Lashley (1917), and more recently to Henry and Rogers (1960), Keele (1968), Schmidt (1976a), and Brooks (1979). Early thinking on how motor programs might be structured is credited to Karl Lashley (see “K.S. Lashley on ‘Motor Programs’” on p. 198; see also Rosenbaum, Cohen, Jax, Weiss, & van der Wel, 2007).
As a result of this thinking, at least two levels can be distinguished in the motor system: (a) an executive level (including the information-processing stages) for selecting, organizing, and preparing and initiating a complex pattern of muscular activities and (b) an effector level (motor programs and the muscular system) for actually controlling or producing the patterns as they unfold. We can further distinguish these two levels by examining two distinct types of errors that can occur in performance, which we describe later in the chapter. That said, however, we have already presented evidence that such a view can explain only a limited set of movement situations, as many examples can be cited in which feedback processes seem to interact with open-loop processes in the production of movement. A more complete approach to motor programming would be to ask how the sensory processes operate together with (i.e., cooperate with) these open-loop processes to produce skilled actions. We turn to some of these ideas in the next sections.
Sensory Information and Motor Programs
The next sections deal with various functions of feedback in movement control. We can conceptualize these functions as acting before a movement, during a movement, and after a movement.
Prior to the Movement
One of the major roles of sensory information is almost certainly to provide information about the initial state of the motor system prior to the action. Consider this simple example: You must know whether you are standing with your left or right foot forward in order to initiate a walking pattern (Keele, 1973). The spinal frog (figure 6.8) requires sensory information from the forelimb in order to direct the hindlimb to the elbow during the wiping response. Such information is presumably provided by afferent feedback from the various proprioceptors, and it would seem to be critical for the selection and adjustment of the proper action. We argued in chapters 2 and 3 that these processes are very important for open types of skills, for which the nature of the environment is unpredictable or constantly changing.
Polit and Bizzi (1979), using deafferented monkeys, showed that when the initial position of the shoulder changed prior to the elbow action, a systematic error in pointing to the target position occurred. This is understandable from figure 6.15, because changing the shoulder angle as shown necessarily affects the elbow angle (from u1 to u2) required for pointing at a target in a given position in space. If the monkey programmed a given elbow angle, then the equilibrium-point mechanism (chapter 7) would achieve that angle, and the arm would not be pointing to the proper target. These monkeys did not learn to point to the target, even after considerable practice. By contrast, normal, intact monkeys learned in a few trials to compensate for the shifts in the shoulder position. The interpretation is that the intact animals had feedback from the shoulder joint and could adjust the angle at the elbow to compensate for the change in the shoulder angle. Thus, these data suggest that feedback about the initial positions of the joints is required when pointing to a position in space if the environment is not perfectly predictable.
Another role of afferent information involves what a number of authors have called functional tuning (Fitch, Tuller, & Turvey, 1982; Turvey, 1977). Recall that the spinal apparatus and resulting limb force output could be affected by change in the head position, much as would be expected on the basis of the idea that the tonic neck reflex was involved in the action (Hellebrandt, Houtz, Partridge, & Walters, 1956). In this example, afferent information from the neck influences the spinal mechanisms prior to action, thereby facilitating or inhibiting them. But a more compelling reason for assuming that premovement tuning must occur relates to some simple facts about the nature of the motor apparatus. In figure 6.16 are two diagrams of a hypothetical rapid movement. In both cases, the movement involves flexion of the elbow a distance of 45°, beginning with the arm straight. In figure 6.16a, the upper arm is positioned 45° below the horizontal, so that a flexion of the elbow will result in the forearm’s being horizontal at the end. In figure 6.16b, the upper arm is 45° above horizontal, so that the forearm will be vertical at the end. The same command signal delivered to the biceps muscle group will not “work” in both situations, for two reasons. First, a force is required to hold the forearm against gravity at the target position in the first situation, but not in the second. Second, more force is required to move the forearm against gravity in the first example relative to the second. A logical conclusion from this simple example is that the motor system must “know” the position the shoulder is in prior to the action so that the command to the elbow flexors can produce the required 45° movement. How this happens is not entirely clear, but that it happens seems nearly obvious.
Consider another complicating factor facing the motor system in producing a movement. Figure 6.17 is a schematic diagram of the muscle attachments involved in a simple movement. This time, imagine that the movement is an extension movement in which the elbow is to be moved through 45°. Notice that the triceps muscle, which is the primary elbow extensor, is attached to the humerus in two places (internal and external heads) and to the scapula of the shoulder area (the long head). Thus, the triceps muscle performs two actions when it contracts: It extends the elbow and it tends to extend the shoulder joint, pulling the humerus back. Therefore, when the triceps is contracting to produce the 45° movement, one of the muscles that flexes the shoulder must contract so that the shoulder joint is stabilized and only the elbow moves. Thus, during this simple extension movement, the motor system must “know” that a two-jointed muscle is involved and produce some compensatory stabilization. The amount of stabilization will be dependent on the shoulder angle because of the length–tension relation (chapter 7).
The picture that emerges from these observations is that a “simple” 45° movement of the elbow joint is not really that simple at all, at least in terms of the motor system. In addition, other complicated aspects of the muscle need to be considered by the motor system, such as the nonlinear relationship between the muscle force and limb velocity, together with aspects of the contraction process that make the motor system very difficult to predict and control (Partridge, 1979, 1983). We all know that our nervous system controls our limbs beautifully in these “simple” situations. How it does so is exciting to ponder.
One role that feedback seems to have during movement production is a monitoring function, whereby the feedback from the movement is taken in and processed but not necessarily used in the control of the action unless something goes wrong. It is probable that a long string of actions dealing with finger movements in piano playing is programmed and carried out open loop. Feedback from the fingers is returned to the central nervous system for analysis, as if the central nervous system were “checking” for errors. If no errors appear, then the feedback is ignored. But if the feedback indicates that an error has occurred, attention can be directed to that feedback source, and an appropriate correction may be initiated. Reflexive corrections may also be generated, as discussed in chapter 5.
A second way to view feedback is that it may be intricately involved in the physical control of the limb. We mentioned a number of examples of this in the preceding chapter. The possibility exists that a constantly changing reference of correctness is specified by the gamma motor neurons to the muscle spindles and that their actions result in a continuous set of corrections to keep the movement on the proper course. The feedback could be involved in the determination of the end location of a movement if the reference of correctness were set for this position. And in repetitive movements, the feedback from early segments of the sequence can provide adjustments for the later segments.
Following the Movement
Extensive feedback is also delivered to the central nervous system after a movement. Such information can be evaluated, presumably, by the stages of information processing in order to determine the nature (or “quality”) of the movement just made. Information about whether or not the movement achieved the environmental goal and about its smoothness, its level of force or effort, or its form or style is derived from sensory feedback. A major role for such information is in the adjustment of the movement on the subsequent trial, perhaps to reduce the errors made on the present trial. As such, this information has a considerable relevance to the acquisition of skills, as discussed in the final part of this book dealing with motor learning (chapters 12 and 13 in particular).
Types of Motor Program Errors
If, as we suspect, the motor system can be conceptualized as consisting of two major parts (an executive level and an effector level), we can ask about the origins of errors in each—including their causes, their detection, and their correction. We conceptualize the motor system as being capable of making essentially two distinct types of errors. Each of these errors involves feedback in distinctly different ways for their detection and correction. We discuss these two kinds of errors in the next sections.
When a person makes a rapid movement, there are really two goals (Schmidt, 1976a). First, there is an environmentally defined goal, such as changing gears in a standard transmission car or doing a somersault from a diving board. A second goal (or subgoal) can be defined in terms of the muscular activities required to produce the desired outcomes in the environment—that is, to produce the first goal. For example, a person must contract the muscles in the arm and torso in one of a limited number of ways in order to change gears smoothly, and only certain patterns of muscular activity will result in a somersault. Essentially, how to generate such a subgoal is the problem facing the performer.
We can view the subgoal as a pattern of action that is structured in both space and time. Such a pattern of action will determine where a particular part of the body will be at a particular time after the movement starts. If this spatial–temporal pattern in the muscles and limbs (the subgoal) is produced accurately, then the environmental goal will be achieved. This process can go astray in essentially two ways.